Sustainable Water Harvesting and Institutional

Strengthening in Amhara
(SWHISA)

Hydrology Guide Book

Prepared

By

Birhanu Zemadim (PhD)

E-Mail birhanuzem@[Link]
Phone +251 (0) 910457450

January, 2010
Bahir Dar

1
 Table of Contents
Table of Contents .............................................................................................................................2
List of Figures ..................................................................................................................................5
List of Tables....................................................................................................................................8
Summary ..........................................................................................................................................9
CHAPTER ONE ............................................................................................................................10
HYDRO-METEOROLOGICAL DATA MEASUREMENT AND QUALITY CHECK.............10
1.0 Introduction ..........................................................................................................................10
1.1 Data and Data Organization .................................................................................................10
1.1.1 Standardization of data format for easy dissemination; UCG Format ..........................11
1.1.2 Identification of Missing Records .................................................................................13
1.2 Precipitation Data.................................................................................................................13
1.2.1 Precipitation Data Measurement ...................................................................................14
1.2.2 Point precipitation and missing data estimation............................................................16
[Link] Missing rainfall estimation using Arithmetic mean ...............................................18
[Link] Missing rainfall estimation using Normal Ratio Method.......................................18
1.2.3 Consistency of Precipitation Data .................................................................................20
1.2.4 Areal Precipitation........................................................................................................23
[Link] Thiessen method.....................................................................................................25
[Link] Isohyetal method ....................................................................................................28
1.3 Streamflow Data...................................................................................................................32
1.3.1 Selection of sites for stream gauging ............................................................................33
1.3.2 Method of measuring streamflow .................................................................................34
[Link] Current Meter Gaugings .......................................................................................36
[Link] Rating of the Current Meter ...................................................................................36
1.3.3 Stage Discharge Rating Curve ......................................................................................39
[Link] Parameter Optimization..........................................................................................41
[Link] The Paired t-test .....................................................................................................42
[Link] The Sign test...........................................................................................................43
1.3.4 Flow data Synthesis.......................................................................................................49
1.3.5 Flow Duration Curves ...................................................................................................50
[Link] Procedures of FDC construction using Rank-Ordered Method: ............................51
[Link] Procedures of FDC construction using Class Interval Technique: ........................51
1.3.6 Missing data computations............................................................................................55
1.4 Sediment Concentration .......................................................................................................56
1.4.1 Method of Sampling Sediment......................................................................................56
1.4.2 Relationship of Suspended Sediment Transport and Streamflow.................................57
1.4.3 Computation of seasonal distribution (Date Average and Annual) ..............................58
1.4.4 Reduction in Reservoir Capacity...................................................................................59
1.4.5 Prediction of a Reservoir Life .......................................................................................60
1.4.6 Reservoir Sedimentation Control ..................................................................................63
1.5 Climatic Variables................................................................................................................64
1.5.1 Climatic Data Measurement..........................................................................................64
CHAPTER TWO............................................................................................................................67
RAINFALL RUNOFF MODELS..................................................................................................67
2.1. Introduction .........................................................................................................................67

2
 2.2 Time series data availability.................................................................................................67
2.3 General Climate Characteristics...........................................................................................71
2.4 Distribution of Precipitation.................................................................................................73
2.4.1 Temporal Distribution ...................................................................................................73
2.4.2 Spatial Distribution .......................................................................................................74
2.5 Distribution and variability of Streamflow ..........................................................................75
2.6 Streamflow Hydrograph.......................................................................................................77
2.7 Rainfall Runoff Models........................................................................................................78
2.7.1 The Soil and Water Assessment Tool (SWAT) ............................................................79
[Link] SWAT Model Applications in the Ethiopian Catchments .....................................82
2.7.2 The HEC-HMS Model ..................................................................................................85
[Link] Application of HEC-HMS Model in Ethiopian Catchments..................................86
CHAPTER THREE........................................................................................................................87
RESERVOIR EVAPORATION AND SEDIMENTATION ESTIMATION ...............................87
3.0 Introduction ..........................................................................................................................87
3.1 Method of estimating evaporation using climatic data ........................................................87
3.1.1 Methods of Parameters Estimation for Penman Combination ......................................88
3.1.2 Estimation of Actual Evaporation .................................................................................91
3.2 River sediment inflow estimation: Prediction of a Reservoir Life.......................................92
3.2.1 Sediment Load Estimate................................................................................................92
3.3. Predicting Reservoir Life ....................................................................................................93
3.3 Reservoir Sedimentation Control .........................................................................................95
3.4 Reservoir Capacity Estimation.............................................................................................96
CHAPTER FOUR ..........................................................................................................................97
Frequency Analysis of Rainfall and Determination of Peak Flow.................................................97
4.1 Statistical Distributions commonly used in Hydrology .......................................................97
4.2 Local IDF Curve Construction.............................................................................................97
4.2.1 Rainfall Data Analysis ..................................................................................................97
4.2.2 Frequency Analysis .......................................................................................................99
4.3 Methods for Frequency Analysis .......................................................................................100
4.3.1 The Extreme Value Distribution or Gumble Distribution...........................................100
4.3.2 Lognormal Distribution...............................................................................................102
4.3.3 Log Pearson Distribution ............................................................................................102
4.4 Fitting the Rainfall Data Using EVI Type I distribution....................................................103
4.4.1 Fitting the data to Lognormal and Log Pearson type III Distribution.........................104
4.5 Testing the Goodness of Fit of Data to the Probability Distribution .................................107
4.6. Computation of the Probability distribution Function ......................................................108
4.6.1 Computation of Extreme Value (XT) For the Various Rainfall Durations and Return
Periods..................................................................................................................................109
4.7 Determination of Flood Magnitude....................................................................................112
4.7.1 Sheet Flow...............................................................................................................114
4.7.2 Shallow Concentrated Flow ....................................................................................115
4.7.3 Open Channels ........................................................................................................115
CHAPTER FIVE..........................................................................................................................120
Synthetic Flow Generation and Reservoir Simulation.................................................................120
5.0 Introduction ........................................................................................................................120
5.1 Synthetic Flow Data Generation using Thomas Fiering Method.......................................120

3
 5.2 Reservoir Sediment Volume Estimation ............................................................................122
5.3 Distribution of Sediment in the Reservoirs ........................................................................122
5.3.1 Area-Reduction Method of Borland and Miller..........................................................123
5.3.2 Computing the non zero-capacity elevation at the dam/reservoir...............................124
5.3.3. Distribution of Sediment in the Reservoir............................................................125
5.3.4 Reservoir Topography.................................................................................................125
Chapter Six...................................................................................................................................129
Probable Maximum Precipitation (PMP) and Probable Maximum Flows (PMF).......................129
6.1 Probable Maximum Precipitation.......................................................................................129
6.2 Methods of Estimating Probable Maximum Precipitation (PMP) .....................................129
6.2.1 The Hershfield Method ...............................................................................................130
6.3 Transformation of Point PMP to Areal PMP .....................................................................133
6.4 Probable Maximum Flood (PMF) ......................................................................................133
CHAPTER SEVEN......................................................................................................................135
GIS and Remote Sensing Application in Water Resources Applications....................................135
7.0 Introduction to GIS and Remote Sensing...........................................................................135
7.1 Catchment Characterization using GIS Technique ............................................................135
7.1.1 Deriving Catchment Characteristics............................................................................135
7.2 Watershed Delineation and Mapping: HEC-GeoHMS Watershed Processing.................139
7.3 Remote Sensing Applications ............................................................................................142
7.3.1 Use of IDRISI/ILWIS to study Catchment Characteristics ........................................143
7.3.2 Analysis of Catchment Cover Change Detections ......................................................149
CHAPTER EIGHT.......................................................................................................................150
PREDICTION IN UNGAGED CATCHMENT ..........................................................................150
8.0 Introduction ........................................................................................................................150
8.1 Characterization of Gaged and Ungaged Catchments........................................................150
8.2 Methods of Ungaged Catchment Analysis.........................................................................150
8.2.1 Use of statistical Rules in Data Transfer-Regionalization Approach .........................150
8.2.2 Use of Hydrologic Models ..........................................................................................152
[Link] The Pitman Hydrologic Model.............................................................................152
[Link] Application of Pitman Model to Ungaged Catchments .......................................154
References ....................................................................................................................................165

4
 List of Figures
Figure 1.1 Raw suspended sediment data ......................................................................................11
Figure 1.2 Suspended sediment data (For Koga Watershed) ........................................................12
Figure 1.3 Presentation of Percentage Missing from data of Koga watershed .............................13
Figure 1.4 Examples of non-recording and recording rain gauges ................................................14
Figure 1.5 Coordinate system representing the location and distance relationship of Point A and
its nearby gauging sites, Points B, C, D, E, and F..........................................................................17
Figure 1.6 DMC indicating the accumulated annual precipitation of a station X and that of
surrounding stations .......................................................................................................................21
Figure 1.7 Double Mass Curve for station A and surrounding stations.........................................23
Figure 1.8 Corrected Double Mass Curve and presentation of an outline for areal precipitation
computation....................................................................................................................................25
Figure 1.9 Representation of point rainfall in a watershed for areal computation using Thiessen
polygon method..............................................................................................................................26
Figure 1.10 Polygonizing point rainfall data for areal precipitation estimation using Thiessen
Polygon method..............................................................................................................................26
Figure 1.11 Perpendicular bisectors along the triangle sides. ........................................................27
Figure 1.12 Polygons drawn to indicate influence areas of point rainfall data..............................27
Figure1.13 Methods of computing areal average precipitation for a specific area, or drainage
basin. ..............................................................................................................................................29
Figure 1.14. Representation of point rainfall for areal estimation using Isohyetal method...........29
Figure 1.15 Isohyets drawn to for areal estimation........................................................................30
Figure 1.16 Example data file .......................................................................................................31
Figure 1.17 View of stream cross-section showing the location of points of observation ............32
Figure 1.18 Velocity profile along a cross-section (a) and along a vertical (b).............................35
Figure 1.19 Typical picture of a current meter (Price current meter) and Current meter gauging
(Source: Raghunath (2006) ............................................................................................................36
Figure 1.20 Mid section (a) and Mean Section (b) Methods .........................................................37
Figure 1.21 Discharge Computation ..............................................................................................38
Figure 1.22 Gauge posts on river banks.........................................................................................40
Figure 1.23 Rating Curve and Equation.........................................................................................41
Figure 1.24 Arithmetic plot of stage and discharge data................................................................47
Figure 1.25. Log-Log Plot of stage and discharge data .................................................................48
Figure 1.26 Manual computation of rating curve equation............................................................48
Figure 1.27 Flow Duration Computations......................................................................................54
Figure 1.28 FDC indicating the 50% and 95% flow exceedence probability ................................55
Figure 1.29 Computation of missing flow records in the Tana basin system ................................55
Figure 1.30 Sediment accumulations in a reservoir .......................................................................58
Figure 1.31 Date averaged values of suspended sediment for Koga watershed ............................58
Figure 1.32 Annual Distribution of Suspended Sediment for Koga watershed (1960 - 2002) ......59
Figure 1.33 Reservoir trap efficiency vs. capacity-inflow ratio (after Brune, 1953).....................60
Figure 1.34 MET station for climatic variables measurement .......................................................64
Figure 1.35 Class A Evaporation pan............................................................................................65
Figure 1.36 Wind Speed measuring device installed at 1st class meteorological station in
Haromaya University .....................................................................................................................66

5
Figure 1.37 Sunshine hour recorder ...............................................................................................66
Figure 2.1 Location of the Blue Nile Basin in Ethiopia together with climatic stations ...............68
Figure 2.2 Details of meteorological stations in the Tana subbasin system ..................................69
Figure 2.3 Presentation of time series distribution of climatic variables in Tana basin system ....71
Figure 2.4 Batch file prepared for the analysis of rainfall .............................................................72
Figure 2.5 Status of rainfall data in the Tana basin system............................................................72
Figure 2.6 Seasonal distribution of rainfall in the Tana basin system ...........................................74
Figure 2.7 Spatial rainfall distribution of rainfall over Ethiopia....................................................74
Figure 2.8 Spatial and temporal distribution of rainfall over the Tana basin system ....................75
Figure 2.9 Computation of missing flow records in the Tana basin system .................................75
Figure 2.10 Seasonal flow variations in the Tana Basin Catchment..............................................76
Figure 2.11 Components of streamflow Hydrograph.....................................................................77
Figure 2.12 Streamflow Hydrograph at (a) Abay at Bahir Dar and (b) at Gumara gauging station
........................................................................................................................................................78
Figure 2.13 Processes simulated by SWAT ...................................................................................79
Figure 2.14 Schematic of pathways available for water (After Neitsch et al., 2002) ....................81
Figure 2.15 . Flow hydrograph of observed and simulated flow from Anjeni watershed during
calibration period............................................................................................................................83
Figure 2.16. Sediment hydrograph of observed and simulated sediment from Anjeni watershed
during calibration period ................................................................................................................83
Figure 2.17 Comparison of daily predicted and measured discharge during model calibration
period..............................................................................................................................................84
Figure 2.18 Observed and Simulated Blue Nile daily discharge ...................................................84
Figure 2.19: Runoff generation process in HEC-HMS ..................................................................86
Figure 2.20 Observed and simulated (modelled) hydrographs for Laelay Wukro watershed .......86
Figure 3.1 Suspended Sediment Rating Curve of Bilate River at Alaba Kulito ............................93
Figure 4.1Fitting the Observed values of 10min and 30 min data ...............................................104
Figure 4.2 Fitting the Observed values of 1hour and 24 hour......................................................104
Figure 4.3 Fitting the Observed rainfall data of 10 minute and 30 minute duration using Log
Pearson Type III distribution........................................................................................................106
Figure 4.4 Fitting the Observed values of 1 hour, and 24 hour data using Log Pearson Type III
distribution. ..................................................................................................................................106
Figure 4.5 Fitting the Observed values of 10min, and 30 min data using lognormal distribution.
......................................................................................................................................................106
Figure 4.6 Fitting the Observed values of 1 hour, 24 hour data using lognormal distribution. ...107
Figure 4.7 IDF Curves Plotted On a Log-Log Graph. .................................................................112
Figure 4.8 SCS Relation between Direct Runoff, Curve Number and Precipitation ...................117
Figure 4.9 Unit Peak Discharge, applicable to Ethiopian catchments (Type II rainfall). Source
(ERA Manual) ..............................................................................................................................119
Figure 5.1 Depth-capacity curve of Dendo Reservoir for the estimation of shape factor, m
(Borland and Miller method)........................................................................................................124
Figure 5.2 Topographic map of Laelay Wukro reservoir.............................................................126
Figure 5.3 Area capacity curve for Laelay Wukro reservoir........................................................127
Figure 6.1 The Hershfield’s chart for determination of frequency factor Km .............................131
Figure 6.2 Depth area reduction curve .........................................................................................133
Figure 7.1 Schematic representation of DEM preparation...........................................................136
Figure 7.2 Digital Elevation Model for Blue Nile Basin Watershed ...........................................137

6
Figure 7.3 Distribution of slopes across the watershed................................................................138
Figure 7.4 : Flow chart to derive catchment characteristics using HEC-GeoHMS .....................140
Figure 7.5 Procedures adopted in deriving catchment characteristics for Gilgel Abay Catchment
......................................................................................................................................................141
Figure 7.6 GIS Application to identify DAM site and catchment river network for Gilgel Abay
Catchment.....................................................................................................................................142
Figure 7.7 Methodology adopted in digital image processing .....................................................144
Figure 7.8 An example of Image attributes (P166R68) ...............................................................145
Figure 7.9 Example of Resampling procedure in IDRISI ANDES and Reference Parameters
window .........................................................................................................................................146
Figure 7.10 Example of geo-referenced true color composite of a satellite image: band 345 of
1990 with X, Y coordinates and overlaid catchment. ..................................................................147
Figure 7.11 Example of a mosaicked and geo-referenced true color composite of satellite images
P166R65 and P167R65 overlaid with catchments. .....................................................................147
Figure 7.12 Training sites development.......................................................................................148
Figure 7.13 Preparation of spectral Signature Files for the trained sites. ....................................148
Figure 8.1 Pitman Catchment Model Flowchart and Parameters for optimization (Pitman, 1973)
......................................................................................................................................................154
Figure 8.2 Illustration of the default basin property and parameter estimation program.............156
Figure 8.3 Conceptualization of the subsurface drainage that determines the Interflow process
from the unsaturated zone. ...........................................................................................................157
Figure 8.4 Estimation of the vector slope using parameter estimation software. ........................158
Figure 8.5 Estimation of Parameter K..........................................................................................159
Figure 8.6 Estimation of Parameter T (m2/day) .........................................................................159
Figure 8.7 Illustration of the concept of using a frequency distribution to describe the spatial
distribution of soil moisture for different mean moisture contents..............................................160
Figure 8.8 Estimation of POW using default estimation programme ..........................................161
Figure 8.9 Estimation of Infiltration parameters..........................................................................162

7
 List of Tables

Table 1.1 Guidelines to establishing a rain gauge..........................................................................15

Table 1.2 Recommended minimum densities of meteorological network (WMO, 1994) .............16
Table 1.3 Method of estimation for point precipitation from nearby gauging sites.......................18
Table 1.4 Details of rainfall data at station A and at surrounding .................................................22
Table 1.5 Correction applied to the annual rainfall data of station A ............................................24
Table 1.6 Details of rainfall data over a catchment........................................................................28
Table 1.7 Data for areal estimation using Isohytal method............................................................31
Table 1.8 Recommended minimum densities of streamflow stations............................................34
Table 1.9 Current meter gauging of River .....................................................................................39
Table 1.10 Water level and current meter data ..............................................................................43
Table 1.11 Test statistics results for rating curve determination....................................................49
Table 1.12 Flow data for FDC construction...................................................................................52
Table 1.13 Recommended minimum densities for sediment stations............................................57
Table 1.14 Relationship of trap efficiency and capacity inflow ratio ............................................61
Table 1.15 Minimum density of evaporation stations (WMO, 1994)............................................65
Table 2.1 Summary of Meteorological data (After Dereje et al., 2006) ........................................67
Table 2.2 Locations of meteorological stations in the Tana basin system.....................................69
Table 2.3 Summary of Key Climatic Characteristics for Bahir Dar Station..................................70
Table 3.1 Relationship of trap efficiency and capacity inflow ratio ..............................................94
Table 4.1 Annual Maximum Rainfall Data at Bahir Dar Observatory Station ..............................98
Table 4.2 Maximum Rainfall Intensity of 10 Minute Rainfall Duration. ......................................98
Table 4.3 Maximum Rainfall Intensity of 30 Minute Rainfall Duration. ......................................99
Table 4.4 Maximum Rainfall Intensity of 1 hour Rainfall Duration. ............................................99
Table 4.5 Maximum Rainfall Intensity of 24 hour Duration. ........................................................99
Table 4.6 Plotting Position and reduced Variate for EVI distribution .........................................103
Table 4.7 The standard normal variables, Z for Lognormal and Log Pearson type III distributions.
......................................................................................................................................................105
Table 4.8 Mean and Standard deviations of rainfall data.............................................................108
Table 4.9 Fitting an EVI distribution for 10 minute Annual maximum rainfall..........................109
Table 4.10 Mean and Standard deviations of rainfall durations ..................................................110
Table 4.11 Computation of the 30 Minute Rainfall Intensity Using Lognormal and Log Pearson
Type III Distribution ....................................................................................................................110
Table 4.12 Rainfall Intensities as Computed using Log Pearson Type III. .................................112
Table 4.13 Ia Values for Runoff Curve Numbers ........................................................................118
Table 5.1 Estimated Monthly Parameter Values of the Thomas-Fiering Model .........................121
Table 5.2 Classification of Reservoirs (Ghosh, 1999) .................................................................123
Table 5.3 Sediment Distribution for 50 years of operation with 95% trap efficiency (example at
Dendo Reservoir on Bilate River)................................................................................................128
Table 6.1 Procedure for PMP estimation at Mertolemariam station for 24 hrs duration.............132
Table 7.1 Catchment characteristics of catchments in the Tana basin system.............................139
Table 7.2 Maximum and minimum coordinates of the image .....................................................145
Table 8.1 Soil texture classes according to USDA (1969), based on percentage volumes of sand,
silt, clay and quartz content..........................................................................................................156
Table 10.9 Typical catchment characteristics and Pitman model parameters..............................163

8
 Summary

The present hydrology guide book was prepared primarily to be used as a quick
reference for water resources professionals of the ANRS water bureau for the
planned training program. The presentation of the book starts with hydro-
meteorological data types that are widely used in water resources practices
and their availability in the region both spatially and temporally. Methods of
data analysis including consistency check-ups and standard data transmission
mechanisms are outlined and discussed. The use of hydro-meteorological data
for various practices that mainly include flooding, reservoir sedimentation and
ungaged catchment analysis are briefly and widely captured in the hydrologic
guide book. The use of the state of the art technology that involves GIS and
Remote Sensing applications in water resources practices are also given due
emphasis in the guide book. Mechanisms of deriving physiographic information
that involves catchment characteristics and their relationship with hydrologic
model parameters are discussed. In the absence of measured hydrologic data
from catchments, recommendations are made to develop synthetic information
from standard design documents in the region and other literatures that are
widely applied for various physiographic and climatic regions. Principles of
Probable Maximum Precipitation (PMP) and Probable Maximum Flow (PMF) that
are widely used for design are discussed and method of their computation is
presented. Synthetic flow generating mechanisms are also presented together
with practical examples. In each topic of the book several examples are
provided that enable the reader of the book to easily capture the principles and
be able to adopt the techniques easily.
It is understood that hydrology and hydrological analysis are complex and the
uncertainty in hydrologic variables prediction is quite high particularly in areas
where network gaging stations are limited in number and lack representativness.
Hence this report outlined techniques based on standard approaches for hydro-
meteorological network density setups at various physiographic and climatic
conditions. Moreover as it is difficult to measure every hydro-meteorological
variable in the region, for reliable hydrologic variable predictions, hydrologic
models that are widely accepted and being used for various applications are
discussed and their use in the region is recommended. The choice of the model,
however, is based on the problem statement.
In this work the consultant tried to capture wider aspects of water resource
practices and common principles of hydrology that will give invaluable
information for the intended training program. However, the consultant admits
that the work is not a full-fledged by its own and can be amendable further.

9
 CHAPTER ONE

HYDRO-METEOROLOGICAL DATA MEASUREMENT AND QUALITY CHECK

1.0 Introduction
This chapter is devoted to assess the mechanism of hydro-meteorological data
measurement and processing. Means of data collection, data organization, and
transmission and quality checks are outlined. The practice helps the choice of
quality data to estimate hydrological variables in hydrological processing.
Hydro-meteorological data refers to hydrological data and meteorological
data. Hydrological data includes data of streamflow, sediment flow and flow
gauge heights and current meter readings. Meteorological data refers to the
average state of the atmosphere during a period of time (days, weeks,
decades, years or millennia), and among the meteorological data, solar
radiation is the major external energy source for the climatic system (Rasmusson,
et al., 1993).

For hydrological applications, the most significant meteorological variables are

rainfall, temperature, evaporation, and evapotranspiration in the order of
progressive levels of processing complexity (WMO, 1994). Hence proper data
quality assessment and mode of transmission needs to be practiced to establish
the best estimate of the climate during some future period of interest.

The steps followed to conduct preliminary checking of hydrological data

collected manually include the following according to WMO standard:

a) Log in data at the time of the receipt of the report form;

b) Ensure completeness and correctness of the information, i.e., dates,
station name, and station identification if required in subsequent machine
processing;
c) Ensure completeness of the data;
d) Check the observer’s arithmetic; and
e) Compare the observer’s report with the recorded data.

1.1 Data and Data Organization

Hydro-meteorological data can be collected and is available in hard copy and
soft copy formats. Hard copy data formats need to be converted into softcopy
formats for easy dissemination and analysis. The most common and widely
applied soft copy format of data storage is an excel sheet. However to avoid
too many complications data needs to be stored in standard formats that
presents the detail of the data.
In this chapter the consultant will establish a standard data storage format that is
easily understood by the user and help easy dissemination. An example file

10
(suspended sediment data) from Koga watershed was used for the
demonstration purpose. The remaining hydro-meteorological data can also be
prepared following the same procedure.
Suspended sediment data was obtained in disorganized manner (as shown in
Figure 1.1) and attempt was made in this work to standardize the data. Thus the
daily data of suspended sediment was organized in a manner that is easy for
dissemination. The procedure followed is as follows:
1. Convert the excel file into Notpad format,

2. Convert the Notepad data into standard file format

Figure 1.1 Raw suspended sediment data

Following the above procedures, the following tasks have been done:
1. Standardizing data format for easy dissemination;

2. Identification of Missing Records and

3. Computation of seasonal distribution (Date Averaged Mean and Annual)

1.1.1 Standardization of data format for easy dissemination; UCG Format

The raw data was formatted as shown in Figure 1.2. The first three lines in the file
store miscellaneous information regarding the data series. The actual records of
the data series start from the fourth line in a sequential order.

The interpretations of the information and the numbers to be entered on the first
three lines are described as follows: The first line contains the title of the data file

11
indicating the contents of the file, the catchment name and the type of
hydrological variable. In the second line the first integer variable defines the file
type. It is always 1 for time series data. The second integer variable is set equal
to 1 if the data in the file starting from the fourth line onwards is in free format. If
the data starting from the fourth line in the file are formatted using a format (2X,
6E13.5), according to FORTRAN programming language convention, then this
second number is set to zero.

Figure 1.2 Suspended sediment data (For Koga Watershed)

In the third line the first number indicates the length of data. This number is
15,705 in the data file (Figure 1.2). The second number represents the year
corresponding to the first element of data in the series of data. This is 1960 in the
example file. The third number represents the month of the first element of data.
This is 1 in both of the example files, indicating the month of January.
The fourth number represents the day of the first element of data. This is 1 in the
case of the data series shown as examples, which indicates that the starting day
of the data series is “the first day of the month”. Thus taken together, the second,
the third and the fourth number in the example files indicate the starting date for
the data series, which is “the First of January, 1960”. The fifth, the sixth and the
seventh numbers in the third line represent respectively the hour, the minute and
the second corresponding to the first record in the data series. These three
numbers are zero in each of the example files, which indicate that the first
record in the data series corresponds to the time of 0hr/0min/0sec. The eighth
number represents the time interval of data in seconds. If the data are daily,
then this number is to be entered as 86,400. In the example files, the data
represent daily records. The number in the ninth place corresponds to the
position of the first occurrence of a 29th February value in the data series. For the

12
example data file, this number is 60, which is the sum of the numbers 31 (January
and 29 (February). The tenth number indicates the data type. It is zero for
sampled data and 1 for data integrated over the time step (i.e. the data
interval). In the time series shown in Figures 1 this number is ‘1’ for each case
indicating data averaged over a day. The last (eleventh) number in the third line
is the code for indicating data units. Standard coding is as follows:

1 for millimetre
2 for cumec (metre3/sec)
3 for inch and
4 for cusec (ft3/sec)
5. Other unit (depending on data type)

The actual record of the data series starts from the fourth line.

1.1.2 Identification of Missing Records

Missing values were identified using a FORTRAN code and results are presented
below. The advantage of using Fortran programming language is multiple files
can be processed at once by having one batchfile. This avoids the tiresome now
and then computation using for example, excel file.

Figure 1.3 Presentation of Percentage Missing from data of Koga watershed

1.2 Precipitation Data

Precipitation is an important hydrological variable that is notoriously variable
spatially. In the tropical regions precipitation is equal to rainfall since there is no
snowfall. Information required in estimating the influence of precipitation ranges
from general regional and seasonal variability to the frequency and magnitude
of individual storm events.
Rainfall runoff modelling depends heavily on the records from point rain gauges
and the estimation of rainfall is very important, since, no model, however well-
founded in physical theory or empirically justified by past performance, will be
able to produce accurate hydrograph predictions if the inputs to the model do
not adequately characterise the rainfall inputs (Beven, 2006).
Rainfall volumes and intensities can vary rapidly in space and time. Thus, as well
as an interpolation of rainfall in volumes in time to produce the storm profile, it
may also be necessary to interpolate in space since rain gauges measurements

13
represent only point measurements. Techniques that are quite common for
spatial interpolation of rainfall include Kriging, Thiessen polygons, Inverse
distance weighting and arithmetic mean. But none of these techniques can be
more than an approximation to the actual volume of rainfall over the
catchment, and accuracy of a particular technique is likely to change from
storm to storm.

1.2.1 Precipitation Data Measurement

The most common means of measuring precipitation is a rain gauge, either
storage gauges or recording gauges. A storage gauge is one that captures and
stores precipitation until a measurement is made, often daily. Each time a
measurement is made the daily gauge is emptied and again situated for
precipitation capture. On the other hand, there are various types of recording
gauges, which either measure or weigh precipitation and automatically registers
or records the data on a chart or in a computerized data logger. Recording
gauges are capable of recording intensity (rate of rainfall accumulation) in
addition to total rainfall over a period of time. Figure 1.4 illustrates different
varieties of both types of rain gauges.

a b c

Figure 1.4 Examples of non-recording and recording rain gauges: a) Standard non-
recording precipitation gauge. b) Recording, weighing-type precipitation
gauge. C) Recording, Tipping bucket gauge

For weight recording gauges, the mass of rainfall must be converted to a

representative depth, and for a tipping bucket, the volume or mass capacity of

14
a single tip of the bucket must be calibrated to a depth. In addition, recording
rain gauges can be calibrated for particular time intervals between
measurements. Other calibrations or adjustments may be necessary to
compensate for wind, slope, and other environmental factors.
In order to determine the correct depth of precipitation in a catchment, it is
important to determine optimum number of rain gauges. The optimum number
of rain gauges (N) can be determined using equation 1.

2
⎡ Cv ⎤
N =⎢ ⎥ [1]
⎣E⎦
Where Cv is a coefficient of variation of rainfall based on the existing rain gauge
stations, E is the allowable percentage error in the estimate of basic mean
rainfall.
The particular location of a rain gauge is important for the collection of accurate
data. Table 1.1 describes some of the guidelines to follow when sitting a
precipitation gauge based on WMO Standard. For example, wind creates the
most serious disturbances in and around gauges and often affects the capture
of rainfall. Obstacles, such as trees and buildings, and slope also affect
measurements.

Table 1.1 Guidelines to establishing a rain gauge.

Environmental Guideline for rain gauge establishment

effect
Install as close to the ground as possible. (Suggestion for
extremely windy locations: Install rain gauge in pit dug into
Wind ground with rim of gauge level with ground).
Locate near low bushes and shrubs that may serve as wind
breaks but will not deflect rainfall.
Equip gauge with a wind shield or deflector.
Locate gauges away from buildings and trees (the
distance should be equal or greater than 2 times the
Obstructions height of the obstacle, 4 times is preferable).
Check gauges frequently for debris, which prevents
continuous precipitation capture
Locate on level ground with top of gauge horizontal (a
surveyor’s level can be used to level the gauge).
Slope On steep slopes, install gauge with the top of the gauge
parallel to slope. Several gauges may be necessary to
control extensive variability.

Considerations must be made to determine the number of gauges necessary to

provide enough reliable and accurate data. Reasons for measuring

15
precipitation, area represented by individual gauge, allowable error for
measurements, topography, and climate are all factors to consider when
determining the number of gauges to establish. WMO (1994) presents a
recommendation on the minimum densities of meteorological network (Table
1.2).

Table 1.2 Recommended minimum densities of meteorological network (WMO,

1994)

Region Minimum Density (km2/station)

Non-Recording Recording

Mountainous 250 2500

Interior plains/ Flat regions 575 5750

Under difficult condition

Hilly/undulating 575 5750

Arid Zones* 10,000 100,000

1.2.2 Point precipitation and missing data estimation

A network of gauges providing point precipitation data is beneficial because it is

sometimes necessary to estimate point precipitation at a given site from the
data of nearby gauges. Data may be missing from a particular gauge site or
representative precipitation is necessary at a point of interest. For the missing
data to be estimated or representative precipitation to be determined, the
location of the site must be known in relation to nearby sites. For example in the
Figure 1.5 of the Tana basin system the mean annual precipitation (MAP) at a
point X with geographic coordinates of 370124mE and 1315778mN is to be
determined from the MAP of the nearest six gauging sites.

Procedures to estimate missing information:

1. Develop the coordinate system of point X and other rain gauge stations and
determine the absolute coordinates of available rain gauges with respect to the
unknown point X.

2. Construct a table with a column of the gage and point identification, a

column for rainfall data from each gauging stations and two columns for each

16
of the points change in X (∆X) and change in Y (∆Y) in relationship to point X
(Table 1.3).

3. The estimated rainfall at X is a weighted average of the other points, and the
weights (W) are reciprocals of the sums of the squares (S2) of ∆X and ∆Y.

S 2 = ∆X 2 + ∆Y 2 [1.2]

W = 1/ S 2 [1.3]

4. Then, the estimated rainfall at a point of interest is given by Σ (P x W)/ Σ W

where P is the known precipitation at a point.

Figure 1.5 Coordinate system representing the location and distance relationship of
Point A and its nearby gauging sites, Points B, C, D, E, and F.

17
Table 1.3 Method of estimation for point precipitation from nearby gauging sites.

X Y DX DY S W W*MAP
Rain gauge MAP Projection Projection
Addis
Zemen 1101.31 377036.28 1340077.7 6912.278 24299.67 25263.68 3.95825E-05 0.043593
- -
Bahir Dar 1420.88 327732.7 1282800.7 42391.31 32977.26 53707.75 1.86193E-05 0.026456
- -
Gumara 1549.49 350756.82 1308121.3 19367.18 7656.673 20825.76 4.80175E-05 0.074403
-
Yifag 1027.22 360685.46 1334619.6 9438.537 18841.56 21073.45 4.74531E-05 0.048745
-
Woreta 1392.63 356257.01 1319155.5 13866.99 3377.519 14272.39 7.00654E-05 0.097575
Debre
Tabor 1504.4 394370.69 1317892.1 24246.69 2114.111 24338.68 4.10869E-05 0.061811
PX ?? 370124 1315778 Sum 159481.7 0.000264825 0.352582
PX 1331.379

[Link] Missing rainfall estimation using Arithmetic mean

The simplest and most direct approach to determine missing data and average
depth of precipitation is by taking the average of the data of the various nearby
gauging sites.
P + P + P3 + ...Pn
Pav = 1 2 [1.4]
n

Pav is average rainfall over the area, Pi is rainfall at different gauges, n is number
of gauges in the area. The method is used for flat areas, uniformly distributed
gauges over the area and for rain collected at different stations not varying
widely from the mean.

[Link] Missing rainfall estimation using Normal Ratio Method

Data may be missing from a particular gauge site or representative precipitation
is necessary at a point of interest. The method described in section 1.2.2
concerned the distance relationship of nearby rain gauging stations. However,
determining the arithmetic mean (or, finding the average) of the data from the
nearby gauging stations may also be applied as presented above. In addition
to the arithmetic mean, another method called the normal ratio method can be
utilized to determine missing data. This method can be used over the arithmetic
mean if the percent difference between the normal annual precipitation of the
missing data station and each of the normal annual precipitation of the nearby
stations is greater than 10%.

18
 ⎛ N − Ni ⎞
%difference = ⎜⎜ x ⎟⎟ x100
⎝ Nx ⎠ [1.5]
where Nx is the normal annual rainfall amount from the missing data station
Ni is the normal annual rainfall amount from the one of the nearby
stations.
The Nx – Ni must be positive. If Ni >Nx the numerator will become Ni – Nx. Then,
the mean of the nearby stations’ differences is determined.
If the mean difference is less than 10%, the arithmetic mean is sufficient for
estimating the missing data. On the other hand, if the mean difference is
greater than 10%, the normal ratio method should be used.

⎛ 1 ⎞⎡ N N N ⎤
Px = ⎜ ⎟ ⎢ x P1 + x P2 + .... + x Pn ⎥
⎝ n ⎠ ⎣ N1 N2 Nn ⎦
[1.6]

where Px =missing rainfall data at station x, Nx is the missing data station’s normal
annual rainfall, Ni=normal annual rainfall at station i, and n is the number of
nearby gauges.

Example 1
Determine whether the missing data (grey shaded cell) should be estimated
using the arithmetic mean method or the normal ratio method. According to
the appropriate method for each data set, estimate the particular missing data.

Example Data 1: Data from Region 03 Data 2: Data from Region 01

Station Rainfall at Normal Station Monthly Normal
ID one storm annual ID data, July, annual
event, rainfall (mm) 1980 (mm) rainfall
(mm) (mm)
Ribb 107 1120 Hawzien 195.7 446
[Link] 89 935 Ilala 186.5 452
Gumara 122 1200 Adigrat 243.3 512
B. Dar ? 978 Quiha ? 472

Solution:
The initial step of solving this problem is determining the difference between
each of the stations normal annual rainfall (Ni) to the missing data’s station
normal annual rainfall (Nx). Determining the percent difference is determined
using equation 1.4..

19
 Data 1: The three stations’ Data 2: The three stations’
differences, the mean of the differences and the mean of the
differences, and the necessary differences, and the necessary
method to use for missing data. method to use for missing data.
Ribb: ((1120 – 978)/978)* 100 = Hawzien: ((472 – 446)/472)*100 =
14.5% 5.5%
[Link]: ((978 – 935)/978)* 100 = Ilala: ((472 – 452)/472)*100 = 4.2%
4.3% Adigrat: ((512 – 472)/472)*100 =
Gumara: ((1200 – 978)/978)* 100 8.5%
= 22.7% Mean difference: (5.5 + 4.2 +
Mean difference: (14.5 + 4.3 + 8.5)/3 = 6.1%
22.7)/3 = 13.8%
Difference < 10% Î Arithmetic
Difference > 10% Î Normal Ratio mean method
method

Data 1: The normal ratio method: Equation 1.4 is employed with Nx = 978mm, n =
3, N1 = 1120mm, N2 = 935mm, N3 = 1200, P1 = 107mm, P2 = 89mm, and P3 =
122mm.
⎛ 1 ⎞ ⎡ 978mm 978mm 978mm ⎤
Px = ⎜ ⎟ ⎢ x107mm + x89mm + x122mm⎥ = 95.3mm
⎝ 3 ⎠ ⎣ 1120 935mm 1200mm ⎦

For comparison, the missing data in Example Data 1 determined by the

arithmetic mean is 106mm.

Data 2: The arithmetic mean method: For the 3 monthly precipitation depths, the
average is determined and applied as the missing data.
Px = Missing monthly data from Quiha = (195.7mm + 186.5mm + 243.3mm) / 3 =
208.5mm

For comparison, the missing data in Example Data 2 determined by the normal
ratio method is 208.7mm

1.2.3 Consistency of Precipitation Data

Numerous factors could affect the consistency of rainfall record at a given
station. Among the factors are:
• Damage and replacement of a rain gage
• Change in the gage location or elevation
• Growth of high vegetation or construction of a building
• Change in measurement procedure, or
• Human, mechanical, or electrical error in taking readings

20
A method called Double Mass Curve Analysis (DMC) is used to adjust
inconsistent data. The procedure consists of comparing the accumulated
annual (or seasonal, monthly, weekly, daily, or hourly) precipitation at the station
in question with the accumulated annual (or seasonal, monthly, weekly, daily, or
hourly) precipitation for a group of surrounding stations. Figure 1.6 shows an
example of the accumulated annual precipitation of a station X and that of
surrounding stations.

Figure 1.6 DMC indicating the accumulated annual precipitation of a station X and
that of surrounding stations

A break in the double mass curve (DMC) of Figure 1.6 indicates that a change
occurred in that year. Adjusting the data to the condition before change is
equivalent to bringing the observed points on the line after the change vertically
to the line that is the extension of the line before the break. If the slope of the line
prior to the change is m1, and the slope after the change is m2, then the
adjustment factor will be m1 / m2 . The multiplier will be applied to the post-
change observed data to obtain the adjusted data.

Example 2
Suppose that the annual precipitation of station A and mean annual
precipitation of stations in the Tana basin system are given in Table 1.4. Analyse
the consistency of the rainfall data at station X and adjust wherever appropriate.

21
Table 1.4 Details of rainfall data at station A and at surrounding

Annual Mean Yearly of Cumulative of

rainfall of Cumulative Surrounding Surrounding
Year A of A stations stations

1996 1430 1430 1410 1410

1995 1100 2530 1260 2670

1994 1170 3700 1100 3770

1993 1100 4800 1230 5000

1992 1200 6000 1150 6150

1991 1220 7220 1430 7580

1990 1280 8500 1150 8730

1989 750 9250 950 9680

1988 1120 10370 1230 10910

1987 1250 11620 1350 12260

1986 1380 13000 1440 13700

1985 1210 14210 1360 15060

1984 1760 15970 1730 16790

1983 1400 17370 1080 17870

1982 1240 18610 970 18840

1981 1760 20370 1320 20160

1980 1480 21850 1350 21510

1979 1740 23590 1410 22920

1978 1420 25010 1270 24190

1977 1580 26590 1260 25450

22
Solution: From the data of cumulative annual rainfall and cumulative mean
annual rainfall of surrounding stations, DMC is plotted as shown in Figure 1.7 as a
first step. Then analyse the plot to indicate any change in slope of the line. As
shown in Figure 1.7, there is a change in the slope of the line and the adjustment
will be as follows:

Slope of Original line = (15970-1430)/ (16790-1410) = 0.945

Slope of Deviated line= (26590-15970)/ (25450-16790) = 1.226
Difference in slope = (1.226-0.945)/0.945 = 29.72% > 10%. Hence correction is
necessary.
Correction = 0.945/1.226 = 0.771
The correction applied on the annual rainfall data of station X is shown in Table
1.5 and Figure 1.8.

Double mass curve

30000

25000
Cummulative of doubtful station X

20000

15000

10000

5000

0
0 5000 10000 15000 20000 25000 30000
Cummulative of Surrpounding stations

Figure 1.7 Double Mass Curve for station A and surrounding stations

1.2.4 Areal Precipitation

While point precipitation is important for estimating missing data or checking the
consistency of data from an individual gauge, it is also valuable in
understanding areal precipitation distribution and developing information over a
large area compared to a single location.
Average precipitation depths for a watershed or other division of land area are
often determined through the use of point data from multiple, nearby sites
(Figure 1.8).

23
Table 1.5 Correction applied to the annual rainfall data of station A

Annual rainfall of
Year A corrected cumulative

1996 1430 1430 1430

1995 1100 1100 2530

1994 1170 1170 3700

1993 1100 1100 4800

1992 1200 1200 6000

1991 1220 1220 7220

1990 1280 1280 8500

1989 750 750 9250

1988 1120 1120 10370

1987 1250 1250 11620

1986 1380 1380 13000

1985 1210 1210 14210

1984 1760 1760 15970

1983 1400 * 0.771 1079 17049

1982 1240 * 0.771 956 18005

Correction
1981 1760 * 0.771 1357 19362

1980 1480 * 0.771 1141 20503

1979 1740 * 0.771 1341 21844

1978 1420 *0.771 1095 22939

1977 1580 * 0.771 1218 24157

24
 Double Mass Curve

30000 P1
Cumulative of X after

25000

20000 P3
correction

P5 Average Precipitation
15000 P2
over Rain Gauge
10000 Network
5000 P4
0 P6
0 5000 10000 15000 20000 25000 30000
Cummulative of Surrounding Stations

Figure 1.8 Corrected Double Mass Curve and presentation of an outline for areal
precipitation computation.

The three methods that are commonly used to determine average rainfall depth
are:
• Arithmetic mean,
• Thiessen method, and
• Isohyetal method.
Arithmetic mean method is explained in section [Link] and descriptions of the
other two are explained below:

[Link] Thiessen method

Handles non-uniformly distributed gauges and makes use of weighing factors for
each gauge. This method is less effective for mountainous terrain.
Steps of determining areal rainfall using Thiessen polygon method are as follows:

Step 1: Draw the area concerned to a suitable scale, showing its boundary,
locations of the rain gauges in the area and outside but close to the boundary.

25
 P1

P2

P4 P5
P3

P6
P7

Figure 1.9 Representation of point rainfall in a watershed for areal computation

using Thiessen polygon method.

Step 2: The area of concern is divided into polygons with rain gauges as centers.
Join location of the rain gauges to form a network of triangles

P1

P2

P4
P5
P3

P6

P7

Figure 1.10 Polygonizing point rainfall data for areal precipitation estimation using
Thiessen Polygon method

Step 3:Draw perpendicular bisectors to the triangle sides. These bisectors form
polygons around the stations.

26
 P1

P2

P4 P5
P3

P6
P7

Figure 1.11 Perpendicular bisectors along the triangle sides.

Step 4: Delineate the formed polygons and measure their areas using a
planimeter or by converting them into smaller regular geometric shapes (i.e.
triangles, squares, rectangles, etc.)

P1
A1

A2 P2

A4

P4
A3 P5
A5
P3

A6

P6 A7

P7

Figure 1.12 Polygons drawn to indicate influence areas of point rainfall data

Step 5:
Compute the average rainfall using the following formula:

P1 * A1 + P2 * A2 + ..... + Pn * An
Pav = [1.7]
A1 + A2 + ...... + Pn

27
Example 3
Suppose that the area of each polygon in Figure 1.10 together with magnitudes
of each point rainfall data is shown in Table 1.6. Determine the areal rainfall over
the catchment.

Table 1.6 Details of rainfall data over a catchment

Observed Polygon Pi*Ai

Rainfall (mm) Area, Sqkm
Pi Ai

10 25 250

15 125 1875

20 80 1600

17 90 1530

25 120 3000

40 115 4600

12 130 1560

Total 685 14415

The rainfall from each gauge is weighted according to the extent of its
coverage of the area of concern, such as a watershed.
Hence the mean areal rainfall is 14415 / 685 = 21.043mm

[Link] Isohyetal method

Isohyetal comes from the Greek roots –iso meaning equal and hyet meaning
rain; therefore, isohyetals are lines of equal rainfall. They are drawn the same
way as topographic contour lines are drawn. The rain gauges provide the initial
data (benchmark or control) necessary to develop isohyetals. The controls are
mapped accordingly, and between the control data, the location of isohyetals
at specific intervals (10mm, 20mm, 1cm, 2cm etc.) are interpolated. This
method, although more difficult, is often the most accurate. However, those
creating the isohyetals should know the terrain of the area to effectively locate
the isohyetals according to natural conditions. This method is preferred on
rugged, mountainous terrain.

28
Figure1.13 Methods of computing areal average precipitation for a specific area,
or drainage basin.

The numbers in Figure 1.13 represent precipitation data measured at each of the
rain gauges. a) Area outline and data values, available for arithmetic mean, b)
Thiessen polygons, and c) isohyets, or lines of equal precipitation. Steps followed
in estimation of areal rainfall using Isohyetal method are as follows:

Step 1: Draw the area under study to scale and mark rain gauges on it. Put at
each of the rain gauge location the recorded values of rainfall at the station, for
the period within which the average is required to be determined (Figure 1.14).

Figure 1.14. Representation of point rainfall for areal estimation using Isohyetal
method

29
Step 2: Draw the isohyets of various values by considering the point rainfall data
as guidelines and interpolating between them. Also, incorporate the knowledge
of orographic effects.

Figure 1.15 Isohyets drawn to for areal estimation

Step 3: Determine the area between each pair of the isohyets lines, either by a
planimeter or by converting the areas into smaller regular geometric shapes.

Step 4: Calculate the average rainfall using the following formula:

A ( P + P2 ) / 2 + A2 ( P2 + P3 ) + ....... + An −1 ( Pn −1 + Pn ) / 2
Pav = 1 1 [1.8]
A1 + A2 + ...... + An
Pi = Value of Isohyet lines
Ai = Area between pair of isohyet lines.

Exercise:
Calculate the average rainfall over the area given in the Figure 1.16 using
Isohyetal method.

30
 16.5 25
50

48.7 75
37

68.3 100

125
39 75.7 125

49.5

44.5

Figure 1.16 Example data file

Table 1.7 Data for areal estimation using Isohytal method

Isohyte Area Between Average Rainfall

Isohytes ([Link]) Rainfall Volume
(mm)
(mm)

125 33.28 125.0

100 197.12 112.5

75 296.96 87.5

50 501.76 62.5

25 494.11 37.5

31
1.3 Streamflow Data

Streamflow data, which is expressed as volume per unit time, is the rate at which
water flows through a cross-section. Discharge at a given time can be measured
by several different methods, and the choice of methods depends on the
conditions encountered at a particular site.
Continuous streamflow records are necessary in the design of water supply and
waste systems, in designing hydraulic structures, in the operations of water
management systems, and in estimating the sediment or chemical loads of
streams, including pollutants.
Since continuous measurement of discharge is not usually feasible, records of
discharge are computed from the relationship between stage and discharge, as
defined by periodic discharge measurements and a systematic record of stage,
or from a measuring structure that has been calibrated in either a laboratory or
the field.
Flow in a stream is measured by the combination of current meter and staff
gauge. The use of current meter and staff gauge is explained by the velocity
area method of stream discharge computation. In the velocity area method the
depth of flow in the cross-section is measured at verticals with a rod or sounding
line. As the depth is measured, observations of velocity are obtained
with a current meter at one or more points in the vertical. The measured widths,
depths, and velocities permit computation of discharge for each segment of the
cross-section. The summation of these segment discharges is the total discharge
(Figure 1.17).

Figure 1.17 View of stream cross-section showing the location of points of

observation

32
1.3.1 Selection of sites for stream gauging
Suitable sites should be identified for stream gauging. Sites selected for flow
measurements should ideally have the following characteristics (WMO, 1994).

a) The velocities at all points are parallel to one another and at right angles
to the cross-section of the stream;
b) The curves of distribution of velocity in the section are regular in the
vertical and horizontal planes;
c) The velocities are greater than 0.150 m s-1;
d) The bed of the channel is regular and stable;
e) The depth of flow is greater than 0.300 metre;
f) There is no aquatic growth; and
g) There is minimal formation of slush or frazil ice.

Similar to the above criterion, the selection of a particular site for the gauging
station on a given stream should be guided as follows:

a) The general course of the stream is straight for about 100 metres upstream
and downstream from the gauge site;
b) The total flow is confined to one channel at all stages and no flow
bypasses the site as sub-surface flow;
c) The stream bed is not subject to scour and fill and is free of aquatic
growth;
d) Banks are permanent, high enough to contain floods, and are free of
brush;
e) Unchanging natural controls are present in the form of a bedrock outcrop;
f) A site is available, just upstream from the control, for housing the stage
recorder where the potential for damage by water-borne debris is minimal
during flood stages.
g) The elevation of the stage recorder itself should be above any flood likely
to occur during the life of the station;
h) The gauge site is far enough upstream from the confluence with another
stream or from tidal effect to avoid any variable influences which the
other stream or the tide may have on the stage at the gauge site;
i) A satisfactory reach for measuring discharge at all stages is available
within reasonable proximity of the gauge site. It is not necessary that low
and high flows be measured at the same stream cross-section;
j) The site is readily accessible for ease in the installation and operation of
the gauging station; and
k) Facilities for telemetry or satellite relay can be made available, if required.

In many instances, it may be impossible to meet all of these criteria. Judgement

33
is then required to select the most suitable site for the gauge. The recommended
minimum densities of streamflow stations are presented in Table 1.8

Table 1.8 Recommended minimum densities of streamflow stations

1.3.2 Method of measuring streamflow

The most satisfactory determination of the runoff from a catchment is by
measuring the discharge of the stream draining it, which is termed as stream
gauging. A gauging station is the place or section on a stream where discharge
measurements are made. Some of the usual methods of stream gauging are
given below:
(a) Venturi flumes or standing wave flumes (critical depth meter) for small
channels.
(b) Weirs Q = CLH 3 / 2 [1.9]

where Q = stream discharge, C = coefficient of weir, L = length of weir, H = head

(depth of flow) over the weir crest

(c) Slope-area method:

Q = AV
V = C RS
1
V = R 2 / 3 S 1/ 2
n
1 A [1.10]
C = R1/ 6 , R =
n P

34
Where C is Chezy’s constant, M is Manning’s coefficient of roughness, R is the
hydraulic mean radius, A is the cross-sectional area of flow, P is the wetted
perimeter, S is the water surface slope (=bed slope).

(d) Contracted area methods:

The drop in the water surface in contracted sections as n bridge openings, canal
falls etc. is measured and the discharge is approximately given by:

Q = C d A1 2 g (∆h + ha ) [1.11]
Where C d is coefficient of discharge, A1 area of the most contracted section, ∆h is
difference in water surface between upstream and downstream ends, ha is the
head due to the velocity of approach.

(e) Area Velocity Methods

The area of cross-section of flow may be determined by sounding and plotting
the profile. The mean velocity of flow (V) may be determined by making velocity
measurements. The velocity distribution in a cross-section and in a vertical stream
is shown in Figure 1.18 (a) and (b) and as such the velocity measurement is as
follows:
Velocity measurements are done using current meters (Pigmy, Propeller and
Price types). The measure of mean velocity in a strip of cross-section is done by
three methods.

(a) One Point Method V = v0.6d [1.12]

v + v0.8 d
(b) Two Point Method V= 0.2 d [1.13]
2
(c) Mean depth in the strip

Figure 1.18 Velocity profile along a cross-section (a) and along a vertical (b)

35
[Link] Current Meter Gaugings
The current meter is an instrument, which has a rotating element which when
placed in flowing water; the speed of revolutions has a definite relation with the
velocity of flow past the element.
There are three types of current meters:
(i) Pigmy current meter, whose cup vane assembly is about 5 cm in
diameter and is used for measuring velocities in streams of depth 15 cm
or less,
(ii) The cup type, which consists of a wheel with conical cups revolving on
a vertical axis, and
(iii) The screw or propeller type consisting of vanes revolving on a horizontal
axis.

[Link] Rating of the Current Meter

The relationship between the revolutions per second (N, rps) of the meter and
the velocity of flow past the meter (v, m/sec) has to be first established, or if the
rating equation is given by the maker, it has to be verified. This process of
calibration of the meter is called rating of the current meter. The rating equation
is of the form:

V = aN + b [1.13]
Where a and b are constants (determined from rating of the current meter).

Once the rating equation of the current meter is known, actual stream gauging
can be done from bridges, cradle, boat or launch. The cross section of the
stream at the gauging site is divided into elemental strips of equal width b and
the current meter is lowered to a depth of 0.6d below water surface in shallow
depths (one-point method) and to depths of 0.2d and 0.8d (two-points method)
in deep waters (Figure 1.18b). Typical setup of current meter gauging is shown in
Figure 1.19

Figure 1.19 Typical picture of a current meter (Price current meter) and Current
meter gauging (Source: Raghunath (2006)

36
There are two methods of determining the discharge in each elemental strip:

(i) Mid-section method:

In this method, the vertical in which the velocity measurements are made (by
one-point or two-point method) is taken as the middle of the strip, and the water
depth (d) in the vertical (determined by sounding) is taken as the mean depth
of the strip, Figure. 1.20 (a). If b is the width of strip (usually same for all strips) then
the discharge in the elemental strip is given by:

∆Q = (bd )V 0.6d in the shallow strips [1.14]

v 0 .2 d + v 0 .8 d
∆Q = (bd ) x in deeper water strips [1.15]
2
Stream discharge, Q = ∑ ∆Q [1.16]

In this method, the discharge in two-triangular bits near the ends is not included
in the discharge computation.

a
b

Figure 1.20 Mid section (a) and Mean Section (b) Methods

(ii) Mean-section method:

The cross section is regarded as being made up of a number of segments
bounded by two adjacent verticals. If v1 is the mean velocity at the one vertical
and v2 the mean velocity at the adjacent vertical, and if d1 and d2 are the total
depths measured at verticals 1 and 2, and b is the horizontal distance between
verticals, then the discharge q of the segment is:

⎛ v1 + v 2 ⎞⎛ d1 + d 2 ⎞
b

q=⎜ ⎟⎜ ⎟ [1.17]
⎝ 2 ⎠⎝ 2 ⎠
The total discharge is obtained by adding the discharge from each segment;

Example 4

37
The following data was collected from a river catchment. Compute the river
discharge.

Rating equation of current meter:

v = 0.3 N + 0.05, N = rpm, v = velocity , ( m / sec), Re v. − Re volutions , Sec − timein sec onds

Solution:
The discharge in each strip, ∆Q = (bd )V , where V is the average velocity in each
strip, Figure 1.21. In the first and last strips (near the banks) where the depth is
shallow,
V = v 0.6 d , and in the other five intermediate strips (with deep water),
v0.2 d + v0.8 d
V= . Width of each strip, b=3m, mean depth of strip=d, and the total
2
discharge, Q = ∆Q = 20.6cumecs as computed in Table 1.9.

Figure 1.21 Discharge Computation

38
Table 1.9 Current meter gauging of River

1.3.3 Stage Discharge Rating Curve

A river is gauged by current meter throughout the rainy season (for about 3
months) at different stages (water levels) of the river. The water stage can be
read on the enamel painted staff gauges (gauge posts) erected at different
levels at a gauging station (Figure 1.22).

A curve is drawn by plotting ‘stream discharge ‘Q vs. gauge height h’ which is

called the ‘stage discharge rating curve’ (Figure 1.23).
From the rating curve, the stream discharge corresponding to staff gauge
readings taken throughout the year/s can be obtained, as long as the section of
the stream at or near the gauging site has not materially altered. Periodical
gaging (say, once in three years) are conducted to verify the rating curve, or to
revise the rating curve if any change in section has been noticed.

39
Figure 1.22 Gauge posts on river banks
(Photo by Birhanu, 2006)

40
 3.00
Q = 0.0168(H+0.98)6.6186
R2 = 0.987,
2.50

2.00
Water level (m)

1.50

1.00 Rating curve

Q(observed -1990 - 2000)

0.50

0.00
0 10 20 30 40 50 60
Discharge (m3/sec)

Figure 1.23 Rating Curve and Equation

The general rating curve equation is represented using equation (1.18).

Q = K ( H + Ho) n [1.18]
Where Q the discharge in cumecs, Ho is the height of zero measurement. K and
n are parameters that are constant and vary depending on stream
characteristics. Determination of rating equation at a known stream gauge can
be developed using equation 1.18 and parameters K, Ho and n can be
optimized using a computer algorithm. The objective function used in parameter
optimization is the sum of squared residuals technique. Statistical tests (the
paired t-test and the sign test) are applied to the developed rating equations for
absence of bias. Procedures of rating equation determination are presented
below:

[Link] Parameter Optimization

The optimization algorithm used to estimate parameters of the rating curve
equation is the Sum of Squared Residuals (SSQ) as shown in equation 1.19. The
constants of the rating curve were worked out according to the method of the
Least Squares. That is, the sum of the squares of the deviations between the
discharges measured and estimated by a rating curve equation should be
minimum.

41
 N
SSQ = ∑ ( X i ,measured − X i , simulated ) 2 [1.19]
I =1

Where N is the number of pairs of measured ( X measured ) and estimated

variables ( X estimated ) .

[Link] The Paired t-test

The paired t-test of the differences between the discharges measured and the
discharges estimated by the rating curve is used to check whether a rating
curve, on an average, gives significant overestimates or underestimates on
−
which the curve is based. The mean difference in percent, P , is tested against
its standard error to see if it is significantly different from zero. The assumptions
underlying this test are that the percentage differences between the measured
values and the values estimated by the rating curve equation are independent
of the magnitude of the discharge and normally distributed around a mean
value of zero.
The percentage deviation of the measured discharge and estimated by the
rating curve, and the standard deviation of the percentage deviation is given
by equations 1.20 and 1.21 respectively.

Qm − Qr
P= 100% [1.20]
Qr
⎛ −
⎞
∑ ⎜⎝ P − P ⎟⎠
SD = [1.21]
n −1

−
Where P is the percentage deviation, P is the mean percentage deviation,
Qm is the measured discharge, Qr is discharge estimated by the rating curve, SD
is percentage standard deviation, n is number of discharge measurements.
The standard error of mean deviation and the t-test statistics are given by
equations 1.22 and 1.23 respectively.

SD
SE = [1.22]
n
−
P
t= [1.23]
SE

The t-test statistics obtained by equation 2.9 was compared to the critical paired
t at 5% level of significance.

42
[Link] The Sign test

The sign test is used to check if the curve has been drawn in a sufficiently
balanced manner so that the two sets of discharge values, those measured and
those estimated from the rating curve equation, may be reasonably supposed
to represent the same population. The sign test is given by equation 1.24.

n1 − np − 0.5 *
t= [1.24]
npq
Where t is the sign test criterion, n1 is number of positive signs, n total number of
observations, p and q are probability of sign being positive and negative
respectively and 0.5 * shows continuity correction.

The t-test statistics obtained by equation 2.9 was compared to the critical paired
t at 5% level of significance.

Example 5

The data in Table 1.10 was taken from a certain river gauging station. Develop a
stage discharge (rating curve) equation.

Table 1.10 Water level and current meter data

Water DISCHARGE Water DISCHARGE Water DISCHARGE

Date Level Q observed Date Level Q observed Date Level Q observed

01-Jan-95 0.43 10.858 16-Mar-95 1.43 20.346 29-May-95 2.595 32.599

02-Jan-95 0.44 10.863 17-Mar-95 1.55 20.847 30-May-95 2.285 28.834

03-Jan-95 0.43 10.867 18-Mar-95 1.13 18.02 31-May-95 1.745 24.04

04-Jan-95 0.455 11.115 19-Mar-95 0.89 14.948 01-Jun-95 1.485 21.113

05-Jan-95 0.43 10.315 20-Mar-95 0.7 14.294 02-Jun-95 1.37 21.034

06-Jan-95 0.44 10.662 21-Mar-95 0.64 13.618 03-Jun-95 1.295 19.652

07-Jan-95 0.41 10.214 22-Mar-95 0.58 11.571 04-Jun-95 1.245 19.653

08-Jan-95 0.545 12.182 23-Mar-95 0.965 16.417 05-Jun-95 1.19 19.228

43
09-Jan-95 0.495 10.963 24-Mar-95 0.995 16.245 06-Jun-95 1.145 18.167

10-Jan-95 0.48 11.434 25-Mar-95 1.08 18.339 17-Jul-95 0.63 13.193

11-Jan-95 0.48 11.334 26-Mar-95 1.2 17.742 24-Jul-95 0.58 12.587

12-Jan-95 1.63 20.516 27-Mar-95 1.455 19.53 29-Jul-95 0.57 12.276

13-Jan-95 0.545 12.272 28-Mar-95 0.85 14.805 03-Aug-95 0.55 11.66

14-Jan-95 0.455 10.888 29-Mar-95 0.815 14.678 07-Aug-95 0.6 12.787

15-Jan-95 0.42 9.991 30-Mar-95 1.19 17.998 14-Aug-95 0.52 11.043

16-Jan-95 0.47 19.89 01-Apr-95 1.085 17.468 22-Aug-95 0.505 10.921

17-Jan-95 0.51 11.612 02-Apr-95 0.86 15.45 24-Aug-95 0.485 11.91

18-Jan-95 0.5 11.762 03-Apr-95 1.255 18.06 01-Sep-95 0.46 11.63

19-Jan-95 0.495 11.619 04-Apr-95 1.13 17.008 06-Sep-95 0.45 11.33

20-Jan-95 0.555 12.256 05-Apr-95 0.9 15.97 07-Sep-95 0.475 11.716

21-Jan-95 0.57 12.534 06-Apr-95 0.985 14.95 11-Sep-95 0.46 11.204

22-Jan-95 0.5 12.083 07-Apr-95 1.28 18.503 14-Sep-95 0.45 11.462

23-Jan-95 0.67 13.675 08-Apr-95 1.24 17.69 21-Sep-95 0.43 10.623

24-Jan-95 0.56 12.266 09-Apr-95 1.33 18.625 25-Sep-95 0.42 10.443

25-Jan-95 0.45 11.056 10-Apr-95 1.34 20.078 28-Sep-95 0.41 10.347

26-Jan-95 0.5 11.805 11-Apr-95 1.525 19.941 05-Oct-95 0.4 10.287

28-Jan-95 0.49 11.46 12-Apr-95 1.285 19.464 06-Oct-95 0.41 10.131

29-Jan-95 0.47 12.098 13-Apr-95 1.475 20.59 28-Oct-95 0.36 9.17

30-Jan-95 0.48 13.303 15-Apr-95 1.1 16.688 30-Oct-95 0.35 9.043

31-Jan-95 0.465 12.526 17-Apr-95 0.915 16.745 07-Dec-95 0.3 8.383

01-Feb-95 0.565 14.021 18-Apr-95 1.76 23.193 15-Dec-95 0.32 8.772

02-Feb-95 0.585 13.964 19-Apr-95 2.18 27.681 21-Dec-95 0.32 8.81

44
03-Feb-95 0.48 12.822 20-Apr-95 2.835 36.411 23-Dec-95 0.32 8.552

04-Feb-95 0.42 11.226 21-Apr-95 2.685 31.89 25-Dec-95 0.31 8.514

05-Feb-95 0.4 11.01 22-Apr-95 2.71 33.406 01-Jan-96 0.405 10.866

06-Feb-95 0.42 11.146 23-Apr-95 2.82 36.183 02-Jan-96 0.465 11.669

07-Feb-95 0.455 12.138 24-Apr-95 2.73 34.982 03-Jan-96 0.37 10.155

08-Feb-95 0.45 11.978 25-Apr-95 2.845 39.688 04-Jan-96 0.34 9.485

09-Feb-95 0.5 13.128 26-Apr-95 2.7 35.251 05-Jan-96 0.415 9.518

10-Feb-95 0.42 11.301 27-Apr-95 2.665 32.347 06-Jan-96 0.51 11.454

11-Feb-95 0.6 14.563 28-Apr-95 2.52 29.538 07-Jan-96 0.46 11.6

12-Feb-95 0.48 12.226 29-Apr-95 2.425 29.79 08-Jan-96 0.515 12.062

13-Feb-95 0.48 12.411 30-Apr-95 2.245 27.41 09-Jan-96 0.74 14.131

14-Feb-95 0.45 12.123 01-May-95 2.05 25.421 10-Jan-96 0.66 13.696

15-Feb-95 0.57 14.022 02-May-95 1.895 23.059 11-Jan-96 0.44 11.458

16-Feb-95 0.44 10.962 03-May-95 1.785 22.919 12-Jan-96 0.39 10.778

17-Feb-95 0.435 10.586 04-May-95 1.73 22.713 13-Jan-96 0.415 11.161

18-Feb-95 0.48 11.815 05-May-95 1.655 22.86 14-Jan-96 0.38 10.287

20-Feb-95 0.49 12.436 06-May-95 1.59 21.729 15-Jan-96 0.37 9.987

21-Feb-95 0.59 13.22 08-May-95 1.51 19.618 19-Jan-96 0.765 15.809

22-Feb-95 0.46 11.707 09-May-95 1.51 21.542 20-Jan-96 0.52 12.394

23-Feb-95 0.455 11.508 10-May-95 1.44 20.514 21-Jan-96 0.425 11.21

24-Feb-95 0.43 10.983 11-May-95 1.365 19.826 27-Jan-96 0.385 10.142

25-Feb-95 0.455 11.991 12-May-95 1.32 19.388 28-Jan-96 0.39 10.184

26-Feb-95 0.615 13.278 13-May-95 1.305 18.648 03-Feb-96 0.435 11.396

27-Feb-95 0.52 12.258 14-May-95 1.26 19.265 04-Feb-96 0.4 11.222

45
28-Feb-95 0.53 12.873 15-May-95 1.23 18.303 10-Feb-96 0.4 10.822

01-Mar-95 0.58 13.094 16-May-95 1.2 17.311 11-Feb-96 0.375 9.933

02-Mar-95 0.46 11.643 17-May-95 1.265 18.015 17-Feb-96 0.53 12.796

03-Mar-95 0.76 15.144 18-May-95 1.875 25.18 18-Feb-96 0.585 12.691

04-Mar-95 0.785 15.573 18-May-95 2.09 26.007 20-Feb-96 0.54 11.975

05-Mar-95 0.715 14.151 19-May-95 1.505 20.637 21-Feb-96 0.44 10.634

06-Mar-95 0.58 13.542 20-May-95 1.235 19.152 22-Feb-96 0.46 10.674

07-Mar-95 0.585 13.108 21-May-95 1.16 18.615 26-Feb-96 0.535 12.346

08-Mar-95 0.665 14.702 22-May-95 1.245 19.215 02-Mar-96 0.425 11.248

09-Mar-95 0.955 15.655 23-May-95 1.14 18.42 03-Mar-96 0.455 11.935

10-Mar-95 2.13 25.55 24-May-95 1.045 18.068 09-Mar-96 0.41 11.148

10-Mar-95 1.72 21.959 25-May-95 1.045 16.772 10-Mar-96 0.37 10.269

11-Mar-95 1.51 18.415 26-May-95 1.7 22.414 16-Mar-96 0.665 13.016

12-Mar-95 1.245 18.272 27-May-95 1.45 20.408 17-Mar-96 0.495 12.595

13-Mar-95 0.86 14.582 28-May-95 1.965 25.967 20-Mar-96 0.89 14.337

14-Mar-95 0.66 13.458 28-May-95 2.145 29.153 21-Mar-96 0.82 13.328

15-Mar-95 0.76 14.706 29-May-95 2.46 31.585 23-Mar-96 1.16 16.355

Solution:

The steps followed in constructing rating curve for the above data are described
below:

1. The required number of discharge measurements in order to obtain a

reliable rating curve is checked using the formula shown in 1.25.
2
⎛ 2 SD ⎞
n>⎜ ⎟ [1.25]
⎝ E ⎠
Where n is the number of required measurements, SD is standard
deviation in percent (2SD is allowable width of scatter band), E is a
specified precision expressed as a percentage, usually 5% (Tilrem, 1979).

46
 2. All discharge measurements are plotted on ordinary arithmetical graph
paper, gauge height on vertical scale and discharge on horizontal scale.
A mean curve was fitted to the discharge-gauge height data and
measurements which deviate more than 5 percent of the mean curve
were considered as outliers.

3. The stage discharge relation was plotted on log-log graph paper in order
to establish whether the rating curve to be developed is composed of one
or several straight line segments.

4. Rating curve is estimated using an optimization program that optimized

parameters of the rating curve equation (K, Ho and n).

5. Statistical tests, i.e. the paired t-test and sign- test were applied to the
rating curves for absence of bias.

Data number check-up:

2
⎛ 2 SD ⎞
n>⎜ ⎟
⎝ E ⎠

n=219, SD=6.22, E=0.05, hence 219> 6.2.

Rating Curve
5
1995-1996
4
Gauge height (m)

0
0 10 20 30 40 50
Discharge (cumecs)

Figure 1.24 Arithmetic plot of stage and discharge data

47
 RAW DATA RATING

10
GAUGE HEIGHT

0.1

0.01
1 10 100
DISCHARGE

Figure 1.25. Log-Log Plot of stage and discharge data

Manual computation of the rating curve equation using excel sheet is presented
below:

Figure 1.26 Manual computation of rating curve equation.

The test statistics result is presented in the Table 1.11

48
Table 1.11 Test statistics results for rating curve determination

STDEV 8.974
Paired Stnd error 0.548
t test Test Statistic t -0.541
5% level of sig(table) 1.96
Conc rating curve free of bias
Total No. of Pos signs 119
probab. Of sign being Pos 0.5
sign [Link] sign being Neg 0.5
test Exp No. of Pos signs np 59.5
Stand error of np 8.185
Test criterion t 1.894
5% level of sig(table) 1.96
Conc rating curve free of bias
Number of changes in sign
n1 40
sqrt((n-1)pq) 8.17
t=abs((n1-(n-1)p-0.5/sqrt(n-
1)pq 11.51
The run
of t test 11.51
Sign
test 5% level of sign 1.96
There is no random
Conc fluctuation There is no systematic trend
in the deviation with time
indicating that the rating curve doesn't need
Adjustment for shift control

Hence satisfying the test statistics the rating curve equation is presented as
equation 1.26:

Q = 7.25( H + 1)1.14 [1.26]

1.3.4 Flow data Synthesis

It is not quite common that the starting and ending period of flow data at the
gauging stations that are located in a watershed are the same. For long term
flow analysis to indicate water resources availability and hydrological modelling
both forward extension and backward extension of flow data are important.
Similarly missing flow records in time series needs to be filled-in before any further
use of the data. Hence it is important to develop techniques that serve as a
guideline for flow reconstruction. Generally, the following four methods are used
to reconstruct flow data.

49
Method 1: Linear Interpolation.
If the missing data event was of shorter duration i.e. less than 5 days, then the
missing data points were estimated by linear interpolation between the data
point observed just before and after the missing spell. This method was tested
and found to be applicable for filling shorter missing records of flow.

Method 2: Cross Correlation.

The correlation model shown by equation 1.27 is used to extend data form a
neighbouring station.
Ydep = α * Yindep
[1.27]
where Ydep is the daily flow of the dependent catchment; Yindep is the daily flow of
the independent catchment; and α is the cross correlation coefficient.

Method 3: Seasonal Model (SM).

When the missing flow data is between five days and three months, a seasonal
model as shown in equation 1.28 was used to fill-in the missing records. The
Seasonal Model forecast yd, for each date (d) of the year (1-365) is defined as
the mean of the flows on date d of the data period under study.

Λ 1 L
Yd = ∑ Yd , r
L r =1 [1.28]

Where Yd , r is the observed discharge on date d in year r and L is the number of

years.

Method 4: Rainfall Runoff Modelling.

Linear Perturbation Model (LPM) in its Linear Transfer Function (LTF) form is used to
fill-in missing records of greater than three months. A LPM- LTF model was run
using an areal rainfall at the subcatchments level. New data set should be
prepared with the available recorded data and the extracted simulated data
that represent the missing records. This method was recommended to fill-in
missing flow records from measured rainfall data when the missing flow was
greater than three months.

1.3.5 Flow Duration Curves

Flow Duration Curves (FDCs), which are cumulative frequency curves compiled
by ranking all daily stream flows in order of magnitude is plotted by computing
the percentage of time each flow is equalled or exceeded. The method is a
useful way of treating the time variability of water discharge data in hydropower
studies or other water resources practices. A flow duration curve is plotted as

50
flow in ordinate and the percent of time a particular flow can be expected to
be exceeded in abscissa. The basic advantage of flow duration curve is it allows
the characterizing of the flow over long periods of time to be presented in one
compact curve.
Methods of Deriving FDCs are Rank-Ordered Technique and Class Interval
Technique. Procedures of FDC construction using the two methods is outlined
below.

[Link] Procedures of FDC construction using Rank-Ordered Method:

(i) Rank data in acceding/descending order;
(ii) Assign individual order number for each ranked values, the largest
beginning with 1;
(iii) The order numbers are then divided by the total number in the record
and multiplied by 100 to obtain the percent of time that the mean flow
has been equaled or exceeded during the period of record being
considered;
(iv) The flow value is then plotted versus the respective computed
exceedance percentage;
(v) References to the flow duration values at specific exceedance value
are usually made as Q50, Q30, QI0, and So on, indicating the flow
value at the percentage point subscripted. Naturally, the longer the
record, the more statistically valuable the information that results.

[Link] Procedures of FDC construction using Class Interval Technique:

(i) The time series of flow values are categorized into class intervals
(ii) The classes range from the highest flow value in the series to the
lowest value in the time series
(iii) A tally is made of the number of flows in each, and by summation
the number of values greater than a given upper limit of the class
can be determined
(iv) The number of flows greater than the upper limit of a class interval
can be divided by the total number of flow values in the data series
to obtain the exceedance percentage
(v) The value of the flow for the particular upper limit of the class
interval is then plotted versus the computed exceedence percent

Example 6: Construct a FDC for the river flow data shown in table 1.12.

51
 Table 1.12 Flow data for FDC construction

Date Flow Date Flow Date Flow Date Flow Date Flow Date Flow
Data Data Data Data Data Data

06/01/1974 15.74 04/04/1974 15.23 01/07/1974 14.15 14/09/1974 11.63 30/10/1974 11.27 22/03/1975 12.35

07/01/1974 14.66 05/04/1974 14.60 02/07/1974 14.15 15/09/1974 11.63 31/10/1974 11.36 23/03/1975 13.25

08/01/1974 14.39 06/04/1974 17.66 03/07/1974 13.97 18/09/1974 11.99 01/11/1974 11.27 24/03/1975 11.99

09/01/1974 14.92 07/04/1974 31.34 04/07/1974 13.97 19/09/1974 11.99 02/11/1974 11.09 25/03/1975 12.89

10/01/1974 16.87 08/04/1974 31.16 05/07/1974 13.97 20/09/1974 11.81 03/11/1974 11.09 26/03/1975 12.08

11/01/1974 16.87 09/04/1974 29.63 06/07/1974 13.61 21/09/1974 11.81 04/11/1974 11.27 27/03/1975 13.16

12/01/1974 15.46 10/04/1974 29.54 07/07/1974 13.61 22/09/1974 11.81 05/11/1974 11.54 28/03/1975 13.61

13/01/1974 14.92 11/04/1974 30.80 08/07/1974 13.70 23/09/1974 11.81 06/11/1974 11.09 29/03/1975 12.53

14/01/1974 14.92 12/04/1974 31.34 09/07/1974 13.97 24/09/1974 11.81 07/11/1974 11.09 30/03/1975 12.53

15/01/1974 20.51 13/04/1974 27.38 10/07/1974 13.97 25/09/1974 11.81 08/11/1974 11.09 31/03/1975 12.35

16/01/1974 17.45 14/04/1974 27.02 11/07/1974 13.79 26/09/1974 11.81 09/11/1974 11.09 01/04/1975 13.88

17/01/1974 18.64 15/04/1974 22.25 12/07/1974 13.61 27/09/1974 11.63 10/11/1974 10.91 02/04/1975 14.15

18/01/1974 17.74 01/05/1974 31.88 13/07/1974 13.97 28/09/1974 11.63 11/11/1974 10.91 03/04/1975 13.97

19/01/1974 17.15 02/05/1974 33.77 14/07/1974 13.88 29/09/1974 11.63 12/11/1974 10.91 04/04/1975 16.31

20/01/1974 16.58 03/05/1974 33.50 15/07/1974 13.88 30/09/1974 11.63 13/11/1974 10.91 05/04/1975 15.05

21/01/1974 18.94 04/05/1974 26.48 16/08/1974 12.71 01/10/1974 11.45 14/11/1974 10.91 06/04/1975 17.21

22/01/1974 18.04 05/05/1974 25.31 17/08/1974 12.71 02/10/1974 11.45 15/11/1974 10.73 07/04/1975 18.11

23/01/1974 17.74 06/05/1974 24.50 18/08/1974 12.89 03/10/1974 11.45 16/11/1974 10.73 08/04/1975 16.31

24/01/1974 17.15 07/05/1974 22.61 19/08/1974 12.71 04/10/1974 11.45 17/11/1974 10.73 09/04/1975 27.83

25/01/1974 16.58 08/05/1974 21.80 20/08/1974 12.71 05/10/1974 11.45 18/11/1974 10.73 10/04/1975 22.07

26/01/1974 16.02 09/05/1974 21.98 21/08/1974 12.53 06/10/1974 11.45 19/11/1974 10.73 11/04/1975 21.53

27/01/1974 16.02 10/05/1974 29.72 22/08/1974 12.71 07/10/1974 11.45 20/11/1974 10.73 12/04/1975 18.65

52
28/01/1974 18.04 11/05/1974 30.08 23/08/1974 11.99 08/10/1974 11.45 21/11/1974 10.73 13/04/1975 17.03

29/01/1974 17.15 12/05/1974 26.48 24/08/1974 12.89 09/10/1974 12.17 22/11/1974 10.73 14/04/1975 16.67

30/01/1974 17.15 13/05/1974 26.48 25/08/1974 12.89 10/10/1974 11.63 23/11/1974 10.73 15/04/1975 16.67

31/01/1974 15.74 14/05/1974 27.38 26/08/1974 12.89 11/10/1974 11.45 24/11/1974 10.73 16/04/1975 16.31

01/02/1974 21.48 15/05/1974 24.14 27/08/1974 12.71 12/10/1974 11.45 25/11/1974 10.73 17/04/1975 15.23

02/02/1974 19.25 13/11/1974 10.91 28/08/1974 12.89 13/10/1974 11.45 26/11/1974 10.73 18/04/1975 14.33

03/02/1974 20.19 14/11/1974 10.91 29/08/1974 12.89 14/10/1974 11.45 27/11/1974 10.55 19/04/1975 14.15

04/02/1974 25.62 15/11/1974 10.73 30/08/1974 12.89 15/10/1974 11.27 28/11/1974 10.55 20/04/1975 14.33

05/02/1974 33.86 16/11/1974 10.73 31/08/1974 12.89 16/10/1974 11.45 29/11/1974 10.55 21/04/1975 15.95

06/02/1974 32.19 17/11/1974 10.73 01/09/1974 11.81 17/10/1974 11.45 30/11/1974 10.55 22/04/1975 23.69

07/02/1974 30.17 18/11/1974 10.73 02/09/1974 11.81 18/10/1974 11.27 01/12/1974 10.55 23/04/1975 15.59

08/02/1974 28.23 19/11/1974 10.73 03/09/1974 11.81 19/10/1974 11.27 02/12/1974 10.55 24/04/1975 30.89

09/02/1974 27.09 20/11/1974 10.73 04/09/1974 11.81 20/10/1974 11.27 03/12/1974 10.55 25/04/1975 28.91

10/02/1974 25.62 21/11/1974 10.73 05/09/1974 11.81 21/10/1974 11.27 04/12/1974 10.55 26/04/1975 28.91

11/02/1974 22.48 22/11/1974 10.73 06/09/1974 11.81 22/10/1974 11.27 05/12/1974 10.55 27/04/1975 31.79

12/02/1974 20.51 23/11/1974 10.73 07/09/1974 11.81 23/10/1974 11.27 06/12/1974 10.64 28/04/1975 34.85

13/02/1974 18.04 24/11/1974 10.73 08/09/1974 11.81 24/10/1974 11.27 07/12/1974 10.73 29/04/1975 29.81

14/02/1974 15.74 25/11/1974 10.73 09/09/1974 11.81 25/10/1974 11.09 08/12/1974 10.73 30/04/1975 26.75

15/02/1974 14.92 26/11/1974 10.73 10/09/1974 11.81 26/10/1974 11.09 09/12/1974 11.63 01/05/1975 26.75

01/04/1974 12.53 27/11/1974 10.55 11/09/1974 11.81 27/10/1974 11.09 10/12/1974 11.00 02/05/1975 29.81

02/04/1974 11.99 28/11/1974 10.55 12/09/1974 11.81 28/10/1974 11.09 11/12/1974 11.54 03/05/1975 27.65

03/04/1974 16.13 29/11/1974 10.55 13/09/1974 11.63 29/10/1974 11.27 12/12/1974 11.36 04/05/1975 19.55

53
Solution:

Following the procedures presented in section 1.3.5, one can establish an excel
sheet (Figure 1.27) and derive the FDC curve.

Figure 1.27 Flow Duration Computations.

The Flow data in column B and Percent Exceeded in column E are plotted as
shown in Figure 1.28 to indicate the FDC.

54
 Flow Duration Curve

60

50

40
Discharge(cumecs)

30

20

10

0
0 10 20 30 40 50 60 70 80 90 100
Percent of time that indicated discharge was equaled or exceeded

Figure 1.28 FDC indicating the 50% and 95% flow exceedence probability

1.3.6 Missing data computations

A seasonal model shown in equation 1.28 can be used to compute seasonal
flow (365 values each). Before the computation of seasonal flow variations,
basic hydrological analysis that includes percentage missing computation and
filling the missing record is important. The methodology presented in Precipitation
Data was used here to evaluate flow data in the Tana basin system. The missing
records are filled-in using the technique presented in section 1.3.4.

Figure 1.29 Computation of missing flow records in the Tana basin system

55
1.4 Sediment Concentration

This section introduces the mechanism of sediment transport and the nature of
deposition. The principle helps to understand the occurrence of sediments in a
watershed and their depositions in either flood plains and/or streams.

As the silt originates from the watershed, the characteristics of the catchment
such as its areal extent, soil types, land slopes, vegetal cover and climatic
conditions like temperature, nature and intensity of rainfall, have a great
significance in the sediment production in the form of sheet erosion, gully erosion
and stream, channel erosion. In regions of moderate rainfall, sheet erosion is the
dominant source of total sediment load while in arid and semi-arid regions,
gullying and stream-channel erosion furnish the greater part of the load.

Experiments have shown that the erosive power of water, flowing with a velocity
of V , varies as V 2 while the transporting ability of water varies as V 6 . Sediment
moves in the stream as suspended load (fine particles) in the flowing water, and
as bed load (large particles), which slides or rolls along the channel bottom.
Sometimes, the particles (small particles of sand and gravel) move by bouncing
along the bed, which is termed as ‘saltation’, which is a transitional stage
between bed and suspended load. The material, which moves as bed load at
one section may be in suspension at another section.
The suspended sediment load of streams is measured by sampling the water,
filtering to remove the sediment, drying and weighing the filtered material.

Sediment load, (ppm) =Weight of sediment in the sample x 106

Weight of sediment laden water sample [1.29]

1.4.1 Method of Sampling Sediment

There are two ways of sampling sediment; The ‘Depth Integrating Type’ and the
‘Point Sampling’. Point samplers are used only where it is not possible to use the
depth integrating type because of great depth of high velocity, or for studies of
sediment distribution in streams. The sample is usually collected in ‘pint bottle’
held in a sample of stream-lined body so as not to disturb the flow while
collecting a representative sample. The general guideline to establish stations for
sediment sampling is presented in Table 1.13.

56
Table 1.13 Recommended minimum densities for sediment stations

(Source: WMO 1994)

Physiographic unit Minimum density per station
(area in km2 per station)

Coastal 18300

Mountaneous 6700

Interior Plains 12500

Hilly/Undulating 12500

Small Islands 2000

Polar/arid 200000

1.4.2 Relationship of Suspended Sediment Transport and Streamflow

The relation between the suspended-sediment transport Qs and stream flow Q is
given by:

Qs = KQ n [1.30]
LogQs = LogK + nLogQ [1.31]

Sediment rating curve is developed from a continuous record of stream flow and
it provides a rough estimate of sediment inflow to reservoirs. The total sediment
transport may be estimated by adding 10-20% to the suspended sediment
transport to allow for the bed load contribution.

When the sediment-laden water reaches a reservoir, the velocity and turbulence
are greatly reduced. The dense fluid-solid mixture along the bottom of the
reservoir moves slowly in the form of a density current or stratified flows, i.e., a
diffused colloidal suspension having a density slightly different from that of the
main body of reservoir water, due to dissolved minerals and temperature, and
hence does not mix readily with the reservoir water (Figure 1.30). Smaller particles
may be deposited near the base of the dam. Some of the density currents and
settled sediments near the base of the dam can possibly be flushed out by
operating the sluice gates. The modern multipurpose reservoirs are operated at
various water levels, which are significant in the deposition and movement of silt
in the reservoir.

57
Figure 1.30 Sediment accumulations in a reservoir
(Source: Hydrology Principles, Analysis and Design, Raghunath, 2006)

1.4.3 Computation of seasonal distribution (Date Average and Annual)

The total amount of sediment that passes any section of a stream at any time is
referred to as the sediment yield or sediment production. Seasonal distribution of
sediment data helps to analyse the magnitude of suspended sediment with
different time steps for management purposes. An example data set for Koga
watershed is presented here to indicate the seasonal (date averaged values),
monthly values and annual values of sediment deposition. Results are presented
in Figures 1.31 and 1.32. The computed mean daily value and annual values are
162.42unit and 59284.6unit respectively. In Figure 1.31 the first day starts with
January 1 and ends on December 31st.

1800
Date Averaged Mean
1600

1400
Suspended Sediment Load

1200

1000

800

600

400

200

0
01/01/01

15/01/01

29/01/01

12/02/01

26/02/01

12/03/01

26/03/01

09/04/01

23/04/01

07/05/01

21/05/01

04/06/01

18/06/01

02/07/01

16/07/01

30/07/01

13/08/01

27/08/01

10/09/01

24/09/01

08/10/01

22/10/01

05/11/01

19/11/01

03/12/01

17/12/01

31/12/01

Date

Figure 1.31 Date averaged values of suspended sediment for Koga watershed

58
As shown in Figure 1.31, significant suspended sediment loading into the Koga
reservoir starts in June and continuous to November, peaking in August.

Annual Values of Suspended Sediment

300000

250000
Suspended Sediment Load

200000

150000

100000

50000

0
1960

1962

1964

1966

1968

1970

1972

1974

1976

1978

1980

1982

1984

1986

1988

1990

1992

1994

1996

1998

2000

2002
Year

Figure 1.32 Annual Distribution of Suspended Sediment for Koga watershed (1960 -
2002)

As shown in Figure 1.32, huge volume of suspended sediment was deposited in

1975 (253718). This magnitude is an exaggerated value which might be an error
in data recording. Apart from the data of 1975 there is no real trend that is
observed in the data series. Long term trend analysis can be evaluated based
on the Mann-Kendal’s parametric statistical test.

1.4.4 Reduction in Reservoir Capacity

The useful life of a reservoir gets reduced due to sediment deposition causing a
decrease in its storage capacity. The factors affecting the pattern of sediment
deposition in reservoirs are:

(i) Sediment load (i.e., sediment inflow rate)

(ii) Sediment size (i.e., gradation of silt)
(iii) Compaction of sediment
(iv) River inflow pattern
(v) River valley slope
(vi) Shape of reservoir
(vii) Capacity of reservoir (its size and storage period)
(viii) Vegetal growth at the head
(ix) Outlets in the dam (their types, location and size)

59
(x) Reservoir operation
(xi) Upstream reservoirs, if any.

The percent of the inflowing sediment, which is retained in a reservoir is called

the trap efficiency and it is a function of the ratio of reservoir capacity to total
annual sediment inflow, since a small reservoir on a large stream passes most of
its inflow quickly (giving no time for the silt to settle) while a large reservoir allows
more detention time for the suspended silt to settle.
The relation between trap efficiency of reservoir vs. capacity-inflow ratio is shown
in Figure 1.33 (Brune, 1953). The rate at which the capacity of a reservoir is
reduced by sediment deposition depends on
(i) The rate of sediment inflow, i.e., sediment load.
(ii) The percentage of the sediment inflow trapped in the reservoir, i.e., trap
efficiency.
(iii) The density of the deposited sediment.

Figure 1.33 Reservoir trap efficiency vs. capacity-inflow ratio (after Brune, 1953)

1.4.5 Prediction of a Reservoir Life

In estimating the useful life of a reservoir, the correct prediction of the density of
the deposited sediment is an important factor. Lane and Koelzer (1943) gave the
equation for the dry specific weight γ t after time t years as:
γ t = γ i + K log10 t [1.32]
Where γ i is initial specific weight and K is a constant for the rate of compaction
The useful capacity of reservoir lost each year by sediment deposition is:
Vs = Q sη trap [1.33]
Where Vs is volume of useful capacity of reservoir lost each year

60
Qs is annual sediment inflow into the reservoir
η trap trap efficiency of the reservoir
While allocating space for the dead storage in the reservoirs (i.e., to provide
space for sediment deposition during the life of the project) the trap efficiency is
taken as at least 95% and rarely below 90%. Sediment deposits in the upper end
of the reservoirs generally become covered by vegetation resulting in heavy
evapotranspiration loss of the available water, which is more critical in arid
regions.

Example 7
A proposed reservoir in the Tana basin system has a capacity of 400 ha-m. The
catchment area is 130 km2 and the annual stream flow averages 12.31 cm of
runoff. If the annual sediment production is 0.03 ha-m/km2, what is the probable
life of the reservoir before its capacity is reduced to 20% of its initial capacity by
sediment deposition. The relation between trap efficiency and capacity inflow
ratio is given below.

Table 1.14 Relationship of trap efficiency and capacity inflow ratio

Capacity Trap Capacity Trap Capacity Trap

Inflow Efficiency Inflow Ratio, Efficiency Inflow Efficiency
Ratio, C/I C/I Ratio, C/I
η trap η trap η trap

1 97.5 0.06 80 0.006 31

0.7 97 0.05 77 0.005 27

0.6 96.5 0.04 74 0.004 20

0.5 96 0.03 68 0.003 13

0.4 95.5 0.02 60 0.002 2

0.3 95 0.015 52

0.2 93 0.01 43

0.1 87 0.008 38

0.07 82 0.007 36

61
Solution:
The useful life may be computed by determining the number of years required
for each incremental loss of reservoir capacity (i.e., for the decreasing values of
capacity-inflow ratios) up to the critical storage volume of 400 × 0.20 = 80 ha-m
as tabulated below:

Capacity Capacity* Trap efficiency Annual** Loss of No. of

C Inflow ηtrap (%) Sediment Reservoir Years for
Trapped Capacity The
capacity
∆C
loss

( ha − m ) C for the Ave. for Vs = Qs xη trap ( ha − m ) ∆C ÷ Vs

I C *Increment
ratio
I

400 0.25 94

320 0.2 93 93.5 3.64 80 22

240 0.15 90 91.5 3.57 80 22.4

160 0.1 87 88.5 3.45 80 23.2

80 0.05 77 82.0 3.2 80 25

Total= 92.6 say,

93 yrs

12.31 130 x10 6

*Average annual inflow, I = x = 1600ha − m
100 10 4
C 400
For reservoir capacity C=400ha-m, = = 0.25
I 1600
** Annual sediment inflow into the reservoir

Qs = 0.03x130 = 3.9ha − m

62
1.4.6 Reservoir Sedimentation Control

Sediment deposition in reservoirs can not be actually prevented but it can be

retarded by adopting some of the following measures:

(i) Reservoir sites, which are prolific sources of sediment, should be avoided.
(ii) By adopting soil-conservation measures in the catchment area, as the silt
originates in the watershed.
(iii) Agronomic soil conservation practices like cover cropping, strip cropping,
contour farming, suitable crop rotations, application of green manure
(mulching), proper control over graze lands, terracing and benching on
steep hill slopes, etc. retard overland flow, increase infiltration and reduce
erosion.
(iv) Contour trenching and afforestation on hill slopes, contour bunding gully
plugging by check dams, and stream bank stabilisation by the use of
spurs, rivetments, vegetation, etc. are some of the engineering measures
of soil conservation.
(v) Vegetal cover on the land reduces the impact force of rain drops and
minimises erosion.
(vi) Sluice gates provided in the dam at various levels and reservoir operation,
permit the discharge of fine sediments without giving them time to settle to
the bottom.
(vii) Sediment deposits in tanks and small reservoirs may be removed by
excavation, dredging, draining and flushing either by mechanical or
hydraulic methods and sometimes may have some sales value.

63
1.5 Climatic Variables
This section deals with meteorological data that include temperature, wind
speed, sunshine hours, relative humidity, solar radiation and pan evaporation.
Climatic data measurement mechanisms will be discussed. In most parts of
catchments, climate in the highlands is strongly influenced by the effects of
elevation. This complicates the mechanism of data measurement and reliability
of the available data. Only few areas are properly gauged in the temperate
regions at higher elevations and tropical regions of the lower elevations.
Approximately there are 45 weather stations in the Blue Nile basin with varying
data record length (1954 to 2006).

1.5.1 Climatic Data Measurement

Climatic data measurement involves the measurement of rainfall, sunshine hours,
relative humidity, wind speed, solar radiation, pan evaporation and maximum
and minimum temperatures. Data measurements for rainfall are discussed in the
previous section. Figure 1.34 shows a MET station for the measurement of climatic
variables directly. Figures 1.35, 1.36, and 1.37 shows equipments for measuring
evaporation, wind speed and sunshine hours respectively. Apart from the direct
measurement of evaporation using class A pan and lysimeters, evaporation or
evapotranspiration from large water or land surfaces can be measured indirectly
using indirect methods that give acceptable results. Evaporation pans and
lysimeters are used in networks. For existing reservoirs and for plots or small
catchments, estimates of evaporation can be made by water-budget, energy-
budget, and aerodynamic approaches.

Figure 1.34 MET station for climatic variables measurement

64
In the long term analysis of water regime of lakes and reservoirs, evaporation
plays an important role in the water management. In such cases, the number
and distribution of evaporation stations are determined according to the area
and configuration of the lakes and the climatic region or regions involved. The
minimum densities of evaporation stations as recommended by the WMO
standard are outlined in Table 1.15.

Table 1.15 Minimum density of evaporation stations (WMO, 1994)

Physiographic unit Minimum density per station

(area in km2 per station)

Coastal 50000

Mountainous 50000

Interior plains 50000

Hilly/Undulating 50000

Small Islands 50000

Polar/arid 100000

Figure 1.35 Class A Evaporation pan

65
Figure 1.36 Wind Speed measuring device installed at 1st class meteorological
station in Haromaya University

Figure 1.37 Sunshine hour recorder

66
 CHAPTER TWO

RAINFALL RUNOFF MODELS

2.1. Introduction
This chapter discusses about the rainfall-runoff models that are suitable for
hydrological processing and various water resources practices. Time series and
spatial data availability in the Blue Nile basin are also presented and discussed.

2.2 Time series data availability

Generally there are 75 river flow measuring stations, 71 rainfall gauging stations,
44 stations that record temperature data and approximately 23 stations record
pan evaporation data in the Blue Nile Basin of Ethiopia (Table 2.1). However not
all data are readily available and used for hydrological modeling due to
inconsistency in records and huge chunk of missing data.

Table 2.1 Summary of Meteorological data (After Dereje et al., 2006)

[Link] Drainage Stream Meteorological Data

Basin gauges

Rain gauges Temperature Evaporation

1 Anger 3 2 1 1

2 Beles 2 1 3 1

3 Beshilo 6 3 3

4 Dabus 8 8 4 3

5 Dinder 2 - - -

6 Didessa 6 8 6 2

7 Fincha 1 5 3 2

8 Guder 4 3 3 -

9 Jemma 6 10 4 4

10 Muger 5 5 2 1

11 [Link] 10 5 2 -

67
12 Rahad - - -

13 S. Gojjam 15 7 2 2

14 Tana 12 9 10 4

15 Weleka 1 1 1 -

16 Wenbera 1 - -

Total Sum 75 71 44 23

The spatial distribution of climatic stations in the Blue Nile basin is presented in
Figure 2.1

Figure 2.1 Location of the Blue Nile Basin in Ethiopia together with climatic stations.
(After Semu et al., 2009)

In the Tana basin system there are six climatic stations. The spatial distribution
and geographic location of these stations are presented in Table 2.2 and Figure
2.2 respectively. All of these stations record rainfall data. However the synoptic
stations are limited in number. With the presence of synoptic station (for
example, Bahir Dar station), better meteorological data capturing mechanism is
available. The Bahir Dar meteorological station (located at 37o25’ E and 11o36’
N) records continuous rainfall, minimum and maximum temperatures, wind

68
speed, sunshine hours, relative humidity, and evaporation. These key climatic
characteristics at Bahir Dar station are presented in Table 2.3.

Table 2.2 Locations of meteorological stations in the Tana basin system

X Y Projection Elevation
CODE Longitude Latitude Projection
Addis Zemen 37.87 12.12 377036.278 1340077.671 2105
Bahir Dar 37.42 11.60 327732.695 1282800.742 1828
Gumara 37.63 11.83 350756.820 1308121.327 1797
Yifag 37.72 12.07 360685.463 1334619.562 1838
Woreta 37.68 11.93 356257.012 1319155.519 1799
Debre Tabor 38.03 11.92 394370.689 1317892.111 2314

Figure 2.2 Details of meteorological stations in the Tana subbasin system

The time series distribution of climatic variables in the Tana basin system is
presented in Figure 2.3.

69
Table 2.3 Summary of Key Climatic Characteristics for Bahir Dar Station. (Source :
KOGA IRRIGATION AND WATERSHED MANAGEMENT PROJECT (Hydrology
Factual Report)

Climatic Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov Dec

Variable

Mean monthly rainfall 2.6 1.8 7.9 24.2 83.9 187.2 424.3 393.1 198.0 93.4 19.6 3.2
(mm)(1)

80% probable rainfall 0 0 0 0.6 20.3 121.7 314.2 272.0 147.2 33.9 0.9 0
(mm)(1)

Daily maximum rainfall 16 16.6 32.5 52.8 116 94.8 133 200 70.4 88.5 38.8 15
(mm)

Number of rainy days ( 0 0 0 1 2 6 15 13 6 3 1 0

> 10 mm)

Mean monthly 26.5 27.9 29.4 29.8 28.7 26.5 23.9 23.8 25.1 26.2 26.3 26.1
maximum temperature
(oC)

Mean monthly minimum 7.3 8.9 11.8 13.0 14.2 13.8 13.7 13.4 12.7 12.3 10.2 7.8
temperature (oC)

Average monthly 16.9 18.4 20.6 21.4 21.5 20.2 18.8 18.6 18.9 19.3 18.3 17.0
o
temperature ( C)

Mean monthly wind 121 121 147 156 138 130 104 95 95 104 112 104
speed (km/day)(2)

Mean monthly relative 59 52 49 49 59 72 81 82 78 71 66 62

humidity (%)

Mean daily sunshine 9.6 9.6 9 9.2 8.2 7 5.1 4.8 6.7 8.7 9.5 9.6
duration (hrs)

Evaporation (mm) 170 189 254 255 208 130 73 67 79 121 137 151
(Piche Evaporimeter)

70
2.3 General Climate Characteristics
Based on traditional Ethiopian climate classification, the climate of Upper Blue
Nile basin can be described as follows:

(i) The Kolla zone below 1800 m has mean annual temperatures in the range
20-28°C;
(ii) The Woina Dega zone between 1800-2400m has mean annual
temperatures in the range 16-20°C; and
(iii) The Dega zone above 2400 m has mean annual temperatures in the
range 6-16°C

Bahir Dar Maximum and Minimum Temperature Addis Zemen Maximum and Mininum Temperature

40.0 MAX MIN 40.00

35.0 35.00

30.0 30.00

25.00

Temperature
25.0
Temperature

MAX
20.0 20.00
MIN

15.0 15.00

10.0 10.00

5.0 5.00

0.0 0.00
01/01/92

01/01/94

01/01/96

01/01/98

01/01/00

01/01/02
01/01/92

01/01/94

01/01/96

01/01/98

01/01/00

01/01/02

Date Date

Wind Speed (m/s) at Bahir Dar Station

Relative Humidity at Bahir Dar Station
2.50
120.00

100.00 2.00
Relative Humidity (%)

80.00
1.50
Wind Speed

60.00
1.00

40.00

0.50

20.00

0.00
0.00
01/01/92
01/07/92
01/01/93
01/07/93
01/01/94
01/07/94
01/01/95
01/07/95
01/01/96
01/07/96
01/01/97
01/07/97
01/01/98
01/07/98
01/01/99
01/07/99
01/01/00
01/07/00
01/01/01
01/07/01
01/01/02
01/07/02
01/01/03
01/07/03
01/01/92

01/07/92

01/01/93

01/07/93

01/01/94

01/07/94

01/01/95

01/07/95

01/01/96

01/07/96

01/01/97

01/07/97

01/01/98

01/07/98

01/01/99

01/07/99

01/01/00

01/07/00

01/01/01

01/07/01

01/01/02

01/07/02

01/01/03

01/07/03

Date
Date

Figure 2.3 Presentation of time series distribution of climatic variables in Tana basin
system

For the analysis or rainfall in the Tana basin system concurrent data period from
1992 to 2003 was analysed. The batch file prepared to assess the status of rainfall
data for the Tana basin system is presented in Figure 2.4.

71
Figure 2.4 Batch file prepared for the analysis of rainfall

Figure 2.5 presents the percentage missing computed from the six rainfall stations
in the Tana basin system.

Figure 2.5 Status of rainfall data in the Tana basin system

72
 2.4 Distribution of Precipitation

2.4.1 Temporal Distribution

Rainfall varies due to geographical location and features. Mountainous areas
experience great variability within relatively short distances and within elevation
changes. Coastal regions and immediate inland areas are subject to distinctive
variability in rainfall. In Ethiopia and throughout Africa, annual rainfall may be
unimodal or bimodal. Unimodal describes rainfall that occurs during one
consecutive season while bimodal rainfall occurs in two relatively separate
seasons (belg and kremt).
Seasonal distribution of rainfall helps in assessing the hydrological behaviour of
catchments and is a vital input for water resources assessments. The seasonal
distribution of rainfall data in the Tana basin system is presented in Figure 2.6.
40 40
Bahir Dar Debre Tabor
35 35

30 30
Sesonal Flow (Cumecs)

Sesonal Flow (Cumecs)

25
25

20
20

15
15

10
10

5
5

0
0
Jan Jan Mar Apr May May Jun Jul Aug Sep Oct Nov Dec
Date Jan Jan Mar Apr May May Jun Jul Aug Sep Oct Nov Dec
Date

40 40
Gumara Woreta
35 35

30 30
Sesonal Flow (Cumecs)
Sesonal Flow (Cumecs)

25 25

20 20

15 15

10 10

5 5

0 0

Jan Jan Mar Apr May May Jun Jul Aug Sep Oct Nov Dec Jan Jan Mar Apr May May Jun Jul Aug Sep Oct Nov Dec
Date Date

40
40
Yifag Addis Zemen
35
35

30
30
Sesonal Flow (Cumecs)
Sesonal Flow (Cumecs)

25
25

20
20

15
15

10 10

5
73 5

0 0

Jan Jan Mar Apr May May Jun Jul Aug Sep Oct Nov Dec Jan Jan Mar Apr May May Jun Jul Aug Sep Oct Nov Dec
Date Date
Figure 2.6 Seasonal distribution of rainfall in the Tana basin system

As shown in Figure 2.6 the main rain season starts from mid of May peaking in July
and August. The rainfall extends to September and in all stations a unimodal
rainfall distribution was observed.

2.4.2 Spatial Distribution

Spatial distribution of rainfall shows the variation of rainfall depth with space in
the catchment. Depth measured by two or more rain gages of the same type
will not be the same. In a spatial distribution of rainfall it is possible to determine a
spatial average of the rainfall depth over the catchment. An example of
average spatial rainfall distribution in Ethiopia is shown in Figure 2.7. Some times it
is possible to present spatial and temporal variation of rainfall over a catchment.
Figure 2.8 presents both spatial and temporal rainfall distribution over the Tana
basin system.

Figure 2.7 Spatial rainfall distribution of rainfall over Ethiopia

74
Figure 2.8 Spatial and temporal distribution of rainfall over the Tana basin system

2.5 Distribution and variability of Streamflow

A seasonal model presented in chapter one can be used to compute seasonal
flow (365 values each). Before the computation of seasonal flow variations,
basic hydrological analysis that includes percentage missing computation and
filling the missing record are important. The methodology presented in chapter
one was used here to evaluate flow data in the Tana basin system. The missing
records are filled-in using the techniques presented in chapter one, section 1.3.4,
flow data analysis.

Figure 2.9 Computation of missing flow records in the Tana basin system

75
 Figure 2.10 shows hydrological diagrams for the Tana basin catchment at five
gauging stations: Abay at the outlet, Ribb, Gumara, Gilgel Abay and Megech
gauging stations. Smoothening of the seasonal hydrological data was done
using 30 days moving average. The analysis helps designers in the field of water
resources practices to evaluate the availability and distribution of flows
seasonally and is a direct input in rainfall-runoff modelling techniques.
100 400
Megech Abay Flow
360

80 320

Sesonal Flow (Cumecs)

Sesonal Flow (Cumecs)

280

60 240

2.8 Base flow characteristics

200

40 160

120

20 80

40

0 0

Jan Jan Mar Apr May May Jun Jul Aug Sep Oct Nov Dec Jan Jan Mar Apr May May Jun Jul Aug Sep Oct Nov Dec
Date Date

200
Gumara
Ribb
80
160
Sesonal Flow (Cumecs)
Sesonal Flow (Cumecs)

120

40
80

40

0
0
Jan Jan Mar Apr May May Jun Jul Aug Sep Oct Nov Dec
Jan Jan Mar Apr May May Jun Jul Aug Sep Oct Nov Dec
Date
Date

250

Gilgel Abay
200
Sesonal Flow (Cumecs)

150

100

50

Jan Jan Mar Apr May May Jun Jul Aug Sep Oct Nov Dec

Figure 2.10 Seasonal flow variations in the Tana Basin Catchment

76
2.6 Streamflow Hydrograph
A streamflow hydrograph is a graph showing discharge (i.e., stream flow at the
concentration point) versus time. The various components of a natural
hydrograph are shown in Figure 2.11. At the beginning, there is only base flow
(i.e., the ground water contribution to the stream) gradually depleting in an
exponential form. After the storm commences, the initial losses like interception
and infiltration are met and then the surface flow begins.

Figure 2.11 Components of streamflow Hydrograph

(Source: Raghunath (2006), Hydrology: Second Revised Edition)

In gauged streams the hydrograph may have a single peak or multiple peaks
according to the complexity of storms. For flood analysis and derivation of unit
hydrograph, a single peaked hydrograph is preferred. Typical hydrographs at
two gauging station in the Tana basin system is shown in Figure 2.12 (a and b).

77
 800
Abay Flow at Bahir Dar
700

600 a
Flow (Cumecs)

500

400

300

200

100

0
01/01/1976

01/01/1978

01/01/1980

01/01/1982

01/01/1984

01/01/1986

01/01/1988

01/01/1990

01/01/1992

01/01/1994

01/01/1996

01/01/1998

01/01/2000

01/01/2002

01/01/2004

01/01/2006
Date

Gumara Flow

450

400

350

300 b
Flow (Cumecs)

250

200

150

100

50

0
01/01/1976

01/01/1978

01/01/1980

01/01/1982

01/01/1984

01/01/1986

01/01/1988

01/01/1990

01/01/1992

01/01/1994

01/01/1996

01/01/1998

01/01/2000

01/01/2002

01/01/2004

01/01/2006

Date

Figure 2.12 Streamflow Hydrograph at (a) Abay at Bahir Dar and (b) at Gumara
gauging station

2.7 Rainfall Runoff Models

The common rainfall runoff models that are widely used in water resources
practices include the conceptual models that require readily available data.
Hydrological model types and their applications have been reported in many
publications. Thus the reader is advised to refer various documents that are
available in the water bureau and postgraduate thesis works. Similarly enormous
information is available at the web page of the International Water
Management Institute, IWMI web page ([Link]). Here two very common
types of hydrologic models that are widely used in water resources practices are
presented. These include the Soil and Water Assessment Tool (SWAT) and HEC-
HMS.

78
2.7.1 The Soil and Water Assessment Tool (SWAT)
SWAT model was developed to predict the impact of land management
practices on water, sediment and agricultural chemical yields in large complex
watersheds with varying soils, landuse and management conditions over long
periods of time (Neitsch et al., 2002). SWAT allows a number of different physical
processes to be simulated in a watershed as shown in Figure 2.13.
Sub-basin Outlet
River
Weather
Flow
crop growth SENSITIVITY
flow
Precipitation O2 ANALYSIS
evaporation
sediments
Temperature NH4
infiltration
transport HNO3 MULTI-
Solar radiation
OBJECTIVE
percolation Org. N AUTO-
Humidity water quality
CALIBRATION
H2PO4
groundwater processes
Wind speed
Org. P

PET nutrients •QUAL2E

sediments

sediment •RWQM Algae

Figure 2.13 Processes simulated by SWAT

The climatic variables required by SWAT consist of the following:

• Daily precipitation,
• Maximum/minimum air temperature,
• Solar radiation,
• Wind speed
• Relative humidity.
The model allows the above information to be input from records of observed
data or generated during the simulation (Neitsch et al., 2002). Input information
for each subbasin is grouped or organized into the following categories: climate;
hydrologic response units or HRUs; ponds/wetlands; groundwater; and the main
channel, or reach, draining the subbasin. Hydrologic response units are lumped
land areas within the subbasin that are comprised of unique land cover, soil, and
management combinations.
Water balance is the driving force behind everything that happens in the
watershed (Neitsch et al., 2002) and SWAT uses equation 2.1 to simulate the
same.
t
SWt = SW0 + ∑ (Rday − Qsurf − E a − wseep − Q gw )
i =1 [2.1]

79
where SWt is the final soil water content (mm H2O), SW0 is the initial soil water
content on day i (mm H2O), t is the time (days), Rday is the amount of
precipitation on day i (mm H2O), Qsurf is the amount of surface runoff on day i
(mm H2O), Ea is the amount of evapotranspiration on day i (mm H2O), wseep is
the amount of water entering the vadose zone from the soil profile on day i (mm
H2O), and Qgw is the amount of return flow on day i (mm H2O).
Surface runoff was estimated using the Soil Conservation Service Curve Number
procedures (SCS, 1972). Surface runoff occurs whenever the rate of water
application to the ground surface exceeds the rate of infiltration, and the SCS
curve number is a function of the soil’s permeability, land use and antecedent
soil water conditions (Neitsch et al., 2002).
The SCS curve number equation is (SCS, 1972):

Qsurf =
(R day − Ia )
2

(Rday − Ia + S)
[2.2]
where Qsurf is the accumulated runoff or rainfall excess (mm H2O), Rday is the
rainfall depth for the day (mm H2O), Ia is the initial abstractions which includes
surface storage, interception and infiltration prior to runoff (mm H2O), and S is
the retention parameter (mm H2O). The retention parameter varies spatially due
to changes in soils, land use, management and slope and temporally due to
changes in soil water content. The retention parameter is defined as:
⎛ 1000 ⎞
S = 25.4⎜ − 10 ⎟
⎝ CN ⎠ [2.3]
where CN is the curve number for the day. The initial abstractions, Ia, is
commonly approximated as 0.2S and equation 8.2 becomes

=
(Rday − 0.2 S )
2

(R + 0.8S )
Qsurf
day
[2.4]
Runoff will only occur when Rday > Ia.
The schematic representation of SWAT model is presented in Figure 2.14.
Why SWAT model is chosen a widely applicable model world wide and in the
Ethiopian catchments is described below:
ƒ Impressive tool for distributed modeling!
ƒ Great manual/documentation (not often the case for comparable
models)
ƒ Good support while using it
ƒ Physically based, but full of assumptions and many parameters that are
often not/hardly measurable
-> very strong assumptions about the system under investigation!
ƒ Limitations through semi-distributed model structure

80
 ƒ Relatively simple, empirical approaches to describe complex hydrological
processes. Example: Green & Ampt approach somewhat more process
based than SCS approach, but still far away from reality
ƒ Great tool for Applied Hydrology and Research

Precipitation

Irrigation Rain Snow

Snow cover

Snow melt

Infiltration Surface Runoff

Transmission
Soil Storage Streamflow
Losses

Soil water Pond/Reservoir

Routing Water Balance
(10 layers) Irrigation
Diversion
P/R Evap.

Soil Evap. Irrigation

Transmission
Losses
P/R Outflow
Plant Uptake
and Route to
P/R Seepage
Tranpiration Next
Reach or
Reservoir
Lateral Flow

Shallow Aquifer
Percolation

Irrigation Revap Seepage Return

Flow

Deep Aquifer

Irrigation

Figure 2.14 Schematic of pathways available for water (After Neitsch et al., 2002)

81
[Link] SWAT Model Applications in the Ethiopian Catchments
SWAT hydrological model was applied in the Ethiopian catchments for various
purposes and its suitability in predicting streamflow, sediment yield and further
analysis of impacts of landuse and land cover change was reported.
Formulation of best management option for a watershed using SWAT model was
reported in the work of Biniam Biruck (2009) for the Anjeni Watershed in the Blue
Nile Basin. The report indicates the effective use of SWAT model to investigate
the flow and sediment processes in the watershed and comparison of different
potential land management options on alleviation of soil erosion. The model
predicted the soil loss from the catchment to be 550tons/year and 30% of the
loss was from the gully. In the study the management aspect of SWAT modelling
indicated that the presence of terraces in the watershed can save
2046tons/year. The model predicted that further construction of terraces can
save 932tons/year despite reducing the water availability in reaches for the
livelihood during the dry season by 30%. The management scenario indicated
that forestation of degraded areas and bush lands in the Anjeni watershed
prevent soil erosion with an amount of 333tons/year.

Application of SWAT model was extended to study the impacts of changes on

landuse/climate on reservoir storage on flooding or water availability (Zeray et
al.,2007, Tadele and Forch, 2007) in the Lake Ziway/ Southern Rift Valley Lakes
basins in Ethiopia. The SWAT modelling study by Zeray et al.,(2007) aimed at
quantifying the possible impacts of climate change on water resources
availability from the year 2001 to 2099 and to suggest possible adaptation
measures against this impact. In the study of climate change scenario the SWAT
model was initially calibrated and validated for the base period (1981-2000) and
computed the total inflow volume into the lake. The total inflow volumes of the
future periods were simulated by applying future changes of temperature and
precipitation (Zeray et al., 2007). In another study, the dynamics of Landuse
change in the Hare River basin located in the Southern Rift Valley Lakes Basin of
Ethiopia (Tadele and Forch, 2007) was studied using SWAT model. The purpose of
the study was to investigate the consequent impacts of landuse change on
streamflow.
Characterization and relationships among rainfall, runoff, and soil loss for Andit-
Tid watershed located 190km north of Addis Ababa was developed using SWAT
model. The study indicated that infiltration excess runoff from upslope watershed
is almost nonexistent. The model indicated that lower slopes produced high
runoff amount compared to those located at high slope. This gives new insights
about the effect of slope on runoff generation rather than the common belief
that the higher the slope the higher the runoff is. In the study the SWAT model
predicted the daily, weekly and monthly time steps streamflow at a reasonable
accuracy. The statistical model evaluation criteria, the coefficient of
determination, the Nash-Sutcliffe coefficient, and Root Mean Square Error values

82
showed good prediction efficiency of the model. Using SWAT model developing
predictable relations of hydrological behaviour and erosion hot spot areas will
help to further strengthen the importance of this tool in addition to saving
resource and time (Tegenu Ashagire, 2009). Some of the modelling results using
SWAT model at micro watersheds are presented in the Figures 2.15, 2.16 and 2.17
for the Anjeni and Andit Tid watersheds in the Amhara region.

Figure 2.15 . Flow hydrograph of observed and simulated flow from Anjeni
watershed during calibration period (After Biniam Biruck, 2009)

Figure 2.16. Sediment hydrograph of observed and simulated sediment from Anjeni
watershed during calibration period (After Biniam Biruck, 2009)

83
Figure 2.17 Comparison of daily predicted and measured discharge during model
calibration period (a) and model validation period (b) for the Andit Tid
watershed. (After Tegenu Ashagire, 2009).

Figure 2.18 shows the SWAT hydrologic simulation results at the outlet of the Blue
Nile river (Ethiopia Sudan border)

Observed Simulated

Observed Simulated
a 7000 b
10000 6000
9000
8000 5000
Flow (m3/s)
Flow ( m3/s)

7000
4000
6000
5000 3000
4000
3000 2000
2000
1000 1000
0
0
81

82

83

84

84

85

86

Jul-81

Jul-82

Jul-83

Jul-84

Jul-85

Jul-86
Jan-81

Jan-82

Jan-83

Jan-84

Jan-85

Jan-86
19

19

19

19

19

19

19
1/

1/

1/

1/

2/

2/

2/
/0

/0

/0

/0

/1

/1

/1
01

01

01

01

31

31

31

Time (day) Time(month)

Figure 2.18 Observed and Simulated Blue Nile daily discharge (a) and monthly
discharge (b) for calibration period between 1981 and 1986, at El Deim station
(Ethiopia/Sudan border).

84
SWAT modelling results shown above indicate the model simulated the flow very
well. Similarly the model could reproduce peak flows quite well. The model could
capture dry period characteristics well. The over all performance of the model
during calibration and validation was found to be very promising. These indicate
that the model can be a dependable tool for hydrological studies in the
Ethiopian catchments.

2.7.2 The HEC-HMS Model

HEC-HMS is the US Army Corps of Engineers’ Hydrologic Modeling System
computer program developed by the Hydrologic Engineering Center (HEC). The
program simulates precipitation-runoff and routing processes, both natural and
controlled. The model is developed over 30years continuous research and
improvement. It is available freely in public domain.
The rainfall-runoff simulation process is achieved through four major
components; runoff volume component, runoff transform, routing and base flow
components. In each component there are several models that will allow the
user alternative options that suit the watershed conditions. The model can be run
as lumped or distributed model. The runoff generating mechanism of HEC-HMS
model is depicted in Figure 2.19 below.

RAINFALL

Rainfall loss Direct runoff component

component
PERVIOUS SURFACE IMPERVIOUS SURFACE

LOSSES
DIRECT RUNOFF

RIVER CHANNEL
BASEFLOW
RESERVIOUR ROUTING
Baseflow component
Routing component

BASIN OUTLET

85
Figure 2.19: Runoff generation process in HEC-HMS

The major component in the rainfall-runoff transformation is the runoff volume

component. The runoff volume is computed with the Natural Resources
Conservation Centre Curve Number (NRCS-CN) and formerly known as the Soil
Conservation Service (SCS) model developed by the United States Department
of Agriculture (USDA, 1972). The method is commonly called SCS method and is
presented in equations 2.3 and 2.4. It computes direct runoff through an
empirical equation that requires the rainfall and a watershed coefficient inputs
(Nayak et al., 2003).

[Link] Application of HEC-HMS Model in Ethiopian Catchments

The HEC-HMS model has been applied in various catchments of Ethiopia with
different climatic conditions. Figure 2.20a and 2.210b show the simulation result
of HEC-HMS model for Lailay Wukro and GumSelassa watersheds in Tigay regions.

60 30

50
a 25 b
Discharges (m3/s)
Discharge (m3/s)

40 20

30 15

20 10

10 5

0 0

Time (Days) Time (Days)

Observed Discharges Simulated Discharges Observed Discharge Simulated Discharge

Figure 2.20 Observed and simulated (modelled) hydrographs for Laelay Wukro
watershed (After Mohammod, 2009).

As presented in Figure 2.20a and 2.20b the flow graph comparison shows the
fitness of the simulated hydrograph with the observed hydrograph. For most of
the events the shape of the hydrograph has been replicated successfully,
except at the start and end of the hydrograph for some events. This indicates its
suitability for event based and continuous simulation of flow.

86
 CHAPTER THREE

RESERVOIR EVAPORATION AND SEDIMENTATION ESTIMATION

3.0 Introduction
This chapter presents the basic concept of evaporation and parameters
required to estimate evaporation. Method of computing free water surface
potential evaporation and evaporation from watersheds that use crop
characteristics will be presented. Simpler methods to compute actual
evaporation for the case of limited data case will be briefly presented.
Direct measurement of evaporation or evapotranspiration from large water or
land surfaces is not possible, but lysimeters are used to measure it in limited-
volume containers. Several indirect methods have been developed that give
acceptable results (WMO, 1994), evaporation pans and lysimeters are used in
networks, and for reservoirs and small catchments, estimates can be made by
water-budget, energy-budget and aerodynamic approaches (WMO, 1994).
Evaporation from the land surface comprises evaporation directly from the soil
and vegetation surface, and transpiration through plant leaves, in which water is
extracted by the plant’s roots, transpired upwards through its stem, and diffused
into the atmosphere through tiny openings in the leaves called stomata. The
process of evaporation from the land surface and transpiration from vegetation
are collectively termed evapotranspiration. The potential evapotranspiration is
the evapotranspiration that would occur from a well vegetated surface when
moisture supply is not limiting. Actual evapotranspiration drops below its
potential level as the soil dries out (Chow, 1988).

3.1 Method of estimating evaporation using climatic data

The potential evaporation can be estimated in several ways based on observed
meteorological parameters. Methods commonly applied (Dingman, 2002;
Chow, 1988) in hydrological studies, are as follows:

Energy balance method

Evapotranspiration process is determined by energy exchange at the
vegetation surface and limited by the amount of energy available.
Evapotranspiration therefore can be determined by applying the principle of
conservation of energy. It states that energy into the surface equals energy out
at the same time period. The equation for the conservation of energy is in the
form;
Rn − G − λET − H = 0 [3.1]
Where Rn is net radiation, H is sensible heat, G is the soil heat flux and λET is latent
heat flux.

87
Aerodynamic method
The aerodynamic method for estimating potential evapotranspiration was first
applied by Thornthwaite and Holzman (1942). A relationship involving the
gradients of specific humidity q and the logarithmic wind profile was proposed.
The equation is of the form
(q − q1 )(U 2 − U 1 )
ET = ρ a k 2 2 [3.2]
2
⎛ z2 ⎞
In⎜⎜ ⎟⎟
⎝ z1 ⎠
Where ρa is density of moist air, k is Von Karman’s constant; U1 and U2 are wind
speeds at heights Z1 and Z2. Over a rough cropped surface z-d is substituted for z.
Wind speed and humidity at a number of heights above the surface as well as
temperature are requirements for this approach.

Temperature Method
In certain regions of the world, meteorological and climatological data may be
quite limited. Models based almost solely on air temperature may be used in
such cases to provide estimates of ET (Chong-Yu Xu, 2002). The temperature
methods are some of the earliest method for estimating evapotranspiration
(Jensen et al., 1990) and if estimates are made for periods of several weeks or a
month, reasonable approximations are possible (Chong-Yu Xu, 2002). Most
temperature based equations take the form of:
ET = cT a [3.3]
Where ET is Potential evapotranspiration, a and c are constants and T is air
temperature.

Penman Combination method

The Penman combination method is suited for application for small catchments
areas with detailed Climatological data. The data required are daily net
radiation, relative humidity, wind speed, air pressure and temperature. The
Penman combination is given as,
∆ γ
E= Er + Ea [3.4]
∆ +γ ∆ +γ
Where E is evapotranspiration (mm/day), ∆ is gradient of saturated vapor
pressure curve at air temperature Ta, γ is psychrometric constant, Er is radiation
term and Ea is aerodynamic term.

3.1.1 Methods of Parameters Estimation for Penman Combination

1. Saturation vapor pressure gradient, ∆
For the estimation of evapotranspiration, the slope of the relation between
saturation vapor pressure and temperature ∆ is required. This is estimated by
4098es
∆= (Pa/oC) [3.5]
(237.3 + T )2

88
Where T is temperature in oC, es is saturation vapor pressure estimated by
⎛ 17.27T ⎞
es = 611exp⎜ ⎟ Pa (N/M2) [3.6]
⎝ 237 . 3 + T ⎠
2. Actual vapor pressure, ea
This is estimated by ea = RHes (Pa) [3.7]
Where RH is relative humidity (0≤RH≤1) and es is saturation vapor pressure.
3. Psychrometric constant, γ
The psychrometric constant, γ is estimated by
Cp Kh P
γ= (Pa oC) [3.8]
0.622lv K w
K
Where ratio h of heat and vapor diffusivities is 1 (Priestley and Taylor, 1972)
Kw
P is pressure in Pa, Cp is specific heat at constant pressure and equals to 1005
J/Kg.K
lv is latent heat of vaporization and is expressed as
lv = 2.501 × 10 6 − 2370T (J/Kg) [3.9]
4. Atmospheric pressure, P

The atmospheric pressure, P is the pressure exerted by the weight of the earth’s
atmosphere and is expressed as
5.26
⎛ 293 − 0.006 z ⎞
P = 101.3⎜ ⎟ (KPa) [3.10]
⎝ 293 ⎠
Where z is elevation above the sea level (m)

5. Radiation term, Er
Radiation term Er is calculated from the equation
R
E r = n (mm/day) [3.11]
lv ρ w
Where Rn is net radiation W/m2 and ρw is density of water (997Kg/m3).

6. Aerodynamic term Ea
Aerodynamic term Ea is calculated from the equation
E a = B(es − ea ) (mm/day) [3.12]
0.102u 2
Where B = 2
mm/[Link] [3.13]
⎡ ⎛ z 2 ⎞⎤
⎢ln⎜⎜ ⎟⎟⎥
⎣ ⎝ z 0 ⎠⎦
u2 is wind velocity (m/s) measured at height Z2 (cm) and Zo is roughness height of
natural surface.
For the evapotranspiration estimation Doorenbos and Pruitt (1977)
recommended

89
 ⎛ u ⎞
B = 0.0027⎜1 + ⎟ mm/[Link] [3.14]
⎝ 100 ⎠
Where u is the 24 hour wind run in km/day measured at height of 2 metres. In this
study approximate values of the roughness height of natural surfaces were taken
referring to Chow (1988).
Penman for East Africa
McCulloch (1965) modified the Penman equation to estimate
evapotranspiration in any part of East Africa as
E=
∆ ⎡
⎢
∆ +γ ⎣
⎛ n ⎞⎤
S A (1 − ρ )⎜ 0.29Cos (ϕ ) + 0.52 ⎟⎥ −
∆ ⎡ 4⎛
⎢
N ⎠⎦ ∆ + γ ⎣
n⎞
( ) ⎤
σTa ⎜ 0.1 + 0.9 ⎟ 0.56 − 0.8 e ⎥
⎝ ⎝ N⎠ ⎦
γ ⎡ ⎛ h ⎞ ⎤
+ ⎢ 0.26⎜1 + ⎟(es − e )⎥ [3.15]
∆ +γ ⎣ ⎝ 20000 ⎠ ⎦
Where SA is short wave radiation at the edge of the atmosphere, ρ is albedo
(0.25 for grassland and 0.05 for open water surface), φ is latitude in degrees
decimal points, n is the mean observed sunshine hours (Hours), N is the possible
maximum sunshine duration (hours), σ is the Stefan Boltzmann constant (5.67 *10-
8 W/m2.K4), T
a is the mean air temperature (Kelvin), h is the station altitude
(meters), es is saturation vapor pressure (KPa) and e is actual vapor pressure
(KPa). The value of ∆ is computed using the equation (3.5) and the psychometric
constant (γ) in Pa/oC is estimated using equation (3.8)
Doorenbos and Pruitt (1977) compared the computed and measured values of
evapotranspiration at many locations and concluded that the Penman
combination method is the best approach, especially if the vapor transport
coefficient B in equation (3.13) is calibrated for local conditions.

PENMAN-MONTEITH METHOD
The Penman-Monteith equation combines components that account for energy
needed to sustain evaporation, the strength of the mechanism required to remove the
water vapor and aerodynamic and surface resistance terms. The Penman-Monteith
equation is:

λE =
[ ]
∆ ⋅ (H net − G ) + ρ air ⋅ c p ⋅ e zo − e z ra
∆ + γ ⋅ (1 + rc ra ) [3.16]
where λE is the latent heat flux density (MJ m-2 d-1), E is the depth rate evaporation
(mm d-1), ∆ is the slope of the saturation vapor pressure-temperature curve, de/dT (kPa
˚C-1), Hnet is the net radiation (MJ m-2 d-1), G is the heat flux density to the ground (MJ
m-2 d-1), ρair is the air density (kg m-3), cp is the specific heat at constant pressure (MJ
o
kg-1 ˚C-1), ez is the saturation vapor pressure of air at height z (kPa), ez is the water
vapor pressure of air at height z (kPa), γ is the psychrometric constant (kPa ˚C-1), rc is
the plant canopy resistance (s m-1), and ra is the diffusion resistance of the air layer
(aerodynamic resistance) (s m-1).

90
3.1.2 Estimation of Actual Evaporation
Actual evaporation is estimated using data of rainfall, flow, potential
evaporation, maximum and minimum temperature, humidity and wind speed
using methods described below. The availability of concurrent data will assist the
estimation of actual evaporation for catchments under question. The rainfall
and discharge data are also required to estimate the long term water loss from
catchment.

Actual Evaporation using water budget method

The Water budget approach is based on the principle of conservation of mass or
the continuity equation as often referred to by hydrologists. The principle requires
that for any given volume and during any period of time, the difference
between the inflow and outflow rates is equal to the rate of change of the water
stored. This is given mathematically as:
⎛
⎜⎜ P +
(Qri + Q gi ) ⎞ ⎛ (Qro + Q go )
⎟⎟ − ⎜⎜
( ⎞)
+ E ⎟⎟ ± δ =
dS
dt [3.17]
⎝ A ⎠ ⎝ A ⎠ A
Where P is mean precipitation rate directly on the given volume, Qri is surface
inflow rate, Qro is surface outflow rate, Q gi is groundwater inflow rate, Q go is
groundwater outflow rate, E is mean evaporation rate, A surface area under
consideration, ds is the rate of change of storage, and δ is measurement
dt
errors.

Re-arranging the above equation, we have

P + qr − E = q g +
(dS )
dt ± δ [3.18]
A
Where q r = (Qri − Qro ) / A is the mean net surface inflow rate per unit area, and
qg = (Qgo − Qgi ) / A is the mean net groundwater outflow rate per unit area.
On monthly basis, the q g term of equation (3.18) could be assumed to be
negligible compared to the other terms. Hence, assuming measurement errors
equal zero, the equation on monthly basis becomes

E = P + qr −
(
dS
dt
) [3.19]
A
If the factors in equation (3.19) are averaged over a long period and, the
conditions else where in the catchment remain unchanged, then the net
change in storage can usually be considered as zero (Dingman, 2002), thus on
annual basis it becomes:

E = P + qr [3.20]

91
For smaller catchments/basins without water bodies such as reservoirs, wetlands
and /or lakes within them, equation (3.19) reduces to equation (3.20).
Actual Evaporation using Modified Turc and Pike’s (1964)
The formula for Turc and Pike (Pike, 1964) for estimating actual
evapotranspiration is expressed as:
P
AE PE
= [3.21]
PE 2
⎛ P ⎞
1+ ⎜ ⎟
⎝ PE ⎠
Where AE is Actual Evapotranspiration, PE is Potential Evapotranspiration and P is
precipitation.

3.2 River sediment inflow estimation: Prediction of a Reservoir Life

Determination of sediment loads, both bed and suspended loads, is important in
all water resources studies. It is important for (i) evaluating potential
sedimentation of reservoirs, lakes & ponds; (ii) designing water treatment plants,
and (iii) designing reservoirs.
Various methods could be employed for estimation of sediment loads in rivers,
provided that adequate and reliable records of flow and sediment data are
available. In the absence of complete and long period data, the flow duration
curve sediment rating curve method is used to better approximate the sediment
loads.

3.2.1 Sediment Load Estimate

Usually there is limited number of suspended sediment load measurements while
the discharge measurement is often having long years of record. On the basis of
limited suspended sediment concentration data available, develop sediment
rating curve in the form of equation 3.22. The plot of the corresponding rating
curve is also shown in Figure 3.1.

Qs = ∂ Q β [3.22]
Where Qs is the suspended sediment discharge in tons/day and Q is the
discharge in m3/s.

92
 1.0E+05

1.0E+04
SS Discharge (t/da

1.0E+03

1.0E+02

Qs = 124.02Q1.25
1.0E+01 R2 = 0.52

1.0E+00
0.1 1 10 100

Flow (m^3/s)

Figure 3.1 Suspended Sediment Rating Curve of Bilate River at Alaba Kulito

From the long series of flow data, a flow-duration curve is developed. It is

therefore possible to estimate the suspended sediment by associating it to the
already developed sediment rating curve equation. Eventually, we can estimate
the annual suspended sediment load in tons/year.
As is well known, the river sediment load is also comprised of the bed load
component, which is sometimes referred to as unmeasured load. Whenever
there is no measured bed load data it is usually assumed to be a certain
percentage of the suspended load. In literature (Yang, 1996), it stated that the
bed load varies from 5 to 25% of the suspended load; the higher percentages
applying for mountain and coarse-bed rivers.

3.3. Predicting Reservoir Life

In estimating the useful life of a reservoir, the correct prediction of the density of
the deposited sediment is an important factor. Lane and Koelzer (1943) gave the
equation for the dry specific weight γ t after time t years as:
γ t = γ i + K log10 t [3.22]
Where γ i is initial specific weight and K is a constant for the rate of compaction
The useful capacity of reservoir lost each year by sediment deposition is:
Vs = Q sη trap [3.23]
Where Vs is volume of useful capacity of reservoir lost each year
Qs is annual sediment inflow into the reservoir
η trap trap efficiency of the reservoir

93
While allocating space for the dead storage in the reservoirs (i.e., to provide
space for sediment deposition during the life of the project) the trap efficiency is
taken as at least 95% and rarely below 90%. Sediment deposits in the upper end
of the reservoirs generally become covered by vegetation resulting in heavy
evapotranspiration loss of the available water, which is more critical in arid
regions.

Example
A proposed reservoir in the Tana basin has a capacity of 400 ha-m. The
catchment area is 130 km2 and the annual stream flow averages 12.31 cm of
runoff. If the annual sediment production is 0.03 ha-m/km2, what is the probable
life of the reservoir before its capacity is reduced to 20% of its initial capacity by
sediment deposition. The relation between trap efficiency and capacity inflow
ratio is given below.

Table 3.1 Relationship of trap efficiency and capacity inflow ratio

Capacity Trap Capacity Trap Capacity Trap

Inflow Efficiency Inflow Ratio, Efficiency Inflow Efficiency
Ratio, C/I C/I Ratio, C/I
η trap η trap η trap

1 97.5 0.06 80 0.006 31

0.7 97 0.05 77 0.005 27

0.6 96.5 0.04 74 0.004 20

0.5 96 0.03 68 0.003 13

0.4 95.5 0.02 60 0.002 2

0.3 95 0.015 52

0.2 93 0.01 43

0.1 87 0.008 38

0.07 82 0.007 36

94
Solution:
The useful life may be computed by determining the number of years required
for each incremental loss of reservoir capacity (i.e., for the decreasing values of
capacity-inflow ratios) up to the critical storage volume of 400 × 0.20 = 80 ha-m
as tabulated below:

Capacity Capacity* Trap efficiency Annual** Loss of No. of

C Inflow ηtrap (%) Sediment Reservoir Years for
Trapped Capacity The
capacity
∆C
loss

( ha − m ) C for the Ave. for Vs = Qs xη trap ( ha − m ) ∆C ÷ Vs

I C *Increment
ratio
I

400 0.25 94

320 0.2 93 93.5 3.64 80 22

240 0.15 90 91.5 3.57 80 22.4

160 0.1 87 88.5 3.45 80 23.2

80 0.05 77 82.0 3.2 80 25

Total= 92.6 say,

93 yrs

12.31 130 x10 6

*Average annual inflow, I = x = 1600ha − m
100 10 4
C 400
For reservoir capacity C=400ha-m, = = 0.25
I 1600
** Annual sediment inflow into the reservoir
Qs = 0.03x130 = 3.9ha − m

3.3 Reservoir Sedimentation Control

Sediment deposition in reservoirs can not be actually prevented but it can be
retarded by adopting some of the following measures:
(viii) Reservoir sites, which are prolific sources of sediment, should be avoided.
(ix) By adopting soil-conservation measures in the catchment area, as the silt
originates in the watershed.

95
(x) Agronomic soil conservation practices like cover cropping, strip cropping,
contour farming, suitable crop rotations, application of green manure
(mulching), proper control over graze lands, terracing and benching on
steep hill slopes, etc. retard overland flow, increase infiltration and reduce
erosion.
(xi) Contour trenching and afforestation on hill slopes, contour bunding gully
plugging by check dams, and stream bank stabilisation by the use of
spurs, rivetments, vegetation, etc. are some of the engineering measures
of soil conservation.
(xii) Vegetal cover on the land reduces the impact force of rain drops and
minimises erosion.
(xiii) Sluice gates provided in the dam at various levels and reservoir operation,
permit the discharge of fine sediments without giving them time to settle to
the bottom.
(xiv) Sediment deposits in tanks and small reservoirs may be removed by
excavation, dredging, draining and flushing either by mechanical or
hydraulic methods and sometimes may have some sales value.

3.4 Reservoir Capacity Estimation

The reservoir capacity is calculated using the formula:
h
V= ⋅ ( Ares1 + Ares 2 + ( Ares1 × Ares 2 ) 0.5 )
3 [3.24]
Where:
V 1,2 reservoir capacity (volume) between two successive elevations (m3)
h elevation difference between successive contours (m)
Ares1 and Ares 2 area of reservoir water spread at elevation h1 and h2 (m2)

96
 CHAPTER FOUR

Frequency Analysis of Rainfall and Determination of Peak Flow

4.1 Statistical Distributions commonly used in Hydrology

One of the first steps in many hydrologic design projects, such as in urban
drainage design and design of hydraulic structures involves the determination of
the rainfall event or events to be used. The most common approach is to use a
design storm or event that involves a relation ship between rainfall intensity
(depth), duration, and the frequency or return period appropriate for the facility
and site location. In the present work the commonly used statistical distribution
techniques are discussed. Different temporal resolutions of data that are
pertinent to develop Intensity Duration Frequency (IDF) relationship are
discussed. As an example of IDF development, local IDF curve is established for
the Bahir Dar synoptic station using the annual maximum series of the short
duration data obtained from National Metrological Service Agency (NMSA).
Three methods i.e. the Extreme Value Type I (EVI) or Gumble distribution, Log
Pearson Type III and Lognormal frequency distributions were applied for analysis
in developing the IDF relationship. Finally hydrologic methods are discussed to
estimate flood magnitudes for the desired frequency of occurrence.

4.2 Local IDF Curve Construction

4.2.1 Rainfall Data Analysis

A rainfall data of short time durations (10 minute, 30 minutes, 1 hour and 24 hour)
data were collected form NMSA. A concurrent data period from 1971 to 1982
was chosen for analysis. Close examination of the raw data shows that there are
some errors in the recorded data. While choosing for the annual maximum data
it was observed that the 30 minute rainfall data for the year 1972 in the month of
July and August (31.1mm and 23.4mm) is greater than that of the hourly
maximum rainfall data (16mm and 15.2mm) which is an error and is difficult to
trace back the reason for some 20 to 30 years back. Thus during IDF generation
the data of the 30 minute rainfall duration was used in both cases for the month
of July and August in the year 1972. The annual maximum data used for rainfall
analysis from the period 1971 to 1982 for the different rainfall durations is
presented in Table 4.1.

The Extreme Value Type I, Log normal and Log Pearson Type III frequency
distributions were used to select which distribution fits the data in a best way and
thus, to determine the extreme values for the different return periods. Finally the

97
IDF curves were plotted in Log-Log paper from which the rainfall intensities of the
required time of concentration would be derived.

Table 4.1 Annual Maximum Rainfall Data at Bahir Dar Observatory Station

Year Maximum Annual Rainfall(mm)

for the Duration Di
10minute 30minute 1hour 24hour
1971 6 20.3 27.3 72
1972 9.7 31.1 31.1 62
1973 10 20 49.8 116.3
1974 16.8 30 44.9 77
1975 18 44 44.7 64.3
1976 14 32 45.1 73.4
1977 20 36.5 46.2 133.2
1978 20 40 44.6 61.1
1979 19 33.5 39.7 56.2
1980 19 38.3 44.6 79.9
1981 19 57 85.5 108.7
1982 7.7 12.2 25.9 35.9
Mean 14.93 32.91 44.12 78.33
[Link] 5.21 11.92 15.24 27.79

Maximum 10 Minutes Rainfall

For the 10 minute rainfall duration, the highest rainfall for the recording period is
20mm and was occurred during June 1977 and September 1978 which amounts
to 26% and 20% of the annual value of the same year. The mean and standard
deviation for the recording period as computed in Table 4.1 are 14.93 and 5.21
respectively. The derived rainfall intensity for the 10 minute rainfall duration is
given in Table 4.2.

Table 4.2 Maximum Rainfall Intensity of 10 Minute Rainfall Duration.

Duration Intensity of Rainfall(mm/hr) for the indicated frequency T (yrs)
(minute) 2 5 10 25 50
10 89.220 119.468 134.428 148.859 157.573

Maximum 30 Minutes Rainfall

For 30 minutes rainfall duration, the highest rainfall for the recording period is
57mm which occurred in August 1981 and amounts to 67% of the annual value in
the same year. The mean and standard deviation for the recording period as

98
shown in Table 4.1 are 32.91 and 11.92 respectively. The derived rainfall intensity
for the 30 minute rainfall duration is given in Table 4.3.

Table 4.3 Maximum Rainfall Intensity of 30 Minute Rainfall Duration.

Duration Rainfall Intensity (mm/hr) for the indicated frequency T (yrs)
(minute) 2 5 10 25 50
30 65.457 87.240 98.094 108.696 114.844

Maximum One Hour Rainfall

For 1 hour rainfall duration, the highest rainfall for the recording period is 85.5
which occurred in July 1981 and amounts to 40% of the annual value. The mean
and standard deviation for the recording period as shown in Table 4.1 are 44.12
and 15.24 respectively. The derived rainfall intensity for the 1 hour rainfall
duration is given in table 4.4.

Table 4.4 Maximum Rainfall Intensity of 1 hour Rainfall Duration.

Duration Rainfall Intensity (mm/hr) for the indicated frequency T (yrs)
(minute) 2 5 10 25 50
60 42.425 54.997 64.349 77.361 87.976

Maximum Daily Rainfall

Maximum daily rainfall data were obtained from NMSA for the period of 1971-
2004. In choosing for the concurrent data record only the data from the year
1971 to 1982 were used to fit with the other rainfall duration data. The mean and
standard deviation for the recording period as shown in Table 4.1 for the
maximum daily rainfall are 78.33 and 27.79 respectively. The adopted rainfall
intensity for the 24 hour rainfall duration is given in Table 4.5.

Table 4.5 Maximum Rainfall Intensity of 24 hour Duration.

Duration Rainfall Intensity (mm/hr) for the indicated frequency T (yrs)
(minute) 2 5 10 25 50
1440 3.174 4.183 4.756 5.391 5.811

4.2.2 Frequency Analysis

Hydrologic systems are some times impacted by extreme events, such as severe
storms, floods and droughts. The magnitude of an extreme event is inversely
related to its frequency of occurrence. Very severe events occurring less
frequently than more moderate events. The objective of frequency analysis is to
relate the magnitude of extreme events to their frequency of occurrence
through the use of the probability distributions.

99
The study of extreme events in this study involves the selection of largest value
form the recorded period. In which case the study of IDF curve utilizes, the
largest extreme value of rainstorms observed during the specified duration in
each year at a given meteorological station.

There are a number of frequency distributions in use. Annual maximum short time
rainfall durations and daily amounts ordinarily conform to a Gumbel Type I, Log
Pearson, Lognormal and Gamma distributions (Linsley et al., 1982).Among these,
the most commonly used distributions namely are Extreme value Type I or
Gumbel distribution, Log Pearson Type III and Log normal distributions.

4.3 Methods for Frequency Analysis

4.3.1 The Extreme Value Distribution or Gumble Distribution

The Gumble extreme-value distribution is the most popular distribution and has
received the highest application for estimating large events in various parts of
the world. This distribution has been used for rainfall depth-duration-frequency
studies (Hershfield, 1961).
Extreme value distribution has been also been widely used in hydrology. Storm
rainfall analyses have been carried out most commonly by the EVI distribution
(Chow, 1988, and Thomilison, 1980).

The probability density function, P ( x ) for type I distribution is (Haan,1977):

⎧⎛ X − β ⎞ ⎛ X −β ⎞⎫
⎨⎜ ± ⎟ − exp⎜ m ⎟⎬
⎩⎝ α ⎠ ⎝ α ⎠⎭
Px ( X ) = exp [4.1]
α
− ∞ < X < ∞,−∞ < β < ∞andα > 0

The negative and positive signs apply to maximum and minimum values (events)
respectively. Here, the parameters α and β are scale and location factors with β
being the mode of the distribution.
⎛X −β ⎞
Using a transformation Y = ⎜ ⎟ ,the extreme value (type I) distribution of
⎝ α ⎠
Equation 4.1 becomes:

Py (Y ) = exp(± y − exp( m y ) ) [4.2]

And the cumulative distribution F(Y) will be:

100
 y

∫ exp(m t − exp(m t ))dt ,−∞ < Y < ∞

−∞

F (Y ) = exp( − exp( − y )) for maximum event (extreme values) [4.3a)

F (Y ) = 1 − exp( − exp( − y )) for minimum events (extreme values) [4.3b]

The probability that an observation will exceed the mean value of the
distribution is 1 − F (Y ) .Regarding the estimation of parameters (α and β), Lowery
and Nash (1970) concluded that the method of moments has satisfactory results
as the other methods. Therefore, using this method:
S
α= = 0.7794 * S [4.4]
1.283
−
And β = X − 0.45 * S [4.5]
−
Where S and X are the standard deviation and mean of the sample
Similarly, Chow (1988) has given the type I (EVI) cumulative distribution function
in the same was as follows:
⎛ ⎛ ⎛ X − µ ⎞⎞⎞
F ( X ) = exp⎜⎜ − exp⎜⎜ − ⎜ ⎟ ⎟⎟ ⎟⎟,−∞ ≤ X ≤ ∞ [4.6]
⎝ ⎝ ⎝ α ⎠⎠⎠

Where α and µ are the parameters and estimated from the following
relationships:

6 *S −
α= = 0.7797 * S And µ = X − 0.5772 * α [4.7]
π

Where µ is the mode of the distribution point of maximum probability density and
‘X’ is the variant for the historiacally observed data)
X −µ
A reduced Variate ‘Y’ can be defined as Y = and substituting Y into
α
equation (4.6) yields:

F ( X ) = exp( − exp( −Y )) [4.8]

101
Simplifying and solving equation (4.8) for Y gives:
⎛ ⎛ 1 ⎞⎞
Y = − ln⎜⎜ ln⎜⎜ ⎟⎟ ⎟⎟ [4.9]
⎝ ⎝ F (X ) ⎠ ⎠
T −1
The value F ( X ) = [4.10]
T
Therefore, for the EVI distribution, XT is related to YT as follows:

X −µ XT − µ
Y= , YT = [4.11]
α α
Hence, X T = µ + α * YT

4.3.2 Lognormal Distribution

The two parameters (the mean and standard deviation) lognormal distribution
has been some times used for rainfall intensity duration analysis. If the random
variable Y = Log ( X ) is normally distributed, then X is said to be log normally
distributed. This logarithmic transformation of the normal distribution (i.e. the
probability density function) is given as follows:

1 ⎛ ( y − µ y )2 ⎞
f ( x) = exp⎜ − ⎟ [4.12]
xδ y 2π ⎜ 2δ y
2 ⎟
⎝ ⎠
Y = log( X ), µ y δy
Where and are the mean and standard deviation of the
−
population which are equivalent to X and S of the sample.

4.3.3 Log Pearson Distribution

Log Pearson distribution is a logarithmic transformation of the Gamma
distribution. If log X follows a Pearson Type III distribution, then X is said to follow a
log Pearson distribution. Log Pearson distribution has got a wide application in
the analysis of rainfall intensities. The fit of the distribution to data can be
checked by using the Chi-Square test or probability plotting.
Chow (1954) has given that the following general equations for any distribution
from which the T –year event magnitude can be computed.

102
 −
X T = X + KS x [4.13]
−
Where x is the event magnitude of the record, X and Sx are the mean and
standard deviation of the sample series, K is the frequency factor defined by a
specific distribution, and is the function of the probability level x.

4.4 Fitting the Rainfall Data Using EVI Type I distribution

Probability plotting position formula, which refers to the probability assigned to
each piece of data to be plotted, is used to fit the rainfall data to EVI Type I
distribution.
The plotting position formula for EVI distribution is using the formula developed by
Gringorten (1963) (Linsley, et al.1982) and is given as:

m − 0.44
P= [4.14]
n + 0.12

Where m and n are the rank and size of the data. The reduced Variate YT is
calculated as given in equation:

⎛ ⎡ T ⎤⎞
YT = − ln⎜⎜ ln ⎢ ⎥ ⎟⎟ [4.15]
⎝ ⎣ T − 1⎦ ⎠
Where T = 1 / P is the reciprocal of the exceedance probability of an event. The
computation of the probability plotting position is shown in Table 4.6. The
probability plotting is shown in Figures 4.1 and 4.2.

Table 4.6 Plotting Position and reduced Variate for EVI distribution

Annual Maximum Rainfall(mm) Rank P=(m-0.44) Red

for the indicated Duration Di (m) /(n+0.12) T=1/P Variate
10minute 30minute 1hour 24hour (YT)
20 57 85.5 133.2 1 0.046 21.79 3.058
20 44 49.8 116.3 2 0.128 7.82 1.989
19 40 46.2 108.7 3 0.210 4.77 1.446
19 38.3 45.1 79.9 4 0.292 3.43 1.064
19 36.5 44.9 77 5 0.374 2.68 0.759
18 33.5 44.7 73.4 6 0.456 2.19 0.497
16.8 32 44.6 72 7 0.538 1.86 0.259
14 31.1 44.6 64.3 8 0.620 1.61 0.034
10 30 39.7 62 9 0.702 1.43 -0.190
9.7 20.3 31.1 61.1 10 0.784 1.28 -0.426
7.7 20 27.3 56.2 11 0.866 1.16 -0.697
6 12.2 25.9 35.9 12 0.948 1.06 -1.081
103
 61

51
10 Minute
Rainfall (mm)

41 30 Minute

31

21

11

1
-1.08 -0.70 -0.43 -0.19 0.03 0.26 0.50 0.76 1.06 1.45 1.99 3.06
Reduced variate (YT)

Figure 4.1Fitting the Observed values of 10min and 30 min data

150
140
130
120 1 Hour
110
Rainfall (mm)

100 24 Hour
90
80
70
60
50
40
30
20
-1.08 -0.70 -0.43 -0.19 0.03 0.26 0.50 0.76 1.06 1.45 1.99 3.06
Reduced variate (YT)

Figure 4.2 Fitting the Observed values of 1hour and 24 hour

4.4.1 Fitting the data to Lognormal and Log Pearson type III Distribution
Plot of the annual maximum rainfall were also made against lognormal and Log
Pearson type III distribution for comparison purpose to decide which frequency
distribution best fits the data. The value of the standard normal variable, Z, was
used as horizontal axis to linearize the plot, which is equivalent to using normal
probability plotting paper.
In plotting the distributions Blom’s (1958) plotting position with value of b equal to
3/8 for lognormal and 4/8 for Pearson type III were used. The standard normal

104
variable, Z was related to the probability plotting position, P as follows (Chow et
al, 1988)
2.516 + 0.8028w + 0.0103w 2
Z = w−
1 + 1.4328w + 0.1893w 2 + 0.0013w 3 [4.16]
1/ 2
⎛ 1 ⎞
w = ln⎜ 2 ⎟
Where ⎝P ⎠ 0<P<0.5
And P=1-P where P>0.5 (In which case the value of Z is given a negative sign)
m − 3/8 m − 0 .5
P= P=
n + 0.25 for Lognormal and n for log Pearson type III distribution.

Table 4.7 The standard normal variables, Z for Lognormal and Log Pearson type III
distributions.

Maximum Annual Rainfall(mm) Rank Log Pearson type III Lognormal

for the indicated Duration Di (m) P= p=(m-0.375)
10min 30min 1hour 24hour (m-0.5)/n W Z /(n+0.25) W Z
20 57 85.5 133.2 1 0.042 2.521 1.73 0.051 2.439479 1.64
20 44 49.8 116.3 2 0.125 2.039 1.15 0.133 2.009984 1.11
19 40 46.2 108.7 3 0.208 1.771 0.81 0.214 1.755246 0.79
19 38.3 45.1 79.9 4 0.292 1.570 0.55 0.296 1.560559 0.54
19 36.5 44.9 77 5 0.375 1.401 0.32 0.378 1.395743 0.31
18 33.5 44.7 73.4 6 0.458 1.249 0.10 0.459 1.247642 0.1
16.8 32 44.6 72 7 0.542 1.249 -0.10 0.541 1.247642 -0.1
14 31.1 44.6 64.3 8 0.625 1.401 -0.32 0.622 1.395743 -0.31
10 30 39.7 62 9 0.708 1.570 -0.55 0.704 1.560559 -0.54
9.7 20.3 31.1 61.1 10 0.792 1.771 -0.81 0.786 1.755246 -0.79
7.7 20 27.3 56.2 11 0.875 2.039 -1.15 0.867 2.009984 -1.11
6 12.2 25.9 35.9 12 0.958 2.521 -1.73 0.949 2.439479 -1.64

70

60 10 Minute

50
30 Minute
Rainfall (mm)

40

30

20

10

0
-2.00 -1.50 -1.00 -0.50 0.00 0.50 1.00 1.50 2.00
Standard norm al variable,Z
105
Figure 4.3 Fitting the Observed rainfall data of 10 minute and 30 minute duration
using Log Pearson Type III distribution.

150
140
130 1 Hour
120
110
100 24 Hour
90
Rainfall (mm)

80
70
60
50
40
30
20
10
0
-2.00 -1.50 -1.00 -0.50 0.00 0.50 1.00 1.50 2.00
Standard norm al variable,Z

Figure 4.4 Fitting the Observed values of 1 hour, and 24 hour data using Log Pearson
Type III distribution.

70

60 10 Minute

50
30 Minute
Rainfall (mm)

40

30

20

10

0
-2 -1.5 -1 -0.5 0 0.5 1 1.5 2
Standard norm al variable,Z

Figure 4.5 Fitting the Observed values of 10min, and 30 min data using lognormal
distribution.

106
 160
150
140
130 1 Hour
120
110 24 Hour
Rainfall (mm) 100
90
80
70
60
50
40
30
20
10
0
-2 -1.5 -1 -0.5 0 0.5 1 1.5 2
Standard norm al variable,Z

Figure 4.6 Fitting the Observed values of 1 hour, 24 hour data using lognormal
distribution.

The probability plotting of the rainfall data for different duration is shown from
Figures 4.1 to 4.6 for different frequency distributions to select which frequency
distribution fits the data in a better way. Thus the following conclusion was made
by observing the plotting position. The annual maximum 10 minute rainfall data
was best fitted by EVI Type I distribution. The annual maximum of 30 minute, 1
hour and 24 hour rainfall data were best fitted by Log Pearson Type III
distribution.
The probability plotting of the rainfall data for the Log Pearson and Log normal
distribution is almost identical, but since the annual maximum rainfall data is
skewed either positively or negatively, the Log Pearson Type III distribution was
given a priority in selecting the distribution type.

4.5 Testing the Goodness of Fit of Data to the Probability Distribution

The goodness of fit of a probability distribution was tested by comparing the
theoretical and sample values of the relative frequency or the cumulative
frequency function by developing a null hypothesis at some level of significance
and degree of freedom. In the case of the relative frequency function the Chi-
Square test (Χ2) is used at 0.05 significance level.

The Χ2 test statistic is given by

n[ f s ( xi ) − p ( xi )]
m 2

X =∑
2

i =1 P ( xi ) [4.17]

107
Where m is the number of intervals. f s ( xi ) is the sample value of the relative
frequency and np ( xi ) is the expected number of occurrences in interval i.

The Chi-Square test for the rainfall durations is computed as follows:

Table 4.8 Mean and Standard deviations of rainfall data

Rainfall Durations
10 Minute 30 Minute 1 Hour 24 Hour
Mean(Xav) 14.93 32.91 44.12 78.33
Standard Dev(S) 5.21 11.92 15.24 27.79

For Chi-Square distribution the interval I is estimated using

1 / 4( S ) ≤ I ≤ 1 / 2( S ) suggested by Steel and Torrie (Chow 1988). Thus the interval
selected for 10 minute, 30 minutes, 1 hour and 24 hour respectively will be 2mm,
5mm, 6mm, and 10mm.

4.6. Computation of the Probability distribution Function

The value of F(xi) is computed for the respective Xi values from;

⎛ ⎛ ⎛ x − µ ⎞⎞⎞
F (Y ) = exp⎜⎜ − exp⎜⎜ − ⎜ ⎟ ⎟⎟ ⎟⎟
⎝ ⎝ ⎝ α ⎠⎠⎠

6 *S −
α= = 0.7797 * S µ = X − 0.5772α [4.18]
π
Where α and µ are parameters and estimated from the relations:
Hence for 10 minute rainfall duration the parameters α and µ were computed
from
α = 0.7797 *10.41 = 4.06
_
µ = X − 0.5772 * α = 14.93 − 0.5772 * 4.06 = 12.58
⎛x−µ⎞
The reduced Variate Y is calculated from: Y = ⎜ ⎟ for the upper limit of the
⎝ α ⎠
interval.
Thus after substituting µ and α the equation for Y becomes:
⎛ x − 12.58 ⎞
Y =⎜ ⎟
⎝ 4.06 ⎠

108
Table 4.9 Fitting an EVI distribution for 10 minute Annual maximum rainfall

2
Class No. Data ni fs(xi) Y F(xi) Ps(xi) Xi
Range
1 2 3 4 5 6 7
1 <=6 1 0.083 -1.621 0.006 0.006 11.164
2 6.2-7.1 0 0.000 -1.350 0.021 0.015 0.177
3 7.2-8.1 1 0.083 -1.103 0.049 0.028 1.319
4 8.2-9.1 0 0.000 -0.857 0.095 0.046 0.548
5 9.2-10.1 2 0.167 -0.611 0.159 0.064 1.994
6 10.2-11.1 0 0.000 -0.365 0.237 0.078 0.942
7 11.2-12.1 0 0.000 -0.118 0.324 0.088 1.050
8 12.2-13.1 0 0.000 0.128 0.415 0.090 1.084584
9 13.2-14.1 1 0.083 0.374 0.503 0.088 0.002789
10 14.2-15.1 0 0.000 0.621 0.584 0.081 0.977271
11 15.2-16.1 0 0.000 0.867 0.657 0.073 0.87292
12 16.2-17.1 1 0.083 1.113 0.720 0.063 0.077721
13 17.2-18.1 1 0.083 1.360 0.774 0.054 0.199119
14 18.2-19.1 3 0.250 1.606 0.818 0.045 11.35129
15 >=19.2 2 0.167 -1.621 0.006 -0.812
SUM 12 1.000 31.760

The value of Xc2 for 10 minute rainfall duration as computed in Table 4.9 is 31.76.
On the other hand the value of X2v, 1-α for the cumulative probability of 1- α=0.95,
or for the 95% confidence interval and the degree of freedom
v = m − p − 1 = 15 − 2 − 1 = 12 is X24, 0.95=21.02

Thus the conclusion to be drawn here is that, the null hypothesis is not satisfied
(Xc2> X2v, 1-α) for 10 minute rainfall durations and hence the EVI distribution is not
acceptable to derive the extreme values at different return periods. With the
same procedure applicability of other distribution techniques can be tested.

4.6.1 Computation of Extreme Value (XT) For the Various Rainfall Durations and
Return Periods.
Once the type of distribution of the observed data is known, the rainfall events
(XT) exceeding the observed values would be estimated numerically. In this work
Log Pearson Type III distribution method was applied.

109
Table 4.10 Mean and Standard deviations of rainfall durations

Rain fall Mean Standard α µ

duration Xav deviation(S)
10min 14.93 5.21 4.06 12.59
20min 19.83 10.41 8.12 15.15
30min 32.91 11.92 9.29 27.55
1hour 44.12 15.24 11.88 37.26
24hour 78.33 27.79 21.67 65.82

Table 4.11 Computation of the 30 Minute Rainfall Intensity Using Lognormal and Log
Pearson Type III Distribution
YEAR ANNUAL MAXIMUM RAINFALL AND LOG VALUES
10 Minute Y=Log(Xi) 30 Minute Y=Log(Xi) 1 Hour Y=Log(Xi) 24 Hour Y=Log(Xi)
1971 6 0.778 20.3 1.307 27.3 1.436 72 1.857332
1972 9.7 0.987 31.1 1.493 31.1 1.493 62 1.792392
1973 10 1.000 20 1.301 49.8 1.697 116.3 2.06558
1974 16.8 1.225 30 1.477 44.9 1.652 77 1.886491
1975 18 1.255 44 1.643 44.7 1.650 64.3 1.808211
1976 14 1.146 32 1.505 45.1 1.654 73.4 1.865696
1977 20 1.301 36.5 1.562 46.2 1.665 133.2 2.124504
1978 20 1.301 40 1.602 44.6 1.649 61.1 1.786041
1979 19 1.279 33.5 1.525 39.7 1.599 56.2 1.749736
1980 19 1.279 38.3 1.583 44.6 1.649 79.9 1.902547
1981 19 1.279 57 1.756 85.5 1.932 108.7 2.03623
1982 7.7 0.886 12.2 1.086 25.9 35.9 1.555094
Mean 1.143 1.487 1.643 1.869
Std Deviation 0.183 0.179 0.124 0.155
Coefficient of -0.976 -0.947 0.751 -0.148
Skewness

For 10 Minute rainfall duration

Freqency LOGNORMAL LOGPEARSON TYPE III
YT= XT= Rainfall Rainfall
YT YT
T KT Yav+KT*Sy 10 Intensity(mm/hr) KT YT=Yav+KT* XT=10 Intensity(mm/hr)
2 0 1.143 13.900 83.397 0.160 1.172 14.870 89.220
5 0.842 1.297 19.819 118.915 0.853 1.299 19.911 119.468
10 1.282 1.378 23.856 143.139 1.133 1.350 22.405 134.428
25 1.751 1.463 29.069 174.415 1.375 1.395 24.810 148.859
50 2.054 1.519 33.028 198.168 1.510 1.419 26.262 157.573

For 30 Minute rainfall duration

110
Freqency LOGNORMAL LOGPEARSON TYPE III
YT= XT= Rainfall Rainfall
YT YT
T KT Yav+KT*Sy 10 Intensity(mm/hr) KT YT=Yav+KT*Sy XT=10 Intensity(mm/hr)
2 0 1.487 30.690 61.380 0.156 1.515 32.728 65.457
5 0.842 1.638 43.423 86.846 0.853 1.640 43.620 87.240
10 1.282 1.716 52.057 104.114 1.138 1.691 49.047 98.094
25 1.751 1.800 63.158 126.316 1.387 1.735 54.348 108.696
50 2.054 1.855 71.559 143.119 1.520 1.759 57.422 114.844

For 1 Hour rainfall duration

Freqency LOGNORMAL LOGPEARSON TYPE III
YT= XT= Rainfall YT= Rainfall
YT YT
T KT Yav+KT*Sy 10 Intensity(mm/hr) KT Yav+KT*Sy XT=10 Intensity(mm/hr)
2 0 1.643 43.954 43.954 -0.124 1.628 42.425 42.425
5 0.842 1.747 55.900 55.900 0.785 1.740 54.997 54.997
10 1.282 1.802 63.382 63.382 1.335 1.809 64.349 64.349
25 1.751 1.860 72.464 72.464 1.980 1.889 77.361 77.361
50 2.054 1.898 79.013 79.013 2.430 1.944 87.967 87.967

For 24 Hour rainfall duration

Freqency LOGNORMAL LOGPEARSON TYPE III
YT= XT= Rainfall YT= Rainfall
YT YT
T KT Yav+KT*Sy 10 Intensity(mm/hr) KT Yav+KT*SyXT=10 Intensity(mm/hr)
2 0 1.869 73.961 3.082 0.083 1.882 76.184 3.174
5 0.842 2.000 99.887 4.162 0.856 2.002 100.388 4.183
10 1.282 2.068 116.872 4.870 1.216 2.057 114.151 4.756
25 1.751 2.140 138.167 5.757 1.567 2.112 129.385 5.391
50 2.054 2.187 153.947 6.414 1.777 2.144 139.455 5.811

Thus taking the rainfall intensities as computed from the Log Pearson Type III
distribution the plot of the Intensity duration curve using the return period 2, 5, 10,
25 and 25 is as follows:

111
Table 4.12 Rainfall Intensities as Computed using Log Pearson Type III.
Duration Rainfall Intensity (mm/hr) for the indicated frequency T (yrs)
(minute) 2 5 10 25 50
10 89.220 119.468 134.428 148.859 157.573
30 65.457 87.240 98.094 108.696 114.844
60 42.425 54.997 64.349 77.361 87.976
1440 3.174 4.183 4.756 5.391 5.811
Rainfall Intensity (mm/hr)

100.000
T=2 YEAR
T=5 YEAR
T=10 YEAR
T=25 YEAR
T=50 YEAR

1.000
1 10 100 1000 10000
Duration in Minutes

Figure 4.7 IDF Curves Plotted On a Log-Log Graph.

4.7 Determination of Flood Magnitude

Before determining flood magnitudes, the following factors shall be evaluated

and included according to the Ethiopian Road Authority (ERA) standard
manual:

• Drainage basin characteristics including: size, shape, slope, land use,

geology, soil type, surface infiltration, and storage;
• Stream channel characteristics including geometry and configuration,
natural and artificial controls, channel modification, aggradation -
degradation, and debris;
• Flood plain characteristics; and
• Meteorological characteristics such as precipitation amounts and type
(rain, hail, or combinations), storm cell size and distribution characteristics,
storm direction, and time rate of precipitation (hyetograph).

112
 Once the determination of design storm for the required frequency of
occurrence is over, the flood magnitude is determined using standard
hydrologic procedures. There are two kinds of general classification of
hydrologic methods applied in estimating flood magnitudes.

I. Peak discharge method: This method gives only peak discharges at a location,
which includes rational method, SCS-graphical peak discharge method (TR-55)
etc. They are mostly used for conveyance and cross drainage structural
designs.

II. Hydrograph method: This method has a capacity of generating the

hydrograph points for a location. This includes Unit hydrograph method, SCS-
Tabular hydrograph method (TR-20). The method besides its significance in the
design of conveyance structures plays a major role in the design of storage
and energy dissipater. The unit hydrograph method can conveniently be
applied to areas greater than 25km2 and less than 500km2.

The SCS Runoff Curve Number (CN) method is represented by Equation 4.19
(P − I a ) 2
Q=
(P − I a ) + S [4.19]
Where: Q = runoff (mm)
P = rainfall (mm)
S = potential maximum retention after runoff begins (mm) and Ia is initial
abstraction (mm). Initial abstraction (Ia) is all losses before runoff begins. It
includes water retained in surface depressions, water intercepted by vegetation,
evaporation, and infiltration. Ia is highly variable but generally is correlated with
soil and cover parameters. Through studies of many small agricultural
watersheds, Ia was found to be approximated to be 0. 2S.

Thus the runoff equation 4.19 becomes:

Q=
( P − 0.2 S )
2

P + 0 .8 S [4.20]
S is related to the soil and cover conditions of the watershed through the CN. CN
has a range of 0 to 100, and S is related to CN by:
⎛ 1000 ⎞
S = 25.4⎜ − 10 ⎟
⎝ CN ⎠ [4.21]
Where S is in mm, the curve numbers (CN) for normal antecedent moisture
conditions (AMCII). For dry condition (AMC I) or wet condition (AMCIII),
equivalent curve numbers can be computed by
⎛ 4.2CN ( II ) ⎞
CN ( I ) = ⎜⎜ ⎟⎟
⎝ 10 − 0.058CN ( II ) ⎠ [4.22]

113
 ⎛ 23CN ( II ) ⎞
CN ( III ) = ⎜⎜ ⎟⎟
⎝ 10 + 0.13CN ( II ) ⎠ [4.23]

Practical Example: Peak Flow Determination using SCS Methods

From the statistical result presented in Table 4.12, for a 60 minutes time of
concentration and 10 years return period compute peak flow for the design of
small dam in the Tana basin system.

Step 1: Determination of design storm

The SCS method is based on a 24-hour storm event; hence the design storm is
4.8mm/hr*24hr=114.14mm. For comparison purpose the ERA standard manual
was checked and for 10 year return period the corresponding figure for Bahir Dar
town is 131mm, which is over predicted. Using Type II storm design curve which is
applicable for Ethiopian catchments (according to ERA’s recommendation), the
daily rainfall is disintegrated over 24hrs time.

Step 2: Determination of Time of Concentration

Water moves through a catchment area as sheet flow, shallow concentrated
flow, open channel flow, or some combination of these. The time of
concentration is the sum of Tt values for the various consecutive flow segments:

Tc = Tt1 + Tt 2 + ....Ttm (4.24)

Where:
Tc = time of concentration, hr
m = number of flow segments.

4.7.1 Sheet Flow

Sheet flow is flow over plane surfaces. It usually occurs in the headwater of
streams. With sheet flow, the friction value (Manning's n) is an effective
roughness coefficient that includes the effect of raindrop impact; drag over the
plane surface; obstacles such as litter, crop ridges, and rocks; and erosion and
transportation of sediment. These n values are for very shallow flow depths of
about 0.03 m or so.

ERA standard manual suggests for sheet flow of less than 100 meters to use
Manning's kinematic solution (Overton and Meadows 1976) to compute Tt:

[
Tt = 0.091(nL ) / (P 2 ) S 0.4
0 .8 0 .5
] [4.25]
Where:
Tt = travel time, hr
n = Manning's roughness coefficient (obtained from standard tables)

114
 L = flow length, m
P2 = 2-year, 24-hour rainfall, mm
s = slope of hydraulic grade line (land slope), m/m

In the above formula P2 can be obtained from Table 4.12 and hence becomes
3.714mm/hr *24hr= 76.2mm. Other parameters like L and S are obtained from
catchment characteristics.

4.7.2 Shallow Concentrated Flow

After a maximum of 100 meters, sheet flow usually becomes shallow
concentrated flow. The average velocity for this flow can be determined from
equations 4.26 and 4.27, in which average velocity is a function of watercourse
slope and type of channel.

Unpaved V = 4.9178( S ) 0.5 [4.26]

Paved V = 6.1961( S ) 0.5 [4.27]

Where:
V = average velocity, m/s
S = slope of hydraulic grade line (watercourse slope), m/m

These two equations are based on the solution of Manning's equation with
different assumptions for n (Manning's roughness coefficient) and r (hydraulic
radius, meters). For unpaved areas, n is 0.05 and r is 0.12; for paved areas, n is
0.025 and r is 0.06. After determining average velocity, equation 4.28 is used to
estimate travel time for the shallow concentrated flow segment.

L
Tt = [4.28]
(3600V )

Where: Tt is travel time in hr, L s flow length in m, V is average velocity in m/s and
3600 is a conversion factor from seconds to hours.

4.7.3 Open Channels

For open channels average flow velocity is usually determined using Manning's
equation or water surface profile information (4.29)
1
V = r 2 / 3 s1/ 2 [4.29]
n
Where: V is average velocity (m/s), r is hydraulic radius in m (equal to
a/pw), a is cross sectional flow area in m2, Pw is wetted perimeter in m, S is

115
 slope of the hydraulic grade line, m/m, n is Manning’s roughness
coefficient.

After average velocity is computed using equation 4.29, Tt for the channel
segment can be estimated using equation 4.28.

Step 3: Physiographic Information

Watershed area is given as 100ha. From physiographic data Soils are Eutric and
Humic Cambisols with a silt loam of Hydrologic Soil Group B, and sandy clay
loam of Hydrologic Soil Group C respectively. Approximately 30% (30 ha) is silt
loam and 70% (70ha) is sandy clay loam. The cover type is pasture in poor
hydrologic condition. The Curve Runoff Number from standard hydrologic tables
is as follows:
Pasture, poor condition, Soil Group B CN = 79
Pasture, poor condition, Soil Group C CN = 86. The weighted curve number is:

CN = (0.30 x 79) + (0.70 x 86) = 84

Once the curve number is known and design storm ( P10 ) is determined using
statistical methods, direct runoff in mm ( Q10 ) is determined using Figure 4.8 from
the SCS relation between Direct Runoff, Curve Number and Precipitation chart.

116
Figure 4.8 SCS Relation between Direct Runoff, Curve Number and Precipitation

Step 4: Computation of Unit Peak Discharge ( qu )

After computing time of concentration unit peak discharge is computed. Initial
abstraction is computed using equation 4.21 once the curve number is
specified. Alternatively use Table 4.13 to compute initial abstraction from a wide
range of
curve number values. Using information of initial abstraction, design storm and
time of concentration Figure 4.17 is used to determine unit peak discharge in
m3/s/100ha/mm.

117
Table 4.13 Ia Values for Runoff Curve Numbers

Curve Number Ia (mm) Curve Number Ia (mm) Curve Number Ia (mm)

40 76.2 60 33.9 80 12.7
41 73.1 61 32.5 81 11.9
42 70.2 62 31.1 82 11.2
43 67.3 63 29.8 83 10.4
44 64.6 64 28.6 84 9.7
45 62.1 65 27.4 85 9.0
46 59.6 66 26.2 86 8.3
47 57.3 67 25.0 87 7.6
48 55.0 68 23.9 88 6.9
49 52.9 69 22.8 89 6.3
50 50.8 70 21.8 90 5.6
51 48.8 71 20.6 91 5.0
52 46.9 72 19.8 92 4.4
53 45.1 73 18.8 93 3.8
54 43.3 74 17.9 94 3.3
55 41.6 75 16.9 95 2.7
56 39.9 76 16.1 96 2.1
57 38.3 77 15.2 97 1.6
58 36.8 78 14.3 98 1.0
59 35.3 79 13.5 99 0.4

118
Figure 4.9 Unit Peak Discharge, applicable to Ethiopian catchments (Type II
rainfall). Source (ERA Manual)

Step 5: Computation of Peak Discharge (Q )

Using watershed area (in ha) as an input and information of direct runoff (in mm)
and peak runoff rate (m3/s/100ha/mm), use equation 4.30 to determine peak
discharge in m3/s.

Q = qu * A * Q10 [4.30]

In many hydrologic models like SWAT and HEC-HMS, the above procedures are
incorporated as separate algorithms and can be run as appropriate. This is to
facilitate the tedious hydrologic calculations as presented above.

119
 CHAPTER FIVE

Synthetic Flow Generation and Reservoir Simulation

5.0 Introduction
Some of the rivers and/or stations in the any one study area do no not have flow
and sediment data at all or they have very limited number of actually measured
data. In such cases, there should be some means of estimating the flow and
sediment load at respective locations. In the absence of any developed
regional method, for example, a general relation is derived by correlating the
available sediment data of stations or rivers with the corresponding catchment
area in and around the project area.
For instance, to estmate the dam site sediment load at Bilate River, sediment
data and catchment areas of stations Bilate at Alaba Kulito, Gelana at Tore,
Gidabo at Aposto, Kola at Aleta Wondo, and that of Bedessa near Dilla the
following exponential relation was derived.

Q s = 7298 e 0.0025 A [5.1]

Where Qs is suspended sediment load in tons per year and A is the catchment
area in square kilometers.. This equation, in the absence of any measured
sediment data, may be used to estimate the suspended sediment load at a
given station in the region for which it is derived. In the presence of measured
data in the Amhara region it is wise to evaluate the applicability of equation 5.1.

5.1 Synthetic Flow Data Generation using Thomas Fiering Method

Long term data sequence of flow is useful for various hydrological and water
resources planning purposes. Long term inflow to the reservoir can be used
mainly for two purposes:

(1) When the available data is less than the life span of the reservoir (which is
50 years in most medium dams), characteristics of water demand and
supply can be simulated by extending inflow data up to and even
beyond the life of the reservoir,
(2) To account year to year and season to season carry over storage and size
the reservoir accordingly. Without carry over storage analysis, the size of
the reservoir may be unnecessarily bigger.

The stochastic Thomas-Fiering monthly flow generation model can be given as

follows. The main advantage of the model is to capture the seasonal variability
of flows. The model can either be computed using Excel or can be programmed
using any programming language. In this training, it will be demonstrated using
Fortran codes:

120
 − −
Q p , j +1 = Q j +1 + B j (Q p , j − Q j ) + t j s j +1 (1 − r j2 ) 0.5 [5.1]
Where:
Qp,j+1 = generated monthly flow of year p and month j+1
j = the months; j =1 corresponds to January-1, and j =36 December-3
−
Q j = mean monthly flow of the month j over N years for which the record is
available
sj+1 = standard deviation of the monthly flow of the month j+1 over N years for
which the record is available
s j +1
B j = rj ( ) [5.3]
sj
t j = normal random independent aviate with zero mean and unit variance
generated using the Box and Muller technique.
r j = cross correlation between Qj and Q j+1 monthly flows over N years.

Procedure of generation
1. Provide the short record length monthly flows at the desired location
−
2. Compute the seasonal values of the mean ( Q j ), the standard deviation
( s j ) and the cross correlation coefficient of rank 1 ( r j ) between
consecutive months. The following Table provides an example of the
calculation of the 36 parameters

Table 5.1 Estimated Monthly Parameter Values of the Thomas-Fiering Model

Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov Dec
Qj, Mean(MCM) 5.66 3.18 2.96 8.18 26.40 19.77 14.70 15.86 15.81 21.17 18.04 7.82
Sj, Std. Dev. 5.89 3.84 2.68 6.41 11.72 11.03 11.76 15.73 14.11 12.61 15.01 5.70
(MCM)
(rj), Cross 0.54 0.65 0.76 0.45 0.25 0.32 0.41 0.92 0.75 0.42 0.29 0.68
Correlations
(Bj), Slopes 0.43 0.34 0.79 1.18 0.46 0.28 0.49 1.21 0.67 0.39 0.33 0.25

4. Generate tj = normal random independent variate with zero mean and

unit variance generated using the Box and Muller technique.
5. Finally generate the Q j+1 series and adjust for negative flows.

121
5.2 Reservoir Sediment Volume Estimation
In order to estimate sediment distribution the mean density of deposited
sediment and reservoir trap efficiency has to be first determined. The density of
deposited material is used to convert the total sediment inflow to a reservoir from
a mass to a volume.
Basic factors influencing the density of sediment deposited in a reservoir are (1)
the manner of reservoir operation, (2) the texture and size of deposited sediment
particles, and (3) the compaction and consolidation rate.
Density of initial sediment deposits is estimated from the relation below (Yang,
1996).
W 0 = W c Pc + W m Pm + W s Ps [5.3]
Where Wo is the initial density of deposit; Wc, Wm, and Ws are initial unit weights of
clay, silt and sand, respectively; and Pc, Pm, and Ps are percentages of clay, silt,
and sand, respectively of the incoming sediment.

The initial unit weights of clay, silt and sand can be taken to be 561 kg/m3, 1140
kg/m3, and 1550 kg/m3, respectively assuming Type 2 reservoir operation
(Normally moderate to considerable reservoir drawdown) (Yang, 1996). The
percentages of clay silt and sand in the incoming sediment can be observed for
the respective rivers. If we assume the values to be considered as Pc = 20%, Pm =
60% and Ps = 20%, the initial mean density is estimated to be 1106 kg/m3. The
average unit weight is then determined for a 50-years of consolidation using the
following relation.
⎛ T ⎞
WT = W 0 + 0.4343 K ⎜ ln T − 1 ⎟ [5.4]
⎝T −1 ⎠
Where WT= the average density after T years of reservoir operation;
W0 = Initial unit weight (density)
K = constant based on the type of reservoir operation and sediment size
analysis.
If the K value is to be 44.4 based on the selected type of reservoir operation and
sediment size, after 50 years of reservoir operation (T = 50), the average unit is
computed as 1164 kg/m3.
In order to estimate the amount of sediment retained, out of that comes from
upstream (inflow sediment), the trap efficiency is determined from Brune’s curve
(Brune method).
Making use of the above information (average density and trap efficiency), the
total volume of sediment to be deposited in each reservoir in a period of Y years
is estimated. Accordingly, the total sediment deposit of a reservoir site can be
calculated for Y year and distributed according to area reduction.

5.3 Distribution of Sediment in the Reservoirs

When the sediment-laden water is impounded the sediment is distributed
throughout the slope of the reservoir. At the beginning of the impounding the

122
coarser portion of the sediment load is deposited near the head of the reservoir
and the fine particles are transported by density currents towards the dam. As
deposition continues the resulting delta front moves progressively down into the
reservoir. Very little of the sediment really passes through the outlets.
There are several ways of estimating sediment distribution in a particular
reservoir, this training will demonstrate, using the empirical Area-reduction
method developed by Borland and Miller (1960) as a result of detail analysis of
the deposition patterns of several reservoirs in USA.

5.3.1 Area-Reduction Method of Borland and Miller

The method recognizes that sediment distribution in a reservoir depends on (1)
the manner of reservoir operation, (2) the texture and size of deposited sediment
particles, (3) the shape of the reservoir, and (4) the volume of sediment
deposited in the reservoir. In the method reservoirs are classified in four broad
types depending on the approximate slope of the line of reservoir capacity vs.
depth plotted on a log-log scale as shown in Table 5.2. The shape factor, m, of a
reservoir is defined as the reciprocal of the slope of the depth vs. capacity plot.

Table 5.2 Classification of Reservoirs (Ghosh, 1999)

Reservoir Type Classification M

I Lake 3.5 – 4.5
II Flood plain, foothill 2.5 – 3.5
III Hill 1.5 – 2.5
IV Gorge 1.0 – 1.5

Example of reservoir classification exercise:

Based on the available data, capacity and elevation, at the dam site the shape
of the Dendo reservoir on Bilate River is defined by the depth-capacity
relationship as shown in Figure 5.1. According to this method, the shape factor is
estimated to be 2.30 (m = 1/0.435 = 2.30), which makes the reservoir to be
classified Hill Type (Type III) for Dendo Reservoir.

123
 1.80

1.60
y = 0.435x - 2.1067
1.40

1.20
Log (depth),

1.00

0.80

0.60

0.40

0.20

0.00
4.00 4.50 5.00 5.50 6.00 6.50 7.00 7.50 8.00 8.50 9.00

Log (capacity), MCM

Figure 5.1 Depth-capacity curve of Dendo Reservoir for the estimation of shape
factor, m (Borland and Miller method).

5.3.2 Computing the non zero-capacity elevation at the dam/reservoir

Using the original elevation area and capacity curves to compute the value of
the dimensionless function F at several different pool elevations in the deeper
part of the reservoir:
S − Vh
F= [5.5]
HAh
Where:
S = total sediment deposition (in MCM)
Vh = reservoir capacity (m3) at each elevation, h
H = original depth of reservoir below normal pool
Ah = reservoir area (m2) at a given elevation h
Also compute the decimal values for the relative depth p as
h − hmin
p=
H
Where hmin = original bottom elevation
Plot the resulting F and p values on the type curve for determining the new zero
depth. The intersection of the plotted F values with the type curve selected for
the reservoir defines the po value for the new zero-capacity elevation at the
dam.

124
5.3.3. Distribution of Sediment in the Reservoir
The specific volume of the trapped sediment is distributed within the reservoir
according to the selected curve. Compute the values for relative sediment area
a for each relative depth p using the appropriate equation.
TypeI : a = 5.047 p 1.85 (1 − p ) 0.36
Type II : a = 2.487 p 0.57 (1 − p ) 0.41
Type III : a = 16.967 p 1.15 (1 − p) 2.32
Type IV : a = 1.486 p −0.25 (1 − p)1.34 [5.6]
Compute the relative sediment area a at the new zero elevation and proceed
computing the area correction factor as Ao/a.
Compute the area at each pool elevation occupied by sediment by multiplying
the area correction factor by the relative sediment area at each level above
the new zero-capacity elevation. Note that in the fully sedimented part of the
reservoir extending from the new zero-capacity elevation down to the original
bottom, the sediment pool area equals the original pool.
Compute the sediment volume for each stage increment above the new zero-
capacity elevation using the end area method which is given in equation 5.7
( A + A2 )
S hnew = 1 * (h2 − h1 ) [5.7]
2
From the zero-capacity elevation to the reservoir bottom, the sediment volume
equals the original reservoir capacity, since this zone is entirely sedimented. The
cumulative volume of deposited sediment is computed by summing the
sediment volume at each pool level. The total sediment volume should match
the pre-determined sediment volume during the sediment load computation
within about 1% marginal error. Finally compute the revised area and capacity
by subtracting the sediment area and capacity volumes (see example Table
5.3).

5.3.4 Reservoir Topography

The reservoir volume and elevation relationships are used as a basis for reservoir
water balance analysis. To avoid deviation of reservoir capacity with time due to
sediment accumulation in the reservoirs, new topographic surveys should be
carried out for the selected reservoirs and accordingly the newly developed
reservoir water level and storage relationship are used for further analysis. As an
example the topographic map and area capacity of Laelay Wukro dam is
presented in Figures 5.2 and 5.3 respectively.

The reservoir capacity is calculated using the formula:

h
V= ⋅ ( Ares1 + Ares 2 + ( Ares1 × Ares 2 ) 0.5 ) (2.1)
3
Where

125
V 1,2 reservoir capacity (volume) between two successive elevations (m3)
h elevation difference between successive contours (m)
Ares1 and Ares 2 area of reservoir water spread at elevation h1 and h2 (m2)

Figure 5.2 Topographic map of Laelay Wukro reservoir (After Mohammod, 2009)

126
 Capacity
Elevation
Curve

Area
Elevation
Curve

Figure 5.3 Area capacity curve for Laelay Wukro reservoir

127
Table 5.3 Sediment Distribution for 50 years of operation with 95% trap efficiency
(example at Dendo Reservoir on Bilate River)

Water Sediment Sediment Accumulated

Surface Original Relative area volume sediment
Elev. Original area capacity Depth depth, p Ap K Ap 0.5(A1+A2)*H/100 volume
No. (m) (ha) (MCM) (m) (5)/42 Type III (ha) (MCM) (MCM)
(1) (2) (3) (4) (5) (6) (7) (8) (9) (10)
1 1600 1931.81 324.662 42.0 1.00 0.000 0.00 0.015 171.947
2 1599 1864.79 305.679 41.0 0.98 0.003 2.99 0.087 171.932
3 1598 1802.21 287.344 40.0 0.95 0.014 14.33 0.249 171.846
4 1597 1633.03 270.168 39.0 0.93 0.034 35.41 0.511 171.597
5 1596 1528.89 254.359 38.0 0.90 0.064 66.69 0.874 171.086
6 1595 1456.31 239.433 37.0 0.88 0.103 108.20 1.339 170.212
7 1594 1366.22 226.320 36.0 0.86 0.153 159.69 1.902 168.872
8 1593 1317.53 211.901 35.0 0.83 0.211 220.69 2.557 166.971
9 1592 1257.51 199.026 34.0 0.81 0.278 290.62 3.297 164.414
10 1591 1204.94 186.714 33.0 0.79 0.352 368.73 4.115 161.117
11 1590 1161.71 174.880 32.0 0.76 0.434 454.20 5.002 157.003
12 1589 1121.26 163.466 31.0 0.74 0.522 546.12 5.948 152.001
13 1588 1075.47 152.482 30.0 0.71 0.615 643.48 6.944 146.053
14 1587 1028.91 141.960 29.0 0.69 0.712 745.24 7.978 139.109
15 1586 993.49 131.848 28.0 0.67 0.813 850.27 9.038 131.132
16 1585 957.41 122.094 27.0 0.64 0.915 957.41 9.398 122.094
17 1584 922.15 112.696 26.0 0.62 1.018 922.15 9.038 112.696
18 1583 885.51 103.657 25.0 0.60 1.121 885.51 8.660 103.657
19 1582 846.45 94.998 24.0 0.57 1.222 846.45 8.285 94.998
20 1581 810.53 86.713 23.0 0.55 1.321 810.53 7.918 86.713
21 1580 773.08 78.795 22.0 0.52 1.415 773.08 7.529 78.795
22 1579 732.72 71.266 21.0 0.50 1.505 732.72 7.116 71.266
23 1578 690.39 64.150 20.0 0.48 1.587 690.39 6.721 64.150
24 1577 653.79 57.429 19.0 0.45 1.661 653.79 6.348 57.429
25 1576 615.90 51.081 18.0 0.43 1.726 615.90 5.877 51.081
26 1575 559.59 45.203 17.0 0.40 1.781 559.59 5.300 45.203
27 1574 500.34 39.904 16.0 0.38 1.823 500.34 4.846 39.904
28 1573 468.82 35.058 15.0 0.36 1.852 468.82 4.481 35.058
29 1572 427.30 30.577 14.0 0.33 1.867 427.30 4.119 30.577
30 1571 396.56 26.458 13.0 0.31 1.865 396.56 3.714 26.458
31 1570 346.21 22.744 12.0 0.29 1.846 346.21 3.418 22.744
32 1569 337.38 19.456 11.0 0.26 1.809 337.38 3.244 19.326
33 1568 311.44 16.675 10.0 0.24 1.753 311.44 2.960 16.082
34 1567 280.62 14.454 9.0 0.21 1.675 280.62 2.626 13.122
35 1566 244.66 12.707 8.0 0.19 1.576 244.66 2.282 10.495
36 1565 211.79 10.899 7.0 0.17 1.455 211.79 2.057 8.213
37 1564 199.61 8.896 6.0 0.14 1.310 199.61 1.942 6.156
38 1563 188.83 7.143 5.0 0.12 1.142 188.83 1.752 4.214
39 1562 161.64 4.648 4.0 0.10 0.950 161.64 1.557 2.461
40 1561 149.78 1.558 3.0 0.07 0.735 149.78 0.800 0.904
41 1560 10.18 0.104 2.0 0.05 0.499 10.18 0.073 0.104
42 1559 4.46 0.031 1.0 0.02 0.246 4.46 0.031 0.031
43 1558 1.75 0.000 0.0 0.00 0.000 1.75 0.009 0.000

128
 Chapter Six

Probable Maximum Precipitation (PMP) and Probable Maximum

Flows (PMF)

6.1 Probable Maximum Precipitation

WMO (1986) defines Probable Maximum Precipitation (PMP) as the greatest

depth of precipitation for a given duration that is meteorologically possible over
a given station or a specified area. The term, probable maximum precipitation, is
well established and is widely used to refer to the quantity of precipitation that
approximates the physical upper limit for a given duration over a particular
basin. It is used to predict the Probable Maximum Flood (PMF) values, which
intern shows when an area will be flooded, or to pinpoint when a certain rainfall
rate or a specific volume of flow will recur in the future.

6.2 Methods of Estimating Probable Maximum Precipitation (PMP)

PMP estimation methods fall into the following general categories: the storm
model approach; the maximization and transposition of individual observed
storms; generalized (regionalized) methods; theoretical or empirical methods
derived from maximum depth, duration, and area observations; and statistical
methods.
The storm model approach uses physical parameters, such as surface dew point,
height of storm cell, and inflow and outflow, to represent the precipitation
process (Collier and Hardaker, 1996). Storm transposition involves translating
observed storm characteristics from one or more gauged locations to the
location where the PMP estimation is required (typically an ungauged location).
Storm maximization consists of adjusting observed precipitation amounts upward
to account for maximum atmospheric moisture convergence. Generalized PMP
methods are often developed by maximizing and translating classes of storms
over a broad region; storm classification in turn is based on the storm type,
and/or storm efficiency defined as the ratio of maximum observed rainfall to the
amount of precipitable water in the storm column (Collier and Hardaker 1996).

Although the PMP has a theoretical exceedence probability of zero, meaning

that it is so large that it will never be exceeded, this is not the case in reality.
Consequently, a few studies have sought to assign a risk statement to PMP
estimates. The National Research Council (NRC,1994) estimates the return period
of the PMP in the United States as between 105 and 109 yr. Koutsoyiannis (1999)
developed a rather straightforward method for assigning a return period to PMP
values estimated using the frequency factor method (Hershfield 1961, 1965):
Koutsoyiannis (1999) fit a generalized extreme value (GEV) distribution to the

129
frequency factors computed from the 2645 record used by Hershfield and found
that the largest factor (km = 15) corresponds to a return period of about 60,000
yr, which falls at the low end of the NRC (1994) range. Foufoula-Georgiou (1989)
investigated a storm transposition approach for assessing the frequency of
extreme precipitation depths, but stressed the need for further research before
applying the method to the PMP and PMF.

6.2.1 The Hershfield Method

Hershfield (1961, 1965) based on a general frequency equation given by Chow

(1951) suggested that PMP for a station can be estimated from the following
Equation
X pmp = X n + K mσ n [6.1]
Where X pmp is the PMP for a given station for a specific duration and X n and n σ
are the mean and standard deviation for a series of n annual maximum rainfall
values of a given duration respectively. Km is the frequency factor and is the
largest of all the calculated K values for all stations in a given area. The value of
K is calculated using the following equation.

K =( X − X 1 n −1
) σ n −1
[6.2]

Where X1, X n−1

and σ n −1
are the highest, mean and standard deviation
respectively excluding the X1 value from the series. In a survey of more than 2600
stations world over, Km values as calculated from equation 6.2 vary from less
than 3 to a highest value of 14.5. Hershfield adopted the highest value rounded
to 15 for estimating PMP,

i.e. X pmp
= X n + 15σ n [6.3]

Later Hershfield found that use of Km=15 was not appropriate. He noted that K
varies inversely with mean annual maximum rainfall at any station and presented
a chart for determining Km for 5-min, 1-h, 6-h, and 24-h durations of mean
annual maximum rainfall as shown Figure 6.1 below.

130
Figure 6.1 The Hershfield’s chart for determination of frequency factor Km

The curves of Figure 6.1 are based on observed data from 2650 stations 90
percent of which were in the United States, where observations were at least
daily for a period of at least 10 years. As a matter of fact, there are several
measurements of rainfall in the United States made at locations other than
where there are official gauges that exceed the PMP values calculated from this
statistical procedure (Riedel, 1977). Computation of Km for Canada (McKay,
1965) indicated a maximum value of 30 associated with a mean annual
maximum 24-hours rainfall amount of 15mm.

Recent research outputs of countries like China and Romania also showed that
their frequency factor/Km/ value varies between 6 and 8.5 and rejected the
Hershfields chart as it over estimates the PMP.( Desa et al., 2001) and M.N. Desa
et al.,(2003))

Example of PMP calculation procedure

The PMP value at Merto lemariam station for 24 hours duration is taken to
demonstrate the procedures of PMP estimation.

131
Table 6.1 Procedure for PMP estimation at Mertolemariam station for 24 hrs duration

No. Descriptions Symbol Values Remarks

10.82N,38.23E
Mertolemariam
1 Station Name and 2110m Alt.
2 Sample Size N 18
3 Mean(mm) Xn 38.1
4 Standard Deviation(mm) σn 11.5
5 Mean after Excluding the
maximum rainfall depth from
the series(mm) Xn-1 36.7
6 Standard deviation after
Excluding the maximum
rainfall depth from the
series(mm) σn-1 10.2
7 The ratio of 5 and 3 Xn-1/Xn 0.96
8 The ratio of 6 and 4 σn-1/σn 0.88
Adjustment of mean for the
9 maximum observed series 1.01
Adjustment of mean for
10 length of record 1.02
After
11 Adjustment of standard
Hershfield,1961,b
deviation for the maximum
observed series 1.03
Adjustment of standard
12 deviation for length of record 1.07
adj.
13 Adjusted mean(mm) mean 39.3
14 Adjusted stdev(mm) [Link] 12.7
15 Frequency factor km 7 See section 6.2.1
1-day probable maximum 1-day
16 precipitation(mm) PMP 128.3
24 hrs probable maximum 1.13*daily After
17 precipitation(mm) PMP 145.0 Hershfield,1961,b

As shown in the above table the 24 hours PMP at Mertolemariam station is

145mm. Different adjustments like adjustment for record length were made for
different stations in the basin for different durations. Include all PMP vales
estimated by the above method and the chart method. show percent of
deviation or discrepancy as the last column.

132
6.3 Transformation of Point PMP to Areal PMP

Design rainfall values like PMP are generally expressed in the form of point rainfall
intensity values which is the rainfall depth at a location. In order to obtain areal
average values for an area, hydrologists and engineers require techniques
whereby point rainfall amounts can be transformed to average rainfall amounts
over a specified area. This problem of point-to-area rainfall conversion can be
addressed using depth–area curves such as that of Figure 6.2 which require the
use of areal reduction factors. The two types of areal reduction factors
commonly in use are Geographically Fixed and Storm Centered relationships
(U.S. Weather Bureau, 1957, 1958a, 1958b; Miller et al., 1973; Srikanthan, 1995).

Figure 6.2 Depth area reduction curve

When the probable maximum precipitation is to be applied to an area larger

than about 25 km2, it should be reduced. No modification is considered
necessary for smaller areas.

6.4 Probable Maximum Flood (PMF)

The Probable Maximum Flood is defined as the flood that may be expected
from the most severe combination of critical meteorological and hydrologic
conditions that are reasonably possible in the drainage basin under study.

133
The PMF is generally viewed as the flood resulting from a PMP, applied to
assumed antecedent basin conditions. Procedures for selecting antecedent
conditions and transforming water input to river flow vary among different
agencies and hydrologists, allowing significant variation in the PMF computed
from given PMP. The PMF is normally quantified in terms of peak discharge and
associated hydrograph. The relative engineering importance of peak discharge
versus volume depends on characteristics of the engineering works in question.

The calculation of the maximum probable flood is based on the integration of

probable maximum precipitation obtained before with the help of a rainfall-
runoff deterministic model in which the basin is taken as a system that has as
input the rain histogram and as output the flood hydrograph. The transformation
of the net rainfall in discharge hydrograph on each basin (tributary) was based
on the unit hydrograph method, which is one of the most important investigation
instruments of the rainfall-runoff process, and for the flood propagation in the
riverbed and its combining with floods formed on the main tributaries.

134
 CHAPTER SEVEN

GIS and Remote Sensing Application in Water Resources Applications

7.0 Introduction to GIS and Remote Sensing
With the availability of digital and remote sensing data, the application of GIS in
the field of hydrology is becoming apparent. The most frequent use of GIS in
hydrology has been in input/output data handling for modelling purposes, as
well as in the derivation of flow direction, flow length and slope maps from
Digital Elevation Models (DEMs) (SKOPE,1998). For this purpose primary data is
collected, organized and synthesized in a way that makes them ready to use for
any water resources practices. Many of the input data preparations using GIS
and remote sensing are referred as watershed processing. The basic software
packages deployed for this purpose are Arcview GIS 3.3, Arc GIS 9.1, ILWIS
3.3 Academic and the Hydrologic Engineering Centre Geospatial Hydrologic
Modelling Extension (HEC-GeoHMS), which is an Arcview extension.

7.1 Catchment Characterization using GIS Technique

GIS is a useful tool in deriving catchment characteristics. The process of deriving
catchment characteristics for water resources practices starts with identifying
the drainage network, delineating the watershed and sub-basin boundaries,
followed by analyzing and preparing the input catchment parameters for
simulation. The total watershed area was initially delineated on topographic
maps and areal photos of 1:50,000 scale. Later the drainage area was verified
on ground and corrected accordingly. The sub-watershed area delineation and
extraction is done considering the drainage system. The HEC-GeoHMS was used
to carry out most of these activities. It processes the terrain and spatial
information to generate a number of hydrologic inputs. A Digital Elevation Model
(DEM) is required as input for watershed processing.

7.1.1 Deriving Catchment Characteristics

Topographic and Physiographic Information

Scanned topographic maps scaled 1: 50,000 can be used as input for
generating the DEM for the catchment. The maps were digitized to prepare
layers of drainage, contour, roads and other important topographic or physical
features. Then ArcView GIS 3.3 for digitizing and ILWIS 3.3 Academic is used to
prepare the digital elevation model.

135
Scanned GIS
Topographic Digitization Digital Processing DEM
map Maps

Figure 7.1 Schematic representation of DEM preparation

There are several different methods used to create DEM series. From this several
techniques the two more known methods are discussed as follows:

1. Using Contour Maps

The use of contour maps for DEM generation is the most common technique. The
process involves the following:
ƒ Existing plates used for printing maps are scanned,
ƒ The resulting raster is vectorized and edited,
9 contours are "tagged" with elevations
ƒ Additional elevation data are created from the hydrography layer,
9 i.e. shorelines provide additional contours
ƒ Finally, an algorithm is used to interpolate elevations at every grid point
from the contour data (Example Surfer Software, or Kriging Algorithm).

2. Using Photogrammetric Technique

The use of photogrammetric method to create DEM can be done manually or
automatically:
ƒ Manually, an operator looks at a pair of stereophotos through a
stereoplotter and must move two dots together until they appear to be
one lying just at the surface of the ground,
ƒ Automatically, an instrument calculates the parallax displacement of a
large number of points

136
Figure 7.2 Digital Elevation Model for Blue Nile Basin Watershed
Watershed slopes
Slope is highly variable with in a basin, and hence no single measure of slope is
commonly agreed upon. Before the advent of DEMs, it was almost practically
impossible to derive slopes for all the landscapes within a basin (Mazvimav,
2006). DEM can be used to estimate slopes for all pixels within a catchment.
From multiple pixels in a DEM and cumulative frequency distribution of slopes,
slope indices S x are derived. S x denotes a slope value for which x % of the pixels
in a basin are equal to or less than this value. In some studies (Entekhabi, 2001;
Mazvimav, 2006) median slope ( S 50 ) is used instead of average slopes for the
watersheds.

137
Figure 7.3 shows the distribution of slopes across the watershed. It can be seen
that the watershed is dominated with flat slopes accounting about 88% less than
5% and Laelay Wukro with steep slopes. Watershed topography can be also
compared taking different indexes like ( S 25 , S 50 and S 75 ). The median slopes are
25.7%, 5.6% and 3.2% for Laelay Wukro, Haiba and GumSelassa watersheds
respectively. The median slopes determined for the total and sub-watersheds
are used to estimate watershed lag.

50
45
Percentage of total

40
35
watershed
area (%)

30
25
20
15
10
5
0
<3% 3- 5% 5-8% 8-16% 16-32% 32-45% >45%

Slope (%)
GumSelassa Haiba Laelay Wukro

Figure 7.3 Distribution of slopes across the watershed (Adopted from Mohammod,
2009)
An example of catchment characteristics that was derived for the Tana basin
system is presented in Table 7.1.

138
 Table 7.1 Catchment characteristics of catchments in the Tana basin system

Catchm Perimet Catchm Longest Draina Circulari Elong Longes Avera Ave % % % %
ent er ent Area Drainag ge ty ation t flow ge rag flat undula rollin hilly
name (m) (km²) e Density index(CI Ratio path eleva e ting g
tion slop
Length (m/km² ) (EL) length(
e
(m) ) m)
GRANO 45515.91 55.65 17188.7 427.73 37.23 2.04 22375. 2233 27.0 6.5 31.15 27 35.3
13 0 3
KOGA 123718.4 327.27 46086.5 306.48 46.77 2.25 51024. 2418 23.7 27.3 53.27 9.77 9.62
2 85 0
GEMER 98975.62 305.18 30430.8 361.06 32.1 1.54 34965. 2324 33.4 11.3 41.37 24.48 22.8
O 39 9 2
MEGEC 129976.5 562.2 47499.0 326.38 30.05 1.77 52421. 2394 37.3 1.8 26.10 33.30 38.7
H 2 91 4 5
GUMERA 230046.3 1355.77 84608.8 303.54 39.03 2.03 89584. 2715 33.7 8.3 26.64 32.73 32.3
2 40 7
GILGEL 224596.7 1667.87 77079.9 334.60 30.24 1.67 81779. 2675 36.9 9.89 45.10 26.20 18.8
ABAY 7 13 8 1
RIBB 250406.8 1489.89 98558.2 324.01 42.09 2.26 101758 2912 44.7 9.15 33.89 29.10 27.8
6 0 .2 2 5

7.2 Watershed Delineation and Mapping: HEC-GeoHMS Watershed Processing

HEC-GeoHMS is used to prepare the input catchment parameters for the

hydrological simulations following step by step procedures incorporated in the
model. The program features are Terrain Preprocessing, Basin Processing,
Hydrologic parameter estimation and HEC-HMS model support. The first two
features are accomplished through a number of procedural steps. Terrain
preprocessing includes filling sinks, assigning flow direction and flow
accumulation, defining stream and sub-watershed area sizes. In the basin
processing parameters like river slope, river length, watershed centroid, and
longest and centroidal flow path can be derived. Figure 7.3 depicts the general
procedure for deriving catchment characteristics using a computer software
and Figure 7.4 indicates the step by step outline procedure adopted for Gilgel
Abay catchment.

139
 Input
DEM
DEM

Filling sinks,
DEM Reconditioning depressions and flat
area

Determines flow direction, flow

Terrain Analysis accumulation and area
contributing to flow

Extracts watershed and

stream network based
Watershed Stream Network on user-defined
Extraction Extraction thresholds

Determines basin input

Basin Processing parameters for
hydrologic modelling
(slopes, river length,
basin centroid, etc…)

Watershed models Watershed model that

can be retrieved in HEC-
HMS

Figure 7.4 : Flow chart to derive catchment characteristics using HEC-GeoHMS

140
Fill Sink Map Flow Direction Map Flow Accumulation Map

Catchment Extraction Drainage Network Extraction Map

Map Drainage Network
Ordering Map

Catchment Merge Map

Figure 7.5 Procedures adopted in deriving catchment characteristics for Gilgel

Abay Catchment

The above procedure helps to identify catchment characteristics like area,

perimeter, average slope in %, length of the longest stream, slope length,
average catchment elevation etc… The practice helps to establish a possible
dam site location in the Gilgel Abay catchment (Figure 7.6).

141
Figure 7.6 GIS Application to identify DAM site and catchment river network for
Gilgel Abay Catchment

7.3 Remote Sensing Applications

Remote sensing applications to hydrology are relatively new but are rapidly
becoming an important information source for practicing hydrologists. Remote
sensing uses measurements of the electromagnetic spectrum to characterize
the landscape, or infer properties of it, or, in some cases, to actually measure
hydrologic state variables. Areal photography in the visible wave lengths is a
remote sensing technique frequently used by hydrologists; however, modern

142
remote sensing is centerd around satellite systems (Engman, 1992). Remote
sensing data are available in three forms: as an image analogues to an aerial
photograph, in an analogue format, and in digital format. The conversion from
imagery or analogue data to digital data, and vice versa, is based on
separating the measured intensity of reflectance into increments, usually based
on byte word lengths, with 0 representing the lowest(darkest) level of
reflectance and 255 the highest (brightest) reflectance. The pixel format of
digital remote sensing data makes it ideal for merging with GIS. Remote sensing
data provide suitable information for hydrological studies involving landuse
change studies, climate modeling, flood zonation studies and the like. This
implies landuse data in most hydrological studies is derived from remotely sensed
data whose spatial resolution has to be small enough to map the spatial
heterogeneity of the surface conditions.
The most common sources of remotely sensed data are the LANDSAT images
available from different missions. And two types of images are available from
LANDSAT namely, MSS (Multi spectral Scanner) and TM (Thematic Mapper). The
resolutions of the images are 80 and 30 m respectively. Using these images it is
possible to determine at 100 to 30 m spatial resolution the type of land cover. This
is done by image classification techniques which interprets the radiometric
measurements into specific land use or land cover types depending on spectral
signature of different surfaces.

7.3.1 Use of IDRISI/ILWIS to study Catchment Characteristics

With the aid of IDRISI 15 software (The Andes Edition) or, ILWIS, which is a public
software the LANDSAT image obtained from remotely sensed data is enhanced
using linear contrast stretching and histogram equalization technique to increase
the volume of image information. All images will be rectified to a common
Universal Transversal Mercator (UTM) coordinates system based on the 1:50,000
scale topographic maps.

In order to increase the visual interpretability, the color composite of bands 2, 3

and 4 can be used and a supervised classification with the maximum likelihood
algorithm can be conducted to classify the Landsat images. The maximum
likelihood classification system is widely applied in remotely sensed image
analysis (Wang et al., 2007; Mutie, 2006; Gadain, 2005; Braimoh, 2004). The
accuracy of the classified image was verified from the ground truthing.

The process of determination of landuse/land cover from satellite images is

based on the remote sensing procedures and processes described on image
processing laboratory by using IDRISI and Arc View GIS (Yeung et al., 2003), as
well as procedures described in various publications (Wang et al., 2007; Mutie,

143
2006; Gadain, 2005; Braimoh, 2004) and the reference manual of IDRISI Andes
(Eastman, 2006). These include Pre- processing, Image classification and Post-
processing. The details are presented in Figure 7.7.

LANDSAT Geo-registration Mosaicking

Image using RESAMPLE

Signature Training sites Compositing

development development

Maximum likelihood Study Area Land cover

classification Extraction Classes

Figure 7.7 Methodology adopted in digital image processing

In the first step, the simple image pre-processing was carried out including image
acquisition and importing into IDRISI-ANDES together with Polygon rasterizing of
the catchments. The imported images were enhanced, and geometrically
corrected by geo-referencing option. Enhancement techniques were applied to
increase visual distinctions between features and increase the amount of
information that can be visually interpreted from the data. Each Landsat image
was enhanced using linear contrast stretching and histogram equalization
techniques to increase the volume of image information and to improve
interpretability of image.

To locate ground features on imagery, or to compare a series of images, a

geometric correction procedure was used to register each pixel to real world
coordinates (Jensen, 1996). Remotely sensed images in raw format had no
coordinates and all images were rectified to a common UTM coordinates system
based on the 1:50,000 scale topographic maps and to ensure the accurate
identification of temporal changes and geometric compatibility with other
sources of information, the images were geo-coded to the co-ordinate and
mapping system of the national topographic maps such that UTM coordinates.
Image rectification was done using linear transformation and nearest-neighbor
re-sampling and the Root Mean Square Error (RMSE) was used to measure the
efficiency of rectification as described in Lu et al, (2003) and Wang et al. (2007).
An acceptable RMSE is one half of the unit of the pixel size (Wang et al 2007,
Eastman, 2006).

144
The detailed image re-sampling and geometric correction processes are as
follows:

a) The corresponding file (the file that contain both coordinates: image’s
coordinates and ground coordinates) was prepared by selecting the
ground control points (points from the topographic map which was also
supplemented by delineated rivers in and around the catchment area)
and record the images coordinates (in terms of rows and column) of the
same point.

b) The maximum and minimum points of the images were identified from the
attributes of the images as shown in Figure 7.8 and converted into UTM
coordinate system as shown in Table 7.2.

Figure 7.8 An example of Image attributes (P166R68)

(Adopted from Birhanu, 2008)

Table 7.2 Maximum and minimum coordinates of the image

Year X (meter) Y (meter)

Year Minimum Maximum Minimum Maximum
1990 667505.957500 883719.937500 8941536.000000 9136995.000000

c) The number of columns and rows were calculated according to equation

7.1 and 7.2 respectively.

145
 MaximumX − MinimumX
Column = [7.1]
Re solution
MaximumY − MinimumY
Row = [7.2]
Re solution

d) The Root Mean Square Error (RMSE) is calculated and, finally image
registration was done to the ground control points. From the
correspondence file the image was geo-referenced by applying the
Resample function as shown in Figure 7.9. Figures 7.10 and 7.11 show an
example of geo-referenced image with overlaid catchments.

Figure 7.9 Example of Resampling procedure in IDRISI ANDES and Reference

Parameters window

146
Figure 7.10 Example of geo-referenced true color composite of a satellite image:
band 345 of 1990 with X, Y coordinates and overlaid catchment.
(After Birhanu, 2008)

After image Resampling and geometric correction, image classification was

performed to convert image data in to thematic data. A supervised
classification with the maximum likelihood algorithm can be conducted to
classify the Landsat images using six bands (band 1 to 5 and then 7) leaving
band 6 because it is a thermal mapping band.

In the supervised classification of Landsat images, pixels were identified for

reclassification and polygons were digitized around each training site, assigning
a unique identifier to each cover type. Figure 7.12 shows training site prepared
for a Landsat image and after training sites are developed a signature file an be
prepared as shown in Figure 7.13.

Figure 7.11 Example of a mosaicked and geo-referenced true color composite of

satellite images P166R65 and P167R65 overlaid with catchments.
(After Birhanu, 2008)

147
 Wooded Grassland

Bushland and Scattered

cropland
Grassland and Scattered
cropland
Natural Forest

Figure 7.12 Training sites development

Figure 7.13 Preparation of spectral Signature Files for the trained sites.

The last step of image processing is post processing. The post processing involves
output display and change detection of the analyzed images. The GIS Analysis
option of the IDRISI ANDES gives an over lay of the rasterized catchment polygon
and reclassified image.

148
7.3.2 Analysis of Catchment Cover Change Detections
At present several successful methods are used to describe the patterns and
processes of landuse and land coverage changes quantitatively based on
satellite remote-sensing data (Liu and Buheoser, 2000; Liu et al., 2002). The IDRISI
ANDES gives the area of each landuse/land cover of the extracted image and
from the series of images, a percentage change was computed. The change
of land use and land cover will be estimated by subtracting the areas
difference.

Area − Area yearx + t

= × 100 %
yearx
% change [7.3]
Area yearx
Where: Area yearx = area of cover i at the first year
Area yearx +t = area of cover i at the second year
t years = period in years between the first and second image scene
acquisition

149
 CHAPTER EIGHT

PREDICTION IN UNGAGED CATCHMENT

8.0 Introduction
Prediction in ungauged catchment using regionalization technique is a top
research priority in the fields of applied hydrology as it is difficult to gage many
watersheds due to complex physiographic conditions.
One of the most important applications for the case of ungaged catchment
goes to determining the size of the water harvesting structure, that is, the volume
of water that will be expected to be generated from the catchment. Availability
of data is more serious problem to determine flow quantity and its variation in
time at the point of interest in most catchments of the Amhara region. Hence
different statistical and hydrologic techniques have been developed to predict
flow for ungaged catchments.
The SCS method discussed in chapter four is entirely dependent on physical
features of the watershed such as hydrologic soil group, landuse and catchment
treatment which are necessary for computing the losses. Watershed transform
can be also modelled with due consideration of river length, catchment slope
besides to soil and landuse information of the watershed. Therefore the SCS
method is applicable for ungaged catchments. SCS model features and
parameter calibration were discussed in depth in chapter four.

8.1 Characterization of Gaged and Ungaged Catchments

Catchments are said to be gaged if enough hydrological information (Flow,
Gage height, Rating curves, and Sediment data) are obtained at a point of
interest where river or watershed structures are to be constructed. Information
that are obtained for the gaged catchments can easily be incorporated into
many hydrological modelling for facility design and decision support system.
Hence hydrological model parameters are derived from gaged catchments by
calibration and validation of flow and sediment procedures and can easily be
transferred to ungaged catchments. For ungaged catchments, the parameters
of rainfall-runoff models cannot be obtained by calibration and hence need to
be obtained by other methods.

8.2 Methods of Ungaged Catchment Analysis

8.2.1 Use of statistical Rules in Data Transfer-Regionalization Approach

The regionalization approach is based on regionally available streamflow
records and regional analysis of unit catchment runoff (specific yield of
catchments). The flow magnitudes for ungaged catchments can be estimated
by relating catchment characteristics with known discharge values at gaged

150
sites on a definite statistical relationship. Logarithmic values of catchment
characteristics and discharge can be related using the following relationship:
LogQ = fLogA [8.1]
Where Q is the discharge at gaged site and A is catchment area.

The use of regression analysis will provide the following expression:

LogQ = nLogA + C [8.2]

Using the above relationships the most familiar exponential relationship is

established as shown below.

Q = CA n [8.3]
Where Q is unit catchment runoff ( l / sec/ km 2 or m 3 / km 2 )
A is drainage area ( ( km 2 )
C and n are coefficient and index respectively for the region under
consideration.
The above relationship can further be strengthened by incorporating other
relevant catchment characteristics that include: mean annual rainfall of the
catchment, average slope and average elevation in mountainous areas, stream
length, drainage density and other pertinent information.

With the inclusion of multiple catchment characteristics the general form of the
multiple regression equation for the case of ungaged catchment is as follows:

Q = CX 1 1 CX 2 2 CX 3 3 ............CX m m [8.4]
b b b b

Where Q is the streamflow characteristics, CX 1 , CX 2 , CX 3 ....CX m are catchment

characteristics (rainfall, area, slope, altitude, stream length, drainage density
etc..), b1 , b2 , b3 .....bm are coefficient of the independent variables
CX 1 , CX 2 , CX 3 ....CX m respectively. The above model can be linearized by a
logarithmic transformation as shown below.

LogQ = B1 log CX 1 + B2 log CX 2 + ......... + Bm log CX m [8.5]

Significant parameters to be included in the above model need to be specified

and one has to cross check whether the data are adequately described by the
regression model or not. Statistical measure of efficiency, the coefficient of
determination, R 2 , is used as a measure of the ability of the regression line to
explain variation in the dependent variable. Information regarding rainfall, area,
elevation, stream length, slope of the main stream and shape index could be
made available. The physiographic catchment characteristics are measured
from a topographical map of scale 1:50,000.

151
An example of practical approach for the estimation of peak flow for ungaged
catchments for the Ethiopian catchments was done by Gebeyehu (???). After
multiple regression analysis of 78 catchments with drainage area size from 20km2
to 66,000km2, a formula indicated by equation 8.6 was used for the estimation of
peak flow for ungaged catchments based on least square regionalization
approach.

Qmax = 0.87 ( Area ) 0.7 [8.6]

Where Area is catchment area in km2 and Qmax mean annual flood in m3/s.

8.2.2 Use of Hydrologic Models

Establishing rainfall-runoff relationship of a given watershed requires good set of
input and observed flow data either at the outlet of the watershed or its sub-
watersheds. Hydrological models predominantly based on physical parameters
are often used to determine the flow characteristics of ungaged catchments.
The most commonly used approach to ungaged catchments modelling is to
relate model parameters and catchment characteristics in a statistical manner,
through regression models. Catchment characteristics are those that can be
observed or estimated directly from measurements of catchment or channel
characteristics. This study considered the catchment characteristics that can be
derived from sources that are readily available to practicing hydrologists, i.e.
maps, GIS processing, and field measurements.

Flow from ungaged catchments can be estimated from conceptual hydrologic

models in either of two ways. First, successful calibration and validation of a
conceptual hydrologic model helps to create regional parameter sets which are
referred as ‘model parameters mapping’ (Pitman, 1973, Migdley et al, 1994;
Kapangaziwiri, 2007). The second method is to develop conceptual hydrologic
model parameters from catchment characteristics which is done by creating
regression relationships between catchment characteristics and model
parameters.
This chapter presents prediction of hydrologic variables for the ungaged
catchments using the Pitman hydrologic model. The Pitman model is suitable for
long-term hydrological data synthesis and parameter estimation for ungaged
catchments (Pitman, 1973).

[Link] The Pitman Hydrologic Model

The essential elements of the model are precipitation and potential
evapotranspiration. The original model was designed to accept input data of
coarse time resolution (one month). But provision was made to solve the water
balance of the catchment at much smaller time intervals.

152
The Pitman model was developed in South Africa for regional water resources
assessment (Pitman, 1973). In the Pitman model calculations begin from
assumed soil moisture conditions and are terminated when input data are
exhausted. Precipitation is stored as interception and as soil moisture and this is
subject to evaporation and transpiration. The quantity of precipitation that is not
absorbed by the soil is the source of surface runoff. A portion of the precipitation
held as soil moisture finds its way to the river system. By suitably lagging the
various components, one may compute the total runoff volume at the
catchment outlet. Figure 8.1 presents the Pitman catchment model flow chart.

Parame Units Description

ter
POW - Power of Soil Moisture-Runoff
equations
SL Mm Soil Moisture Storage below which
no runoff occurs(Wilting point)
ST Mm Maximum Soil Moisture Capacity
FT mm/mo Runoff from soil when Soil Moisture
nth is at full capacity
GW mm/mo Maximum Groundwater runoff

153
 nth
Al % Impervious portion of the
catchment
ZMIN mm/mo Minimum catchment absorption
nth rate
ZMAX mm/mo Maximum catchment absorption
nth rate
PI Mm Interception storage
TL months Lag of surface runoff
GL months Lag of runoff from soil moisture
R - Evaporation-Soil Moisture storage
relationship
Figure 8.1 Pitman Catchment Model Flowchart and Parameters for optimization
(Pitman, 1973)

The Pitman model has been kept to as simple a form as possible so as to

minimise the number of parameters to be evaluated (Pitman, 1973). It has been
tested on about fifty gaged South African catchments representing a wide
variety of hydrological regimes (Pitman, 1973; Hughes, 1996; Jeremy Meigh and
Matt Fry, 2004; Koblenz, 2006). The model has gained widespread use in SADC
countries (Hughes, 1996; Mwelwa, 2004). And it has been successfully
implemented in Pangani basin, Rufiji basin and Mkomazi catchment in Tanzania
(Birhanu, 2008, PBWO/IUCN, 2006) to support water resources management and
water capacity decisions. The model’s monthly time step lends itself to the
availability and quality of hydro-met data in most African catchments.

[Link] Application of Pitman Model to Ungaged Catchments

There are two approaches to parameter regionalization: the first is to relate
parameter values to measured physical basin characteristics and develop
predictive equations for parameter values (Hughes et al., 2006); the second is to
identify regions where calibrated parameter values are broadly similar
(parameter mapping). The physical basin characteristics of the Blue Nile basin
catchments are so complex that relationships with the Pitman model parameters
could not be easily established. However an attempt is made in this hydrologic
guide book to relate few model parameters with catchment characteristics.

Most recent studies in the parameterization of Pitman model (Huges et al., 2006;
Huges, 2004a) indicated methods to estimate parameters of the soil moisture
accounting and subsurface runoff from catchment characteristics. There is lack
of systematic database on catchment characteristics in the Blue Nile basin
catchments unlike the widely available database in South Africa where initial

154
concepts of parameterization of Pitman model from catchment characteristics
were made. However, attempts are made to estimate parameters that control
the soil moisture accounting and subsurface runoff from properties derived from
catchment characteristics (soil, soil texture, physiographic conditions, landuse,
hydro meteorological data and geology).

Maximum Soil Moisture Storage (ST)

The maximum moisture storage (ST) parameter of the Pitman model is assumed
to be made up of two components; the soil component and the unsaturated
ST = STsoil + STunsat
zone component (i.e., soil ( mm ) ).

STsoil
Estimating
STsoil is the soil storage depth (mm) at saturation representing the immediate
store of infiltrated rainfall before it is lost to either evapotranspiration or to
percolation and runoff. The maximum amount of moisture of the ‘soil’
ST
component soil is estimated as follows:

STsoil ( mm ) = POR (%) * VVAR (%) * SoilDepth ( m) / 10

[8.7]

Where POR is the soil porosity which is a measure of the moisture holding
capacity, VVAR represents a correction factor for vertical variations in porosity.

The estimate of porosity is based on the soil texture class as presented in Table
8.1. There are no clear methods to estimate the VVAR (%) of porosity in the
catchment as it is essentially subjective and guided by the extent of the
variation. Practical experiences indicated that VVAR estimates of 62% for a
steep, moderately deep sandy loam and 85% for an undulating topography
having deep sandy soils of granite-gnessis types of geology are recommended.

155
Table 8.1 Soil texture classes according to USDA (1969), based on percentage
volumes of sand, silt, clay and quartz content.

Texture Class Sand (%) Silt (%) Clay Quartz Assumed

(%) (%) porosity
(%)

Sand 92 5 3 92 42

Loamy sand 82 12 6 82 40

Sandy clay 58 15 27 60 33
loam

Sandy clay 52 6 42 52 32

Clay 22 20 58 25 39

Pitman model parameter estimation software was provided by the Institute of

Water Research (IWR) from Rhodes University, South Africa as shown in Figure 8.2
to estimate soil parameters. Soil depth data can be obtained from SOTER
database (Dijkshoorn, 2003).

Figure 8.2 Illustration of the default basin property and parameter estimation
program.

156
Estimating STunsat
In estimating STunsat , the basic assumption that is widely applied is that water
percolating downwards in the unsaturated zone will have two directional
components; a vertical one contributing directly to recharge of the saturated
ground water zone and a lateral one that could contribute to the re-emergence
of subsurface water at a spring or seep. The vector result of these two
components is referred to here as the drainage vector slope (VS in Figure. 8.3),
which is estimated in the default procedure using % values for the vertical and
horizontal components.

Area between surface and the

drainage vector slope

Area between drainage vector

slope and ground water slope

Figure 8.3 Conceptualization of the subsurface drainage that determines the

Interflow process from the unsaturated zone.

If the total unsaturated zone potential storage (mm depth) is expressed as the
product of the mean depth to ground water (DGW m) and the storativity (S) of
ST
the unsaturated zone material, then the final estimate of unsat becomes:

ST (mm) = DGW * 1000 * S * Ratio

If BS > VS then unsat [8.8]
ST (mm) = 0
If BS ≤ VS then unsat [8.9]

Vector Slope (VS) of 4.2% was determined from geology of catchment using the
default parameter estimation software as shown in Figure 8.4.

157
Figure 8.4 Estimation of the vector slope using parameter estimation software.

The information on depth to ground water needed for this estimation may be
reasonably accurate in areas where comprehensive borehole drilling records
exist. Estimations for areas without this kind of information may introduce some
uncertainties. Obtaining representative values of storativity was also a problem
and may introduce a further source of uncertainty.

Estimating Runoff from the soil moisture store (FT)

Runoff from soil moisture at full capacity was assumed to be the sum of runoff
from the soil ( FTsoil ) and runoff from unsaturated zone ( FTunsat ). FTsoil is the
maximum subsurface outflow when the basin’s soils are at saturation and is
assumed to occur through the banks of the active channel. The contributing
area (CA, in km2 km-2) is given as follows:

CA = 2 * DD(kmkm −2 ) * FTSoilDepth(m) / 1000 [8.10]

where DD is the basin’s drainage density which is estimated from topographic
maps. The soil depth value used should be based on the soil depths in the lower
topographic units of the basin (i.e FTsoil depth).
FT −1
The monthly depth of interflow from the soil ( soil , in mmmonth ) was thus
assumed to be adequately explained as a function of CA , saturated hydraulic
−1
conductivity of the basin soils, K ( md ) and the mean basin slope in decimal
( BS ) and expressed as follows;

FTsoil = CA * K * BS * 30 *1000
[8.11]

Based on area weighted soil texture classes and characteristics of surface cover
conditions, Parameter estimation software provides the value of K as shown in
Figure 8.5.

158
Figure 8.5 Estimation of Parameter K

Estimating the outflow from the unsaturated zone ( FTunsat ) is by far a greater
challenge. The estimation approach assumes either saturated flow in the
fracture zones or a perched water table and is based on defining a
2 −1
representative Transmissivity (T in m d )

FTunsat ( mm ) = 2 * DD * T * VS * 30 / 100 [8.12]

−2
Where DD is drainage density ( kmkm ), T is Transmissivity, VS is a drainage vector
slope. The values of Transmissivity currently used within the default estimation
2 −1
program vary from 0.5 to 5 m d .

Figure 8.6 Estimation of Parameter T (m2/day)

159
Estimating POW

In the absence of detailed field data, simpler approach based on the

probability distributed principle of Moore (1985) would produce an estimate of
POW.

The concept is illustrated in Fig. 8.7. The four lines represent cumulative Normal
distribution frequency curves for mean basin moisture contents of 0.2 to 0.8,
each having a different standard deviation. If a method of estimating the
variation in the standard deviation with mean moisture content can be found, it
follows that a relationship between mean moisture content and relative runoff
(i.e. runoff relative to the maximum at full basin saturation) can be developed.
The approach adopted uses quite arbitrary equations to achieve these
principles and is based on a SDEV parameter that is assumed to vary with basin
properties:

If RAT > 0.75 then SD = (1.1 − RAT ) * SDEV /(1.1 − 0.75) [8.13]
If RAT ≤ 0.75 then SD = ( RAT + ( 0 . 75 − RAT ) * 0 . 2 ) * SDEV / 0 . 75 [8.14]
where RAT is mean relative moisture content, SDEV is maximum standard
deviation, and SD is standard deviation at RAT.

100
Mean 0.2, SD 0.25
Mean 0.4, SD 0.38
80 Mean 0.6, SD 0.50
Mean 0.8, SD 0.51
Frequency

60

40

20

0
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
Relative moisture content

Figure 8.7 Illustration of the concept of using a frequency distribution to describe

the spatial distribution of soil moisture for different mean moisture
contents (After Kapangaziwiri, 2007).

The resulting relationship between mean relative basin moisture content and
relative runoff is then identical to the format of the Pitman model ‘soil’ moisture
S / STsoil for the
runoff function if this is expressed in non-dimensional terms (i.e.

160
 Q / FTsoil for the vertical axis). It is assumed that the standard
horizontal axis and
deviation (SDEV) at a mean moisture content of 0.75 can be established from
the physical attributes of the basin.
Appropriate values of POW in the Pitman model function ( Q / FT = S / ST ) have
been manually calibrated to reproduce similar shaped curves ( Figure 8.8).

Figure 8.8 Estimation of POW using default estimation programme

Estimation of soil surface Infiltration Parameters

The approach is the use of soil properties to define the parameters of a modified
form of the Kostiakov equation (Hughes and Sami, 1994), basin hydro-
meteorological characteristics to disaggregate monthly rainfall and to apply the
infiltration equation to estimate surface runoff for a range of monthly rainfalls.
The infiltration parameters are then manually fitted to match the infiltration
equation based estimates of runoff for different monthly rainfalls. The Kostiakov
equation (Kostiakov, 1932) to estimate surface infiltration rate is given as follows:

Infiltrationrate = (mmh −1 ) = k * C * T k −1 [8.15]

where k and C are parameters and T is cumulative time in minutes from the start
of the storm. The daily input rainfalls are disaggregated into 5 time periods based
on a parameter representing the expected mean storm duration, MSD (h) using
the following equation.

RINT = rain * st var(k ) * 60 / MSD [8.16]

where RINT is the rainfall intensity (mm h-1), rain is the daily rainfall (mm),
and stvar is a distribution constant for each of the 5 time periods (0.045, 0.184,
0.383, 0.300, and 0.088 for time periods 1 to 5 respectively). Thus the parameters

161
(ZMIN and ZMAX) are then adjusted and the function plotted as a cumulative
curve to be similar to the infiltration function results (Figure.8.9).

Figure 8.9 Estimation of Infiltration parameters

According to the above descriptions Pitman model parameters can be related

to catchment characteristics and are presented in Table 8.2.

162
Table 10.9 Typical catchment characteristics and Pitman model parameters

BASIN 1 2 3 4 ………………….. N

MAP (mm)

Basin Area
(Km2)

Drainage
Density
(Km/Km2)

Mean Basin
Slope (%)

Drainage
Vector Slope
(%)

Mean Monthly
Rainfall (mm)

Mean Number
of Rainy Days
(Month)

Mean Annual
Evaporation
(mm)

Mean Soil
Depth (mm)

Soil Porosity (%)

Vertical
Variation
Factor (%)

Soil
Permeability
(m/day)

ST (mm/Month)

FTsoil(mm/Mont

163
h)

FTunsat(mm/M
onth)

FT (mm/Month)

POW

ZMIN(mm)

ZMAX(mm)

164
 References

Beven, K.J., (2006). Rainfall-Runoff Modelling, The Primer. John Wiley & Sons Ltd,
The Atrium.
Bilate River Basin Development Study (1974), Volume I, National Water Resources
Commission, Addis Ababa.
Biniam Biruck (2009). Formulation of best management option for a watershed
using SWAT (Anjeni Watershed, Blue Nile Basin, Ethiopia). [Link] Thesis. Cornell
University
Chow, Ven Te., (1988). Applied Hydrology, McGraw- Hill series in water resources
Dereje et al., 2006 from semu’s paper
Dijkshoorn, J. A. (2003) SOTER database for Southern Africa (SOTERSAF) Technical
Report. Wageningen, International Soil Reference and Information Centre.
Dingman, S.L., (2002). Physical Hydrology, Second Edition. (2nd ed.) Upper Saddle
P P

River, N.J.: Prentice Hall.

Doorenbos, J., and W.O. Pruitt., (1977). Crop water requirements, irrigation and
Drainage paper 24, U.N. Food and Agriculture Organization, Rome, Italy, 1977.
field manual for conservation practices.
Gebeyehu(1989) Regional flood frequency analysis. Hydraulics laboratory, The
Royal Institute of Technology of Stockholm, Bulletin No. TRITA-VBI-148, Sweden.
Ghosh, S.N. (1999). Flood Control and Drainage Engineering, 2nd Edition. A.A.
P P

Hughes, D.A., (1996). Southern Africa Friend-Rainfall –Runoff Modelling. Institute

for Water Research. Rhodes University, Grahamstown, South Africa.
Jeremy Meigh and Matt Fry., (2004). Southern Africa FRIEND Phase II 2000-2003.
IHP-VI Technical Documents in Hydrology/No.69 UNESCO, Paris.
Kassa Tadele and Gerd Forch (2007) Impact of Landuse/ Land Cover change on
streamflow: A case study of Hare watershed, Ethiopia. FWU Water Resources
Koblenz.,(2006) FRIEND- a global perispective [Link] Regimes from
International Experimental and Network Data. Edited by Eric Servat and Siegfried
Demuth.
KOGA Irrigation and Watershed Management Project (2004). Hydrology Factual
Report
McCulloch,J. S.O., (1965). Tables for the rapid computation of the Penman
estimate of evaporation. East Afr. [Link]. J.30: 286:295.
Mohammod (2009). Draft PhD Thesis, submitted for external examination.
Moore, R.J., (1985). The probability-distributed principle and runoff production at
point and basin scales. Hydrological Sciences Journal, vol. 30(2), pp. 273-297.
Mwelwa, E.M., (2004). The Application of the monthly time step Pitman Rainfall-
Runoff Model to the Kafue River Basin of Zambia. [Link] Thesis, Rhodes University,
Grahamstown, South Africa
Neitsch, S.L., Arnold, J.G., Kiniry, J.R., Williams, J.R., King, K.W., (2002). Soil and
Water Assessment Tool, Theoretical Documentation and user’s manual, Version
2000.

165
Nigussie, T.G.(2005). Investigation on Sediment Transport Characteristics and
Impacts of Human Activities on Morphological Processes of Ethiopian Rivers:
Case Study of Kulfo River, Southern Ethiopia. PhD dissertation, Dresden University
of Technology, Dresden, Germany.
PBWO/IUCN, (2006) Hydrology and System Analysis. Volume 1 of 2:The Hydrology
of the Pangani River Basin. Report 1:Pangani River Basin Flow Assessment
Initiative, Moshi
Pike, J.G., (1964) The estimation of annual runoff from meteorological data in a
tropical climate. Journal of Hydrology 2, 116-123.
Pitman, W.V (1973) A Mathematical Model for Generating Monthly River Flows
from Meteorological Data in South Africa. Report No.2/73 Hydrological Research
Unit.
Raghunath (2006), Hydrology: Second Revised Edition
Rasmusson, E.m., Dickinson, R.E., Kutzbatch, J.E., Cleaveland, M.K., (1993).
Climatology, Hand Book of Hydrology/ David [Link], editor in chief. Pp 2.1-
2.11.
Sir Alexander GIBB & Partners (1988). Gelana Irrigation Project Feasibility Study
Final Report, volume 2.
Soil Conservation Service (SCS), (1972): National Engineering Handbook, Sect. 4,
Hydrology, Chapt. 10, Estimation of direct runoff from storm rainfall by Victor
Mockus, 1972.
Tegenu Ashagire (2009). Modeling Rainfall, Runoff And Soil Loss Relationships In
the Northeastern Highlands Of Ethiopia, Andit Tid Watershed
Tilrem, O.A., (1979). Stage discharge relations at stream gauging stations.
Manual on procedures in Operational hydrology. Ministry of water energy and
minerals. Tanzania. Norwegian agency for International development (NORAD).
WMO, (1994). Guide to Hydrological Practices, Data Acquisition and processing,
analysis, forecasting and other applications. WMO-No. 168. Fifth edition, 1994.
Yang, C.T. (1996). Sediment Transport: Theory and Practice, McGraw-Hill
Companies, Inc., New York.
Zeray, L., Roehrig, J. and Chekol, D.A (2007) Climate Change Impact on Lake
Ziway Watershed Water Availability, Ethiopia. FWU Water Resources Publications.
Volume No: 06/2007. ISSN No. 1613-1045. pp 18-23.

166