See discussions, stats, and author profiles for this publication at: [Link]

net/publication/287224448

Remote sensing of aboveground forest biomass: A review

Article in Tropical Ecology · May 2016

CITATIONS READS

114 7,268

5 authors, including:

Onisimo Mutanga Cletah Shoko

University of KwaZulu-Natal University of the Witwatersrand
446 PUBLICATIONS 16,148 CITATIONS 83 PUBLICATIONS 1,392 CITATIONS

SEE PROFILE SEE PROFILE

Sam Adewale Adelabu

University of the Free State
91 PUBLICATIONS 1,200 CITATIONS

SEE PROFILE

All content following this page was uploaded by Tsitsi Bangira on 03 March 2020.

The user has requested enhancement of the downloaded file.

Tropical Ecology 57(2): 125-132, 2016 ISSN 0564-3295
© International Society for Tropical Ecology
[Link]

Remote sensing of aboveground forest biomass: A review

1* 1 1 2
DUBE TIMOTHY , MUTANGA ONISIMO , SHOKO CLETAH , SAMUEL ADELABU &
1
BANGIRA TSITSI

1
Discipline of Geography, School of Agricultural, Earth and Environmental Sciences, University
of KwaZulu-Natal, P/Bag X01, Scottsville, Pietermaritzburg 3209, South Africa
2
Department of Geography, University of the Free State QwaQwa Campus, Kestell Road, Private
Bag X13, Phuthadijhaba 9866, South Africa

Abstract: Forests are central to global carbon cycle, therefore, accurate inventorying and
monitoring of forest aboveground biomass in local to regional scales is critical in understanding
their role as atmospheric carbon sinks or sources. This article provides a review of various
remote sensing applications in forest aboveground biomass inventorying and monitoring as well
as highlights the associated challenges and opportunities. The review concluded that the use of
remote sensing in large-scale forest aboveground biomass quantification provides plausible
alternatives, when compared to the use of conventional approaches, which are labour-, cost-,
and time-intensive and sometimes inapplicable due to poor accessibility. It was noted that
although remote sensing provides reasonably accurate forest aboveground biomass estimates,
active sensors, such as LiDAR and radar are not fully operational as yet due to complex pre-
processing and high cost of data acquisition.

Key words: Aboveground biomass, allometric equations, biomass expansion factor,

forest, LiDAR, multispectral remote sensing.

Handling Editor: S.P.S. Kushwaha

Introduction biomass (AGB), and forest carbon stocks (Hyyppä

et al. 2000; Ketterings et al. 2001; Steininger 2000).
Considering the natural and fundamental role Repeated monitoring of AGB is an important task
in the basic functioning of the biosphere by for sustainable and effective management of
regulating global carbon cycle, forests unequi- forests (Joshi et al. 2014). This involves data on
vocally reduce atmospheric carbon content various forest parameters such as height, basal
considerably. Despite the ravaging impacts of area, diameter-at-breast height (dbh), etc. (Field et
global climate change, forests are capable of al. 1998). It is also important to have up-to-date
stabilizing atmospheric carbon dioxide concen- knowledge on available low-cost forest AGB
trations thereby mitigating the global warming estimation methods (Mutanga et al. 2012;
and the climate change. It is estimated that 2-4 Gt Muukkonen & Heiskanen 2007).
C of atmospheric carbon can be sequestered by Conventionally, the AGB monitoring requires
forests annually (Lu et al. 2010; Qureshi et al. intensive field inventory, including taxonomical
2012). Hence, understanding of the global climate information, collateral and ancillary data analysis
change necessarily involves the quantification of and mapping of the forest cover using remote
forest parameters such as volume, aboveground sensing or other means. The conventional moni-
*Corresponding Author; e-mail: timothydube3@[Link]
126 REMOTE SENSING OF ABOVEGROUND FOREST BIOMASS

toring of AGB is labour-, cost-, and time-intensive stratification by ecological zones for forest AGB
and sometimes inapplicable due to poor estimation does not accurately reflect the actual
accessibility, making it practical only in relatively tree AGB in a specific area or region. Remote
smaller areas (Mutanga et al. 2012). Remote sensing methods have provided better alternative
sensing, on the other hand, offers an efficient and for forest AGB monitoring in terms of accessibility
economical means for AGB monitoring by facili- and the cost, most importantly in the developing
tating forest type and canopy density stratification, world. While the use of conventional methods for
which greatly helps in field inventory. Its repeated the monitoring of AGB is not totally rejected by
coverage offers historical data required for change the authors of this review, integrating them with
detection, while its digital data format can be modern remote sensing data would considerably
easily integrated in a geographic information help in quantifying, monitoring, and under-
system (GIS) for further analysis. Many standing forest AGB at various scales (e.g.,
researchers have used remote sensing data to Chinembiri et al. 2013; Cryus & Tanja 2004; Min
monitor forest AGB in different parts of the world et al. 2009; Mitchard et al. 2009; Patenaude et al.
(e.g., Cryus & Tanja 2004; Min et al. 2009; 2005; Straub et al. 2009; Zheng et al. 2004).
Mitchard et al. 2009; Patenaude et al. 2005; Straub
et al. 2009; Zheng et al. 2004). Satellite remote sensing of forest AGB
The underlying goal of this article is to: (i)
The fast developments in remote sensing
review and synthesize different forest AGB esti-
technology have provided multispectral, hyper-
mation methods, using remotely-sensed data and
spectral, light detection and ranging (LiDAR), and
(ii) highlight the limitations and successes in the
radio detection and ranging (radar) data. The most
application of the remote sensing-based methods.
In order to address these aims, we firstly present commonly used ones are the multispectral sensors,
an overview of different conventional methods which measure reflectence by ground features in
available for accurate estimation of forest AGB the visible, near-, middle-, and far-infrared
and secondly provide different AGB remote portions of the electromagnetic spectrum. These
sensing variables as well as highlight the effects of sensors can be used for many forest-related
bioclimatic factors (i.e. age, species/forest type, studies. A survey of the forest studies that used
rainfall and topography). Finally, the strengths the remote sensing, shows that remote sensing can
and weaknesses of different remote sensing data provide information critically required for forest
used for AGB monitoring with specific emphasis on AGB assessment.
those applicable to South Africa are discussed.
Forest AGB estimation using optical remote
Conventional AGB estimation techniques sensing data
Optical remote sensing makes use of natural
The Intergovernmental Panel on Climate
radiation from sun and provides a two-dimensional
Change (IPCC) has aggregated AGB estimation
view of forests and other earth surface features.
methods into three different groups i.e., Tier-1, -2
Due to easy accessibility and affordability, a num-
and -3. Each of these tiers represent the level of
ber of studies have employed optical remote sensing
methodological complexity in AGB estimation
for forest AGB estimation (Basuki et al. 2013;
(Henry et al. 2011). Following the above grouping,
Kajisa et al. 2009; Lu 2006; Lu et al. 2012) (see also
literature shows that forest AGB and changes can
Table 1). One of the strengths of optical remote
be estimated directly or indirectly from forest
sensing data both is that it is operational at local to
inventory data by either using allometric
equations, biomass expansion factors (BEF) or global scales with sensors, such as Landsat TM,
conversion factors such as wood density (Chave et AVHRR, and MODIS providing globally consistent
al. 2008; Dovey 2009; Muukkonen & Heiskanen spatial data. In contrast, a study by Lu (2006)
2007). Although Henry et al. (2011), reported that demonstrated that the use of coarse spatial reso-
conventional techniques generally provide accurate lution sensors (i.e., Landsat, MODIS etc.) for AGB
estimates of forest AGB, it is still regarded as estimation resulted in poor prediction accuracy due
environmentally unfriendly, time-consuming to the presence of mixed pixels together with a
method, largely applicable to small sample sizes. mismatch between the size of compartments and
Segura & Kanninen (2005) argued that the use of the pixel. However, Basuki et al. (2013) observed
generic equations in conventional methods and that the application of advanced techniques, such
 DUBE et al. 127

Table 1. A summary of the use of multispectral data in estimating biomass.

Sensor Study area Approach Findings Reference

LiDAR& Gansu province, Stepwise multiple regressions Integration of LiDAR and Qisheng
SPOT-5 western China were used. SPOT-5 data can increase (2012)
HRG biomass estimation accuracy
imagery than use of only LiDAR data (R2
of 0.736 and 18.64 ton ha-1)
Landsat East Discrete wavelet transforms Biomass estimates ranged Basuki et
PALSAR Kalimantan, (DWT), Brovey transform were between 0.70–0.75 r2 values. al. (2013)
Indonesia used.
Landsat Georgia forest Vegetation indices and Hardwoods biomass was Min et al.
land multiple regression analyses estimated with R2 of 0.52, 0.30 (2009b)
were used to develop AGB for softwoods and 0.66 for mixed
estimation models. forests.
Landsat Kampong Thom Object-based approach was ABG estimates ranges between Kajisa et al.
(ETM+). Province in used. 0.67 and 0.76 R2. (2009)
central
Cambodia
SPOT-5 Sun Yat-Sen, Gray Level Co-occurrence The results showed that ABG Li et al.
HRG Nanjing, China Matrix was applied. was poorly correlated with most (2008)
imagery textures.
Landsat TM Western Biomass from Cluster BioCLUST offered plausible Luther et
imagery Newfoundland, Labeling based on Structure results. al. (2006)
Canada and Type (BioCLUST), was
used.

as spectral mixture analysis to decompose a a good correlation (r2 = 0.87) between AGB
mixture of spectral components of Landsat ETM+ estimates derived from the AISA Eagle and LiDAR
into vegetation, soil, and shade fractions and the data. Treuhaft et al. (2003) reported that forest
integration of these components with radar data, AGB estimated from forest structure profiles using
using the discrete wavelet transform (DWT), hyperspectral data may be more accurate when
improves forest AGB estimation accuracy compared to estimates obtained from microwave
significantly when compared to previous studies. power or optical radiance measurements. The
Kajisa et al. (2009), who applied an object-based importance of different wave-lengths in AGB
and statistical methods together with textural estimation has been investigated by different
attributes derived from coarse images, observed several researchers with plausible results. Hongrui
significant improvements in AGB estimation et al. (2011) estimated biomass of desert steppe in
accuracies. More recent research has focused on inner Mongolia based on red-edge reflectance
the use of hyperspectral data for forest AGB curve area method. The results illustrated that the
estimation. red-edge reflectance curve area (between norma-
lized reflectance curve and wavelength in 680 - 780
Forest aboveground biomass estimation using nm region) performed better when compared with
hyperspectral remote sensing data the use of conventional vegetation indices or the
Sevral studies have used hyperspectral remote red-edge position alone. This method resulted in
sensing to map forest AGB (Clark et al. 2011; lower standard error of prediction (26.4 g m-2)
Goodenough et al. 2008; Koch 2010; le Maire et al. compared to that from optimal narrow-band ratio
2008; Treuhaft et al. 2003). Goodenough et al. vegetation index (37.4 g m-2).
(2008) compared the use of hyperspectral AISA Swatantran et al. (2011) observed that hyper-
Eagle data to AVIRIS data and found that both spectral data is prone to saturation in dense forest
images could be successfully employed in deriving canopies, such as in the tropical rain forest. Clark
plausible forest AGB estimates. The study showed et al. (2011) estimated forest AGB using the 1.6 m
128 REMOTE SENSING OF ABOVEGROUND FOREST BIOMASS

spatial resolution hyperspectral imagery from the demonstrated accurate estimation of forest
HYDICE (Hyperspectral Digital Imagery Collection structural attributes, the most limiting factors in
Experiment) airborne sensor with 210 bands applying this method are: the (i) saturation level
covering the full range of the electro-magnetic at various wavelengths (C, L and P bands) (ii)
spectrum. The results yielded lower forest AGB polarizations (e.g., HV and VV), and the (iii)
estimates when compared to estimates derived vegetation stand characteristics including ground
using LiDAR metrics with an r2 value of 0.90 and conditions. Due to these limitations, the backs-
RMSE of 38.3 Mg ha-1. Despite the lower estimates catter technique can not be applied on any
by the hyperspectral sensors, they concluded that vegetation type without taking note of the stand
airborne LiDAR and hyperspectral sensors can characteristics and ground conditions. The other
play an important role in the development of limitation of the radar remote sensing methods,
future sensors (Clark et al. 2011). A study by based on the evaluation of backscattering
Anderson et al. (2008) demonstrated that inte- amplitudes for the quantification of forest AGB, is
grating hyper-spectral and waveform LiDAR data that they saturate in forest areas characterized by
provide improved AGB estimates when compared biomass above 150 t ha-1 (Cutler et al. 2012). As a
to the use of these data sets independently. In all result, research has now shifted towards the use of
forest conditions, 8 - 9 % more variation in basal synthetic aperture radar (SAR) interferometry,
area (BA), forest AGB and quadratic mean stem polarimetry alone or a combination of both.
diameter (QMSD) was explained by the use of the However, Pulliainen et al. (2003) observed that the
multi-sensor data than either AVIRIS or LVIS accuracy of this approach is mostly dependant on
alone, with the estimated error ranging from 5 to the number of images used or site conditions viz.,
8 % (Anderson et al. 2008). wind speed, moisture and whether the tempe-
Although hyperspectral data demonstrates rature was below freezing point.
some successes in AGB estimation, the data also Uncertainties such as complexity of atmo-
suffers from band redundancy for specific appli- sphere, course temporal and spatial resolutions,
cations. For example, neighbouring bands or even saturation problems in deriving forest AGB
bands from different parts of the spectrum may be estimates associated with broad optical sensors
strongly correlated and contain highly similar and radar data has led to the introduction of
information. Furthermore, the use of non-para- LiDAR sensor to improve on these limitations.
metric techniques, such as spectral similarity Thus, the advent of LiDAR imaging techniques
measures, sub-pixel classification techniques, coupled with advanced statistical techniques
machine learning methods and decision tree classi- ranging from simple regression between LiDAR-
fication techniques, make no assumption of data derived height metrics and forest AGB to methods
distribution, despite them being robust in reducing including automated tree crown delineation,
data dimensionality (Cho et al. 2010). In fact, these stochastic simulation and machine learning
methods are computationally intensive, especially approaches, have resulted in a number of different
when applied on high spatial and spectral studies exploring their potential in deriving
resolution data over large areas. accurate biomass among other forests studies (Cho
et al. 2012; Gleason & Im 2012; Lefsky et al. 2001a
Forest aboveground biomass estimation using & 2005; Lima et al. 2003; Weishampel et al. 1996;
active remote sensing approaches Wulder 1998).
Gleason & Im (2012) studied the effectiveness
Forest aboveground biomass estimation using of four modelling techniques, namely linear mixed-
radar data involves the use of either backscatter effects (LME) regression, random forest (RF),
values or interferometry technique (Ghasemi support vector regression (SVR), and Cubist in
2011). The backscatter values provide the most deriving forest AGB in moderately dense forest (40
convenient and accurate method for biomass - 60 % canopy closure) at both tree and plot levels
estimation as they can be compared to field using LiDAR. The results from the four methods
biomass measurement using regression analysis were almost similar at individual tree level AGB
especially in coniferous forests. Literature shows estimates (RMSE 505, 506, 457, and 502 kg tree-1).
that the HH, VV, and HV backscatter are related Furthermore, forest aboveground biomass esti-
to tree stuctural variables, such as trunk and mation accuracy improved when modelled at the
crown biomass (Beaudoin et al. 1994; Ghasemi plot level, whereas support vector regression (SVR)
2011; Santoro et al. 2006). Although this approach produced the most accurate biomass model (RMSE
 DUBE et al. 129

671 kg per 380 m2 plot when forest plots were (2012) the proportion of most tree AGB
modelled as a collection of trees). Kaiguang et al. components vary strongly with stand age, and this
(2009) investigated the effectiveness of canopy introduces errors especially when age dependent
height distributions (CHD) and canopy height expansion factors are not available for the
quantile functions (CHQ) as LiDAR metrics for conversion. For example, Dovey (2009) constructed
estimating AGB as well as the capability of simple multipliers, and later used them together
LiDAR for mapping AGB at a range of scales. with merchantable plantation timber volumes to
Their results demonstrated that the models can estimate AGB for different Eucalyptus spp., Pinus
accurately predict forest AGB and yield consistent patula and Acacia mearnsii components. These
predictive performances across a variety of scales limitations, therefore, provide opportunities for
with an r-square ranging from 0.80 to 0.95 (RMSE studies that will estimate AGB in South Africa
ranging from 14.3 to 33.7 Mg ha-1) among all the with the continuous availability of remote sensing
fitted models. A meta-analysis of more than 70 data and advanced techniques for estimating
studies on terrestrial AGB estimation using AGB.
LiDAR remote sensing by Zolkos et al. (2013)
indicated that (i) biomass models developed from Conclusions
airborne LiDAR metrics are significantly more
accurate than those using radar or passive optical The present study has reviewed the various
data alone, (ii) biomass models developed from remote sensing applications in AGB monitoring,
multi-sensor metrics were more variable than emphasizing the limitations and the prospects
LiDAR-only models and they did not improve linked to these techniques. The review established
biomass estimates, (iii) model accuracies varied that remote sensing applications provide a
with forest type, and the underlying causes for the plausible AGB estimates when compared to
observed differences were multi-faceted, and (iv) labour-intensive, costly, and time consuming
relative to the magnitude of field biomass, model traditional techniques. We, therefore, conclude
errors declined with increasing plot size. that more research is needed on the application of
remote sensing for estimating the AGB, especially
Successes and limitations of forest AGB in developing world in order to meet the Kyoto
estimation in South Africa Protocol objectives.
Few attempts have been made in estimating
AGB in South African forests especially on
Acknowledgements
indigenous forests and partly on plantation forests
Authors would like to thank the anonymous
using field inventory data (Christie & Scholes
reviewers.
1995; Dube et al. 2014a, 2014b, 2015, 2016; Dube
& Mutanga 2015a, 2015b; Dovey 2009; Schönau &
Boden 1982) with limited remote sensing appli-
References
cations. The limited number of studies on AGB
Anderson, J. E., L. C. Plourde, M. E. Martin, B. H.
estimation are as a result of inadequate number of
Braswell, M. L. Smith, R. O. Dubayah, M. A. Hofton
allometric equations for tree species in Sub-
& J. B. Blair. 2008. Integrating waveform lidar with
Saharan Africa and South Africa in particular. For
hyperspectral imagery for inventory of a northern
example, a close analysis of Chave et al. (2005)
temperate forest. Remote Sensing of Environment
study clearly indicated that none of the trees from
112: 1856-1870.
African forests were used to develop allometric
Basuki, T. M., A. K. Skidmore, Y. A. Hussin & [Link]
equations. Zianis & Mencuccini (2004) reported ca.
Duren. 2013. Estimating tropical forest biomass
279 allometric equations from all continents,
more accurately by integrating ALOS PALSAR and
except for Africa. Therefore, this clearly demon-
Landsat-7 ETM+ data. International Journal of
strates that little has been done to understand
Remote Sensing 34: 4871-4888.
AGB and carbon stocks in Sub-Saharan Africa in
Beaudoin, A., T. Le Toan, S. Goze, E. Nezry, A. Lopes,
general and South Africa in particular. Available
E. Mougin, C. C. Hsu, H. C. Han, J. A. Kong & R. T.
forest AGB studies in South Africa mostly utilized
Shin. 1994. Retrieval of forest biomass from SAR
merchantable volume together with expansion
data. International Journal of Remote Sensing 15:
factors and this has been identified as a critical
2777-2796.
source of uncertainties. According to Tsui et al.
Chave, J., R. Condit, H. C. Muller-Landau, S. C. Thomas,
130 REMOTE SENSING OF ABOVEGROUND FOREST BIOMASS

P. S. Ashton, S. Bunyavejchewin, J. K. Zimmerman Aquatic Science 39: 89-95.

& E. C. Losos. 2008. Assessing evidence for a Dube, T., O. Mutanga, A. Elhadi & R. Ismail. 2014b.
pervasive alteration in tropical tree communities. Intra-and-inter species biomass prediction in a
PLoS Biol 6: e45. plantation forest: testing the utility of high spatial
Chave, J., C. Andalo, S. Brown, M. Cairns, J. Chambers, resolution spaceborne multispectral RapidEye
D. Eamus, H. Folster, F. Fromard, N. Higuchi & T. sensor and advanced machine learning algorithms.
Kira. 2005. Tree allometry and improved estimation Sensors 14: 15348-15370.
of carbon stocks and balance in tropical forests. Dube, T. & O. Mutanga. 2015a. Evaluating the utility of
Oecologia 145: 87-99. the medium-spatial resolution Landsat-8 multi-
Chinembiri, T. S., M. C. Bronsveld, D. G. Rossiter & T. spectral sensor in quantifying aboveground biomass
Dube. 2013. The precision of C stock estimation in in uMgeny catchment, South [Link] Journal
the Ludhikola watershed using model-based and of Photogrammetry & Remote Sensing 101: 36-46.
design-based approaches. Natural Resources Research Dube, T. & O. Mutanga. 2015b. Investigating the
22: 297-309. robustness of the new Landsat-8 Operational 601
Cho, M. A., R. Mathieu, G. P. Asner, L. Naidoo, J. V. Land Imager derived texture metrics in estimating
Aardt, A. Ramoelo, P. Debba, K. Wessels, R. Main, I. plantation forest aboveground biomass in resource
P. J. Smit & B. Erasmus. 2012. Mapping tree species constrained areas. ISPRS Journal of Photo-
composition in South African savannas using an grammetry and Remote Sensing 108: 12-32.
integrated airborne spectral and LiDAR system. Dube, T., O. Mutanga, E. Abdel-Rehman, R. Ismail & R.
Remote Sensing of Environment 125: 214-226. Slotow. 2015c. Predicting eucalyptus spp. stand
Cho, M., P. Debba, R. Mathieu, L. Naidoo, J. Van Aardt volume in Zululand, South Africa: an analysis using
& G. Asner. 2010. Parameters monitoring of the a stochastic gradient boosting regression ensemble
enviroment and retrieval of biogeophysical with multi-source data sets. International Journal
improving discrimination of savnna tree species of Remote Sensing 36: 3751-3772.
through a multiple-endmember spectral angle Dube, T., O. Mutanga & R. Ismail. 2016. Quantifying
mappe approach : canopy-level analysis. IEEE aboveground biomass in African environments: A
Transactions on Geoscience Remote Sensing 48: review of the trade-offs between sensor estimation
4133. accuracy and [Link] Ecology 57(3).
Christie, S. I. & R. J. Scholes. 1995. Carbon storage in Dube, T. & O. Mutanga. In press. Quantifying the
eucalyptus and pine plantations in South Africa. variability and allocation patterns of aboveground
Environmental Monitoring and Assessment 38: carbon stocks across plantation forest types,
231-241. structural attributes and age in sub-tropical coastal
Clark, M. L., D. A. Roberts, J. J. Ewel & D. B Clark. region of Kwa Zulu Natal, South Africa using
2011. Estimation of tropical rain forest aboveground remote [Link] of Applied Geography.
biomass with small-footprint lidar and hyper Field, C. B., M. J. Behrenfeld, J. T. Randerson & P.
spectral sensors. Remote Sensing of Environment Falkowski. 1998. Primary production of the
115: 2931-2942. biosphere: Integrating terrestrial and oceanic com-
Cryus, S. & K. Tanja. 2004. Biomass estimation using ponents. Science 281: 237-240.
Landsat-TM and -ETM+. Towards a regional model Ghasemi N., M. R. Sahebi & A Mohammadzadeh. 2011.
for Southern Africa? Geo Journal 59: 177-187. A review on biomass estimation methods
Cutler, M. E. J., D. S. Boyd, G. M. Foody & A. Vetrivel. using synthetic aperture radar data. International
2012. Estimating tropical forest biomass with a Journal of Geomatics and Geosciences 1: 776-788.
combination of SAR image texture and Landsat TM Gleason, C. J. & J. Im. 2012. Forest biomass estimation
data: An assessment of predictions between regions. from airborne LiDAR data using machine learning
ISPRS Journal of Photogrammetry and Remote approaches. Remote Sensing of Environment 125:
Sensing 70: 66-77. 80-91.
Dovey, S. B. 2009. Estimating biomass and macro- Goodenough, D. G., K. O. Niemann, A. Dyk, G. Hobart,
nutrient content of some commercially important P. Gordon, M. Loisel & C. Hao. 2008. Comparison of
plantation species in South Africa. Southern Forests AVIRIS and AISA airborne hyperspectral sensing
71: 245-251. for above-ground forest carbon mapping. Geo-
Dube, T., W. Gumindoga & M. Chawira. 2014a. Detection science and Remote Sensing Symposium, 2008.
of land cover changes around Lake Mutirikwi, IGARSS 2008. IEEE International, IEEE, [Link]-
Zimbabwe, based on traditional remote sensing 129-II-132.
image classification techniques. African Journal of Henry, M., N. Picard, C. Trotta, R. J. Manlay, R.
 DUBE et al. 131

Valentini, M. Bernoux & L. Saint-André. 2011. productivity. Remote Sensing of Environment 95:
Estimating tree biomass of sub-Saharan African 549-558.
forests: a review of available allometric equations. Li, M., Y. Tan, J. Pan & S. Peng. 2008. Modeling forest
Silva Fennica 45: 477-569. aboveground biomass by combining spectrum,
Hongrui, R., Z. Guangsheng & Z. Xinshi. 2011. textures and topographic features. Frontiers of
Estimation of green aboveground biomass of desert Forestry in China 3: 10-15.
steppe in Inner Mongolia based on red-edge Lima, K., P. Treitz, M. Wulder, B. St-Onge & M. Flood.
reflectance curve area method. Biosystems Engi- 2003. LiDAR remote sensing of forest structure.
neering 109: 385-395. Progress in Physical Geography 27: 88-106.
Hyyppä, J., H. Hyyppä, M. Inkinen, M. Engdahl, S. Lu, D. 2006. The potential and challenge of remote
Linko & Y. H. Zhu. 2000. Accuracy comparison of sensing‐based biomass estimation. International
various remote sensing data sources in the retrieval Journal of Remote Sensing 27: 1297-1328.
of forest stand attributes. Forest Ecology and Lu, D., Q. Chen, G. Wang, E. Moran, M. Batistella, M.
Management 128: 109-120. Zhang, G. Vaglio Laurin & D. Saah. 2012.
Joshi, H. G. & M. Ghose. 2014. Community structure, Aboveground forest biomas estimation with Landsat
species diversity, and aboveground biomass of the and LIDAR date and uncertanity analysis of the
Sundarbans mangrove [Link] Ecology 55: estimates. International Journal of Forestry
283-303. Research 2012. 16 pages.
Kaiguang, Z., P. Sorin & N. Ross. 2009. Lidar remote Lu, X. T., J. X. Yin, M. R. Jepsen & J. W. Tang. 2010.
sensing of forest biomass: A scale-invariant Ecosystem carbon storage and partioning in a
estimation approach using airborne lasers. Remote tropical seasonal forest in Southwestern China.
Sensing of Environment 113: 182-196. Forest Ecology and Management 260: 1798-1803.
Kajisa, T., T. Murakami, N. Mizoue, N. Top & S. Luther, J. E., R. A. Fournier, D. E. Piercey, L. Guindon
Yoshida. 2009. Object-based forest biomass esti- & R. J. Hall. 2006. Biomass mapping using forest
mation using Landsat ETM+ in Kampong Thom type and structure derived from Landsat TM
Province, Cambodia. Journal of Forest Research 14: imagery. International Journal of Applied Earth
203-211. Observation and Geoinformation 8: 173-187.
Ketterings, Q. M., R. Coe, M. van Noordwijk, Y. Min, L., J. J. Qu & H. Xianjun. 2009. Estimating
Ambagau’ & C. A. Palm. 2001. Reducing un- aboveground biomass for different forest types
certainty in the use of allometric biomass equations based on Landsat TM measurements, Geoin-
for predicting aboveground tree biomass in mixed tormatics, 2009. 17th International Conference on
secondary forests. Forest Ecology and Management Geoinformatics. pp. 1-6. Proceeding of a meeting
146: 199-209. held on 12-14 August 2009, Fairfax. Virginia.
Koch, B. 2010. Status and future of laser scanning, Mitchard, E. T. A., S. S. Saatchi, I. H. Woodhouse, G.
synthetic aperture radar and hyperspectral remote Nangendo, N. S. Ribeiro, M. Williams, C. M. Ryan,
sensing data for forest biomass assessment. ISPRS S. L. Lewis, T. R. Feldpausch & P. Meir. 2009.
Journal of Photogrammetry and Remote Sensing 65: Using satellite radar backscatter to predict above-
581-590. ground woody biomass: A consistent relationship
le Maire, G., C. François, K. Soudani, D. Berveiller, J. Y. across four different African landscapes. Geo-
Pontailler, N. Bréda, H. Genet, H. Davi & E. physical Research Letters 36: L23401.
Dufrêne. 2008. Calibration and validation of hyper- Mutanga, O., E. Adam & M. A. Cho. 2012. High density
spectral indices for the estimation of broadleaved biomass estimation for wetland vegetation using
forest leaf chlorophyll content, leaf mass per area, WorldView-2 imagery and random forest regression
leaf area index and leaf canopy biomass. Remote algorithm. International Journal of Applied Earth
Sensing of Environment 112: 3846-3864. Observation and Geoinformation 18: 399-406.
Lefsky, M. A., W. Cohen & T. Spies. 2001. An evaluation Muukkonen, P. & J. Heiskanen. 2007. Biomass esti-
of alternate remote sensing products for forest mation over a large area based on standwise forest
inventory, monitoring, and mapping of Douglas-fir inventory data and ASTER and MODIS satellite
forests in western Oregon. Canadian Journal of data: A possibility to verify carbon inventories.
Forest Research 31: 78-87. Remote Sensing of Environment 107: 617-624.
Lefsky, M. A., D. P. Turner, M. Guzy & W. B. Cohen. Patenaude, G., M. Ronald & P. D. Terence. 2005. Syn-
2005. Combining lidar estimates of aboveground thesis of remote sensing approaches for forest carbon
biomass and Landsat estimates of stand age for estimation: reporting to the Kyoto Protocol.
spatially extensive validation of modeled forest Environmental Science and Policy 8: 161-178.
 132 REMOTE SENSING OF ABOVEGROUND FOREST BIOMASS

Pulliainen, J., M. Engdahl & M. Hallikainen. 2003. Swatantran, A., R. Dubayah, D. Roberts, M. Hofton & J.
Feasibility of multi-temporal interferometric SAR B. Blair. 2011. Mapping biomass and stress in the
data for stand-level estimation of boreal forest stem Sierra Nevada using lidar and hyperspectral data
volume. Remote Sensing of Environment 85: 397- fusion. Remote Sensing of Environment 115: 2917-
409. 2930.
Qisheng, H. 2012. Estimation of coniferous forest above- Treuhaft, R. N., G. P. Asner & B. E Law. 2003.
ground biomass using LiDAR and Spot-5 data, Structure-based forest biomass from fusion of radar
Remote Sensing, Environment and Transportation and hyperspectral observations. Geophysical Research
Engineering (RSETE), 2012. 2nd International Letters 30: 1472.
Conference, pp. 1-4. Tsui, O. W., N. C. Coops, M. A. Wulder, P. L. Marshall &
Qureshi, A., B. R. Pariva & S. A. Hussain.2012. A review A. McCardle. 2012. Using multi-frequency radar
of protocols used for assessment of carbon stock in and discrete-return LiDAR measurements to
forested landscapes. Ennvironmental Science and estimate aboveground biomass and biomass
Policy 16: 81-89. components in a coastal temperate forest. ISPRS
Santoro, M., L. Eriksson, J. Askne & C. Schmullius. Journal of Photogrammetry and Remote Sensing 69:
2006. Assessment of stand‐wise stem volume 121-133.
retrieval in boreal forest from JERS‐1 L‐band SAR Weishampel, J. F., K. J. Ranson & D. J. Harding. 1996.
backscatter. International Journal of Remote Remote sensing of forest canopies. Selbyana 17:
Sensing 27: 3425-3454. 6-14.
Schönau, A. & D. L. Boden. 1982. Preliminary biomass Wulder, M. 1998. Optical remote-sensing technique for
studies in young eucalypts. South African Forestry the assessment of forest inventory and biophysical
Journal 120: 24-28. parameters. Progress in Physical Geography 22:
Segura, M. & M. Kanninen. 2005. Allometric models for 449-476.
tree volume and total aboveground biomass in a Zheng, D., J. Rademacher, J. Chen, T. Crow, M. Bresee,
tropical humid forest in Costa Rica1. Biotropica 37: J. Le Moine & S. R. Ryu. 2004. Estimating above-
2-8. ground biomass using Landsat 7 ETM+ data across
Steininger, M. 2000. Satellite estimation of tropical a managed landscape in northern Wisconsin, USA.
secondary forest above-ground biomass data from Remote Sensing of Environment 93: 402-411.
Brazil and Bolivia. International Journal of Remote Zianis, D. & M. Mencuccini. 2004. On simplifying allo-
Sensing 21: 1139-1157. metric analyses of forest biomass. Forest Ecology
Straub, C., M. Dees, H. Weinacker & B. Koch. 2009. and Manegement 187: 311-332.
Using airborne laser scanner data and CIR Zolkos, S. G., S. J. Goetz & R. Dubayah. 2013. A meta-
orthophotos to estimate the stem volume of forest analysis of terrestrial aboveground biomass esti-
stands. PFG Photogrammetrie, Fernerkundung, mation using lidar remote sensing. Remote Sensing
Geoinformation 2009: 277-287. of Environment 128: 289-298.

(Received on 29.04.2014 and accepted after revisions, on 04.10.2014)

View publication stats