Theoretical and Applied Climatology (2019) 135:463–472

[Link]

ORIGINAL PAPER

Factors affecting yearly and monthly visits to Taipei Zoo

Ai-Tsen Su 1 & Yann-Jou Lin 2

Received: 22 March 2016 / Accepted: 18 January 2018 / Published online: 1 February 2018
# Springer-Verlag GmbH Austria, part of Springer Nature 2018

Abstract
This study investigated factors affecting yearly and monthly numbers of visits to Taipei Zoo. Both linear and nonlinear regression
models were used to estimate yearly visits. The results of both models showed that the Bopening effect^ and Banimal star effect^
had a significantly positive effect on yearly visits, while a SARS outbreak had a negative effect. The number of years had a
significant influence on yearly visits. Results showed that the nonlinear model had better explanatory power and fitted the
variations of visits better. Results of monthly model showed that monthly visits were significantly influenced by time fluctua-
tions, weather conditions, and the animal star effect. Chinese New Year, summer vacation, numbers of holidays, and animal star
exhibitions increased the number of monthly visits, while the number of days with temperatures at or below 15 °C, the number of
days with temperatures at or above 30 °C, and the number of rainy days had significantly negative effects. Furthermore, the
model of monthly visits showed that the animal star effect could last for over two quarters. The results of this study clarify the
factors affecting visits to an outdoor recreation site and confirm the importance of meteorological factors to recreation use.

1 Introduction visitors annually. Its visitor numbers fluctuate over different

time periods as many recreation sites. However, zoos have
Number of visitors is important for recreation planning and different characteristics than other recreation sites. There are
management. Therefore, recreation demand analysis is an im- many kinds of attractive animals kept in the outdoor display
portant research topic, and many studies have explored the areas. Visitors have to spend more time staying outdoors for
influence of various factors on variation in recreational visits. watching the animals in the zoos, so the factors affecting their
Taipei Zoo is Taiwan’s most famous zoo, attracting millions of visitor numbers may thus differ as well. This study aimed to
analyze the factors that affect yearly and monthly visits fluc-
tuations of Taipei Zoo and to investigate the relative effects of
Study highlights each factor.
• Weather conditions influence visitor numbers to the zoo. High numbers
of days with excessively high or low temperatures or rainfall lower the Temporal fluctuations in number of visits are common
number of monthly visits. (Ritchie and Beliveau 1974, Butler 1980). Many studies ex-
• The number of visits to the zoo is significantly affected by the animal amined the temporal variations in number of visitors to scenic
star effect. Exhibition of rare animals can attract a large number of areas (Li and Lin 2012), urban forests (Dwyer 1988), and
visitors to the zoo, and this effect can last for up to two quarters.
• Pandemics of contagious disease such as SARS can cause a significant beaches (Finger and Lehmann 2012, Marvasti 2013).
drop in the number of visits to the zoo. Periodic fluctuation in visits can be found at many recreation
• The zoo presents a significant opening effect. A large number of visitors sites (Ritchie and Beliveau 1974) and include seasonal
appear in the opening year. (Bartczak et al. 2012, Li and Lin 2012, Marvasti 2013),
• There are more visitors during holidays than weekdays, and the number
of holidays in a month is positively correlated with visitor numbers. monthly (Finger and Lehmann 2012, Aylen et al. 2014), and
weekly fluctuations (Dwyer 1988, Arnberger and
* Yann-Jou Lin Hinterberger 2003, Ploner and Brandenburg 2003,
yannjlin@[Link] Arnberger 2006, Arnberger and Eder 2007, Finger and
Ai-Tsen Su
Lehmann 2012, Marvasti 2013, Su et al. 2014). For example,
su@[Link] the number of visitors to beaches is higher during summer
holidays (Finger and Lehmann 2012, Marvasti 2013), while
1
Department of Social and Regional Development, National Taipei in autumn, more people engage in outdoor hiking (Li and Lin
University of Education, 134, Sec. 2, Heping E. Rd, Taipei, Taiwan 2012). Nonperiodic fluctuations include special events and
2
Department of Horticulture and Landscape Architecture, National holidays (Siderelis and Moore 2011, Finger and Lehmann
Taiwan University, 138, Sec. 4, Keelung Rd, Taipei, Taiwan
464 A.-T. Su, Y.-J. Lin

2012, Su et al. 2014). Visitor numbers are higher on long in Asia had an immense impact on tourism in all affected
weekends, national holidays, or the Taiwanese Lunar New countries, including Taiwan (Wang 2009, Mao et al. 2010,
Year than on normal days (Finger and Lehmann 2012, Su Min et al. 2011), China (Beutels et al. 2009, Yang and Wong
et al. 2014). To analyze the effects of weekends, holidays, 2012), and Cambodia (Tang and Wong 2009). Studies by Kuo
and special events on monthly visits, the number or ratio of et al. (2008) and McAleer et al. (2010) showed that SARS and
holidays in each given month is used (Aylen et al. 2014). Avian Flu both had long- and short-term effects on tourism.
Visitor numbers are also affected by weather conditions The 2009 H1N1 outbreak also affected the travel intentions of
(Dwyer 1988, Ploner and Brandenburg 2003, Finger and Koreans (Lee et al. 2012) and led to a loss of one million
Lehmann 2012, Li and Lin 2012, Marvasti 2013, Su et al. overseas visitors to Mexico (Rassy and Smith 2013). These
2014). Previous studies demonstrated that participation of out- studies have all highlighted the fact that outbreaks of serious
door recreation activities is influenced by weather conditions, contagious disease are an important factor influencing travel
including air temperature, sunshine, rainfall, snowfall, and intentions and number of visitors to recreational areas.
wind field (Dwyer 1988, Ploner and Brandenburg 2003, Animal exhibitions are attractions specific to zoos. In the
Scott et al. 2008, Shih et al. 2009, Finger and Lehmann motion picture studies, the participation of movie stars posi-
2012, Li and Lin 2012, Price et al. 2012, Becken 2013, tively affected the movies’ revenues (Elberse 2007). Movie
Marvasti 2013, Aylen et al. 2014, King et al. 2014, Su et al. stars can enhance box office receipts during the opening week
2014, Hewer et al. 2015). and increase the public’s anticipation (Karniouchina 2011).
Different indicators of air temperature have been used in The phenomenon can be recognized as the star power (Liu
previous studies to examine its effect on visitor numbers. et al. 2014). Some evidence shows that the presence of animal
Maximum temperature is the most commonly used indicator. stars can similarly attract large numbers of visitors, such as the
When temperature rises, visitor numbers increase accordingly pandas at the San Diego Zoo and Edinburgh Zoo, the gorilla at
(Dwyer 1988, Ploner and Brandenburg 2003, Finger and Higashiyama Zoo, San Diego Zoo and Knoxville Zoo, and
Lehmann 2012, Becken 2013, Aylen et al. 2014, Su et al. meerkat pups at Edinburgh Zoo. In Taiwan, the first panda
2014). However, these studies have shown that the relation- cub of Taipei Zoo attracted record crowds to the Panda Hall
ship is nonlinear. Once the temperature either exceeds or falls during the cub’s first weekend. Although the presence of an-
below a certain point, visitor numbers begin to decline. For imal stars at zoos often attracts more visitors, this effect has
example, the number of visitors to Chester Zoo increases as not yet been verified by empirical research.
the temperature rises but begins to decrease once the temper- In summary, in addition to temporal fluctuations, number
ature exceeds 21 °C (Aylen et al. 2014), while temperatures of visitors to zoos is affected by weather conditions, outbreak
above 23 °C have a negative influence on visitor number at of contagious disease, and the animal star effect in particular.
two lidos in Zurich (Finger and Lehmann 2012). It is evident Therefore, the purpose of this study is to analyze in-depth the
that excessively high or low temperature has a negative effect effects of temporal fluctuations, weather conditions, and out-
on visitor numbers. Therefore, the number of days with tem- break of contagious disease on the yearly and monthly num-
peratures higher or lower than a specific value is used in some bers of visitors to Taipei Zoo. In addition, this study also
studies analyzing its effect on monthly visitor numbers. For further investigates the effects and duration of effects of the
example, if the number of days with temperatures below 10 °C zoo’s opening and animal star displays.
increases in Kuanyin Mountain, Taiwan, the hiker numbers
decrease (Li and Lin 2012).
The effect of sunshine on visitor numbers is often positive 2 Study site
and measured using sunshine hours (Ploner and Brandenburg
2003, Becken 2013, Su et al. 2014, Creemers et al. 2015) or Taipei Zoo is located in the southeastern corner of Taipei (24°
the percentage of sunny days (Dwyer 1988). Rainfall, con- 59′ 56″ N, 121° 34′ 52″ E). In September 1986, it was moved
versely, has a negative effect on visitor numbers. Daily rainfall to the present site in Mucha and was officially opened in 1987.
(Becken 2013, Aylen et al. 2014), hours of rainfall (Su et al. Taipei Zoo covers an area of 165 ha. There are eight existing
2014), percentage of rainy days (Dwyer 1988), and presence outdoor display areas and six indoor display areas.
of rainfall (Finger and Lehmann 2012, Marvasti 2013) are Transportation to Taipei Zoo is very convenient from all dis-
often used for analyzing the influence of rainfall on daily tricts of Taipei, and a nearby highway provides direct access to
visits. When investigating fluctuations in monthly visitor other cities and counties. Thus, the zoo is highly accessible.
numbers, the number of days of rainfall (Li and Lin 2012) The average air temperature in Taipei from 2005 to 2014
and the amount of rainfall (Becken 2013, Aylen et al. 2014) was 23.3 °C. The average number of days per month with a
have both been used as explanatory variables. maximum temperature lower than or equal to 15 °C was
Natural disasters and outbreaks of contagious disease can 1.1 days, and the average number of days per month with a
also influence travel intentions. The 2003 outbreak of SARS maximum temperature higher than or equal to 30 °C was
Factors affecting yearly and monthly visits to Taipei Zoo 465

11.8 days. Average monthly sunshine duration was 117.0 h, explanatory variables included time periods, important events,
and the average percentage of sunshine was 31.45%. Average and weather conditions. The processing of these variables is
monthly precipitation was 205.1 mm, average hours of rainfall shown in Table 1. Time period included the number of years
were 113.0 h, and average number of rainy days was and the number of years squared. Important events included
13.8 days. Average monthly wind speed was 2.5 m/s, and the opening of the zoo, the post-SARS period, and the animal
the average maximum wind speed was 18.0 m/s. star effect. The variables following the opening of the zoo
included the opening year (1987) and the first and second year
after opening (1988 and 1989). The post-SARS period was
3 Modeling the yearly visits from 2003 (the outbreak) to 2014. The animal star effect re-
ferred to the number of species debuting within a year follow-
3.1 Fluctuation in yearly visits ing the first exhibition. If the debuts took place during the
second half of the year, the following year was also treated
Since moving to its present location, Taipei Zoo has had sev- as being within the period of the animal star effect. Animal
eral million visitors each year. For the first 3 years after open- stars included koalas, king penguins, giant pandas, and giant
ing, yearly visitor numbers were over three million. The low- panda cubs, which debuted in 1999, 2000, 2009, and 2014,
est number of yearly visitors occurred in 1990, which was respectively. The koalas and king penguins were both
followed by a steady growth in visitor numbers. When displayed in the second half of the year; therefore, the year
Taipei Zoo first moved to its new location, its most well- 2000 was taken as having an animal star effect of 2, as it
known animal star was Lin Wang the elephant. After Lin benefitted from the effect of both koalas and king penguins.
Wang, a new generation of animal stars followed. For exam- The years 1999, 2001, 2009, and 2014 each had an animal star
ple, in 1999, the koalas Harley and Patrick led an upsurge in effect of 1, while all the other years had an effect of 0. To
visits; in 2000, the king penguins saw visitor numbers rise to a reflect the potential lags, two lagged variables, including the
new record high, and the following year visitor numbers were visits lagged by 1 and 2 years, were considered. Weather data
just below the number of visits seen in the opening year. Each was obtained from the Taipei Meteorological Station of the
animal star gave rise to a new topic and facilitated high visitor Central Weather Bureau (25° 02′ 23″ N, 121° 30′ 24″ E; ele-
numbers. The 2003 SARS outbreak caused a sharp decline in vation 5.3 m) and included average air temperature, sunshine
visitor numbers; this was followed by a slow increase. The duration (including number of sunshine hours and percentage
2009 debut of the giant pandas Tuan Tuan and Yuan Yuan of sunshine), rainfall (including sum of precipitation, hours of
caused a sudden, small increase in visits. In 2014, the giant rainfall, and number of rainy days), and wind field (including
panda cub Yuan Zai caused another peak in visitor numbers average wind speed and maximum wind speed).
(Fig. 1).
3.3 Effect of events on yearly visits
3.2 Measurement of variables
The fluctuation in yearly visits showed more visitors during
In the modeling of yearly visits, the dependent variable was the opening year and the following 2 years after opening. A
the number of visitors each year provided by Taipei Zoo. The one-way analysis of variance (ANOVA) is used to test the

Fig. 1 Comparison of the yearly

model results vs. Taipei Zoo
visitor data
466 A.-T. Su, Y.-J. Lin

Table 1 Variables used in

modeling yearly visits to Taipei Variable Details
Zoo
Data of visits Yearly visits Yearly visits from 1987 to 2014
Date information Years Numbers of years since 1987
Event information Opening year 1 = opening year; 0 = other years
1st and 2nd year after 1 = 1st and 2nd year after opening; 0 = other years
opening
After SARS 1 = the years after SARS; 0 = other years
Animal star effect 1 = one species debuting; 2 = two species debuting; 0 =
other years
Meteorological Average temp. Yearly average air temperature (°C)
information Sunshine hours Sum of sunshine hours in a year
Percentage of sunshine Percentage of sunshine possible for a year
Precipitation Sum of precipitation in a year (mm)
Rainfall hour Rainfall in a year (hours)
Days of precipitation Days of precipitation in a year
Average wind speed Yearly average wind speed (m/s)
Maximum wind speed Maximum wind speed for a year (m/s)

hypothesis that several means are equal (SPSS Inc. 2007). The variables for regression analysis of yearly visits. The ordinary
F value in ANOVA is a ratio of variation between sample least squares method was employed to estimate the relation-
means to variation within the samples. The results showed a ships among variables. The opening year and the first and
significant difference in visits among the opening year and the second years after opening were set as dummy variables,
first and second year after opening and other years (F = 2.732, and all other years were used as reference categories. The
p ≤ .10). To compare the visits before/after SARS and the years after SARS were also set as dummy variables, while
visits with/without animal stars debut, the t tests were used. all other years were used as reference categories. Results in-
The t test is used to compare means for two groups of samples dicated that the number of years, opening year, first and sec-
(SPSS Inc. 2007). The t value is the calculated difference ond years after opening, and animal star effect all had a sig-
represented in units of standard error. The results suggested nificantly positive influence on yearly visits, while the post-
that there were less visitors after SARS than those before SARS period had a negative influence on yearly visits. The
SARS (t = 2.280, p ≤ .05). However, the numbers of visitors linear model established the statistical relationships between
were significantly higher in the years with animal stars debut yearly visit and a set of explanatory variables as follows:
than those without animal stars debut (t = − 4.338, p ≤ .001).
YVisit ¼ 2684281 þ 79322Year þ 2762944Open
3.4 Effect of weather conditions on yearly visits þ 824235AfterO–1412837SARS

Correlational analysis showed that all the variables of weather þ 1021449Animal

conditions, including average air temperature (r = .06,
where YVisit is the yearly visit, Year is the numbers of years
p > .05), sunshine hours (r = − .31, p > .05), percentage of
since 1987, Open is the dummy variable of opening year,
sunshine (r = − .28, p > .05), precipitation (r = .10, p > .05),
AfterO is the dummy variable of first and second year after
rainfall hour (r = .08, p > .05), average wind speed (r = − .20,
opening, SARS is the years after SARS, and Animal is the
p > .05), and maximum wind speed (r = − .01, p > .05), did not
animal star effect. The number of visitors increased by
show significant correlation with yearly visits.
79,322 each year; the opening year had an additional increase
of 2,762,944 visitors, and the 2 years after opening had an
3.5 Linear models for predicting yearly visits additional increase of 824,235 visitors compared with
nonopening years. When animal stars debuted, there was an
The linear regression is used to estimate the coefficients of the increase of 1,021,449 visitors for each additional species of
linear equation that predict the value of the dependent variable animal star. After the SARS outbreak, visitor numbers de-
from one or more independent variables (SPSS Inc. 2007). creased by 1,412,837 per year compared with the visits before
The linear model used the number of years, the opening year, the SARS outbreak. The linear model explained 83.9% of the
the first and second years after opening, the post-SARS peri- variance (Table 2). By comparing the estimated yearly visits
od, the lagged visits, and the animal star effect as explanatory
Factors affecting yearly and monthly visits to Taipei Zoo 467

of the linear model with the actual visits, the linear model reflect the effects of zoo openings, SARS, and animal star
results were shown to be consistent with the trend of growth debuts on fluctuations in yearly visits (Fig. 1), and the explan-
in visitor numbers as the years increased and able to reflect the atory power of the nonlinear model was higher than that of the
increase in visits created by animal star debuts (Fig. 1). linear model.

3.6 Nonlinear models for predicting yearly visits 4 Modeling the monthly visits
To examine whether yearly visits followed a nonlinear trend 4.1 Fluctuation in monthly visits
of fluctuation, number of years squared (Years2) was added as
an independent variable to the linear model in regression anal- Although overall yearly visits to Taipei Zoo are subject to
ysis of yearly visits. Results indicated that the opening year fluctuation, a detailed analysis of monthly visits over the last
and the first and second years after opening all had significant- decade revealed that within a given year, the month with the
ly positive effects on yearly visits, whereas the post-SARS highest number of visits was usually February, coinciding
period had a significantly negative effect. The nonlinear func- with winter vacations and the Taiwanese Lunar New Year.
tion of yearly visit could be described as: June generally had low visitor numbers, whereas visits in-
YVisit ¼ 1499698 þ 267645Year−5889Year2 creased in July and August (summer vacations), showing a
clear relationship with school holidays. Number of visitors
þ 3765093Open þ 1576292AfterO–1399092SARS decreased again in September (Fig. 2).
þ 932846Animal
4.2 Measurement of variables
2
where Year is the square numbers of years since 1987. The
number of years and number of years squared both had sig- The regression models of monthly visits used monthly
nificant effects on yearly visits. Compared with nonopening visits of the past 10 years as the dependent variable.
years, the opening year had an additional 3,765,093 visitors, The explanatory variables included time period, special
while the first and second years after opening had an addition- events, and weather conditions. The data concerning
al 1,576,292. The animal star effect had a significantly posi- monthly visits were recorded by ticket sales between
tive influence on yearly visits. When animal stars debuted, 2005 and 2014. Time period variables included the
there was an increase of approximately 932,846 visitors for month of Chinese New Year (CNY), summer vacation,
each additional species of animal star. After the SARS out- and number of holidays. The CNY month referred to
break, yearly visits decreased by 1,399,092 compared with the the month during which the CNY holiday occurred;
numbers before the SARS outbreak. The explanatory power summer vacation referred to July and August; and total
of the nonlinear regression model was 92.3% (Table 2). By number of holidays included all regular holidays and
comparing the nonlinear model to the actual visitor numbers, national holidays. There were two debuts of animal stars
the nonlinear model was shown to be able to appropriately in these 10 years, so special events included the four

Table 2 Models of yearly visits

to Taipei Zoo Linear model Nonlinear model

B VIF B VIF

(Constant) 2,684,281 *** 1,499,698 ***

Years 79,322 ** 7.299 267,645 *** 42.813
Years2 – − 5889 *** 31.015
Opening year 2,762,944 *** 1.296 3,765,093 *** 1.827
1st and 2nd years after opening 824,235 * 1.407 1,576,292*** 1.983
After SARS − 1,412,837 *** 5.737 − 1,399,092*** 5.738
Animal star effect 1,021,449 *** 1.357 932,846*** 1.386
R2 .839 .923
Adjusted R2 .803 .902

B regression coefficient, VIF variance inflation factor, R2 coefficient of determination, Adjusted R2 R-squared that
has been adjusted for the number of predictors in the model
*p ≤ .05; **p ≤ .01; ***p ≤ .001
468 A.-T. Su, Y.-J. Lin

Fig. 2 The monthly visits to

Taipei Zoo from 2005 to 2014

quarters of 2009 (giant panda debut) and 2014 (giant temperature at or below 15 °C and number of days with
panda cub debut). Weather data was also obtained from a maximum temperature at or above 30 °C was used as
the Taipei Meteorological Station and included air tem- explanatory variables for regression analysis. Results
perature (including average temperature, number of days showed that both variables had a negative influence on
with maximum temperature at or below 15 °C, and visitor numbers. Correlational analyses showed that the
number of days with maximum temperature at or above sum of precipitation (r = − .34, p ≤ .01), hours of rainfall
30 °C), sunshine duration (including number of sun- (r = − .25, p ≤ .01), and days of precipitation (r = − .44,
shine hours and percentage of sunshine), rainfall (in- p ≤ .01) were significantly negatively correlated with
cluding sum of precipitation, hours of rainfall, and num- monthly visits, while the correlation coefficient of days
ber of rainy days), and wind field (including average of precipitation was relatively high. The number of sun-
wind speed and maximum wind speed) (Table 3). shine hours (r = .07, p < .05), percentage of sunshine
(r = .14, p < .05), average wind speed (r = .04, p < .05),
4.3 Effect of vacations on monthly visits and maximum wind speed (r = − .11, p < .05) did not
show significant correlation with monthly visits.
The fluctuation in monthly visits showed more visitors
during the CNY month and summer vacation. The vis- 4.5 Regression models for predicting monthly visits
itor number on the month of CNY or summer vacation
was analyzed. The results of a one-way ANOVA The independent variables for a regression analysis
showed a significant difference between the CNY month (SPSS Inc. 2007) of monthly visits included month cat-
and summer vacation compared with other months. The egories, number of holidays, number of days with tem-
Scheffé method is a post hoc test used for multiple perature at or below 15 °C, number of days with tem-
comparisons in ANOVA (SPSS Inc. 2007). Results from perature at or above 30 °C, days of precipitation, panda
Scheffé’s method showed that visitor numbers were sig- effects, and panda cub effects. CNY month and summer
nificantly higher during the CNY month than during vacation were set as dummy variables for monthly
summer vacation and other months. Further analysis of categories, and all other months were used as
holidays in each month showed a significantly positive reference categories. The panda and panda cub effects
correlation between the number of holidays and monthly were divided into first, second, third, and fourth
visits (r = .42, p ≤ .01). quarters and used as dummy variables, while all other
months were used as reference categories. To diagnose
4.4 Effect of weather conditions on monthly visits the presence of heteroscedasticity, Koutsoyiannis (1977)
suggests computing the Spearman rank correlation be-
To explore the effects of air temperature on monthly tween residual and explanatory variables. The low rank
visits, a quadratic function of average temperature was correlation coefficients suggested that there was no
used, and the results suggested that there was no rela- heteroscedasticity. Results showed that the CNY month,
tionship between the average temperature and monthly summer vacation, number of holidays, pandas’ first and
visits. The number of days with a maximum second quarters, and panda cub’s first and second
Factors affecting yearly and monthly visits to Taipei Zoo 469

Table 3 Variables used in

modeling monthly visits to Taipei Variable Details
Zoo
Data of visits Monthly visits Monthly visits from 2005 to 2014
Date information CNY month 1 = the month of Chinese New Year; 0 = other months
Month of summer 1 = July or August; 0 = other months
vacation
No. of holidays Number of holidays
Event information Panda Q1 1 = month during the pandas’ 1st quarter; 0 = other
months
Panda Q2 1 = month during the pandas’ 2nd quarter; 0 = other
months
Panda Q3 1 = month during the pandas’ 3rd quarter; 0 = other
months
Panda Q4 1 = month during the pandas’ 4th quarter; 0 = other
months
Panda cub Q1 1 = month during the panda cub’s 1st quarter; 0 = other
months
Panda cub Q2 1 = month during the panda cub’s 2nd quarter; 0 = other
months
Panda cub Q3 1 = month during the panda cub’s 3rd quarter; 0 = other
months
Panda cub Q4 1 = month during the panda cub’s 4th quarter; 0 = other
months
Meteorological Average temp. Monthly average air temperature (°C)
information Days of highest temp. Days of maximum temperature lower than or equal to
≤ 15 °C 15 °C
Days of highest temp. Days of maximum temperature higher than or equal to
≥ 30 °C 30 °C
Sunshine hours Sum of sunshine hours in a month
Percentage of sunshine Percentage of sunshine possible for a month
Precipitation Sum of precipitation in a month (mm)
Rainfall hour Rainfall in a month (hours)
Days of precipitation Days of precipitation in a month
Average wind speed Monthly average wind speed (m/s)
Maximum wind speed Maximum wind speed for a month (m/s)

quarters had positive influence on monthly visits, while equal to 15 °C, Temp30°C is the days of maximum tem-
the number of days with maximum temperature at or perature higher than or equal to 30 °C, Rain is the days
below 15 °C, the number of days with maximum tem- of precipitation, PandaQ1 to PandaQ4 are the months
perature at or above 30 °C, and days of precipitation all during the pandas’ first to fourth quarters, and CubQ1
had negative effects on monthly visits. The function of to CubQ4 are the months during the panda cub’s first to
monthly visit could be described as: fourth quarters. CNY month showed an increase of
66,584 visitors in comparison with other months, and
MVisit ¼ 233165 þ 66584CNY þ 110728Summer þ 14242Holiday summer vacation months increased additional 110,728
–12357Temp15°C –4071Temp30°C –4960Rain þ 82014PandaQ1 visitors. For each additional day of holidays, monthly
visits showed an increase of approximately 14,242. For
þ 61235PandaQ2 þ 26145PandaQ3−2373PandaQ4 each additional day with maximum temperature at or
þ 169429CubQ1 þ 98233CubQ2 þ 9581CubQ3 þ 21721CubQ4 below 15 °C, visitor numbers decreased by 12,357; for
each additional day with maximum temperature at or
where MVisit is the monthly visit, CNY is the month of above 30 °C, visitor numbers decreased by 4071. For
Chinese New Year, Summer is the month of summer each additional rainy day, visitor numbers decreased by
vacation, Holiday is the numbers of holidays, Temp15°C 4960. If the month was in the first or second quarter of
is the days of maximum temperature lower than or the pandas’ debut, monthly visits increased by 82,014
470 A.-T. Su, Y.-J. Lin

and 61,235, respectively; if the month was in the first enough. Further studies on appropriate parameters of weather
or second quarter of the panda cub’s debut, monthly conditions for yearly visits would be useful.
visits increased by 169,429 and 98,233, respectively.
The explanatory power of the model was 74.0% 5.2 Meteorological and other parameters for monthly
(Table 4). The model also showed that the pandas’ visits
and panda cub’s effects on monthly visits could last
for approximately two quarters. Previous studies have already shown a positive relationship
between visitor numbers and air temperature (Dwyer 1988,
Ploner and Brandenburg 2003, Finger and Lehmann 2012,
Aylen et al. 2014, Su et al. 2014), and that temperatures ex-
5 Discussion ceeding a specific level can cause a reduction in visits (Finger
and Lehmann 2012, Aylen et al. 2014). The monthly visit
5.1 Comparisons between models of yearly visits model in this study also showed that the number of days with
maximum temperature at or below 15 °C, the number of days
Both linear and nonlinear models were used to estimate yearly with maximum temperature at or above 30 °C, and days of
visits. Number of years, opening year, first and second years precipitation had negative influence on monthly visits.
after opening, post-SARS period, and animal star effect were Therefore, it is evident that excessively low or high air tem-
all used in both models. The number of years implied an peratures and rainfall all discourage visits.
integrated fluctuation of many demographic variables, includ- The model of monthly visits also considered CNY month,
ing the growth of population or per capita income. The effect summer vacation, and number of holidays as temporal vari-
of each variable was statistically significant, and the explana- ables. CNY month, summer vacations, and number of holi-
tory power of the nonlinear model was better. Therefore, we days all had a positive effect on monthly visits. The result
suggest that the nonlinear model is a better fit to the actual showing that summer vacation caused a greater increase in
variation of visits. However, the variables of weather condi- visits is consistent with research by Finger and Lehmann
tions were not correlated with yearly visits in our study. The (2012), Marvasti (2013), and Aylen et al. (2014). Studies by
reason might be that the duration of yearly data was not long Su et al. (2014) also showed a positive effect of the CNY
period on visitor numbers.
Table 4 Model of monthly visits to Taipei Zoo

Monthly model 5.3 The opening effect and animal star effect on visits

B VIF Fluctuation in visitor numbers to Taipei Zoo over the years

showed that the opening year attracted a higher number of
(Constant) 233,165 ***
visitors. The linear and nonlinear models verified the opening
CNY month 66,584 ** 2.054
effect in zoos. When the zoo moved to its current location, a
Month of summer vacation 110,728 *** 2.091
grand animal parade and a zoo-opening ceremony was also
No. of holidays 14,242 *** 1.736
held, but the official opening of the zoo did not occur imme-
Days of highest temp. ≤ 15 °C − 12,357 *** 1.620
diately. This increased public anticipation and caused a large
Days of highest temp. ≥ 30 °C − 4071 *** 2.817
surge in the numbers of visitors after the official opening. This
Days of precipitation − 4960*** 1.258
phenomenon resembles the results of delayed visit plans fol-
Panda Q1 82,014 ** 1.043
lowing zoo closures that had lasted 41 days due to foot and
Panda Q2 61,235 * 1.047
mouth disease and bad weathers at Chester Zoo in England,
Panda Q3 26,145 (n.s.) 1.097
presented by Aylen et al. (2014). In these cases, planned visits
Panda Q4 − 2373 (n.s.) 1.040
could not be implemented, and caused an accumulation of
Panda cub Q1 169,429 *** 1.126
visitor numbers, or created a compensating Bbounce-back.^
Panda cub Q2 98,233 *** 1.050
Animal exhibits distinguish zoos from other recreation
Panda cub Q3 9581 (n.s.) 1.132 sites. Modeling yearly visits verified that visitor numbers were
Panda cub Q4 21,721 (n.s.) 1.067 affected by animal star debuts, and this effect was also seen in
R2 .740 the model of monthly visits. Without the animal star effect,
Adjusted R2 .706 visits in each month generally showed a fixed fluctuation
B regression coefficient, VIF variance inflation factor, R2 coefficient of
trend. However, the giant pandas’ debut in early 2009 and
determination, Adjusted R2 R-squared that has been adjusted for the num- the giant panda cub’s debut in early 2014 both had significant-
ber of predictors in the model ly positive effects on visits in both the first and second quar-
*p ≤ .05; **p ≤ .01; ***p ≤ .001; n.s. not significant ters, with greater influence in the first quarter than in the
Factors affecting yearly and monthly visits to Taipei Zoo 471

second. The panda cub’s debut had a greater effect on visits Austria. J Nat Conserv 11(4):260–267. [Link]
1617-1381-00057
than the pandas’ debut.
Aylen J, Albertson K, Cavan G (2014) The impact of weather and climate
on tourist demand: the case of Chester Zoo. Clim Chang 127(2):
183–197. [Link]
5.4 Applications to zoo management Bartczak A, Englin J, Pang A (2012) When are forest visits valued the
most? An analysis of the seasonal demand for forest recreation in
A better understanding of factors influencing visits can allow Poland. Environ Resour Econ 52(2):249–264. [Link]
for better preparation for periods with high visitor numbers 1007/s10640-011-9527-1
Becken S (2013) Measuring the effect of weather on tourism: a destina-
and appropriate allocation of labor and resources. This will
tion and activity-based analysis. J Travel Res 52(2):156–167. https://
enable zoos to better cope with large numbers of visitors. In [Link]/10.1177/0047287512461569
addition, during times when high numbers of visitors might Beutels P, Jia N, Zhou Q-Y, Smith R, Cao W-C, de Vlas SJ (2009) The
cause overcrowding or conflicts of use, jeopardizing service economic impact of SARS in Beijing, China. Tropical Med Int
Health 14:85–91. [Link]
quality and customer satisfaction, staff could direct crowds
x
along different routes through the zoo, thereby providing a Butler RW (1980) The concept of a tourist area cycle of evolution: im-
higher level of customer service. Conversely, it will also allow plications for management of resources. Can Geogr 24(1):5–12.
for anticipation of periods with lower visitor numbers, during [Link]
Creemers L, Wets G, Cools M (2015) Meteorological variation in daily
which various marketing tactics aimed at attracting visitors
travel behaviour: evidence from revealed preference data from the
could be adopted, thus making effective use of recreation Netherlands. Theor Appl Climatol 120(1–2):183–194. [Link]
resources. org/10.1007/s00704-014-1169-0
Dwyer JF (1988) Predicting daily use of urban forest recreation sites.
Landsc Urban Plan 15(1–2):127–138. [Link]
5.5 Limitations and future research 0169-2046(88)90021-7
Elberse A (2007) The power of stars: do star actors drive the success of
Although this study considered temporal fluctuations, special movies? J Marketing 71(4):102–120. [Link]
71.4.102
events, and weather conditions, subsequent research could Finger R, Lehmann N (2012) Modeling the sensitivity of outdoor recre-
still investigate other factors that may affect visits. First, more ation activities to climate change. Clim Res 51(3):229–236. https://
samples might be needed for out-of-sample modeling to val- [Link]/10.3354/cr01079
idate the models. Second, the animal star effect could also Hewer MJ, Scott D, Gough WA (2015) Tourism climatology for
camping: a case study of two Ontario parks (Canada). Theor Appl
include the corresponding level of media reporting, so it Climatol 121(3-4):401–411. [Link]
would be possible to use media exposure to estimate the ani- 1228-6
mal star effect on visits or the duration of the effect by collat- Inc SPSS (2007) SPSS statistics base 17.0: user’s guide. SPSS Inc,
ing data regarding past media coverage. Furthermore, over the Chicago
Karniouchina EV (2011) Impact of star and movie buzz on motion picture
years, Taipei Zoo has successively constructed new outdoor distribution and box office revenue. Int J Res Mark 28(1):62–74.
and indoor display areas on its premises and has closed certain [Link]
exhibition areas for refurbishment. If data on the total areas King MA, Abrahams AS, Ragsdale CT (2014) Ensemble methods for
involved were obtained, it could offer a greater understanding advanced skier days prediction. Expert Syst Appl 41(4):1176–1188.
[Link]
of the extent of the relationship between visits and environ- Koutsoyiannis A (1977) Theory of econometrics. Macmillan, London.
mental attributes. Additionally, outbreaks of other diseases or [Link]
epidemics except SARS may well have an effect on visitor Kuo H-I, Chen C-C, Tseng W-C, Ju L-F, Huang B-W (2008) Assessing
numbers. Defining the time scope of each disease or epidemic impacts of SARS and Avian Flu on international tourism demand to
Asia. Tour Manag 29(5):917–928. [Link]
outbreak would be required. Future research with further in- tourman.2007.10.006
depth investigation is awaited in this respect. Lee C-K, Song H-J, Bendle LJ, Kim M-J, Han H (2012) The impact of
non-pharmaceutical interventions for 2009 H1N1 influenza on trav-
el intentions: a model of goal-directed behavior. Tour Manag 33(1):
89–99. [Link]
Li C, Lin S-H (2012) Influence of weather conditions on hiking behavior.
References Int J Biometeorol 56(4):777–781. [Link]
011-0459-9
Liu A, Liu Y, Mazumdar T (2014) Star power in the eye of the beholder: a
Arnberger A (2006) Recreation use of urban forests: an inter-area com- study of the influence of stars in the movie industry. Mark Lett
parison. Urban For Urban Gree 4(3–4):135–144. [Link] 25(4):385–396. [Link]
1016/[Link].2006.01.004 Mao C-K, Ding CG, Lee H-Y (2010) Post-SARS tourist arrival recovery
Arnberger A, Eder R (2007) Monitoring recreational activities in urban patterns: an analysis based on a catastrophe theory. Tour Manag
forests using long-term video observation. Forestry 80(1):1–15. 31(6):855–861. [Link]
[Link] Marvasti A (2013) Estimating outdoor recreation demand with aggregate
Arnberger A, Hinterberger B (2003) Visitor monitoring methods for man- data: a revealed preference approach. Ocean Coast Manage 71:170–
aging public use pressures in the Danube Floodplains National Park, 175. [Link]
472 A.-T. Su, Y.-J. Lin

McAleer M, Huang B-W, Kuo H-I, Chen C-C, Chang C-L (2010) An Scott D, Gössling S, de Freitas CR (2008) Preferred climates for tourism:
econometric analysis of SARS and Avian Flu on international tourist case studies from Canada, New Zealand and Sweden. Clim Res
arrivals to Asia. Environ Model Softw 25(1):100–106. [Link] 38(1):61–73. [Link]
org/10.1016/[Link].2009.07.015 Shih C, Nicholls S, Holecek DF (2009) Impact of weather on downhill ski
Min JCH, Lim C, Kung H-H (2011) Intervention analysis of SARS on lift ticket sales. J Travel Res 47(3):359–372. [Link]
Japanese tourism demand for Taiwan. Qual Quant 45(1):91–102. 0047287508321207
[Link] Siderelis C, Moore RL (2011) A mixed logit model of visitors’ national
Ploner A, Brandenburg C (2003) Modelling visitor attendance levels park choices. Soc Nat Resour 24(8):799–813. [Link]
subject to day of the week and weather: a comparison between linear 1080/08941921003731009
regression models and regression trees. J Nat Conserv 11(4):297– Su A-T, Cheng C-K, Lin Y-J (2014) Modeling daily visits to the 2010
308. [Link] Taipei International Flora Exposition. Urban For Urban Greening
Price AE, Reed JA, Hooker SP (2012) The association of natural ele- 13(4):725–733. [Link]
ments and trail use by adults. Prev Chronic Dis 9:110177. https:// Tang TC, Wong KN (2009) The SARS epidemic and international visitor
[Link]/10.5888/pcd9.110177 arrivals to Cambodia: is the impact permanent or transitory? Tour
Rassy D, Smith RD (2013) The economic impact of H1N1 on Mexico’s Econ 15(4):883–890. [Link]
tourist and pork sectors. Health Econ 22(7):824–834. [Link] 000000009789955206
10.1002/hec.2862 Wang Y-S (2009) The impact of crisis events and macroeconomic activity
on Taiwan’s international inbound tourism demand. Tour Manag
Ritchie JRB, Beliveau D (1974) Hallmark events: an evaluation of a
30(1):75–82. [Link]
strategic response to seasonality in the travel market. J Travel Res
Yang Y, Wong KKF (2012) A spatial econometric approach to model
13(2):14–20. [Link]
spillover effects in tourism flows. J Travel Res 51(6):768–778.
[Link]