Received: 18 December 2017 Accepted: 21 June 2018
DOI: 10.1111/conl.12592
LETTER
Incorporating biotic interactions reveals potential climate
tolerance of giant pandas
Fang Wang1,2 Qing Zhao3 William J. McShea1 Melissa Songer1 Qiongyu Huang1
Xiaofeng Zhang4 Lingguo Zhou4
1National Zoological Park, Smithsonian
Conservation Biology Institute, Front Royal,
Virginia
2Michigan State University, East Lansing,
Michigan
3School of Natural Resources, University of
Missouri, Columbia, Missouri
4Shaanxi Forestry Department, Xi'an,
Shaanxi, China
Correspondence
FangWang,National Zoological Park, Smith-
sonianConservationBiology Institute, Front
Royal,VA22630.
Email:Wangfang.vic@gmail.com
Editor
Lu Zhi
Abstract
Many studies have overestimated species’ range shifts under climate change because
they treat climate as the only determinant while ignoring biotic factors. To assess the
response of giant pandas to climate change, we incorporated spatial effects in model-
ing bamboo distributions, which in turn was incorporated to represent giant panda–
bamboo biotic interactions in predicting giant panda distribution. Our study revealed
potential tolerance of giant pandas to climate change. We found significant residual
spatial correlation in the bamboo models. The biotic interactions with bamboo under-
stories and anthropogenic activities had large effects on panda distribution, which
lowered the relative importance of climatic variables. Our results are fundamentally
different from previous studies that used climate-only and nonspatial approaches,
which may have overestimated the effects of climate change on panda and lead to
inappropriate conservation recommendations. We strongly advocate that giant panda
conservation planning continues to focus on protecting bamboo forest and reducing
anthropogenic interferences.
KEYWORD S
bamboo, biotic interaction, China, climate change, conservation planning, giant panda, spatial autocorre-
lation, species distribution model, wildlife conservation
1 INTRODUCTION
Climate change is challenging the conservation planning of
governments and natural resource organizations (Bernazzani,
Bradley, & Opperman, 2012). However, forecasts based on
species distribution models (SDMs) are often criticized for
being too simplistic if they assume that climate and few abi-
otic factors are the only determinants of a species’ geograph-
ical range (Harris et al., 2014). Biotic interactions such as
resource–consumer interactions and interspecific competition
are also essential factors that drive species’ distributions, and
incorporating these factors can improve forecasts of the eco-
This is an open access article under the terms of the Creative Commons Attribution License, which permits use, distribution and reproduction in any medium, provided the original
work is properly cited.
© 2018 The Authors. Conservation Letters published by Wiley Periodicals, Inc.
logical consequences of climate change on species (Wisz
et al., 2013). However, most studies have adopted a climate-
only modeling approach and ignored important biotic factors,
even when such information was available (Dormann, 2007;
Pacifici et al., 2015).
Another critical but often-ignored issue is spatial auto-
correlation (SAC). SAC can derive from biotic interactions,
biotic traits such as dispersal limits and narrow ecophysical
niche (e.g., certain soil type), and other specialized habitat use
(Merckx, Steyaert, Vanreusel, Vincx, & Vanaverbeke, 2011).
While incorporating biotic traits such as slow migration
can improve the performance of SDMs in mapping species’
Conservation Letters. 2018;e12592. wileyonlinelibrary.com/journal/conl 1 of 9
https://doi.org/10.1111/conl.12592
2 of 9 WANG ET AL.
realized niche spaces (Botkin et al., 2007), data representing
biotic interactions and biotic traits may not always be avail-
able, and residual SAC needs to be accounted for. Failure to
account for SAC can lead to overstated predictions of species’
habitat loss when extrapolated to future conditions (Crase,
Liedloff, Vesk, Fukuda, & Wintle, 2014; Zhao, Boomer,
Silverman, & Fleming, 2017).
Because the giant panda (Ailuropoda melanoleuca) is
specialized to feed on bamboo, it is important to consider
its biotic interaction with bamboo for conservation planning.
Recent studies that directly connected giant panda distribu-
tions with climatic metrics predicted a severe habitat loss
of 37–62% (Fan et al., 2014), 60% (Songer, Delion, Biggs,
& Huang, 2012), or 53–71% (Li et al., 2015). However,
ingoring giant panda's interaction with bamboo as well as
other habitat preferences may result in overrated importance
of climatic variables. Some studies have included biotic
interactions (e.g., bamboo distributions) in their models,
but did not consider bamboo's dispersal limit caused by its
unique clustered distribution pattern and/or ignored the effect
of critical anthropogenic variables. The recommendations
from most of these studies is to establish new nature reserves
outside of the current network to mitigate the threats of
climate change (Fan et al., 2014; Songer et al., 2012; Tuanmu
et al., 2013). These recommendations can be costly and risky,
however, if the models used in these studies overestimated
shifts in giant panda or bamboo distributions under climate
change.
The goal of our study is to evaluate the response of giant
pandas to future climate change. Our objectives are to: (1)
examine the effects of climate on bamboo distributions while
accounting for residual SAC; (2) identify the relative con-
tributions of biotic interactions, anthropogenic disturbances,
and climate in driving giant panda distributions; (3) predict
future distributions of bamboos and giant panda under climate
change, and (4) provide recommendations for conservation
strategies. This study has strong implications to the conserva-
tion of giant panda, as well as other species that are predicted
to experience a significant shift in their critical resources as a
result of climate change.
2 METHODS
2.1 Study area
We used the distribution of giant pandas in the Qinling Moun-
tains (hereafter referred to as Qinling) with a 10 km buffer
zone as our study area. Two species, wood bamboo (Basha-
nia fargesii) and arrow bamboo (Fargesia qinlingensis), are
the main diet of giant pandas in Qinling. Both bamboo species
have long flowering intervals and, between flowering events,
they use asexual reproduction to spread outward along rhi-
zomes at a rate of approximately <10 m per year (Sun, 2011).
2.2 Species data
For giant pandas, we obtained distribution information
from the Shaanxi Forestry Department (SFD). The SFD
recorded giant panda signs along 424 transects (approxi-
mately 1,360 km total length) distributed across both pro-
tected and unprotected habitat in Shaanxi Province during
2010–2012 (SFD, 2017). The study area covered all exist-
ing or potential habitat for giant pandas and included forests
inside and outside of eighteen nature reserves (Figure 1).
Giant pandas were labeled “present” if giant panda signs (e.g.,
fecal and foraging site) were recorded.
For arrow and wood bamboo, 5,998 vegetation plots were
surveyed across the Qinling. The plot locations were along the
giant panda transects and regularly spaced across the entire
potential range of giant pandas (Figure 1). Field staff recorded
a bamboo species as present when it occupied an area larger
than 10 m × 10 m (SFD, 2017).
Because both the giant panda and bamboo surveys yielded
presence-only data, in order to use statistical algorithms
requiring both presence and absence values, we generated
random pseudo-absent locations (Iturbide et al., 2015).
Pseudo-absent locations for giant panda were randomly
generated at 2 km from the presence points according to its
home range size (approximately 5 km2; Pan et al., 2014). Due
to the relatively low dispersal ability of bamboos, pseudo-
absence sites for bamboos were randomly selected without
distance limitation from presence sites. We acknowledge
that such presence–absence data represent indices of giant
panda or bamboo distributions rather than true occupancy
status.
2.3 Climate data
We constructed models using current climatic conditions
(average for 1950–2000, Supporting Information Table S1)
and projected to the future (given by WorldClim for the range
2061–2080, hereafter referred to as future). We selected two
widely used Representative Concentration Pathways (RCP)
scenarios for our study: RCP4.5, an optimistic scenario where
carbon emissions peak around 2040, resulting in 4.5 W/m2
radiative forcing by 2100; and RCP8.5, a pessimistic scenario,
which reflects high carbon, resulting in 8.5 W/m2 radiative
forcing by 2100 (Moss et al., 2010). We used three global cli-
matemodels (hereafter referred to as GCM) for future climatic
conditions: ACCESS1.0, CCSM4, and HadGEM2-AO (here-
after referred to as AC, CC, and HD, respectively (Hijmans,
Cameron, Parra, Jones, & Jarvis, 2005).
2.4 Environment data
We reviewed previous studies for the giant panda and the
two bamboo species (Hull et al., 2014; Pan et al. 2014; Wang
et al., 2014; Zhang et al., 2011, 2017), and identified abiotic
WANG ET AL. 3 of 9
F IGURE 1 Sampling design for giant panda and bamboo species distribution in the Qinling Mountains. Government staff conducted 424 line
transects across 18 nature reserves across the Qinling Mountains, covering the known giant panda distribution area with an approximate 10 km buffer.
See methods for transect details. Known distribution of giant pandas is indicated in insert map
(e.g., elevation and slope), biotic (e.g., bamboo presence
for giant panda), and anthropogenic (e.g., road transporta-
tion and construction for giant panda) variables that have
been shown to affect their occupancy (Supporting Informa-
tion Table S2). We used a 30-m resolution digital elevation
model (Global ASTER, 2009) to delineate the slope, aspect,
and terrain ruggedness using ArcToolbox in ArcGIS 10.2
(ESRI, 2011). Georeferenced data of nature reserves, human
residences roads, and mining locations were obtained from
the SFD. All the layers (Table S2) were finally standardized
to 500 m × 500 m spatial resolution using ArcToolbox in
ArcGIS 10.2.
2.5 Modeling current and future bamboo
distributions
Prior to modeling bamboo and giant panda distribution, we
conducted a four-step variable selection to reduce the mul-
ticollinearity of remaining climatic and nonclimatic candi-
date variables (see details in Supporting Information 1) using
a Variance Inflation Factor (VIF) method (García, García,
López Martín, & Salmerón, 2015). We excluded any vari-
able that had a VIF value greater than 5 from further analyses
(Shiu, 2006).
To model the current bamboo distribution, we first com-
pared the discriminative performance of an environmental-
only model (hereafter ENV) that do not account for residual
SAC, and a residual autocovariated model (hereafter RAC),
which included an autocovariate term derived from the resid-
uals of the ENVmodel (Crase et al., 2014).We used an ensem-
ble modeling approach (Pliscoff, Luebert, Hilger, & Guisan,
2014), in which six modeling algorithms were used: artifi-
cial neural network (ANN), generalized linear model (GLM),
boosted regression tree (BRT), maximum entropy model-
ing (MAXENT), multivariate adaptive regression splines
(MARS), and random forest (RF). Discriminative perfor-
mance was assessed under a 10-fold process using two cross-
validated performance metrics, the AUC (area under curve of
the receiver operating characteristic [ROC]; Fawcett, 2006)
and the TSS (true skill statistic; de Oliveira, Rangel, Lima-
Ribeiro, Terribile, & Diniz-Filho, 2014). We plotted Moran's
I correlogram (Legendre & Legendre, 2012) to further quan-
tify the remaining spatial autocorrelation in model residuals
and validate the model performance.
To forecast future bamboo distributions, we modeled bam-
boo distributions using three GCMs and two RCPs (IPCC,
2012). Since the asexual reproductive dispersal for our focal
bamboo species was relatively low, we used the current
4 of 9 WANG ET AL.
autocovariate in the predictive functions for RAC models. To
better demonstrate the species’ distribution change, we used
a threshold that maximized the sum of modeling sensitivity
and specificity to transform the species occurrence probabil-
ities to binary presences/absences predictions (Cantor, Sun,
Tortolero-Luna, Richards-Kortum, & Follen, 1999).
2.6 Modeling current and future giant panda
distributions
With the bamboo model outputs, we used the same ensemble
modelingmethod, and constructed threemodeling approaches
to predict current giant panda habitat: climate-only models
that connect giant panda distribution with only climatic vari-
ables; bamboo-ENV model that incorporates ENV bamboo
model output as well as other critical habitat preferences; and
bamboo-RAC model in which RAC bamboo model output
and other habitat preferences were added. To predict future
giant panda distributions, we used the same three GCMs and
two RCPs.
3 RESULTS
3.1 Bamboo–environment relationship
Five climate variables and three nonclimate environmental
variables were included in bamboo models after the collinear-
ity test (Supporting Information Table S1). The correlogram
and map of model residuals revealed a significant nonrandom
pattern for arrow and wood bamboo ENV models (Sup-
porting Information Figure S1). The Moran's I (p < 0.01)
indicated higher similarities among survey locations within
20 km for arrow bamboo and 25 km for wooden bamboo
(Supporting Information Figure S2). Incorporating the RAC
term significantly improved the model's discriminative ability
(Figure 2), and reduced residual SAC (Supporting Informa-
tion Figure S2).
For both arrow and wood bamboo, BIO13 (precipitation of
the wettest month) had the highest contribution in modeling
species distribution (Table 1), followed byBIO1 (annualmean
temperature) and BIO6 (min temperature of coldest month)
for arrow bamboo and BIO6 and BIO15 (precipitation season-
ality) for wood bamboo. Though the species–climate associa-
tions were similar between the ENV and RAC models, includ-
ing RAC reduced the importance of climatic variables for both
bamboo species (Table 1).
3.2 Current and future bamboo distribution
Both modeling approaches estimated a similar expected
number of occupied cells for current bamboo distributions
(Figure 3). However, the occurrence probability under cli-
mate change scenarios diverged (Figure 3). For the wood
bamboo, the ENV model predicted a 36–85% loss (RCP4.5:
36–85%, RCP8.5: 46–88%) in its distribution under cli-
mate change, while the RAC model predicted less decrease
(RCP4.5: 0%; RCP8.5: 0–31%) than the ENV model. For the
arrow bamboo, the ENV model predicted more habitat loss
under both climate change scenarios (RCP4.5: 52–70% and
RCP8.5: 62–85%) than the RAC model (RCP4.5: 0–39% and
RCP8.5: 24–51%). According to RAC models, bamboo cov-
erage in the central Qinling where the four most important
nature reserves are located remains primarily bamboo covered
(>85%).
3.3 Biotic interactions in giant panda models
Seven nonclimatic variables, five climatic variables, and a
bamboo layer were included in giant panda models (Sup-
porting Information Table S2). The biotic interactions intro-
duced into the giant panda models significantly improved the
model's discriminative ability (p < 0.01; Figure 2).
According to both biotic models, four predictors besides
climatic variables (i.e., bamboo distribution, distance to road,
distance to large residences, and distance to nature reserves)
had a model weight higher than 0.1 (Table 1). The occupancy
probabilities of giant pandas were higher in areas with bam-
boo understory in or adjacent to nature reserves. Being close
to residential areas and major roads significantly reduced the
occupancy probabilities of giant pandas, which suggested the
negative associations between giant panda and human infras-
tructure. Including bamboo distributions and other noncli-
mate variables lowered the importance of climate variables in
explaining giant panda distributions, despite that these noncli-
mate variables are not highly correlated to climate variables
(Table S1 and S2).
3.4 Future giant panda distribution
Climate-only models of giant panda distribution predicted
results similar to previous studies that giant pandas would
lose 49–85% of their current habitat under a range of cli-
mate change scenarios (Figure 4). By contrast, the biotic mod-
els predicted less habitat loss (bamboo-ENV model: mean
42%, range 33–65%; bamboo-RAC model: mean 16%, range
12–34%). Though the results diverged, all three modeling
approaches predicted major habitat loss in eastern Qinling
Mountains.
4 DISCUSSION
Our study revealed potential tolerance of giant pandas to
future climate change. . Previous studies either used climate-
only models (Fan et al., 2014; Songer et al., 2012) or non-
spatially modeled bamboo distributions and climate variables
WANG ET AL. 5 of 9
F IGURE 2 The cross-validation results using AUC and TSS to compare the performance of different modeling approaches. Each cross indicates
the mean and SD of AUC and TSS tests for giant panda (A: climate-only model; B: bamboo-ENV and bamboo-RAC models), arrow bamboo (C: ENV
model; D: RAC model), and wood bamboo (E: ENV model; F: RAC model) bamboo modeling. Higher values for both tests represent improved model
performance when biological traits and a spatial term was incorporated in the modeling
(Li et al., 2015; Tuanmu et al., 2013) to model giant panda
distributions. Both these methods produced dire forecasts and
emphasized the effects of climate. Our climate-only giant
panda models and ENV bamboo models predicted similar
species’ distribution changes to these previous studies. How-
ever, given their relatively poor discriminative performance
and ignorance of important biotic interactions and residual
SAC, we believe that they have overemphasized the effects
of climate change on giant panda distribution, and may lead
to inappropriate recommendations for conservation actions.
We found that accounting for consumer–resource biotic
interactions and residual SAC improved model performance
and changed the forecasts. We acknowledge that our results
are fundamentally different from previous studies (Fan et al.,
2014; Li et al., 2015; Tuanmu et al., 2013), including some
of the present authors (Songer et al., 2012). We found that
6 of 9 WANG ET AL.
TABLE 1 The relative importance of the climatic and nonclimatic variables in modeling bamboo and giant panda distribution. Bold numbers
indicate variables had a relative importance >0.1
Arrow bamboo Wood bamboo Giant panda
Variable type Variable ENV RAC ENV RAC Climate-only Bamboo-ENV Bamboo-RAC
Climate Bio1 0.21 0.15 0.06 0.04 0.26 0.05 0.04
Bio6 0.14 0.10 0.19 0.10 0.05 0.01 0.01
Bio11 0.11 0.09 0.14 0.06 0.03 0.03 0.03
Bio13 0.32 0.25 0.39 0.19 0.36 0.05 0.04
Bio15 0.11 0.08 0.16 0.08 0.32 0.09 0.08
Land feature Aspect 0.02 0.01 0.03 0.02 – 0.00 0.00
Slope 0.06 0.04 0.02 0.02 – 0.01 0.01
Ruggedness 0.01 0.00 0.00 0.00 – 0.00 0.00
Biotic Bamboo – – – – – 0.36 0.38
Anthropogenic Residential area – – – – – 0.11 0.11
Road – – – – – 0.14 0.15
Mining site – – – – – 0.02 0.01
Nature reserve – – – – – 0.15 0.15
F IGURE 3 The occurrence probability for the arrow and wood bamboo under climate change scenarios. The ENV model forecasts a major
decrease in both arrow and wood bamboo distribution under different GCMs and RCPs (indicated in right; see Methods for details). Combining the
wood and arrow bamboo, the RAC model forecasts a more stable distribution
WANG ET AL. 7 of 9
F IGURE 4 The occurrence probability of giant pandas under climate change predicted by climate-only, bamboo-ENV, and bamboo-RAC mod-
eling approach. Climate-only models predicted similar results to previous studies under different three GCMs (AC, CC, and HD) and two RCPs (RCP
4.5 and 8.5), with giant pandas losing 49–85% of its current habitat under various climate change scenarios. The bamboo-ENV model predicted a mean
habitat loss of 44% (33–65%), and the bamboo-RAC model predicted a habitat loss of 16% (12–34%), with new habitat patches located in northern
Qinling Mountains
species environmental envelop (niche breadth) to be wider
than the projected temperature/precipitation changes, so the
species can potentially persist under the projected climatic
conditions.
An advantage of our RAC approach is that the RAC term is
calculated from the residuals of nonspatial models, and thus
represents factors other than the covariates already included in
the models such as land facet (Brost & Beier 2012; Wessels,
Freitag, & Van Jaarsveld, 1999), tourism, and species interac-
tions (e.g., livestock grazing) (Wang, McShea, Wang, & Li,
2015; Zhang et al., 2017), for which data are not available
for the current study. In addition, the underground rhizome
system of bamboos (He et al., 2000) may also cause residual
SAC in the models, but such effects are difficult to quantify
and need to be accounted for using the RAC term. In contrast
to the stems and leaves that might be more vulnerable to tem-
perature change, the rhizome system is belowground, enabling
the lateral buds to produce either canes or new rhizomes with
less impact from aboveground temperatures. The asexual dis-
persal characteristics of bamboos may provide resilience of
these species against unsuitable climatic conditions, a pattern
that is consistent with the forecast of our RAC models. Due
to the complex characteristics of the RAC term, future studies
that focus on the effects of anthropogenic factors such as agri-
culture, livestock grazing and tourism on bamboo and giant
panda distributions are warranted.
8 of 9 WANG ET AL.
One of the most important principals in climate change
mitigation is that the decision-making process should be
based on the most comprehensive data and robust models
(Nicholson & Possingham 2007). Other than proposing new
nature reserves and planting bamboos in areas without cur-
rent giant panda distributions, we suggest that the future con-
servation plans focus on reinforcing current strategy, with
special emphasize on the adaptive management of fast devel-
oping tourism and other anthropogenic activities (e.g., farm-
ing and livestock grazing) in bamboo forest. For example,
though the current habitats at lower elevations may remain
suitable for giant panda if bamboo and forests remain, farm-
land moves up under warmer environment could be an emerg-
ing threat which warrants further attention. A great opportu-
nity to better target our results in conservation practices lies
in the Overall Plan of Ecological Civilization Systems Reform
recently announced by Chinese government. This plan intro-
duced major changes in the way natural resources are man-
aged, including nationwide transfer payment for ecosystem
service (PES), key ecological function regions zoning, and
the establishment of three huge giant panda national parks
(Ouyang et al., 2016). We strongly advocate that the newly
proposed national parks as well as existing national reserves
establish a comprehensive, adaptive framework of monitor-
ing, modeling, and managing natural resources and human
activities (including proposed tourism projects) (Xu et al.,
2017). In addition, areas that are predicted suitable for giant
panda and bamboo species, for example, the northern Qin-
ling Mountains, should be identified as key ecological func-
tion regions with higher PES rates (Yang et al., 2018). We
feel that these efforts would bear more positive results for
climate change mitigation, for vulnerable giant pandas and
beyond.
Despite the fast development of SDMs, many scientists
and conservation practitioners still estimate species’ range
shifts based on the assumption that climate and few abiotic
factors are the only determinants. We believe that this study
has strong implications to establish a better understanding of
climate-mediated range shifts for many other species around
the world. Armed with such knowledge, scientists and con-
servation practitioners may be able to better identify conser-
vation priorities to ensure the long-term survival of wildlife
species.
ACKNOWLEDGMENTS
We thank the staff of Huangbaiyuan Nature Reserve,
Pingheliang Nature Reserve, Niuweihe Nature Reserve, and
Changqing Nature Reserve for their assistance in the field-
work. The Shaanxi Forestry Department helped in logistical
details and permit applications.
REFERENCES
Aster, GDEM. (2009). ASTER GDEM is a product of NASA and METI.
NASA EOSDIS Land Processes DAAC, USGS Earth Resources
Observation and Science (EROS) Center, Sioux Falls, South Dakota
(https://lpdaac.usgs.gov), accessed March 4, 2017, at https://doi.org/
10.5067/ASTER/ASTGTM.002
Bernazzani, P., Bradley, B. A., & Opperman, J. J. (2012). Integrating
climate change into habitat conservation plans under the U.S. endan-
gered species act. Environmental Management, 49, 1103–1114.
Botkin, D. B., Saxe, H., Araújo, M. B., Betts, R., Bradshaw, R. H. W.,
Cedhagen, T.,… Faith, D. P. (2007). Forecasting the effects of global
warming on biodiversity. Bioscience, 57, 227–236.
Brost, B. M., & Beier, P. (2012). Use of land facets to design linkages
for climate change. Ecological Applications, 22, 87–103.
Cantor, S. B., Sun, C. C., Tortolero-Luna, G., Richards-Kortum, R., &
Follen, M. (1999). A comparison of C/B ratios from studies using
receiver operating characteristic curve analysis. Journal of Clinical
Epidemiology, 52, 885–892.
Crase, B., Liedloff, A., Vesk, P. A., Fukuda, Y., & Wintle, B. A. (2014).
Incorporating spatial autocorrelation into species distribution mod-
els alters forecasts of climate-mediated range shifts. Global Change
Biology, 20, 2566–2579.
Dormann, C. F. (2007). Effects of incorporating spatial autocorrelation
into the analysis of species distribution data.Global Ecology and Bio-
geography, 16, 129–138.
ESRI. (2011). ArcGIS Desktop: Release 10. Redlands, CA: Environmen-
tal Systems Research Institute
Fan, J., Li, J., Xia, R., Hu, L., Wu, X., & Li, G. (2014). Assessing
the impact of climate change on the habitat distribution of the giant
panda in the Qinling Mountains of China.EcologicalModelling, 274,
12–20.
Fawcett, T. (2006). An introduction to ROC analysis. Pattern Recogni-
tion Letters, 27, 861–874.
García, C. B., García, J., López Martín, M. M., & Salmerón, R. (2015).
Collinearity: Revisiting the variance inflation factor in ridge regres-
sion. Journal of Applied Statistics, 42, 648–661.
Harris, D. B., Gregory, S. D., Brook, B. W., Ritchie, E. G., Croft, D.
B., Coulson, G., & Fordham, D. A. (2014). The influence of non-
climate predictors at local and landscape resolutions depends on the
autecology of the species. Austral Ecology, 39, 710–721.
He, Q., Wang, K., Wu, R., Weng, P., Zhang, P., Wu, Z., & Hu, K. (2000).
Investigation on the rhizome and root system of different manage-
ment types of bamboo shoot forest of Phyllostachys heterocycla cv.
Pubescens. Journal of Zhejiang Forestry Science and Technology, 20,
31–34.
Hijmans, R. J., Cameron, S. E., Parra, J. L., Jones, P. G., & Jarvis, A.
(2005). Very high resolution interpolated climate surfaces for global
land areas. International Journal of Climatology, 25, 1965–1978.
Hull, V., Roloff, G., Zhang, J., Liu, W., Zhou, S., Huang, J., … Liu, J.
(2014). A synthesis of giant panda habitat selection. Ursus, 25, 148–
162.
Iturbide, M., Bedia, J., Herrera, S., del Hierro, O., Pinto, M., & Gutiér-
rez, J. M. (2015). A framework for species distribution modelling
with improved pseudo-absence generation. Ecological Modelling,
312, 166–174.
WANG ET AL. 9 of 9
Leguendre, P. & Leguendre, L. (2012). Numerical Ecology, 3rd Edn.,
Vol. 24. Amsterdam: Elsevier.
Li, R., Xu, M., Wong, M. H. G., Qiu, S., Li, X., Ehrenfeld, D., & Li, D.
(2015). Climate change threatens giant panda protection in the 21st
century. Biological Conservation, 182, 93–101.
Moss, R.H., Edmonds, J.A., Hibbard, K.A., Manning, M.R., Rose, S.K.,
Van Vuuren, D.P., Carter, T.R., Emori, S., Kainuma, M. & Kram, T.
(2010). The next generation of scenarios for climate change research
and assessment. Nature, 463, 747.
Merckx, B., Steyaert, M., Vanreusel, A., Vincx, M., & Vanaverbeke,
J. (2011). Null models reveal preferential sampling, spatial autocor-
relation and overfitting in habitat suitability modelling. Ecological
Modelling, 222, 588–597.
Nicholson, E., & Possingham, H. P. (2007). Making conservation deci-
sions under uncertainty for the persistence of multiple species. Eco-
logical Applications, 17, 251–265.
de Oliveira, G., Rangel, T. F., Lima-Ribeiro, M. S., Terribile, L. C.,
& Diniz-Filho, J. A. F. (2014). Evaluating, partitioning, and map-
ping the spatial autocorrelation component in ecological niche mod-
eling: A new approach based on environmentally equidistant records.
Ecography, 37, 637–647.
Ouyang, Z., Zheng, H., Xiao, Y., Polasky, S., Liu, J., Xu, W., … Rao,
E. (2016). Improvements in ecosystem services from investments in
natural capital. Science, 352, 1455–1459.
Pacifici, M., Foden, W. B., Visconti, P., Watson, J. E. M., Butchart, S.
H. M., Kovacs, K. M.,… Akcakaya, H. R. (2015). Assessing species
vulnerability to climate change. Nature Climate Change, 5, 215.
Pan, W., Lu, Z., Zhu, X., Wang, D., Wang, H., Long, Y., … Zhou, X.
(2014). A chance for lasting survival: Ecology and behavior of wild
giant pandas. Washington, DC: Smithsonian Institution Press.
Pliscoff, P., Luebert, F., Hilger, H. H., & Guisan, A. (2014). Effects of
alternative sets of climatic predictors on species distribution models
and associated estimates of extinction risk: A test with plants in an
arid environment. Ecological Modelling, 288, 166–177.
Shaanxi Forestry Department. (2017).Giant pandas of Qinling: A report
of the fourth giant panda census. Xi'an: Shaanxi Science and Tech-
nology Press.
Shiu, H. -J. (2006). The application of the value added intellectual coef-
ficient to measure corporate performance: Evidence from technolog-
ical firms. International Journal of Management, 23, 356.
Songer, M., Delion, M., Biggs, A., & Huang, Q. (2012). Modeling
impacts of climate change on giant panda habitat. International Jour-
nal of Ecology, 2012, http://doi.org/10.1155/2012/108752.
Sun, Y. (2011). Reassessing Giant Panda Habitat with Satellite-derived
Bamboo Information: A Case Study in the Qinling Mountains, China
(Master dissertation). University of Twente, Enschede. Retrieved
from https://webapps.itc.utwente.nl/librarywww/papers_2011/msc/
nrm/sun.pdf.
Tuanmu, M. -N., Viña, A., Winkler, J. A., Li, Y., Xu, W., Ouyang, Z.,
& Liu, J. (2013). Climate-change impacts on understorey bamboo
species and giant pandas in China's Qinling Mountains. Nature Cli-
mate Change, 3, 249–253.
Wang, F., McShea, W. J., Wang, D., & Li, S. (2015). Shared resources
between giant panda and sympatric wild and domestic mammals.
Biological Conservation, 186, 319–325.
Wang, F., McShea, W. J., Wang, D., Li, S., Zhao, Q., Wang, H., & Lu, Z.
(2014). Evaluating landscape options for corridor restoration between
giant panda reserves. Plos One, 9, e105086.
Wessels, K. J., Freitag, S., & Van Jaarsveld, A. S. (1999). The use of
land facets as biodiversity surrogates during reserve selection at a
local scale. Biological Conservation, 89, 21–38.
Wisz, M. S., Pottier, J., Kissling, W. D., Pellissier, L., Lenoir, J.,
Damgaard, C. F., … Guisan, A. (2013). The role of biotic interac-
tions in shaping distributions and realised assemblages of species:
Implications for species distribution modelling. Biological Reviews,
88, 15–30.
Xu, W., Viña, A., Kong, L., Pimm, S. L., Zhang, J., Yang, W.,… Liu, J.
(2017). Reassessing the conservation status of the giant panda using
remote sensing. Nature Ecology & Evolution, 1, 1635.
Yang, H., Yang, W., Zhang, J., Connor, T. & Liu, J. (2018). Revealing
pathways from payments for ecosystem services to socioeconomic
outcomes. Science Advances, 4, eaao6652.
Zhang, J., Hull, V., Ouyang, Z., Li, R., Connor, T., Yang, H., … Liu,
J. (2017). Divergent responses of sympatric species to livestock
encroachment at fine spatiotemporal scales.Biological Conservation,
209, 119–129.
Zhang, Z., Swaisgood, R.R., Zhang, S., Nordstrom, L.A., Wang, H., Gu,
X., Hu, J. & Wei, F. (2011). Old-growth forest is what giant pandas
really need. Biology Letters, 7, 403–406.
Zhao, Q., Boomer, G. S., Silverman, E., & Fleming, K. (2017). Account-
ing for the temporal variation of spatial effect improves inference and
projection of population dynamics models. Ecological Modelling,
360, 252–259.
SUPPORTING INFORMATION
Additional supporting information may be found online in the
Supporting Information section at the end of the article.
How to cite this article: Wang F, ZhaoQ,McSheaWJ,
et al. Incorporating biotic interactions reveals potential
climate tolerance of giant pandas. Conservation Let-
ters. 2018;e12592. https://doi.org/10.1111/conl.12592