B I O D I V E R S I T Y R E S E A R C H
Spatial distribution as a measure of conservation needs: an example with Asiatic black bears in south-western China
Fang Liu1*, William McShea2, David Garshelis3, Xiaojian Zhu1, Dajun Wang1, Ji’en Gong4and Youping Chen5
INTRODUCTION
Changes in the spatial distribution of large mammals may be a sensitive measure of human impacts on ecosystems (Morrison et al., 2007).
In particular, large carnivores, which are often considered flagship, umbrella or indicator species, have been used as conservation tools for gauging and preventing loss of biodi- versity (Ray, 2005). Bears, as omnivores, appear to be more
resistant to environmental perturbations than other large Carnivora that depend on a much narrower meat diet (Fredriksson et al., 2007). As such, the loss of bears could signify significant environmental degradation and/or heavy poaching. In Asia, bears are poached for traditional medicine (mainly bile from their gallbladder), food (meat and paws) and pets (Servheen, 1999; Shepherd & Nijman, 2008). A high proportion of the users and consumers of these bear products live in China.
1The Center of Nature and Society, College of Life Sciences, 5 Yiheyuan Road, Haidan, Peking University, Beijing, 100871, China,
2Conservation Research Center, Smithsonian National Zoological Park, Front Royal, VA, 22630, USA,3Minnesota Departmant of Natural Resources, Ground Rapids, MN, 55744, USA,4Sichuan Forestry Department, 15, Renminbeilu Road, Chengdu, Sichuan, 61008, China,5Wanglang National Nature Reserve, 88 South Street, Long’an Town, Pingwu, Sichuan, 622550, China
*Correspondence: Fang Liu,
The Center of Nature and Society, College of Life Sciences, 5 Yiheyuan Road, Haidan, Peking University, Beijing 100871, China.
E-mail: [email protected]
ABSTRACT
AimTo create a fine-scale map of the distribution of Asiatic black bears, identify landscape variables affecting the spatial range of this species and assess population trends using presence–absence data and opinions of local villagers.
LocationSichuan Province, south-western China.
MethodsWe divided the province into 15·15 km cells, stratified them by forest cover, elevation and road density and randomly selected 494 cells (21% of province) for surveys. In each cell, we interviewed villagers and ground-verified their reports of bear presence. We ground-truthed reports of bear absence by conducting transects for bear sign in the best available habitat. We used logistic regression to identify key variables affecting presence of bears and predict their occurrence in unsampled cells.
ResultsWe detected bears in 360 cells (73%). Models correctly predicted bear occurrence in 90.3% of cells where we detected bears and 84.5% of sampled cells where bears were absent. Models predicted 42.7% of Sichuan to be occupied by bears. Bear occurrence was strongly related to forest cover throughout the province. Roads had a negative effect in western region of province. Agricultural lands had a negative effect only when they were distant from forests. Villagers were accurate in their knowledge of bear presence or absence. Interviewed villagers (n= 1816) thought that bears were increasing in 32%, stable in 10%, and decreasing in 58% of cells with bears. Where bear populations were perceived to be declining, villagers identified poaching as the most common cause.
Main conclusionsOur fine-scale distribution map can be used for future monitoring and the key landscape factors related to occupancy by bears can be used in management plans for this species. Interviewing local villagers is an efficient and reliable means of assessing distribution, and changes therein, for animals such as bears that often interact with people and leave obvious signs.
Keywords
Asiatic black bear, China, distribution map, poaching, presence–absence data.
Killing bears for their gallbladder and paws dates back at least 2000–3000 years in China (Xu, 1992; Wu, 1994).
Although four species of bears exist in China [brown bear (Ursus arctos), Asiatic black bear (Ursus thibetanus), sun bear (Helarctos malayanus) and giant panda (Ailuropoda melanol- euca)], the gallbladders of Asiatic black bears and, to a lesser extent, brown bears have been most heavily sought for Chinese traditional medicine (Wu, 1994; Xuet al., 1998). Asiatic black bears are included in CITES Appendix I, Vulnerable in the IUCN Red List and under Class 2 of China’s wildlife protection law.
China still contains the largest portion of the distributional range of Asiatic black bears (Garshelis & Steinmetz, 2007).
Black bears were once distributed throughout north-eastern, eastern, and central China and absent only in the steppe and semi-desert habitats of Inner Mongolia and western China, where the generally open habitat is more conducive for brown bears (Ma et al., 1994). Asiatic black bears occupy varied broad-leaf, coniferous and mixed forests (Reid et al., 1991;
Wanget al., 1995; Izumiyama & Shiraishi, 2004). Decades ago, a rapid increase in the human population and intensive development of agriculture caused the extirpation of Asiatic black bears from large portions of the original range (Ma & Li, 1999). In 1998, China initiated two of the world’s largest ecological rehabilitation projects – the Natural Forest Conser- vation Program and the Sloping Land Conversion Program (Xuet al., 2006; Renet al., 2007), both of which were expected to increase potential black bear habitat and thereby the abundance of bears.
No reliable population estimates exist for Asiatic black bears in China. Values that have been postulated, ranging from 17,000 to 47,000, were based on incidence of bear sign, interviews with local people and estimates by local government officials (Fan & Song, 1997; Hou & Hu, 1997; Maet al., 2001).
The large uncertainty in these estimates makes them of little value in monitoring population trends (Garshelis, 2002).
In the absence of reliable population estimates, trend monitoring may be more effectively accomplished by ascer- taining changes in distribution (i.e. presence–absence; Pollock, 2006), as has been done for other species of bears (Mattson &
Merrill, 2002; Posillico et al., 2004; Ratnayeke et al., 2007).
However, whereas recent distribution maps for Asiatic black bears in China may be more reliable than population estimates, they are of too coarse a scale (Yang & Li, 1996; Ma & Li, 1999;
Zhang, 1999) to be useful for assessing changes in distribution.
Our objectives were 4-fold. First, we sought to create a ground-truthed map of the distribution of Asiatic black bears using presence–absence data at a scale equivalent to the average home range of the species. An accurate, fine-scale map of bear distribution would be valuable as a baseline for future trend monitoring. Second, we used this occupancy map to identify landscape variables that have affected and will continue to affect occurrence of this species. This would be helpful in assessing the effectiveness of current land-use programmes.
Third, we ascertained population status by conducting semi- structured interviews of local villagers and asking them about
their perceptions of population change (including recent extirpation or appearance) and major factors causing such change. Finally, we used our map to assess the conservation needs of black bears with respect to giant pandas and other endangered species living the same habitats.
METHODS Study area
Our survey was conducted during non-winter months from 2005 to 2007 in Sichuan Province, south-western China (9721¢–10831¢E, 2603¢–3419¢N). This province occupies an area of 485,000 km2and supports a human population of 87 million (Sichuan Statistics Dept, 2007). Sichuan also is believed to contain the largest provincial geographical distri- bution of Asiatic black bears in China (Yang & Li, 1996; Gong
& Harris, 2006).
Sichuan Province has been divided into six climate-physio- graphical regions (Zhao et al., 1989; Fig. 1). The eastern region, consisting of the Sichuan Basin and its surrounding low mountains, is one of the most productive agricultural areas of China, with more than 90% of the human population of the province (Sichuan Statistics Dept, 2007). Around the north- eastern and south-eastern margins of the basin are low mountains (1000–2000 m), whereas the western flank of Sichuan Basin contains rugged mountains (Minshan and Qionglai Mountains), which are extensively covered with natural forests. The vegetation in the eastern mountains is deciduous, evergreen or mixed broadleaved-conifer forests. For the purposes of our sampling and analysis, we combined these four areas (see areas a, b, c, and d in Fig. 1) into an eastern region. The southern region of Sichuan has deep river valleys and high plateaus (c.3000 m) in the south-western corner, the Liangshan Mountains in the north-east and the agricultural plains in the south. The natural vegetation shifts, with increasing elevation, from evergreen oak forest to mixed conifer and oak, to conifer, evergreen shrub, and grasslands.
Extensive plantations of conifer, mainly Pinus yunnanesis(Li et al., 1997), have been planted in the lower elevations. The western region encompassed a portion of the Tibetan Plateau (3500–4500 m). The vegetation in this region is a mixture of coniferous forest, evergreen shrub and grasslands. The exten- sive forests remaining in the mountains of Sichuan Province provide highly suitable habitats for Asiatic black bears.
Sampling method
We divided the province into 15·15 km cells (n= 2337;
Fig. 1) using ArcGIS [version 9.1; Environmental Systems Research Institute (ESRI), Inc., Redlands, CA, USA]. This grid- cell size was selected as being roughly the maximum home range size observed for Asiatic black bears (Hwang, 2003;
K. Yamazaki, pers. comm., 2008). We derived a landuse map from a supervised classification of Landsat 7 TM images (29 October 1999–15 July 2002) usingerdas imagine8.6 (version
8.6; ERSI, Inc.). Landuse was classified into seven classes:
forest, shrub, grassland, agriculture, water, cloud and other non-forest. We used ArcGIS 9.1 to calculate the percentage of each land cover in each grid cell. We obtained a mean elevation for each grid cell from 1 : 250,000 DEM [National Geomatics Center of China (NGCC), 1999, unpublished] and the amount of primary and secondary roads and number of townships from a digitized 1 : 250,000 topographical map (NGCC, 1999).
We used the Near function of ArcGIS 9.1 to derive the distances from the centre point of cells to major roads and the edge of the closest nature reserves.
We initially focused sampling on grid cells with > 20%
forest cover (862 cells), and divided the cells into classes on the basis of three factors: percentage forest cover, mean elevation and length of roads, with three ranks of each factor (total of 27 classes). We chose 40% of each class for sampling (338 cells).
After 2005, we added cells with forest cover < 20%, for a total survey of 494 cells or 21% of the cells of the province.
Field survey
Our field survey included village interviews and observation of bear sign. Upon arrival in a survey cell, we interviewed local villagers who regularly used the forest and enquired whether bears were present in the area. We also raised related questions about population trend over the past 10 years. For verification
of occupancy, we requested knowledgeable villagers to lead us to nearby evidence of bear activity (Fig. 2). We considered claw marks on trees, feeding platforms, faeces and footprints to be positive evidence of bear activity (Hwang et al., 2002;
Huygens et al., 2003). Where bear activity was evident, we recorded the type of bear sign, its GPS location and marked the cell as occupied.
Figure 1 Cells (15·15 km) where Asiatic black bears were detected (n= 360) and deemed to be absent (n= 134) during our 2005–07 survey (blank cells were not surveyed) within six climate-physiographical regions of Sichuan Province, China. For analysis, we combined region a (Sichuan Basin) with regions b, c and d into a single Eastern region. The two other regions were the Western Plateau (e) and Southern Mountains and Plains (f) (using WGS 84 projection).
Figure 2 Research team verifying bear sign with local villagers. In this case, the villager brought us to a tree with a feeding platform created by an Asiatic black bear.
A
15x15 Km cells Bear absent cells Bear present cells Border of physiographic areas Border of Sichuan province
100 50 100 Km
If local villagers could not show us bear sign, we conducted our own semi-random sign surveys within the best suitable forest within the cell. We preferentially chose oak forests because black bears climb oak trees in the fall to consume acorns, and their claw marks on tree trunks, and the branches that they break in the canopy remain evident for a year or more. We searched for sign along five transects (100·20 m) in the best bear habitat that we could find. Transects were conducted along elevational contours. Three people, spread across the width of the transect, walked up to and closely inspected every tree within 10 m of the transect line, looking for bear claw marks and also examined the ground for scats, tracks or digging signs. If we found positive evidence of bear activity along any transect, we considered the cell occupied;
otherwise, we considered it unoccupied.
Distribution model
We used binary logistic regression inspss(version 15.0; SPSS Inc, Chicago, IL, USA) to identify environmental factors associated with occupied vs. unoccupied cells. Due to the size of Sichuan Province and the large regional differences in topography and vegetation, we derived separate models for each of our three defined physiographical regions (Fig. 1). We constructed models using the following attributes of each cell:
percentage coverage of forest, agriculture, grassland and shrub, mean elevation, total road length, number of townships and distance from the centre of each cell to the closest major road and the closest nature reserve. The environmental variables (cover and elevation) were selected based on findings of previous Asiatic black bear studies (Reidet al., 1991; Izumiy- ama & Shiraishi, 2004). Road length, distance to the closest major road and number of townships were the best available (surrogate) measures we could obtain for human density. We considered distance from the cell centre to the boundary of the closest nature reserve as the best available surrogate measure for conservation effort within the cell. We considered all biologically relevant combinations of these variables and used Akaike’s Information Criterion (AIC) (Akaike, 1981) within the logistic regression as an indicator of fit. We ordered models by change in AIC (DAIC) and considered models with DAIC < 2 to have a substantial and equivalent level of empirical support (Burnham & Anderson, 2002). We evaluated the number of occurrences for each variable in the top models (i.e. withDAIC < 2) and selected the most common variables to build a candidate model for each region. We assessed the effect of our covariates on bear occurrence by examining the parameter estimates of each covariate within the final model for each region.
To predict occupancy of unsurveyed cells using the models, we determined a cut-off value for probability of occurrence, above which we considered the cell to be occupied. We identified the optimum cut-off value separately for each region by applying the models to surveyed cells and calculating and graphing correct classification rates for potential cut-off values at 0.1 increments (0.1–1.0). Low values would maximize
correct classification rates for cells where bears were known to be present (because none were falsely excluded), but incor- rectly classify many absence cells as being occupied; conversely, high cut-off values would correctly classify most absence cells, but incorrectly exclude some presence cells. We graphed the correct classification rates for both presence and absence cells and considered the point where these curves intersected to be the optimum cut-off value for bear occurrence.
We used 10-fold cross validation (Burman, 1989) to test the predictive performance of the model. First, we randomly divided the observed data into 10 equal-sized subsamples. We then used nine subsamples to develop each model and predicted the probability of occurrence of the remaining subsample. We conducted this process 10 times for each model and calculated the overall classification rates for both presence and absence cells in each region: comparable classification rates would suggest that the model and predictions were fairly stable for the data.
We applied these predictive models to estimate the prob- ability of occurrence for bears in each cell within each region.
We input the attribute data from 2025 cells, after excluding 291 cells on the edges of the province that lacked complete information, and 21 cells that had excessive cloud cover on the satellite image.
Population status and trend
If bears were present, we asked villagers in each cell (X¼4 people per cell) their opinion of the population trend, as well as likely reasons for this trend. If bears were not present, we asked when they had become extirpated and the likely reasons for extirpation.
We scored each villager’s opinion of population trend as decreasing ()1), stable (0), or increasing (1) and then averaged these scores for all interviews within each cell (excluding those who offered no opinion). The range of mean scores was)1 to 1. We considered cells with mean scores > 0.1 as having increasing bear populations (as perceived by local people), cells with scores from)0.1 to 0.1 as having stable bear populations and cells with mean scores <)0.1 as having declining bear populations.
RESULTS
Distribution map and landscape variables
We conducted 1816 interviews of villagers and found evidence of Asiatic black bears in 360 grid cells and no evidence of bears in 134 cells (Fig. 1). We used 334 occupied and 118 unoccupied cells with adequate habitat information to con- struct logistic regression models for all relevant combinations of variables (see Appendix S1 in Supplementary Material), and examined the composition of top models (DAIC values < 2, Table 1). We selected the following variables to build the final models for the eastern region: forest cover (%), mean elevation and distance to the closest major road and nature reserve.
We selected the following variables for the final model in the southern region: forest cover, agricultural cover, shrub cover and distance to the closest nature reserve. We constructed the final model for cells in the western region with the following variables: forest cover, agricultural cover, length of road, number of towns and distance to the closest major road and nature reserve. Omnibus tests of final models for each region indicated that the joint predictive ability of all the covariates was significant (eastern region: v2= 172.93, d.f. = 4, P< 0.001; western region: v2= 86.67, d.f. = 6, P< 0.001;
southern region: v2= 58.60, d.f. = 4, P< 0.001). Goodness- of-fit tests for each model revealed no significant differences between predicted and observed occurrence of black bears for the eastern (Hosmer and Lemeshow v2= 2.00, d.f. = 8, P= 0.998), the southern (Hosmer and Lemeshow v2= 8.74, d.f. = 8,P= 0.365) and the western (Hosmer and Lemeshow v2= 5.67, d.f. = 8,P= 0.684) models. A 10-fold cross-valida- tion indicated that these models correctly classified 90.3% of cells that we knew to be occupied by bears and 84.5% of unoccupied cells (Table 2).
Bear occurrence was positively related to forest cover in all three regions (Table 3). In the eastern region, cells in higher elevation also were more apt to be occupied by bears. In the southern region, in addition to forest cover, shrub cover was a positive predicator of the presence of bears, whereas agricul- tural cover had the strongest negative effect on bear occur- rence. Conversely, in the western region, agriculture was positively related to the presence of bears. In the western region, total length of roads within cells negatively affected bear occurrence, whereas the distance to the closest major road positively affected bear occurrence (Table 3). We found no significant effect of the proximity of nature reserves on bear occurrence in any region, although the relationship was marginally close to significant in the eastern and western regions (Table 3). This result may be partially explained by the tight relationship between distance to the closest reserve and forest cover. In the eastern region, most reserves were formed to conserve forests used by giant pandas and less forest is found distant from reserve boundaries (Pearson correlation =)0.57, P< 0.001). In the western region, most reserves were formed to conserve grassland species on the Tibetan Plateau (Fig. 4) and the forest along the eastern edge of the region are far from these reserves (Pearson correlation = 0.155,P= 0.047). There are few reserves within the southern region (Fig. 4), most forests are coniferous plantations and we found no strong relationship between forest cover and distance to a reserve (Pearson correlation =)0.065,P> 0.1).
We estimated that 42.7% of the land area of the province was occupied by black bears on the basis of our calculated optimal cut-off values for probability of occurrence: 0.63 for both eastern and southern regions, 0.71 for the western region (Fig. 3); the core range for bears was located within the Minshan and Qionglai Mountains (region d in Fig. 1).
Probability of occurrence was very low for the Sichuan basin and the surrounding low hills in the eastern portion of the province (Fig. 3). Among 864 cells predicted to be occupied by Table 1 Logistic regression models for landscape and environ-
ment variables used to predict the presence or absence of Asiatic black bears in 15·15 km cells in three regions of Sichuan Province, China, 2005–07. Models are ranked in order of their AIC.
Area Model variables AIC DAIC K
Eastern region
FOR, ELE, TW, RDD, NRD 44.411 0 5
FOR, ELE, RDD, NRD 44.883 0.472 4
FOR, ELE, TW, RDD 45.411 1 4
FOR, GRS, ELE, TW, RDD, NRD 45.843 1.432 6 FOR, AGR, ELE, TW, RDD, NRD 46.209 1.798 6
FOR, ELE, RDD 46.262 1.851 3
FOR, SHB, FOR, ELE, TW, RDD, NRD
46.315 1.904 6 FOR, ELE, RDL, RDD, NRD 46.316 1.905 5 FOR, SHB, FOR, ELE, RDD, NRD 46.39 1.979 5 FOR, GRS, ELE, RDD, NRD 46.406 1.995 5 FOR, ELE, RDL, TW, RDD, NRD 46.41 1.999 6 Southern
region
FOR, AGR, SHB, NRD 82.180 0.000 4
FOR, GRS, SHB, NRD 82.464 0.284 4
FOR, AGR, SHB 82.472 0.292 3
FOR, AGR, GRS, SHB 82.730 0.550 4
FOR, AGR, GRS, SHB, NRD 82.746 0.566 5
FOR, GRS, SHB 82.880 0.700 3
FOR, GRS, SHB, ELE, NRD 83.311 1.131 5 FOR, AGR, SHB, ELE, NRD 83.352 1.172 5
FOR, GRS, SHB, ELE 83.396 1.216 4
FOR, AGR, SHB, ELE 83.532 1.352 4
FOR, AGR, SHB, TW, NRD 84.155 1.975 5 FOR, AGR, SHB, RDD, NRD 84.168 1.988 5 Western
region
FOR, AGR, GRS, SHB, RDD 98.201 0.000 5 FOR, AGR, GRS, SHB, RDL, TW,
RDD, NRD
98.423 0.222 8 FOR, AGR, GRS, SHB, RDD,
NRD
98.937 0.736 6 FOR, AGR, RDL, TW, RDD, NRD 98.938 0.737 6 FOR, AGR, GRS, SHB, RDL, TW,
RDD
99.083 0.882 7 FOR, AGR, SHB, RDL, TW, RDD,
NRD
99.127 0.926 7 FOR, AGR, GRS, SHB, TW, RDD 99.236 1.035 6 FOR, AGR, GRS, SHB, RDL, RDD,
NRD
99.244 1.043 7 FOR, AGR, RDL, RDD, NRD 99.269 1.068 5 FOR, AGR, GRS, SHB, RDL, RDD 99.284 1.083 6 FOR, AGR, GRS, SHB, ELE, RDD 99.856 1.655 6 FOR, AGR, GRS, RDL, TW, RDD,
NRD
99.856 1.655 7 FOR, AGR, GRS, SHB, TW, RDD,
NRD
99.951 1.750 7 FOR, AGR, GRS, RDL, RDD, NRD 99.987 1.786 6 FOR, AGR, SHB, RDL, RDD, NRD 100.071 1.870 6 FOR, AGR, RDL, RDD 100.177 1.976 4 AGR, % agriculture cover; ELE, mean elevation; FOR, % forest cover;
GRS, % grassland cover; NRD, distance from centre of cell to edge of the closest nature reserve; RDD, distance from centre of cell to the closest major road; RDL, total length of road; SHB, % shrub cover;
TW, number of towns;K, number of variables in the model.
bears, 32.4% had at least 10% of the cell area within a nature reserve (Fig. 4).
Status and trend of population
Local people we interviewed perceived black bear populations to be increasing in 32%, stable in 10% and decreasing in 58%
of cells that were occupied (n= 288 cells, 855 respondents).
A higher percentage of cells with black bear populations that were perceived to be increasing or stable occurred in the western region (n= 53 cells; 51%), than in either the eastern (n= 36 cells; 34%) or southern (n= 31 cells; 46%) regions, but this difference was not significant (v2= 4.51, d.f. = 2, P= 0.11). People we interviewed indicated that bears in the local area had been extirpated sometime during their lifetimes for 33% of the cells where bears were absent during our survey (n= 130 cells); in the remainder, bears had been extirpated
prior to the residency of current villagers. The cells where bears had been recently extirpated were mostly distributed in the north (both eastern and western corners) and south-central portions of the province. No local people reported the expansion of bears into their area within their lifetime.
Although forest cover was strongly related to bear occur- rence, it was poorly associated with the current population trend. A mix of increasing, stable and declining bear popula- tions (as perceived by local people) existed at all levels of forest cover (Fig. 5a). Moreover, bears had been extirpated in some areas with high (up to 49%) forest cover. Likewise, although agriculture was negatively associated with bear occurrence, some populations that were perceived to be increasing or stable existed in cells with significant cropland (up to 55%; Fig. 5b);
90% of purportedly increasing populations occurred in cells with < 15% agriculture, but 85% of declining populations occurred in cells with little agriculture and extirpation occurred at all levels of agricultural coverage (0–84%). There was no association between perceived bear population trend and the distance to the closest nature reserve or road (ANOVA, F2,262= 0.15,P= 0.863).
Villagers who believed that bear populations were increasing in their local area (n= 350 respondents) commonly attributed the increases to the enactment of a 1996 gun prohibition (49%), more stringent wildlife protection laws and increased protected areas (35%) and reforestation (13%). Villagers who perceived bear populations to be declining (n= 343 respon- dents) indicated that poaching activities (72%) and habitat loss due to logging, mining and new roads (23%) were the primary causes. In 43 cells where bears had been extirpated during the villagers’ lifetime, interviewees (n= 127) most often thought that logging (35%), poaching (30%), conversion of forest into cropland (22%) or building of roads (13%) were the reasons.
DISCUSSION
We succeeded in creating a reliable map of the distribution of Asiatic black bears within Sichuan Province at a scale (15·15 km cells; Figs 3 & 4) more appropriate for conser- vation than previous maps, which recorded bear presence at the scale of the county (Jianget al., 1994; Yuet al., 1994; Yang
& Li, 1996; Zhang, 1999) or even coarser (Ma & Li, 1999). Due to the generally large and highly variable size of the counties of Table 2 Estimates of correct classification rates of logistic regression models to predict occupancy of Asiatic black bears in Sichuan Province, China, based on 10-fold cross-validation.
Area
Mean % correct prediction rate for occupied cells Mean % correct prediction rate for unoccupied cells
N XSE Range N XSE Range
Eastern region 122 95.8 ± 0.001 95.1–96.7 51 94.7 ± 0.005 92.2–96.1
Southern region 82 90.1 ± 0.007 85.9–93.8 31 73.9 ± 0.008 71.0–77.4
Western region 130 85.3 ± 0.006 82.3–88.5 36 79.2 ± 0.008 72.5–80.6
Overall 334 90.3 ± 0.002 89.4–91.3 118 84.5 ± 0.005 82.0–86.4
N, number of cells.
Table 3 Parameter estimates of logistic regression models to relate landscape and environment variables to the presence or absence of Asiatic black bears in 15·15 km cells in three regions of Sichuan Province, China, 2005–07.
Area Variable
Parameter estimate
Standard
error Waldv2 P-value Eastern
region
FOR 14.106 5.262 7.186 0.007
ELE 0.005 0.002 8.085 0.004
RDD )8.409 4.669 3.243 0.072
NRD )1.906 1.090 3.060 0.080
Southern region
FOR 5.238 1.906 7.552 0.006
AGR )9.418 2.852 10.905 0.001
SHB 12.794 4.644 7.589 0.006
NRD )2.837 1.880 2.279 0.131
Western region
FOR 17.065 3.941 18.749 < 0.001
AGR 86.282 33.868 6.490 0.011
RDL )3.575 1.333 7.192 0.007
TW 0.088 0.058 2.259 0.133
RDD 4.633 1.815 6.516 0.011
NRD 5.960 3.193 3.485 0.062
AGR, % agriculture cover; ELE, mean elevation; FOR, % forest cover;
NRD, distance from centre of cell to the edge of the closest nature reserve; RDD, distance from centre of cell to the closest major road;
RDL, total length of road; SHB, % shrub cover; TW, number of towns.
Sichuan Province (range 167–20,994 km2), unclear or unspec- ified methodology of these surveys, and inconsistencies among these various reports, the previous maps are of little practical value for conservation planning or monitoring. Our map is a lower cost alternative to population estimates (Joseph et al., 2006) and a useful and reliable tool for large-scale monitoring, as area of occupancy has been positively linked to abundance through a range of spatial scales (Holtet al., 2002). For future monitoring programmes, we recommend attention to the edge of the present distribution, as the distribution edge is usually
dynamic (Murcia, 1995; Revillaet al., 2001) and will probably be most sensitive to changes in bear population status.
The most important predictors of bear occurrence were somewhat different in each region (Table 3). Forest cover was the only significant variable that positively affected bear occurrence in all three regions, indicating the importance of forest retention for conservation of black bears.
Unexpectedly, agricultural cover was positively associated with bear occurrence in the western region, but negatively related to bear presence in the southern region. This contra- Figure 3 Estimated probability of black
bear occurrence (2005–07) in 15·15 km cells in Sichuan Province, China, based on logistic regression models (using WGS 84 projection).
Figure 4 Predicted distribution map of Asiatic black bears in Sichuan Province (cells) relative to nature reserves, and range of giant pandas (using WGS 84 projection).
A
Border of Sichuan
Probability of bear occurrence 0. 00 - 0.10
0.10 - 0. 20 0. 20 - 0. 30 0. 30 - 0. 40 0. 40 - 0. 50 0. 50 - 0. 60 0. 60 - 0. 70 0. 70 - 0. 80 0. 80 - 0. 90 0. 90 - 1. 00
A
100
• Km
Predicted occupied cells of black bear I | Nature reserves
Range of giant panda Border of Sichuan
dictory result may have been due to the differing location of agricultural fields with respect to forests in each region: in the southern region, agricultural areas are mainly distributed at lower elevations distant from forest, whereas crops in the western region existed adjacent to forests.
Shrub cover was a positive predicator of bear presence in the southern region, possibly because the soft mast from these plants is especially important in a region dominated by coniferous forests, with less hard mast. The natural forest in this region was converted to coniferous plantations during the 1950–90s; a new forest protection policy instituted in 1998 has enabled regeneration of young natural forests (Xuet al., 2006), some of which might have appeared to be shrubs on satellite images.
The two measures of roads were only significant in the western region, indicating that recent construction and improvement of roads within the region are likely to have negative impacts on bears, as have been shown for many other threatened species (Wilcoveet al., 1998).
We were surprised that distance to reserves was not a significant predictor of bear occurrence in our models.
However, as we describe in the results, there was a strong relationship between forest cover and the location of reserves, such that reserves, at least in the eastern region, encompass much of the large tracts of remaining forest (Fig. 4). In the western region, reserves are mostly confined to the higher elevations of the Tibetan Plateau, beyond the distribution of trees and therefore beyond the distribution of bears. Reserves may serve to conserve forest, but we could find no impact of their conservation effort on bear distribution beyond the boundaries of the reserves.
It is important to note some caveats related to our map.
First and foremost, much of our map is predicted from a model and models are always imperfect; the only truly known occupancy was in the cells we actually sampled (cf. Fig. 1 vs.
Fig. 3). Predicted distribution maps may be prone to error, especially for generalist species like a black bear (Jime´nez- Valverdeet al., 2008). Notably, our habitat classification did not differentiate between coniferous and deciduous forest due to the limitations of satellite imagery in complex topography.
Coniferous forest, both plantation and natural, was usually not a suitable habitat for black bears on the basis of our observations and village interviews. Accordingly, the logistic regression model for the southern region had the lowest correct classification rate of the three regions (Table 2), indicating that we may have overestimated the amount of bear habitat in the south, where there are extensive stands of plantation conifer (e.g.Pinus yunnanesis; Liet al., 1997). The model also may have overestimated bear distribution in the north-west, where there are extensive natural conifer stands (e.g.Picea asperataandPicea balfourian; Jia, 1989). However, black bears do occur in portions of the western plateau, probably due to the occurrence of alpine oaks (Quercusspp.), which were intermixed with the coniferous forest in some regions (Peng, 1998).
We also may have overestimated the amount of bear habitat simply because of the cell size that we chose for sampling. We observed in many cases that bears only lived in a small portion of a cell (e.g. at the edge, adjacent to another occupied cell), yet for our analysis, the entire cell was categorized as occupied.
Although our cell size made sense in terms of bear biology and the practicality of sampling, a smaller cell size almost certainly would have indicated a smaller occupied area.
We consider it improbable that we underestimated occu- pancy by failing to find evidence of bears that lived within cells (unless they were present in only a tiny portion of a cell). In a companion study, we observed that if bears were present, the probability of not detecting their sign in five 100·20 m transects in a hardwood forest was < 1% (Andrew Trent, Virginia Tech, unpubl. data). In this study, we found that villagers were accurate in the information they provided on bear presence. In all cases where people indicated that bears existed in the local area, we found recent sign (Fig. 2) and we encountered only a single case where local people indicated bears were absent, but we found sign of their existence.
(a)
(b)
Figure 5 Percentage of cells with increasing (n= 90), stable (n= 29), decreasing (n= 165) (based on perceptions of villagers) and extirpated (n= 35) populations of Asiatic black bears in categories of different (a) forest cover and (b) agricultural cover, Sichuan, China, 2005–07 (using WGS 84 projection).
45 Increase Stable 40
35 m 30
CD
B
25I 20
oa. 15 10
5 0
100 90 80 70 60 50 40 30 20 10 0
Decrease Extirpated
0-15 16-30 31-45 46-100 Percent of forest cover
Increase Stable Decrease Extirpated
0-15 16-30 31-45 46-100 Percent of agricultural cover
Monitoring techniques based on indices are notoriously variable (Gibbs, 2000), but a combination of interviews with local people and ground-truthed surveys can result in near- perfect detection of animals such as bears that are large, destroy crops and leave obvious signs in the forest. These data can thus yield accurate distribution maps, including informa- tion on limiting factors (Chauhan & Singh, 2006; Ratnayeke et al., 2007) and assessments of population trend (Steinmetz et al., 2006).
We relied on the opinion of villagers for population trends because villagers, as a sedentary population, have a long history of interactions with wildlife and often have a sense of their commonness or rarity (Berkeset al., 2000; Huntington, 2000).
We are aware that perceptions of population trend can be flawed: villagers may have perceived bears to be declining mainly because they became aware of bears recently being killed by poachers or as nuisances; conversely, they may have perceived populations to be increasing mainly because they saw more bears in their cornfields, or they thought that government reforestation programmes should be working.
This may explain why the habitat features that explained occupancy did not correlate well with perception of population trend. On the other hand, poaching is likely to be a key issue affecting population trend independent of habitat, so impres- sions of population trend based on bear interactions with local people may have been reasonably accurate. In areas where bear populations were thought to be increasing or stable, people ascribed this improved status mainly to the control of guns, despite the continued use of traps and poison.
Our results indicate that a majority of local extirpations of bears occurred before the establishment of the present forest conservation laws and protected area system. Natural forest conservation and reforestation programmes are likely to have a favourable impact on black bears; however, we found no evidence of range extension for bear populations in the 10 years since the implementation of these programmes and declining bear populations occurred in many areas with good habitat. We suggest two explanations for what appears to be a lack of response of black bears to these conservation initiatives: (1) nut-bearing trees (principally oaks), which are vital for these bears (Reid et al., 1991; Hashimoto et al., 2003), either were not planted (most of the planted trees were coniferous species andEucalyptus; Renet al., 2007) or have not reached maturity in the short time since these reforestation programmes were implemented; and (2) poaching continues to threaten bear populations. Poaching for the trade of wildlife parts has caused dramatic declines for many large mammal species, such as tigers (Pantbera tigris; Kenney et al., 1995), rhinos (e.g. Diceros bicornis and Ceratotherium simum; Emslie & Brooks, 1999), elephants (e.g.Loxodonta africana; Blake & Hedges, 2004) and Tibetan antelope (Pantholops hodgsoni; Harris et al., 1999). Our survey indicates that any potential improvement in habitat due to recent forestry policies is not yet sufficient to overcome human-caused mortality in many places through- out the province.
Despite their important conservation status both nationally and internationally, no reserves have been established in Sichuan Province specifically for the management of Asiatic black bears. Fortunately, black bears and giant pandas are sympatric in the mountainous forests along the eastern edge of the Tibetan Plateau (Schalleret al., 1989; Fig. 4); thus, nature reserves established to protect giant pandas and their habitat (Louckset al., 2001; Lu¨ & Liu, 2004) also afford some security for black bears. A network of 32 nature reserves protects more than 95% of the remaining panda habitat in Sichuan Province (State Forestry Administration, 2006). Our data indicating that Asiatic black bears are widespread in the central portion of Sichuan Province, where the majority of these reserves are located, support the premise that creating more reserves for flagship species (in this case the giant panda) aid in the persistence of other threatened species (Leader-Williams &
Dublin, 2000). By the same token, extending active protection to black bears in southern and western Sichuan Province, through an expanded reserve system and better law enforce- ment against poaching, would better conserve a host of other species in these high diversity ecosystems (Fig. 4; Myerset al., 2000) and would compliment the current emphasis on giant pandas within the province.
ACKNOWLEDGEMENTS
We appreciate the financial support provided by the Smithso- nian Institution, Friends of the National Zoo, World Society of Protected Animals (WSPA), the International Association of Bear Research and Management (IBA), EarthWatch and Animals Asia Foundation. We thank K. Ding (Sichuan University) andL. Zhang(Peking University) for assisting with the fieldwork and Andrew Trent (Virginia Tech) for providing the data on detection probability of bear signs. We thank Sichuan Forestry Department for giving us permission to do the survey and for logistical support. All the forestry bureaus of counties helped us during the fieldwork. Several graduate students in the Center of Nature and Society of Peking University, particularlyS. Li,X. LiuandJ. Liu, helped us with interpretation of satellite images and data analysis. We thank DrsS. RatnayekeandM.-H.Hwang for helpful comments on the manuscript and analysis that significantly improved it. We could not have accomplished this work without the help from the directorS. Jiangand the staff of Wanglang Nature Reserve, notably,L. Zhao,L. Shao, L. Yan, Y. Zhang,J. Liu, H. Zhou, S. Zhang,Y. Wen,G. Lan,Z. Yuan,J. Zhao,C. Liang,A. Long, J. CuiandZ. Zhou; they were our best comrades in the forest.
REFERENCES
Akaike, H. (1981) Likelihood of a model and information criteria.Journal of Econometrics,16, 3–14.
Berkes, F., Colding, J. & Folke, C. (2000) Rediscovery of tra- ditional ecological knowledge as adaptive management.
Ecological Applications,10, 1251–1262.
Blake, S. & Hedges, S. (2004) Sinking the flagship, the case of forest elephants in Asia and Africa.Conservation Biology,18, 1191–1202.
Burman, P. (1989) A comparative study of ordinary cross- validation, V-fold cross-validation and the repeated learn- ing-testing methods.Biometrika,76, 503–514.
Burnham, K.P. & Anderson, D.R. (2002)Model selection and multimodel inference: a practical information-theoretic approach, 2nd edn. Springer Press, New York.
Chauhan, N.P.S. & Singh, R.K.J. (2006) Status and distribution of sun bears in Manipur, India.Ursus,17, 182–185.
Emslie, R. & Brooks, M. (1999)Status survey and conservation action plan: African rhino. IUCN, Gland, Switzerland.
Fan, Z. & Song, Y. (1997) Bears present status and con- servation, and bear farms of China. Proceedings of the Second International Symposium on the Trade of Bear Parts (ed. by D.F. Williamson and A.L. Gaski), pp. 6–19.
TRAFFIC (USA)/World Wildlife Fund (Washington, DC), Seattle, WA.
Fredriksson, G., Steinmetz, R., Wong, S. & Garshelis, D.L.
(2007)Helarctos malayanus, sun bear. 2007 IUCN Red List of Threatened Species. Available at: http://www.iucnredlist.org/
(last accessed February 2008).
Garshelis, D.L. (2002) Misconceptions, ironies, and uncertain- ties regarding trends in bear populations.Ursus,13, 321–334.
Garshelis, D.L. & Steinmetz, R. (2007)Ursus thibetanus, Asiatic black bear. 2007 IUCN Red List of Threatened Species.
Available at: http://www.iucnredlist.org/ (last accessed Feb- ruary 2008).
Gibbs, J.P. (2000) Monitoring populations.Research techniques in animal ecology (ed. by L. Boitani and T.K. Fuller), pp.
213–252. Columbia University Press, New York.
Gong, J. & Harris, R. (2006) The status of bears in China.
Understanding Asian bears to secure their future(ed. by Japan Bear Network), pp. 96–101. Japan Bear Network, Ibaraki, Japan.
Harris, R., Pletscher, D., Loggers, C. & Miller, D. (1999) Status and trends of Tibetan plateau mammalian fauna, Yeniugou, China.Biological Conservation,87, 13–19.
Hashimoto, Y., Kaji, M., Sawada, H. & Takatsuki, S. (2003) Five-year study on the autumn food habits of the Asiatic black bear in relation to nut production.Ecological Research, 18, 485–492.
Holt, A.R., Gaston, K.J. & He, F. (2002) Occupancy-abundance relationships and spatial distribution: a review. Basic and Applied Ecology,3, 1–13.
Hou, W. & Hu, J. (1997) The present situations of China’s bear resources and protection.Journal of Sichuan Teacher’s Col- lege (Natural Science),18, 287–291 (in Chinese).
Huntington, H.P. (2000) Using traditional ecological knowl- edge in science: methods and applications.Ecological Appli- cations,10, 1270–1274.
Huygens, O.C., Miyashita, T., Dahle, B., Carr, M., Izumiyama, S., Sugawara, T. & Hayashi, H. (2003) Diet and feeding habits of Asiatic black bears in the Northern Japanese Alps.
Ursus,14, 236–245.
Hwang, M.-H. (2003)Ecology of Asiatic black bears and people- bear interactions in Yushan National Park, Taiwan. PhD Dissertation, University of Minnesota, Minneapolis.
Hwang, M.-H., Garshelis, D.L. & Wang, Y. (2002) Diets of Asiatic black bears in Taiwan, with methodological and geographical comparisons.Ursus,13, 111–125.
Izumiyama, S. & Shiraishi, T. (2004) Seasonal changes in ele- vation and habitat use of the Asiatic black bear (Ursus thi- betanus) in the Northern Japan Alps. Mammal Study, 29, 1–8.
Jia, M. (1989) The Pinaceaein Sichuan and its geographical distribution of the coniferous forests.Journal of Chongqing Teachers College (Natural Sciences Edition), 6, 61–71 (in Chinese).
Jiang, S., Ao, L., Yang, J., Fan, J. & Li, Y. (1994) The survey of resource of black bear of Leshan Prefecture, Sichuan.Sichuan Journal of Zoology,13, 175–177 (in Chinese).
Jime´nez-Valverde, A., Lobo, J.M. & Hortal, J. (2008) Not as good as they seem: the importance of concepts in species distribution modeling. Diversity & Distributions, 14, 885–
890.
Joseph, L.N., Field, S.A., Wilcox, C. & Possingham, H.P.
(2006) Presence-absence versus abundance data for mon- itoring threatened species.Conservation Biology, 20, 1679–
1687.
Kenney, J.S., Smith, J.L.D., Starfield, A.M. & McDougal, C.W.
(1995) The long-term effects of tiger poaching on population viability.Conservation Biology,9, 1127–1133.
Leader-Williams, N. & Dublin, H. (2000) Charismatic mega- fauna as ‘‘flagship species’’.Priorites for the conservation of mammalian diversity. Has the panda had its day? (ed. by A. Entwistle and N. Dunstone), pp. 53–81. Cambridge University Press, Cambridge, UK.
Li, X., Hu, T., Yang, D., Jiang, X., Jiang, X., Zhang, Q., Liu, Z.
& Wang, F. (1997) Research on stand structure classification of low-benefit protection standPinus yunnanesisin the Panxi Region.Journal of Sichuan Agriculture University, 15, 236–
240 (in Chinese).
Loucks, C., Lu¨, Z., Dinerstein, E., Wang, H., Olson, D., Zhu, C.
& Wang, D. (2001) Giant pandas in a changing landscape.
Science,294, 1465.
Lu¨, Z. & Liu, Y. (2004) China’s national plan for conservation of the giant panda. Giant pandas: biology and conservation (ed. by D.G. Lindburg and K. Baragona), pp. 226–227.
University of California Press, Berkeley, CA.
Ma, Y. & Li, X. (1999) Status and management of the Asiatic black bear in China. Bears: status survey and conservation action plan(ed. by C. Servheen, S. Herrero and B. Peyton), pp. 200–202. IUCN, Gland, Switzerland.
Ma, Y., Hu, J. & Zhai, Q. (1994) Bears of China. Sichuan Science and Technology Press, Chengdu, China (in Chinese).
Ma, Y., Xu, L. & Hu, J. (2001) On the resources and conser- vation of bears in China.Proceedings of the Third Interna- tional Symposium on the Trade in Bear Parts(ed. by D.F.
Williamson and M.J. Phipps), pp. 38–43. TRAFFIC East Asia, Hong Kong, China.
Mattson, D.J. & Merrill, T. (2002) Extirpations of grizzly bears in the contiguous United States, 1850–2000. Conservation Biology,16, 1123–1136.
Morrison, J., Sechrest, W., Dinerstein, E., Wilcove, D. &
Lamoreux, J. (2007) Persistence of large mammal faunas as indicators of global human impacts.Journal of Mammalogy, 88, 1363–1380.
Murcia, C. (1995) Edge effects in fragmented forest: implications for conservation.Trends in Ecology and Evolution,10, 58–62.
Myers, N., Mittermeier, R.A., Mittermeier, C.G., da Fonseca, G.A.B. & Kent, J. (2000) Biodiversity hotspots for conser- vation priorities.Nature,403, 853–858.
Peng, J. (1998) Investigation on Selenarctos thibetanus mu- pinensisresources in the Ganzi Prefecture of Sichuan.Acta Theriologica Sinica,18, 311–313 (in Chinese).
Pollock, J.F. (2006) Detecting population declines over large areas with presence-absence, time-to-encounter, and count survey methods. Conservation Biology, 20, 882–
892.
Posillico, M., Meriggi, A., Pagnin, E., Lovari, S. & Russo, L.
(2004) A habitat model for brown bear conservation and land use planning in the central Apennines.Biological Con- servation,118, 141–150.
Ratnayeke, S., van Manen, F.T., Pieris, R. & Pragash, V.S.J.
(2007) Landscape characteristics of sloth bear range in Sri Lanka.Ursus,18, 189–202.
Ray, J.C. (2005) Large carnivorous animals as tools for con- serving biodiversity: assumptions and uncertainties. Large carnivores and the conservation of biodiversity(ed. by J.C. Ray, K.H. Redford, R.S. Steneck and J. Berger), pp. 34–59. Island Press, Washington, DC.
Reid, D., Jiang, M., Teng, Q., Qin, Z. & Hu, J. (1991) Ecology of the Asiatic black bear (Ursus thibetanus) in Sichuan, China.Mammalia,55, 221–237.
Ren, H., Shen, W., Lu, H., Wen, X. & Jian, S. (2007) Degraded ecosystems in China: status, causes, and restoration efforts.
Landscape and Ecological Engineering,3, 1–13.
Revilla, E., Palomares, F. & Delibes, M. (2001) Edge-core ef- fects and the effectiveness of traditional reserves in conser- vation: Eurasian badgers in Don˜ana national park.
Conservation Biology,15, 148–158.
Servheen, C. (1999) The trade in bears and bear parts.Bears:
status survey and conservation action plan(ed. by C. Servh- een, S. Herrero and B. Peyton), pp. 33–38. IUCN, Gland, Switzerland.
Schaller, G.B., Teng, Q., Johnson, K.G., Wang, X., Shen, H. &
Hu, J. (1989) The feeding ecology of giant pandas and Asi- atic black bears in the Tangjiahe Reserve, China.Carnivore behavior, ecology, and evolution, Vol. 1 (ed. by J.L. Gittle- man), pp. 212–240. Cornell University Press, New York.
Shepherd, C.R. & Nijman, V. (2008) The trade in bear parts from Myanmar: an illustration of the ineffectiveness of enforcement of international wildlife trade regulations.Bio- diversity Conservation,17, 35–42.
Sichuan Statistics Dept (2007) 2007 Sichuan statistical year- book. China Statistics Press, Beijing, China (in Chinese).
State Forestry Administration (2006)The 3rd national survey report on giant panda in China. Science Press, Beijing, China (in Chinese).
Steinmetz, R., Chutipong, W. & Seuaturien, N. (2006) Col- laborating to conserve large mammals in Southeast Asia.
Conservation Biology,20, 1391–1401.
Wang, W., Wei, F. & Hu, J. (1995) A preliminary study of the diet habit of the black bear in Mabian nature reserve.Journal of Sichuan Teacher’s College (Natural Science),16, 100–104 (in Chinese).
Wilcove, D.S., Rothstein, D., Dubow, J., Phillips, A. & Losos, E.
(1998) Quantifying threats to imperiled species in the United States.BioScience,48, 607–615.
Wu, J. (1994) Herbal textual research on bear bile.Journal of Chinese Medical Materials,17, 41–43 (in Chinese).
Xu, X. (1992) Historical and present status of the Asiatic black bear in northeast China. International Conference on Bear Research and Management,9, 53–55.
Xu, H., Sun, X., Wen, F., Zhou, J., Li, S., Ding, T. & Liu, L. (1998) Research and outlook on bear bile.Chinese Journal of Infor- mation on Traditional Chinese Medicine,5, 18–20 (in Chinese).
Xu, J., Yin, R., Li, Z. & Liu, C. (2006) China’s ecological rehabilitation: unprecedented efforts, dramatic impacts, and requisite policies.Ecological Economics,57, 595–607.
Yang, X. & Li, H. (1996) The study on the status of wild black bear resource and conservation strategy of Sichuan Province.
Journal of Sichuan Forestry Science and Technology,17, 58–61 (in Chinese).
Yu, X., Wei, S., Zheng, X., Lan, Z. & Chen, H. (1994) Distri- bution and conservation of resource of black bear of Deyang Prefecture.Sichuan Forestry Science and Technology,15, 63–
65 (in Chinese).
Zhang, R. (1999)Zoogeography of China. Science Press, Beijing, China (in Chinese).
Zhao, E., Wu, G. & Inger, R.F. (1989) Ecological and geo- graphic distribution of the amphibians of Sichuan, China.
Copeia,1989, 549–557.
Editor: Jon Paul Rodriguez
SUPPORTING INFORMATION
Additional Supporting Information may be found in the online version of this article:
Appendix S1 Top 30 logistic regression models for landscape and environment variables used to predict the presence of Asiatic black bears in 15·15 km cells in three regions of Sichuan Province, China, 2005–07. Models are ranked in order of their AIC.
Please note: Wiley-Blackwell is not responsible for the content or functionality of any supporting materials supplied by the authors. Any queries (other than missing material) should be directed to the corresponding author for the article.