Long‐term ecological data for conservation: Range change in the black‐billed capercaillie (Tetrao urogalloides) in northeast China (1970s–2070s)

Abstract Long‐term ecological data can be an effective tool to help ecologists integrate future projections with historical contexts and provide unique insights into the long‐term dynamics of endangered species. However, hampered by data limitations, including incomplete and spatially biased data, relatively few studies have used multidecadal datasets or have examined changes in biogeography from a historical perspective. The black‐billed capercaillie (Tetrao urogalloides) is a large capercaillie (classified as Least Concern [LC] on the IUCN red list) that has undergone a dramatic decline in population during the late 20th century and is considered endangered. Its conservation status is pessimistic, and the species requires immediate protection. Therefore, we supplemented a historical dataset to identify changes in this bird's range and population in northeast China over the long term. The study area spanned Heilongjiang Province, Jilin Province, and the northeast corner of Inner Mongolia in northeast China. We integrated an ecological niche model (BIOMOD2) with long‐term ecological data on this species to estimate the magnitude of change in distribution over time. Our results revealed a 35.25% reduction in the current distribution of this species compared to their potential distribution in the 1970s. This decline is expected to continue under climate change. For example, the future range loss was estimated to be 38.79 ± 0.22% (8.64–90.19%), and the actual state could be worse, because the baseline range of the model was greater than the real range in the 2000s, showing a 12.39% overestimation. To overcome this poor outlook, a conservation strategy should be established in sensitive areas, including the southwestern Greater Khingan Mountains and northern Lesser Khingan Mountains. Actions that should be considered include field investigations, establishing a monitor network, designing ecological corridors, and cooperating with local inhabitants, governments, and conservation biologists to improve the conservation of the black‐billed capercaillie.

Historical ecological data can be used to document species records over time (McClenachan, Ferretti, & Baum, 2012;Turvey et al., 2015;Yang et al., 2016) and help ecologists integrate future perspectives within historical contexts to provide unique insights into the long-term dynamics of endangered species (Beaugrand et al., 2015;Scherrer et al., 2017).
In fact, the application of long-term ecological data, particularly historical data, is often hampered by data limitations, including incomplete and spatially biased datasets (Boakes et al., 2010;Hortal, Jimenez-Valverde, Gomez, Lobo, & Baselga, 2008;McClenachan et al., 2012). Despite recognizing the considerable potential benefits of long-term ecological archives to conservation research, policy, and practices, relatively few studies have analyzed multidecadal datasets or have examined changes in biogeography from a historical perspective (Davies et al., 2014;Li, Waller, & Syphard, 2017;McClenachan et al., 2012;Turvey et al., 2015). Therefore, in conservation efforts, there is a need to provide effective methods to assess the utility and potential limitations of long-term ecological data to develop a meaningful understanding of population dynamics from the past to the future (McClenachan et al., 2012;Yang et al., 2016;Zhang et al., 2016Zhang et al., , 2017. Multiple resources and ecological niche model have proven to be effective methods of reconstructing historical distributions from longterm ecological data (Zhang et al., 2017). Multiple cross-checked data resources can reduce nonstandardized sampling and errors and provide sufficient data for further analysis (Boakes et al., 2010;Fonzo et al., 2013;Ren et al., 2016). Meanwhile, ecological niche modeling (ENM) is useful for constructing species distributions to support conservation efforts (Guisan et al., 2013;Luo, Jiang, & Tang, 2015;Veloz et al., 2015). Discrepancies between different species distribution models (SDMs) can be large, making the choice of the appropriate model difficult (Cuyckens et al., 2016;Elith & Leathwick, 2009;Elith et al., 2011;Renner & Warton, 2013;Veloz, 2009). In this context, ensemble forecasting approaches may be an appropriate choice (Thuiller, Lafourcade, Engler, & Araújo, 2009). BIOMOD is considered a suitable platform for ensemble forecasting of species distributions (Cianfrani et al., 2011;Thuiller et al., 2009). Combining multiple data resources with ecological niche modeling can be used to evaluate a species sensitively and comprehensively by revealing changes that occur over decades.
IUCN's assessment information of black-billed capercaillie (Tetrao urogalloides) is Least Concern (LC) on the red list, its low conservation status is due to the wider distribution of the species in Siberia, and assessment report also clearly pointed out that the population may have declined due to over-hunting in northeast China (http://www. iucnredlist.org/details/22679491/0) (González et al., 2012;Moss et al., 2014;Wang, Ren, He, & Zhu, 2008). Tetrao urogalloides urogalloides is one of three subspecies and is mainly distributed in northeast Asia, where northeast China is the southernmost distribution of this species, and it is a large forest-dependent species associated with conifer-dominated forests, and considered as a significant species for boreal and montane forests (He, Wan, Tian, & Ge, 2004;Ren et al., 2016;Sykes, Neiffer, Terrell, Powell, & Newton, 2013;Yin, Ge, Guan, Sun, & Li, 2009;Zhang, Ding, Ding, & Zheng, 2003). Previous research has implied that population declines have been driven by habitat destruction, with a 53% decrease in observation records between about 1950 and 2010 . Given that climatic regimes are expected to change further in coming years, as noted in the Paris Climate Agreement and by the Intergovernmental Panel on Climate Change (Huang, Yu, Dai, Wei, & Kang, 2017;IPCC, 2015), such changes may force animals to move toward higher elevations and latitudes, leading to habitat loss and fragmentation and range contraction. Therefore, the conservation situation is considered pessimistic for the black-billed capercaillie in the near future. To support conservation efforts, we should strive to understand the changes in its range and associated driving factors based on long-term dynamics that may not be available from short-term ecological studies.
In this study, we integrated ENM's with long-term ecological records of black-billed capercaillie to estimate the magnitude of changes in their distribution over time. To this goal, we cross-checked datasets from previous research and added new records collected from multiple resources (e.g., gazetteers and journal articles); reconstructed climate data to represent environmental variables, including historical climate data and general circulation models (GCMs); estimated the potential habitat for each period using BIOMOD2; and evaluated changes in the distribution range and possible impact factors. Our results offer an understanding of the ecological and biogeographic characteristics of the black-billed capercaillie population decline over a long period of time, which can be used to improve the power of conservation management for the species and other species in similar ecologic niches.

| Study area
The study area spanned Heilongjiang Province, Jilin Province, and
To eliminate potential bias caused by clustered occurrences, we removed duplicate records within the same cell (size ~1 km 2 ). The total records selected and classified into different decades (Table 1) after this elimination included 376 records from the 1970s, 278 from the 1980s, 227 from the 1990s, and 199 from the 2000s. The records in the 2000s were selected to derive future climate projections and were considered the baseline records. and Food Security (http://www.ccafs-climate.org/) (Appendix S1).

| Environmental variable selection
Topography data were obtained from the SRTM 90 m Digital Elevation Database (http://srtm.csi.cgiar.org/), which were used to calculate slope and aspect data. In total, 22 variables were obtained for further research (Appendix S1). All ecological data were obtained in a raster structure with a cell size of 1 km 2 .

F I G U R E 1 Study area in northeast China
To reduce collinearities among the variables used for modeling and avoid inappropriately complex models, we built a Spearman correlation matrix for all variables (Ranjitkar et al., 2016;Zhang et al., 2017).
We drew a tree diagram (Appendix S1) using the Hmisc package in R to facilitate variable selection. In the tree diagram, correlated variables were linked by lines and clustered together. For any coefficient >0.70, we removed the variable with a lower value in the percentage importance as calculated by Maxent3.3.3k (Elith et al., 2011;Phillips, Anderson, & Schapire, 2006). Finally, each historical period had special variables for ecological niche modeling, and current conditions (baseline) had special variables for ensemble forecasting of the species distribution.

| Species distribution modeling analyses
We estimated the potential distribution of the black-billed capercaillie in northeast China over time using BIOMOD2. Implemented SDMs from the historical climate (including 1970s, 1980s, 1990s, and 2000s), five SDMs under current conditions (baseline), and 60 SDMs for MIROC5, with four RCPs in each period (including the 2030s, 2050s, and 2070s). The final logistic results for each period were weighted based on the TSS. To transform the models from environmental suitability into presence-absence distributions, we used the threshold (P) calculated with the weighted average cutoff calculated by BIOMOD. From this, all outputs were divided into two groups; outputs above the threshold (>P) were grouped as "present," while all other values (<P) were considered "absent" (see details in Appendix S2).
It is important to quantify the spatial pattern of changes in distributions. Therefore, the predicted current range/historical range (CR), potential range loss (RL), and potential range gain (RG) were assessed by summing the numbers of the respective related pixels. Then, the range turnover (RT) ([RG + RL]/[CR + RG]) was calculated (Luo et al., 2015). The change in range was calculated with BiodiversityR (Ranjitkar et al., 2016). The elevation and center of the distribution in different periods were extracted. Spatial analyses were performed Occurrence data were collected from five sources and verified (see Section 2-Data). Records were obtained from occurrence points, and duplicate records within the same cell were removed. All results were derived from the distribution models. Model accuracy was determined based on the TSS (mean ± SD). The ranges were determined as the total of the number of related pixels (~1 km 2 ).

| Database construction results
A total of 358 records from previous research , 75 from articles and local historical documents (He et al., 2004;Yin et al., 2009;Zhang et al., 2003;Zhao, 2001), five from bird records, four from specimens, and two from nature reserve scientific surveys were collected. Conflicting records with unsubstantiated metadata, such as those lacking relevant or detailed descriptions, were excluded from the analysis.

| Model performance
The number of available records of the black-billed capercaillie in each decade ranged from 199 to 376 (   Figure 2). In the 1970s, their potential distribution area was 46.28% of the total study area (938,000 km 2 ). Compared to the potential distribution in the 1970s, the range decreased by 21.89% and 44.33% in the 1980s and 1990s, respectively. The elevation increased compared to the 1970s, with an increase of 2.40% and 6.34% in the 1980s and 1990s, respectively (see Appendix S3). There may have been regional extinctions in the Changbai and Wanda Mountains in the late 1980s, in agreement with previous research . In the 2000s Overall, compared to the 1970s, RL and RG for all periods (1980s, 1990s, and 2000s)

| DISCUSSION
Our investigation supports the use of long-term ecological data to understand the dynamics of species responses to human pressures and climate change. The analyses controlled or tested for multiple issues affecting data quality, resolution, incompleteness, and biases that have not been addressed in previous studies, including the creation of a long-term ecological dataset of the black-billed capercaillie from multiple cross-checked resources, construction of spatiotemporal geographical changes with species distribution models, and integration of changes in the near future with a historical context.
These methods can be used to track range changes across longer timescales than those usually addressed in ecology or conservation biology. We suggest that future research addressing climate change should consider the differences between the baseline and more realistic current conditions (i.e., the 2000s in this study).

| Model limitations
Our research contained several limitations, including common limitations associated with SDMs, such as data accuracy, variable representativeness, and spatiotemporal bias (Beck, Böller, Erhardt, & Schwanghart, 2014;Boakes et al., 2010;Ranjitkar et al., 2016;Rowe et al., 2014). In addition, integrating the future changes within a historical context was only based on the difference between the range from the baseline and the 2010s, because the GCMs were calculated from the baseline (Hijmans et al., 2005).
However, historical records of black-billed capercaillie presence already contain potential information on human influences and habitat information, which could reduce bias caused by missing predictors or variables representative of the model. The difference between the range from the baseline and the 2010s also offers an opportunity for qualitative analysis and provides a more realistic assessment under climate change.

| Threats and conservation
Previous research has shown that anthropogenic influences, including deforestation, agricultural expansion, and overexploitation of terrestrial ecosystems, have occurred throughout northeast China in the past decades, leading to distribution declines (Moss et al., 2014;Ren et al., 2016). However, this influence is difficult to quantify and analyze due to the availability of historical data on human activities (e.g., land cover and vegetation cover  (He et al., 2004;Yin et al., 2009;Zhang et al., 2003). This is in agreement with our results of a range increase in the 2000s compared to the 1990s. Environmental protection programs and population trends (i.e., southern shift and urbanization) suggest that anthropogenic influences may not dramatically increase in the near future, and that the range change will be mainly driven by climate change.
China has established nature reserves network in our study area, In addition, community co-management should be established with the cooperation of local inhabitants, governments, and conservation biologists. Ecological corridors should be designed to maximize habitat connectivity. Meanwhile, to facilitate acceptance by local communities whose lands are affected by such conservation networks, an incentive program should be launched to encourage locals to take part in conservation efforts.

CONFLICT OF INTEREST
None declared.

AUTHOR CONTRIBUTIONS
All authors designed the experiment. Li Yang and Xiaofeng Luan performed the data collection, the structure of manuscript, and drafted the first version of the manuscript. Chao Zhang performed the data collection, experiments, and data analysis. Minhao Chen performed data collection and interpretation. Jingxin Li and Lei Yang worked in the data analysis. Zhaomin Huo and Shahid Ahmad worked in the data collection. All co-authors participated in the scientific discussions and commented on the manuscript.