台灣黑熊之棲息地利用及分布預測模式

外文標題: 
Habitat Use and Distribution Model of Formosan Black Bears (Ursus thibetanus formosanus)
校院系所: 
屏東科技大學野生動物保育研究所
指導教授: 
黃美秀
出版年份: 
2007年
主題類別: 
摘要: 

棲息地的喪失及破碎化為現今保育和經營管理大型食肉目動物最主要的議題,有效的經營管理野生動物遂繫於瞭解動物對棲息地的需求和預測其分佈。台灣黑熊(Ursus thibetanus formosanus)為亞洲黑熊(Ursus thibetanus)於台灣之特有亞種,也是瀕臨絕種的保育類動物。本研究旨在收集近年來(1990~2006)發現台灣黑熊的紀錄,搭配台灣地區生態與環境因素資料庫,以瞭解黑熊棲息地利用特色,並進一步利用基因演算(Genetic algorithm for rule-set prediction,GARP)法和邏輯回歸分析(Logistic regression),建立台灣黑熊分佈預測模式。

本研究總計收集589筆台灣黑熊紀錄點位資料,涵蓋344個1*1公里網格(簡稱“有熊格”)。於本島除高雄市及臺北市以外海拔一百公尺以上的研究樣區中,黑熊紀錄出現地所屬的環境因素,包括海拔、坡度、植生指標、離溪流最近距離、自然度指標、道路密度、離道路最近距離,皆與樣區呈現顯著差異(Mann-Whitney U test,p<0.001),黑熊對於這些棲地因素亦皆具特殊之選擇性(chi-square test,p<0.05)。

與研究樣區相較之下,黑熊偏好海拔1000~3500公尺地區,並對於<1000公尺的低海拔利用程度偏低。在國有林地區,台灣黑熊則偏向出現於針葉林與針闊葉混合林(39.79% vs.29.41%),且植生指標(Normalized Difference Vegetation Index)大於0.4(69.77%)之區域。台灣黑熊對於坡度15~45度,以及距溪流較>2500公尺區域的利用程度亦比預期中的高,而較少利用接近溪流<500公尺的區域。在人為干擾因素方面,黑熊會避開人為活動頻繁地區,偏好自然度指標高的原始森林,以及離道路>1公里的區域,而迴避道路密度>3公尺/公頃的區域。

台灣黑熊記錄出現於高山型國家公園以外的地區,於上敘七項環境因素屬性與樣區皆呈顯著差異(p<0.001,X2皆>29.04);然於國家公園內,黑熊出現地所屬的環境特性,僅於坡度、植生指標、距溪流或道路之最近距離因素,與樣區呈現顯著差異(p<0.001,X2皆>7.13)。台灣黑熊在較缺乏經營管理之國家公園以外的地區的整體環境選擇程度較國家公園內高,顯示人為干擾因素對於黑熊棲地選擇之重要影響。

台灣黑熊分佈預測模式中,邏輯迴歸法因為沒有熊出現紀錄的精準度較低,預測結果與實務經驗為基礎的預期分佈差距頗大,無法提供合理的預測結果。反之,基因演算法的預測能力較邏輯迴歸分析高(sensitivity =0.893,Kappa=0.360);此模式預測結果顯示,台灣黑熊主要分佈於中央山脈,隔離的海岸山脈亦有零星分佈。利用切點0.525預測的黑熊潛在棲息地為9587平方公里,佔台灣本島25.5%,並有45.34%於國家公園內,突顯出加強非國家保護區系統棲息地的經營管理,對於保育使小族群物種的重要性。

本研究顯示台灣黑熊與早期研究的地理分佈上有縮減的趨勢,其棲息地的選擇除了受到食物資源豐富度以及相關物候環境影響之外,人為干擾因素更是關鍵因素。故建議未來的預測模式,除了應該加強建立有熊及無熊資料庫的精準度之外,同時應該整合人類活動類型及位置,尤其是非法狩獵壓力、食物豐富度(包括植物及動物),以及不同類型道路系統,以達更精確的分佈預測。

外文摘要: 

Habitat loss and fragmentation have become the primary issues for conserving and managing large carnivores. Successful wildlife management depends on understanding habitat requirement and predicting distribution of these animals. Formosan black bears (Ursus thibetanus formosanus), the endemic subspecies of Asiatic black bears in Taiwan, are listed as an endangered species locally. The study objective was to collect occurrence records of Formosan black bears throughout the island during 1990~2006, and then integrate the ecological and environmental database to understand the characteristics of bear habitat use. Moreover, the Logistic regression and the Genetic algorithm for Rule-set Prediction (GARP) were also applied to establish predictive models for the geographic distribution of this species.
I collected 589 bear locations, including 344 1*1 km grids, i.e., bear presence grids. Compared with the study area (> 100 m in elevation but except for Taipei and Kaohsiung cities), bear presence grids in all seven environment variables used, including elevation, slope, NDVI (Normalized Difference Vegetation Index), shortest distance to river, natureness, road and road density, showed significant differences (Mann-Whitney U test, p<0.001). The result also indicated bears showed specific preference toward these environment variables (chi-square test, p<0.001).

Formosan black bears preferred areas of elevations of 1000~3500 m, but used areas <1000 m than expected. Within national forests, bears preferred broad-leaved and coniferous mixed forests and coniferous forests (39.79% vs. 29.41%), along with areas with Normalized Difference Vegetation Index >0.4 (69.77%). They also preferred regions with slopes of 15~45 degree, distance to roads >2500 m, and distance > 2500m away from rivers, but seemed to avoid areas close to rives (< 500m). Copying with human disturbances, bears avoided areas with frequent human activity, preferred areas with high natureness index of natural forests, and > 1 km away from roads, but avoided areas of high road density, i.e., >3 m/ha.

Outside the mountainous national parks, there were significant differences between bear presence grids and the study area for all the seven environment variables used. However, within the mountainous national parks, significant differences between bear presence grids and the study area only occurred in slope, NDVI, and shortest distance to roads and rivers, respectively. Bears showed a relatively higher level of habitat selectiveness outside the national parks which generally with less management, than within national parks, indicating that the critical impact of human disturbance on habitat selection of Formosan black bears.
Due to biased bear absence data used in the logistic regression model, the predictive distribution model extremely differed from empirical observation of bears, and could hardly provide reasonable results. Alternatively, the GARP resulted in relatively high prediction accuracy (sensitivity =0.893, Kappa=0.360). This model predicted that bears mainly occurred along the Central Mountain Ranges, and few in the isolated Eastern Coastal Mountain Range. By applying the cut-point of 0.525, the model also indicated 9587 km2 of potential bear habitat, i.e., 25.5 % of the Taiwan island, and 45.34% was within national parks. This indicated the importance of enhancing habitat management among areas outside the national parks for preserving the small population of Formosan black bears throughout the island.

The result indicated the decreasing geographic distribution of Formosan black bears. Moreover, their habitat use is not only limited by food abundance and related physical environmental factors, but also human disturbance. The study suggests that in addition to improving the accuracy of bear presence-absence data, types and locations of human activities, especially illegal hunting, food abundance, and various road networks should be incorporated into the distribution model of Formosan black bears for better prediction.