臺灣黑熊(Ursus thibetanus formosanus)族群相對豐富度及分布預測模式

The Relative Abundance and the Distribution Prediction Models of Formosan Black Bears (Ursus thibetanus formosanus)

蔡幸蒨(Shin-Chien Tsai)

指導教授 ： 黃美秀 ； 陳朝圳

國立屏東科技大學/農學院/野生動物保育研究所/碩士(2011年)

https://doi.org/10.6346/NPUST.2011.00219

摘要

物種的地理分布及棲地環境特色是制定有效野生動物保育計畫及經營管理所不可或缺的資訊。近年來電腦科技、遙測技術及地理資訊系統的發展，對於物種分布範圍、影響因素、棲地利用等研究之效率得到提升。本研究旨在藉由痕跡調查法瞭解目前未知是否有臺灣黑熊（Ursus thibetanus formosanus）出現之區域，瞭解其野外的分布及相關環境。利用黑熊有/無之空間資料，以Maxent與二元邏輯迴歸建構臺灣黑熊預測模式，瞭解及釐清黑熊於不同年代之地理分布和棲地選擇的變化，以及探討影響分布的重要因素。
2007年3月至2009年5月進行野外黑熊穿越帶痕跡調查，20個區域共發現236筆熊痕跡，以爪痕最常見（71.2%），排遺次之（16%）。各區域痕跡密度為0-3.8/ha不等，整體平均值為0.7/ha，或遭遇率為0.42/km。本痕跡調查結果與國外研究相較，均有偏低的情況。
利用本研究所收集的野外黑熊痕跡點位資料，以及歷年文獻訪談黑熊資料庫（黃美秀等，2010），經篩選後共計有熊點位1,017筆，野外資料占12%。除2筆年代不明資料外，第一時期（1990-1999年）和第二時期（2000-2010年）分別有300筆、715筆有熊點位。Maxent所建構的的二元性預測分布圖顯示，第一時期和第二時期分別為5,853 km2和8,588 km2，為臺灣本島面積的15.9%和23.3%。分布範圍主要位於中央山脈，海岸山脈只有極少且零星之區塊。黑熊在兩時期間對各項環境因素之選擇偏好的梯度大致上類似。相較於第一期的預測分布圖，黑熊於後期除了在中部丹大野生動物重要棲息環境一帶的範圍較小且零碎化之外，其他地區的分布似乎有擴張之趨勢。
利用原住民訪談、文獻資料、野外黑熊痕跡調查三類資料來源篩選出470筆無熊網格。邏輯迴歸以海拔、道路密度、距道路之最近距離、距國家公園之最近距離、坡度五個因素為最佳模式。預測圖顯示黑熊集中分布於三個高山型國家公園及其鄰近區域，然 Hosmer-Lemeshow擬合優度指標顯示最佳模式的五個環境變因並無法有效的預測黑熊的分布有無，推測主要可能因無熊偏差或尚未考量到的其他環境因素所致。
本研究顯示除了海拔之外，道路、距國家公園遠近等人為活動相關因子是影響黑熊分布的重要因素。根據野外調查紀錄和預測模式分布圖的結果，本研究建議北部的雪霸-太魯閣國家公園，以及南部的玉山國家公園和關山野生動物重要棲息地範圍為目前臺灣黑熊分布的重要熱點，並建議各區應該視族群狀況設立不同類型基準（benchmark），以進行有效的監測和經營管理。

關鍵字

臺灣黑熊 ； 分布 ； 痕跡調查 ； Maxent ； 邏輯迴歸 ； 棲地 ； 有-無出現

並列摘要

Geographic distribution and habitat use of animals are essential information for wildlife conservation and management. Due to the recently rapid development of computer, GIS (Geography Information Systems) and remote sensing techniques, the power and efficiency of studies in species distribution and habitat evaluation has been greatly improved. The study objective was to investigate the distribution status and habitat characteristics of the locally endangered Formosan black bear (Ursus thibetanus formosanus) through sign surveys. Maxent and logistic regression models were also applied by using presence/absence data to establish the distribution prediction models of different time periods.
Country-wide sign surveys were conducted in 20 survey areas from Mar 2007 to May 2009. Bear claw marks were most frequently founded (71.2%, n=236), and then scats (16 %). The sign densities varied by areas and ranged from 0 to 3.8/ha, with an overall average of 0.7/ha or an encounter rate of 0.42/km. Compared to the data of related studies in other bear countries, this study seemed to have a relatively low bear density.
Among 1,017 presence locations, mainly based on the literatures (Hwang et al., 2010), the sign survey accounted for 12%. After excluding 2 locations with unknown dates, the data were divided into two period: the first period (1990-1999, n=300) and the second period (2000-2010, n=715). The Maxent prediction model showed that the distribution range covered 5,853 km2 and 8,588 km2 (i.e., 15.9% and 23.3% of the island area) for the first and second period, respectively. Both ranges were mainly located in Central mountain range, but with extremely small and fragmented patch in the Coastal mountain range. The habitat selection of bears revealed a similar pattern between both periods. Moreover, compared to the bear range in the first period, the distribution range seemed to expand in the second period, expect for the Danda Major Wildlife Habitat area.
Through three data sources, including local interviews, paper reviews and sign survey, 470 absence grids were selected. The best-fitted logistic regression model included 5 factors of elevation, road density, distance to the nearest road, distance to the nearest national park and aspect. Although the prediction map showed that bears concentrated in three mountainous national parks, the Hosmer-Lemeshow index indicated a low prediction power of the model. It may be related to the precision of absence data and other variables which was not applied for simulations.
The result indicated that elevation, road density, distance to the nearest road and distance to the nearest national park were important variables for bear distribution. According to field sign surveys and prediction models, the study suggested that the Sheipa and Taroko National Parks, and Yushan National Park and Guanshan Wildlife Habitat were potential hot spots of bear habitats in the northern and southern Taiwan, respectively. The study further suggested that different types of benchmark should be applied in different areas while local population status and threats were considered in order to enhance effective monitoring and management.

並列關鍵字

Ursus thibetanus formosanus ； distribution ； sign survey ； Maxent ； logistic regression ； habitat ； presence-absence

參考文獻

王穎，2006。丹大野生動物重要棲息環境野生動物調查研究計畫（2/2）。

行政院農委會 79 年生態研究第 010 號。17 頁。

重要棲息環境分區規畫及動物監測（第二年）。行政院農業委員

環境動物監測及原住民狩獵利用之調查。行政院農業委員會林

行政院農委會 80 年生態研究第 014 號。44 頁。