基于核密度估计的动物生境适宜度制图方法

Mapping wildlife habitat suitability using kernel density estimation

生境适宜度制图能提供动物适宜生境的空间分布信息,对野生动物种群管理、保护地规划等非常重要。生境适宜度制图的关键是构建生境适宜度模型(habitat suitability model,HSM),只基于动物出现位置数据构建HSM的方法在实践中得到了非常广泛的应用。然而现有的只基于动物出现位置数据构建HSM的方法还不能很好地直接表达动物生境适宜度和环境因子之间具有生态学意义的数量关系,因此也就不能很好地体现环境因子对动物生境利用的生态学作用。提出了一种基于核密度估计构建HSM的方法,在地理信息系统技术支持下,通过运用核密度估计从代表性的动物出现位置数据中估计出动物出现对各个环境因子的概率密度函数来直接表达生境适宜度与各个环境因子之间的数量关系,以体现环境因子对动物生境利用的生态学作用,在此基础上对生境适宜度与各个环境因子之间的数量关系进行综合构建了具有明确生态学意义的HSM用于动物生境适宜度制图。以美国Voyageures国家公园的白尾鹿(Odocoileus virginianus)生境适宜度制图为例,基于365个出现位置点位数据并结合积雪深度、地表覆被类型、森林边界长度和坡度等环境因子数据,开展了该方法的案例研究。通过交叉验证计算连续Boyce指数对制图结果进行评价,结果表明:基于核密度估计方法构建的HSM预测能力强,所得出的生境适宜度图经10次交叉验证,连续Boyce指数平均值为0.75,标准差为0.11,达到了较高精度。此外,由于基于核密度估计的方法能以"生境适宜度和环境因子之间具有生态学意义的数量关系"的形式来直接体现环境因子对动物生境利用的生态学作用,就模型的可解释性而言,该方法要优于现有的其他构建HSM的方法。

Habitat suitability mapping can provide information on spatial variation of wildlife habitat suitability which is useful for wildlife population management and conservation planning. The key issue of habitat suitability mapping is how to construct habitat suitability model （HSM）. Existing presence-only based methods of constructing HSM cannot explicitly express the quantitative relationship between wildlife habitat suitability and environmental factors. So the resulted HSMwould be insufficient to express the ecological effects of environmental factors on wildlife habitat use. In this paper we proposed a kernel density estimation-based method of using wildlife occurrence data to construct HSM for wildlife habitat suitability mapping, which can obtain the quantitative ecological relationship between habitat suitability and environmental factors. Under the assumption that habitat should be more suitable where wildlife presented more frequently, the presence probability density function （PDF） estimated from wildlife occurrence data was used to express the quantitative ecological relationship between habitat suitability and each individual environmental factor. Then, assuming that habitat suitability is determined by the limiting factor, a minimum operation was applied to construct the final HSM by synthesizing all these relationships on an environmental factor set. In the case study we applied the proposed kernel density estimation-based method to habitat suitability mapping of the White-tailed deer （ Odocoileus virginianus） in Voyageures National Park based on 365 deer occurrence points from aerial survey and environmental data characterizing its living environment, with the aid of geographic information system techniques. The environmental factors selected were snow depth, land cover type, forest edge length, and slope gradient. Cross validation was used to evaluate the performance of the proposed method. Results showed that the mean of continuous Boyce Indexes calculated from 10 repetitions of 2-fold cross validation was 0.75 with a standard deviation of 0.11. This suggests that the HSM from the proposed method has a good predictive capability. The proposed kernel density estimation-based method well captures the ecological effects of environmental factors on the wildlife habitat use in the form of quantitative relationship between wildlife habitat suitability and environmental factors. From the perspective of model interpretability, the proposed method is better than other HSM-constructing methods.