基于BIOMOD的青海湖流域高原鼠兔分布模拟

随着统计模型及空间信息数据的不断发展和完善,物种分布模型已经成为全球变化背景下研究大尺度物种分布情况的重要工具。高原鼠兔(Ochotona curzoniae)是青藏高原特有的关键物种,在青藏高原生态系统中占有重要地位。通过采集高原鼠兔的分布点数据及环境变量数据,基于R语言中BIOMOD包中的7个模型对其在青海湖流域的分布进行了模拟。结果表明,高原鼠兔主要分布于青海湖西岸和北岸、天峻县周边及布哈河流域上游,影响高原鼠兔分布的主要环境因子为距道路距离、距居民点距离、最暖月最高气温、NDVI标准差、最冷季和最干季降水量。BIOMOD组合模型中,推进式回归树模型(GBM)和最大熵模型(MAXENT)的模拟效果最好,广义线性回归模型(GLM)结果较差。而优化后的结果显示,模拟结果的集成和筛选能有效提高模型的精度和效果。

Modelling the Distribution of Plateau Pika (Ochotona curzoniae) in Qinghai Lake Basin Using BIOMOD

Species distribution model has become an important tool to study the species distribution at large-scale in the context of global change due to the development and improvement of statistical models and spatial information data. Plateau Pika (Ochotona curzoniae) is a keystone species in the Qinghai-Tibet Plateau and plays an important role in the entire ecosystem. The Qinghai Lake Basin is located in the northeast of the Qinghai-Tibetan Plateau and is a typical closed inland basin with a watershed area of approximately 29 661 km2 that (Fig. 1). This research aimed to model the distribution of Plateau Pika in the Qinghai Lake Basin using seven models from BIOMOD package in R with occurrence data and environmental variables. AUC (area under the curve) and TSS (true skill statistic) based on confusion matrix (Table 1) were chosen to evaluate the performance of different models. The results showed that the Plateau Pika mainly distributed in the west and north bank of Qinghai Lake, around Tianjun county and in the upstream of the Buha River (Fig. 3 and Fig. 4). The most important environmental factors affecting the distribution of Plateau Pika were the distance to road and, to the settlement of people, the air temperature of the warmest month, the NDVI standard deviation, and the precipitation of the coldest and driest season (Table 2). The Boost Regression Tree model (GBM) and Maximum Entropy model (MAXENT) make the best predictions, while the Generalized Linear Model (GLM) gives a poor result (Fig. 2). The optimized result shows that the integration and selection of can improve the accuracy and performance of the model effectively (Fig. 5 and Fig. 6).