延河流域本氏针茅(Stipa bungeana)分布预测——广义相加模型及其应用

The predictive distribution of Stipa bungeana in Yanhe River catchment:GAM model and its application

物种分布预测,对于物种的保护、利用和恢复具有重要意义。利用广义相加模型(GAM,Generalized Additive Model),对延河流域典型地带性物种本氏针茅(Stipa bungeana)的空间分布预测进行研究,以期为该流域本氏针茅草地的保护、恢复等提供依据。结果表明,本氏针茅分布的环境梯度较广,在坡度、坡向、温度与降雨的各个梯度上都有分布,除高平地和侵蚀剧烈的沟道外,各种地形部位上亦可以存在。建立的广义相加模型表明,本氏针茅的分布主要取决于年均蒸发量和温度季节变化两个因子,而非单纯的降雨、温度因素。从其分布概率看,本氏针茅在延河流域大部分地区都有可能分布,但其分布集中区主要在中北部,与实际观测相符。模型检验表明,建立的模型满足统计要求。

The predictive distribution of species is of great importance not only to the conservation, utilization and restoration of grassland with these species as the dominant species, but also to the predictive distribution of vegetation communities in ecological restoration practices. In recent years much progress has been made in predictive models for species; models like GLM （Generalized Linear Model ）, GAM （Generalized Additive Model ） and VGAM （Vector Generalized Additive Model） become more and more popular and useful, provide greater impact in planning and decisionmaking in nature conservation and ecological restoration. Comparative studies showed that GAM, as a data-driven model, could give more accurate prediction of species distribution. This paper employs the non-parametric GAM model to explore the potential distribution of the Stipa bungeana in Yanhe River catchment. As per the data requirement of GAM, 3 topographic and 9 climatic indexes were first extracted and analyzed using ARCGIS and ANUSPLIN, then a GAM model for Stipa bungeana was established using these data and GRASP （Generalized Regression and Spatial Prediction） module for S-PLUS. The gradient analysis in this model showes that the Stipa bungeana can distribute in a wide range of environments, in different gradients of slope, aspect, temperature and precipitation, in all land positions expected for high flat land and extensively eroded gullies. However it does not mean the distribution of Stipa bungeana is equally affected by each factor, or has an uniform distribution probability in the whole environmental range. The GAM modeling indicates the distribution of Stipa bungeana is mainly controlled by the average annual evaporation and temperature seasonality but not the rainfall and temperature as commonly reported. A map is produced in ARCVIEW using a lookup table generated in GRASP module, showing the distribution probability of the Stipa bungeana in Yanhe River catchment. From this map it can be seen that Stipa bungeana can distribute in most areas in Yanhe River catchment, and its distribution centre locates in the middle and north part of the catchment. This result is in agreement with the reported point distribution of Stipa bungeana, proves that the GAM model well fits Stipa bungeana. However, as this is the first time the model being used in the study of vegetation-environment relations in China, there are still problems that need further study.