柴达木盆地植被覆盖度的分析与模拟——空间自相关方法的运用(英文)

Relationship between vegetation and environmental factors has always been a major topic in ecology, but it has also been an important way to reveal vegetation’s dynamic response to and feedback effects on climate change. For the special geographical location and climatic characteristics of the Qaidam Basin, with the support of traditional and remote sensing data, in this paper a vegetation coverage model was established. The quantitative prediction of vegetation coverage by five environmental factors was initially realized through multiple stepwise regression (MSR) models. However, there is significant multicollinearity among these five environmental factors, which reduces the performance of the MSR model. Then through the introduction of the Moran Index, an indicator that reflects the spatial auto-correlation of vegetation distribution, only two variables of average annual rainfall and local Moran Index were used in the final establishment of the vegetation coverage model. The results show that there is significant spatial autocorrelation in the distribution of vegetation. The role of spatial autocorrelation in the establishment of vegetation coverage model has not only improved the model fitting R2 from 0.608 to 0.656, but also removed the multicollinearity among independents.

Analyzing and modeling the coverage of vegetation in the Qaidam Basin of China: The role of spatial autocorrelation

Relationship between vegetation and environmental factors has always been a major topic in ecology, but it has also been an important way to reveal vegetation’s dynamic response to and feedback effects on climate change. For the special geographical location and climatic characteristics of the Qaidam Basin, with the support of traditional and remote sensing data, in this paper a vegetation coverage model was established. The quantitative prediction of vegetation coverage by five environmental factors was initially realized through multiple stepwise regression (MSR) models. However, there is significant multicollinearity among these five environmental factors, which reduces the performance of the MSR model. Then through the introduction of the Moran Index, an indicator that reflects the spatial autocorrelation of vegetation distribution, only two variables of average annual rainfall and local Moran Index were used in the final establishment of the vegetation coverage model. The results show that there is significant spatial autocorrelation in the distribution of vegetation. The role of spatial autocorrelation in the establishment of vegetation coverage model has not only improved the model fitting R² from 0.608 to 0.656, but also removed the multicollinearity among independents.