Study of the spatiotemporal variability in agricultural drought vulnerability based on a dynamic classification projection pursuit model

Abstract

Studies of the temporal and spatial variabilities in agricultural drought vulnerability can help researchers better understand the potential risks of regional agricultural drought. In this paper, projection points with approximate numerical values are grouped into one class, and the degree of dispersion is measured by the information entropy rather than the standard deviation. Then, the projection pursuit model is improved. Finally, the model is applied to assess the spatiotemporal variability in agricultural drought vulnerability for 10 counties in Qiqihar, a typical semi-arid region in China. The model analysis shows that the improved model avoids the cutoff radius, which is one of the difficulties of traditional models. Moreover, information entropy is superior to the standard deviation in measuring data. The vulnerability results show that regional drought vulnerability exhibits an increasing trend over time, the vulnerability in the northern part of the region is relatively high, and the vulnerability in the central and southern parts of the region is relatively low. The improved model can enrich the theory of projection pursuit, and the temporal and spatial variability of vulnerability results can provide a reference for resisting agricultural droughts in semi-arid areas.