Prediction of spring agricultural drought in cold and arid regions on the basis of soil freeze thaw processes

Qiqihar city, Heilongjiang Province, China The prespring period was divided into four periods according to the freezing and thawing of the soil: the prefreeze period, rapid freezing period, stable freezing period and thawing period and select relevant factors as explanatory variables. PPR model, PLS model, PLS-PPR model, LSTM model, RF model and PCR model were used to predict agricultural drought in spring. The standardized soil moisture index at different depths was used to characterize regional drought as an explained variable. Comparative analyses revealed that the projection pursuit regression model was significantly better than the other five models were.The results show that the accuracy of drought prediction in the shallow layer was greater than that in the deep layer, and the prediction results in dry years are better than those in humid years. Subregionally, the prediction effect in the southern region was better than that in the central region, and the prediction effect in the central region was better than that in the northern region. These results were consistent with the spatial distribution characteristics of the regions;In this work, a method for predicting spring agricultural drought in cold and dry areas based on the freezing and thawing of soil during the prespring period was developed and provide a reference for the study of spring drought in cold areas. • Explanatory variables are selected via freeze-thaw process, drought identified by SSMI serves as the dependent variable. • The paper integrates cold-region spring drought prediction with the PPR model and performs comparative analysis. • The paper researches and discusses spatial-scale prediction accuracy of soil depth models ranging from 10 cm to 200 cm.