Spatio-temporal model of wheat yield in Shandong Province based on Bayesian Kriging

In order to solve the issue of dependence in crop yield prediction, on the basis of Bayesian hierarchical model, GB2 distribution is used instead of normal, logistic and other distribution forms, time effect t and Kriging method for spatial effect are also introduced into the scale, shape, kurtosis and skewness parameters of GB2 distribution, at the same time, a variety of covariates are added to the model to explore the factors that affect the yield. The empirical results based on wheat yield data in Shandong Province show that, compared with the two-step method which fits time trend and yield distribution separately, the embedded Bayesian model better simulates the characteristic of spatio-temporal dependence in crop yield and reduces the superimposition of prediction errors; The introduction of GB2 distribution provides greater flexibility for fitting crop yield; The introduction of the Kriging method improves the prediction of the model; The embedded spatio-temporal model based on the Kriging method can effectively improve the accuracy of wheat yield prediction and reduce the area yield insurance premium.