| Cultivated area in China is accounted for 8% of the world, but the proportion of the population has reached 20%. Therefore, food production is a top priority and factors that influence food production have been the focus of academic attention. At present, many scholars have studied the impact of climate change on food production. However, there are some deficiencies in the current achievements, for example, complex relationships between climatic factors and yield analysis process are not taken into account or there does not have a systemic analysis in climatic factors change and anomalous climate change on food production. From a statistical point of view, this article put forward a new method for analyzing the impact of climate change on grain yield, based on the China’s yearly grain yield and yearly averaged climatic factor data from 1961 to 2014.This study focuses on the following three aspects. The first part mainly separates the trend yield and the climatic yield using HP filter, forecasts the climatic yield combining BP filter and Fourier model, and forecasts the trend yield by polynomial distribution lag model. Then the results are compared with the traditional BP neural network and grey model method. Considering the multicollinearity problems among the climatic factors and the non-linear relationship between climatic factors and grain yield, second part studies the relationship between climatic factors and grain yield using two types of nonlinear partial least squares regression (PLSR) method. One is using cubic B-spline function to transform the non-linear relationship among the original variables into the quasi linear relationship before building PLSR model. And in order to achieve non-linear effect, the other is adding generalized regression neural network (GRNN) model into PLSR model. The third part discusses the future development trend of agro-meteorological disaster by R/S analysis method, and conducts risk assessment of grain yield loss in China on the basis of the fuzzy mathematic method based on information diffusion.This article has reached the following three conclusions from the main analysis of three aspects above:(1)Polynomial lag model has been used to predict the trend yield and Fourier model based on BP filter has been used to predict the climatic yield. Then we conclude that the relative errors are small and the most extreme difference and the variance are only 3.11% and 1.02%, respectively. The results show that the model in this article can be preferably applied to predict the grain yields, and it reflects accurately the impact of future climate change on grain yield fluctuation.(2)The non-linear Partial Least Squares Regression model, based on cubic B spline function to study the relationship between climatic factors and climatic yield, has a favorable result and high forecast accuracy. Compared with the conventional C-D production function method, the factors demanded by the Spline-PLSR model are less, the operating process is easier and the forecast accuracy is higher. Compared with the non-linear Partial Least Squares Regression which is internal embedded by GRNN, the Spline-PLSR model is more stable.(3)R/S method analyzes the trends of low temperature freezing-disaster’s affected area ratio may continue to rise and the disaster-affected area ratio of drought and hail and wind damage may continue to decline. However, the disaster-affected area ratio of flood will exhibit greater volatility trend. Then on the basis of calculating the relative disaster losses, a method based fuzzy mathematics is introduced in this article to assess the risk of yield losses due to agro-meteorogical disasters. The results show that when the probability increases, the risk of yield losses due to agro-meteorogical disasters may reduce. The high risk values of flood, drought, hail and wind damage and low temperature freezing-disaster may appear in the disaster loss rates of 5%-40%,5%-60%,5%-15% and 5%-10%, respectively. |
PDF Full Text Request