Research On The Relationship Of Climatic Factors And Grain Yield In China

China is a country with a huge population. Ensuring food security has always been one of the most important objectives in China’s agricultural policy. Among the food security issues, the foremost is the impact factor research of food production, which has also been the focal point of academia concerns. In order to study the impact of climatic factor change on food production, a lot of scholars have considered this. However, there are some shortages in the existing achievements, for example, there were barely studies considered the complex relationship of climatic factors themselves or had a systemic analysis in climatic factors on food production. Using a statistical point of view, this article proposed a new method in assessing the relationship between climatic factors and food production, based on the China’s yearly grain yield and yearly averaged climatic factor data from1961to2011.The main point in this article concentrated in two aspects below. First, the article applied the HP filter decomposition method to separated the long time-series of yearly grain yield into a trend and a fluctuation, with the trend caused by technology factors, while the fluctuation by climatic factors. The results were also compared with those obtained by the traditional,5-year moving average and Logistic fitting methods. Second, the article selected the theory of Partial Least Squares Regression to model the relationship between climatic factors and food production, so as to solve the problem of Multi-Collinearity in climatic factors. Considering the relations between climatic factors and food production were not simply linear, the article further introduced Gauss kernel function. Using Gauss kernel function, the article changed the non-linear relationship of climatic factors into a quasi-linear one. Then using the methods above, the article carried out a regression analysis and forecast for the relations between climatic factors and food production. These were also compared with results from ordinary Partial Least Squares Regression and CD production function.The article has reached the following conclusions by the methods described above:(1) When the yearly grain yield data is a long time-series, the application of the HP filtering method is simple, the process is easy and it is suitable for all kinds of flexible date with great inclusiveness. Meanwhile, the results of the analysis not only reflect very well the societal development trend, but also rightly demonstrate the impact of climate change on food production fluctuations. The method is scientifically effective.(2)The modeling study of the relationship using the non-linear Partial Least Squares Regression model, based on Gauss kernel function, has favorable results and high forecast accuracy. Compared with ordinary Partial Least Squares Regression, the un-linear Partial Least Squares Regression model could not only solve the problem of Multi-Collinearity in climatic factors, but also revealed the non-linear relationship between climatic factors and food production. Compared with CD production function, the factors needed by the non-linear Partial Least Squares Regression model are less and the operating process is easier.At last, the article introduced Multiple Regression for determining the relative importance among the factors, and attempted to apply it to the model of climatic factors on food production.