Study And Application On The Occurrence Prediction Model Of Paddy Stem Borer (Scirpophaga Incertulas) And Liriomya Huidobresis (Blanchard)

In agricultural production, perennial occurrence of crop pests seriously affected the crop production. In order to control the pests effectively, human beings have done a long-term research on the way of pest prediction. The experience prediction, experiment prediction, statistic prediction and information prediction are mainly used at present. In recent years, with the development of nonlinear science, information technology and artificial intelligence technology, the prediction research of nonlinear pest, one of the important ways for the prediction of crop pests, has aroused greater attention.Considering the prediction on the crop pests occurrence, this paper uses the technologies of Stepwise Regression Analysis, Cluster Analysis, Principal Components Analysis(PCA), Back Propagation(BP) Artificial Neural Network(ANN), Temporal Geographical Information System(TGIS), etc., and takes the paddy stem borer (Scirpophaga incertulas), a common pest in rice and Liriomya Huidobresis(Blanchard) and in vegetable, as the research objects to study the prediction model of the pest and prediction technology of the pest information. It proposes the pest prediction models based both on the Stepwise Regression Analysis and the PCA/BPANN, and studies preliminarily the pest occurrence information prediction system based on TGIS. The main research steps and results are as follows:1. First, based on the linear science, a pest prediction model is built by using the Stepwise Regression Analysis, and the model is tested by using the occurrence data of paddy stem borer and Blanchard which are found in Jianshui County, Yunnan Province. The results show that:(1) With Stepwise Regression Equation, we can find two main meteorological factors which influence the two pests' occurrence so that the pest occurrence trend can be analyzed in a certain degree. (2) The multiple correlation coefficients of the Stepwise Regression Equation in both two pests are not more than 0.7, so we can get the result that, it has the limitation by using the Stepwise Regression Equation to quantitatively predict the paddy stem borer and Blanchard.2. Because the original data in the pest prediction model may be unreliable, this paper uses Cluster Analysis to process the original data so as to maintain its validity; Second, due to the problem of computing complexity and non-convergence when excessive factors are input in BPANN, this paper uses the dimension reduction by the way of PCA, and transforms the multi factors to minority comprehensive factors, to guarantee the comprehensiveness of influence factors, to reduce the data input of BPANN, and to improve the operating rate and the prediction accuracy of the prediction model.3. With reference to the findings in the regulations of the pest occurrence, the correlations of the influence factors, reliability of the original data, the calculation complexity when excessive factors are input in BPANN, and etc., this study combines the technology of Cluster Analysis, PCA and BPANN, and proposes the pest prediction model based on PCA and BPANN. This model shows its better predicting results and stronger stability by testing with the occurrence of two pests which are paddy stem borer and Blanchard in Jianshui County, Yunnan Province.4. Comparative studies are made between the pest occurrence prediction models based on the Stepwise regression analysis and those based on PCA/BPANN. This paper builds the pest prediction model by using the ways of Stepwise Regression Analysis and PCA/BPANN respectively, and selects paddy stem borer and Blanchard in Jianshui County of Yunnan Province to test the two prediction models as mentioned above respectively. The study shows, according to the real occurrence data, the pest prediction model based on PCA/BPANN with consideration to nonlinear science has stronger generality than the pest prediction model based on the Stepwise Regression Analysis with consideration to linear science.5. TGIS technology is applied to the pest prediction system research. Based on the research of spatiotemporal in the occurrence of crop pests, time is introduced as the same important factor as space into the design of pest prediction system. According to the characteristic that the distribution area of the crop pests is related to the changes of field parcel, the article describes the transformation of the spatiotemporal on crop pests to the spatiotemporal on the corresponding of the changing information in field parcel, and gives a spatiotemporal data model which fits the changes of field parcel. Finally, preliminarily studies are made the building of pest prediction system based on TGIS.