Development of the long-term forecast system based on the neural network of the potato late blight in Liangshan Prefecture

In this article, the relationship between occurrence area of potato late blight and climatic conditions is studied to find the leading influencing climatic factors based on the occurrence area and grading materials of late blight in potato production in Liangshan Prefecture from 1992 to 2007. The result shows that in the past years, the occurrence area and degree of late blight is positively correlated to the average temperature in January, May and June; at the same time, it shows positive correlation with the rainy days and relative air humidity in May, and shows relatively weak positive correlation with the rainy days and sunshine hours in June. Based on leading factor analysis, the average temperature in January, May and June, sunshine hours, relative humidity and rainy days in May and June are selected as input factors, late blight occurrence grading is selected as the output factor and a three-hierarchy BP network architecture is used to develop the long-term forecasting system for potato late blight in Liangshan Prefecture based on neuron network. According to the application in production from 2008 to 2009, the forecasting accuracy rate of the system is over 85%, with favorable social and economic benefits.