基于GM（1，1）模型的新疆棉叶螨预测研究

Predictive Study on Xinjiang Cotton Spider Mites Based on GM （ 1,1 ） Models

影响棉叶螨发生发展的因子众多，有些影响因子的数据不易采集，给建立科学实用的螨害预测预报模型带来很大困难。为尝试解决这一问题，本文将灰色系统理论应用于新疆亚热带棉区棉叶螨预测预报研究，基于历史螨害发生数据，建立了GM（1，1）灾变预测模型。检验结果表明，该模型的平均相对精度为97．1％，小误差概率P值为1，均方差比值C为0．0627，各项指标均达到Ⅰ级预测精度。使用2014-2016年度螨害发生数据对模型进行验证，预测结果与螨害实际发生情况相吻合。研究结果表明：GM（1，1）灾变预测模型可用于棉叶螨的中长期预测，且具有建模所需数据量少、数据易于获取、预测精度高等优点，在实际农业生产中具有更好的实用性。同时由于农作物虫害具有相似的规律性，因此，本文的研究方法能够为其他类似的生物灾害预测提供参考和借鉴。

Cotton Spider Mites is one of the major biological disasters in cotton production and making heavy loss in agricul- tural economy. However, there are many factors could affect the occurrence and development of cotton spider mites. Besides, the acquisition of some data by using the traditional method such as manual inspection and field sampling is difficult, which poses a great difficulty in the normal application in actual process of the prediction and forecast model of mite damage. In order to solve this problem and to improve the accuracy of the prediction of cotton spider mites, this article applies grey system theory to the fore-cast, prediction and study of cotton spider mites in Xinjiang subtropical region. Through analyzed the historical data of mites occurrence from 2004 to 2013, we established the GM （ 1,1 ） disaster prediction model. Test results show that the average relative precision of the model is 97.1%, small error probability P value is l, and mean variance ratio C is 0.062 7, with all indexes have reached grade I prediction accuracy. Result of verification on the model by using the data of annual mite pest occurrence from 2014 to 2016 shows that the predicted result coincides with the actual occurrence of cotton spider mites. The research results show that： The GM （ 1,1 ） disaster prediction model can be used for medium and long term prediction of cotton spider mites, and the model was built with high prediction precision. Data required by GM （ 1,1 ） model, single with easy access, has better practicability in the actual agriculture process. At the same time, because the crop pests has a similar regularity, so that this research method can provide reference for other kinds of similar bio-disaster prediction and forecast.