甘肃省小麦条锈病发生面积预测

甘肃是中国小麦条锈菌重要的周年循环发生区之一，是重要的菌源基地。准确预测甘肃省小麦条锈病的发生面积，对甘肃及中国小麦条锈病的科学防控具有重要意义。利用2001-2021年甘肃省小麦条锈病秋苗发生面积、温度、相对湿度、降雨量和日照时数等资料，通过Pearson相关性分析筛选到影响甘肃小麦条锈病流行的4个关键因子，即小麦条锈病秋苗发生面积、上年8月最低气温、1月平均相对湿度和3月日照时数，并采用全子集回归和BP神经网络算法对甘肃小麦条锈病发生面积进行预测。结果表明，全子集回归模型1和模型2对2020-2021年甘肃小麦条锈病发生面积预测准确度分别为94.63%和88.81%；BP神经网络模型1和模型2的预测准确度分别为98.25%和94.03%。综上可知，BP神经网络模型1是最佳预测模型，其预测2022年甘肃省小麦条锈病发生面积为10.03万hm²。

Prediction Area for Occurrence of Wheat Stripe Rust in Gansu

Gansu province is one of the important annual circulation areas of wheat stripe rust in China,so it is an important pathogenic base.Accurate prediction area for occurrence of wheat stripe rust in Gansu province is of great significance to the scientific prevention and control of wheat stripe rust in Gansu province and even in China.In this study,the area for occurrence of wheat stripe rust in autumn,temperature,relative humidity,rainfall and sunshine duration from 2001 to 2021 were used to screen 4 key factors affecting the prevalence of wheat stripe rust in Gansu province by Pearson correlation analysis.The four factors are the area for occurrence of wheat stripe rust in seedlings in autumn,minimum temperature in August of last year,average relative humidity in January and sunshine duration in March.The area of occurrence of wheat stripe rust in Gansu province was predicted by using full subset regression and BP neural network algorithm.The results showed that the prediction accuracy of total subset regression model 1 and model 2 for the area of occurrence of wheat stripe rust in Gansu from 2020 to 2021 was 94.63% and 88.81%,respectively.The prediction accuracy of BP neural network model 1 and model 2 was 98.25% and 94.03%,respectively.It can be seen from the above that BP neural network model 1 is the best prediction model,which predicted that the occurrence area of wheat stripe rust in Gansu province in 2022 will be 100 300 hm².