• 论文 •

### 基于气候适宜度指数预报玉米产量时旬权重系数的确定方法

1. 1. 中国农业科学院农业环境与可持续发展研究所/作物高效用水与抗灾减损国家工程实验室/农业部农业环境重点实验室，北京 100081；2. 吉林省气象科学研究所，长春 130062；3. 沈阳农业大学农学院，沈阳 110061
• 出版日期:2018-10-20 发布日期:2018-10-16
• 作者简介:邱美娟（1987-），女，博士生，从事农业气象灾害与产量预测研究。E-mail：qmjcams@163.com]
• 基金资助:
中国农业科学院科技创新工程（CAAS-ASTIP-2014-IEDA）；农业农村资源等监测统计经费（2130111-20147-2018）；国家重点研发计划“重大自然灾害监测预警与防范”重点专项（2017YFC1502804）

### Determination Methods of Weight Coefficient in Spring Maize Yield Prediction Based on Climatic Suitability Index

QIU Mei-juan, LIU Bu-chun, YUAN Fu-xiang, LIU Yuan, ZHANG Yue-ying, WU Xin-yue, XIAO Nan-shu

1. 1. Institute of Environment and Sustainable Development in Agriculture, CAAS/National Engineering Laboratory of Efficient Crop Water Use and Disaster Reduction/Key Laboratory of Agricultural Environment, MOA, Beijing 100081,China; 2. Institute of Meteorological Sciences of Jilin Province, Changchun 130062; 3. College of Agronomy, Shenyang Agricultural University, Shenyang 110161
• Online:2018-10-20 Published:2018-10-16

Abstract: The climatic suitability models of each 10-day in the growth season of spring maize were constructed by using crop data of spring maize from 1980 to 2016 and daily meteorological data of 50 meteorological stations in Jilin province. In order to calculate climatic suitability index in different times (from early April to the 10-day before forecast day), methods of absolute, normalization, and correlation were used respectively to determine the weight coefficient of climatic suitability of each 10-day. Then, the relevance between meteorological influence index for maize yield and climatic suitability index obtained by different methods has been analyzed. A yield dynamic prediction model was established by regression analysis, and was used to forecast the spring maize yield in Jilin province. The results showed that, there were some differences between the three weight coefficient determination methods, but on the whole, the variation trend with the growth period were basically the same. The yield bumper or poor harvest prediction model established by regression analysis using materials from 1981 to 2012 most passed the 0.05 level effective test and the historical fitting average accuracy was all above 85.0%, the normalized root mean square error NRMSE was all less than 17.0%, and the accuracy of the bumper or poor harvest trend was generally in 60.0%-80.0%.The difference between the three methods was not obvious. The results of maize yield extrapolation forecast from 2013 to 2016 showed that, the yield prediction accuracy in each forecast times had fluctuant, but the average accuracy of methods of absolute, normalization, and correlation was 93.5%, 90.8% and 87.2%, respectively，and the standard deviation of the forecast results was 32.6, 69.4 and 116.1, respectively. Moreover, the average accuracy of each prediction time for the method of absolute was above 85.0%. It showed that the accuracy and stability of the prediction result of absolute were all high, which could meet the needs of business services.