基于气候适宜度指数的吉林省大豆单产动态预报研究

Study of Soybean Yield Forecast in Jilin Province Based on Climate Suitability Index Method

为及时、准确地进行大豆产量预报,给相关部门提供可靠依据保障吉林省粮食安全,利用吉林省1980-2016年大豆产量、生育期、逐日气象数据,基于大豆生长发育的生物学特性(最适温度、上限温度、下限温度、需水量、需光性等),构建大豆生长季逐旬温度、降水、日照时数及综合气候适宜度模型,通过与相对气象产量进行相关和回归分析,建立基于气候适宜度指数的7-8月逐旬产量动态预报模型,对吉林省大豆产量进行动态预报。结果表明:各时段建立的产量预报模型均通过0.05水平的有效性检验,能够客观反映大豆生长期内气象要素状况;各预报模型的历史回代检验平均准确率均大于85.0%,均方根误差小于20.0%;历史回代拟合的气象产量与实际气象产量在年际变化上具有较好的一致性,两者相关性通过0.05水平的显著性检验;在1981-2014共34年中各旬单产趋势预报准确的年份均在24年以上。各时段预报模型对2015-2016年的外推预报准确率分别在92.0%和81.4%以上,趋势预报不稳定,可能与模型没有考虑灾害影响有关。总体上,构建的产量预报模型可为吉林省大豆产量预报提供参考依据。

In order to forecast the yield of soybean in time# and accurately and provide reliable basis for the relevant departments to ensure the food security of Jilin province, the temperature, precipitation, sunshine and comprehensive climatic suitability models of eachten-day in growth stages of soybean were constructed by using the yield material, developmental stages and daily meteorological data, from 1980 to 2016, and considering the upper limit temperature, the lower temperature, the optimal temperature, the water and sunshine demand for the growth and development of soybean. And via the correlation and regression analysis with meteorological yield, dynamic yield forecast model based on climate suitability index of each ten-day during July to August were established to predict the soybean yield of Jilin province dynamically. The results showed that the yield forecast models in different periods all passed the validity test of 0. 05 level, and could reflect the status of meteorological factors in growing stages of soybean objectively. The mean accuracy rate of the yield dynamic prediction model for historical fitting test was all above 85.0%, and the nonmalized root mean square error n-RMSE was less than 20. 0%. Inter-annual variation between actual relative meteorological yield and relative meteorological yield fit by yield dynamic prediction model had good consistency, and the correlation passed the significant test of 0. 05 level. In the 34 years from 1981 to 2014, the accurate yield tendency forecast of each ten-day was all above 24 years. The extrapolation forecast accuracy of 2015-2016 in each period was above 92.0% and 81.4% respectively and the tendency prediction was unstable. It may be for the reason that the influence of disaster was not considered in the model. Generally, the production forecast model was set up to provide important basis for agricultural meteorological yield forecast of Jilin province.