农业气象产量预报方法中权重系数的确定

Determination of weight coefficient in agro-meteorological yield prediction

基于气候适宜度指数的作物产量预报中,各旬气候适宜度加权集成构成了气候适宜度指数,权重系数的确定对预报的准确性至关重要.本文以山东省冬小麦为例,分别采用绝对值法、归一化法和相关系数法确定各旬产量预报的权重系数,分析不同权重系数确定方法得到的气候适宜度指数之间的相关性及气候适宜度指数与农业气象产量的相关性,并利用相对较优的方法进行冬小麦产量动态预报.结果表明:与绝对值法和归一化法相比,采用相关系数法确定的权重系数计算的气候适宜度指数与气象产量的相关系数多数均通过了P<0.01水平的显著性检验,可较全面的反映冬小麦生长发育随气候条件的变化关系.建立的冬小麦产量动态预报模型历史回代检验的准确率基本均为94.5%以上,标准化均方根误差n-RMSE均小于7.0%,表明建立的模型模拟性能较好.对2010—2011年山东省冬小麦产量的外推预报表明,预报平均准确率为93.0%以上,多数预报准确率为96.0%以上,可满足冬小麦产量预报的业务需求.

In the crop yield prediction based on climate suitability index, the determination of weight coefficient was of vital importance to the results. Using winter wheat in Shandong province as a case study,weight coefficient of each ten-day was determined by methods of an absolute value,normalization and a correlation coefficient. Correlations among climate suitability indexes calculated by the above methods and with meteorological yield were analyzed,and the optimal one was used to dynamically predict wheat yield. The results indicate that the correlation between meteorological yield and climate suitability index obtained by the correlation coefficient method almost all passes a significant test of the 0. 01 level compared with the other two methods,which can reflect the relationship between the development of winter wheat with change of climatic conditions. The accuracy rate of the historical fitting for the established model is above 94. 5%,and the standardized root mean square error（ n-RMSE） is less than7. 0%,showing the excellent performance of the model. The results of yield dynamic prediction during 2010 to2011 suggest that the mean accuracy rate is above 93. 0% and most are above 96. 0%,which can meet the service needs.