气温和降水预报准确率对土壤水分模拟精度的影响

Effects of temperature and precipitation forecast accuracy on the precision of soil moisture simulation

农田土壤水分预报精度与天气预报准确率关系密切。为了了解天气预报准确率对土壤水分预报精度的影响,本文利用2010—2020年郑州土壤水分自动观测站土壤水分资料和逐日气象数据,基于农田土壤水分平衡模型模拟了土壤水分逐日变化,计算该模型在不同月份及不同作物下的模拟精度。通过分别设置日平均气温±1~2℃和降水量±10%~20%的变化幅度,分析气温和降水预报偏差对土壤水分模拟精度的影响。结果表明:(1)土壤水分模拟R~2年内表现为“大-小-大”的趋势,而NRMSE和MARE的趋势为“小-大-小”;11月—翌年3月的模拟效果优于其他月份。冬小麦的R~2为0.910,NRMSE为15.14%,MARE为8.57%;夏玉米分别为0.841、27.62%和15.49%,模型对冬小麦土壤水分的模拟精度高于夏玉米。(2)当气温预报偏差±1℃/2℃,对NRMSE和MARE略有影响,但R~2均未发生变化;气温预报值偏高,土壤水分模拟精度偏高;反之,模拟精度偏低。(3)当降水预报值偏多10%~20%,土壤水分模拟精度比降水正常时低;预报值偏少,则模拟精度提高,但NRMSE和MARE最大变幅分别为6.16%和3.23%,均小于降水预报偏差幅度。(4)当气温和降水预报同时存在偏差,土壤水分模拟精度的变化趋势主要由降水预报偏差所决定,即降水预报值比实际值偏多,模拟精度下降,预报值偏少,模拟精度上升;NRMSE和MARE最大变幅分别为6.19%和3.23%。

The precision of cropland soil moisture prediction is closely related to the accuracy of weather forecast.To clarify the influence of weather forecast accuracy on the precision of soil moisture prediction,we used soil water balance model of cropland to simulate the daily soil moisture in Zhengzhou,based on soil moisture data and daily meteorological data of automatic observation station from 2010 to 2020.We analyzed the simulation precision of the model in different months and different crops.The variations of daily average temperature(±1-2 ℃) and precipitation(±10%-20%) were set up to analyze the effects of temperature and precipitation forecast bias on the precision of soil moisture simulation.The results showed that annual variation of R~2 presented a trend of “big-small-big”,while that of NRMSE and MARE presented a “small-big-small” pattern.The simulation performance was better in November to the next March than in other months.Simulation precision for winter wheat soil moisture was higher than that of summer maize,with R~2 of 0.910,NRMSE of 15.14% and MARE of 8.57% for winter wheat and 0.841,27.62%,and 15.49% for summer maize,respectively.A deviation of ±1 ℃/2 ℃ in temperature forecast had a slight impact on NRMSE and MARE,but with stable R~2.The simulation precision of soil moisture was higher under the higher forecast value of temperature,and vice versa.When the precipitation forecast value was more than 10%-20%,the precision of soil moisture simulation was lower than that when the precipitation was normal.Conversely,the simulation precision was improved,but the maximum amplitudes of NRMSE and MARE were 6.16% and 3.23%,respectively,both were smaller than the bias range of precipitation forecast.When there were deviations in both temperature and precipitation forecasts,the variation of soil moisture simulation precision was mainly determined by the bias of precipitation forecast,i.e.,when the forecast value of precipitation was higher than the actual value,the simulation precision decreased,and vice versa.The maximum variations of NRMSE and MARE were 6.19% and 3.23%,respectively.