基于随机森林算法的冻融期土壤蒸发预报模型研究

以山西省晋中盆地大田土壤蒸发实测数据和气象站监测资料为例,采用2017~2018年冻融期土壤蒸发实测数据作为因变量、影响冻融期土壤蒸发的9个因素作为自变量,利用随机森林算法构建了冻融期土壤蒸发预报模型,并对冻融期三个阶段影响土壤蒸发的因素进行了重要度评价。结果表明,冻融期土壤蒸发的预测值与实测值的平均相对误差为5.7%,均方误差为0.055,模型预测精度较高,拟合效果较好,所建立的模型适用于冻融期土壤蒸发的预报;不稳定冻结期对土壤蒸发影响较大的为地表土壤含水率和太阳辐射,稳定冻结期为降水和地表土壤含水率,消融解冻期为降水、地表土壤含水率和水面蒸发量。

Prediction of Soil Evaporation in Freeze-thaw Period Based on Random Forest Model

Based on the soil evaporation data and weather station monitoring data measured in the field of Jinzhong Basin in Shanxi Province,the soil evaporation monitored data in the freeze-thaw period of 2017-2018 were used as the dependent variable,and nine meteorological factors affecting the soil evaporation in the freeze-thaw period were used as the independent variables,random forest algorithm was used to establish the prediction model of soil evaporation in the freeze-thaw period.The importance of the factors affecting the soil evaporation in the freeze-thaw period was evaluated.The results show that the average relative error is 5.7%,RRMSEis 0.055,and the prediction effect of the model is good,which is suitable for the prediction of soil evaporation in the freeze-thaw period.The main factors affecting soil evaporation are surface water content and solar radiation during the unstable freezing period,and the main factors are precipitation and surface water content during the stable freezing period;The main factors are precipitation,soil water content and water surface evaporation during the thawing period.