论文部分内容阅读
水资源一直是制约我国西北干旱区农业发展的关键因素。本文以新疆渭库绿洲为研究区域,选取41个土壤含水量与干旱区绿洲植被实测高光谱样本,以植被指数为桥梁,采用支持向量机回归(SVR)方法,建立干旱区绿洲土壤含水量与植被指数之间的拟合方程模型,并与多元回归(MLSR)、偏最小二乘(PLS)回归2种模型进行对比。实验结果表明:不同模型的精度各异,拟合效果由优到劣为:改进的SVR模型、PLS模型、MLSR模型,其中基于干旱区绿洲实测的植被光谱数据改进的SVR模型对土壤含水量具有较好的拟合效果,通过最优参数的定值与最优测试集的抽取,R2高达0.8916,RMSE仅为2.292,在干旱区绿洲的土壤含水量拟合中获得比较高的预测精度。而MLSR模型与PLS模型,R2值分别为0.630、0.655,RMSE为3.002与2.749。研究结果表明,因地制宜开展合理的土壤含水量反演模型规则制定是提高干旱区绿洲土壤浅层含水量监测精度的有效手段,也可为干旱区农业作物生长提供更精准的数据积累。
Water resources have always been the key factor restricting the agricultural development in the arid region of northwestern China. In this paper, 41 counties of soil moisture and oasis vegetation in arid area were selected for the study area. Using the vegetation index as the bridge, SVR (Support Vector Regression) method was used to establish the relationship between the soil moisture content Vegetation index between the fitting equation model, and with multiple regression (MLSR), partial least squares (PLS) regression model for comparison. The experimental results show that the precision of different models is different, and the fitting effect is from superior to inferior: improved SVR model, PLS model and MLSR model. SVR model based on vegetation spectral data measured in the oasis of arid area has the significant effect on soil water content The best fitting result is R2 = 0.8916 and RMSE = 2.292, which is obtained by the setting of the optimal parameters and the optimal test set. The prediction accuracy of soil moisture in the oasis of arid areas is relatively high. The MLSR model and PLS model, R2 values were 0.630,0.655, RMSE was 3.002 and 2.749. The results show that the development of a reasonable rule of soil moisture inversion model according to local conditions is an effective method to improve the monitoring accuracy of shallow soil moisture in oasis area in arid area, and provide more accurate data accumulation for the growth of agricultural crops in arid area.