基于高光谱特征的土壤水力特性参量反演方法研究

打开文本图片集

中图分类号：TV131 文献标识码：A 文章编号：1001-9235（2025）07-0101-14

Abstract：Toenrichemoteinversionmethodsofsoilhdraulicparametersandachieveafastandnondestructiveprediction，59high spectrali-situsoilsamplesinXiangzhouDistrict，ZuhaiCityrecollctedusingFieldSpec4spectrometer.Theremotehdraulic parameter models were established for two land types： grassland and bare land（saturated water content θε ，residual water content θr， inverseofairentryvalue α ，shapecoefficient n ，saturatedhydraulicconductivity Ks， and soil moisture content θ ）.Theresultsare as follows. ① The spectral curves exhibit clearlinearityin three bands：700～750nm，830～1100 nm，and1520～1 620 nm，with the mean values of coefficient of determination for the linear fitting R2all over 0. 94. ② Among gradient boosting regression（GBR）， partial leastsquaresregesionPLSR）andandoforest（RF）GBRpeforstheestandsoshighsensitivitytotelinearfitigater lga2（the logarithm of the slope） in the 83O-11OO nm band. It accurately predicts θε ， α n ，and θ forgrassland types（Relative Percent Deviation， RPD > 1.4）.For the bare land types，the coefficient of determination Rp2 for all hydraulic characteristic parameters， except soil moisture content θ ，is greater than O.9，and RPD exceeds 2.0. ③ In most hydraulic characteristic models predicting the grassland types，thefeatureimportanceofteleafareaindex（LA）isteghestTerefore，anectieinversionmodelisuiltbasedonfield spectraldataandLAI，withsimpleaccsstoiputparametersandgoodpredictionefect，providingapotentialforwideapplicatioin the future.

Keywords： unsaturated soils; hydraulic parameters; spectrum; machine learning;remote sensing inversion

土壤水力特性在研究土壤水分运动、溶质运移等方面具有重要作用[1]。（剩余17148字）