基于高光谱与电磁感应技术的干旱区绿洲土壤含水量反演研究
Inversion of soil moisture content in the oasis of arid areas using hyperspectral and electromagnetic induction instruments

为实现干旱区绿洲土壤含水量的快速、准确监测，利用采集自渭干河-库车河绿洲的84个表层（0~10cm）土壤样本，通过利用电磁感应仪（EM38）将所测解译后数据代替实测土壤含水量数据，将高光谱反射率重采样为Landsat8卫星遥感波段反射率，在选取光谱特征参数、提取敏感波段的基础上，利用偏最小二乘回归（PLSR）方法建立土壤含水量模型，将最优估算模型应用于遥感影像，实现研究区土壤含水量遥感反演。研究结果表明：（1）利用EM38所测水平模式土壤表观电导率与土壤含水量拟合效果最优，能够代替实测土壤含水量进行后续建模分析。（2）相比3种单一的光谱特征指数，利用多种光谱特征指数所建土壤含水量估算模型的建模效果更优，其干、湿各季建模集决定系数R2大于0.7，均方根误差（RMSE）均小于0.5%，RPD均大于2，能够作为有效手段估算干旱区绿洲土壤含水量。（3）不同季节土壤含水量遥感反演值与实测值决定系数R2均大于0.6，均方根误差（RMSE）均小于0.6%，显示了较高的预测精度，证明利用电磁感应技术与高光谱相结合能够实现对干旱区绿洲土壤含水量的精准、高效监测。
In order to realize rapid and accurate monitoring soil moisture content in the oasis arid areas, 84 soil samples from top soil (0~10cm) were collected from Weigan-Kuche oasis. Based on sensitive feature spectrum parameters selected by the spectral correlation with soil moisture content that measured by visible-near-infrared (VNIR) spectroscopy. The interpretation data by using the electromagnetic conduction (EM38) replaced the soil moisture content and the spectral reflectance of hyper spectral bands fit by simulation Landsat8 satellite remote sensing spectral reflectance as soil spectral reflectance. Models were established and the soil moisture content in the oasis arid areas of Weigan and Kuqa Rivers was estimated by using partial least squares (PLSR) analysis. The results showed following. (1) Soil apparent conductivity of EM38 level pattern had high correlation with soil moisture content and ECH could be used as a surrogate for moisture content. (2) The optimal estimation model was variety of spectral characteristic index. The determination coefficient R2 in dry and wet seasons were greater than 0.7, the root mean square error (RMSE) was less than 0.5% and RPD was greater than 2,this model can be used as effective method to estimate soil water moisture of oasis in arid areas. (3) The correlation with soil moisture content of remote sensing inversion and laboratory measured was high, their determination coefficients R2 were greater than 0.6, the root mean square error (RMSE) was less than 0.6%. It would be a precise and efficient method to monitor the soil moisture content in the oasis arid areas by combining with electromagnetic induction technology and hyperspectral.