TM影像反射率对干旱区土壤含水量的响应特征

The Response Characteristics of TM Image Reflectance to the Arid Region of Soil Moisture

以TM影像对塔里木河中游土壤含水量的响应特征为研究对象，选取影像光谱（R）、光谱倒数（1／R）、光谱倒数之对数lg（1／R）和去除归一化植被指数（Rc）四种光谱指标，分别建立土壤含水最的顶测模型，并利用方差检验验证模型的显著性，采用后验差法划分模型精度级别。结果表明：光谱倒数之对数lg（1／R）预测土壤含水量的模型精度较高，且对0～30cm土壤含水量的预测精度最高，达到良好级别，适用于研究区土壤含水量的监测；影像光谱（R）和光谱倒数（1／R）的模型精度次之，仅有个别层（0-30，0～50cm等）达合格或免强合格水平；而去除归一化植被指数的模型精度较差。另外，各预测模型的最伟预测深度为0～30cm，土层深度过大或过小，其预测精度均降低。

The response characteristics of TM image to the soil moisture in the Tarim River are the research object. Selected the image spectrum （R）, spectrum reciprocal （l/R）, the logarithm of reciprocal spectrum lg（l/R） and removal normalized difference vegetation index （Re） of four spectral index were selected to establish the soil moisture content prediction model, the variance test was used to validate the model significance, the model accuracy level was divided by the posterior variance examination. The results showed that： the model accuracy of the logarithm of reciprocal spectrum ig（1/R） prediction of soil moisture is the highest, and achieved a good level for the monitoring of soil moisture content （0- 30 cm）. The model accuracy of the spectral （R） and spectral reciprocal （l/R） prediction of soil moisture is lower than logarithm of reciprocal spectrum with only the individual layers （0-30, 0450 cm, etc. ） reaching the qualified level or narrouly qualified level. The model accuracy of the removal normalized difference vegetation index （Re） prediction of soil moisture is the lowest. Besides, the best prediction depth of every model is the depth of 0-30 cm, and if the soil depth is too deep or too shallow, the prediction accuracy will decrease.