土壤水分是土壤的重要组成部分，它在陆地表层和大气之间的物质和能量交换方面扮演着重要角色，寻求快速而准确的方法估算土壤水分具有重要意义。通常，从可见光一近红外对土壤表层水分的估计多是建立在土壤水分与反射率的关系之上的。而在土壤水分含量不高时，土壤水分的增加使土壤光谱反射率在整个波长范围内降低，尤其在760nm，970nm，1190nm，1450nm，1940nm和2950nm等水分吸收波段，而在土壤水分含量较高时，土壤水分的增加会使土壤光谱反射率在某些光谱波段升高。而土壤水分的估计往往是基于土壤水分与土壤水分吸收波段的吸收强度之间的线性关系上，虽然这些经验的方法对于估算某些土壤的表层水分含量是有效的，但这些关系应用于其它条件(如不同种类土壤、土壤湿度变化范围很大的情况)时却面临很多困难，这与土壤的光谱反射率是由土壤的组成成分(土壤水分、有机质、氧化铁和粘土矿物等)的含量和它们在土壤中的分布密切相关。微分技术处理“连续”的光谱是遥感中常用的数学方法，微分技术能部分消除低频光谱成分的影响。现在微分光谱已广泛地应用于研究植被的生物物理参数、矿物和有机质等。然而利用微分光谱对土壤水分反演的研究却鲜见报道。本文通过对实验室中多种不同类型的土壤进行光谱与土壤表层水分含量进行观测，探讨了通过土壤反射率与微分光谱对土壤表层水分的反演方法。4种类型的土壤光谱数据(反射率(R)，反射率倒数的对数(log(1／R))，反射率的一阶微分光谱(dR／dλ)，反射率倒数的对数的一阶微分光谱(d(log(1／R))／dλ))与土壤表层水分之间的关系在本文中得到分析，R与log(1／R)对于不同土壤类型与土壤表层水分都很敏感，说明通过R与log(1／R)反演土壤表层水分受土壤类型的影响很大，而dR／dλ，d(log(1／R))／dλ)对土壤类型却不敏感，对土壤表层水分较为敏感，说明dR／dλ和d(log(1／R))／dλ)对于反演不同类型土壤具有很大的潜力，微分光谱与土壤水分在某些波段具有显著的相关性。通过随机对9种土壤(各具有4个土壤水分)的数据建立反演土壤水分的模型，并其他9种土壤(各具有4个土壤水分)的数据进行验证模型，结果表明，dR／dλ和d(log(1／R))／dλ)能够显著提高R与log(1／R)对于不同土壤类型土壤表层水分的反演精度，由于吸收过程是非线性的，在四种类型的土壤光谱数据中，总体来说，d(log(1／R))／dλ)具有最好的能力预测不同类型土壤的表层水分含量。

Soil moisture is a very important variable in hydrologic cycle and exchange of matter and energy near ground boundary. It was investigated by the means of remote sensing due to a number of reasons. In this study, under experimental conditions, relationship between hyperspectral data and soil surface moisture has been investigated. Correlation between soil surface moisture for nine soil samples and four sets of spectral data of them (prototype reflectance, absorbance, the first-order derivative of reflectance and the first-order derivative of absorbance) was analyzed. We found that it has no obvious correlation between soil surface moisture and prototype reflectance and absorbance for all samples, while absorbance has higher correlation than reflectance. The first-order derivative of reflectance and the first-order derivative of absorbance have the obvious correlation near wavebands 1844 nm, and the first-order derivative of absorbance reflectance has more obvious correlation than the first-order derivative of reflectance. We choose these bands with high square of correlation coefficient to creat liner regression forecasting equation. Another nine soil samples were used to verify the precision of estimation equation. Results show that the first-order derivative of absorbance has the capability to estimate soil surface moisture of the four sets of data. It shows the great potential to estimate soil surface moisture within a large area for different soil types.