Calibration and Validation of Salt-resistant Hyperspectral Indices for Estimating Soil Moisture in Arid Land

Soil moisture plays a very important role in the exchange of mass and energy between the surface and the atmosphere. For this reason, monitoring of soil moisture over extended areas is highly desirable. Conventionally, evaluation of soil moisture has been based on local measurements. The microwave domain is currently the privileged spectral region exploited to derive soil moisture. Now, hyperspectral remote sensing renders a different approach in image processing techniques. Although some studies have shown that hyperspectral data can be used to quantify characteristics of soil moisture, very few researchers have exploited this potential. The overall trend of soil spectral reflectance varying with soil moisture has been studied previously by numerous researchers who concluded that the surface water content in soils can be cautiously estimated using hyperspectral data, but this approach has not yet been fully understood in saline soils of arid ecosystem and developed into a more innovative direction for use in remote sensing.

Therefore, researchers examined the sensitivities of eight different index types of soil moisture hyperspectral indices to salt content based on laboratory controlled experiments with a wide range of soil moisture content and salt content (dataset I, 196 spectra in total). The main objectives of this research were: (1) investigating the spectral reflectance changes with soil surface moisture varying from very low to very high moisture contents for different wavelengths and for a range of soil salt contents; (2) identifying soil water hyperspectral indices that are insensitive to soil salinity; and (3) developing a model using the best soil moisture hyperspectral index for estimating soil water content in the field.

They were then validated using field in situ measurements (dataset II) as well as data extracted from the same locations from a Hyperion image (dataset III). The dD_{1300, 1970} was identified as the robust salt-resistant index to estimate soil moisture, with R² of 0.57 and 0.65 corresponding to Datasets I and II, respectively. However, the performance of the index somehow dropped when Dataset III (R² = 0.29, P < 0.01) was used, mostly due to the quality of Hyperion data resulting from the propagated errors from atmospheric corrections. With improved accuracy from atmospheric corrections and development of new satellite-borne sensors, it is foreseen that the dD_{1300, 1970} index identified in this study should have wide application in the mapping of soil moisture in saline soils of arid lands.

This work has been published on Journal of Hydrology, 2011, 408(3-4): 276-285. It can be linked from: http://www.sciencedirect.com/science/article/pii/S0022169411005385. This study was supported by the 100 Talents Program of The Chinese Academy of Sciences, the National Science Foundation of China (Grant No. 41071238), and the JSPS Project (No. 21403001).

Reflectance spectra of soil samples (air-dried) as affected by soil salt content (from 0.8% to 20%) in the laboratory experiment (dataset I), soil moisture content was close to 0%.