Estimating Surface Soil Water Content in the Red River Valley of the North using Landsat 5 TM Data

Due to poor drainage and flat terrain, a prolonged wet cycle since the early 1990s in the Red River Valley (RRV) of the North has caused frequent flooding in most farmland, delaying or even preventing spring planting. To combat high soil water content, farmers have adopted tile drainage practices. To schedule crop planting or to evaluate the efficiency of tile drainage and its potential impacts on water balance and quality in the watershed, soil water content and its changes need to be monitored. Compared to in situ measurements, the use of remote sensing for soil water content monitoring is affordable and can be applied over a large area repetitively. In this study multispectral reflectance of the soil sample from RRV at the bands of Landsat 5 TM sensor were evaluated for various soil water contents in the laboratory experiments. Empirically, the soil water contents at 5 and 15 cm were found to be best predicted using an exponential model based on the difference of bands 1 and 5. While the 5 cm model better represents remotely sensed soil water content, 15 cm model better represents root zone condition and therefore is more relevant for supporting field management decision. Because it was challenging to measure water content accurately at 5-cm depth in the fields, only 15 cm model was validated. Validation using a total of 70 observations over nine different fields in the RRV showed that the model compared well with the field measurements (r = 0.94) with an average difference less than the model uncertainty of 0.02 cm(3)/cm(3). The 15 cm model has an application range for soil water content between 0.20 and 0.40 cm(3)/cm(3).