Estimating soil moisture content using laboratory spectral data

Monitoring soil moisture is important for agriculture and forestry and plays an essential role in land surface processes as well as providing feedback among the earth's surface ecosystems. Large-scale regional soil moisture spatial data can be obtained with a reliable and operational approach using remote sensing. In this paper, we provide an operational framework for retrieving soil moisture using laboratory spectral data. The inverted Gaussian function was used to fit soil spectral data, and its feature parameters, including absorption depth (AD) and absorption area (AA), were selected as variables for a soil moisture estimate model. There was a significant correlative relationship between soil moisture and AD, as well as AA near 1400 and 1900nm. A one-variable linear regression model was established to estimate soil moisture. The model was evaluated using the determination coefficients (R-2), root mean square error and average precision. Four models were established and evaluated in this study. The determination coefficients of the four models ranged from 0.794 to 0.845. The average accuracy for soil moisture estimates ranged from 90 to 92%. The results prove that it is feasible to estimate soil moisture using remote sensing technology.