Developing a virtual sensor (VS) for mapping soil moisture at high spatial and temporal resolution.

摘要

Mapping soil moisture at both high spatial and temporal resolution has not been possible due to lack of sensors with these combined capabilities. We transformed the Moderate Resolution Imaging Spectroradiometer (MODIS) into a virtual sensor (VS) for quantitative soil moisture mapping and monitoring at 1 km and 250 m resolution daily. The Vegetation Index (VI) - Land Surface Temperature (LST) triangle model was used as the governing algorithm for VS. We used a time series of 13 data sets from August 01, 2006 to November 06, 2006 of MODIS reflective and thermal imagery and AMSR-E Level 3 soil moisture imagery to develop the VS in the semi-arid environment of southeastern New Mexico. We used Synthetic Aperture Radar (SAR) derived soil moisture imagery for five corresponding dates of the MODIS/AMSR-E imagery to evaluate the performance of VS for soil moisture estimation along with near real time in situ soil moisture measurements.;In situ soil moisture measurements, vegetation density/distribution maps, digital elevation model (DEM), soil type map and soil salinity measurements were used in both linear and non-linear numerical models with the Radarsat 1 SAR fine imagery.;The numerical models based on multiple linear regressions improved soil moisture estimation for the entire study site. We found, however, that vegetation, soil type and elevation have stronger combined effect on microwave soil moisture remote sensing by non-linear regressions (neural networks).;The accuracy of the soil moisture data was evaluated using Kappa statistics. A soil moisture prediction surface prepared by kriging the in situ soil moisture 2 measurements was used as the reference. We obtained the overall accuracy of 75.67% and 77.67% with a Kappa coefficient of 0.43 and 0.61 for the August 02 and November 06 data sets of 2006, respectively. We evaluated the application of VS generated soil moisture data in mapping the spatio-temporal variation in soil moisture in southeastern New Mexico.;The virtual sensor developed in this research has made the AMSR-E 25 km soil moisture information suitable for more local and watershed level applications by disaggregating it to 1 km and 250 m soil moisture data using MODIS reflective and thermal imagery.