Different responses of MODIS-derived NDVI to root-zone soil moisture in semi-arid and humid regions

Surface representation of the root-zone soil moisture is investigated so that feasibility of using optical remote sensing techniques to indirectly map root-zone soil moisture is assessed. Specifically, covariation of root-zone soil moisture with the normalized difference of vegetation index (NDVI) from Moderate Resolution Imaging Spectroradiometer (MODIS) is studied at three sites (New Mexico, Arizona, and Texas) selected from the Soil Climate Analysis Network (SCAN). The three sites represent two types of vegetation (shrub and grass) and two types of climate conditions: semi-arid (New Mexico and Arizona) and humid (Texas). Collocated deseasonalized time series of soil moistures at five depths (5 cm, 10 cm, 20 cm, 50 cm, and 100 cm) and NDVI (8-day composite in 250 m resolution) during the period of February 2000 through April 2004 were used for correlation analysis. Similar analysis was also conducted for the raw time series for comparison purposes. The linear regression of both the deseasonalized time series and the raw time series was used to estimate root-zone soil moisture. Results show that (1) the deseasonalized time series results in consistent and significant correlation (0.46–0.55) between NDVI and root-zone soil moisture at the three sites; (2) vegetation (NDVI) at the humid site needs longer time (10 days) to respond to soil moisture change than that at the semi-arid sites (5 days or less); (3) the time-series of root-zone soil moisture estimated by a linear regression model based on deseasonalized time series accounts for 42–71% of the observed soil moisture variations for the three sites; and (4) in the semi-arid region, root-zone soil moisture of shrub-vegetated area can be better estimated using NDVI than that of grass-vegetated area.