Skip to main content
ARS Home » Research » Publications at this Location » Publication #285506

Title: Relative skills of soil moisture and vegetation optical depth retrievals for agricultural drought monitoring

item HAN, E - Collaborator
item Crow, Wade
item Holmes, Thomas
item BOLTEN, J - National Aeronautics And Space Administration (NASA)

Submitted to: American Geophysical Union
Publication Type: Abstract Only
Publication Acceptance Date: 11/15/2012
Publication Date: 12/3/2012
Citation: Han, E., Crow, W.T., Holmes, T.R., Bolten, J. 2012. Relative skills of soil moisture and vegetation optical depth retrievals for agricultural drought monitoring[abstract]. American Geophysical Union. 2012 CDROM.

Interpretive Summary:

Technical Abstract: Soil moisture condition is an important indicator for agricultural drought monitoring. Through the Land Parameter Retrieval Model (LPRM), vegetation optical depth (VOD) as well as surface soil moisture (SM) can be retrieved simultaneously from brightness temperature observations from the Advanced Microwave Scanning Radiometer (AMSR-E). This study aims to investigate added skills of VOD in addition to SM for agricultural drought monitoring using monthly LPRM-SM and VOD products from 2002 to 2011. First, the lagged rank cross-correlation between Normalized Difference Vegetation Index (NDVI) and the SM/VOD retrievals is used to evaluate the skills of the SM and VOD for drought monitoring. Interestingly, the highest rank cross-correlation between NDVI and VOD is found with lag of (+1) month (temporally lagged behind ranks of NDVI by 1 month), while the highest rank cross-correlation coefficient of SM is found with lag (-1) month (temporally precedes the ranks of NDVI by 1 month). Lagged responses of plants to the available water capacity in the root zone may explain this lagged peak of correlation of VOD. In order to understand this finding more systematically, additional analysis on the microwave polarization difference index and vertical/horizontal brightness temperature are conducted. Next, different types of observations (SM, VOD and NDVI) and hydrologic model results (Palmer model) are merged to improve predictive power. We adopt two different merging approaches (simple weighting method and auto-regressive model) to quantify the added skills of those different drought-related indices. The results show that adding more information rather than using solely SM observation increases lag (-1) month cross-correlation coefficient with NDVI. This result indicates that different observations/models have independent information to some degree. Therefore further analysis on error-correlations between the observations/model results is also conducted. This study suggests necessity of incorporating additional available information (e.g. VOD) for advanced agricultural drought monitoring techniques. In addition, in-depth analysis on the error correlations between different drought-related observations is expected to provide useful insights for data assimilation studies.