|Chen, D - UNIVERSITY OF MANCHESTER|
|Huang, J - UNIVERSITY OF MANCHESTER|
Submitted to: Remote Sensing of Environment
Publication Type: Peer Reviewed Journal
Publication Acceptance Date: September 1, 2005
Publication Date: November 1, 2005
Citation: Chen, D., Huang, J., Jackson, T.J. 2005. Vegetation water content estimation for corn and soybean using spectral indices derived from MODIS near- and short-wave infrared bands. Remote Sensing of Environment. 98:225-236. Interpretive Summary: Remotely sensed Moderate Resolution Imaging Spectroradiometer (MODIS) data obtained from the Terra satellite was evaluated for the estimation of vegetation water content (VWC) over croplands. For three decades the Normalized Difference Vegetation Index (NDVI) has been used to estimate VWC with limited success. The limitation is related to its saturation when vegetation coverage is dense. With MODIS, new approaches based on indices utilizing the longer wavelength reflective infrared range are possible. Data collected in a field experiment were used to evaluate alternative formulations and to validate retrievals. The results showed that improved, reliable retrievals are possible. This is a positive step towards the development of more robust techniques based upon a data source that is globally available on a routine basis that will be of value agricultural managers and modeling.
Technical Abstract: Vegetation water content (VWC) information can be valuable in hydrology, forestry, agriculture and meteorology. The estimation of VWC over a full crop-growing period was performed here using the near infrared (NIR) and short-wave infrared (SWIR) bands of the Terra - MODerate Resolution Imaging Spectroradiometer (Terra-MODIS), as a practical attempt of utilizing MODIS bands in evaluating VWC to date. The study was conducted in the Walnut Creek watershed, which was the focus of the Soil Moisture Experiments 2002 (SMEX02), during the crop-growing season (June 14-July 9, 2002). To improve the quality of the MODIS data products, additional processing was performed that included pixel quality inspection against quality assessment (QA) flags, geolocation correction, and mixed pixels removal. It is worthwhile to mention that for practical studies using coarse resolution data, mixed pixels need to be eliminated to accord with accuracy estimation requirements. The resulting MODIS-derived reflectance values of the NIR bands and the SWIR bands over corn and soybeans fields were validated using atmospherically corrected Landsat Thematic Mapper (TM) / Enhanced Thematic Mapper (ETM) data. Comparable results were obtained when each MODIS band was compared to its equivalent Landsat band, although slightly larger temporal fluctuations exist for some MODIS bands. General features of the MODIS reflective infrared bands (700-3000 nm) during vegetation growth were examined. Reflectance for the SWIR (1640 nm) band and the SWIR (2130 nm) band decreased as the vegetation grew but the reflectance for the NIR (1240 nm) band showed little sensitivity to vegetation growth. Furthermore the SWIR (1640 nm) band and the SWIR (2130 nm) band both had stronger correlations to VWC than the NIR (1240 nm) band during the corn-growing period. These effects are attributed to soil effects that offset the response of NIR (1240 nm) to the corn growth under partially vegetated conditions during the early growth period. It indicates that the NIR (1240 nm) bands may be less valuable in relation to corn VWC estimations. All possible combinations of the 7 MODIS bands have been used to construct VIs. Evaluation of performance have been conducted by comparing their correlations with corn VWC. The insensitivity of NIR (1240 nm) is confirmed and the Normalized Difference Vegetation Index (NDVI) and the Normalized Difference Water Index (NDWI) were found to be the best candidates. The NDVI was saturated to a point in the growing season. In this investigation it was observed that the MODIS SWIR-based VI for corn saturated at a later date than NDVI. A similar later saturation was observed for soybeans with a lag of about 10 days. This analysis of temporal features of various vegetation indices (VI) indicated that those based on SWIR appeared to be more water absorption dominated and, therefore, more promising for VWC estimation. As a result, the SWIR bands were proposed for VWC estimation using operational satellites for corn crop areas. Linear relationships between the SWIR-based VI and VWC were developed using the MODIS data and ground based measurements of VWC. MODIS-derived Normalized Difference Water Indices (NDWI) using SWIR (1640 nm) or SWIR (2130 nm), namely NDWI1640 or NDWI2130, all showed potential in estimating VWC.