Monitoring of surface water quality in large rivers with satellite imagery - Application to the Amazon basin

Monitoring river water quality with satellite data has been limited by the lack of sensors offering high revisit frequency and suitable spatial and radiometric resolutions. In this communcation, it is shown that the last generation of moderate resolution sensors makes possible to monitor efficiently large tropical basins characterized by reduced field monitoring and strong cloud coverage. Emphasis is placed on monitoring suspended sediment concentration (SSC) at the river surface and we use a water quality database available for different locations along the Amazon River in Brazil and Peru. Six years of satellite data (MODIS) are reviewed and in particular, the 250 and 500-meter surface reflectance products are considered. It is shown that MODIS coarse resolution results in significant spectral mixing between river stream pixels and river sides covered by vegetation or sand banks. An automated reflectance retrieval algorithm has been developed to assess the water endmember pixels. The retrieved reflectance allows the seasonal cycle of the SSC to be reliably monitored along the Amazon River. Using the field measurement database for validation purposes, the SSC retrieval performance can be assessed. Prediction error of the SSC lies between 40 and 60 % depending on the site considered. The retrieval performance is reviewed as a function of MODIS product resolution, site dependence and river width for the whole 2000-2006 period.