Satellite Remote Sensing of Suspended Sediment Concentrations in Turbid Rivers

Abstract

Suspended sediment concentration (SSC) is a critical parameter in the study of river sediment transport and water quality variation, but traditional measurements of SSC through in-situ sampling in rivers are expensive and time-consuming to perform. Thus, these methods cannot provide continuous SSC records. Although remote sensing has been applied to estimate the SSC of sea waters as well as low turbid inland waters like lakes, reservoirs and short river reaches visible within a single Landsat satellite image coverage, rivers, especially highly turbid large rivers, have largely been ignored. In fact, only few previous studies used water samples larger than 1300 mg l-1, and all used samples were smaller than 2225 mg l-1, although SSC values over 2225 mg l-1 are very common for a large number of rivers in the world.

The current research aimed to examine the potential of estimating SSC of turbid rivers using satellite data with higher spectral, spatial or temporal resolutions, respectively. The current research investigated the spectral signature of turbid river waters (up to 7468 mg l-1). Using three large Chinese rivers as examples, the current study improved understanding of the relationship between spectral data and SSC in turbid river waters through both field surveys and analysis of satellite spectra data. It was found that SSC could be estimated relatively accurately from spectral data in turbid river waters if appropriate SSC indicators were used, even when some crucial influencing factors such as in situ atmospheric conditions, types and particle sizes of suspended sediments and concentrations of other constituents were unknown.

For large SSC values up to 7468 mg l-1, appropriate SSC indicators could be single bands, band ratios and inherent optical properties (IOPs). Near-infrared (NIR) bands had much more significant relations with SSC than visible bands did in turbid waters. Although band ratios showed more significant relation with SSC, this could possibly be hindered by insufficient atmospheric correction at the visible bands. In addition, the backscattering coefficient at 850 nm seemed able to generate more accurate estimates of SSC for turbid river waters with a wide SSC range of 77¿7468 mg l-1, than single bands or band ratios. As one of IOPs, the backscattering coefficient's relation with SSC should be less site-specific.

Whereas previous studies only covered one Landsat scene, the current research found that developed regression relations could be used to estimate SSC from a number of images covering a huge area (e.g. over 1000 km long). Using high spatial resolution images which are suitable for relatively narrow river reaches, spatial distribution of SSC along a river could be mapped out to help improve the catchment management such as by studying sediment sources, effects of reservoir trapping and channel erosion/deposition as well as monitoring illegal sand dredging in river channels.

In addition, the current research found that developed regression relations could be used to estimate SSC from time-series satellite images that were acquired on a large number of days. This is beneficial for understanding dynamic changes of SSC in a river system due to climate change and human activities. Satellite data with higher temporal resolution like Terra MODIS could provide frequent information on SSC. Further research on remote sensing of SSC in turbid waters may help to monitoring SSC variations of rivers in a changing environment.