Assessment of suspended sediment concentration in the Padma river using satellite remote sensing

Abstract

Variation of Suspended Sediment Concentration (SSC) is an important parameter in the hydrologic, morphologic and ecosystem studies of large alluvial rivers; especially in the Ganges-Brahmaputra-Meghna (GBM) delta. Traditional in situ measurement of SSC in the large Padma River is challenging in terms of time, cost, skilled personnel and spatial coverage. Moreover, there are limitations in terms of spatial and temporal acquisition of reliable data. Satellite remote sensing offers convenient assessment and spatio-temporal mapping of suspended sediments in large alluvial rivers. This study investigated the applicability of open-access Landsat Enhanced Thematic Mapper (ETM+) images in estimating the SSC of the Padma. Multiple- temporal Landsat 7 ETM+ images were processed to extract Digital Numbers (DN) of pixels corresponding to Bangladesh Water Development Board (BWDB)’s river measurement station, Mawa (SW93.5L). The DNs were converted to radiance and ultimately to top-of-atmosphere (ToA) reflectance. Since mostly clear scenes were used, in situ atmospheric correction was ignored. The ToA values for Landsat-7 bands 1-4, which sense electromagnetic radiation of 0.45-0.52, 0.52-0.60, 0.63-0.69 and 0.76-0.90 μm respectively, were combined with corresponding measured values of SSC, procured from historical data archives of BWDB, between the years 2000 to 2010 for determination of statistical relationship between them. R2 for bands 1, 2, 3 and 4 were 0.64, 0.51, 0.44 and 0.67 respectively. The results from analysis showed that Coefficient of Determination (R2) value of band 4 (Near Infrared) presented the best relationship - therefore chosen as the best SSC indicator. Scatter plot of predicted SSC values from a polynomial equation based on band 4 against in situ values of SSC with 1:1 fit line generated strong positive coefficient of determination of 0.89 and Root Mean Square Error (RMSE) of 88.3 ppm. Using a polynomial model based on the band 4 data, spatial distribution maps of SSC, between the years 2000 and 2010, for monsoon and post-monsoon seasons were demonstrated. SSC levels appeared to be generally higher in monsoon and flood seasons compared to post-monsoon season. However, there were exceptions in this observation too. Rise in discharge, water level and flow velocity increased the overall SSC. During cross-section analysis, it was generally observed that rise in bed level also caused small jumps in SSC levels. Using statistical correlation analysis of measured values of SSC and corresponding in situ values of flow velocity, a logarithmic relationship model was derived. Using these models and SSC spatial distribution maps, spatial variation maps of water flow velocity were created.