Developing an empirical model from Landsat data series for monitoring water salinity in coastal Bangladesh.

This study aims to develop an empirical model from Landsat data series to monitor the water salinity of coastal Bangladesh efficiently. Such a model can substitute expensive conventional techniques for assessing remote water quality. A set of equations connecting sensors 5 TM and 8 OLI were generated using multiple regression analysis. Radiometric and atmospheric corrections were carried out to enhance the quality of satellite images. Total 13 compositions of different bands including blue, green and red were considered to find the Coefficient of Determination (r ² ) with the field level EC (electrical conductivity) values collected from 74 sampling locations. Salinity data mainly EC values of coastal water were collected from primary and secondary sources. Considering the r ² values, significant band compositions were identified and then employed to generate linear equations. Such equation for Landsat 5 TM could detect water salinity (i.e. EC) accurately of around 82%. Similarly, the r ² value for Landsat 8 OLI was found as 0.76 that can confirm the applicability of Landsat data series to detect the change of salinity level of coastal water for a long period. The availability of coastal water was delineated by NDWI whereas salinity level was assessed using the developed equations for the year 2001 and 2019. Interestingly, it was observed that coastal areas having lower level of EC almost vanished whereas those of having higher level of EC were increased significantly between 2001 and 2019. Such increase in coastal water salinity is the result of combined effects of climatic and anthropogenic factors, which can pose a considerable risk to the coastal inhabitants including freshwater scarcity, food insecurity, and health hazard.
(Copyright © 2019 Elsevier Ltd. All rights reserved.)