A Model for remotely estimating water quality parameters in inland water bodies based on Landsat ETM+ data

Abstract

Monitoring of water quality parameters of inland water bodies and small inland water bodies specifically can be operationally expensive and involving. This work tries to explore how remote sensing could be used as a tool in monitoring selected water quality parameters in small inland water bodies in the Czech Republic. Models were created based of Landsat 7 ETM+ imagery based on acquisitions from autumn of 2011 until the Spring of 2015 to estimate water quality parameters. The images used were scenes of WRS-2, path and row 191/25 as well as 190/25 respectively. Samples were taken from 13 water bodies (Bunkov, Melice, Jezero, Oplatil, Pohranovsky rybnik, Velka Cerna, Opatovicky pisnik, Ujezdsky rybnik, Bohumilecsky rybnik, Spravcicky pisnik, pisnik Hradek, Bohdanecsky rybnik, Placicky rybnik), with water body area between 8-90 ha, around Pardubice and Hradec Kralove in the Czech Republic. The samples were analysed for chlorophyl-a, Total Carbon, Total Nitrogen, Total Organic Carbon, Temperature and Secchi Disk Depth. The 3×3 average window was used to limit the effect of noise on the images and water only mask was used in order to limit the process to open water areas specifically. Models were developed based on best fit charts of water quality parameter vs. satellite band combination using linear function in the regression analysis. Models created for the water quality parameters had their performance tested based on r2, RMSE and NRMSE. The best models found are (WQ parameter | formula | r2 | RMSE | NRMSE). Band combination L3/L1 or the vice versa had best fit for most the models created. The exception was temperature which was based on surface radiance (brightness temperature in [K]). Where L1-L3 are visible bands of atmospherically corrected Landsat 7 SR product. The models created are intended to help institutions that are mandated in monitoring water bodies.