Statistical analysis of earth observing data for physicochemical water quality parameters estimation for Lake Beseka, Northern main Ethiopian rift, Ethiopia

Abstract

Water quality deterioration in the Main Ethiopian Rift Lakes is one of the problems affecting the health and socioeconomic development of the area. This study was aimed to develop a method to assess the water quality change of Lake Beseka using satellite image reflectance data and observed data. Multi-temporal Landsat imagery and three variables were used to assess the physicochemical environment: pH, EC, and TDS. Linear regression and correlation were accomplished between observed water quality parameters and surface reflectance of different band operations. The best-fitted linear regression equation for pH was found on SWIR1 while for EC and TDS, the blue/green band ratio was highly fitted. The Pearson coefficients of correlation for pH, EC, and TDS were − 0.79, 0.86, and 0.85, respectively. The RMSE and p-value for validation analysis were 0.25/0.005, 548/0.003, and 367/0.004 for pH, EC, and TDS, respectively. The estimated pH, EC, and TDS were higher in the central portion of the lake grouped under brackish water and lower in the southwestern, southern, and northeastern shores. The estimated EC and TDS for the years 2018 and 2021 were relatively lower than in 2007 and 2023. The spatiotemporal lake water variations were due to anthropogenic and geogenic factors including hot springs and groundwater discharge to the lake, water-level rise and depth variation, and evaporation. These findings are helpful for understanding linear regression equations that developed from surface reflectance and observed data, which are essential to estimating and monitoring water quality change in a cost-effective and time-saving manner.