ASSESSMENT OF EUTROPHICATION STATUS OF SELECTED LAKES OF DHAKA CITY USING GIS AND MACHINE LEARNING ALGORITHM

Abstract

Eutrophication in Lakes are caused by the confluence of numerous nutrients (nitrogen, phosphorus etc), temperature, sunshine, dissolved oxygen, land use/land cover, socioeconomic condition, and other biophysical processes. Excess nutrients lead to algal blooms, which in turn cause fish mortality, abnormal lake conditions, and an overall disruption of the aquatic ecosystem. Dhaka city's lakes were seen decline in water quality because of excessive urbanization, uncontrolled sewerage disposal that led to Eutrophication. This research set out to evaluate the current and prospective eutrophication levels in five selected lakes in Dhaka city, Bangladesh namely Uttara, Mirpur, Baridhara, Gulshan and Hatirjheel areas. Primary water quality data were measured form collected water samples and secondary satellite imagery data were used to evaluate Trophic State Index (TSI). Calculated TSI from water quality parameters (Chl-a test & TN test) were used to compare and validate the TSI value computed from NDCI. Subsequently, Artificial Neural Network (ANN) model of machine learning algorithm was developed incorporating land use land cover (LULC) and different normalized satellite indices using 1990, 2000 and 2010 imagery that predicted the NDCI of 2021. Predicted NDCI of 2021 was validated using TSI calculated from primary data. ANN model was trained to predict the NDCI value for 2030 and 2040 to evaluate the TSI. From Carlson’s trophic state equation, TSI for Chl-a were calculated as 81.64, 82, 88.61, 88.44 and 90.99, and TSI for TN were found to be 98.53, 94.46, 102.19, 92.42 and 98.39 indicating “Hypertrophic” state for all five lakes. Calculated NDCI and corresponding TSI value from satellite imagery showed 84 to 95 percent similarity with the field measurement TSI values for 2021. Individual calculated TSI from NDCI of 1990 and 2000 imagery revealed “Supertrophic” and 2010 and 2021 imagery showed “Hypertrophic” state for all five lakes indicating lakes are in declined condition. Trained ANN model from 1990, 2000 and 2010 predicted TSI value of 2021 with 81 percent accuracy. This calibrated model further predicted NDCI and corresponding TSI for 2030 as 82.85, 82.28, 84.11, 76.38 & 79.22 and for 2040 as 87, 85.85, 88.07, 79.92 & 84.56 for all five lakes. These predicted results indicate all lakes are “Hypertrophic” condition with higher TSI values. Therefore, these lake water quality monitoring and subsequently proper management should be ensured immediately.