Determination and Estimation of Water Surface Change With Landsat Data and Machine Learning Algorithms; A Case Study in Lake Marmara

Abstract

Water resources play an important role in the continuity of life. Therefore, it is necessary to map water resources and monitor changes. Remote sensing technologies provide important data in the monitoring, control, and protection studies of water resources. These data are important for planners in studies related to water bodies. In this study, the change of the water surface of Marmara Lake, located in Gölmarmara district, 70 km from Manisa, was determined. In addition, an estimation study of the future spatial change of Marmara Lake was carried out. In this direction, the surface areas were obtained as a result of the analysis of the Landsat 7 images of the study area for the years 2002-2021 with the unsupervised classification method. In addition, precipitation, temperature, and LST data of the area were obtained with the help of Google Earth Engine. RBF Regressor, Linear Regression, Additive Regression, and MultiLayer PerceptronCS methods were used to make the most accurate estimation using the data obtained. Using the data between 2002 and 2012, the change between 2013 and 2021 was determined. When the results were examined, it was observed that the best estimation was obtained with MultiLayer Perceptron CS with R2= 0.91. As a result of the estimation study carried out for the years 2022 and 2026 with this method, it is predicted that the lake will shrink much more and reach 1.56 km2.