Automatic detection and analysis of changes in water bodies

Remote sensing is widely used in agriculture, forestry, vegetation monitoring, and water resources management. Changes in lake surface areas are important indicators for ecosystem assessment and sustainable water management. This study aims to automatically detect water surface area changes in 10 lakes (Akşehir, Eber, Tuz, Salda, Manyas, İznik, Van, Beyşehir, Eğirdir, and Burdur) with different formation origins and water chemistry characteristics using satellite imagery from 1984 to 2025. The main data sources were Landsat 5 TM, Landsat 8 OLI/TIRS, Landsat 9 OLI-2/TIRS-2, and Sentinel-2 MSI images. CHIRPS precipitation and ERA5-Land potential evaporation data were used to support hydroclimatic assessments. All data were processed on the Google Earth Engine platform. Water surfaces were classified using spectral indices and Otsu thresholding, including the Modified Normalized Difference Water Index (MNDWI), Normalized Difference Vegetation Index (NDVI), Normalized Difference Water Index (NDWI), and Automated Water Extraction Index (AWEI). Random Forest (RF), Support Vector Machine (SVM), Classification and Regression Tree (CART), and K-Nearest Neighbors (KNN) algorithms were also applied for the analysis years 1984, 1990, 1995, 2000, 2005, 2010, 2015, 2020, and 2025. Significant decreases were detected in lake water surface areas. Lake Akşehir lost its entire water surface area (100%), while Lake Tuz lost 93.68% of its area. Among the machine learning algorithms, RF achieved the highest performance with an overall accuracy of 99.25%. The study presents a generalizable and updatable remote sensing workflow for monitoring lakes with different formation processes and characteristics and provides a scientific basis for different disciplines involved in sustainable water resources management.