The use of geographic information technologies in watershed management: Porsuk basin change analysis example

Natural watershed management is a critical strategy to support sustainable use of natural resources. In this study, Land Use and Land Cover (LULC) change research in the Porsuk Basin was conducted using Landsat satellite images on the Google Earth Engine (GEE) platform from 2000 to 2020. For change analysis in the Porsuk Basin, Landsat-8, Landsat-5, and Landsat-7 images were used in 2020, 2009, and 2000, respectively. For this purpose, four machine learning algorithms were used: Classification and Regression Trees (CART), Random Forest (RF), Support Vector Machine (SVM), and Gradient Boosted Trees (GBT). The main purpose of this study is to compare the performance of four different machine learning models. Additionally, to investigate the impact of using Landsat satellite's thermal bands and spectral indices on classification, two different datasets were employed, and among the methods utilized, the best results were achieved with the Random Forest method. For the first data set, the overall accuracy values obtained with the Random Forest algorithm were 85%, 89.38%, and 87.5% for the years 2000, 2009, and 2020, respectively. Landsat's thermal bands, NDVI, NDBI, NBI, and NDWI were added to the initial data set to enhance the classification results. For the second data set, the overall accuracy values obtained with the Random Forest algorithm were 89.35%, 91.88%, and 90.62% for the years 2000, 2009, and 2020, respectively. Porsuk Basin is divided into eight different classes: forests, sparse forests, artificial surfaces, water areas, soil, pastures, wet agricultural areas, and dry agricultural areas. It was determined that between 2000 and 2020, the residential areas in the Porsuk Basin increased by 101.02 square kilometers and the forest areas increased by 1003.19 square kilometers.