Özet:

Detection of land use and land cover (LULC) based on detailed, current, and accurate information plays an important role in natural resource management and spatial planning processes. However, there are some limitations such as data processing and storage requirements that hinder monitoring of land use dynamics at the regional scale. GEE is an open-source, free cloud platform that enables processing of geographic data at the global scale. The aim of this research is to obtain the LCLU map of Izmir province using different machine learning algorithms on GEE and to compare the results of the classification algorithms used. Sentinel-2 multi-band satellite images with a spatial resolution of 10m for the year 2022 and various UA indices were used in the study. The broad-scale LCLU classes used in the study were determined as 'Agricultural Area', 'Forest Area', 'Human Surfaces', 'Open Surfaces' and 'Water Surfaces'. Classification and Regression Tree (CART), Support Vector Machine (SVM), and Random Forest (RF) machine learning algorithms were used in the study, and the Producer's Accuracy (PA), User's Accuracy (UA), Overall Accuracy (OA), and Kappa Coefficient of each classifier were calculated. As a result, the RF classification algorithm with a GD of 97.2% and a Kappa value of 95.7% had the highest classification accuracy. The SVM algorithm with a GD of 96.1% and a Kappa value of 94.9% was the second highest accuracy algorithm. The GD of the CART algorithm was calculated as 93.3% and the Kappa value was 91.4%. Therefore, it was found that the RF method produced better results than the CART and SVM methods. On the other hand, it is seen that the overlap of reflection between open surfaces and human surfaces and bare agricultural areas especially in classification models makes it difficult to distinguish these classes.

Anahtar Kelimeler: