Document Type : Original Article

Authors

• Rasoul Lotfi ¹
• Hamed Faroqi ²

¹ Master Graduate, Civil Engineering School, Engineering Department, University of Kurdistan, Sanandaj, Iran

² Assistant Professor, Civil Engineering School, Engineering Department, University of Kurdistan, Sanandaj, Iran

Abstract

Accurate knowledge of Land Use and Land Cover (LULC) changes play a vital role in sustainable natural resource management, urban planning, and environmental monitoring. This study aims to evaluate and compare the performance of machine learning algorithms and statistical methods for land use classification in the Naysar area, a suburb of Sanandaj city.
In this study, Sentinel-2 satellite data, along with a set of input features including spectral bands, spectral indices (NDVI, MNDWI, SAVI, and NDBI), and topographic information, were utilized. Three classification approaches—Random Forest (RF), Maximum Likelihood Classification (MLC), and unsupervised K-means clustering—were implemented. Training and validation samples were collected through visual interpretation of high-resolution imagery and field surveys, and were used for model development and accuracy assessment.
The results showed that the Random Forest algorithm achieved the best performance, with an overall accuracy of 98% and a Kappa coefficient of 0.95, while the Maximum Likelihood method and K-means algorithm achieved overall accuracies of 95% and 91%, respectively.
The findings indicate that the integration of spectral and topographic features with the Random Forest algorithm provides an efficient and reliable approach for accurate land use classification and environmental change monitoring in urban and peri-urban areas.

Keywords

• Land Use and Land Cover Changes (LULC)
• Sentinel-2
• Random Forest
• Machine Learning
• Satellite Image Classification

Main Subjects

• Road and transportation