|
A Comparative Analysis of Machine Learning Algorithms for Monitoring Land Cover Changes (Case Study: Eyvan County, Ilam)
|
Ali Mahdavi  |
|
|
|
Abstract: (18 Views) |
Monitoring land use and land cover (LULC) changes is crucial for sustainable natural resource management, particularly in fragile ecosystems like the Zagros forests. This study aimed to model and analyze LULC changes in Eyvan County, located in the Zagros forests of Ilam Province, using machine learning algorithms and Landsat satellite imagery. Landsat satellite images (TM and OLI sensors) from three time periods (2000, 2010, and 2020) were utilized. Following image preprocessing, land cover classification into five classes (water, orchard, forest, rangeland, and cropland) was performed using two machine learning algorithms: Artificial Neural Network (ANN) and K-Nearest Neighbors (KNN), incorporating 14 spectral indices (including NDVI, SAVI, and NDWI). The models were trained with 795 training samples and their accuracy was evaluated using overall accuracy and Kappa coefficient metrics. Accuracy assessment clearly demonstrated the superior performance of the Artificial Neural Network algorithm (with an overall accuracy of 83% and a Kappa coefficient of 0.78) over the K-Nearest Neighbors algorithm (with an overall accuracy of 74% and a Kappa coefficient of 0.63). Trend analysis using the ANN model over the 20-year period (2000-2020) indicated a significant 9% decrease in forest cover alongside a substantial expansion of cultivated areas, with cropland and orchard areas increasing by 23% and 18%, respectively. Water bodies also showed a 23% increase, primarily associated with the construction of the Kengir Dam in the region. In contrast, the KNN algorithm estimated changes more conservatively and with lower intensity, and in some cases, even showed trends opposite to those identified by ANN. This notable discrepancy in results clearly demonstrates the superiority of the non-linear ANN model in simulating the complex dynamics of land use changes compared to the inherent simplicity of the KNN model. The findings confirm the high effectiveness of integrating remote sensing and machine learning, particularly the Artificial Neural Network algorithm, for monitoring LULC changes. This approach can serve as a powerful tool for developing conservation strategies and sustainable land management in the Zagros region. The use of higher spatial resolution data (e.g., Sentinel-2) and hybrid algorithms is recommended for future studies.
|
|
| Keywords: Satellite imagery, Artificial Neural Network, Land use, Landsat satellite, Land management |
|
|
|
Type of Study: Research |
Subject:
Special
Received: 2025/10/29 | Revised: 2026/02/15 | Accepted: 2026/02/15 | Published: 2026/02/15 | ePublished: 2026/02/15
|
|
|
|
|
|
|
| Add your comments about this article |
|
|
|
