Extended Abstract
Introduction: Wetlands are among the most ecologically significant yet vulnerable ecosystems, playing a pivotal role in biodiversity conservation, water regulation, and climate stabilization. However, these ecosystems are increasingly threatened by both anthropogenic pressures and climate change, leading to widespread degradation. According to global assessments, more than 50% of wetlands have been lost since the early 20th century. This alarming trend underscores the urgent need for effective monitoring and management strategies, particularly in semi-arid regions where wetlands are critical for maintaining environmental balance. The Salehiyeh Wetland, located in Alborz Province, Iran, has experienced substantial environmental changes over recent decades, largely attributed to human interventions such as the construction of drainage systems. This study aims to investigate land use/land cover (LULC) changes in the Salehiyeh Wetland between 1988 and 2021 (1367–1400 in the Iranian calendar) by employing satellite remote sensing data, NDWI and NDVI indices, and the Random Forest (RF) machine learning algorithm.
Materials and Methods: This research utilized Landsat 5 TM and Landsat 8 OLI/TIRS imagery to analyze temporal variations in land cover within the wetland and its surrounding areas. The satellite data underwent standard preprocessing, including geometric and radiometric corrections, to ensure optimal data quality. To quantify water bodies and vegetation cover, the Normalized Difference Water Index (NDWI) and the Normalized Difference Vegetation Index (NDVI) were computed. These indices were employed to distinguish between water surfaces and vegetation, providing critical insights into the wetland’s hydrological and ecological dynamics. For land cover classification, the Random Forest (RF) algorithm was applied, a robust and widely-used machine learning technique known for its high classification accuracy and ability to handle large, multidimensional datasets. The classification results were evaluated using traditional accuracy assessment metrics, including Overall Accuracy (OA) and Kappa Coefficient (Kappa), which are standard indicators for the reliability of remote sensing-based classifications.
Results and Discussion: The analysis of NDWI revealed that the highest water extent in the study area occurred in 1988, with a peak NDWI value of 0.64, indicating a substantial water coverage. However, over the 42-year period, a marked decline in aquatic areas was observed. Specifically, the extent of water bodies decreased from 2.64% of the total land area in 1988 to 0.05% in 2021, signifying a dramatic reduction in wetland habitats. In terms of land cover dynamics, rangelands predominated the landscape in 1988, 1998, and 2008, occupying over 49% of the study area. However, this trend reversed significantly by 2021, as large portions of rangeland were converted into agricultural lands. By 2021, agricultural areas expanded to cover 2,817.84 km², accounting for over 45% of the total area, marking a clear transition towards more intensive land use. Urbanization and the expansion of built-up areas also showed a significant upward trend. From 1988 to 2021, urban and residential areas increased from 46.25 km² to 770.21 km², reflecting growing human encroachment on natural ecosystems. This expansion further exacerbates the pressures on the wetland, as urban sprawl often leads to the drainage of surrounding water bodies. The performance of the RF classifier was exemplary, with overall classification accuracy exceeding 93%, and the Kappa coefficient exceeding 0.90, confirming the high reliability of the model for accurately detecting long-term LULC changes in the region.
Conclusion: This study provides compelling evidence of the profound environmental transformations occurring in the Salehiyeh Wetland over the past four decades. The construction of drainage systems, agricultural intensification, and urban sprawl have all contributed to the drastic reduction of natural water bodies and rangelands. The integration of satellite remote sensing data with advanced machine learning techniques, such as the Random Forest algorithm, offers a highly effective framework for monitoring LULC changes over time. The high accuracy of the results highlights the potential of these methodologies for large-scale environmental monitoring and decision-making. Given the ecological importance of wetlands, it is imperative that policymakers adopt comprehensive conservation and restoration strategies to mitigate further degradation. A balanced approach that integrates urban development, agricultural needs, and ecological preservation is crucial for ensuring the long-term sustainability of wetland ecosystems. Continuous remote sensing monitoring should be prioritized to support evidence-based decision-making and adaptive management in the face of ongoing environmental challenges.