| چکیده انگلیسی مقاله |
Introduction and Goal
The analysis of land use and land cover patterns plays a significant role in resource management and conservation planning, while providing a foundation for systematic approaches to environmental structures. Urban development has notably impacted land use and land cover. As a result, over half of the Earth's surface has undergone changes in recent years, with more than one-third of the land area dedicated to agricultural use. Given these substantial transformations, land use managers and experts have examined the hydrological impacts of land use changes. In this context, machine learning methods such as Artificial Neural Networks (ANN), Support Vector Machines (SVM), Random Forest (RF), Decision Trees (DT), and other models have received considerable attention for land use/land cover (LULC) classification. Planners and managers can utilize predicted LULC changes to promote sustainable land management and reduce adverse consequences. Consequently, detecting and predicting land use changes (LULC) driven by rapid urbanization can lead to disruptions in environmental sustainability. Population growth, economic development, and agricultural expansion are factors contributing to various changes in land covers, including vegetation and water. These persistent land use changes may result in environmental degradation. On the other hand, the intensity of these changes, in response to global population growth and increasing food demands, further emphasizes the need for precise research in this area. Therefore, this study aimed to observe land use changes in 2000, 2014, and 2024 in the Darab region. Additionally, the driving forces behind LULC changes were identified, and the Cellular Automata-Artificial Neural Network (CA-ANN) model was used to analyze predicted land use patterns and trends from 2034 to 2044.
Materials and Methods
Land use classification of satellite images was performed using a pixel-based supervised classification approach in the Google Earth Engine environment. The Support Vector Machine (SVM) model was employed for land use classification. Following the analysis, the study area was categorized into five distinct land use classes: rangelands, barren lands, orchards, agricultural lands, and urban areas. Since modeling land use transitions and predicting future scenarios is essential for spatiotemporal change assessment, the Cellular Automata-Artificial Neural Network (CA-ANN) method was used to forecast land use changes. The MOLUSCE plugin in QGIS was applied to analyze spatiotemporal changes (2034–2044) and compute Land Use/Land Cover (LULC) transitions for generating change maps. A transition potential matrix was also created for the years 2000–2014 to produce a change map. The Multilayer Perceptron Neural Network (ANN-MLP) was utilized for transition potential modeling, with slope, aspect, elevation, and distance from roads, faults, and rivers serving as spatial input metrics. The ANN-MLP structure processed input data through hidden layers, with the output layer containing reclassified LULC categories.
Results and Discussion
The LULC change maps (2000–2024) revealed a significant expansion of agricultural land in the region. Findings indicated that the natural ecosystems of the plain would face serious challenges due to increasing land use transitions. These changes stem from unsustainable exploitation of resources, driven by human activities such as urbanization, improper farming, excessive well-digging, and overuse of groundwater for irrigation, which may lead to soil erosion and desertification. Additionally, agricultural land conversion was accompanied by substantial changes in barren lands. The 2034–2044 prediction maps showed a continuous rise in agricultural land use, increasing from 65.455 km² (18.52%) in 2034 to 81.708 km² (28.81%) in 2044. The simulation accuracy was 82.43%, with a Kappa coefficient of 0.72. The study confirmed that physical and socio-economic factors significantly influenced landscape patterns. Geographical variables were selected for model calibration due to their strong correlation with LULC changes. Physical factors (e.g., terrain, climate) were the primary drivers of human activities, while socio-economic factors (e.g., population growth, GDP) also impacted LULC dynamics.
Conclusion and Suggestions
LULC pattern changes negatively affect groundwater quality and threaten food security. The SVM model proved effective for precise land cover monitoring. The study highlights that any land use modifications must follow scientific, data-driven planning, incorporating remote sensing and GIS technologies to prevent irreversible environmental damage (e.g., groundwater depletion, desertification, land subsidence) in Darab Plain. The projected LULC maps should serve as a guideline for land-use planners and policymakers to regulate agricultural expansion, urbanization, and orchard development. The decline in barren lands in simulated maps reflects rapid population growth, rising demand, and land conversion to farms, orchards, and residential zones as key drivers in Darab. |