نوع مقاله : علمی - پژوهشی
نویسندگان
1 دانشجوی دکتری مدیریت حوزه های آبخیز، دانشگاه ارومیه،ارومیه، ایران.
2 دانشیار، گروه مرتع و آبخیزداری، دانشکده منابع طبیعی، دانشگاه ارومیه، ارومیه، ایران
3 محقق پسا دکتری، دانشگاه جزیره پرنس ادوارد، کانادا
چکیده
کلیدواژهها
عنوان مقاله [English]
نویسندگان [English]
Background and Objective: Meteorological drought, as one of the most catastrophic types of drought, has direct and indirect destructive impacts on society and can even influence the occurrence of other natural phenomena. Iran, due to its geographical position in the arid and semi-arid belt, is highly prone to drought, having experienced frequent and severe droughts in watersheds and forested areas over the past three decades. Accurate spatial prediction (identification of starting points) and temporal forecasting of potential drought damages at both global and national levels is highly challenging. However, in a general assessment, research has shown that damages caused by meteorological drought are more pronounced in forested areas compared to watersheds. Identifying influencing factors, measuring, predicting, and defining a natural hazard such as drought is often very difficult and impractical due to its complexity. Therefore, implementing adaptive measures tailored to the region requires the identification of key components, recognition of complex relationships, and a comprehensive understanding of influencing factors, along with the assessment of drought risks through the preparation of drought vulnerability and sensitivity maps.
Material and Methods: The geographical location of Kurdistan Province is between 34°44′ and 36°30′ N latitude and 45°31′ to 48°16′ E longitude from the Greenwich meridian. A map of drought distribution points was prepared using the Drought Severity Index (DSI) in the Google Earth Engine environment. Natural, anthropogenic, and morphometric factors involved in drought occurrence were considered. The digital elevation model was downloaded from the USGS website and used to prepare information layers of elevation, slope, and morphometric indices. Layers of precipitation, temperature, evapotranspiration, and climate were prepared from data of rain gauge, synoptic, and climatological stations across the province and interpolated using the kriging method. The Normalized Difference Vegetation Index (NDVI) was extracted from Sentinel-2 imagery, and the soil map of Kurdistan Province was obtained from the FAO soil database. Information layers were prepared in ArcGIS and SAGA_GIS software. The drought sensitivity map was generated using a hybrid machine learning model (Bagging–Random Forest). The contribution of each factor to drought occurrence was evaluated using the permutation index, and the model performance was assessed with the ROC curve in the R programming environment.
Results: The results of model evaluation using the ROC curve after preparing the drought sensitivity map with the Bagging–Random Forest method show that the model has excellent performance (AUC > 0.9) in identifying sensitive areas. The analysis of the percentage of areas with sensitivity and vulnerability within the study area indicated that about 60% of watersheds and forested areas in the province fall within the range of moderate to very high vulnerability. The results of the permutation analysis for the model in determining the factors influencing drought occurrence in Kurdistan Province revealed that climatic factors such as precipitation, temperature, climate, and evapotranspiration had the highest influence, whereas the NDVI and the mass balance index had the lowest impact.
Conclusion: The results of the hybrid Bagging–Random Forest model for Kurdistan Province showed that highly drought-sensitive areas are mainly located in the eastern and partially southern parts of the province. Since most of the forested watersheds and valuable rangelands are situated in the western parts of the province, their sensitivity to drought is relatively lower compared to other regions. However, the evaluation of the model’s predictive map in western areas, which include high-quality forests and rangelands, indicated that forested watersheds are in a more critical condition than other watersheds. Therefore, priority in future planning should be given to these areas.
کلیدواژهها [English]
)