Document Type : Original Research
Authors
1
Ph.D Student in Desert Management and Control,Gorgan University of Agricultural Sciences and Natural Recourses, Gorgan, Iran.
2
2. Associate Professor, Department of Arid zone Management, Gorgan University of Agricultural Sciences and Natural Resources, Gorgan, Iran.
3
Associate Professor, Department of Arid zone Management, Gorgan University of Agricultural Sciences and Natural Resources, Gorgan, Iran
4
PhD in Remote Sensing and Geographic Information Systems, National Cartographic Center (NCC), Tehran, Iran.
5
Professor, Department of Remote Sensing and GIS, Faculty of Geography, University of Tehran, Tehran, Iran
Abstract
The Kara-Bogaz Gol Basin, in addition to the Karakum Desert, has been one of the primary sources of dust entering Golestan Province in recent years. This research evaluates the performance of advanced machine learning algorithms in identifying dust emission sources within the Kara-Bogaz Gol Basin. In this study, a comprehensive analytical framework based on the integration of remote sensing data and machine learning methods was employed. Environmental datasets, consisting of nine parameters - dusty days, soil moisture, soil texture, precipitation, wind speed, Normalized Difference Vegetation Index (NDVI), DEM, air temperature, and land cover - were extracted and processed using the Google Earth Engine platform for the period 2003-2023. A total of 340 dust emission sources were identified through visual interpretation of MODIS satellite imagery and used as training points for the machine learning algorithms.The results showed the following accuracy in detecting high-potential dust areas Random Forest 91.8% , Artificial Neural Network 70.9%, XGBoost 89.9% , Gradient Boosting 87.9% , Bagged CART 89.9% , LightGBM 91.8%. Among these, Random Forest and LightGBM demonstrated the best performance in identifying dust sources. The review of explainability methods revealed that three variables—vegetation index with a contribution of 27%, soil moisture (23%), and DEM (19%)—had the greatest impact on predicting dust-prone areas.The primary innovation of this research lies in developing a hybrid machine learning framework capable of identifying dust-prone areas and determining the contribution of environmental parameters to the formation of these storms.
Keywords
Subjects
The Journal of Spatial Planning and Geomatics