Mapping of dust source susceptibility by remote sensing and machine learning techniques (case study: Iran-Iraq...

Area of Science:

• Environmental Science
• Geospatial Analysis
• Remote Sensing

Background:

• Dust storms pose significant environmental challenges in arid regions globally.
• Identifying dust source areas (DSA) is crucial for mitigating dust storm impacts.

Purpose of the Study:

• To develop a spatial map of land susceptibility to dust emissions in the Iran-Iraq border region.
• To combine remote sensing (RS) and statistical predictive models for enhanced dust source identification.

Main Methods:

• Utilized remote sensing techniques and machine learning algorithms: multivariate adaptive regression spline (MARS), random forest (RF), and logistic regression (LR).
• Prepared 152 dust source areas (DSA) data from 2005-2020, with 70% for training and 30% for validation.
• Incorporated six effective variables: soil, lithology, slope, normalized difference vegetation index (NDVI), geomorphology, and land use units.

Main Results:

• Land use was identified as the most significant factor influencing dust source areas across all models.
• The random forest (RF) model demonstrated superior performance with an Area Under the Curve (AUC) of 0.92, outperforming MARS (0.89) and LR (0.78).
• High and very high susceptibility classes covered substantial areas, indicating significant dust emission potential.

Conclusions:

• The developed susceptibility maps provide valuable insights for dust storm management.
• The findings support planners and managers in implementing strategies to control and reduce dust-related risks.
• The integration of RS and machine learning offers a robust approach for mapping dust emission susceptibility.