Spatial prediction method for dust storm source area susceptibility based on sp-gsh-el model
By using the sp-GSH-EL model, combined with aerosol optical thickness and rate of change to screen active dust sources, and by combining multiple covariates and ensemble learning algorithms, the limitations of coverage and insufficient handling of spatial heterogeneity in existing dust storm source area identification technologies have been solved, achieving more accurate prediction of dust storm source area susceptibility.
Patent Information
- Authority / Receiving Office
- CN Β· China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- INSTITUTE OF GRASSLAND RESEARCH OF CAAS
- Filing Date
- 2025-06-30
- Publication Date
- 2026-07-21
AI Technical Summary
Existing technologies for identifying dust storm source areas suffer from limitations in the coverage of ground-based observation methods and untimely data updates. Remote sensing technology has also failed to effectively handle spatial heterogeneity, resulting in low modeling accuracy across multi-scale spatial data.
The sp-GSH-EL model was adopted to screen active dust sources by aerosol optical thickness (AOD) threshold and rate of change. Combined with multiple covariates and ensemble learning algorithms, spatial weighted ensemble neural network (SWENN) was used to optimize spatial heterogeneity and improve the dynamic monitoring and modeling accuracy of the model.
It enables dynamic monitoring and precise identification of dust storm source areas, improves the modeling accuracy and interpretability of multi-scale spatial data, and provides more accurate predictions of susceptibility to dust storm source areas.
Smart Images

Figure CN121074694B_ABST