Publications

Wang, W; Samat, A; Abuduwaili, J; Wang, C; De Maeyer, P; van de Voorde, T (2022). Automatic Identification of Sand and Dust Storm Sources Based on Wind Vector and Google Earth Engine. IEEE GEOSCIENCE AND REMOTE SENSING LETTERS, 19, 1006805.

Abstract
Sand and dust storms (SDS) are both symptoms and causes of desertification. As one of the essential parts of desertification control, SDS source identification can be readily carried out using remote sensing data. This letter proposes an automatic SDS source identification method based on ERA5 surface wind direction and MODIS daily surface reflectance. This is also the first time Google Earth Engine (GEE) has been used to fully automate SDS source mapping, from spatial data processing to results visualization. In this study, we use the zero-crossing edge detection algorithm to extract the dust plume edge based on the enhanced dust index (EDI) and then trace the point sources through the upwind direction. We evaluate the model performance based on 12 SDS events in Arid Central Asia (ACA) and validate against manually labeled SDS point sources. The results showed that the mean accuracy of SDS sources estimation was greater than 67%. To demonstrate the geographic scalability of the method, we also investigate the spatial pattern of SDS point sources in ACA from 2000 to 2021. Last, we adopt land cover data to discuss its applicability in time-series studies. Experimental results have demonstrated that our proposed method has the ability to identify SDS source points accurately and efficiently. We anticipate that this method will play an essential role in dust risk assessment and desertification control.

DOI:
10.1109/LGRS.2022.3199463

ISSN:
1558-0571