Publications

Zheng, ZJ; Yu, JSD; Zhang, XY; Du, SH (2024). Development of a 30 m resolution global sand dune/sheet classification map (GSDS30) using multi-source remote sensing data. REMOTE SENSING OF ENVIRONMENT, 302, 113973.

Abstract
Accurate information of sand dune/sheet (SDS) spatial distribution is required by global-scale environmental assessment. However, the great spatio-temporal heterogeneity of SDS and the confusion of SDS to similar land cover types lead to variant feature representations and inadequate classification samples; thus, existing SDS mapping focuses on the regional scale, and the high-quality global SDS map is still lacking. In this study, we proposed a classification strategy and for the first time developed a 30 m resolution global SDS map (GSDS30) in the year 2017 that contained two SDS types, i.e., shifting SDS and semi-fixed/fixed SDS. The proposed strategy started by determining the initial mapping extent of SDS. Then, multi-source features were extracted to capture SDS variations and enlarge the separability of diverse land covers. Thirdly, we developed a prior-constraining Knearest neighbor method to collect global SDS samples. Finally, a local random forest classifier was applied to generate classification results. The evaluation based on two validation sample sets showed good performance on the overall classification accuracy (88.73% and 87.92%), the Kappa coefficient (0.852 and 0.843), and the producer's and user's accuracies for two SDS types. Based on GSDS30, we found that global SDS occupied an area of 10.43 million km2, in which shifting SDS accounted for 60.41% and semi-fixed/fixed SDS accounted for the remaining 39.59%. The majority of SDS was distributed in Africa, Asia, and Oceania, where Australia had the greatest area of SDS. The combination of GSDS30 and population data showed that global SDS was inhabited by 14.5 million people, and 94.95% of them lived in semi-fixed/fixed SDS. This study is the first attempt to automatically produce global SDS map, which is not only necessary for representing the distribution of SDS, but also valuable in managing aeolian desertification and promoting sustainable development.

DOI:
1879-0704

ISSN:
10.1016/j.rse.2023.113973