GWL_FCS30: a global 30 m wetland map with a fine classification system using multi-sourced and time-series remote sensing imagery in 2020

[摘要] Wetlands, often called the “kidneys of the earth”, play an important rolein maintaining ecological balance, conserving water resources, replenishinggroundwater and controlling soil erosion. Wetland mapping is verychallenging because of its complicated temporal dynamics and large spatialand spectral heterogeneity. An accurate global 30 m wetland dataset that cansimultaneously cover inland and coastal zones is lacking. This studyproposes a novel method for wetland mapping by combining an automatic sampleextraction method, existing multi-sourced products, satellite time-seriesimages and a stratified classification strategy. This approach allowed forthe generation of the first global 30 m wetland map with a fineclassification system (GWL_FCS30), including five inlandwetland sub-categories (permanent water, swamp, marsh, flooded flat andsaline) and three coastal tidal wetland sub-categories (mangrove, saltmarsh and tidal flats), which was developed using Google Earth Engineplatform. We first combined existing multi-sourced global wetland products,expert knowledge, training sample refinement rules and visualinterpretation to generate large and geographically distributed wetlandtraining samples. Second, we integrated the Landsat reflectance time-seriesproducts and Sentinel-1 synthetic aperture radar (SAR) imagery to generate various water-level andphenological information to capture the complicated temporal dynamics andspectral heterogeneity of wetlands. Third, we applied a stratifiedclassification strategy and the local adaptive random forest classificationmodels to produce the wetland dataset with a fine classification system ateach 5 ∘ × 5 ∘ geographical tile in 2020. Lastly,GWL_FCS30, mosaicked by 961 5 ∘ × 5 ∘ regional wetland maps, was validated using 25 708 validationsamples, which achieved an overall accuracy of 86.44 % and a kappacoefficient of 0.822. The cross-comparisons with other global wetlandproducts demonstrated that the GWL_FCS30 dataset performedbetter in capturing the spatial patterns of wetlands and had significantadvantages over the diversity of wetland sub-categories. The statisticalanalysis showed that the global wetland area reached 6.38 million km 2 ,including 6.03 million km 2 of inland wetlands and 0.35 million km 2 of coastal tidal wetlands, approximately 72.96 % of which were distributedpoleward of 40 ∘  N. Therefore, we can conclude that the proposedmethod is suitable for large-area wetland mapping and that theGWL_FCS30 dataset is an accurate wetland mapping product thathas the potential to provide vital support for wetland management. TheGWL_FCS30 dataset in 2020 is freely available at https://doi.org/10.5281/zenodo.7340516 (Liu et al., 2022).

[学科分类] 眼科学