Tidal wetlands, the global hotspots of biodiversity and carbon stocks, are currently experiencing widespread modifications in their composition due to human disturbances and changing climate. Accurate mapping of tidal wetland composition is crucial and urgently required for the conservation and management of coastal ecosystem, as well as for maximizing their associated services. However, remote sensing of tidal wetlands is still challenge due to periodic tidal fluctuations, frequent cloud cover, and similar spectral characteristics with terrestrial land-cover types. Previous approaches to mapping the tidal wetlands have been restricted to small study regions or have focused on an individual tidal wetland type, thus limiting their ability to consistently monitor the composition of tidal wetlands over large geographic extents. To address the above issues, we proposed a novel algorithm on Google Earth Engine, called Multi-class Tidal Wetland Mapping by integrating Tide-level and Phenological features (MTWM-TP), to simultaneously map mangroves, salt marshes and tidal flats for specifying large-scale tidal wetland composition. The MTWM-TP algorithm firstly generates several noise-free composite images with different tide levels and phenological stages and then concatenates them into a random forest classifier for further classification. The usage of tide-level and phenological features eliminates inland landscapes and help to distinguish deciduous salt marshes and evergreen mangroves, leading to a statistically significant improvement in accuracy. We applied the algorithm to 10,274 Sentinel-2 images of East Asia and derived a 10-m-resolution multi-class tidal wetland map with an overall accuracy of 97.02% at a sub-continental scale. We found that tidal wetlands occupied 1,308,241 ha of areas in East Asia in 2020, of which 89.12% were tidal flats, 9.39% were salt marshes, and only 1.49% were mangroves. This spatially explicit map of tidal wetland composition will provide valuable guidance for coastal biodiversity protection and blue carbon restoration. In addition, the proposed MTWM-TP algorithm can serve as a reliable means for monitoring sub-continental- or larger-scale tidal wetland composition more broadly.

潮汐湿地是全球生物多样性和碳储量的热点，目前由于人类干扰和气候变化，其组成正在发生广泛变化。潮汐湿地组成的准确测绘对于沿海生态系统的保护和管理以及最大限度地提高其相关服务至关重要且迫切需要。然而，由于潮汐周期性波动、云层覆盖频繁以及与陆地覆盖类型相似的光谱特征，潮汐湿地的遥感仍然面临挑战。以前绘制潮汐湿地图的方法仅限于小型研究区域或专注于单个潮汐湿地类型，从而限制了他们在较大的地理范围内持续监测潮汐湿地组成的能力。为了解决上述问题，我们在谷歌地球引擎上提出了一种新的算法，称为多类潮汐湿地映射，通过整合潮汐水平和物候特征（MTWM-TP），同时映射红树林、盐沼和潮滩，以指定大-尺度潮汐湿地组成。MTWM-TP算法首先生成多个不同潮位和物候阶段的无噪声合成图像，然后将它们连接成一个随机森林分类器进行进一步分类。潮位和物候特征的使用消除了内陆景观，有助于区分落叶盐沼和常绿红树林，从而在统计上显着提高准确性。我们将该算法应用于 10，274 张 Sentinel-2 东亚图像，并导出了 10 米分辨率的多级潮汐湿地图，在次大陆尺度上的整体精度为 97.02%。我们发现，2020年东亚地区潮汐湿地面积为1308241公顷，其中滩涂占89.12%，盐沼占9.39%，红树林占1.49%。这幅空间明确的潮汐湿地组成图将为沿海生物多样性保护和蓝碳恢复提供有价值的指导。此外，所提出的 MTWM-TP 算法可以作为更广泛地监测次大陆或更大规模潮汐湿地组成的可靠手段。49% 是红树林。这幅空间明确的潮汐湿地组成图将为沿海生物多样性保护和蓝碳恢复提供有价值的指导。此外，所提出的 MTWM-TP 算法可以作为更广泛地监测次大陆或更大规模潮汐湿地组成的可靠手段。49% 是红树林。这幅空间明确的潮汐湿地组成图将为沿海生物多样性保护和蓝碳恢复提供有价值的指导。此外，所提出的 MTWM-TP 算法可以作为更广泛地监测次大陆或更大规模潮汐湿地组成的可靠手段。