Maldivian seagrass aerial extent raster layers 2021 - 2000

Contemporary Seagrass Map (2021)The contemporary product was derived from Sentinel-2 satellite imagery, operated by the European Space Agency (ESA). The imagery, with a spatial resolution of 10 meters was pre-processed in Google Earth Engine (GEE) following established methods for retrieval of benthic signals. A support vector machine (SVM) classifier was used for classification. Training data encompassed three classes: seagrass, non-seagrass (including coral reefs, mangroves, sand/rubble, and macroalgal beds), and optical-deep water (ODW), totaling 25,463 training pixels. Important: the classification output is a binary (seagrass/non-seagrass) class. Validation of the map was conducted independently using 1,019 in-situ field survey points collected from 2017-2023. Mapping accuracy was assessed through an error matrix. Overall accuracy = 82% Historical Seagrass Maps (2000-2021)The historical mapping product is derived from Landsat data spanning 2000 to 2021. The Landsat missions, operated by the United States Geological Survey (USGS) in collaboration with NASA, provide satellite data with a spatial resolution of 30 meters. There are no suitable data for 2010-2011. Each composite, representing a two-year period, underwent radiometric normalisation relative to a reference image from 2020-2021. Training and validation data were designated using an identical methodology as the contemporary maps, with 823 validation points utilised for accuracy assessment from 2017-2023. A fixed pixel approach was adopted to assess accuracy across the entire time series, involving the manual delineation of seagrass and non-seagrass areas. Overall accuracy was >89% in all cases. These data represent GeoTIFF files of seagrass habitat extent (binary classification). Contemporary data come from habitat classification of Sentinel-2 imagery (10 m pixel size). Historical maps come from habitat classification of Landsat data (30 m pixel size). For further details of workflow and data specifications please see the original publication DOI: 10.1038/s41598-024-61088-1《当代海草床分布图（2021年）》 本数据集基于欧洲空间局（European Space Agency, ESA）运营的Sentinel-2卫星影像生成。该影像空间分辨率为10米，依托谷歌地球引擎（Google Earth Engine, GEE），按照成熟的底栖信号提取流程完成预处理。研究采用支持向量机（Support Vector Machine, SVM）分类器完成分类，训练数据集涵盖3个类别：海草、非海草（包含珊瑚礁、红树林、沙/砾石及大型藻床）与光学深水（optical-deep water, ODW），总计25463个训练像素点。重要提示：本次分类输出为二元分类结果（海草/非海草）。该分布图的验证采用2017-2023年采集的1019个独立原位实地调查点完成，通过误差矩阵开展制图精度评估，总体精度达82%。 《历史海草床分布图（2000-2021年）》 历史海草床制图产品基于美国地质调查局（United States Geological Survey, USGS）与美国国家航空航天局（National Aeronautics and Space Administration, NASA）合作运营的Landsat系列卫星2000-2021年的遥感数据生成，影像空间分辨率达30米。2010至2011年无可用有效数据。每幅代表两年时段的影像合成产品，均以2020-2021年的参考影像为基准完成辐射归一化处理。本次研究采用与当代海草床分布图一致的方法划分训练与验证数据集，以2017-2023年的823个验证点位开展精度评估。研究采用固定像素法对全时间序列开展精度验证，即通过手动勾绘海草与非海草区域实现，所有时段的总体精度均高于89%。 本数据集包含海草栖息地范围的GeoTIFF格式文件（二元分类结果）。当代海草床数据来自Sentinel-2影像的栖息地分类（像素尺寸为10米），历史海草床分布图来自Landsat影像的栖息地分类（像素尺寸为30米）。如需了解工作流程与数据规格的更多细节，请参阅原始文献DOI：10.1038/s41598-024-61088-1