2000-2020 年全球 30m 红树林空间分布产品 (GMF30_2000-2020)

红树林是分布在热带和亚热带地区海岸潮间带，以红树植物为主的常绿灌木或乔木组成的潮滩湿地木本植被群落，具有独特的海陆过渡特性，是海岸带具有重要的生态、经济、景观价值的湿地生态系统。该数据集利用无云或云量少于20%的多时相Landsat影像经过大气校正、波段合成等预处理，基于Capsules-Unet深度学习模型，并结合人工解译方法提取了2000、2010、2015和2020年共4期全球红树林分布数据。与国际同类产品相比，本产品时序更长，能够准确反映全球红树林动态变化的特点，验证总体精度大于88%，Kappa系数和F1指标分别大于0.76和0.88。

Mangroves are woody vegetation communities of tidal flat wetlands, primarily composed of evergreen shrubs and trees of mangrove species, distributed in the coastal intertidal zones of tropical and subtropical regions. Featuring unique transitional characteristics between marine and terrestrial ecosystems, mangroves are coastal wetland ecosystems with substantial ecological, economic, and landscape values. This dataset extracted global mangrove distribution data for four periods (2000, 2010, 2015, and 2020) by utilizing multi-temporal Landsat imagery that was cloud-free or had cloud cover less than 20% after preprocessing procedures including atmospheric correction and band synthesis, based on the Capsules-Unet deep learning model combined with manual interpretation methods. Compared with similar international products, this dataset has a longer time series, which can accurately reflect the dynamic changes of global mangroves. Its validated overall accuracy exceeds 88%, while the Kappa coefficient and F1-score are greater than 0.76 and 0.88 respectively.

搜集汇总

背景与挑战

背景概述

该数据集提供了2000年至2020年全球红树林的空间分布产品，包含2000、2010、2015和2020年四个时间点，空间分辨率为30米，基于Landsat影像和深度学习模型生成，验证精度高（总体精度大于88%），旨在支持红树林动态监测和可持续发展目标评估。