Abstract. The extent of surface water has been changing significantly due to climatic change and human activities. However, it is challenging to capture the interannual changes of inland water bodies due to their high seasonal variation and abrupt change. In this paper, a global annual surface water cover frequency dataset (GLOBMAP SWF) was generated from the MODIS land surface reflectance products during 2000–2020 to describe the seasonal and interannual dynamics of surface water. Surface water cover frequency (SWF) was proposed as the percentage of the time period when a pixel is covered by water in a year. Instead of determination of the water directly, the SWF was estimated indirectly by identifying land observations among annual clear-sky observations to reduce the influence of clouds and variability of water bodies and surface background characteristics, which helps to improve the applicability of the algorithm for different regions across the globe. Regional analysis demonstrates that our estimation results show better performances on frozen water, saline lake, bright surface and cloud-frequent regions compared with the two high-frequent surface water datasets derived from MODIS data. Compared with the high-resolution SWF maps extracted from Sentinel-1 data in four regions that cover lake, river and wetland, the R² reaches 0.83 to 0.97, RMSE is ranging from 7.24 % to 11.28 %, and MAE is between 2.07 % and 7.15 %. In 2020, the area of global maximum surface water extent is 3.38 million km², of which the permanent surface water accounts for approximately 54 % (1.83 million km²) and the other 46 % is intermittent surface water (1.55 million km²). The area of global maximum and permanent surface water has been shrinking since 2001, with a change rate of -7577 km²/yr and -4315 km²/yr (p < 0.05), respectively; while the intermittent surface water with the SWF above 50 % has been expanding (1368 km²/yr, p < 0.01). This dataset can be used to analyze the interannual variation and change trend of highly dynamic inland waters extent with consideration of its seasonal variation. The GLOBMAP SWF data are available at https://doi.org/10.5281/zenodo.6462883 (Liu and Liu, 2022).

中文翻译：

---

GLOBMAP SWF：2000-2020 年全球年度地表水覆盖频率数据集，用于内陆水体变化分析

摘要。由于气候变化和人类活动，地表水的范围发生了显着变化。然而，由于内陆水体的高季节性变化和突变性，很难捕捉到内陆水体的年际变化。在本文中，从 2000-2020 年 MODIS 地表反射率产品生成了全球年度地表水覆盖频率数据集 (GLOBMAP SWF)，以描述地表水的季节和年际动态。地表水覆盖频率（SWF）被提出为一年中一个像素被水覆盖的时间段的百分比。代替直接测定水分，SWF通过在年度晴空观测中识别陆地观测间接估计，以减少云的影响以及水体和地表背景特征的可变性，这有助于提高算法在全球不同地区的适用性。区域分析表明，与来自 MODIS 数据的两个高频地表水数据集相比，我们的估计结果在冻水、盐湖、明亮地表和云频繁区域的性能更好。与从湖泊、河流和湿地四个区域的 Sentinel-1 数据中提取的高分辨率 SWF 地图相比，R 区域分析表明，与来自 MODIS 数据的两个高频地表水数据集相比，我们的估计结果在冻水、盐湖、明亮地表和云频繁区域的性能更好。与从湖泊、河流和湿地四个区域的 Sentinel-1 数据中提取的高分辨率 SWF 地图相比，R 区域分析表明，与来自 MODIS 数据的两个高频地表水数据集相比，我们的估计结果在冻水、盐湖、明亮地表和云频繁区域的性能更好。与从湖泊、河流和湿地四个区域的 Sentinel-1 数据中提取的高分辨率 SWF 地图相比，R²达到 0.83 到 0.97，RMSE 介于 7.24 % 到 11.28 % 之间，MAE 介于 2.07 % 到 7.15 % 之间。2020年全球最大地表水面积为338万平方公里，^其中永久地表水约占54%（183万平方公里^），其余46%为间歇性地表水（155万平方公​​里^）。全球最大和永久地表水面积自2001年以来一直在缩小，变化率分别为-7577 km ² /yr和-4315 km ² /yr（p < 0.05）；而SWF高于50%的间歇性地表水一直在扩大（1368 km ² /yr，p< 0.01)。该数据集可用于分析高度动态的内陆水域范围的年际变化和变化趋势，同时考虑其季节变化。GLOBMAP SWF 数据可在 https://doi.org/10.5281/zenodo.6462883 获得（Liu 和 Liu，2022 年）。