ABSTRACT The dynamic nature of continental water body covers is a main indicator of the status of this ecological capital and its trends on the land surface. With synthetic aperture radar (SAR), various water coverage processes can be identified using satellite data. This study proposes the use of a basin-level analysis with SAR images at a medium spatial resolution to perform a regional spatial-temporal analysis for detecting water presence. It develops a methodology to improve the separation of water bodies from other land covers by using a texture analysis, a support-vector machine classification, and the gray level co-occurrence matrix (GLCM). GLCM-mean texture parameter was found to offer sufficient texture information for separating the water class from the land class by providing a more homogeneous measurement of water bodies. The case under study is the Grijalva River Basin, Mexico, a region that requires more in-depth research so that the management of surface water coverage can be integrated with ecosystems, particularly those that are subject to intense human activity. The entire basin was mapped using a series of observations from Sentinel-1. This study produced a total of 36 water body maps with an acceptable classification accuracy of over 90% for each year studied (2016, 2017, and 2018). Water presence was also quantified, resulting in a set of maps containing regional detail, from which temporal data can be obtained on areas with water presence year-round and areas with seasonal flooding. Frequency maps with a 10-km unit cell as the spatial unit were used to detect trends and thereby identify water distribution patterns. The results show the importance of ongoing, basin-level quantifications of zones with dynamic water presence and those with stable water presence.

中文翻译：

---

使用 SAR 数据分析墨西哥盆地尺度上水存在的短期时空动态

摘要 大陆水体覆盖的动态性质是反映这一生态资本状况及其在地表上的趋势的主要指标。借助合成孔径雷达 (SAR)，可以使用卫星数据识别各种水域覆盖过程。本研究建议使用具有中等空间分辨率的 SAR 图像的盆地级分析来执行区域时空分析，以检测水的存在。它开发了一种方法，通过使用纹理分析、支持向量机分类和灰度共生矩阵 (GLCM) 来改善水体与其他土地覆盖的分离。发现 GLCM 平均纹理参数提供了足够的纹理信息，通过提供更均匀的水体测量，将水类与土地类分开。正在研究的案例是墨西哥的格里哈尔瓦河流域，该地区需要更深入的研究，以便地表水覆盖的管理可以与生态系统相结合，尤其是那些受到强烈人类活动影响的生态系统。使用 Sentinel-1 的一系列观测绘制了整个盆地。本研究共制作了 36 张水体图，在研究的每一年（2016、2017 和 2018 年），可接受的分类准确度超过 90%。水的存在也被量化，从而产生了一组包含区域细节的地图，从中可以获得关于全年有水的地区和季节性洪水地区的时间数据。以 10 公里单位单元为空间单位的频率图用于检测趋势，从而识别水分布模式。