High-resolution Glacial Lake Inventory for the Southeastern Tibetan Plateau (2023) derived from Sentinel-2 imagery for "Mapping glacial lakes with Glacier-related Compact Efficient Neural Network (GCE-Net): Perspectives on Balancing Model Size and Performance"
收藏资源简介：
This dataset presents a comprehensive high-resolution inventory of glacial lakes in the Southeastern Tibetan Plateau (SETP) for the year 2023. The inventory identifies a total of 6,455 glacial lakes, covering a combined surface area of 400.18 km². The dataset was generated using the Glacier-related Compact Efficient Neural Network (GCE-Net), a lightweight deep learning framework optimized for accurate glacial lake mapping in complex high-mountain environments. The results have undergone rigorous quality control, including NDWI filtering and manual visual inspection, ensuring high reliability for hydrological and hazard assessment applications. The inventory was derived from multi-source remote sensing data: Satellite Imagery: Sentinel-2 MSI imagery (10 m resolution) acquired in 2023 with <10% cloud cover. Data processing was conducted on the Google Earth Engine (GEE) platform.Topographic Data: Elevation and slope attributes were extracted from the Copernicus DEM (GLO-30) to provide precise terrain information.Model: The boundaries were automatically extracted using the GCE-Net model (ResNet-34 backbone with Ghost Modules and Efficient Channel Attention), followed by morphological post-processing. FILE STRUCTURE AND ATTRIBUTES The dataset is provided in ESRI Shapefile format (.shp). The attribute table includes the following fields: GL_ID: Unique identifier for each glacial lake instance. Longitude: Geodetic longitude of the lake centroid (Decimal Degrees). Latitude: Geodetic latitude of the lake centroid (Decimal Degrees). Area_sqkm: Surface area of the glacial lake in square kilometers. Perimeter: The length of the glacial lake boundary (in meters). Elev_Mean: Mean elevation of the lake surface (in meters), derived from Copernicus DEM. Slope_Mean: Mean slope of the lake area (in degrees), derived from Copernicus DEM. COORDINATE REFERENCE SYSTEM (CRS) The dataset uses the standard WGS 84 geographic coordinate system. Name: WGS 84 (Geographic 2D) EPSG Code: 4326 Datum: World Geodetic System 1984 ensemble Axis Units: degree (North Latitude / East Longitude)
本数据集提供了2023年青藏高原东南部（Southeastern Tibetan Plateau, SETP）高精度冰川湖泊综合清查数据集。本次清查共识别出6455处冰川湖泊，总表面积达400.18平方千米。 本数据集采用冰川相关紧凑型高效神经网络（Glacier-related Compact Efficient Neural Network, GCE-Net）生成，该轻量化深度学习框架针对复杂高海拔山地环境下的冰川湖泊精准制图进行了优化。研究成果经过了严格的质量控制流程，包括归一化差分水体指数（Normalized Difference Water Index, NDWI）筛选与人工目视解译校验，可为水文与灾害评估相关应用提供高可靠性的数据支撑。 本次清查基于多源遥感数据构建： 卫星影像：采用2023年获取的Sentinel-2 MSI影像（空间分辨率10米），云量覆盖率低于10%。数据处理基于谷歌地球引擎（Google Earth Engine, GEE）平台完成。 地形数据：从哥白尼数字高程模型（Copernicus DEM, GLO-30）中提取高程与坡度属性，以提供精准的地形信息。 模型方法：首先通过GCE-Net模型（以ResNet-34为骨干网络，集成Ghost模块（Ghost Modules）与高效通道注意力（Efficient Channel Attention）机制）自动提取湖泊边界，随后进行形态学后处理优化。 ## 文件结构与属性 本数据集以ESRI形状文件（ESRI Shapefile, .shp）格式提供，属性表包含以下字段： - GL_ID：每个冰川湖泊实例的唯一标识符 - Longitude：湖泊质心的大地经度（单位：十进制度） - Latitude：湖泊质心的大地纬度（单位：十进制度） - Area_sqkm：冰川湖泊的表面积（单位：平方千米） - Perimeter：冰川湖泊边界的周长（单位：米） - Elev_Mean：湖泊表面的平均高程（单位：米，源自哥白尼DEM数据） - Slope_Mean：湖泊区域的平均坡度（单位：度，源自哥白尼DEM数据） ## 坐标参考系统（CRS） 本数据集采用标准1984年世界大地测量系统（World Geodetic System 1984, WGS 84）地理坐标系，具体参数如下： - 名称：WGS 84（二维地理坐标系） - EPSG代码：4326 - 基准面：World Geodetic System 1984 ensemble - 轴单位：度（北纬度/东经度）
