China Urban Lake

A high-resolution circa-2020 map of urban lakes (>0.001 km2) in China. The 10-m-resoultion Sentinel-2 imagery as well as a simple but robust water extraction method were used to generate waterbodies in China on Google Earth Engine. After initially filtering out urban water bodies based on the spatial relationships with urban area boundary, we combined high-resolution historical imagery to manually remove non-lake waters such as rivers and paddy fields, and to edit and supplement the missing urban lakes separately. The accuracy of our dataset was evaluated in terms of both area and count of lakes by comparing with manually vectorizing results in randomly sampled urban units. The results showed that our dataset is highly accurate and trustworthy, with the averaged accuracy of 81.85% in area and 93.35% in count.

本数据集为中国境内2020年前后、面积大于0.001平方千米的城市湖泊高分辨率分布图。本研究依托10米分辨率的哨兵二号（Sentinel-2）影像，结合简洁鲁棒的水体提取方法，在谷歌地球引擎（Google Earth Engine）平台上生成中国全域水体范围。随后首先基于与城市建成区边界的空间关系初步筛选出城市水体，再结合高分辨率历史影像开展人工甄别：剔除河流、水田等非湖泊水体，并逐一补全、修正遗漏的城市湖泊。本数据集的精度以随机抽样城市单元的手动矢量化结果为参照，从湖泊面积与数量两个维度进行评估。评估结果显示，本数据集精度优异、可信度高，湖泊面积识别平均精度达81.85%，数量识别平均精度达93.35%。