WADISC: Annual Impervious Surface Data for Ghana, Togo, Benin, and Nigeria from 2001 – 2020

Rapid urbanization in most African countries is increasing impervious surfaces, including building roofs, glass, concrete, asphalt, and paved roads. However, regionally consistent urban data are lacking to support large-scale research and assessment of impervious surface expansion and its impacts on urban heat islands, hydrology and flood risk, infectious disease risk, cropland and biodiversity loss, habitat fragmentation, and carbon sequestration. The West Africa Dataset of Impervious Surface Change (WADISC) uses all available Landsat data to map urban change in Ghana, Togo, Benin, and Nigeria. The approach combines machine learning algorithm with LandTrendr time series analysis to generate annual maps of urban impervious surface cover from 2001 - 2020. The overall mean absolute error was less than 6% cover and the root mean squared error was less than 10% cover, giving us confidence that the predictions can effectively distinguish areas with high versus low impervious cover. We further classified the impervious cover into developed (pixel value is greater than 20%) and undeveloped (pixel value is less than or equal to 20%) with 93% overall accuracy and approximately similar producer (79%) and user (80%) accuracies in developed areas. WADISC is available in two forms: 1) continuous impervious cover with values ranging from 0% – 100% and 2) Developed area classification with pixel values of 1 and 0, representing the presence and absence of developed area. These data can support consistent city, national and regional assessments and research on urbanization and its impacts.WADISC_Impervious.zip contains GeoTIFF files with continuous impervious cover data from 2001-2020.WADISC_Developed.zip contains GeoTIFF files with classified developed area data from 2001-2020.多数非洲国家正经历快速城市化进程，不透水地表（impervious surface）面积持续扩张，涵盖建筑屋顶、玻璃幕墙、混凝土、沥青及铺装道路等类型。然而，当前缺乏具备区域一致性的城市地表数据，难以支撑大规模研究与评估不透水地表扩张及其对城市热岛效应、水文与洪涝风险、传染病传播风险、耕地与生物多样性丧失、生境破碎化以及碳固存的影响。西非不透水地表变化数据集（West Africa Dataset of Impervious Surface Change, WADISC）依托所有可用的Landsat影像数据，完成加纳、多哥、贝宁及尼日利亚的城市地表变化制图。该方法将机器学习算法与LandTrendr时间序列分析相结合，生成2001年至2020年的城市不透水地表覆盖率年度制图产品。整体平均绝对误差低于6%覆盖率，均方根误差低于10%覆盖率，这表明模型预测可有效区分高、低不透水地表覆盖率区域，结果具备较高可信度。在此基础上，我们将不透水地表覆盖率进一步划分为开发区域（像素值大于20%）与非开发区域（像素值小于等于20%），整体分类精度达93%；在开发区域中，生产者精度（79%）与使用者精度（80%）基本持平。WADISC包含两种数据格式：1）连续型不透水地表覆盖率数据，取值范围为0%至100%；2）开发区域分类数据，像素值为1和0，分别代表开发区域的存在与缺失。该数据集可支撑城市、国家及区域尺度下的城市化及其影响相关的一致性评估与研究工作。WADISC_Impervious.zip 包含2001年至2020年连续型不透水地表覆盖率数据的GeoTIFF文件；WADISC_Developed.zip 则包含同期开发区域分类数据的GeoTIFF文件。