Building-level functional maps of 109 Chinese cities

<b>Background</b>As the world’s most rapidly urbanizing country, China now faces mounting challenges from growing inequalities in the built environment, including disparities in access to essential infrastructure and diverse functional facilities. Yet these urban inequalities have remained unclear due to coarse observation scales and limited analytical scopes. In this study, we present the <i>first</i> building-level functional map of China, covering 110 million individual buildings across 109 cities using 69 terabytes of 1-meter resolution multi-modal satellite imagery. The national-scale map is validated by government reports and 5,280,695 observation points, showing strong agreement with external benchmarks. This enables the <i>first</i> nationwide, multi-dimensional assessment of inequality in the built environment across city tiers, geographical regions, and intra-city zones.<b>About data</b>Based on the Paraformer framework that we proposed previously, we produced <b>the first nationwide building-level functional map of urban China</b>, processing over 69 TB of satellite data, including <i>1-meter Google Earth optical imagery (</i><i>https://earth.google.com</i><i>)</i>, <i>10-meter nighttime lights (SGDSAT-1) (</i><i>https://sdg.casearth.cn/en</i><i>)</i>, and <i>building height data (CNBH-10m) (</i><i>https://zenodo.org/records/7827315</i><i>)</i>. Labels were derived from: (1) Building footprint data, including the <i>CN-OpenData (</i><i>https://doi.org/10.11888/Geogra.tpdc.271702</i><i>)</i> and the <i>East Asia Building Dataset (</i><i>https://zenodo.org/records/8174931</i><i>)</i>; and (2) Land use and AOI data used for constructing urban functional annotation are retrieved from <i>OpenStreetMap (</i><i>https://www.openstreetmap.org</i><i>) and EULUC-China dataset (</i><i>https://doi.org/10.1016/j.scib.2019.12.007</i><i>). </i>The first 1-meter resolution national-scale land-cover map used to conduct the accessibility analysis is available in our previous study: <i>SinoLC-1 (</i>https://doi.org/10.5281/zenodo.7707461<i>). </i>The housing inequality and infrastructure allocation analysis was conducted based on the 100-meter gridded population dataset from China's seventh census (<i>https://figshare.com/s/d9dd5f9bb1a7f4fd3734?file=43847643</i>).This version of the data includes (1) Building-level functional maps of 109 Chinese cities, and (2) In-situ validation point sets. The building-level functional maps of 109 Chinese cities are organized in the ESRI Shapefile format, which includes five components: “.cpg”, “.dbf”, “.shx”, “.shp”, and “.prj” files. These components are stored in “.zip” files. Each city is named “G_P_C.zip,” where “G” explains the geographical region (south, central, east, north, northeast, northwest, and southwest of China) information, “P” explains the provincial administrative region information, and “C” explains the city name. For example, the building functional map for Wuhan City, Hubei Province is named “Central_Hubei_Wuhan.zip”.Furthermore, each shapefile of a city contains the building functional types from 1 to 8, where the corresponding relationship between the values and the building functions is shown below:Residential buildingCommercial buildingIndustrial buildingHealthcare buildingSport and art buildingEducational buildingPublic service buildingAdministrative building<b>About validation</b>Given the importance of accurate mapping for downstream analysis, we conducted a comprehensive evaluation using government reports and in situ validation data outlined in the Data Section. This evaluation comprised two parts. First, a statistical-level evaluation was performed for each city based on official reports from the <i>China Urban-Rural Construction Statistical Yearbook (</i><i>https://www.mohurd.gov.cn/gongkai/fdzdgknr/</i><i>sjfb/tjxx/jstjnj/index.html</i><i>)</i> and <i>China Statistical </i><i>Yearbook (</i><i>https://www.stats.gov.</i><i>cn/sj/ndsj/2023/indexch.htm</i><i>)</i>. Second, a building-level geospatial evaluation was conducted by using 5.28 million field-observed points from Amap Inc. (provided in this data version of "Validation_in-situ_points.zip"), and a confusion matrix was calculated to compare the in situ points with the mapped buildings at the same location. The "Validation_in-situ_points.zip" includes the original point sets of each city, named as the city name (e.g., Wuhan.shp and corresponding “.cpg”, “.dbf”, “.shx”, and “.prj” files).