Heat Stress Metrics for US Census Tracts 1998-2020 [DATASET]

Abstract: Extreme heat exposure is a growing public health threat. Heat-health research has commonly used dry-bulb temperature to characterize heat exposure, partly due to limited availability of spatially explicit, public-health-aligned datasets that integrate multiple meteorological factors to quantify heat stress. We address this gap by providing hourly Heat Index (HI), Wet-Bulb Globe Temperature (WBGT), and Universal Thermal Climate Index (UTCI) for U.S. Census tract boundaries across the contiguous United States from 1998–2020. Heat-stress fields were generated by integrating PRISM, ERA5-Land, and National Solar Radiation Database (NSRDB) products, with near-surface temperature and moisture fields reconstructed and ancillary variables interpolated to a harmonized 800-m grid. Heat-stress indices were computed using validated physical models and aggregated to census tracts using area- and population-weighted methods. Validation against station networks shows stable performance for sample year 2010 May-September, with air-temperature RMSE of 1.70 °C, Heat Index RMSE of 3.20 °C, WBGT RMSE of 2.90 °C, and UTCI RMSE of 3.26 °C. These tract-level hourly heat-stress datasets enable direct linkage with public health data.Data Records and Use: We developed an open-access dataset of area-weighted and population-weighted, hourly heat exposure estimates aggregated to U.S. Census Tract boundaries across the contiguous United States. The datasets are stored in parquet format and each file represents two half UTC days and is stored in the following format: heatstress_tract_area_and_popweighted_[DATE]_[DATE + 1]_popy[Population YEAR]_v[Vintage Year].parquet, with total storage of all files around 515GB. Specifically, each Parquet file corresponds to a single PRISM day and contains tract-level hourly heat stress estimates spanning a 24-hour UTC exposure window from 12:00 UTC on the previous calendar day to 11:00 UTC on the labeled day. This convention maintains temporal consistency between daily PRISM products and the sub-daily meteorological inputs used in heat stress modeling while avoiding ambiguities associated with calendar-day aggregation. A complete description of the census-tract parquet file structure, including variable names, units, and definitions, is provided below:GEOID: Census tract identifier (NHGIS-compatible)year: Year of observation (UTC-based)month: Month of observation (UTC-based)day: Day of observation (UTC-based)time (hour): Hour of day (0–23, UTC-based)temp_C_used_area (°C): Area-weighted tract mean air temperaturetemp_C_used_pop (°C): Population-weighted tract mean air temperaturerh_pct_used_area (%): Area-weighted tract mean relative humidityrh_pct_used_pop (%): Population-weighted tract mean relative humidityHI_C_area (°C): Area-weighted Heat IndexHI_C_pop (°C): Population-weighted Heat IndexWBGT_C_area (°C): Area-weighted Wet Bulb Globe TemperatureWBGT_C_pop (°C): Population-weighted Wet Bulb Globe TemperatureUTCI_C_area (°C): Area-weighted Universal Thermal Climate IndexUTCI_C_pop (°C): Population-weighted Universal Thermal Climate IndexNotes: Daily Parquet files are stored in year ZIP folders on figshare. Parquet day files are missing for 2002-01-15, 2002-01-16, 2007-10-10, 2007-10-11, 2007-11-03, and 2007-11-04 due to missing input data.

摘要：极端高温暴露是日益严峻的公共卫生威胁。热健康研究通常采用干球温度（dry-bulb temperature）表征高温暴露，这在一定程度上源于缺乏空间显式、契合公共卫生需求的多气象因子融合数据集以量化热应激。本研究填补了这一空白，提供了1998年至2020年美国本土范围内，覆盖美国人口普查分区（U.S. Census Tract）的逐时热指数（Heat Index, HI）、湿球黑球温度（Wet-Bulb Globe Temperature, WBGT）与通用热气候指数（Universal Thermal Climate Index, UTCI）数据集。热应激场通过融合PRISM、ERA5-Land及国家太阳辐射数据库（National Solar Radiation Database, NSRDB）产品生成，对近地表温湿度场进行重构，并将辅助变量插值至统一的800米网格。热应激指数通过经过验证的物理模型计算得到，并采用面积加权与人口加权方法聚合至普查分区尺度。针对2010年5月至9月的气象站网验证结果显示，模型性能稳定，气温均方根误差（Root Mean Square Error, RMSE）为1.70 °C，热指数均方根误差为3.20 °C，湿球黑球温度均方根误差为2.90 °C，通用热气候指数均方根误差为3.26 °C。该普查分区级逐时热应激数据集可直接与公共卫生数据建立关联。
数据记录与使用说明：本研究构建了一套开放获取的数据集，包含美国本土范围内聚合至美国人口普查分区（U.S. Census Tract）尺度的面积加权与人口加权逐时热暴露估算结果。数据集以Parquet格式存储，每个文件对应两个半天的协调世界时（UTC）时段，命名格式为：heatstress_tract_area_and_popweighted_[DATE]_[DATE + 1]_popy[Population YEAR]_v[Vintage Year].parquet，所有文件总存储量约为515GB。具体而言，每个Parquet文件对应单个PRISM日，包含的普查分区级逐时热应激估算数据覆盖从前一自然日12:00 UTC至标注日11:00 UTC的24小时UTC暴露窗口。该命名约定可保持每日PRISM产品与热应激建模所用次小时气象输入数据的时间一致性，同时避免日历日聚合带来的歧义。下文将完整说明普查分区Parquet文件的结构，包括变量名称、单位与定义：
GEOID：人口普查分区标识符（兼容美国历史地理信息系统（National Historical Geographic Information System, NHGIS））
year：观测年份（基于协调世界时）
month：观测月份（基于协调世界时）
day：观测日期（基于协调世界时）
time (hour)：当日时刻（0–23，基于协调世界时）
temp_C_used_area (°C)：面积加权的普查分区平均气温
temp_C_used_pop (°C)：人口加权的普查分区平均气温
rh_pct_used_area (%)：面积加权的普查分区平均相对湿度
rh_pct_used_pop (%)：人口加权的普查分区平均相对湿度
HI_C_area (°C)：面积加权的热指数（Heat Index, HI）
HI_C_pop (°C)：人口加权的热指数
WBGT_C_area (°C)：面积加权的湿球黑球温度（Wet-Bulb Globe Temperature, WBGT）
WBGT_C_pop (°C)：人口加权的湿球黑球温度
UTCI_C_area (°C)：面积加权的通用热气候指数（Universal Thermal Climate Index, UTCI）
UTCI_C_pop (°C)：人口加权的通用热气候指数
备注：每日Parquet文件以年份为单位打包为ZIP文件夹，存储于figshare平台。由于输入数据缺失，2002年1月15日、2002年1月16日、2007年10月10日、2007年10月11日、2007年11月3日及2007年11月4日的Parquet数据文件缺失。