Population-weighted tract-level warm season daily temperature for the northeastern United States 2003-2019 This repository contains modeled tract-level daily temperature estimates for the northeastern United States for the months May-Sep, 2003-2019. Estimates were spatially aggregated from 1 km predictions from the temperature prediction model described in Carrión et al 2019. We use the ~1 km raster Gridded Population of the World (GPW version 4.11) with the exactextractr package in R to construct population-weighted tract summaries of min/mean/max temperature for all the tracts in the study region (Maine to Virginia), for each day (May-Sep), 2003-2019. Daily modeled 2 m air temperature minimum, mean, and maximum estimates are reported as integers in centiKelvin for each census tract (GEOID field; values for dates 2003-2009 were calculated for the 2000 census boundaries with weights from the 2000 GPW raster, values for 2010-2019 used the 2010 census boundaries with weights from the 2010 GPW raster). Divide by 100 to get the temperature estimate in Kelvin. The resulting file is over 45 million rows and a nicely compressed ~200MB parquet file (read into R, the full dataset takes ~1.26GB of RAM). A glimpse at it looks like this: >head(arrow::read_parquet("summarized_daily_temp_preds.parquet"), 1) GEOID date min_temp_daily_cK mean_temp_daily_cK max_temp_daily_cK 1: 09001010101 2003-05-01 28235 28768 29377