Code and Data for "Spatially concentrated heat emissions from data centers lead to localized urban warming and heat exposure"

Code and Data for Assessing the Heat-Island Effects of Data Centers Overview This Zenodo archive contains the analysis code and supporting data used to quantify the effects of data-center anthropogenic heat emissions on the near-surface environment and population exposure in California. 1 The archive includes Jupyter notebooks, energy-system data, WRF-ready anthropogenic-heat inputs, processed WRF outputs, geographic boundary files, and gridded population data. Directory structure .├── Code_paper/│ ├── Data_Center_Heat_Emission_Generate/│ ├── Model_Evaluation/│ └── WRF_Post_analysis/└── Data_paper/ ├── EIA_Data/ ├── Generated_Heat_Input_Data/ ├── Other/ ├── Population_CA/ └── Processed_WRF_Output/ Code All code is provided as Jupyter notebooks. Code_paper/Data_Center_Heat_Emission_Generate/ 1. Attribute_DC_to_BA.ipynb This notebook: associates data centers with balancing-authority regions using nearby power plants; processes EIA-860 plant information, EIA-923 generation and fuel data, and hourly EIA balancing-authority data; allocates data-center electricity consumption to generation sources; estimates direct Scope 1 and electricity-related Scope 2 heat emissions; and maps the emissions to the WRF domain and exports hourly WRF-SLUCM anthropogenic-heat input files. 2. Generate_Future_DC.ipynb This notebook: creates future data-center scenarios using Monte Carlo placement and/or direct scaling; and supports the higher-demand scenarios used in the study, including the 3× and 5× cases. Code_paper/Model_Evaluation/ This folder contains the following notebooks: Evaluation_AirTemp_AQS_Summer.ipynb Evaluation_RelativeHumidity_AQS_Summer.ipynb Evaluation_WindSpeed_AQS_Summer.ipynb These notebooks compare WRF results with station observations for 2-m air temperature, relative humidity, and 10-m wind speed. They perform observation filtering, spatial and temporal pairing, ensemble averaging, and calculation of performance statistics, including mean bias, mean absolute error, root-mean-square error, normalized errors, and R². The notebooks reference raw observation files and full WRF outputs on the authors’ computing system. These large intermediate files are not part of this archive. Code_paper/WRF_Post_analysis/ 1. Test_AHoption_and_StartTime.ipynb This notebook: tests anthropogenic-heat configurations and simulation start times; constructs paired differences between simulations with anthropogenic heat, AH, and baseline simulations without anthropogenic heat, noAH; processes variables including 2-m air temperature, T2; skin temperature, TSK; relative humidity, RH; 10-m wind speed, WS; and planetary boundary-layer height, PBLH; and creates ensemble-mean, member-level time-mean, and hourly NetCDF products. 2. Impact_of_AH_DC.ipynb This notebook: analyzes the spatial and temporal meteorological response to data-center heat emissions; compares present-day and scaled future scenarios; applies ensemble-based significance testing; and produces the principal analysis figures. 3. Analyze the Health Effect_Exposure.ipynb This notebook: aggregates the 100-m California population raster to the WRF grid; combines population with statistically filtered daily temperature changes; and calculates exposed population and population-day-weighted temperature changes for the modeled scenarios. Data Data_paper/EIA_Data/ This folder contains energy-system data used to attribute data-center electricity demand and associated heat emissions: eia8602023/: U.S. Energy Information Administration Form EIA-860 files for 2023, including utility, plant, generator, ownership, storage, wind, solar, multifuel, and environmental-equipment tables; f923_2023/: Form EIA-923 generation, fuel-consumption, source/disposition, and environmental-information files for 2023; eia_ba_exchange_hourly_2023_07.csv: hourly balancing-authority interchange data for July 2023; eia_generation_type_hourly_2023_07.csv: hourly generation by energy source for July 2023; and eia_rto_hourly_2023_07.csv: hourly balancing-authority/RTO demand and generation data for July 2023. These files retain the structure and naming of their source data products. Users should consult the included EIA documentation and the official EIA metadata when interpreting individual fields. Data_paper/Generated_Heat_Input_Data/ This folder contains WRF-SLUCM anthropogenic-heat inputs for six scenarios: Directory Interpretation DC_AH_total_1_scaled_July_v1_scaleup/ Present-day/reference data-center heat-emission case DC_AH_total_3_scaled_July_v1_scaleup/ 3× scaled data-center case DC_AH_total_5_scaled_July_v1_scaleup/ 5× scaled data-center case Each directory contains 745 sequential files, named 0000 through 0744, representing the hourly sequence used for the July simulation period. Each file is a plain-text, comma-separated 2-D field. Its first line gives the grid dimensions, 291 × 252, followed by the gridded anthropogenic-heat values aligned with WRF domain d02. The scenario labels reflect the naming used during the experiments. Their physical implementation is defined in the heat-emission-generation notebooks and the associated WRF-SLUCM configuration. Data_paper/Processed_WRF_Output/ This folder contains processed NetCDF products derived from paired WRF simulations. The common naming conventions are: AH_minus_noAH: anthropogenic-heat simulation minus its paired baseline simulation; diff_ensmean_*: ensemble-mean hourly difference, generally with dimensions Time × south_north × west_east; diff_time_mean_samples_*: full-period mean difference retained separately for each ensemble member/start time; diff_time_samples_*: hourly paired differences retained for each ensemble member/start time; BASE_* or base_hourly_samples_*: hourly baseline, noAH, fields; 3x_ and 5x_: future data-center scaling scenarios; ALH or ALH0.3: alternative anthropogenic-heat release/configuration experiments; and d02: the inner WRF domain. Variables represented in this folder include: Label Variable T2 2-m air temperature TSK Land-surface skin temperature RH Derived 2-m relative humidity WS 10-m wind speed, derived from U10 and V10 where needed PBLH Planetary boundary-layer height The NetCDF variable attributes provide additional descriptions of the stored quantity, experiment difference, and domain. Temperature differences in kelvin and degrees Celsius have the same numerical magnitude. Data_paper/Population_CA/ This folder contains: CAPOP_2020_100m_TOTAL.tif: 2020 California population raster at approximately 100-m resolution, used for population-exposure calculations. The exposure notebook reprojects and aligns this raster, and aggregates the population to the WRF grid. Data_paper/Other/ This folder contains: geo_em.d02_NLCD_FRC.nc: WRF geographical input/template for domain d02, including model coordinates and land-use/urban-fraction information; ca_counties/: California county boundary shapefile and its companion files; and cb_2018_us_state_500k/: generalized 2018 U.S. state boundary shapefile and its companion files. For a shapefile to work correctly, keep its .shp, .shx, .dbf, and .prj files together.