Supporting data and code for Davenport, Burke, and Diffenbaugh (2021) 'Contribution of historical precipitation change to US flood damages'.

If you find meaningful errors in the code or have questions or suggestions, please contact Frances Davenport at [email protected].

Davenport, F. V., M. Burke, & N. S. Diffenbaugh (2021). Contribution of historical precipitation change to US flood damages. Proceedings of the National Academy of Sciences, 118(4). https://doi.org/10.1073/pnas.2017524118

• data: input data used for analysis (not all raw data is included due to size; see details below)
• analysis_scripts: code to read and analyze data
• processed_data: output from analysis_scripts (used as input to figure scripts)
• figure_scripts: code to create figures

Datasets not included in the repository are available at the following locations:

• PRISM monthly and daily precipitation: available from the PRISM Climate Group, Oregon State University
• CMIP5: information about CMIP5 and how to access the data can be found here
• CLIMDEX HadEX3 historical data: the CLIMDEX monthly maximum 5-day precipitation ("Rx5day") product is available from https://www.climdex.org/access/
• CLIMDEX CMIP5 data: the CLIMDEX CMIP5 datasets are available through Environment Canada
• Spatial Hazard Events and Losses Database for the U.S. (SHELDUS): can be accessed with a subscription through the SHELDUS site). The aggregated (state-month) flood damages used in the analysis are included in the processed_data folder.
• NFIP Redacted Claims Dataset: publicly available from FEMA

Input data included in this repository:

• cb_2017_us_state_20m/: shapefile of state boundaries from the Census cartographic boundary files
• income_state_current.csv: annual state-level income, formatted version of data from the Bureau of Economic Analysis
• fixed_assets_current_dollars.csv: national net stock of reproducible fixed assets from the Bureau of Economic Analysis
• combined_home_value_data.csv: combined data on state-level median home values from the US Census Housing Tables for 1990 and 2000, and from the American Community Survey for 2010
• combined_housing_units_data.csv: combined data on state-level housing units from the US Census for 1990, 2000, and 2010
• WBD_07_HU2_Shape/: shapefile of the Upper Mississippi River basin from the USGS Watershed Boundary Dataset
• WBD_10_HU2_Shape/: shapefile of Missouri River basin from the USGS Watershed Boundary Dataset
• billion_dollar_disasters_Flooding.csv: rainfall and flooding-related events within the Billion Dollar Weather and Climate Disasters

Some scripts take a while to run - approximate times and memory usage are indicated if run time is >1min.

• file_paths.R: file paths for input data and processed data (must be edited for input data not included with repository)
• parameters.R: defines parameter variables used throughout analysis
• func.R: defines functions used throughout analysis

1 - Reading PRISM and CMIP precipitation data:

• calc-monthly-state-PRISM.R: calculate monthly timeseries of average precipitation in each state (CALCULATION TIME: ~10min; MEMORY: 16GB)
• calc-daily-state-PRISM.R: calculate daily timeseries of average precipitation in each state (CALCULATION TIME: ~1hr 15min; MEMORY: 16GB)
• calc-gridded-PRISM-timeseries.R: read regridded (2.5 deg.) monthly PRISM precipitation files and save monthly timeseries (CALCULATION TIME: <5min)
• read-cmip-2pt5-deg.R: read 2.5 deg. CMIP5 data and save monthly timeseries (CALCULATION TIME: <5min per model)
• read-climdex.R: read 2.5 deg. CLIMDEX data and save monthly timeseries (CALCULATION TIME: ~5min)
• calc-watersheds-PRISM.R: calculate monthly timeseries of average precipitation in Missouri River and Upper Mississippi basins (CALCULATION TIME: ~10min)

2 - Combining panel data and running regression models:

• combine-state-panel-data.R: combine monthly flood damages and precipitation data (CALCULATION TIME: <<1min)
• panel-regression-models.R: fit regression models (CALCULATION TIME: <<1min)
• model-bootstrapping.R: bootstrap regression models (CALCULATION TIME: <2min per model)

3 - Calculating precipitation trends and counterfactual damages:

• calc-detrended-precip.R: calculate 10,000 detrended precipitation time series (CALCULATION TIME: each trend time period takes around 15hrs; MEMORY: >12GB)
• calc-random-precip-trends.R: calculate random precipitation trends using moving block bootstrap to determine significance of observed trends (CALCULATION TIME: <5min)
• calc-counterfactual-damages.R: calculate counterfactual damages for each detrended timeseries and each regression model bootstrap replicate (CALCULATION TIME: each model-time period combination takes about 30min when using 10 parallel tasks; MEMORY: 8GB per task)

• monthly_state_precip.Rds: monthly precipitation time series (state-level)
• precip_data_std.Rds: standardized monthly precipitation time series (state-level)
• daily_state_precip.Rds: daily precipitation time series (state-level)
• precip_5daymax.Rds: standardized monthly maximum 5-day precipitation time series (state-level)
• state_panel_data.csv: combined state-month precipitation and flood damages (csv format)
• state_panel_data.Rds: combined state-month precipitation and flood damages (Rds format)
• watershed_precip_data.Rds: monthly precipitation time series for Upper Mississippi Basin and Missouri River Basin
• prism_2pt5deg_precip.Rds: monthly PRISM precipitation time series on 2.5 degree grid
• climdex_monthly_Rx5day.Rds: monthly maximum 5-day precipitation ("Rx5day") time series on 2.5 degree grid
• 2pt5deg_cmip_monthly/: monthly precipitation time series from cmip models on 2.5 degree grid
• 2pt5deg_cmip_rx5day/: monthly max 5-day precipitation ("Rx5day") time series from cmip models on 2.5 degree grid
• regression_models/: regression models (in .Rds format)
• bootstrapped_models/: bootstrapped regression model coefficients
• detrended_precip/: detrended state-month precipitation time series (not included ~30GB)
• counterfactual_damages/: estimated counterfactual damages (not included ~1.5GB)
• quantile_trends_1928-2017.Rds: bootstrapped 50th, 75th, and 95th percentile precipitation trends for each state
• random_trends.Rds: random precipitation trends in each state from moving block bootstrap

files in italics not included in repository due to size

• theme_func.R: defines plot themes and plotting functions used in figure scripts
• fig_1.R: creates Figure 1
• fig_2.R: creates Figure 2
• fig_3.R: creates Figure 3 and Figure S6
• fig_4_5.R: creates Figure 4, Figure 5, and Figure S7
• fig_S1.R: creates Fig. S1
• fig_S2.R: creates Fig. S2
• fig_S3.R: creates Fig. S3
• fig_S4.R: creates Fig. S4
• fig_S5.R: creates Fig. S5
• fig_S8.R: creates Fig. S8
• fig_S9.R: creates Fig. S9
• fig_S10.R: creates Fig. S10
• fig_S11.R: creates Fig. S11

• tidyverse, lfe, fixest, quantreg, raster, ncdf4, sf, velox, RColorBrewer, scales, stringr, ggpubr, stargazer

code was written in R 3.5.1