Weather, Risk, and Resource Orders on Large Wildland Fires in the Western US

Jude Bayham, Erin J. Belval, Matthew P. Thompson, Christopher Dunn, Crystal S. Stonesifer, and David E. Calkin

This repository contains the scripts to construct the dataset and run the analyses to generate figures and tables reported in the manuscript. Most of the analysis is conducted in R. However, the ordered logit analysis of growth potential is conducted in Stata. The raw data is not contained in the repository, but can be requested from the corresponding author at jbayham@colostate.edu.

Project Structure

This section describes the directory structure of the project. The project is divided into two parts. Part 1 builds the datasets used in the analysis phase. Part 2 contains scripts to run the analysis and generate output (tables and figures). Note that the datasets produced in Part 1 is written to the inputs directory in Part 2.

Part 1: Build

inputs directory contains the raw data that should not be modified or overwritten. Unzip the data into this folder
cache directory stores copies of data during intermediate steps in the preprocessing
code directory contains all scripts to read in and preprocess the data

Part 2: Analysis

inputs directory contains the dataset built in Part 1
cache directory stores copies of data or results during analysis steps
code directory contains all scripts to read in and preprocess the data. Tables and figures are saved to a subdirectory in the reports directory.

The functions directory contains all functions specific to this analysis.
The report directory contains the write up of the project. It
The references directory contains bib files for the project.

Building the Project

The project root directory contains a file called project_init.R that initializes the project (installs/loads packages etc.) and should be run each time you open R to begin working on the project. The makefile, build/code/00_build.R, calls the scripts that import the raw data, process, and construct datasets for analysis.

Note that all file references within project are relative to the root directory of the project.

Scripts

Build

The build/code directory contains a series of numbered scripts. These scripts should be run in order to reproduce the dataset.

The script 00_build.R builds the dataset by either loading a cached copy or running the script to build each component.
The script 01_incident.R imports the 209 incident table from Dunn's DB. This appears to be a subset of fires not in complexes and with more than three reports.
The script 02_resources.R imports the resource tables from Dunn's DB. The resource reporting appears to have changed in 2001 or 2002 so I append the tables together. Pre-2002, strike teams and single resources are aggregated.
The script 03_environment.R imports the data from the topography tables and vegetation tables in Dunn's DB. These are all of the time-invariant environmental controls used in subsequent models.
The script 04_weather.R imports weather data from tables both Access DBs. One of the DBs contains tables with BI (burning index), ERC (energy release component), and SFWP (severe fire weather potential). The DB call UofI contains data on max temperature, minimum humidity, precipitation, and windspeed.
The script 05_ross.R imports the ROSS data that Erin extracted. The data contain daily requests, utfs, and assignments for the following resource types: VLAT, Type 1 Airtankers, Type 2-4 Airtankers, Fixed Wing Aircraft, Type 1 Helicopters, Type 2 Helicopters, Type 3 Helicopters, Dozers, Structure Engines, Wildland Engines, Type 1 Crews, Type 2 Crews, and Type 2IA Crews. Several of these categories are collapsed for the analysis.
The script 06_preparedness.R imports the daily national and GACC preparedness level data compiled by Erin.
The script 07_create_temp_ds.R merges the individual components together and cahces an intermediate dataset called temp_ds.Rdata.
The script 08_create_analysis_ds.R creates datasets for each analysis: model of growth, model of growth potential, and model of resource orders. This script also generates an outliers plot of weather and growth observations that are removed. The script also contains code that outputs the list of IC names to Google Open Refine for text cleanup. The names are remerged back with the dataset for use as fixed effect controls when modeling resource orders.

Analysis

The analysis/code directory contains script to conduct analyses and produce figures and tables. Intermediate outputs are cached to the analysis/cache directory.

The script 01_growth.R reads the cached dataset growth.Rdata runs the ananlysis of observed fire growth as a function of weather and other controls.
The script 02_growth_potential.do reads in a cached dataset 02_gp_ds_stata.dta and runs the ordered logit models and generates figures. Note that this analysis is conducted in Stata.
The script 03_resource_orders.R reads the 03_orders.Rdata dataset and runs the FE models of resource orders.
The script 04_simulation.R illustrates how to use the model results together to understand the link between weather and resource orders.
The script 05_map.R generates the leaflet map of fire ignition points
The script 06_tab_gp_evac_forRR.R tabulates growth potential and evacuation status

Report

The report directory contains the manuscript and supporting documents such as tables and figure images.

The script Supplement.Rmd generates the online supplement for the manuscript.

Functions

This directory contains scripts with functions used to process and analyze the data.

The script init_function.R contains a set of functions used to streamline initializing the project.
The script munge_functions.R contains functions to process the data
The script analysis_functions.R contains functions to convert regression output to a table.

jbayham / weather_risk_resource_orders