Skip to contents

Large Hurricane Analysis - Outage

top_15_outage.Rmd

Program Background

  • Script name: Large Hurricane Analysis - Outage

  • Project name: ADRD and Hurricane Exposure

  • Program description:

    Outage measure calculation for 15 previously selected large hurricanes. Only hurricanes in 2014 and after had available outage data.

    Number of customers without electricity is collected by EAGLE-I. County-level electric customer estimates are generating by adding the census business survey and ACS5 population estimates for a county and normalizing to state-wide energy customer estimates from the Energy Information Administration.

  • Data:

  • Restrictions:

  • Outline of program/analysis:

    • Specify all storms

    • Download and process hurdat2 tracks

    • Extract outage data for dates specified by the HURDAT2 plus 2 days after

    • Calculate outage estimates for the maximum percentage of customers without electricity and the average duration of outages

    • Crosswalk county estimates to ZCTA using 2010 ZCTA to County Relationship file

  • Author: Ryan Zomorrodi

  • Revisions

  • Input files

    • All files are downloaded using R functions
    • Cache is stored in raw-data\eaglei

  • Output files

    • tract/{storm id}/outage.csv

    • zcta/{storm id}/outage.csv

      Column Description
      geoid 2010 geographic identifier
      max_pct_out Maximum percentage of customers out of electricity at any point
      avg_customer_hrs_out Average number of hours each customer was out of electricity for

Start Time

start_time <- Sys.time()
cat(start_time, "\n")

The following hurricanes were selected.

hur_info <- readRDS( "raw-data/RF1_Hurricanes_Final.rds")
# get fips codes for each state
fips_codes <- distinct(tigris::fips_codes, state, state_code)
hur_info$fips <- pmap(hur_info, function(states, ...) {
  fips_codes$state_code[match(states, fips_codes$state)]
})

Get hurdat2 info for each storm 

exposhur:::download_newest_hurdat2(path = "raw-data/hurdat2")

hur_info$hurdat2 <- pmap(hur_info, function(hurricane, ...){
  hurdat  <- get_hurdat2(
    storm_id = hurricane,
    path = "raw-data/hurdat2"
  )
  
  list(
    start_date = min(hurdat$datetime),
    end_date = max(hurdat$datetime),
    hurdat = hurdat
  )
})
hur_info <- unnest_wider(hur_info, hurdat2)

Get outage data 

hur_info$eaglei <- pmap(hur_info, function(start_date, end_date, ...) {
  if (year(start_date) < 2014) {
    return(list(NULL))
  }

  get_eaglei(
    interval = interval(start_date, end_date + days(2)),
    path = "raw-data/eaglei"
  )
})

Calculate outage 

hur_info$outage <- pmap(hur_info, function(eaglei, fips, start_date, ...) {
  if (year(start_date) < 2014) {
    return(list(NULL))
  }

  calc_outage(eaglei, fips)
})

pwalk(hur_info, function(outage, hurricane, start_date, ...) {
  if (year(start_date) >= 2014) {
    write_csv(outage, str_c("tract/", hurricane, "/outage.csv"))
  }
})

Crosswalk to zctas 

zcta_to_county <- read_csv("https://www2.census.gov/geo/docs/maps-data/data/rel/zcta_county_rel_10.txt") %>%
  mutate(geoid_z = str_c(STATE, ZCTA5)) %>%
  mutate(geoid_c = str_c(STATE, COUNTY))

outages <- Sys.glob("tract/*/outage.csv") %>%
  map(function(outage_path) {
    outage <- read_csv(outage_path, col_types = cols(geoid = col_character()))
    
    outage %>%
      left_join(zcta_to_county, join_by(geoid == geoid_c)) %>%
      group_by(geoid_z) %>%
      summarise(
        max_pct_out_xwalk = sum(ZPOPPCT * max_pct_out) / 100,
        avg_customer_hrs_out_xwalk = sum(ZPOPPCT * avg_customer_hrs_out) / 100
      )
    
    outage %>%
      write_csv(str_replace(outage_path, "tract", "zcta"))
  })

End Time

end_time <- Sys.time()
cat("End time: ", end_time, "\n")
cat("Duration: ", start_time - end_time, "\n")

Visualizations

Estimated Maximum Percentage of Customers without Electricity
Storm Mean SD Min 25% Quantile Median 75% Quantile Max Distribution
FLORENCE-2018 12.702 20.705 0 2.061 5.185 11.571 100
HARVEY-2017 5.829 14.774 0 0.000 0.317 4.069 100
IRMA-2017 46.973 35.085 0 13.798 38.251 79.503 100
MATTHEW-2016 14.535 22.314 0 0.711 3.642 19.851 100
MICHAEL-2018 20.905 29.487 0 1.390 6.355 28.729 100

Estimated Average Customer Hours without Electricity
Storm Mean SD Min 25% Quantile Median 75% Quantile Max Distribution
FLORENCE-2018 5.238 16.277 0 0.148 0.370 1.122 123.445
HARVEY-2017 1.088 5.680 0 0.000 0.022 0.227 70.360
IRMA-2017 19.436 20.848 0 2.847 10.491 30.496 97.497
MATTHEW-2016 4.386 10.018 0 0.031 0.168 2.005 61.545
MICHAEL-2018 8.329 20.634 0 0.056 0.386 5.300 138.713