OWL-I USA: Kilometer-Scale Power Outage Dataset for the Continental United States 2012-2024
Authors/Creators
Description
Overview
OWL-I (the Outage Framework for Luminosity Interruptions) is the first 1-km resolution, standardized, calibrated, and reproducible global outage detection method derived from satellite nighttime light observations, described in an upcoming publication Rice et al. (in review). This dataset, OWL-I USA, covers the continental United States where robust validation data is available, although further global validaton and data releases are planned.
The use of nighttime light data is becoming increasingly common in the literature for the detection of power outages, due to its global coverage and high spatial resolution. However the commonly-used heuristic methods in the literature suffer two drawbacks: (1) they do not correct for confounding factors which introduce substantial non-outage variability into the nighttime brightness signal, and (2) they are not calibrated against ground-truth outage data. OWL-I addresses both of these problems, removing the influence of confounding factors and calibrating against the EAGLE-I outage database (Brelsford et al. 2024), resulting in a substantial improvement in skill. Our study finds that the standard approach in the literature achieves a Pearson correlation of 0.358 with true outages, while OWL-I doubles that skill with a correlation of 0.720.
OWL-I USA is provided at 30-arcsecond (approximately 1 km) resolution for the period January 19th 2012 (the first date of VIIRS nighttime light availability) through December 31st 2024, at nightly frequency. For every pixel, on every night not obfuscated by clouds, the OWL-I model provides an estimated power outage level and uncertainty.
Data format
Data is provided in 10x10-degree tiles convering the continental United States, following the tilegrid definition of the NASA Black Marble Product, depicted in Figure 2 of the Black Marble User Guide v2.0 (Wang et al. 2024). The OWL-I product is provided at half the spatial resolution of the native Black Marble data, to align it with the resolution of LandScan Global (Lebakula et al. 2025). The 13-year record for each tile is compressed into a .tar.gz archive to comply with Zenodo file number limits. After identifying the tile(s) of interest, you can extract the relevant records in the command line with the following bash command (many applications also exist for extracting .tar.gz files):
for file in owli_h*v*.tar.gz; do tar -xzvf "$file"; done
This will extract a set of 13 NetCDF files, one for each year, of the form owli_YYYY_h*v*.nc where YYYY is the four-digit year. Below is an example abbreviated file structure, from ncdump:
netcdf owli_2015_h5v5 {
dimensions:
lat = 1200 ;
lon = 1200 ;
date = 365 ;
variables:
double landscan(lat, lon) ;
double pixel_customers(lat, lon) ;
double lon(lon) ;
double lat(lat) ;
int64 date(date) ;
float outage_fraction_ensmean(date, lat, lon) ;
float outage_fraction_ensstd(date, lat, lon) ;
}
'outage_fraction_ensmean' provides the OWL-I ensemble mean detection of per-pixel fraction of customers experiencing outage (in the range [0,1]), while 'outage_fraction_ensstd' provides the ensemble standard deviation as a measure of epistemic uncertainty. The associated data from LandScan Global for that year is provided at native resolution, as well as estimated per-pixel customers for that year. Lon/lat coordinates represent pixel centers.
Python package xarray is recommended for interfacing with these NetCDF files (it will automatically parse the date format among other conveniences), though many packages and languages are suitable.
Files
Files
(51.4 GB)
| Name | Size | |
|---|---|---|
|
md5:28d70855410a627a3ec4636b74ced0cb
|
3.9 GB | Download |
|
md5:56f3d7eb68de377614a22eab43f86d52
|
2.2 GB | Download |
|
md5:8a06352ea8c3a2f603c0e276339a3170
|
1.2 GB | Download |
|
md5:a36cb946ad53d88bb438a69a159122d2
|
1.1 GB | Download |
|
md5:821b14d22d9b48dadaa25c931cf4c652
|
871.9 MB | Download |
|
md5:a0e3e14a744a00a5cb7a86544126d9fd
|
1.7 GB | Download |
|
md5:1ff958c647b413541fb54b8509d57372
|
2.4 GB | Download |
|
md5:fb6c1cde2ed132861fcc1bcb49d9da97
|
1.7 GB | Download |
|
md5:2c94b28603a6fc0fe5c913b9e5200ba3
|
2.7 GB | Download |
|
md5:673014afa6fe5e18d42c7a24563c100d
|
6.8 GB | Download |
|
md5:ee7f2e2d16b9eeec2dc48ca2c8218062
|
4.8 GB | Download |
|
md5:63fe50afe06c7bd1f0f38378be9c1285
|
7.0 GB | Download |
|
md5:6240ad3077af83a296a9790a41ada77c
|
2.2 GB | Download |
|
md5:eb766d87b2e88ff6537b19e437de8df4
|
3.7 GB | Download |
|
md5:4cdae45f344af313f3075c09bbb930d6
|
7.8 GB | Download |
|
md5:e9817038786f66cd8d6beccdda5345e9
|
1.5 GB | Download |
Additional details
References
- Wang, Z., Miguel O. Román, Ranjay Shrestha, Tian Yao, & Virginia Kalb. (2024). Black Marble User Guide (Collection 2.0). https://viirsland.gsfc.nasa.gov/PDF/BlackMarbleUserGuide_Collection2.0.pdf
- Lebakula, V., Sims, K., Reith, A., Rose, A., McKee, J., Coleman, P., Kaufman, J., Urban, M., Jochem, C., Whitlock, C., Ogden, M., Pyle, J., Roddy, D., Epting, J., & Bright, E. (2025). LandScan Global 30 Arcsecond Annual Global Gridded Population Datasets from 2000 to 2022. Scientific Data, 12(1), 495. https://doi.org/10.1038/s41597-025-04817-z
- Brelsford, C., Tennille, S., Myers, A., Chinthavali, S., Tansakul, V., Denman, M., Coletti, M., Grant, J., Lee, S., Allen, K., Johnson, E., Huihui, J., Hamaker, A., Newby, S., Medlen, K., Maguire, D., Dunivan Stahl, C., Moehl, J., Redmon, D., … Bhaduri, B. (2024). A dataset of recorded electricity outages by United States county 2014–2022. Scientific Data, 11(1), 271. https://doi.org/10.1038/s41597-024-03095-5