Surrogate flash flooding: Probabilistic excessive rainfall predictions from the High Resolution Ensemble Forecast (HREF) system
Data files
Feb 14, 2025 version files 293.63 GB
-
202105_00.tar.gz
2.82 GB
-
202105_12.tar.gz
2.81 GB
-
202106_00.tar.gz
4.32 GB
-
202106_12.tar.gz
4.59 GB
-
202107_00.tar.gz
5.52 GB
-
202107_12.tar.gz
5.80 GB
-
202108_00.tar.gz
5.40 GB
-
202108_12.tar.gz
5.74 GB
-
202109_00.tar.gz
4.70 GB
-
202109_12.tar.gz
4.80 GB
-
202110_00.tar.gz
4.46 GB
-
202110_12.tar.gz
4.37 GB
-
202111_00.tar.gz
3.68 GB
-
202111_12.tar.gz
3.73 GB
-
202112_00.tar.gz
3.35 GB
-
202112_12.tar.gz
3.42 GB
-
202201_00.tar.gz
3.36 GB
-
202201_12.tar.gz
3.59 GB
-
202202_00.tar.gz
3.05 GB
-
202202_12.tar.gz
3.05 GB
-
202203_00.tar.gz
3.53 GB
-
202203_12.tar.gz
3.52 GB
-
202204_00.tar.gz
3.07 GB
-
202204_12.tar.gz
2.98 GB
-
202205_00.tar.gz
4.46 GB
-
202205_12.tar.gz
4.12 GB
-
202206_00.tar.gz
4.53 GB
-
202206_12.tar.gz
4.72 GB
-
202207_00.tar.gz
5.47 GB
-
202207_12.tar.gz
4.55 GB
-
202208_00.tar.gz
5.83 GB
-
202208_12.tar.gz
5.87 GB
-
202209_00.tar.gz
4.47 GB
-
202209_12.tar.gz
4.36 GB
-
202210_00.tar.gz
4.22 GB
-
202210_12.tar.gz
4.16 GB
-
202211_00.tar.gz
3.73 GB
-
202211_12.tar.gz
3.53 GB
-
202212_00.tar.gz
3.82 GB
-
202212_12.tar.gz
3.80 GB
-
202301_00.tar.gz
3.62 GB
-
202301_12.tar.gz
3.52 GB
-
202302_00.tar.gz
2.98 GB
-
202302_12.tar.gz
3.09 GB
-
202303_00.tar.gz
3.36 GB
-
202303_12.tar.gz
3.61 GB
-
202304_00.tar.gz
3.71 GB
-
202304_12.tar.gz
3.83 GB
-
202305_00.tar.gz
4.65 GB
-
202305_12.tar.gz
4.76 GB
-
202306_00.tar.gz
5.04 GB
-
202306_12.tar.gz
5.25 GB
-
202307_00.tar.gz
5.32 GB
-
202307_12.tar.gz
5.59 GB
-
202308_00.tar.gz
5.32 GB
-
202308_12.tar.gz
5.33 GB
-
202309_00.tar.gz
4.36 GB
-
202309_12.tar.gz
4.37 GB
-
202310_00.tar.gz
3.89 GB
-
202310_12.tar.gz
3.88 GB
-
202311_00.tar.gz
3.21 GB
-
202311_12.tar.gz
3.41 GB
-
202312_00.tar.gz
2.69 GB
-
202312_12.tar.gz
3.62 GB
-
202401_00.tar.gz
3.50 GB
-
202401_12.tar.gz
3.64 GB
-
202402_00.tar.gz
2.56 GB
-
202402_12.tar.gz
2.69 GB
-
202403_00.tar.gz
3.52 GB
-
202403_12.tar.gz
3.41 GB
-
202404_00.tar.gz
2.67 GB
-
202404_12.tar.gz
2.77 GB
-
202405_00.tar.gz
1.56 GB
-
202405_12.tar.gz
1.57 GB
-
grid_stat.zip
8.64 MB
-
lsrflash.zip
6.20 MB
-
lsrreg.zip
6.17 MB
-
README.md
4.72 KB
-
st4gari.zip
6.41 MB
-
st4gffg.zip
6.49 MB
-
usgs.zip
6.09 MB
Abstract
Probabilistic forecasts of excessive rainfall based on the fraction of High Resolution Ensemble Forecast (HREF) members predicting precipitation above a given threshold are used widely in predicting excessive rainfall; however, there is not yet a published study evaluating the skill of these forecasts. In this study, we document the performance of these forecasts over a three-year period, including regional and seasonal variations in skill. We find that there is considerable sensitivity to how excessive rainfall events are defined, especially in regions with large differences in number of excessive rainfall events between different datasets. When verifying against Stage IV exceedances of flash flood guidance (FFG), both the 0000 UTC and 1200 UTC HREF probabilities of exceeding 6-h FFG thresholds exhibit higher Brier Skill Score (BSS) than the operational 0900 UTC day-one excessive rainfall outlook (ERO) in five of eight regions in the CONUS, while probabilities of exceeding fixed 6-h or 12-h precipitation thresholds provide higher BSS than the ERO in another two regions. There is regional variability in the thresholds providing the highest BSS, with FFG (or 75% of FFG) generally providing the best forecasts in the eastern US, but fixed thresholds providing the best forecasts in the western US. Only in the southeastern US are threshold-based HREF forecasts unable to beat the ERO. 1200 UTC HREF-based forecasts using regionally optimal thresholds beat the ERO by 25-30% in terms of BSS. Our results suggest that HREF probabilities of exceeding precipitation thresholds have considerable value for excessive rainfall prediction.
https://doi.org/10.5061/dryad.7m0cfxq4j
Description of the data and file structure
This dataset consists of gridded probabilistic forecasts of precipitation exceeding a large number of precipitation thresholds derived from the operational NCEP High Resolution Ensemble Forecast (HREF) v3 during 2021-2024. The forecasts are on the HREF grid, and are derived using the operational methodology for HREF QPF exceedance probabilities, using a neighborhood radius of 40 km, and weighting time-lagged ensemble members less than non-time-lagged members. The thresholds include fixed precipitation thresholds, average recurrence interval thresholds, and ratios of flash flood guidance.
Files and variables
HREF forecasts are zipped up by month and initialization time (0000 UTC or 1200 UTC). For example, the file named 202106_00.tar.gz contains daily HREF forecasts initialized at 0000 UTC every day during June 2021.
Individual HREF forecasts are in netCDF files. The naming convention is as follows:
href_f12-36${yyyymmdd00}.nc (probabilities of exceeding QPF thresholds during 24-h period contained within the 12-36 h forecast initialized at 0000 UTC)
href${yyyymmdd12}.nc (probabilities of exceeding QPF thresholds during 24-h period contained within the 0-24 h forecast initialized at 1200 UTC)
href12h_f12-36${yyyymmdd00}.nc (probabilities of exceeding 12-h QPF thresholds during 1200-0000, 1500-0300, 1800-0600, 2100-0900, or 0000-1200 UTC, based on forecast initialized at 0000 UTC)
href12h${yyyymmdd12}.nc (probabilities of exceeding 12-h QPF thresholds during 1200-0000, 1500-0300, 1800-0600, 2100-0900, or 000-1200 UTC, based on forecast initialized at 1200 UTC)
href6h_f12-36${yyyymmdd00}.nc (probabilities of exceeding 6-h QPF thresholds during 1200-1800, 1500-2100, 1800-0000, 2100-0300, 0000-0600, 0300-0900, or 0600-1200 UTC, based on forecast initialized at 0000 UTC)
href6h${yyyymmdd12}.nc (probabilities of exceeding 6-h QPF thresholds during 1200-1800, 1500-2100, 1800-0000, 2100-0300, or 0000-0600, 0300-0900, or 0600-1200 UTC, based on forecast initialized at 1200 UTC)
href3h_f12-36${yyyymmdd00}.nc (probabilities of exceeding 3-h QPF thresholds during 1200-1500, 1500-1800, 1800-2100, 2100-0000, 0000-0300, 0300-0600, 0600-0900, or 0900-1200 UTC, based on forecast initialized at 0000 UTC)
href3h${yyyymmdd12}.nc (probabilities of exceeding 3-h QPF thresholds during 1200-1500, 1500-1800, 1800-2100, 2100-0000, or 0000-0300, 0300-0600, 0600-0900, or 0900-1200 UTC, based on forecast initialized at 1200 UTC)
In each file, the probability of exceeding each threshold is encoded as a different variable. The naming convention is as follows:
24-h thresholds:
prob{fixed} where ${fixed} is a number in inches
prob{recurrence}yARI where ${recurrence} is a recurrence interval in years
Other thresholds:
prob${duration}h${fixed} where ${duration} is the duration of the QPF threshold and ${fixed} is a number in inches
prob${duration}h${recurrence}yARI where ${duration} is the duration of the QPF threshold and ${recurrence} is a recurrence interval in years
prob${duration}h${ratio}FFG where ${duration} is the duration of the QPF threshold and ${ratio} is the ratio of flash flood guidance
The verification dataset is shared in the following files:
grid_stat.zip contains all the UFVS events as well as the 0900 UTC day one WPC excessive rainfall outlook
st4gffg.zip contains all Stage IV exceedances of flash flood guidance
st4gari.zip contains all Stage IV exceedances of the 5-year average recurrence interval
lsrreg.zip contains all local storm reports of regular flooding
lsrflash.zip contains all local storm reports of flash flooding
usgs.zip contains all USGS stream gauge indications of flooding
The naming convention of the individual files uses ${yyyymmdd}12, where ${yyyymmdd} is the ending day of the 24-h verification period (1200 UTC to 1200 UTC)
Code/software
The data can be analyzed using python or other open source packages.
Access information
Data was derived from the following sources:
- Raw forecasts (in GRIB2 format) from the HREF members were downloaded from NOMADS in real time, or obtained from the NOAA High Performance Storage System (HPSS).
- Average recurrence interval (ARI) grids were derived from NOAA Atlas 14 and as described in the manuscript.
- Flash flood guidance (FFG) grids were obtained from the NOAA High Performance Storage System (HPSS).
- Observed occurrences of excessive rainfall were obtained from the NOAA Weather Prediction Center.
Probabilistic forecasts of excessive rainfall were constructed based on 0000 and 1200 UTC daily initializations of the High Resolution Ensemble Forecast (HREF) system. We used the standard neighborhood probability approach to create forecasts of exceeding a large number of fixed quantitative precipitation forecast (QPF), average recurrence interval (ARI), and ratios of flash flood guidance (FFG) thresholds. We verified the performance of the resulting day-one forecasts for the 1200 - 1200 UTC period against Stage IV exceedances of FFG as well as against the Unified Flood Verification System (UFVS), and compared against the 0900 UTC day one Weather Prediction Center offical Excessive Rainfall Outlook. More details are provided in the manuscript.