GEFSv12 Quantile Mapping Data & Documentation

The forecasts on this web page are provided to demonstrate the impact of applying a quantile-mapping (QM) and dressing routine to bias correct raw quantitative precipitation forecasts (QPF) from NOAA’s Global Ensemble Forecast System version 12 (GEFSv12, see Hamill et al. 2022, Guan et al. 2022). This routine was designed to improve precipitation in NOAA’s National Blend of Models that incorporates the GEFSv12 30-member ensemble into its forecast guidance.

About These Products

Quantile mapping is a technique that bias corrects the forecast for a given lead time by replacing each ensemble member value of precipitation with observed values having the same quantiles with respect to the observed climatology. The bias correction for the products on this webpage is applied to all 30 members of daily 6-hourly forecasts from the 00 UTC cycle of the GEFSv12. The cumulative distribution functions (CDFs) for building the quantile-mapping relationships between the forecasts and observations are generated using data within the period between 1 January 2002 to 31 December 2019. The CDFs for the observed precipitation combines the Climatology-Calibrated Precipitation Analyses (CCPA; used for values inside CONUS) and Multi-SourceWeighted-Ensemble Precipitation analyses (MSWEP; used for values outside CONUS). The forecast CDFs come from the GEFSv12 reforecast system, which is a five-member ensemble that is available each day from the 00 UTC initial conditions for the same period of analysis as the CCPA/MSWEP. The figure below, from Hamill et al. (2023), shows examples of the CDFs for the forecast and analysis between two different months at a grid point near Ithaca, NY.

figure 1

From Figure 1 of Hamill et al. (2023): (a) August and (b) October empirical cumulative distribution functions of 0–1-day accumulated GEFSv12 reforecast data at the grid point nearest to Ithaca, NY, and 24-h observed total precipitation at the Ithaca Game Farm station using 2000–19 data.

The bias-correction for these products include additional steps of closest-member histogram weighting and dressing with a Guassian error distribution, the details of which are provided in Hamill et al. (2023). The bias-corrected amounts at each grid point are stored for each lead time and day of the year, and then applied to real-time GEFSv12 forecasts for the 00 UTC cycle each day when the data becomes available.

Usage

The dropdowns on the web page contain the 6-hourly ensemble mean of the quantile-mapped (Mean Precipitation) and raw GEFSv12 precipitation (Raw Mean Precipitation) forecasts, as well as probability of precipitation exceedance for thresholds of 0.254, 1, 5, 10, and 25 mm. The slider bar over the image lets the user view the forecasts every 6 hours out to a 240 hour lead time (i.e., 10 days).

An evaluation showing how the QM forecasts perform compared to the raw forecasts for various thresholds, seasons, and storm types are provided by Stovern et al. (2023).

Data Source

The real time GEFSv12 forecasts for the 00 UTC cycle for all 30 members are downloaded from the NOAA Operational Model Archive and Distribution System (NOMADS) every day when the data becomes available.

The five members from the reforecast GEFSV12 can be downloaded from the grib data store at Amazon web Services, https://noaa-gefsretrospective.s3.amazonaws.com/

The data and algorithms used to generated the bias-correction are available for others to use at ftp://ftp.cdc.noaa.gov/Projects/GEFSv12/NBM/ and https://github.com/ThomasMoreHamill/GEFSv12_NBM.

Referencing Forecasts

To reference forecast plots, we ask that you acknowledge PSL as in ”image is provided by the NOAA Physical Sciences Laboratory, Boulder, Colorado, USA, from their website at https://psl.noaa.gov/”. The paper that should be cited along with this link is Hamill et al. (2023) and Stovern et al. (2023), listed in the references list below.

References

  • Stovern, D. R., T. M. Hamill, and L. L. Smith, 2023: Improving National Blend of Models probabilistic precipitation forecasts using long time series of reforecasts and precipitation reanalyses. Part II: Results. Mon. Wea. Rev., 151, 1535–1550, https://doi.org/10.1175/MWR-D-22-0310.1.
  • Guan, H., and Coauthors, 2022: GEFSv12 reforecast dataset for supporting subseasonal and hydrometeorological applications. Mon. Wea. Rev., 150, 647–665, https://doi.org/10.1175/MWRD-21-0245.1.
  • Hamill, T. M., and Coauthors, 2022: The reanalysis for the Global Ensemble Forecast System, version 12. Mon. Wea. Rev., 150, 59–79, https://doi.org/10.1175/MWR-D-21-0023.1.
  • Hamill, T. M., D. R. Stovern, and L. L. Smith, 2023: Improving National Blend of Models probabilistic precipitation forecasts using long time series of reforecasts and precipitation reanalyses. Part I: Methods. Mon. Wea. Rev., 151, 1521–1534, https://doi.org/10.1175/MWR-D-22-0308.1.

Webpage contributors

Tom Hamill developed the code for the QM bias-correction, closest-member histogram weighting, and Gaussian dressing. Diana Stovern developed the methodology to apply the bias correction to real-time forecasts of the GEFSv12, including retrieving the GEFSv12 forecasts from NOMADS and generating the plots daily. Don Hooper and Cathy Smith created the webpage and run the webpage in real-time.

Note that these forecasts are experimental. NOAA/PSL is not responsible for any loss occurred by the use of these forecasts.