NWS Short-range Hydrologic Ensembles

Note

The short-range hydrologic ensembles are an experimental product, are intended to provide an outlook in advance of a hydrologic event, are produced automatically, and are not official forecasts. Official forecasts are provided by NWS Weather Forecast Offices (WFOs) in conjunction with the NWS River Forecast Centers (RFCs). These forecasts are provided on the NWS Advanced Hydrologic Prediction Service web site (AHPS) or via each RFC's own web site.

Background

Short-term hydrologic ensembles are provided by a system named the Meteorological Model-based Ensemble Forecast System, or MMEFS. The objective of MMEFS is to produce short lead-time (< 10 days) hydrologic ensemble forecasts using forcing fields provided by various meteorological ensemble systems as input to and processed by the NWS River Forecast System (NWSRFS) Ensemble Streamflow Prediction (ESP) system. At this time, MMEFS is used experimentally by the 3 NWS Eastern Region RFCs to process model outputs from the National Centers for Environmental Prediction (NCEP) 12-member Global Ensemble Forecast System (GEFS) as provided via AWIPS and the 21-member Short Range Ensemble Forecast (SREF) system produced at NCEP. The design of MMEFS is flexible enough to easily add other meteorological ensemble sources, such as NAEFS (North American Ensemble Forecast System). An online course, designed to help understand numerical weather prediction (NWP) models, which includes a section on ensembles, is available from UCAR's COMET. Another online course from COMET provides an Introduction to NAEFS.

MMEFS was developed for several reasons:

1. Hydrologic forecast uncertainty is closely linked to the uncertainties associated with precipitation and temperature forecasts used by hydrologic simulation models. MMEFS provides a range of hydrologic model results based on differing meteorological numerical weather prediction (NWP) model forecasts used as inputs.
2. MMEFS explicitly uses short-range (days 1-7) NWP model output, eliminating the need for historical precipitation and temperature data for its results.
3. Using the MMEFS provides a means to further users' understanding of the effects of model inputs used in hydrologic simulations.
4. The hydrologic ensembles are useful surrogates for multiple contingency runs that are typically used by river forecast centers to convey quantitative precipitation forecast (QPF) or quantitative temperature forecast (QTF) uncertainty for worst case scenarios.

Hydrologic Ensemble Overview

Presentation Page Overview

Limitations

There are a few significant limitations with the use of the MMEFS for hydrologic forecasts:

• NWP ensemble data is provided as a collection of point data for a grid of points, so data at locations between grid points must be estimated. The estimation methods range from numerical interpolations to ones incorporating terrain, elevation and climatology. At this time, MMEFS uses numerical interpolation.
• The coarseness of each NWP's ensemble grids varies. Samples of the grid resolution for the MARFC area for the GEFS (AWIPS rendition), NCEP GEFS, SREF, and a 12-km version of the WRF are shown here. If you are accustomed to the grid spacing of the GFS and note the difference between it and the GEFS, which is based on the GFS mode, the spacing is increased for the GEFS (greater distance between grid points) due to computational requirements necessary to perform multiple ensemble member simulations.
• The amount of bias in the NWP ensemble outputs affects the hydrologic simulations. Preliminary results have shown an insufficient spread in precipiation amounts, so verification of the processed ensemble results (downscaled precipitation and temperatures and flows/stages) needs to be performed to assess performance and characteristics. NCEP plans to analyze and apply bias correction measures to the NWP ensembles at some time in the future.
• MMEFS does NOT include the effect of hydrologic modeling errors independent of the temperature & precipitation inputs; so, during periods of no forecast precipitation and when temperature differences between NWP ensemble member results either are negligible or result in frozen temperatures, the resulting hydrologic ensembles most likely wil show little spread.
• The component of NWSRFS used in daily RFC operations, the Operational Forecast System (OFS) allows controls to override default model behavior. These modifications can be applied for past and future periods. ESP is affected by the results of these overrides applied to past periods but does not incorporate the controls into the future.
• The snow model used within MMEFS is driven by air temperatures only. As such the precipitation type for winter storms can be incorrect. Also, average melt rates for the different months of the snow season along with air temperatures drive snowmelt. Variations from average conditions, such as high dewpoints and high winds, which amplify melt rates, are not incorporated into MMEFS. Under these conditions snowmelt calculated within MMEFS can be underdone.
• Currently, ESP within NWSRFS is limited to start and end times of 12UTC. NWP forecast cycles do not always provide data matching those requirements. For example, the 00UTC GEFSA forecast cycle's first forecast data is for the interval ending at 06UTC and its last is for 00UTC. With the ESP data requirement, other data must be used to fill in at the start of the simulation for 18UTC and 00UTC, and at the end of the simulation for 06UTC and 12UTC. Observed data (extracted from OFS files) fills in data at the start of ESP, and zero precipitation is used to complete the data at the end of the run. For temperatures, the same approach is used for fill-in at the start of the simulation, but at the end of the simulation, the pattern of temperatures from the previous day is reproduced to supply the required data.
• Tropical systems can be difficult to pick up due to the tight nature of those storms and the coarse resolution of the GEFS and SREF.
• Currently the SWE graphics generation package is hard-wired to display "Snow Cover" along the left axis, even though the quantities produced are water equivalents.
• The current downscaling approach (numerical interpolation) will have difficulty with elevation-dependent significant temperature change scenarios.

Acknowledgements

MMEFS is a joint collaboration of the ER DOHs: Thomas Adams (NWS/OHRFC), Joseph Ostrowski (NWS/MARFC), and Robert Shedd (NWS/NERFC)