ExREF is a near-realtime experimental ensemble used as a proving ground for development and testing of new ensemble methodologies.

It is a testbed for new generations of regional ensembles and it is a collaborative project between WPC, EMC, AWC, DTC, GSD, HMT, and more.

  • Overview
  • Collaborators
  • Config
  • About LAPS
  • Distribution
  • Verif
  • Ref
  • Future

ExREF | Overview

The ExREF has eight model members with 9-km grid spacing over a large portion North America. It uses the Advanced Research Weather Research and Forecasting (WRF-ARW) model (Skamarock et al. 2008) and uses multiple physics and a variety of initial and boundary conditions. Forecasts are initialized four times a day and run out to 84 h.


The Developmental Testbed Center (DTC) is using ExREF to test several innovations, which, if proven successful, will inform the composition of future implementations of the NWS operational Short Range Forecasting System (SREF).

Ongoing experiments involve the importance of microphysics diversity to enhance ensemble spread, use of local-scale perturbations to enhance the initial conditions, and sensitivity to horizontal grid spacing.


ExREF has been a component of the Hydrometeorology Testbed (HMT) NWP suite since the beginning of 2013. Its results are distributed to the NWS, etc.


ExREF has been running in near realtime since the beginning of calendar year 2013.

ExREF | Collaborators

This is a collaborative project between many agencies.

NOAA ESRL Global Systems Division (GSD)

› Run ExREF in near-realtime
› Develop new initialization methods (dynamic downscaling of perturbations)

NOAA Environmental Modeling Center (EMC)

› Use ExREF as a proving ground of innovations for the 16-km operational Short-Range Ensemble Forecast System (SREF)

Developmental Testbed Center (DTC)

› Conduct testing and evaluation of ExREF and SREF

NOAA Weather Prediction Center (WPC)

› Testing ExREF during the July, 2013 Flash Flood Intense Rainfall Experiment

NOAA Hydrometeorology Testbed (HMT)

› Use ExREF for numerical guidance of extreme pcp events

National Weather Service WFOs & RFOs

› Use ExREF in AWIPS as forecaster tool
› Coming soon to: Sacramento, CA and NWS Eastern Region


  Ligia Bernardet1,*
  Steve Albers1,+
  Linda Wharton1,*
  Isidora Jankov1,+
  Zoltan Toth1
  David Reynolds2
  Kelly Mahoney2,*
  Ellen Sukovich2,*
  William Hogsett3
  David Novak3
  Jun Du4

1 NOAA/ ESRL/ Global Systems Division, Boulder, CO
2 NOAA/ ESRL/ Physical Sciences Division, Boulder, CO
3 NOAA Weather Prediction Center, College Park, MD
4 NOAA Environmental Modeling Center, College Park, MD
* Coop Inst for Research in Env Sciences/Univ of CO – Boulder, CO
+ Coop Inst for Research in Atmosphere/CO State Univ - Fort Collins, CO

ExREF | Configuration

The ensemble uses a variety of initial and boundary conditions to represent uncertainty in the large scale fields.

Microphysics diversity is also used as QPF is very sensitive to this parameterization

Highlights include
› 9-km grid spacing over most of North America
› Forecast length 84-h
› Initialization four times a day using
› Global Forecast System (GFS)
› Local Analysis and Prediction System (LAPS; laps.noaa.gov)
› Boundary Conditions: Global Ensemble Forecast System (GEFS)
› Microphysics: Thompson, Ferrier, WSM6

ExREF configuration since June 27, 2013

Model and version

WRF-ARW v3.4.1


North America

Horizontal grid spacing

9 km

Number of grid points

801 x 581

Grid projection


Grid center

lat = 41.02480; lon = -97.95627

Other grid parameters

truelat1  =  30.0 truelat2  =  60.0,;

stand_lon = -106.0

Number of vertical levels



54 s

Microphysics parameterization


Cumulus parameterization


Planetary boundary layer parameterization


Surface layer parameterization

MM5 Monin-Obukhov scheme

Land surface model

Thermal diffusion scheme

Radiation parameterization

RRTM LW / Dudhia SW;  radt=10 min

Frequency of run initialization

00, 06, 12, 18 UTC

Forecast length

84 h



Boundary Conditions


Products delivered

GRIB2 files for members and mean via ftp. A geographical subset over the West Coast, termed DWR domain, is delivered to NWS WR via LDM. Graphics and verification on web.

Member differentiation

The ensemble uses various initial and boundary conditions, as well as multiple microphysics parameterizations to represent uncertainties in various aspects of the forecast system and produce a reasonable ensemble spread.






































GEFS - Global Ensemble Forecasting System

GFS - Global Forecast System

t-LAPS - Traditional LAPS (utilizes Barnes analysis)

v-LAPS - Variational LAPS (utilizes variational analysis)

WRF - Weather Research

WSM6 - WRF single-moment  6-class microphysics scheme

ExREF domain

Larger parent domain

Figure 1. ExREF domain.

Smaller nest domain

Figure 2. ExREF subdomain sent to WFO.

ExREF | About LAPS

LAPS is a data assimilation system that used for analysis, situational awareness, and forecast WRF model member initializations. The GFS global model or from members of the Global Ensemble Forecast System (GEFS) are used for the WRF model member boundary conditions.

Local Analysis and Prediction System (LAPS)

› Data assimilation system with options
  ›› Traditional analyses (Barnes analysis; LAPS-t)
  ›› Variational analyses (Multi-scale variational minimization; LAPS-v)
› Employs a large-scale analysis as a first guess (for ExREF, uses GFS)
› Blends a variety of in-situ and remotely sensed data
› Twenty year old LAPS hot start capability adds clouds and vertical motion at the initial time
› Has 150 users worldwide, including US National Weather Service
› Can be run in regional or global domains

LAPS Diagram

Ensemble product delivery flowchart

ExREF | Distribution

ExRef is run on a super computer at NOAA/ESRL/GSD and the ensemble results are sent to the
› California Sacramento NWS WFO
› NOAA Weather Prediction Center, where it will be used in the summer 2013 Flash Flood and Intense Rainfall Experiment.

› A subset of the forecast fields produced by ExREF is publicly available at: http://laps.noaa.gov/hmt/hmt.html Note: website takes several moments load.

Ensemble product delivery flowchart
› The ensemble statistical mean fields are sent using the Unidata Local Data Manager (LDM) protocol to the NWS Western Region HQ, and distributed to field offices for use in the Graphic Forecast Editor for use, and

ExREF | Verification

NCEP/ EMC Ensemble Verification Package

Ensemble verification package used at EMC has been implemented at GSD for ExREF evaluation and verification.

ExREF Verification (4/10 – 5/10, 2013)

LAPS hot-start leads to large ETS gains at the 3-h forecast lead time for the members that use Thompson and WSM6 microphysics (M1, M3, M4, M6, and M7) over the GFS initialization (M0).

Hot start of members that use Ferrier scheme (M2, M5) problematic – need to reassess how LAPS microphysics maps onto WRF’s.

ETS of ensemble mean is larger than that of the members for most forecast lead times.

Threat Score plot one
Threat Score plot one Threat Score plot one

Verification Continued

Two-meter T, Td compared favorably with observations (dashed line).

T, Td did not display the positive warm moist bias exhibited by the North American Mesoscale (NAM) model in the afternoon hours.

Threat Score plot one
Threat Score plot one

ExREF | References and Contact

Reference Papers

› Albers, S., J. McGinley, D. Birkenheuer, and J. Smart, 1996. The Local Analysis and Prediction System (LAPS): Analyses of clouds, precipitation, and temperature. Wea. Forecasting, 11, 273–287.

› Bernardet, L. et all 2013. The Experimental Regional Ensemble Forecast System. 14th WRF Users’ Workshop, Boulder, CO, June 24-28, 2013.

› Jankov, I., T. Jensen, L. Bernardet, J. Du, D. Novak, Z. Toth, 2013. Initial condition perturbation tests within Experimental Regional Ensemble Forecasting (ExREF) system. 14th WRF Users’ Workshop, Boulder, CO, June 24-28, 2013.

› Toth, Z., S. C. Albers, and Y. Xie, 2012. Analysis of Fine Scale Weather Phenomena. Bull. Amer. Meteor. Soc., 93 (3). ES35-ES38.

› Xie, Y., S. Koch, J. McGinley, S. Albers, P. E. Bieringer, M. Wolfson, and M. Chan, 2011. A space–time multi- scale analysis system: A sequential variational analysis approach. Mon. Wea. Rev., 139, 1224–1240.

Contact Information

For more information, email: Ligia Bernardet

ExREF | Future

The the project includes plans for future development and testing.

› One plan is to enhance the Initial Conditions with the use of dynamic downscaling, a technique to combine the un-certainty represented by the difference between members of the GFS ensemble with local data assimilation.

Below is a design diagram.

› There are plans for high-resolution movable nests in ExREF.