Codes Employed

The components of the end-to-end forecast system used for this test included:

    • WRF Preprocessing System (WPS) (v3.1.1)

    • WRF-ARW model (tag from the WRF Repository, based on v3.1.1        with a considerable number of updates)

    • WRF Post Processor (WPP) (v3.1)

    • Model Evaluation Tools (MET) (v3.0.1) *Includes PB2NC error       described here

    • NCAR Command Language (NCL) for graphics generation

    • Statistical programming language, R, to compute confidence        intervals

Domain Configuration

    • Contiguous U.S. (CONUS) domain with 15-km grid spacing

Click thumbnail for larger image.

    • 403 x 302 gridpoints, for a total of 121,706 horizontal gridpoints

    • 56 vertical levels (57 sigma entries); model top at 10 hPa

    • Lambert-Conformal map projection

Sample Namelists

    • namelist.wps

    • namelist.input

AFWA Reference Configuration

Microphysics:	WRF Single-Moment 5 Scheme
Radiation (LW/SW):	RRTM/Dudhia
Surface Layer:	Monin-Obukhov Similarity Theory
Land Surface:	Noah
PBL:	Yonsei University Scheme
Convection:	Kain-Fritsch Scheme

    • Sample namelist.input

Other run-time settings

    • Long timestep = 90 s; Acoustic step = 4
    • Calls to the boundary layer and microphysics parameterization        were made every time step
    • Calls to the cumulus parameterization were made every 5        minutes
    • Calls to radiation were made every 30 minutes

The end-to-end system for this test did not include a data assimilation component.

Initial and Boundary Conditions

    • Initial conditions (ICs) and Lateral Boundary Conditions
       (LBCs): 0.5 x 0.5 degree Global Forecast System (GFS) model

    • Lower Boundary Conditions (LoBCs): Agricultural
       Meteorological Model (AGRMET) System output

    • SST Initialization: Fleet Numerical Meteorology and        Oceanography Center (FNMOC) daily, real-time sea surface        temperature (SST) product

Cases Run

• Forecast Dates: 2 June 2008 - 31 May 2009

• Initializations: Every 36 hours, including both 00 and 12 UTC

• Forecast Length: 48 hours; output files generated every 3 hours

The tables below list the forecast initializations that failed to complete the end-to-end process due to the reasons described in the table. All incomplete forecasts were due to missing or bad input data sets, not model crashes.

Missing Forecasts:

Affected Case	Missing Data	Reason
2008071000	WRF Output	Missing SST Input Data
2008091512	WRF Output	Bad SST Input Data
2008101512	WRF Output	Bad SST Input Data
2008101700	WRF Output	Bad SST Input Data
2008101812	WRF Output	Bad SST Input Data
2008102112	WRF Output	Missing AGRMET Input Data
2008121112	WRF Output	Bas SST Input Data
2009030100	WRF Output	Missing SST Input Data
2009040112	WRF Output	Bad SST Input Data
2009042212	WRF Output	Bad SST Input Data
2009052400	WRF Output	Missing SST Input Data
2009052512	WRF Output	Missing SST Input Data

Missing Verification:

Affected Case	Missing Data	Reason
2008071300	Incomplete sfc and upper air verification beyond 33 h	Missing Prepbufr data
2008071412	Incomplete sfc and upper air verification beyond 24 h	Missing Prepbufr data
2008101400	Incomplete sfc and upper air verification beyond 39 h	Missing Prepbufr data
2009012700	Incomplete sfc and upper air verification for 24-27 h	Missing Prepbufr data

Verification

The Model Evaluation Tools (MET) package, comprised of:

    • grid-to-point comparisons - utilized for surface and upper air        model data
    • grid-to-grid comparisons - utilized for QPF

was used to generate objective verification statistics, including:

    • Bias-corrected Root Mean Square Error (BCRMSE), Root Mean        Square Error (RMSE) and Mean Error (Bias) for:
        • Surface: Temperature (2 m), Dew Point Temperature (2 m)            and Winds (10 m)
        • Upper Air: Temperature, Dew Point Temperature and Winds

    • Gilbert Skill Score (GSS) and Frequency Bias (FBias) for:
        • 3-hr and 24-hr Precipitation Accumulation intervals

Each type of verification metric is accompanied by confidence intervals (CIs), at the 99% level, computed using a parametric method for the surface and upper air variables and a boostrapping method for precipitation.

Area-averaged verification results were computed for the full domain, as well as 14 sub-domains.