HWRF Model

The Hurricane Weather Research and Forecasting model (HWRF) is a limited-area numerical model used by the National Weather Service (NWS) to provide numerical guidance for operational tropical cyclone forecasting. The atmospheric component of HWRF employs the Non-Hydrostatic Model (NMM) dynamic core of the Weather Research and Forecasting (WRF) model. The oceanic component of HWRF is the Princeton Ocean Model for Tropical Cyclones (POM-TC).

2014 HWRF Configuration

The HWRF model is run with three telescopic atmospheric domains. The parent domain has a horizontal resolution of 27-km, with two storm following inner nests at 9- and 3-km horizontal resolution. This configuration is coupled with the POM-TC ocean model, uses Gridpoint Statistical Interpolation (GSI) 3D-Var hybrid data assimilation, and employs the GFDL radiation parameterization invoked hourly. Other parameterizations used include Ferrier microphysics, SAS convection, GFS planetary boundary layer, and GFDL surface layer. For more information please refer to the 2014 HWRF Scientific Documentation.       

2015 HWRF Configuration

The HWRF model is run with three telescopic atmospheric domains. The parent domain has a horizontal resolution of 18-km, with two storm following inner nests at 6- and 2-km horizontal resolution. This configuration is coupled with the POM-TC ocean model, uses GSI 3D-Var hybrid data assimilation, and employs the RRTMG radiation parameterization with partial cloudiness scheme. Other parameterizations used include Ferrier-Aligo microphysics, SAS convection, GFS planetary boundary layer, and Noah land surface. For more information please refer to the 2015 HWRF Scientific Documentation.       

COAMPS-TC Model

CTCX

The Coupled Ocean-Atmosphere Mesoscale Prediction System (COAMPS) is the Navy high-resolution regional operational prediction system. COAMPS is developed by the Naval Research Laboratory (NRL) and consists of data quality control, data assimilation, initialization, a non-hydrostatic atmospheric model and a hydrostatic ocean model (Hodur 1997). The Arakawa C grid is used for both the atmospheric and ocean models. The atmospheric model utilizes the sigma-z vertical coordinate and the ocean model uses the hybrid Sigma/z. A version of COAMPS has recently been developed which is dedicated to the prediction of tropical cyclones (COAMPS-TC). For the 2014 Retrospective HFIP Testing, the model was run on a Mercator projection with one fixed coarse mesh domain and either two or three moving, two-way interactive nested domains.

The model has a 45-km horizontal resolution parent domain, with storm following inner nests of 15- and 5-km horizontal resolution. There are 40 vertical levels with a 32-km model top.

NHC Official forecasts

The NHC issues TC forecasts for AL and EP basins. The forecasts are provided by hurricane specialists who utilize guidance from a varitey of numerical and statistical model forecasts. The ATCF forecast ID of the forecasts are denoted by OFCL.

Cases Verified

The primary focus of this evaluation is the HWRF model. For purpose of comparison, the official forecasts (OFCL) issued by the NHC and forecasts from the COAMPS-TC model (CTCX) were also evaluated. The tables below show the datasets used for each comparison. Entries denoted by "Retro" indicates retrospective (a.k.a pre-implementation) tests performed by EMC and NRL for the indicated year. Cells containing "Oper" indicate the operational forecasts from HWRF, COAMPS-TC, and NHC. Each configuration is indicated by the ID shown at the top tab of the table.

Verification of model (HWRF and CTCX) and official forecasts (OFCL).

WP evaluations include 2 years of operational HWRF forecasts.

Verification

Rapid Intensification (RI) is defined as the 95th percentile of all 24-h intensity change episodes over water of tropical systems in the Atlantic basin. This translates to an increase of 30 kt (15.4 m/s) in a 24-h period, determined from a sample spanning storms from years 1989-2000. Similar criteria is applied for Rapid Weakening (RW), defined as an intensity decrease of 30 kt in 24-h.

The verification of the RI/RW forecasts for each RI/RW threshold is quantified in terms of a binary event (yes or no) based upon which contingency table statistics can be computed.

Further, the RI/RW criteria are relaxed in time and intensity to determine if small errors in timing or amount are leading to large errors in the event statistics.

In order to determine whether a model is able to properly capture the RI magnitude, the distributions of the difference of change in a-deck values in the last 24-h and change in b-deck values in the last 24-h (herein referred to as ABDEL) are examined. These values are computed separately for the 4 quadrants of the contingency table.

The MET-TC package was used to conduct the enhanced RI/RW verification used for this study.