APPENDIX C: JTWC AND NHC OBJECTIVE AID DESCRIPTIONS

1. JTWC OBJECTIVE AIDS

The JTWC divides objective aids into six categories: extrapolation, climatology and analog, statistical, dynamic, hybrid and empirical or analytical. The following is a brief discussion of the objective aids within each of these categories.

Extrapolation techniques are used at JTWC for both track and intensity forecasting. Forecast speed, direction and intensification are computed by taking the difference between the current working best track position and the 12-hour-old best track position. The accuracy of XTRP is obviously dependent on the accuracy of the working best track positions. Thus, for weak and poorly defined circulations, the performance of XTRP will be poor, whereas for intense systems with an eye, the XTRP forecast should be quite representative of the actual past motion of the storm. Statistically, XTRP performs quite will for the first 12 to 24 hours, and can perform well for later forecast times if the synoptic situation is such that steady movement and intensification are expected.

The historical data base in the Northwest Pacific, is 1945-81, and 1900-90 for the rest of JTWC's AOR. The climatological and analog objective aids use subsets of this database.

(1) Climatology (CLIM)

CLIM selects all historical storms that have best track positions that fall within a 6 x 6 degree box centered on the current position of the storm, and that occur in the month or months covered by the past 48 hours. If no historical storms satisfy both of these criteria, the time window is set at all 12 months.

CLIM then produces 24, 48, and 72 hour forecast positions by an equally weighted averaging of all historical storm best tracks.

(2) TYAN93

An analog forecast is a weighted average of the past motion of the current storm and a selected sample of historical TC tracks. The historical data base is the same as that used by CLIM. Historical storms are included in the computation of a forecast if they are the best match of the past 12 hour and 24 hour motion vectors and have one or more best track positions that meet a date and time criteria. This aid produces a list of the top five matching storms along a straight or recurving track for 12 and 24 hour past motion vector matches and for a blend of the straight and recurving storms.

A statistical objective aid applies regression techniques/predictions to a given situation based on a statistical characteristic of the current storm motion, past storm motion and intensity, or the surrounding environment. The following is a brief description of the three statistical aids used at JTWC.

(1) Climatology and Persistence (WPCLIPER or CLIP)

A statistical regression technique that is based on climatology, current position and 12-hour and 24-hour past movement. This technique is used as a crude baseline against which to measure the forecast skill of other more sophisticated techniques. CLIP in the Northwest Pacific uses third-order regression equations and is based on the work of Xu and Neumann (1985). This model consists of 12 regression equations, six of which are used to give predicted zonal speed in 12 hour increments, and the other six used to give predicted meridional speed in 12 hour increments. The resulting speed predictions are integrated to give forecast positions at 24, 48, and 72 hours.

(2) Colorado State University Model (CSUM)

A statistical-dynamical technique based on the work of Matsumoto (1984). It actually consists of three sets of independent regression equations that can be used to generate the 24, 48, or 72 hour forecast position. The equation set used for a particular forecast depends on whether the present (for the 24 hour forecast) or forecast positions of the storm are deemed to be below, on, or above the subtropical ridge. This determination is made on the basis of the direction of motion over the preceding 24 hours according to the following definitions:

(a) below the ridge: set initially, unless b) or c) apply,

(b) on the ridge: direction of motion 330 to 030 degrees,

(c) above the ridge: direction of motion 031 to 120 degrees.

Synoptic analysis and numerical prognosis inputs to the regression equations are interpolated 500-hPa height values taken at + or - 40 degrees longitude and +35 and - 10 degrees latitude relative to the present or predicted location of the storm according to the following schedule:

(a) 24 hr forecast: current and 24 hour-old-analyses,

(b) 48 hr forecast: current analysis and 24 hour prognosis,

(c) 72 hr forecast: 24 and 48 hour prognoses.

NOGAPS 200-hPa heights are substituted for the 500-hPa values for storms that are north of the ridge and have a wind intensity of greater than 90 knots. The 24 hour forecast equations also include the current and 24-hour-old position of the storm.

(3) JTWC92

A statistical-dynamical technique based on the work of C. J. Neumann (Englebretson, 1992). Predictor parameters include current and 12-hour-old position of the storm, a version of WPCLIPER, deep layer mean analysis and prognostic fields, and regression coefficients needed to blend the output of numerous iterations of the statistical prediction technique employed to develop the track forecasts.

The following are brief descriptions of the seven dynamic aids used at JTWC.

(1) NOGAPS Vortex Tracking Routine (NGPS)

This objective technique follows the movement of the vortex as analyzed and predicted in the NOGAPS 1000-hPa wind field using an isogon fix method developed at Fleet Numerical Oceanography Center. A search for the vortex is conducted every six hours in the vicinity of the storm through 72 hours even if the vortex is temporarily lost.

(2) One-Way Influence Tropical Cyclone Model (OTCM)

This technique is a coarse resolution (205 km grid), three layer, primitive equation model with a horizontal domain of 6400 x 4700 km. OTCM is initialized using six hour or 12 hour prognostic fields from the latest NOGAPS run, and the initial fields are smoothed and adjusted in the vicinity of the storm to induce a persistence bias into OTCM's forecast. A symmetric bogus vortex is then inserted, and the boundaries are updated by NOGAPS fields as the integration proceeds. The bogus vortex is maintained against frictional dissipation by an analytical heating function. The forecast positions are based on the movement of the vortex in the lowest layer of the model (effectively 850 hPa).

This model is set up as follows:

a) 3-layers: 1000-700, 700-400, and 400-100 hPa

b) horizontal resolution: 205 km

c) approximate domain: 6000 km east-west by 4800 km

d) initialized off the appropriate prognostic fields (usually six or 12 hr progs) from latest NOGAPS run.

e) boundaries updated by NOGAPS field value every 12 hours during run.

f) bogus with a symmetric, medium-sized, deep, mass-balanced bogus vortex with size and intensity adjustable based on best track information.

The synoptic wind field in the vicinity of the current storm position is smoothed and adjusted to provide a steering that corresponds with 12-hour past movement to give OTCM an initial persistence bias. The bogus vortex is then added and model integration begins.

(3) Full Beta and Advection Model (FBAM)

This model is an adaption of the Beta and Advection model used by NMC. This model combines steering based on smoothed NOGAPS deep layer mean (DLM) wind fields without an empirical "propagation" correction. The DLM mean wind field is computed using all NOGAPS pressure levels from 1000 hPa to 100 hPa, thus giving a deep definition of environmental steering. Maximum weight is given to the 700-hPa level. The DLM fields are also smoothed by retaining only NOGAPS spectral wave numbers less than 18, or effectively, wavelengths greater than about 2000 km.

The steering is assumed to be horizontally uniform, and is computed by averaging smoothed DLM values at 400 km N/S/E/W of the current position of the TC at each point in the forecast. New storm positions are generated in one-hour steps based on the total steering propagation velocity vector. The effective radius and inflow angles are re-computed with each time step, since they will vary as beta changes with latitude. A linear time interpolation is made between the appropriate DLM fields to provide smoother variation in steering with time. A persistence feature is included, which weighs 12-hour past motion out to the 12-hour forecast position using a cosine weighting function.

(4) Medium Beta and Advection Model (MBAM)

This model is an adaptation of the FBAM model with the steering flow derived from the 850-, 700- and 500-hPa layers' weighted mean winds.

(5) Shallow Beta and Advection Model (SBAM)

This model is an adaptation of the FBAM model with the steering flow derived from the 850- and 700-hPa layer weighted mean winds.

(6) Japanese Typhoon Model (JTYM)

The Japanese Typhoon Model is a limited-area, grid-point model with the following specifications:

(1) 50 km horizontal resolution

(2) 109x109 grid points (approximately 5500x5500 km)

(3) Lambert conformal projection for storm latitude 20N; mercator projection otherwise

(4) 8 vertical (sigma) levels

(5) boundaries updated by Japanese Global Spectral Model (GSM) once a day.

(6) provides forecast positions only out to 60 hours.

(7) NOGAPS Steering Model (NSM)

The NOGAPS Steering Model (NSM) is being developed and tested at JTWC. It differs from FBAM in four aspects:

a) It uses only 500 and 700 hPa, unsmoothed windfields downloaded from FNMOC via TYMNET. The data are digital and on a 2.5 X 2.5 degree grid.

b) No propagation component is added.

c) No persistence component is added.

d) Forecasts based on single-level steering are available (NSM7 and NSM5) in addition to a pressure-weighted layer average forecast (NSML).

The average steering at any time is computed by averaging wind field values around an annulus approximately centered on the storm, and approximately 6.25 to 8.25 degrees in radius.

Two hybrid forecast aids used at JTWC are:

(1) Half-Persistence and Climatology (HPAC)

Forecast positions are generated by equally weighting the forecasts given by XTRP and CLIM. Forecast positions are computed from a direct interpolation between corresponding forecast positions of XTRP and CLIM.

(2) Combined Confidence Weighted Forecasts (CCWF)

An optimal blend of objective techniques, CCWF blends selected techniques (currently OTCM, CSUM and HPAC) by using the inverse of the covariance matrices computed from historical and real-time cross-track and along-track errors as the weighting function.

Two empirical forecast aids used at JTWC are:

(1) DVORAK An estimation of the TC's current and 24 hour motion is made from the interpretations of satellite imagery (Dvorak, 1984).

(2) Typhoon Acceleration Prediction Technique (TAPT)

This technique (Weir, 1982) utilizes upper-tropospheric and surface wind fields to estimate acceleration associated with the mid-latitude westerlies. It includes guidelines for the duration of acceleration, upper limits and probable paths of the cyclone.

2. NHC OBJECTIVE AIDS

The following, extracted from Sheets (1990), are brief reviews of the characteristics of objective aids used at NHC.

Extrapolation techniques are used for both track and intensity forecasting.

All climatological aids used at NHC are based on the historical data base in the Atlantic, which consists of historical tracks from 1886 to the present. The climatological and analog (HURRAN) aids at NHC are very similar to those from JTWC.

Three statistical objective aids used at NHC are:

(1) Climatology and Persistence (CLIPER) model

The CLIPER (Neumann, 1972) is a statistical model based on climatology and persistence. It consists of prediction equations that relate future zonal and meridional displacements of a TC to a set of predictors. These predictors include initial and previous 12-hour positions, initial an previous 12-hour storm motion vectors, day number of the year (from 1 to 365) and the estimated maximum surface wind.

The prediction equations are derived using linear regression. The developmental data consist of a set of best tracks of TCs in the Atlantic Ocean, Caribbean Sea and Gulf of Mexico for the period 1931-70. Only TCs of tropical storm intensity or greater were included.

(2) NHC83 model

NHC83 (Neumann 1988) is a statistical-dynamical model which uses the perfect prog method to derive statistical relationships between TC motion and geopotential height fields. The developmental dataset included deep-layer mean geopotential height analyses for the period 1962-81 and Atlantic TC tracks for the same period. Two separate regressions are performed for storms initially north or south of 25 N. The geopotential height grids are rotated so that the axes are along and perpendicular to the initial direction of motion of the TC. When NHC83 is run operationally, the geopotential heights are obtained from the National Meteorological Center's (NMC) 18-layer, 80-wave global spectral model (the aviation model). The final NHC83 forecast track statistically combines the predictions using the geopotential heights with a CLIPER-type prediction so that the model makes use of the initial motion estimate. The initial motion information is also used in the grid rotation.

Although no major changes were made to NHC83 during the verification period, the model which drives NHC83 has changed (Bonner, 1989). Prior to 1987, the 12-layer, 40-wave spectral model was used. Beginning in 1988, a correction to the geopotential height input was made to account for the differing biases in the 12- and 18-layer versions of the NMC spectral model.

(3) NHC90

NHC90 is a statistical-dynamical model that is essentially an update to the NHC-83 model.

The four dynamic aids used at NHC are:

(1) Sanders Barotropic (SANBAR) Model

The SANBAR (Sanders et al., 1975; Goldenberg et al., 1987) is a barotropic model that uses an equivalent barotropic vorticity equation to forecast a vertically averaged, pressure weighted, deep-layer wind field. The initial wind field includes the synoptic scale and the vortex scale, where the storm circulation is represented by an idealized axisymmetric vortex. The wind field is adjusted so that the initial motion of the vortex is approximately equal to the operational initial motion estimate.

The SANBAR model offers simplicity and efficiency in a dynamical model, and the barotropic assumption allows the initial motion vector to be included in a relatively straightforward manner. However, the skill of SANBAR is limited by the accuracy of the barotropic approximations and synoptic analysis.

(2) Moveable Fine Mesh (MFM) Model

The MFM (Hovermale and Livezey, 1977) is a multilevel baroclinic model. MFM covers a area about 3000 X 3000 km. The horizontal grid distance of MFM model is 60 km with 10 vertical levels. The MFM model domain moves in order to remain centered on the tropical cyclone center location. The model includes cumulus parameterization and boundary layer processes. Lateral boundary conditions are obtained from the aviation model, but were provided by the limited-area fine mesh (LFM) model prior to 1984. Before 1983, the storm circulation was determined from a simulation with an axisymmetric version of the model. The axisymmetric vortex was added to the large scale analysis to provide the model with initial conditions. Beginning in 1983, the vortex was determined from a three dimensional model run which included the variation of the Coriolis parameter with latitude. All of the MFM forecasts in the comparison included the three- dimensional vortex initialization. Unlike SANBAR or any of the other models discussed above, the MFM does not make use of the initial motion estimates in any way.

(3) Quasi-Lagrangian Model (QLM)

The QLM (Mather, 1988) is a multilevel baroclinic model which includes parameterizations of physical processes. Lateral boundary conditions for the QLM are obtained from the aviation model, similar to the MFM, although the model domain does not move to follow the storm. It is not necessary to move the QLM domain since it covers a larger area than the MFM (about 4400 X 4400 km compared with 3000 X 3000 km for the MFM). The QLM uses 18 vertical levels and has a horizontal resolution of 40 km. The tropical cyclone circulation is represented by an idealized vortex in gradient balance which is merged with a large scale analysis. The QLM does not make use of the initial motion estimate.

(4) Beta and Advection Model (BAM)

The BAM model (Marks, 1989) is a barotropic vorticity model, although it uses vertically averaged horizontal winds which were predicted from the NMC global spectral model. The basic idea of the BAM is that the tropical cyclone motion is determined by absolute vorticity conservation principle.

Appendix B References

Chapter 5

Pfad - The Proxy pFad of © 2024 Garber Painting. All rights reserved.

Note: This service is not intended for secure transactions such as banking, social media, email, or purchasing. Use at your own risk. We assume no liability whatsoever for broken pages.

Alternative Proxies:

Alternative Proxy

pFad Proxy

pFad v3 Proxy

pFad v4 Proxy