Table 4. Table displaying operational models at use at NHC and a description. Timeliness is given as ‘E’ for early and ‘L’ for late. & Canadian GEM model uses a 4D-VAR schemes that makes use of in situ flight data
Increasing the number of observations in the often data-sparse TC environment has resulted in improvements to the analysis and forecasting of TCs. The NOAA’s Hurricane Research Division (HRD) flew a large number of “synoptic flow” missions near and around TCs in the Atlantic between 1982 and 1996, and these data resulted in significant reductions in forecast errors (Burpee et al. 1996). These research flights led to the procurement of the NOAA G-IV aircraft, which now conducts operational “synoptic surveillance” missions for TCs threatening land areas over the western part of the Atlantic basin (Aberson 2002). Figure 39 illustrates a typical flight path in the near-storm environment of Hurricane Dolly in 2008. Note that in this case G-IV flight path is designed to gather data ahead and on the periphery of Dolly, augmenting the routine radiosonde data from sites in the southern United States. The impact of these data gathered relatively close to the TC is largest in the first 48 hours of the forecast and decrease at later times (Fig. 40).
Figure 40. Average track errors (km) for a homogeneous sample of GFS model runs and a homogeneous sample of GFDL model runs (ALL) and without (NO) surveillance data initialized only a mission nominal times, and improvements over the former and not the latter (%). Sample sizes for each model at each forecast time are provided above the graph. Reproduced from Aberson (2010).
While these data largely improve the forecast, Aberson (2008) identified several cases during the 2004 and 2005 Atlantic hurricane seasons, where the inclusion of the dropsonde data actually degraded model performance. The degradation in skill stemmed from several bad observations which were accepted by the data assimilation system that normally inspects and then excludes uncertain or poor data (Aberson 2008).
Figure 39. Flight path for synoptic surveillance mission at 0000 UTC 22 July for Hurricane Dolly. Solid burgundy dots indicate the flight path of the NOAA G-IV aircraft around Dolly, starting at observation 1 and ending at observation 31. The pink triangle indicates the mission where the flight originated.
IV. Verification Methods
There are several motivating factors for developing a data base of verification statistics. Studying forecast errors can help forecasters and modelers understand the errors and identify biases in model guidance that can be targeted for improvement. Verification results can also aid in the development of new forecast products or tools. For example, wind speed and storm surge probabilities issued by the NHC are generated through random sampling of a data base of track and intensity errors from official NHC forecasts (DeMaria 2009). The evaluation of forecast error also provides a foundation from which forecast uncertainty can be estimated, a critical factor for determining the extent of watches and warnings providing estimates for the time of arrival of tropical storm or hurricane conditions. Finally, a reliable set of verification statistics can enable decision-makers to make better use of existing products and understand their strengths and limitations.
All global TC warning centers issue an “official” forecast of position and intensity for all operationally-designated TCs. Forecasts are issued every 6 hours at the standard synoptic times of 0000, 0600, 1200, and 1800 UTC for 12, 24, 36, 48, 72, 96, and 120 hours (some global centers only issue forecasts out to three days). At the conclusion of the season forecasters should review operational best track positions with an aggregate of data from a variety of sources to develop a final “best track.” To be included in the verification database, NHC requires that the system should be a TC at both the forecast’s initial and verification times; no other stages of development are considered.
It is important to differentiate between forecast error and forecast skill. Track forecast error is defined as the great-circle distance between a TC’s forecast position and its post-storm best track position at that time. Forecast skill represents a normalization of forecast error against some common metric or baseline, with forecast skill being positive (negative) when the forecast error is less (more) than the associated baseline error. The most common baseline against which forecast skill can be evaluated is a model based solely on climatology and persistence, such as CLIPER5. Since CLIPER5 contains no information whatsoever on the TC’s current environment, it represents a “no-skill” level of accuracy to which other models can be compared. Thus, model guidance which consistently outperforms CLIPER5 is said to be skillful, while guidance that performs worse than CLIPER5 is unskillful. It is worth noting that the baseline for forecast skill varies from season to season. In some seasons, CLIPER5 errors may be small, indicating that tracks were inherently “easier” to predict and vice versa.
Construction of a Best Track
To construct the “best track” of a TC one should first begin with the operational track analysis completed and then satisfy the basic components of an accepted “best track” definition uch as the one currently in place at the NHC:
“A subjectively-smoothed representation of a TC’s location and intensity over its lifetime. The best track contains the cyclone’s latitude, longitude, maximum sustained surface wind, and minimum sea-level pressure at 6-hourly intervals. Best track positions and intensities, which are based on a post-storm assessment of all available data, may differ from values contained in storm advisories. They also generally will not reflect the erratic motion implied by connecting individual center fixes.”
A fundamental part of the “best track” definition is “6-hourly representative estimates of the cyclone’s center location.” In post-analysis the forecaster can optimally re-construct the TC track with the benefit of inspecting all available data. This is not always possible in real-time, as the “working best track” is developed under tight time constraints, when some data may be unavailable.
A major consideration is the degree of detail to furnish when preparing the “best track.” Plotted raw fix data often reveal a series of irregular movements, which do not generally persist for more than a few hours (previously discussed in the short-term motion section). Short-term track wobbles or bends are also not representative of the overall TC motion, and therefore a subjectively-smoothed “best track” that does not focus on short period, transient motions is ideal. Some forecasters have interpreted this to mean that, if a TC is located at a given position at 06 UTC, the “best track” need not lie exactly at that spot in post-analysis. From a sampling perspective, the re-positioning that results is part of a filtering procedure that could be administered to avoid aliasing small-scale noise. Small-scale oscillations are not indicative of mean TC motion. For a time series with data points ΔT apart, the smallest wavelength which can be depicted accurately is about 4*ΔT (reference?). Since the typical advisory analysis times are 6 hours apart, the smallest periods which can be adequately represented are on the order of 24 hours. Thus the analyst should try to avoid analyzing oscillations with a period less than 24 hours.
Therefore the best track estimate often differs a little from the advisory estimate because the former is meant to represent a particular time, while the latter is meant to be valid at a particular time.
Estimating Forecast Error
Track error can be decomposed into along- and cross-track components to obtain additional information related to forecast track. Along-track error is the component parallel to the forecast track, and the component of the track error perpendicular to the forecast track is the cross-track error. The along-track error indicates forecast timing issues, while the cross-track error is indicative of error in forecasting direction. Figure 41 depicts along- and cross-track errors for a westward-moving TC over the Gulf of Mexico where the cross-track error shows that the official track was biased toward the right or north of the observed track.
Figure 41. Illustration of the definition of along- and cross-track error. The actual or best-track of the TC appears in boldface black, and the official forecast track is in dashed blue. The along- and cross-track errors are vectors, the along-track given by the purple arrow and the cross-track given by the red arrow.
Figure 42 shows a comparison of along- and cross- track errors from the Atlantic basin from 2004 to 2008. There is a close correspondence between the two error components through around 48 hours, suggesting that both timing and direction errors are similar in the short range. However, at longer ranges in the Atlantic basin, the along-track error grows larger relative to the cross-track error, indicating that it is more difficult to forecast the timing or speed of motion than the direction of motion. This analysis could be performed for other basins to identify systematic errors. Along- and cross-track errors can also be examined regionally in an effort to localize characteristic track forecasting biases in different parts of a basin.
Figure 42. NHC official track errors for 2004-08. The red curve represents along-track error, and the purple curve represents the cross-track error.
Summary and Conclusions
This chapter describes current methods for estimating TC position and motion and forecasting TC track. The techniques described to estimate TC location and motion primarily rely on geostationary satellite data. The recent proliferation of the number of polar-orbiting satellites with microwave sensors and scatterometers has also provided more resources to analyze TC location, particularly when the TC is not easily observed in geostationary imagery. These tools are supplementary to in situ observations provided by aerial reconnaissance aircraft when TCs threaten land areas (in the Atlantic and occasionally the eastern North Pacific). However, TC center fix location should be determined using data from all available observational platforms.
Improvements to TC track prediction have accelerated as numerical model resolution, computational resources and data sources increase, and data assimilation methods become more sophisticated. In an effort to extract as much forecast skill from the models as possible, multi-model consensus methods have become the most skillful predictors of TC track.
Future work in numerical modeling is expected to lead to a continuation of improvements in TC track prediction. Further advancements will be gained by applied research, improved observations through existing platforms, new observing technologies, increases in computer power and improvements to dynamical models and data assimilation schemes. These enhancements should largely come about as a result of increasing model resolution and computational efficiencies, faster and more accurate assimilation methods, the proliferation of vast amounts of observational data, including further strides in ensemble prediction. Utilizing numerical model and statistical forecasts together optimally should also help forecasters to continue to reduce TC track prediction error.
The authors are grateful for the comments and other support provided by Chip Guard, Jeff Hawkins, Jim Goerss, Eric Blake, James Franklin, Jack Beven, Richard Pasch, Dan Brown, Robbie Berg, John Cangialosi, and Chris Landsea. The authors are especially appreciative of the advice and other useful comments submitted by David P. Roberts regarding the Microwave section, by Roger Edson regarding the Microwave and QuikSCAT sections, and Jim Goerss concerning the model consensus section. We also are in debt to Colin McAdie who helped developed the radar section. Special recognition is also given to UCAR COMET for allowing the publication of several figures from COMET modules and to the Naval Research Laboratory in Monterrey, CA, for the use of archived microwave imagery for several TCs. We also thank Joan David who assisted in the modification of several of the figure images for publication.
Aberson, S.D., 1998: Five-day TC track forecasts in the north Atlantic basin, Wea. Forecasting, 13, 1005-1015.
Aberson, S.D., 2010: Ten years of hurricane synoptic surveillance. Mon. Wea. Rev., xxx, xxx-
Aberson, S.D., 2002: Two years of Operational Hurricane Synoptic Surveillance. Wea. Forecasting, 17, 1101-1110.
Aberson, S.D., 2008: Large forecast degradations due to synoptic surveillance during the 2004 and 2005 hurricane seasons, Mon. Wea. Rev., 136, 3138-3150.
Barrett, Bradford S., Lance M. Leslie, and Brian H. Fiedler, 2005: An example of the value of strong climatological signals in TC track forecasting: Hurricane Ivan (2004), Mon. Wea. Rev., 134, 1568-1577.
Bender, M.A., R.E. Tuleya, and Y. Kurihara, 1987: A numerical study of the effect of island terrain on tropical cyclones. Mon. Wea. Rev., 115, 130-155.
Bender, M.A, Ginis, I., B. R. Tuleya, B. Thomas, and T. Marchok, 2007: The operational GFDL coupled-hurricane ocean prediction system and a summary of its performance, Mon. Wea. Rev., 135, 3965–3989.
Bessafi, M., A. Lasserre-Bigorry, C.J. Neumann, F. Pignolet-Tardan, D. Payet, and M. Lee-Ching-Ken: Statistical prediction of TC motion: An analog-CLIPER approach, Wea. Forecasting., 17, 821-831.
Burpee, R.W., J.L. Franklin, S. J. Lord, R.E. Tuleya, and S.D. Aberson, 1996: The impact of Omega dropwindsondes on operational hurricane track forecast models. Bull. Amer. Soc., 77, 925-933.
Brand, S, 1970: Interaction of binary TCs in the western Pacific ocean, J. Appl. Meteor., 9, 433-431.
Brand, S., and J. Blelloch, 1974: Changes in the characteristics of typhoons crossing the island of Taiwan. Mon. Wea. Rev.,102, 708-713.
Brennan, M.J., C.C. Hennon, and R.D. Knabb, 2009: The operational use of QuikSCAT ocean surface vector winds at the National Hurricane Center. Wea. Forecasting, 24, 621-645.
Carr, L.E., III and R.L. Elsberry, 1997: Models of tropical cyclone wind distribution and Beta-effect propagation for application to TC track forecasting, Mon. Wea. Rev., 125, 3190-3208.
Carr, L.E., III and R.L. Elsberry, 1998: Objective diagnosis of binary TC interactions for the western North Pacific basin, Mon. Wea. Rev., 126, 1734-1740.
Carr, L.E., III, R.L. Elsberry, and J.E. Peak, 2001: Beta test of the systematic approach expert system prototype as a TC forecasting aid. Wea. Forecasting, 16, 355-368.
Chan, J.C.L., F.M.F. Ko, and Y.M. Lei, 2002: Relationship between potential vorticity tendency and TC motion, J. of Atmos. Sci., xx, 1317-1336.
Chang, S. W., 1982: The orographic effects induced by an island mountain range on propagating tropical cyclones. Mon. Wea. Rev.,110, 1255-1270.
DeMaria, M., J.A. Knaff, R. Knabb, C. Lauer, C.R. Sampson, and R.T. DeMaria, 2009: A new method for estimating tropical cyclone wind speed probabilities, Wea. Forecasting, in press.
Dong, K. and C.J. Neuman, 1983: On the relative motion of binary TCs, Mon. Wea. Rev., 111, 945-953.
Dunn, G.E., and B.I. Miller, 1964: Atlantic hurricanes. Louisiana State Press, pp. 326.
Dvorak, V.F., 1984: TC intensity analysis using satellite data. NOAA Tech. Report NESDIS 11. Available from NOAA/NESDIS, 5200 Auth Rd., Washington, D.C., 20233, 47 pp.
Edson, R.T., xxxx: Normalized radar cross-section (NRCS) patterns from QuikSCAT-A new analysis tool over tropical oceans. Proceedings of the xx Conference on Hurricanes and Tropical Meteorology
Ellrod, G. P., 1995: Advances in the detection and analysis of fog at night using GOES multispectral infrared imagery. Wea. Forecasting, 10, 606-619
Elsberry, Russell L., and Les E. Carr III, 2000: Consensus of dynamical TC track forecasts – errors versus spread. Mon. Wea. Rev., 128, 4131-4138.
Elsberry, R.L, 2005: Achievement of USWRP Hurricane Landfall Research Goal, Bull. of Amer. Met. Soc., 86, 643-645.
Franklin, J.L., cited 2008: 2008 National Hurricane Center Forecast Verification Report. [Available online at http://www.nhc.noaa.gov/verification/pdfs/Verification_2008.pdf.]
Fujiwhara, S., 1921: The natural tendency toward symmetry, etc., Quart. J. Roy. Meteor. Soc., 47, 287-293.
----------------, 1923: On the growth and decay of vortical systems, Quart. J. Roy. Meteor. Soc, 49, 75-104.
----------------, 1931: Short note on the behavior of two vortices. Proc. Phys. Math. Soc. Japan, Ser. 3, 13, 106-110.
George, J.E. and Gray, W.M.: TC motion and surrounding flow relationships, J. Appl. Meteor., 15, 1252-1264.
Goerss, J.S., 2000: TC track forecasts using an ensemble of dynamical models. Mon. Wea. Rev., 128, 1187-1193.
Holland, G.J. and G.S. Dietachmayer, 1993: On the interaction of tropical-cyclone-scale vortices. III. Continuous barotropic vortices. Q. J. Roy. Meteor. Soc., 119, 1381-1398.
Holland, G.J., and M. Lander, 1993: The meandering nature of TC tracks, J. Atmos. Sci., 50, 1254-1266.
Hope, J.R., and C.J. Neumann, 1970: An operational technique for relating the movement of existing TCs to past tracks. Mon. Wea. Rev., 98, 925-933.
Lander, M.A. and G.J. Holland, 1993: On the interaction of tropical-cyclone scale vortices. I: Observations, Quart. J. Roy. Meteor. Soc., 119, 1347-1361.
Lander, M.A., 1995: The merger of two TCs, Mon. Wea. Rev., 123, 2260-2265.
Lawrence, M.B., and B.M. Mayfield, 1977: Satellite observations of trochoidal motion in hurricane Belle 1976. Mon. Wea. Rev., 105, 1458-1461.
Lin, Y.L, N.C. Witcraft, and Y.H. Kuo, 2006: Dynamics of track deflection associated with the passage of TCs over a meso-scale mountain. Mon. Wea. Rev., 134, 3509-3538.
Neumann, C.J., 1983: The National Hurricane Center NHC83 Model. NOAA Technical Mem. NWS NHC 41, pp.44.
Prieto, R., B.D. McNoldy, S.R. Fulton, and W.H. Schubert: A classification of binary TC-like vortex interactions, Mon. Wea. Rev., 131, 2656-2666.
Rappaport, E.N., J.L. Franklin, L.A. Avila, S.R. Baig, J.L. Beven. E.S. Blake, C.A. Burr, Jiann-Gwo Jiing, C.A. Juckins, R.D. Knabb, C.W. Landsea, M. McInerney-Mainelli, B.M. Mayfield, C.J. McAdie, R.J. Pasch, C. Sisko, S.R. Stewart, and A.N. Tribble, 2008. Advances and challenges at the National Hurricane Center, Wea. Forecasting, 24, 395-419.
Ritchie, E.A., and R.L. Elsberry, 1993: Simulated impacts of a mesoscale convective system on the track of Typhoon Robyn in TCM-93. Mon. Wea. Rev., 128, 2232-2251.
Ritchie, E.A., and G.J. Holland, 1993: On the interaction of two TC scale vortices. II: Discrete vortex patches. Quart. J. Roy. Meteor. Soc., 119, 1363-1379.
Ross, R.J., and Y. Kurihara, 1995: A numerical study on influences of Hurricane Glorida (1985) on the environment. Mon. Wea. Rev., 123, 332-346.
Sampson, C.R., J.S. Goerss, and H.C. Weber, 2006: Operational performance of a new barotropic model (WBAR) in the Western North Pacific Basin. Wea. Forecasting, 21, 656-662.
Sampson, C.R., J.A. Knaff, and E.M. Fukada, 2007: Operational evaluation of a selective consensus in the western North Pacific basin. Wea. Forecasting, 22, 671-675.
Senn, H.V, and H.W. Hiser, 1959: On the origin of hurricane spiral rain bands. J. Atmos. Sci., 16, 419-426.
Velden, C.S., and L.M. Leslie, 1991: The basic relationship between TC intensity and the depth of the environmental steering layer in the Australian region. Wea. Forecasting, 6, 244-253.
Williford, C.E., T.N. Krishnamurti, R.C. Torres, S. Cocke, Z. Christidis, and T.S. Vijaya Kumar, 2003: Real-time multimodel superensemble forecast of Atlantic tropical systems in 1999, Mon. Wea. Rev., 131, 1878-1894.
Willoughby, H.E., 1988: Linear motion of a shallow-water barotropic vortex. J. Atmos. Sci., 45, 1906-1928.
---------------------, 1990: Temporal changes of the primary circulation in TCs. J. Atmos. Sci., 47, 242-264.
---------------, F.D. Marks, and R.J. Feinberg, 1984: Stationary and moving convective bands in hurricanes. J. Atmos. Sci., 41, 3189-3211.
Wong, M.L. and J.C.L Chan, 2006: Tropical cyclone motion in response to land surface friction. J. Atmos. Sci., 63, 1324-1337.
Yeh, T.C., and R.L. Elsberry, 1993: Interaction of typhoons with the Taiwan orography. Part I: Upstream track deflections. Mon. Wea. Rev., 121, 3193-3212.