Guide to Tropical Cyclone Forecasting: Tropical Cyclone Motion

NASA’s SeaWinds instrument onboard the QuikSCAT satellite was an active microwave, pencil-beam scatterometer launched in 1999 that failed in late 2009. QuikSCAT estimated near surface (10 m) wind speed and direction over the ocean surface by measuring the backscatter variations from small-scale roughness elements on the ocean surface. The satellite had an 1800-km wide swath and covered about 90% of the global oceans on a daily basis. Compared to the core of a TC, the data were somewhat coarse, having a standard 25-km and a higher 12.5- km product generated by post-processing techniques (Fig. 12-13). Despite the failure of QuikSCAT in 2009, a discussion of its capabilities will be provided since there is a possibility that additional data from recently launched scatterometers similar to QuikSCAT will become available for TC analysis in the coming years.

Figure 12. A 12.5-km QuikSCAT pass of Hurricane Lisa at 2116 UTC on 28 September 2004. A color-coded legend at the bottom of the image indicates the wind speed in knots and agrees with the vector wind barb.

Figure 13. 12.5-km QuikSCAT pass over Hurricane Helene at 0917 UTC on 19 September 2006. A color-coded legend in the upper right indicates the wind speed in knots, and the vector wind barbs are in agreement. Rain-flagged wind barbs are in black.

QuikSCAT often provided a snapshot of the entire TC circulation, which is more complete than even multiple passes of a single reconnaissance aircraft throughout the storm. However, QuikSCAT was limited to two passes (at most) daily over any particular TC. Moreover, some passes result in misses or only partial hits on a TC due to the relatively wide gaps (approaching 1000 km in the deep Tropics) between data swaths. Apart from microwave or conventional satellite data and aerial reconnaissance, QuikSCAT was the most frequently-used data source in operations for TC center identification and location (Brennan et al. 2009).

There were two main limitations of QuikSCAT in estimating TC location. First, the Ku-band signal of QuikSCAT was degraded by the high rainfall rates in TCs, which caused errors in the wind solution. Secondly, there was directional ambiguity associated with the QuikSCAT retrievals, resulting in up to four possible wind direction solutions at each point in the QuikSCAT swath. The automated QuikSCAT solution was selected using a median filter technique that uses a short-term model forecast as a first “guess” and enforces spatial consistency in the chosen wind field (Hoffman and Leidner 2005). In the TC environment the combination of these factors resulted in errors in the ambiguity removal process that often made the automated wind solution unreliable for most TC analysis applications (Brennan et al. 2009).

As a result of these limitations, performing a streamline analysis of all possible QuikSCAT wind solutions (i.e., an “ambiguity analysis”) was the preferred way to identify the circulation center. The forecaster should begin by exploring QuikSCAT ambiguity data in a region where the wind direction was known from other measurements (e.g., on the periphery of the TC circulation). From this starting point, choose “possible solutions” that move in toward the suspected TC center location, remembering that QuikSCAT ambiguities pointed in the direction that the wind is blowing toward. Priority should be given to the “two-“ of “three-” way ambiguities first, since there is less uncertainty in the wind direction at these locations; consult “four-”way ambiguities lastly, as the uncertainty may be the highest there. A good strategy is to try to establish north-south or east-west tangent lines to help estimate the latitude and longitude of the TC center (Fig. 14).

Figure 14. An ambiguity retrieval plot for Tropical Storm Helene from 0900 UTC 14 September 2006 with the procedure for drawing streamlines annotated in several notes sections.

The European Space Agency’s Advanced Scatterometer (ASCAT) is an operational scatterometer that, with similarities to QuikSCAT, is currently flying on the METOP satellite series (three missions planned through 2020) and is an additional source for satellite wind retrievals over TCs.

ASCAT has two 550-km wide swaths and a 720-km nadir gap. Figure 16 below shows the typical coverage of ASCAT data globally relative to QuikSCAT and highlights the large gaps over the Tropics. Additionally, the resolution of ASCAT wind retrievals is only half as fine as that of QuikSCAT, with retrievals available at 50- and 25-km resolution, and the retrievals are noticeably smoother than those of QuikSCAT. One advantage of ASCAT, however, is that its C-band retrievals are less sensitive to rain, which can result in improved retrieval quality in the typically rainy TC areas, particularly in weaker TCs.

Figure 16. Image showing typical, daily coverage of ASCAT (top) and QuikSCAT (bottom). Note the wider QuikSCAT swaths compared to ASCAT with large gaps in coverage over the global Tropics.

Aerial weather reconnaissance has been vitally important in determining the center location of TCs since 1944. In the Atlantic basin specially modified and equipped aircraft of the U.S. Air Force Reserve (AFRES) and the National Oceanic and Atmospheric Administration's Aircraft Operations Center (NOAA/AOC) are used to investigate TCs.

A typical mission consists of the plane generally flying at 10,000 ft (700 hPa) for hurricanes and 5,000 ft (850 hPa) for tropical storms. In disturbances which have not become a TC (i.e., “invest” missions), the typical flight altitude is 1,500 ft. The plane flies a “Figure Four” or “alpha pattern” about the storm (Fig. 17). Missions can last from 10 to 12 hours, with two to six center “fixes” possible (or none if the crew finds no center). Aircraft observations from within the TC are generally restricted to locations along and near the flight path.

Figure 17. Illustration of the standard “Figure Four” or “Alpha” pattern flown by aerial reconnaissance aircraft into TCs. This particular image is the flight track from the mission into Hurricane Georges from 28 September 1998 between 2000 and 2300 UTC.

There are often times when the flight-level center is not coincident with the surface center, as in a sheared vortex. To determine whether this is the case, one should look within the “Remarks” section of the Vortex message to discover whether the meteorologist on the aircraft noted any difference between the two centers. Figure 18 shows a Vortex message from Atlantic Tropical Storm Ernesto in 2006, where the meteorologist on board noted that the surface and flight-level centers were not coincident. In fact, the surface center is displaced 21 nm south of the flight-level center, implying a moderate southerly or south-southeasterly shear over the cyclone.

Figure 18. Vortex message from Tropical Storm Ernesto (2006). In the “Remarks” section the on-board meteorologist notes that the surface center is displaced 21 nm south of the flight-level center.

Proper interpretation of the Vortex message requires careful attention to detail. For example, the center fix listed on the Vortex message does not necessarily designate the flight-level center, but simply reflects the location of the wind speed minimum or wind shift encountered along the flight path. Examination of both the individual HD-obs (high-density observations with a time interval as short as 30 seconds) will help determine whether the flight-level center and reported center fix are coincident. Center fixes can also jump around in poorly organized or weak systems. Position estimates from Tropical Storm Frances (1998) at times bounced around 50 n mi from fix to fix, as the cyclone had a large, non-circular center.

Aircraft reconnaissance data improve track forecasts. Figure 19 shows a plot of track forecast errors with aircraft data and without aircraft data for the period 1989-2002. The plot shows a nearly monotonic

Figure 19. NHC official forecast accuracy with and without reconnaissance data.

degradation in forecast skill out to 72 hours in both the “aircraft” and “non-aircraft” populations. The percentage errors due to the unavailability of aircraft data are largest near the beginning of the forecast period and slowly diminish at later times. This makes intuitive sense, since a majority of track forecast errors in the first few hours are primarily dependent upon accurately locating the center. Track forecast errors later in the forecast period are more influenced by the large-scale steering pattern.

Radar Fixes

Radar is an acronym that stands for Radio Detection and Ranging, a technique where a beam of pulsed electromagnetic energy in the microwave wavelength range is transmitted outward from a (typically) rotating antenna. Backscattered energy from hydrometeors and other atmospheric scatterers is collected by the same antenna and displayed either as a horizontal Plan Position Indicator (PPI) plot, or a vertical Range Height Indicator (RHI) slice. Some radar display systems also have the capacity to integrate data collected from scans at different beam elevations.

Doppler radar can also measure the change in frequency between the original and backscattered beam, which can provide information on whether backscattering particles are moving towards or away from the radar (i.e., radial velocity). Two Doppler radars in proximity can provide full three-dimensional winds in regions where their beams overlap at an appropriate angle. An airborne Doppler radar can also obtain a full wind field by sampling the same volume from different locations. Due to refraction of the radar beam, and the Earth’s curvature, considerable resolution and information are lost as the distance from the radar increases (Fig. 20). This is demonstrated in the radar image of Atlantic Hurricane Lenny in Figure 20, where no deep convection on the southwest side of the eyewall. While it is possible that little to no convection exists on this side of the storm, the most likely explanation is that the radar beam has “climbed over” the deepest convection.

Figure 20) Hurricane Lenny (0212 UTC 17 Nov 1999) south of Puerto Rico.  The effect of the increased height of the beam above the surface with increasing range is demonstrated.   The height of the center of the radar beam above the surface is indicated.

As a TC approaches a radar, the first evidence of its arrival is usually outer rainbands, some of which move ahead of and at roughly the same speed and direction as the cyclone. However, determining an accurate center location by radar is not possible at this stage. Once significant lengths of spiral band are observed, fitting logarithmic curves with a constant crossing angle of 10-20° can provide an initial indication of the cyclone center (Senn and Heiser 1959).

Surface observations are generally not dense enough to accurately track a TC center, especially over the open oceans. When a variety of observational platforms yields an ambiguous analysis, however, surface observations may become more important. Occasionally an unexpected ship report provides the only data report in a data-sparse area, confirming the presence or location of a TC center. The wind speed and direction from surface observations give some indication of how far from the observing site the and in what direction the TC center lies, and the pressure tendency helps determine whether the TC is moving toward or away from the site.

A time series of surface observations updated with regular frequency is even more valuable than a single observation. Veering or backing of the wind direction with time indicates whether a center is approaching or moving away from a given location. Unfortunately, most ship reports are made only at 3- or 6-hourly times. Reports from first-order land sites and mesonets, C-MAN stations, and buoys report hourly, but some stations have data interruptions due to power outages, communication failures, and even station destruction in more severe TCs. Also, the representativeness of surface observations should be considered to ensure that transient features within the TC circulation have not contaminated the data.

In recent years there has been an expansion of land and ocean observations from a variety of new sources. For example, there has been a rapid increase in the number of coastal mesonets and other buoy networks over different parts of the globe. Figure 23 shows the relatively high density of observations in the form of C-MAN stations, and an array of other coastal automated stations in the vicinity of the Florida peninsula, most of them deployed through NOAA’s National Data Buoy Center (NDBC) and National Ocean Services (NOS). In addition, several university research groups coordinate the deployment of observing networks in the path of a TC to supplement the relatively few observations available at a TC landfall. Private companies such as “Weather Underground,” “Weather Bug,” and “Weatherflow” also provide extended data coverage mostly over land areas.

Figure 23. Map showing a subset of the large number of coastal observations in the vicinity of the Florida peninsula in the southern United States.

Fundamentally, TC motion is governed by two mechanisms: advection by the environmental flow of the relative vorticity associated with the TC ,and by advective processes that involve interactions on different scales with the environmental flow, the planetary vorticity gradient, and the TC vortex. The first mechanism can be thought of as TC being analogous to a “block of wood” in a “river of air” or a “cork running through a stream”. For this mechanism, the layer-averaged winds over the depth of the troposphere correlate well with the actual storm motion. A forecaster can develop a simple estimate of the large-scale motion, which could serve as a proxy for the actual TC motion. The calculation for this parameter is known as the deep-layer mean (DLM) and is simply the average wind in discrete layers from the lower- to the upper-levels of the troposphere, weighted by pressure or thickness. If the vertical layer extends from 850 hPa to 200 hPa, the vertical integration can be approximated by the following expression:

650 mb

A direct relationship exists between the intensity and the vertical depth of the TC vortex in the troposphere (Velden et al. 1993). As a result, this relationship implies that a forecaster should consider the intensity of a TC before determining the appropriate steering layer to consider. Figure 26 indicates that stronger TCs typically extend through the depth of the troposphere and are steered by a DLM wind. Weaker storms typically have less vertical depth and a shallower layer of average winds should be examined to assess the potential TC motion. In addition, the forecaster should compare the vertical depth of the vortex in model output to ensure that the model has the proper representation of the TC. Figure 24 shows the forecast scenario at 0600 UTC 5 September 2008 when Hurricane Ike was centered well east of Florida. A majority of the track guidance at this time indicated that Ike would turn from a west-southwest course north of Puerto Rico and Hispañiola to a west-northwest or even northwest course later in the

Figure 24. Plot of model track guidance for Hurricane Ike at 0600 UTC 5 September 2008. The solid white line with hurricane symbols represents the best track, while the multi-colored lines represent individual track model forecasts out to 3 days; multi-colored, dashed lines represent individual track model forecasts from 3 to 5 days. The cyan track is the official forecast, the black line is the GFS model, and the orange line is the ECMWF model.

forecast period. A series of cross-sections from the GFS, one of the models forecasting a more northwesterly course beyond 72 hours, appears on the top part of Figure 25. These cross-sections give some indication of the vertical depth and organization of the TC vortex within the model as well as the DLM flow. In the case of the GFS, the model fails to initialize and later develop a mature TC vortex as evidenced by insufficient depth of the circulation above about 500 hPa, even though Ike was a well-defined hurricane at the initial time. Many of the other

Figure 25. Forecast east-to-west model cross-sections for Hurricane Ike. (a) and (b) are a series of GFS and ECMWF analyses and forecasts through 48 h of vorticity and horizontal winds (every 6 hours) associated with Hurricane Ike. A horizontal distance across the TC is on the x-axis, while pressure (in mb) is indicated on the y-axis. Vorticity is contoured in units of m²s^-1, with the red and purple shadings indicating positive vorticity and blue shadings indicating negative vorticity. Forecast wind barbs are also displayed in knots.

models also forecasting a northwesterly course suffered from poor model representations of the TC vortex initially and through the forecast period. However, several cross-sections from the ECMWF model appear in the lower half of Figure 25 and reveal a deeper TC vortex that is consistent with a mature hurricane. The ECMWF model, with its more accurate structural depiction of Ike, consistently forecast the hurricane to remain farther south and had a better verifying forecast. This result suggests that Ike was being steered by a deep layer mean (DLM) flow from the northeast or east around a strong subtropical high to the north. The GFS, however, steered its weaker representation of Ike in shallower flow around the low- to mid-level Atlantic ridge.

The Beta and Advection Models (BAMs) are simple, two-dimensional trajectory models, which contain vertical layer-averaged horizontal winds in addition to a “beta” correction term. The “beta” correction term is a result of differential vertical advection of planetary vorticity in response to the TC circulation and represents a self-advection. The differential advection induces weak secondary circulations known as “beta gyres” on either side of the TC, which cause a net northwestward to north-northwestward drift of about 1-2 ms^-1. The beta effect propagation is dependent on the outer wind structure of the TC, a part of the storm that is poorly sampled by aerial reconnaissance aircraft (Carr and Elsberry 1997). Apart from the beta drift, the BAM models correspond to different layer means and can be used to assess likely storm motion in different situations: BAM shallow (850-700 hPa), BAM medium (850-400 hPa), and BAM deep (850-200 hPa), known as BAMS, BAMM and BAMD, respectively. For example, for a weak or severely sheared system the BAMS output, would be a better indicator of TC motion than BAMD, when the environmental flow is simple and baroclinic forcing is weak.

Figure 26. Graphic illustrating the relationship between TC intensity and TC environmental steering. From Velden et al. (1993).