Guide to Tropical Cyclone Forecasting: Tropical Cyclone Motion



Download 197.74 Kb.
Page1/5
Date conversion19.10.2016
Size197.74 Kb.
  1   2   3   4   5


Global Guide to Tropical Cyclone Forecasting: Tropical Cyclone Motion

Todd B. Kimberlain and Michael J. Brennan



National Hurricane Center

NOAA/NWS/NCEP, Miami, FL

Draft: 12 January 2010

Corresponding author address: Todd B. Kimberlain, National Hurricane Center,

11691 SW 17th Street, Miami, FL 33165. E-mail: Todd.Kimberlain@noaa.gov



Introduction

Prediction of tropical cyclone (TC) track remains a challenging problem. New tools to observe TCs have emerged in recent years to facilitate the task of locating the TC center and to estimate TC motion. The advent of new technology has coincided with remarkable progress in the field of numerical weather prediction, which have helped reduce typical TC track errors in the Atlantic basin at all lead times by half since 1990 (Rappaport et al. 2009). Part of the success is due to a tremendous increase in remote sensing data from satellites, which are particularly valuable in data-sparse regions such as the Tropics. This increase in observations has allowed data assimilation schemes to greatly improve the analysis of the atmosphere, while model initialization schemes for the TC have also progressed. The development of vortex relocation schemes, which have begun to replace the more archaic vortex bogussing procedures, has also likely contributed to the improvements in track forecasting. Due to this progress, reliance on statistical or simple dynamical models has decreased considerably since the performance of the regional and global dynamical models has far surpassed them.

This chapter describes contemporary approaches to analyzing TC position and forecasting TC track. The first part focuses on position analysis and the various observational platforms used in this task. A discussion of numerical model guidance applications and TC forecasting techniques follows in the second part with a particular emphasis on “consensus” forecasting. The chapter concludes with a description of verification methods and their practical uses in an operational setting.

Position Analysis

Keeping an Operational Best Track

In an operational environment observations arrive from a wide array of platforms at different times. These observations contain analysis errors, vary in spatial and temporal coverage, and are often not representative of the true TC motion. This makes the determination of TC position and motion difficult in real time. For example, Dvorak classifications (Dvorak 1984), remote sensing and radar “fixes,” and surface observations may be unevenly available and are only occasionally supplemented with aerial reconnaissance fixes (primarily near land areas in the Atlantic basin). This can result in contradictory data, which must be carefully considered to determine the current TC motion.

Locating the center of the TC is the first step in the preparation of a track forecast, as this will help establish both the short- and long-term motion of the TC. The motion estimate is also used to initialize some computer models and can help in diagnosing the interaction of the TC with the surrounding environment. A center fix can be estimated using geostationary or microwave satellite data, radar data, and surface observations. In portions of the North Atlantic basin, observations from aerial reconnaissance are also available.

A TC position estimate can be compared to an estimate based on the average motion for the last 6, 12, or 18 hours. The time period for the average motion calculation will vary, depending on the steadiness of the motion and confidence in the accuracy of the center fix, with the goal of minimizing unrepresentative short-term motions. Please refer to the section on “Persistence” for a discussion of averaging periods.

After determining a preliminary position, the latest data are analyzed. The first- guess should be compared to surface observations in the vicinity of a TC in order to avoid obvious errors. However, in the case of a poorly-defined system or a lack of data, the first guess may be the best position estimate. The working “best track” of the TC should then be smoothed over the past 24 hours. This re-analysis can be extended further in time if new data are available. Once a current center position has been analyzed, the bearing and speed of the TC are computed, and the new motion estimate is compared to recent estimates to ensure realistic values.

Satellite Fixes

The most common method of “fixing” a TC’s location is through the use of geostationary satellite imagery. These satellites view TCs with high temporal frequency and are the primary tool for TC center fixing. In addition, geostationary images can be animated and modified with color enhancements that correspond to the various brightness temperatures, to help determine the location and motion of a TC center. Polar-orbiting satellites can also be used, but their utility is limited since they pass overhead the same location at most twice daily in the Tropics.

At night it is often difficult to distinguish between low and high clouds in infrared imagery, which complicates center fixing for poorly-defined systems. Nighttime “visible” imagery (Ellrod 1995) has been developed to supplement conventional infrared (IR) imagery for TC center fixing (Jiann-Gwo Jiing, personal communication). This imagery is created by merging imagery from two different “windows” within the infrared spectrum, the shortwave (3.9 micron) and longwave IR (11 micron) imagery. The radiative properties of the shortwave “window” allow for the identification of low cloud features, while the longwave channel is more sensitive to the tops of deep convective clouds. The merged product can distinguish between low clouds and deeper convective features and is useful for locating the center of sheared or poorly-organized TCs. One weakness is that the low cloud pattern can be obscured by cirrus.

A primary method to locate the TC center using satellite imagery is the Dvorak (1984) technique. After meeting minimum requirements for persistence, incipient disturbances typically grow in organization and develop primitive banding features. Dvorak (1984) indicates how to use banding features to determine the cloud system center (CSC). Another method shows how the center may be located by drawing a curved line down the center of a banding feature through the coldest cloud top temperatures. The center typically lies near the inner edge of the band on the counterclockwise end (comma head) portion of the band. Center fixes from both methods should be in agreement. In cases where the CSC is more difficult to determine, the analyst is instructed to draw lines following the cloud line curvature and then fit circles to the lines of tightest curvature (Dvorak 1984). The CSC generally should lie near the common area where the circles converge. When a dry slot is evident on the concave part of a band, the analyst will often be able to identify a cloud minimum wedge. The CSC should fall at the mid-point of the line drawn from the tip of the wedge to the center of the comma head. The estimates from these methods can then be compared to an extrapolation of the TC position by using past motion estimates.

Using satellite imagery to identify the CSC has weaknesses. It is possible to mistake a mid-level center for the low-level center in weak or sheared systems. Additionally, mid- or high-level clouds may obscure the center. The presence of multiple centers, especially those developing within a broad monsoon disturbance, is also a complicating factor. When there is no dominant center, the analyst should estimate a centroid location and not focus on a single center, as this could result in an unrepresentative estimate of position and motion. The most reliable center is almost always found by animating low cloud motions using visible satellite imagery.

Even though satellite imagery is vital to operational TC forecasting, aircraft reconnaissance data are typically always favored over satellite observations. Operational forecasters tend to rely primarily on in situ measurements of the TC inner core provided by aircraft, since they are more accurate than satellite-based fixes. The results of a six-year study of the accuracy of real-time aircraft and satellite fixes of TC location are plotted in Figure 1 and demonstrate the National Hurricane Center’s (NHC) heavy reliance on aircraft data. While a little more than a quarter of all satellite position estimates were in error by at least 20 nm, only 5% of aircraft reconnaissance estimates were in error by the same amount.



Figure 1. Initial position accuracy of satellite and aerial reconnaissance data compared to the final best track positions. Plot shows the frequency (in %) of initial position errors (nm) for satellite versus aerial reconnaissance. The red curve represents the accuracy of aircraft data, and the blue curve represents accuracy of satellite data.



Microwave Satellite Data

The utility of microwave data for TC center fixing from polar orbiting satellites has increased in recent years. This imagery is especially useful when mid- to high-level clouds obscure the TC center in geostationary imagery. These obscuring clouds are transparent in the microwave portion of the electromagnetic spectrum, and thus microwave radiometers can help to reveal the location of the TC center. An example of deep convective and cirrus clouds obscuring the center of developing eastern North Pacific Tropical Storm Georgette is shown in Figure 2 (a) and (b) from 0000 UTC 27 August 2004. A Special Sensor Microwave/Imager (SSM/I) pass from 0411 UTC the same day (c and d) shows the exposed low-level center of the developing TC at the edge of the convective canopy.

Passive radiometers sense emitted microwave radiation over a wide range of wavelengths, enabling them to see through clouds and discern structures that would be difficult to observe using geostationary satellites. Microwave imagery can provide two-dimensional views of TCs that previously were possible only through land-based radars and can be used to supplement center fixes from geostationary imagery.

georgette.jpg

Figure 2. (a and b) IR satellite imagery of Tropical Storm Georgette from 0000 UTC 27 August 2004, with no enhancement in (a) and the standard BD Dvorak curve in (b). (c) SSM/I 85 GHz H pass from 0411 over Georgette and (d) the composite image for the pass, illustrating the exposed low-level center.

A total of 13 low Earth orbiting (LEO) satellites are currently operational (Table 1). Some of the instruments include the Special Sensor Microwave Imager (SSM/I), the Tropical Rainfall Measuring Mission (TRMM) Microwave Imager (TMI), the Advanced Microwave Scanning Radiometer–Earth Observing (AMSR-E) and the Special Sensor Microwave Imager/Sounder (SSMIS). The Instruments section gives a more detailed description of each sensor, its attributes, and capabilities.

Satellite

Instrument

Operating Since

Scan Strategy/ Type

Frequencies (GHz)

No. of Channels

Resolution (km)

Swath Width (km)

NASA/

DoD Coriolis

Windsat

2003-

conical, imager

6.8-372

22

11-55

1025

NASA TRMM

TRMM (TMI)

1997-

conical, imager

10.65-85.5

9

5-50

780

Metop-A%

AMSU-A

AMSU-B

ASCAT

2006-

cross-track, sounder

scatterometer



23.8-89

85-183.31

5.25


15

5


48

16

50 (25)



1650

1650


2 520

F-13$

F-14

F-15

SSM/I

SSM/T1

SSM/T2

1987-

conical, imager

cross-track, sounder

cross-track, sounder


19.5-85

50.5-59.4

92-183


7

7

5



15-69

*174 (48-120)@

*309 (85-213)#


1400

DMSP

(F-16, F-17, and F-18)

SSMIS

2004-

conical, imager & sounder

19-183

24

12-55

1700

Aqua

AMSR-E

2002-

conical, imager

6.9-89

14

5-50

1600

NOAA-15

NOAA-16

NOAA-17

NOAA-18

AMSU


1998-

2000-


2002-

2005-


cross-track, sounder

50-183

20


**50 (16) @

**150 (50) #



2200

Table 1. Table showing the current suite of microwave polar orbiting instruments along the associated satellite. Format is AMSU-A (AMSU-B/MHS), * Format is SSM/T1 (SSM/T2), 2 WindSat channels at 10.7, 18.7, and 37 GHz are fully polarimetric, @ at nadir, # at limb, $ currently inoperative, and % Metop-B to be launched in 2010

Data obtained at 85 to 91 GHz are primarily used to observe the deep convective clouds associated with TC structure, especially in the TC core. Within this region of the electromagnetic spectrum, water droplets in the part of clouds near the freezing level help deplete upwelled microwave radiation. Since non-precipitating cirrus has little effect on radiation at this wavelength, the remaining upwelled radiation is released to space, where satellites can observe it. As a result, the scattering and absorption effects leave much less microwave radiation aloft, making satellite brightness temperatures appear cold.

At 37 GHz, microwave radiation rising from the ocean surface and the warm cloud region is only slightly diminished after its passage through the remainder of the cloud, making observed brightness temperatures in areas of low clouds and precipitation warm at this frequency.

Frequencies (GHz)

Corresponding Altitudes (km)

85 – 91

8-10

36 - 37

5-8

19

3-5

Table 2. Typical height (in km) for microwave imaging frequencies for analysis (in GHz).

One of the advantages of using passive microwave imagery is that it provides information on TC structure from which a center location can be inferred. Data near 85 GHz can depict the location of a mid-level center in a vertically sheared system, since those frequencies reveal features at middle- to upper-levels. Imagery from 37 GHz, however, can reveal low cloud features indicative of the low-level center and is more suitable for center fixes for weaker TCs, even though it is of lower resolution than the 85 GHz channel. Center fixes from these two channels may not often in agreement, particularly for sheared TCs.

Analysts should interrogate multiple channels/frequencies, polarizations, and composite images to determine the TC center location. Combining and cross-referencing data from multiple images allows the analyst to confidently provide TC position estimates from a few independent center fixes. A brief, step-by-step procedure providing center fixes from an analysis of microwave images follows.

The analyst can obtain a myriad of real-time and archived microwave imagery from the U.S. Navy’s NRL or FNMOC TC web sites located at http://www.nrlmry.navy.mil/tc_pages/tc_home.html and https://www.fnmoc.navy.mil/tcweb/cgi-bin/tc_home.cgi. The analyst can begin by evaluating high and low frequency single polarization images. The 85- 91 GHz images are easy to interpret because the



microwave_example_global guide.jpg

Figure 3. TRMM 85 GHz H image of Hurricane Ike at 1905 UTC 3 September 2008.

cloud liquid water appears as warm (darker blue in Fig. 3) and represents low cloud lines while significantly colder brightness temperatures (red) represent intense convection. For example, Figure 3 is an 85-GHz H-pol TRMM pass over Atlantic Hurricane Ike, which illustrates warm rain processes in low-level bands at outer radii, while the brighter yellows and reds depict radiometrically cold convective cloud tops at a much higher level surrounding the eye. Next, lower frequency polarized images (e.g., Figure 4-5) should be evaluated.

microwave_example_global guide2.jpg

Figure 4. TRMM 37 GHz V pass over Hurricane Ike at 1905 UTC 3 September 2008.

The center of incipient systems can sometimes be difficult to locate since increasing low-level convergence of cloud liquid water near the vortex can initially mask the center. Finally, the analyst should evaluate both high and low frequency color composite images, which use the Polarization Correction Temperature (PCT) solution to discriminate intense convection from both single polarized images. On the PCT image one should focus on the blue-green colors, which primarily denote

microwave_example_global guide3.jpg

Figure 5. TRMM PCT image over Hurricane Ike from 1905 UTC 3 September 2008.

low-level clouds. In stronger storms, a helpful hint for locating the TC center is to look for a “dark” spot on the low frequency color image, which should be the “rain-free” dry region. Figure 5 represents an example of a PCT image from the same TRMM pass over Ike (as in Fig. 4), where blue-green streamers indicate low cloud bands and reds and pinks indicate well-developed convective clouds. A dark green spot, representing an area that is the same temperature of the ocean surface, indicates the cloud-free eye. Analysis of multiple images (preferably within six hours of the analysis time) can increase chances of identifying the TC center.

The scanners present on the various polar-orbiting satellites are either, cross-track or conical. The cross-track instruments scan at varying angles as the instrument points away from nadir. As a result, these imagers have degraded resolution near the edge of the scan, where the footprint size is larger (Fig. 6a.) Conical-scanning instruments look forward at a fixed angle with a rotating antenna, and footprints of equal size provide the same resolution across the entire swath (Fig. 6b.).



cross_track_conical_scanners.bmp

Figure 6. Illustration of (a) the cross-track scanner, which scans at varying angles with higher resolution in the middle of the swath and lower resolution on the edges (courtesy of Chris Velden), and (b) a conical scanner, which has a fixed viewing angle and a constant-sized footprint size across the swath (courtesy of UCAR/COMET).

The difference between the two scanning techniques is apparent over western Pacific Supertyphoon Podul on 24 and 25 October 2001 (Figs. 7-9). The AMSU-B 89 GHz image taken at 0246 UTC 25 October 2001 captures Podul at the edge of the swath, where the limitation of larger footprint size and a coarser resolution leads to an enlargement of the pixel size and distortion of the pixels (Fig. 8). This degraded resolution can severely hamper TC center fix estimates. A comparison with the 2228 UTC 24 October 2001 SSM/I image (Fig. 7) from a few hours earlier before demonstrates the superior quality of the conical scanner, as the TC center fix is more reliable due to the higher resolution imagery. Figure 9 is another image of Podul from 1036 UTC 25 October 2001, illustrating the high resolution of the SSM/I pass even at the edge of the swath.

podul.jpg.jpg

Figure 7. SSMIS 85 GHz H image of Supertyphoon Podul from 2228 UTC 24 October 2001, illustrating the superb resolution of the conical scanner across the entire swath. The colors correspond to brightness temperatures of the legend found below.



podulgross.jpg.jpg

Figure 8. AMSU-B 89 GHz image of Supertyphoon Podul at 0246 UTC 25 October 2001, illustrating the poor resolution of the cross-track scanner at limb. The colors correspond to brightness temperatures in the legend below.



podul3.jpg

Figure 9. SSM/I image of Supertyphoon Podul from 1036 UTC 25 October 2001, illustrating the remarkably good resolution of conical scanner at limb. The various colors correspond to brightness temperature given in the legend below.



The Instruments

  • The Tropical Rainfall Measuring Mission (TRMM) flies in a near-equatorial orbit, measuring rainfall and other rainfall-related products using two different sensors. The passive TRMM Microwave Imager (TMI) detects rainfall by measuring the intensity of backscattered radiation at five different frequencies between 10.7 and 85 GHz. As a consequence of its lower orbital altitude, the satellite has smaller footprint sizes and higher resolution (6 km). The TMI is also frequently used for TC observation, as its mid-Earth orbit is confined to a narrow region between 35° N and 35° S, ideal for observing most TCs.



  • Advanced Microwave Sounding Unit-B (AMSU-B) is another passive instrument on the NOAA series of polar-orbiting satellites. The AMSU instrument looks cross-track, allowing its scan angle to vary as it points away from nadir.



  • Metop-A is the first in a series of three polar-orbiting satellites flown by the European Space Agency (ESA) carrying a high resolution microwave imager similar to that found on the NOAA series.



  • The Special Sensor Microwave Imager (SSM/I) is another instrument on board Defense Meteorological Satellite Program (DMSP) satellites that can detect and track TCs.



  • The SSMIS instrument is a follow-on to the SSM/I T1 and T2 instruments (temperature and water vapor profilers, respectively) and is a replacement to the earlier SSM/I. SSMIS is a passive, conically-scanning radiometer, with more spectral channels and a 1700-km-wide swath, compared to 1400 km for SSM/I.



  • The WindSat instrument is one of two payloads aboard the U.S. Navy’s Coriolis satellite launched in 2003. It can measure wind speed and direction, although not reliably, in the heavy rain environment in the core of TCs. The multi-frequency polarimetric radiometer passively measures emitted microwave radiation At five different frequencies ranging from 6.8 to 37 GHz, is a conical scanner, and has about three times better spatial resolution than the SSM/I, and is available to monitor TCs.



  • The Advancing Microwave Sounding Radiometer – Eos (AMSR-E) is a passive conically-scanning instrument flying on the NASA Aqua research satellite. It senses microwave radiation at 12 different channels on six different frequencies ranging from 6.9 to 89 GHz.

Cautionary Considerations

Microwave imagery should be used with caution to estimate TC location. The slanted viewing geometry of microwave polar-orbiting satellites displaces features slightly askew of their actual location (i.e., parallax error). Figure 10 illustrates the typical parallax error which results in the 85-GHz channel, where scattering of microwave radiation by ice particles is dominant, and in the 37-GHz channel, where that satellite mainly senses microwave radiation emitted from cloud water droplets in the lower part of clouds. Paired with the schematic drawings in Figure 10 are two nearly simultaneous microwave images over Typhoon Jelawat on 8 August 2000, which reveal the discrepancy in TC position between the two frequencies. The image in panel b) shows how Jelawat looks from the lower-level 37 GHz imagery, with the red circle indicating where the analyst has placed the center. In panel d), however, the image reveals a distorted and much broader eyewall surrounding the center, and the yellow circle corresponds to the center fix by the analyst. Parallax error is present in both images, but the 15- to 20-km error in the 85-GHz channel is greater than the 5-km error in the 37-GHz image. Smaller errors are typically seen in the 37-GHz imagery because the particles sensed at that frequency are emitted from a lower-altitude, reducing the distortion caused by the viewing geometry relative to the ice particles at higher altitudes seen at higher frequencies.



parallax_error2.jpg

Figure 10. Illustration of typical parallax error associated with two microwave channels. 6(a) shows an exaggerated cross-section of the parallax error from the 37 GHz channel, while (c) shows the parallax error associated with the 85-91 GHz channels. (b) and (d) show typical parallax error as observed in microwave imagery, with (b) showing minor parallax error at 37 GHz and (d) showing more substantial parallax error at 85 GHz. Figure courtesy of UCAR/COMET.

The availability of microwave data is another limitation which must be considered. Data latency, the time between a satellite overpass and its successful transmission to a ground receiving station for processing, can result in a significant time lag between the pass over the TC and when the imagery is available. Figure 11 indicates the fraction of data available in real-time as a function of data latency. On average nearly 80% of all data are available within five hours of overpass (for perspective, forecasts are produced on a six-hourly cycle), and there are still occasions when data are further delayed. The next generation of U.S. low Earth orbiting satellite, the National Polar-orbiting Operational Environmental Satellite System (NPOESS), is expected to be greatly reduce data latency with additional ground receiving sites, with 95% of all data likely within 30 minutes of overpass (NPOESS COMET module).

latency.gif

Figure 11. Plot of the fraction of microwave polar-orbiting data available to users in real time as a function of time. The x-axis or delay time is in hours, while the y-axis or data fraction available to users is in %. Figure courtesy of UCAR/COMET.


  1   2   3   4   5


The database is protected by copyright ©ininet.org 2016
send message

    Main page