FIFTH INTERNATIONAL WORKSHOP ON TROPICAL CYCLONES
PRESENT AND FUTURE USES OF SATELLITE OBSERVATIONS FOR TROPICAL CYCLONE FORECASTING AND RESEARCH.
Topic 0.1a Forecast office presentations.
Presentor: Jeff Callaghan
Bureau of Meteorology,
295 Ann Street Brisbane
Fax: 617 32214895
Abstract: The availability of satellite imagery has produced substantial improvements in the detection, analyses, and prediction of tropical cyclones in the Australian region over the last few decades. The Dvorak satellite technique has been the backbone of tropical cyclone intensity analyses over this period. However, new technologies have introduced microwave, water vapour and scatterometer data that have provided a vast increase in information in the vicinity of tropical cyclones and ultimately lead to increased warning lead times and reduction of forecast errors. We discuss how introduction of these new data have led to increased understanding of tropical cyclones as well as uncertainty in the analyses of the most intense tropical cyclones. We show how increased data have revealed systems not well described by the usual conceptual tropical cyclone models.
In 1954, a disastrous tropical cyclone (dubbed the Great Cyclone) struck the Australian subtropical east coast and caused around 30 fatalities. The Regional Director of the Bureau of Meteorology in Queensland at the time described the tracking of this cyclone: After leaving Vanuatu the centre of the cyclone swung into the Weather Bureaus blackout zone- 600,000 square miles of the Coral Sea - without even one inhabited atoll to receive reports from. We were not certain of the cyclones path and we only received two sketchy reports from ships but we could only guess as to what it was doing. We call this area No Mans Ocean, as we just dont know what is happening there. The Bureau lost the cyclone for 2 days until its winds started to buffet the Queensland coast.
This illustrates the power of satellite in locating the positions of tropical cyclone from their tell-tale cloud signatures that we tend to take for granted today. Below we describe our use in Queensland of the various satellite products, problems we face in this basin, and products that we would like to receive.
0.1.2. Dvorak Technique
The Dvorak (1984) analysis is the worldwide standard for tropical cyclone intensity monitoring in the absence of aircraft. In the Australian region, this technique has served us admirably for many years and undoubtedly has led to accurate warnings, which have saved many lives in the region. Nevertheless, at times there are difficulties in applying the technique, which we now describe.
Tropical cyclone Justin. The Dvorak technique was difficult to apply during the formation of tropical cyclone Justin and some international agencies were reluctant to call it a tropical cyclone. It was named at 0000UTC 7 March 1997 and soon after a yacht foundered off the North Queensland Coast in huge seas. It was a large storm with an unusual cloud structure and the infrared satellite imagery in Figure 1 shows it at near-peak intensity in the first phase of its life.
The mean sea level (MSL) analyses in Figure 2 were primarily derived from the Coral Sea automatic weather station (AWS) observations and show the pressure gradient was strongest near the centre of Justin at 1200UTC 9 March 1997 when the vessel Osco Star (marked by arrow at this time) reported its maximum wind speed of 80 knots from the northwest. From comparison of its earlier observations with nearby AWS data, we assessed this observation from the Osco Star as a 10-minute wind speed of 65 knots. This maximum wind zone was under deep convection 180 to 190 km northeast of the centre of Justin. During its lifetime, Justin sunk three other vessels and damaged many others including the Osco Star, which had to be towed back to port. Tragically, Justin was responsible for around 40 fatalities in the region.
Impact of 85GHz data on Dvorak analyses. Data at microwave frequencies from polar-orbiting satellites are more directly related to precipitation than are those from visible and IR channels. The upwelling radiation at these microwave frequencies can therefore be used to assess structure of the tropical cyclones precipitation regions. We compare Special Sensor Microwave/Imager (SSM/I) and the higher resolution microwave data from the Tropical Rainfall Measuring Mission (TRMM) satellite with IR imagery for tropical cyclone Vance in Figure 3.
Figure 1. (Left)IR satellite
imagery 1133UTC 9 March 1997.
Figure 2. (Right) Mean Sea level
isobars at 978hPa and 982hPa near
the centre and then 990hPa and
998hPa together with AWS wind observations every 12h from 0000UTC 8 March 1997 (080000) to 1200UTC 9 March 1997 (091200). Hatched areas denote cloud tops colder than -320C and black areas show cloud tops colder than -630C .
Figure 3. Enhanced IR, with corresponding 85GHz imagery SSM/I in the top right frame and TRMM images in the top left and centre bottom frames for tropical cyclone Vance during March 1999.
The TRMM image at 0459UTC 20 March 1999 has what looks like a concentric eyewall pattern with a very small and circular inner eye, which indicates a very intense tropical cyclone. The corresponding IR image at 0502UTC barely resolves the eye and indicates a T number around 5.5 and maximum 10-minute wind speeds of around 90 knots, which agrees with the best track data (central pressure 940hPa). The SSM/I image at 1248UTC does not resolve the inner eye. However the corresponding IR image at 1233UTC has a warmer eye evident and the best track central pressure for this time is 925hPa. The next TRMM image at 1910UTC 20 March 1999 again shows the concentric eyewall pattern with the outer eyewall beginning to supplant the inner eye, which showed signs of weakening. The corresponding IR image at 1932UTC has a better-defined eye and the central pressure in the best track for this time is 910hPa. A subsequent TRMM image at 0322UTC 21 March 1999 showed the inner eye still evident but further weakened.
The question to be resolved here is: While Dvorak IR analyses indicated intensification over the period, do the TRMM images indicate that over the period that Vance weakened from an extremely intense cyclone (central pressure below 900hPa) due to undergoing a concentric eyewall cycle? If this is correct, it would appear that we need TRMM observations to detect some of the most intense tropical cyclones.
Tropical cyclone Kelvin.
Figure 4.Enhanced infrared
satellite imagery at 1231UTC
25 February 1991. Willis Island is denoted by
large solid circle and wind
observations on the Island at the
time were 190/65 knots with gusts
to 84 knots.
Kelvin was a rapidly intensifying system that reached hurricane force intensity while exhibiting the cloud signature (Figure 4) of a sheared system. Prior to verification of its intensity from the Willis Island observations, it was operationally given a T number of 2.5 (below tropical cyclone intensity). It developed south of a strong monsoon flow with the 900 hPa northwesterly winds at Port Moresby in Papua New Guinea increasing to 60 knots (31 m s-1 ). We will show how a marked middle-to upper-level involuted (TUTT type) trough was developing south of Kelvin as it intensified.
0.1.3 Rapidly intensifying small tropical making landfall.
These events are badly forecast by numerical forecast models and their intensity is often underestimated using Dvorak techniques. A recent example is Tropical Cyclone Tessi, which made landfall just to the north of Townsville. European Centre for Medium Range Weather Forecasts (EC) usually handles these systems better than the other forecasting models. However, the EC forecasts covering the period up to the landfall of Tessi (Figure 5) indicated that a weak tropical low would move toward the east coast as it further weakened into an innocuous trough. The mesoscale MSL charts (Figure 6) as Tessi passed to the north of Townsville show an intense small system.
The maximum measured 10-minute wind speed from the AWS on Magnetic Island was 59 knots at 1600UTC 2 April 2000 (see left centre frame on Figure 6). Subsequently, in the lower two frames in Figure 6, Tessi contracted into a small intense system near the point of landfall. The worst wind damage was found in the isolated areas near this point of landfall. Therefore Tessi probably achieved 10-minute mean winds of 65 knots or more (hurricane force). The 85GHz TRMM and SSM/I (Figure 7)images between 2301UTC 1 April 2000 and 2000UTC 2 April 2000 show the rapid development of the eye at landfall. No eye was evident on IR imagery.
Figure 5. EC MSL forecasts covering the period leading up to the landfall of tropical cyclone Tessi.
Figure 7. 85GHZ data at 2310
UTC 1 April 2000 (top) and
2000UTC 2 April 2000 (bottom)
Figure 6. MSL analyses from 1400UTC 2 April 2000 to 2000UTC 2 April 2000 as Tropical cyclone Tessi passed to the north of the City of Townsville.
0.1.4 Vertical wind shear.
Vertical wind shear charts available from the Cooperative Institute for Meteorological Satellite Studies (CIMSS) at the University of Wisconsin web site are widely used in the Australian region. However, here we show an example in which the shear is in the lower half of the troposphere and is missed by the shear charts. Tropical cyclone Vaughan was moving toward the North Queensland coast and underwent a huge convective burst at 1800UTC 5 April 2000(4 am local time). The vertical wind shear chart for that time is shown in Figure 8 and indications were that the convection was in a low shear environment. From McBride and Zehr (1981), a requirement for development of tropical cyclones is for positive (negative) vertical wind shear poleward (equatorward) of the cyclone and poleward (equatorward) vertical wind shear to the west (east) of the centre. This is the case in the shear analyses in Figure 8. However, the middle level circulation of this cyclone was steered onto the coast at 1100UTC 6 April 2000 by a 500hPa ridge (see 500hPa sequence in Figure 9 b)). At the same time, the low-level centre was left behind following collapse of the low-level ridge to its south (850hPa sequence in Figure 9(b)).
At the Workshop, we will discuss another recent case in which the low level centre lagged behind the middle-level centre at landfall.
Figure 8. (Top)CIMSS
150-300mb layer mean
minus 700-925mb layer
mean wind shear
Figure 9(a). (Centre)
Sequence at 850hPa.
2300UTC 4 April 2000 to
1100UTC 6 April 2000.
Figure 9(b). (Right)
Sequence at 500hPa.
2300UTC 4 April 2000 to
1100UTC 6 April 2000.
0.1.5. Upper outflow patterns from water-vapor winds.
The water-vapor winds from the CIMSS site are popular products for forecasters in the Australian region. We use them to diagnose the upper outflow patterns. Intensifying storms are observed to have strong outflow patterns with the more rapid intensifiers having strong upper outflow channels on both the equatorward and poleward sides of the cyclone. Additionally, we often see the formation of upper cyclonic flow near the centre at the commencement of a rapid intensification period. This phenomenon was examined by Titley and Elsberry (2000) for the pre-conditioning period of Supertyphoon Flo. Below we examine the formation of upper cyclonic flow near the centre of tropical cyclone Chris between 0000UTC and 1200UTC 3 February 2002 (Figure 10). The 85GHZ data in Figure 11 shows that soon after this time Chris rapidly formed an eye.
Figure 10. CIMSS water vapor winds for tropical cyclone Chris.
Figure 11. 85GHZ data at 2204 UTC 3 February 2002 (left) and 0505UTC 4 February 2002 (right)
0.1.6 Value of TRMM rainfall data in forecasting flood rains
Tropical cyclone Elaine weakened off the Western Australian coast and by 2137 UTC 18 March 1999 the system was almost devoid of deep convection. By 0319 UTC 19 March 1999 an intense rain band developed south of the low-level centre and TRMM rainfall rates in the band (Figure 12) were off the scale, and thus much in excess of 20 mm/h. The rainband passed over Geraldton and 100 mm was recorded there in the 24 hours to 0100 UTC 20 March 1999. Of this amount, 86 mm fell in the last 6 hours. The southern rainbands evident in the TRMM images in Figure 12 at 2202 UTC 19 March 1999 and at 0251 UTC 20 March 1999 were located over the town of Moora with rainfall rates of 15 mm/h. A much heavier northern rainband was evident on the TRMM images and at 0251 UTC 20 March 1999 this northern rainband was about 100 km north of Moora. At 1900 UTC 20 March 1999 torrents of water swept into the town of Moora and damaged houses and businesses and drove 1800 people from their houses.
Australian hydrologists were greatly impressed by this rainfall data as there was not enough rainfall gauges on the ground to predict the flooding. They questioned why there was not more extended availability of these satellite data.
Figure 12. TRMM rainfall data
associated with ex-tropical cyclone
Elaine, 0319UTC 19 March 1999 (top),
2202 UTC 19 March 1999 (top right)
and 0251UTC 20 March 1999 (lower
0.1.7 Quikscat wind observations.
QuikSCAT is on a polar-orbiting satellite and has an 1800 km wide measurement swath on the earth surface. Generally, this results in twice per day coverage over a given geographic region. SeaWinds is a radar instrument on the Quikscat that sends pulses to the ocean surface and measures the echoes that bounce back to the satellite called backscatter. Winds derived from the Quikscat data for the 4 July 2001 are displayed in Figure 13 and reveal a large fetch of gale-to storm-force winds on the southern side of a tropical low. Large long period swells with significant wave heights over 4 metres were measured on offshore wave rider buoys near Brisbane, Queenslands with periods to 15.9seconds.
Figure 13. QuikSCAT winds 1901 UTC 3 July 2001 (left) and 0601 UTC 4 July 2001 (right).
The long period swells arrived on the east coast of Australia in beautiful weather conditions. Long period swells have a large breaker height to deep-water wave height ratio and also break much closer to the shore than they normally would. Under steeply sloping ocean floor conditions near the coast, these waves can double in height. On the 6, 7 and 8 July 2001, large waves impacted on the Southern Queensland and New South Wales Coasts. Six people were drowned in Queensland waters in three boating disasters on coastal bars while in New South Wales two rock fishermen were swept to their deaths. Forecasters used this QuikSCAT data to warn the public of the large ocean swells, but this did not prevent the loss of life in these abnormal conditions.
We have described how the Brisbane, Queensland office currently uses the various forms of satellite data in tropical cyclone forecasting operations. However, we showed some examples of disastrous tropical storms that are not easily analysed by current satellite techniques. We may have also identified very intense tropical cyclones, which probably need high resolution TRMM data for detection. We showed how TRMM rainfall data have the potential to aid flood forecasting in remote areas and Australian hydrologists have questioned why these data are not more widely available. At the workshop, we will also describe other satellite products that we would like to receive such as deep mean layer winds that are currently available for other basins.
Dvorak, V.F. 1984: Tropical Cyclone Intensity Analysis Using Satellite Data. NOAA Technical Report NESDIS 11. 45pp.
McBride, J.L. and Zehr, R. 1981. Observational Analysis of Tropical Cyclone Formation. Part II: Comparison of Non-Developing versus Developing Systems. Monthly Weather Review, 38, 1132-1151.
Titley, D.W. and Elsberry, R. L. 2000. Large Intensity Changes in Tropical Cyclones: A Case Study of Supertyphoon Flo during TCM-90. Monthly Weather Review: Vol. 128, pp. 3556-3573.
Page # 0
Jeff Callaghan, Bureau of Meteorology, Brisbane.