Challenging Aspects of Severe Tropical Cyclone George s s Track Forecasts Jim DAVIDSON Former Affilliation
Severe Tropical Cyclone George bears down on the Western Australian Coast in early March 2007
Tropical Cyclone George March 2007
Setting the Scene STC George was both very intense and physically large 3 fatalities numerous injuries localised damage 10 minute mean wind = 194 km/h recorded just offshore Equivalent to wind gusts reaching 275 km/h Very destructive winds to about 115 km inland 5 metre storm tide but landfall near the bottom of the tide Substantial but not significant rainfalls occurred No major flooding due lack of previous rainfall and the steady movement of the cyclone Track forecasts were very challenging and the focus of this presentation
4 Panels of Multi Sensor Satellite Imagery close to Landfall
Radar Image with the elliptical eye of George straddling the coast
Very Destructive Wind swath of TC George extended well inland
Track Forecast Performance WA TC Forecast Accuracy : 2006 07 season 600 500 400 Accuracy (km) 300 200 100 0 0 12 24 48 Forecast Tim e (h) George Jacob Kara 01/02 05/06 Average
Selection of Forecast Tracks & Best Track at base time 060000 UTC Best Track in Black with Multi model Forecast Tracks in various colours
Selection of Forecast Tracks & Best Track at base time 061200 UTC Best Track in Black with Multi model Forecast Tracks in various colours
Selection of Forecast Tracks & Best Track at base time 070000 UTC Best Track in Black with Multi model Forecast Tracks in various colours
Selection of Forecast Tracks & Best Track at base time 0712000 UTC Best Track in Black with Multi model Forecast Tracks in various colours
EC Model Track Forecast Ensemble at base time 060000 UTC
Warning Graphic with 24/48 hour Forecasts base time 060000 UTC
One exception to the generalisation The UKMO ensemble did indicate the possibility of a tight recurvature but this prediction was made earlier on 3 March. Amidst other guidance, including later runs of the UKMO, it did not influence the forecast to any degree except maybe to raise the probability of such a scenario from zero to still quite a very low number.
Warning Performance The possibility that George could recurve to the south was identified late morning on 7 March and a Cyclone Watch was issued for coastal and island communities The section of coast where George ultimately made landfall was first put under Cyclone Warning late on 7 March, approximately 24 hours prior to landfall Warnings consistently indicated that George was expected to impact the coast as a Severe TC Moreover the warnings specifically stated that the system was expected to produce very destructive winds for more than 100 km inland
TCWC Methodology Meteorologists in the TCWC construct track forecasts based on a consensus approach, combining available dynamical and statistical forecast aids The forecasts issued by the TCWC drew directly upon guidance from 9 independent models Meteorologists also informed their decision making by accessing output from models run in ensemble mode Despite being the most skilful track forecasting methodology available and practised in all major TCWCs it is still dependent on the skill of the component guidance
TCWC Methodology Remarkedly, the available guidance had a small spread (very little variation in direction and speed of the tracks) None of the available guidance foreshadowed the abrupt southerly shift in motion This was considered a surprising outcome even in 2007 It was extremely rare for all available guidance to fail Small spread lends to higher confidence in the track forecast Hence this is the worst situation a TCWC meteorologist can face namely a small spread leading to high confidence, but ultimately large error
Error Vs Spread 1. Small Spread/Large Error: Nightmare 2. Large Spread/Large Error: Largest opportunity for improvement 3. Small Spread/Small Error: Ideal Case 4. Large Spread/Small Error: Opposing errors cancel each other out 1 2 3 4
Small Model Spread Large Error George the hard one!
A Puzzling Question Assuming a model has a fair and realistic representation of the TC and the environment, why/how does it steer the system in a direction inconsistent with the steering flows it depicts?
TC Steering 101 (largely courtesy of NRL Monterey) TC motion is the result of a complex interaction between a number of internal and external influences Large scale environmental steering is typically the most prominent external influence on a TC, accounting for as much as 70 to 90% of the motion and is computed by separating the TC wind fields from the large scale environmental wind fields (which is easier said than done) Separation of flow is dependent on the situation but can be anywhere from 1 7 degrees from the centre of the TC Deep layer mean steering is obviously best but a single layer computation can sometimes provide a crude estimate Internal influences (such as the Beta Effect) steer the TC poleward and westward in both hemispheres
Possible Significant Factors leading to the Sharp Recurvature TC Humba (undergoing ETT well upstream) TC Jacob (Fujiwhara Effect) Beta Effect Poorly represented monsoonal flow in the models (Assymetrical Convection not a consideration here)
Possible Significant Factors leading to the Sharp Recurvature TC Humba (undergoing ETT well upstream) TC Jacob (Fujiwhara Effect) Beta Effect Poorly represented monsoonal flow in the models
TC Humba was undergoing ETT well upstream (near 90E) and is unlikely to have influenced George s track to any measurable degree
Possible Significant Factors leading to the Sharp Recurvature TC Humba (undergoing ETT well upstream) TC Jacob (Fujiwhara Effect) Beta Effect Poorly represented monsoonal flow in the models
TC Tech Bulletin issued 0112 UTC 07/03/2007 by the Perth TCWC:... No model has effectively captured TC Jacob in the analysis fields, and therefore all guidance is considered with some caution. Interaction between the two systems may imply a greater level of SW motion in the short term, but may also hinder any recurvature in the longer term....
I am almost 100% convinced that this was all down to the model's handling of Jacob. I am giving a talk at a symposium later this week on why tropical cyclone forecasts sometimes go wrong. I plan to make this one of my prime examples! It illustrates that modelling the environment around the tropical cyclone is just as important as modelling the cyclone itself to get a good forecast. Julian Heming Tropical Prediction Scientist UK Met Office
Jacob was relatively weak and moving further away from George during the critical forecast period J G Track of TC Jacob
TC George TC Jacob interaction 300 200 george 100 7/3 09Z 4/3 00Z Jacob 0 500 400 300 200 100 0 100 200 300 400 500 4/3 00Z 100 7/3 09Z 200 300
TC George TC Jacob Interaction TC George was larger and more intense Expect greater effect on TC Jacob s track Break in interaction as cause of S ly motion? Systems don t breach direct interaction threshold From late on the 5 th interaction would have decreased as separation increased Systems continued to separate around time of TC George s abrupt change in direction towards the south Independent or causal? Weak argument given distance and relative strengths
Possible Significant Factors leading to the Sharp Recurvature TC Humba (undergoing ETT well upstream) TC Jacob (Fujiwhara Effect) Beta Effect Poorly represented monsoonal flow in the models
Beta Effect 101 (largely courtesy of NRL Monterey) Local change of the vertical component of relative vorticity maximum value at the equator and zero at the poles Causes TCs to move polewards and westward with a speed of several degrees per day (in the absence of the largescale steering flow) Function of the TC size but not necessarily the TC intensity When the TC size is large (as with George), the Beta Effect may have some impact on the motion and should therefore be considered when developing track forecast strategy
Beta Effect on George Physically large therefore expect greater beta effect Likely to have influenced the motion to some degree but difficult to argue that the beta effect contributed in a significant way to the abrupt change in direction Far more likely that the beta effect would have lead to a more gradual change in direction towards the southwest
Possible Significant Factors leading to the Sharp Recurvature TC Humba (undergoing ETT well upstream) TC Jacob (Fujiwhara Effect) Beta Effect Poorly represented monsoonal flow in the models
300/50 knots 090/50 knots 010/25 190/25 knots
4 Sector Vector Analysis at 071200 UTC EC 500 hpa North 300/50 East 330/30 South 100/60 West 130/20 SUM 025/35
061200 UTC 070000 UTC TXLAPS 500 hpa Wind Analyses 071200 UTC 080000 UTC
TXLAPS 500 hpa Wind Analysis at 080000 UTC
TC Justin in 1997 was centred in the Coral Sea off Queensland. The system was located north of a strong 500 hpa ridge and all computer models forecast it to move under the influence of the middle level easterlies on to the Queensland coast. However the monsoon westerly flow to the north was very strong and deep such that the easterly flow was balanced by the monsoon flow and the system remained stationary.
TC LUA 2012 small spread & small error
Animation of Model & Actual IR Imagery (TC LUA) Animation of Synthetic and Actual IR Imagery (Sun and Rikus, 2004) ACCESS TC Operational Forecast of TC Lua, Base Time 12UTC, 15 March 2012 Courtesy Lawrie Rikus, ESM Program, CAWCR Sun, Z. and L.J. Rikus, 2004: Validating model clouds and their optical properties using Geostationary satellite imagery. Mon. Wea. Rev., 132, 2006 2020
Severe Tropical Cyclone YASI
Track of TC YASI with extent of hurricane, storm & gale force winds Bright RED = Hurricane force winds
TC YASI a really good example of Small Spread & Small Error for almost a week leading up to landfall
TC Yasi Forecast Tracks
TC YASI (& TC ULUI) ) Forecast Performance
Model forecasts for SANDY were exceptional out to 7 days Deterministic EC Multi Model Ensemble Strike Probability Map
In conclusion The influence of strong and deep monsoonal westerlies on the steering flow of TCs is not always fully captured by numerical models (as highlighted with TC George) Consensus track forecasting has been largely a positive experience but still value from detailed synoptic analysis Increased availability of near real time wind field analyses from a suite of numerical models together with remote sensed wind data over the oceans should hopefully lead to careful synoptic analysis being a routine function in TCWCs (if of course forecaster numbers allow)
Thank you Questions?