Comments on Tropical Cyclone Model Forecasts
In the last few years, it has become easy to view on the Internet the forecast tracks of tropical cyclones produced by the various numerical computer models that simulate the atmosphere. Early Alert provides some of these forecasts as part of its graphics display (example below)
Some may ask why there are so many models used to forecast tropical cyclones, which can be confusing at times. Why don’t we just use the model that produces the best forecasts? The simple answer is there is no one perfect model of the atmosphere. The interaction between tropical cyclones and their surrounding environment is very complex and often occurs in areas with few direct measurements of the air from within and just outside the storm. Some models will produce better forecasts than the others depending on their strength, size, shape, location or time of year.
Forecasters at the National Hurricane Center are aware of these specific characteristics of the models and will weigh one model’s forecast over another depending on the situation, then produce the “official” forecast based on this expertise.
These models have been developed by various scientific establishments in the United States and other countries. The models are continually being refined and upgraded based on the priorities and resources of the individual developers. These types of models are known as Global Circulation Models (GCMs), which simulate the behavior of the atmosphere around the world. Examples of GCMs include the GFS from the NOAA/NWS Center for Environmental Prediction, GFDL from the NOAA Geophysical Fluid Dynamics Lab, NOGAPS from the US Navy, the ECMWF from Europe, CMC/GEM from Canada, and UKMET from the UK.
As noted above, there are different characteristics of each model pertaining to the tropical cyclone forecasts. These global circulation models are not set in stone, but typically undergo changes to incorporate the latest improvements in science and computer technology. Therefore, notions of how one model performed over another will not necessarily apply from one year to the next.
Currently, the GCMs most heavily weighted by meteorologists for predictions of tropical cyclones tend to be the GFS and the ECMWF. In May of this year, a major change to the way data from tropical cyclones is handled by the GFS model was executed. This appears to have greatly improved the forecast tracks as demonstrated in the cases of Debby and Ernesto.
Again, even though one model shows superior performance over another in one or two instances that may not always be the case. Know that forecasters at the National Hurricane Center are most knowledgeable in model characteristics and any recent changes to model performance. The official NHC forecast highlighted in Early Alert graphics is the one on which plans should be made. But if you are in the habit of checking the various forecast tracks on the Internet, and want to acquire your own sense of confidence in what is happening, look for trends in the entire set of forecast tracks. If the tracks are clustering closer together, your confidence in the forecast should be higher. If the tracks are spread apart, your confidence level should be lower. Also, a general shift in the entire cluster of forecast tracks should give you a better sense of and impending change with the projected track of the storm.