In a key step toward improving the prediction of hurricanes, scientists at the National Centre for Atmospheric Research (NCAR) have reproduced in a computer model the fine-scale structure that drives the birth and development of tropical cyclones.
The simulation, which used the NCAR/Penn State Mesoscale Model, Version 5 (MM5), is said to mark the first time a cloud-resolving simulation has been able to reproduce the formation of a tropical cyclone.
NCAR is part of a team now building a model similar to the MM5, but with more advanced capabilities, that will generate daily weather forecasts for the NWS beginning in 2004.
‘One of the remaining mysteries about hurricanes is how they form, especially when they’re influenced by midlatitude weather systems that move into the subtropics and tropics,’ said researcher Christopher Davis. ‘We hope that by analysing the mechanisms behind storm formation in these simulations, we can make hypotheses of tropical cyclone formation that can be tested using aircraft, radar, and satellite data. We also hope to understand what is needed to predict storm formation in operational weather forecast models.’
Operational computer models used for day-to-day weather prediction have become increasingly adept at projecting a hurricanes motion.
Yet even the best models have little skill in predicting intensity, especially the rapid strengthening often noted in the most powerful hurricanes.
Part of the problem is that the compact core of a hurricane, including the spiral bands of showers and thunderstorms that gather and focus energy, cant be modelled in sufficient detail on the computers and models used for everyday forecasting.
For instance, the finest horizontal scale resolved in operational computer models is 10-30 miles, but spiral bands can be less than 10 miles wide.
To see the eyewall and precipitation bands within a tropical cyclone, Davis and colleague Jordan Powers turned to the MM5, one of the worlds highest-resolution research models for reproducing storm-scale weather across a large area.
The model’s horizontal distance is as fine as 0.75 miles between computation points.
For their experiment, Davis and Powers studied Hurricane Diana, which struck North Carolina in 1984. Diana was chosen because of ample surface data and because a well-defined non-tropical low preceded its formation.
The MM5 successfully reproduced several stages in Diana’s development, from its original state as a non-tropical low to its intensification to hurricane status more than a day later.
The Weather Research and Forecasting Model, now being developed for future use by the National Weather Service, is designed to regularly operate with resolutions from 0.6 to 6.2 miles.
Together with more powerful computers, this will put the type of fine-scale detail in the MM5 into the hands of daily weather forecasters.