02 April 2020

O/T: Predictions of Weather and Health

On Thursday Prime Minister Trudeau looked evasive on declining to release model results in the face of questions from many of the media. Why?

As someone who spent part of my career making forecasts, in my case of weather, and getting slammed when they didn't pan out, it's interesting to see epidemiologists struggling to explain modelling and why there are such great differences in predictions of the evolution of the present pandemic. In the US is it 100,000 deaths, 200,000 deaths or 2.2 million deaths?

A large part of the problem is not knowing the present situation. For weather forecasting, we can't monitor conditions everywhere. Small-scale atmospheric phenomena not captured by the observing systems can, under the right conditions amplify and change the whole weather map over the course of time.

The health parallel is infections not well captured by the data. Evidence is mounting that many more people have the virus than are identified; they don't have noticeable symptoms. Iceland reports 50% of those tested had the virus but showed no symptoms. In Canada, for the general population we only test those showing symptoms and turn up to be tested, but those mildly infected can pass the disease on to others.

In weather forecasting technology has helped fill the information gap — weather satellites, radar, lightning strike detection systems and low-cost weather stations. Can technology help for health?

There's also uncertainty because of introduced variability. For weather, the behaviour of a butterfly, whether or not it flaps its wings at a particular time and place, could produce a vortex that might amplify into a storm, That puts a limit on how far in the future you can make useful weather predictions.

The health parallel is not knowing how people are going to behave. Will they self-isolate and observe advice on physical distancing? Or will they carry on without regard to the situation?

There are also uncertainties about mechanisms. When electric charge builds up in a thunderstorm why does lightning strike in one location and not another? Why do some people react strongly to the coronavirus while others seem immune?

Despite these problems weather forecasts have been issued for decades because the imperfect information was still of value. Hopefully, the Prime Minister got the message through the media that even imperfect forecasts should be released.

1 comment:

Denis Bourque said...

Thanks, John. Well said.

Like John, I’m also a retired meteorologist. In fact, I worked with John and even for John for a while. The last 15 years of my career were spent in policy trying to explain to non-scientist government official the value of weather forecasting.

What few understand is that weather forecasting came into existence in the 1860s-1870s for the sole purpose of saving lives and protecting property by tracking and forecasting severe weather; in other words, the bad stuff. That continues to be its reason for existence: tracking storms, hurricanes, tornadoes or any weather that could be a threat to life and property. As we know, these don’t happen everywhere everyday - so we benefit, on the side, from fair weather forecasts which we use to plan outings, picnics, etc. But that is a spin-off of the need to be weather-vigilant every day.

I’m still astounded how much the science of weather forecasting improved during my nearly 40 years of involvement. Way back when, weather offices produced a single forecast in the office which they issued to the public. That is no longer the case. These days, weather services produce many versions of the forecast within the computers (too complicated to explain) and then merge them into the one that is issued to the public. But the fact remains that weather services have a worst case scenario forecast, a best case scenario forecast and many in between. Merging them into the “most likely situation” is the skill that comes from experience.

Epidemiologists have similar techniques including models with different assumptions resulting in forecasts of best outcomes to worst outcomes. Most of the difference depends on human behaviour. Meteorological agencies put out their forecasts in the hopes of modifying human behaviour to save lives (e.g., fly around a storm, don’t set out to canoe on a lake when thunderstorms are possible, avoid travelling when a winter storm is coming.) Seeing epidemiological forecasts from reputable sources, e.g., with the endorsement of the government health agencies, would hopefully have the same impact of modifying behaviour on our society in these difficult days, or at the very least confirming to those that have already taken action that they are doing the right thing. - and to stick with it for the long haul. Like John says, I hope the decision is to release these.