Prospective and retrospective predictions

In first-year psychology lectures, I have sometimes referenced a local weather event as a way of introducing the concepts prospective and restrospective prediction. In weather, as in psychology, you are trying to make informed guesses about what will happen in the future, or what has happened in the past. “It will rain tomorrow, because it is raining today” is a prospective prediction; it is about something that has not yet happened but could, and is based on prior observation. “The footpath is wet, so it must have rained earlier” is a retrospective prediction; we already know something has happened, and we are trying to determine what led to its occurrence.

Many areas of Australia are enduring a period of drought. I was pleased to read recently that meteorologists are predicting a wet, La Niña Summer. If their prospective prediction turns out correct, then the likelihood that the Summer-after-next will be wetter than average will change. In other words, weather models modify their predictions in line with new evidence.

Speaking of evidence, how can we know that La Niña has actually happened or not happened? Well, we can compare data collected this year (e.g. rainfall statistics) with data collected in previous years, and search for a significant pattern. In other words, predicting climate change involves both a prospective and retrospective fitting of data to theory. This point is well encapsulated by a comment made by the Queensland Premier, Annastacia Palaszczuk, in the wake of Tropical Cyclone Marcia:

This is something that they have never seen before as well, going from a low pressure system to a (Category) 1 all the way up to a 5… Now, they’re going to go back and look through all the research and try to work out how that happened so quickly. (Brisbane Times, 22nd Feb. 2015)

Indeed, despite early warnings that Marcia was intensifying out at sea and headed for Northern Queensland, the extremity of its intensification was a complete surprise. The aftermath was quantified and used to retrospectively classify the cyclone as Category 5. Despite the inaccurate prediction of Marcia’s intensification, the cyclone’s death toll was zero, because residents were prepared for the worst. Cyclone Marcia teaches us the value of prospective analysis, fraught with error as it may be.

In Australia, the Bureau of Meteorology (BoM) “operates a nationwide network of over 60 Weather Watch radars.” (BoM, 2017). These have varying levels of sophistication, with the best of the lot employing high-resolution Doppler radar. The nearest Weather Watch to me is located just North of Grafton. It looks a little something like this:


Below is a radar image of where I live, courtesy of the Grafton radar. Specifically, I live at the little black square directly to the right of the twin storm, pictured:


Over the past few months, it has hardly rained at all in this area. I worked from my home office last Friday (22nd Sep.), and was surprised to hear thunder at around AEST 13:00. At AEST 13:26, the Bureau issued an amber alert for large hail and damaging winds for people in parts of Northern Rivers, Mid North Coast and Northern Tablelands forecast districts. I had decided to take a lunch break at that point. Around AEST 13:40 it started to pour – this must have been because of the washing that had been put out in the morning. Of course, the rain did not last long enough – by the time my lunch break was over, the sun had come out again. (This explains the scumbag hat.)

Relative to the position of the Grafton radar, I live about 92 kms to the direct South. The BoM’s notes on the Grafton radar suggest that I am well within range for accurate thunderstorm mapping. Even so, the then-prospective prediction of large hail and damaging winds seems to have been a false alarm in retrospect. Accurate diagnosis (there was a pair of storms, as imaged by radar), but inaccurate prognosis. The BoM were probably right that hail and winds happened somewhere in the region at sometime, but the temporal and spatial exactness of this prediction was lacking. Just like the tornado that tore through Kurnell (Sydney) in December 2016, the twin storm’s brief existence was practically impossible to predict beforehand. In the case of the Kurnell tornado, the evidence (such as a 213km/h wind gust – the strongest on record for NSW to date) supported the retrospective hypothesis that a F2 tornado did in fact take place. In the words of NSW severe weather forecaster Michael Logan:

The intensity of the storm as detected on radar, combined with the peak wind gust recorded at Kurnell and narrow damage path, suggests it is extremely likely a tornado was associated with this very severe thunderstorm. (BoM media release,  17th Dec. 2015)

In conclusion, the weather gives us some interesting examples of two general types of hypothesis: prospective and retrospective. Volatile, localised weather phenomena are subjects in the retrospective realm of prediction. We can categorise weather phenomena once the damage has been done, but you can forget about predicting that a F2 tornado – or indeed more than five seconds of rain from two close-by storms – will hit your town on any given day in the future.

Leave a Reply

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s