How Well did Long Range Weather Forecasters Do this Winter?

Spring is finally coming.  As I write, I can see the roses in my garden finally growing, and thumb-sized bumblebees crowd the glass at my window.  Now that the cold weather is retreating, I thought it would be fun to revisit my post from December about the dire weather forecasts made for the UK this winter.

As avid readers will remember, I discussed the Daily Express’ hair-raising headline about “Record Breaking Snow“.  In November, James Madden of Exacta Weather forecasted

a period of prolonged, extreme cold…snow drifts of several feet and long-lasting snow accumulations on a widespread scale.

The Met Office, making a statement around the same time, were more conservative:

While we have seen a return to more normal, cooler temperatures for this time of year, this is no indication of what we can expect over the next four months with regards to temperatures and when we might see snow. It is far too early to tell.

So what weather did we have?

At first glance, it seems obvious that the apocalyptic predictions above did not come to pass.  It is certainly true that the exceptional, extreme cold alluded to above did not materialise.  From the Met Office’s assessment of the UK’s Winter 2013/14:

Mean temperatures over the UK were well above the long-term average for all three months with a mean winter temperature of 5.2 ^oC which is 1.5 ^oC above the average and the fifth highest in the series. There was a notable absence of frosts, and the lowest UK temperature of the winter, -7.7 ^oC at Altnaharra, Sutherland on 17th February was the highest such winter value for at least 50 years.

What is true is that this winter has been one of the wettest on record, with the UK experiencing 1.5 times the expected rainfall, with the vast majority falling in the south-east.  At high ground, this precipitation fell as snow, providing domestic ski resorts with more snow than their international competitors.  The Met Office Outlook for 2013/14 gave a 15% chance for the winter being extremely wet.  Note that this doesn’t mean that the Met were “wrong”.  If they had ascribed a 0% chance of wet weather, then we might be justified in saying they made a mistake.  This distinction goes right to the heart of what “probability” means mathematically (if you want to disappear down that rabbit hole, here is one of many entrances).

This winter has also been exceptionally stormy, with more days of severe gales than any winter since that of 1871.  Combined with heavy rainfall, storm surges and floods have devastated the south of England, with the UK government pledging £340m to help rebuild the shattered infrastructure.

What did the Forecasters say?

James Madden of Exacta has reviewed his forecasting season on his website (pdf).  He notes his earlier prediction of stormy weather, and does address his incorrect prediction of extreme cold:

72 of the last 100 Novembers were also ‘warmer’ than the one we experienced in 2013. So although we never experienced the record-breaking cold and snow episodes that were originally suggested in my long-range forecast for November, we did still experience a rather cold month with a number of snow events across the country.

It’s not clear to which 72 Novembers Madden is referring.  The fact that most forecasts use the 1981-2010 period for comparisons (which comprise 29 of the last 100 winters) is telling, but without further evidence we can’t say much more about the above statement.  The rest of the document still contains only oblique references to the forecaster’s methods of generating predictions, which means we still can’t say anything about their veracity.

The Met Office maintains:

It’s not currently scientifically possible to provide a detailed forecast over these long timescales.

This statement seems to be borne out by the events of the winter.  In the last page of his review, Madden posts a table of his “accuracy” for summer and winter forecasts, to which he assigns a percentage.  Posting metrics for accuracy and inaccuracy is a good thing, and something that many futurists state is necessary if predictions are to improve.  However, without knowing how a metric is calculated, it’s hard to be convinced of its value.  When the majority of weather forecasting is qualitative (“rain heavier than average”, “fair chance of sunshine”), how do you turn that into something quantitative? It would be nice to have some clarity on this point.

What’s the Bottom Line?

As I said last time, long-range weather forecasting is an inexact science.  As Madden says:

To get the weather right all the time would be quite some feat.

At the bottom of their 3 month Outlook, the Met Office say

The Outlook should not be used in isolation but should be used with shorter-range and more detailed (30-day, 15-day and 1-to-5-day) forecasts and warnings available to the contingency planning community from the Met Office.

So, it’s important to make sure when listening to dire forecasts of weather in a few months time, to check back with the forecasters nearer the time, as their predictions become more refined.