‘Bet’ Office Forecasts

Guest post by Charles Duncan

As Christopher Booker fittingly writes:

There could be few more apt epitaphs for the year now ending than a recollection of the headlines in April that greeted a stark warning from the Environment Agency. Fuelled by the predictions of the climate-change-obsessed Met Office (and the the official policy, since 2007, of the similarly fixated EU) that we will have “hotter, drier summers” for decades to come, the agency foretold that the drought conditions of the early spring were likely to last “until Christmas and perhaps beyond”. The prophecy was swiftly followed by the wettest late spring, the wettest summer, the wettest autumn and the wettest Christmas we have ever known – eight months of near-continuous rain and floods amounting to England’s wettest year since records began.

Just because one year fits the ‘drier’ prediction…

UK rain map

2011 and 2012 rainfall: the North-South split flips.

The Met Office have excelled themselves in recent forecasts.  In September 2012 their 3-month outlook for October, November and December forecast a drier than usual period:

For UK-averaged rainfall the predicted probabilities favour below normal rainfall during October. For the period October-November-December as a whole the range of forecasts also favours lower than average rainfall.

The probability that UK rainfall for October-November-December will fall into the driest of our five categories is around 25% whilst the probability that it will fall into the wettest of our five categories is 15-20%, close to the climatological average. (The 1981-2010 probability for each of these categories is 20%).

Now you might think that after the floods we had in October and November they might have reviewed this.  But their outlook issued on 20th November promised more of the same:

Predictions for UK-mean precipitation for December and December-January-February show a slight shift towards below-normal values – consistent with negative North Atlantic Oscillation conditions – although the spread of probabilities is large. Consequently, for the season as a whole the chance of above-average totals remains significant.

The probability that UK precipitation for December-January-February will fall into the driest of our five categories is between 20% and 25% and the probability that it will fall into the wettest category is around 15% (the climatological probability for each of these categories is 20%).

I draw three conclusions from this:

  1. Their confidence in their ability to forecast is woeful.  With five categories the average is going to be 20%, so having 25% as the most likely and 15% as the least likely is a joke.
  2. They are truly dreadful at forecasting – as their September forecast shows.
  3. They don’t learn, as their November forecast shows; don’t these guys have windows in their offices?

[Or we can blame their use of models based on the ‘rules’ of a climate cycle that has now ended, V.]

But, hey don’t worry – they know EXACTLY what it’s going to be in 100 years! They predict that next year will be significantly warmer than this (now there’s a surprise!).

My own prediction is that we are in for 6 years of cooling, and at the end it will be about 0.25°C cooler than now.  After that warming will resume for another 30 years. Fancy hazarding a guess yourself?

This entry was posted in Opinion, Weather and tagged , , . Bookmark the permalink.

11 Responses to ‘Bet’ Office Forecasts

  1. John Robertson says:

    So its clear, their accuracy is only matched by their honesty?
    Happy New Year

  2. Verity Jones says:

    Something like that. LOL. They’re saying ‘warmer and drier’ is the long term trend due to climate models, but weather models seem to be infected with the same biases.

  3. Lars P. says:

    If their forecast is based on models they cannot learn unless they would use models capable of learning – continuously improving those – and the question is what part of the models are they able/allowed to update. Do they try to hindcast past weather, re-run the forecast and re-analyse, check why the models got it wrong?
    I fear their models have a built-in problem and are not capable of longer term forecast. It is the warming bias that makes them regularly run wrong.
    Not being capable to find the issue and fix it – as they do not search the error where they should – they are doomed to repeat again and again the same, irrelevant how big their computer is.

    A Happy New Year Verity!

    • Verity Jones says:

      Happy New Year!
      Models capable of learning are very powerful. I had experience of relatively simple ones back 15 years ago or more for optimizing complex fermentation reactions. They were very powerful, teasing out relationships between parameters that we had not foreseen and were very valuable in increasing understanding of the systems.
      Possibly the complexity of weather/climate models is greater, but I would suspect that the right learning feedback is not built into them. Standard models are only as good as the understanding of those designing them. Intelligent models should be designed with very few inbuilt relationships because they search for the relationships themselves, despite this many parameters should be provided and the computer learning will tease out the relationships. However the temporal feedback may be an issue. We found translating lags challenging.

  4. An idiot is defined as someone who repeats the same process, confidently expecting a different result from the last time.

    A Met Office long-range forecaster is someone who repeats the same process, each time from a different starting point, yet still expects the same result.

    A climate modeller is someone who repeats the same process thousands of times, getting different results each time, averages them, and calls the average “a robust projection”, though one and no more than one result can possibly be “correct”. Any “correct” result would obviously disappear in the averaging process.

    I’ll go with the idiot – at least he has the slim chance of getting his desired result by pure serendipity. The other two have no such chance.

    • Verity Jones says:

      Ah but how does the idiot know if the desired result has any value if he has arrived at it by chance?

      • He doesn’t have to know – he got what he wanted.

        In principle, I’ve got nothing against climate modellers nor their models. I think they’re what they were originally intended to be; useful tools (the models I mean!) for exploring aspects of the climate. What they are most definitely not, is any use whatever in predicting (sorry projecting), long-term climate changes. The vast spread of results over even a relatively short period is a clue writ large. Another clue is that the output from each short period in the run has to be constrained within entirely artificial limits to prevent the series shooting off in a highly unlikely direction.

        There are many papers in which authors, expecting runs of regional models initialised with historic data to produce some semblance of reality, report disappointment with the results. It seems they don’t do temperature, precipitation, wind speed, clouds or climate oscillations like ENSO at all well, even over relatively short periods. GCMs do better but when downscaled to regional level perform badly. If GCMS can’t model regions how can they model the whole?

        I’ve never understood how a regional model could possibly work – any region is bounded by other regions which must have a major influence on that region – winds, currents, cyclones, jet stream etc. Give me Old Moore’s Almanac any day.

  5. Verity Jones says:

    Here we go: Extreme rainfall in UK ‘increasing’ Roger Harrabin – http://www.bbc.co.uk/news/uk-20896049

    Upwards trend

    Extreme rain is defined as the sort of downpour you would expect once in 100 days.

    There are big swings in rainfall from year to year, but the overall trend is upwards since 1960. Last year, for instance, extreme rain fell around once every 70 days.

    The phenomenon of more frequent downpours has already been noted elsewhere, particularly in China and India.

    Scientists say that as the world has warmed by 0.7C, the atmosphere is able to hold 4% more moisture, which means more potential rain.”

Comments are closed.