Met Office: poor accuracy on predicted rainfall

Guest post by Charles Duncan

I have often claimed the Met Office is unable to forecast weather more than two or three days out.

I like to have data to support my assertions, so I thought I would see how well they did predicting last year’s rainfall.

Each month they publish a 3-month outlook, with estimates for rainfall for the first of the three months and the three month total. I took the first month’s data, so the test is of their ability to forecast in a window from one to six weeks out.

I plotted this against their own data for actual UK rainfall as a scatter plot. The result shown in the graph has a correlation coefficient of just 0.055.

Enough said!

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

10 Responses to Met Office: poor accuracy on predicted rainfall

  1. Doug Cotton says:

    [OT – sorry Doug – you don’t get to post here on every thread about your own area of interest when it is not related to the subject area – Verity]

  2. argylesock says:

    This seems like a good example of how climate change means unpredictable weather.

    • Verity Jones says:

      Or it could be that the Met Office’s forecasting takes too much account of patterns they view as ‘the new normal’ under increased CO2, when in fact, in a cooling world, we have a different set of drivers now operating.

  3. Willis Eschenbach says:

    Piers Corbyn is always on about how he can beat the Met … doesn’t look too hard from here.

    For fun, you should compare the actual outcome to the “climatology”, the long-term average for that month. See if you could do better using climatology.

    Or even better, use the three previous overlapping three-month periods as predictive variables, and use the predictions from that “naive” system. See if that can beat the Met.

    Always fun,


  4. Bloke down the pub says:

    Willis Eschenbach says:
    January 17, 2013 at 10:13 pm
    Or you could just use a coin that isn’t weighted to read hot.

    It has been mentioned before, possibly here, that Met office forecasts would benefit from including a measure of how accurate their previous forecast turned out. I think maybe the Canadians do this?

  5. I called a couple of my relatives who live in the UK on this one. One told me that it is bitterly cold in Bournemouth today but she feels quite happy with the Met as this was in their predictions a few days ago. She says that the Met is truly wonderful in the short term (up to five days) but pitiful when forecasting seasonal or longer term trends.

    If the Met diverges from reality when it comes to forecasting weeks or months ahead, why would we give them any kind of credibilty when they claim that the climate will be warmer decades or generations into the future?

  6. Verity Jones says:

    Yes – must commend them for the short term forecasts in the last few days – spot on! It’s the longer terms stuff that is problematic. The issue is perhaps the apparent certainty of the prognostications.
    @Bloke down the pub
    Your suggestion of reflections on accuracy is a good one – I don’t recall it being mentioned here previously.

  7. Pingback: Met Office Accuracy II | Digging in the Clay

Comments are closed.