How well are weather GCM doing?

Posted: March 24, 2015 by tchannon in weather

Some idea can come from forecast synoptic charts so I have put together charts sets for T+120, T+96, T+72, T+48, T+24 and the T+0 analysis.

In an ideal world these would be identical.


Figure 1, set of forecasts for Eclipse day, 20th March 2015, 12 hours.
Click image for full size PNG, 680kB

T+120 is top left, runs left to right then top the bottom, analysis bottom right or see image legends.

These are reasonable results at the same time as showing why even a 24 hour forecast is inadequate for many purposes.

The infamous rain problem tends to be played down by professionals. Test is a three hour forecast… seen that promoted? It is also only applies to an agreed list of stations, not locally to many people. WMO standard.

You might find this Met Office page useful, if hard work decoding the government speak.

Noticed? No show me, is all talking down, without hard evidence. Doesn’t even provide the list of sites.

I don’t know whether the Met Office has separate weather and climate mainframes, I hope so. Finer detail has been mooted but the input data is mostly woefully coarse, is not going to happen and anyway the exponent is high, number crunching needed capability goes crazy. None of this overcomes the basic flaws, not about average, it’s about worst case, the weather doing peculiar things.

These people are trying so knocking them is not good unless excessive claims are made and that mostly is sales droids, politicos.

There are many places where there are local perturbing conditions, traditionally known to real forecasters. These are few today, outposts are very few.

Climate rubbish, get rid of the lot, put them to work on weather or get out. This might address some of the protest within the Met Office over pay. It does say meteorological, nothing to do with climate. Climate, few people in one office.

Somewhere we have an infamous letter from the Met Office in reply to asking why it was raining when the forecaster said it was sunny. Apparently the forecaster was in an office block and didn’t know. Caused mirth here some years ago.

How far are we from useful medium term forecasts, did we see this coming?

Record breaking US winter sees resorts close early

Post by Tim

  1. Stephen Richards says:

    rog, they did say a year or two ago (2009?) that they use the same model for both climate and weather. i can’t believe that they would be running two massive systems;

  2. Stephen Richards says:

    There is also the knowledge that the weather has several long duration modes. Eg easterly winds in winter, atlantic depression runs in spring, anticyclonic summers.
    When inthese modes their weather predictions are more accurate but when the mode changes to ‘unstable’ they can’t seem to forecast ACCURATELY even 6 hours ahead.

    Then there is what they call accurate and what you and I call accurate. If you see a forecast which suggest dry sunny weather where you live and it’s pissing of rain, but locally, you think that is inaccurate.

    Ryan Maue at weatherbell does some accuracy work on medium range forecasts. I have been watching and comparing for years. None of the models is what I would call accurate and certain the long range models are all very poor. Even the UKMO website suggests that monthly and even 10 forecasts are a bit of a guess.

  3. oldbrew says:

    Wasn’t the 2014 El Nino hype based on models? Nothing happened.

    ‘The world is almost certain to be struck by the “El Nino” phenomenon this year [2014], with the potential to induce “major climactic impacts” around the world, the Met Office has warned.

    India and Australia are likely to be hit the hardest but the fallout could also be felt in Britain.’

    ‘almost certain’ turns out to be more Met Office wishful thinking.

  4. oldbrew says:

    More from the same news report – alarmist fear-mongering highlighted.

    ‘“El Nino leads to extreme events and has a pronounced warming effect,” said Michel Jarraud, head of the UN’s World Meteorological Society.

    An El Nino event occurs when the prevailing trade winds that circulate over surface waters in the tropical pacific begin to weaken. This allows warm waters of the western pacific to wash back eastward, dragging precipitation with it, leaving countries like India, Indonesia and Australia drier, and increasing the chances of wildfires and lower crop production.

    Meanwhile, the phenomenon precipitates heavier rainfall in the South American and eastern Pacific nations, raising the prospect of floods and landslides.’

    Did any of that happen?

  5. suricat says:

    Hi TC, you’re right. As soon as something happens that wasn’t in the model’s programme the model diverges from an accurate representation, thus, the model becomes a divergent predictive tool and inaccurate for the purpose that it was designed for.

    Parametrisations and re-runs can put a model ‘back on track’, but the ‘hind sight’ of a comparison with the way that Earth re-configured its systems in ‘real time’ is needed to permit this to be done with any accuracy (assuming that all attractors are understood).

    The next problem is ‘definition’. Its no good looking at a 10 kilometre grid if all the cloud, precipitation and convection only exists at a 1 kilometre diameter. This just gives ‘signal mush’! If a 1 kilometre grid definition is required, the resolution can only be achieved from a ‘grid’ 1/3 (9 ‘pixels’/kilometre ^2) the dimension of what definition is expected. A 10 kilometre grid can only be a ‘statistical assumption’ and ‘grossly unstable’!

    However, I wouldn’t ‘hit on’ the guys that are involved with these models. They are hampered with the ‘run-time’ needed for our current super-computers and doing the best possible job with inadequate tools.

    IMHO these models are totally inadequate for the use of ‘prediction’ for ‘climatological purposes’. The ‘shorter their prediction’, the ‘more accurate their prediction’ per se.

    Best regards, Ray.

  6. tom0mason says:

    Maybe the Met Office should link to here for a better outlook of the weather.