Why are BoM 3 month forecast models so pathetically WRONG

Just taking a quick look at the latest January to March rainfall and temperature three month Outlooks from the Australian Bureau of Meteorology (BoM). I have linked to the originals for rainfall and temperature. Note these forecast maps were issued on 17 Dec 08.

I have made the actual rain and temperature anomaly maps at this useful page. You can choose rain or temperature (max or min) for various periods or parameters. I am not saying medium term forecasting is easy but I am puzzled the BoM maps are often so EXACTLY wrong.

Rain forecast Actual rain deciles, blue wet, red dry
Max temperature forecast Maximum temperature anomalies
Min temperature forecast Minimum temperature anomalies

11 thoughts on “Why are BoM 3 month forecast models so pathetically WRONG”

  1. In Melbourne, four of the last few days of January 2009 were very hot. They set up conditions for the near-Melbourne bush fires on 7th Feb 2009.

    Here are some summary records of the monthly temperatures at the long-serving Melbourne Met office at Latrobe Street.

    To show how illusions can be created, note that:

    1. The minimum temperature for Jan 2009 of 15.7 degrees was the coldest for some years (5 years, 3 shown here).
    2. The maximum temperature for Feb 2007 at 29 degrees was higher than Feb 2009, 28.1 deg, the month that included the fire day.
    3. March 2009 monthly max of 24.3 is a lot cooler than March 2008 or March 2007.

    Of course, this selective extraction of information is nicknamed “cherry picking” and for good reason. It’s not scientifically desirable, but it’s interesting in terms of the perceptions that people gain. And of course, one station at Melbourne Central does not represent the State of Victoria.

    Tables follow.

    Melbourne

  2. Geoff,

    there’s no reason to quote the Latrobe St station at all. Just look at it:

  3. Also, at least one of the fires which killed people was deliberately started. Yes, the bush was bone dry but the rest was murder.

  4. Note that very station 948680 -378 -1450 35 MELBOURNE REGIONAL O AUSTRALIA is still listed by Hadley as being used in HadCRUT3. We have no way of knowing EXACTLY what part of the data they used but Jones et al in 1986 used 1941-1970 – an unusual selection, They specified the 1951-1980 period for many smaller Australian places in their 1986 TR027 documentation, sometimes they used other periods, Rockhampton was 1951-1970 and Kalgoorlie 1941-1980. The latest year, eg 1980 probably means that was the newest data at time of publication of TR027. So I would assume the great bulk of those stations have been updated over the years. But until everything is published, we are guessing.

  5. Do you actually know that all will be published, Warwick, or when? I think that extrapolating the little piece rates, the compendium would cost a buyer millions of $.

  6. Where I live there are 2 weather stations situated within 300m of each other. The manual station which is used to record temps for the historical record is located 5m from a tarred road and surrounded by buildings (a new house has been built in the last 12 months within 20 m). The AWS is situated on a grassed oval with no buildings within 50+m and open to the elements from all sides. The AWS temps are recorded on the recent observations BOM site for each day and the monthly average is recorded on the weather watch site and the monthly/yearly historical summary on the BOM site. I compared the monthly averages of the AWS and manual stations from Mar 08 to Feb 09 and found that there was a av max temp difference of nearly 0.7C higher for the manual station and an av min temp difference of 0.5C higher for the manual station. Thus one could conclude that the urban heat effect for Casino is around 0.6C for the mean temp. That about equals the rise in warming for the past 30 years. Any one have the same readings for similar stations.

  7. Assuming the two instruments are correctly calibrated, you have pinpointed an interesting micro-climate effect Ian. An illustration of how sensitive temperature instruments are to the immediate environment – and at a time of year when the UHI is not strong. I hope you keep up your monitoring through the winter, differences could be greatest on calm cold nights. To assess the size of the Casino UHI; you could compare data from the urban centre with a rural station, over as long a period as possible.
    I am interested too in the problem of a discontinuity throughout our data due to introduction of AWS, which are of course prone to breakdowns too. In many cases when AWS are installed, the old thermometers are discontinued.

    And a note for Geoff. I do not think there are any plans by Hadley or CRU to publish any more details of stations used. If any details have been published it is because they have been somehow shamed into it, eg. by ClimateAudit.s correspondence.

  8. Casino is unusual in that it has 2 stations so close to each other in basically different surrounds. I don’t know if there are any similar station positioning anywhere else but it would be good if someone out there is aware of anything similar and has the data available.
    I checked Casino’s 1911-40 30 yr average with the 1971-2000 average for both max and min temps and found that the mean temp for the former was 0.1C above the long-term average and the latter was -0.1C (thus the mean temp has cooled some 0.2C since 1911-400. I did a quick check for Sydney and Melbourne and found a quite large increase in mean temps during the same periods above.
    Murrurundi (in NSW Hunter) has had a drop of max temp of 1.0C and 0.8C min temp and Bourke has even dropped around 0.2C (mean temp) over this period.
    I would need to check more stations to make any assertions but it appears that larger urban areas have increased but this may not be so in smaller population centres.
    I used the 1911-40 averages as this was a period of rapid warming similar to the 1971-2000. We will need to wait for the 1981 – 2010 period to have a clearer picture.

  9. Most farmers I know find 3-month outlooks next to useless anyway. They would prefer one-month statements of probability. I find Ian Holton’s work much more accurate and reliable than the BoM’s because he takes into account indices of ocean temp anomalies and other factors that the BoM seems to ignore. No, I am not his relative.

Leave a Reply

Your email address will not be published.