RealClimate logo

Greenhouse gases help seasonal predictions

Filed under: — rasmus @ 30 November 2005 - (Français)

The conventional wisdom in meteorology has been that certain factors such as the complete oceanic state and the exact concentrations of greenhouse gases are of minor importance for a normal weather forecast. Moreover, whereas sea surface temperatures (SST) are important, the deep sea temperatures are believed to have little impact for predictions for the next few days. The reason is that the ocean reacts slowly to changes in the atmosphere (has much higher inertia and much higher heat capacity). Hence, the most important information needed for such a weather forecast is the atmospheric initial conditions, a description of what the atmosphere and the SST look like when the weather model starts computing the weather evolution.

If we want to make predictions over longer periods, such as the subsequent months, then the effects of boundary conditions become more important than the atmospheric initial conditions (due to internal chaos and the lack of predictability from the initial state). The oceans, which provide part of the boundary conditions for the atmosphere, also have sufficient time to change appreciably, and the change in the oceanic state must be taken into account. Such long forecasts are often referred to as seasonal forecasting. Seasonal forecasts are often made with coupled ocean-atmoaphere models (more like climate models), as opposed to atmosphere-only models for ordinary weather forecasts. Until now, it has been assumed that factors such as greenhouse gas concentrations play a minor role for the skill of the seasonal forecasts. However, a new study by Doblas-Reyes and co-authors at the European Centre for Medium-Range Weather Forecasts (ECMWF) – the ‘European weather centre’ – suggests that by updating the greenhouse conditions with the observed annual mean values, the skill of the seasonal forecasts improve noticeably.

The implications of these findings are quite subtle, but important: we are already feeling the climatic effects of rising greenhouse gas concentrations. These findings also open for more practical use of our knowlegde about climate change: by taking these trends (i.e. extra information) into account, we improve our ability to make predictions seasons ahead. This is also something that has been observed, e.g. in Norway, where the climatic trends result in more warm than cold seasons.


Doblas-Reyes, F.J., R. Hagedorn, T.N. Palmer and J.-J. Morcrette:
Impact of increasing greenhouse gas concentrations in seasonal ensemble forecasts.
ECMWF Tech. Mem. #476, October 2005

11 Responses to “Greenhouse gases help seasonal predictions”

  1. 1
    Tom Rees says:

    Reminds me of the UK met Office annual predictions, which forecast annual global temperatures based on atlantic multidecadal oscillation, ENSO, solar, recent volcanic activity and, crucially, radiative forcing due to GHG.

  2. 2

    Yes, that is a good way to make a long term forecast, and it appears lacking in some current day models.
    I would force feed NASA GISS GT (or other reliable) prediction as a constant and see if the weather forecast comes up with something better. May be this is done in some cases, I don’t know. What I know is that current seasonal forecasts seem to be oblivious to GT status and projections. Case in point, this year’s all time high GT seems forgotten in latest seasonal forecasts calling for a colder winter than last year.

  3. 3
    Steven Knudsen says:

    Very interesting. So really there is no such thing as climate in the sense of average or mean conditions, but more like a trend in climate. So one must decide how often to take averages. Too long, and the climate trend is missed. Too short, and the data is too noisy.

    Good luck!

    [Response:Basically, the term ‘climatology’ is used to mean the ‘mean climate’ and provide a kind of reference level. But when the climate changes – one definition of a climate change is a change in the probability distribution (pdf) of a climate variable – then it’s clear that a given base line may not always be appropriate. One could use a sliding time window (i.e. always use the last say 30 years), but as you say, one may run into problems as the trend may not be captured or the interval is too short to give a sufficient statistical representation. Take Oslo for instance: the temperature between 1961 and 1990 is not very representative for the last decade (which is warmer). Usually, a pdf assumes an iid (independent and identically distributed) variable, but a non-zero trend implies a non-iid process: the data is neither independent (chronological order matters) not identically distributed. The positive side is that once the trend is known, we have more information about the process than if there were not trend. Hence, the predictive skill is boosted by the trend. -rasmus]

  4. 4
    Pat Neuman says:

    … “by taking these trends (i.e. extra information) into account, we improve our ability to make predictions seasons ahead”.

    Trends in the Midwest in winter have been for warmer temperatures, higher dewpoints and increases in rainfall. Should that extra information be put into probabilistic river forecast products? How?


  5. 5
    Pat Neuman says:

    More questions

    The answer to the first question in #4 seems obvious, yes it should be put in.

    What if the answer to the second question (how ?) is complex.

    If how? is too complex than maybe it should not be put in now, but when?

  6. 6
    Pat Neuman says:

    Weather forecasters may be caught between a rock and a hard place.

    Overnight minimum temperatures rarely get as low as the weather forecasters predict. Oddly enough, that’s been going on for years. Why? … If weather forecasters deviate from model guidance they need to give an explanation. But weather forecasters have been warned not to talk about climate change or global warming. That explains why they’re overnight forecasts have been biased on the low side, for years.

  7. 7
    Pat Neuman says:

    How come so few people are commenting on this topic (Greenhouse Gases Help Seasonal Predictions). Maybe if the topic somehow involved hurricanes then some people would comment. But!! the topic “Greenhouse Gases Help Seasonal Predictions” does involve hurricanes!!

    Hurricane outlooks that are made before hurricane seasons begin ARE seasonal predictions.

    Ken Robinson (#28 of next topic) was correct in his remarks that US government scientists involved in predicting hurricanes are claiming natural cycles are solely responsible for the strong hurricanes in 2005.

    Their claim is not surprising … US government scientist’s involved in predicting weather and hurricanes do not support the evidence that global warming is happening due mainly to greenhouse gases accumulating in the atmosphere from anthropogenic sources.

    So, if Ken heard that US government’s scientists are claiming that global warming is not happening due mainly to greenhouse gases accumulating in the atmosphere from anthropogenic sources; what would Ken say about global warming happening?


    I’m not meaning to be picking on Ken, just trying to bring up a point).

  8. 8
    Pat Neuman says:

    Changes (increases) in rainfall intensity, temperatures, humidity and rate of snowmelt and have occurred in recent decades, due in part to increases in CO2 concentration. Despite the observed changes, there has been a lack of research and development needed to incorporate the changes into hydrologic models for hydrologic forecasting and hydrologic outlooks. See:

    Excerpts from articles (below), have identified significant changes in rainfall intensity, temperatures, humidity and rate of snowmelt:

    … “In general, the results indicated an area of increasing frequency and/or intensity of heavy storms along an axis extending from Missouri north east ward through Illinois to southern Michigan and northern Ohio. The increases appear to be greater than expected from climate variability and sufficiently large to have an impact on water-control structural designs and other aspects of applied climatology”. Rainfall Frequency Atlas of the Midwest by Floyd A. Huff and James R. Angel, Illinois State Water Survey, Champaign, IL, Bulletin 71, 1992

    “Especially since about 1950, that the incidence of these extreme floods has been increasing way more than you’d expect from just chance alone,” Knox said. … He studied the sediments from the Mississippi River beds, which provided him evidence of flood activity as far back as 7,000 years ago. … [James Knox, professor of geography at the University of Wisconsin, Madison]

    … “dust emissions have a wide impact on climate and weather, from modifying rainfall thousands of miles away, to influencing hurricane intensity and affecting air quality”. Heavy Rains Can Make More Dust In Earth’s Driest Spots

    … “Conclusions on the Timing of Snowmelt Runoff and Humidity
    1) Trends were shown for recent earlier in the year annual snowmelt runoff at three river stations within the Northern Great Plains and Upper Midwest.
    2) Trends were shown for recent increasing dewpoint averages for January, February, and March but not April.”

  9. 9
    Michael Jankowski says:

    Re#6, If you squint hard enough, you might be able to get a glimpse of those guys with AK-47s pointed at the Weather Channel people – just in case they slip up.

  10. 10
    Stephen Berg says:

    Re: #9, “Re#6, If you squint hard enough, you might be able to get a glimpse of those guys with AK-47s pointed at the Weather Channel people – just in case they slip up.”

    Now that is out of line. It’s not about AK-47s, but advertising revenues.

  11. 11
    Lynn Vincentnathan says:

    Re #7. I was at a conference for a week, so I missed this post. But I agree, this is an important topic.

    I often wonder, when weathermen say, “…above (or below) the average…” if they are averaging the weather stats up to the present day, which would include the impact of GW. So, I guess the average keeps getting a bit higher. But the viewer probably assumes the average has been a constant over the last 100 or so years, so they have no good comparison point to get a sense GW is happening. And since no one is telling them about GW in the media, they really only have daily weather reports to understand what might be going on. So in a way, if adding in GHGs helps in weather prediction, one negative outcome would be that weathermen will be able to say, “…as predicted…,” giving viewers a false sense that everything is normal & under control & GW is not happening, since I doubt they will mention that GHGs were used to help them make such good predictions.

    We might be better off with weathermen saying, “…well, this is totally unpredicted….this is really weird weather…” Or, better yet, have the media start telling the truth about GW.

    BTW, I know a weatherman whose brother is (was) in oil, which I found out much later. He did mention GW, unlike the other weatherman on the other channel in the area (who said nada about GW) — he constantly said it was NOT happening (at least up through 2002, when I left the area). My friend & I thought he seemed like a very nice man & we wondered why he was telling us garbage about GW, since it is vital that we laypeople know about this serious problem, so we can solve it. Then we found out about his brother….