NOAA tweaks their poorly performing Climate Forecast System model to remove ‘cold bias’

‘Cold bias’ removed from Atlantic

From NOAA/NWS technical information notices (TIN) (h/t) to Dr. Ryan Maue and Larry Kummer. It seems the climate models can’t quite zoom in on the short term future of a region of the globe. Literally, they are “all over the map”:

enso-models-map

What will the future hold for ENSO? Who knows? These models aren’t in good agreement for the next year. NOAA says they are fixing CFSV2 to “reinitialize” it. Now it looks like cooling ahead in the summer with a bounce in winter:

nino12Sea-3-29-16

nino34Sea

The PDF referenced below is really quite something to read. Note the before and after the “fix”.

More at WUWT ENSO Forecast Page: https://wattsupwiththat.com/enso-forecast-page/


 

Here is the technical bulletin describing the change.

Technical Implementation Notice 16-09

National Weather Service Headquarters Washington DC

1205 PM EDT Fri Mar 25 2016

To:       Subscribers:

-NOAA Weather Wire Service

-Emergency Managers Weather Information Network

-NOAAPORT

-Other NWS Partners, Users and Employees

From:     Tim McClung

Chief Operating Officer

NWS Office of Science and Technology Integration

Subject:  Reset of ocean initial conditions in the operational

Climate Forecast System version 2 (CFSv2) to remove

Atlantic Ocean cold bias March 28, 2016

Beginning with the 0600 Coordinated Universal Time (UTC) model run on Monday, March 28, 2016, the National Centers for Environmental Prediction (NCEP) Central Operations will implement a fix to the Global Ocean Data Assimilation system used in the Climate Forecast System (CFSv2). This fix is expected to remove an erroneous cold anomaly in the equatorial and South Atlantic Ocean temperature.

A large temperature cold bias emerged in the ocean initial conditions along the equator in the spring of 2015 and grew to include the South Atlantic in the summer of 2015. These cold biases have translated into cold biases in the same regions in CFSv2 forecasts. The ENSO forecasts from the CFSv2 from December 2015 show unusual behavior, in that the El Nino event does not transition to neutral but rather experiences a double-dip El Nino, which is not realistic. A fix has been tested that re-initialized the ocean initial conditions in the operational CFSv2 from an offline GODAS, which uses a weak relaxation to climatology (the NODC WOD09) to control the noise.

EMC tested this re-initialization strategy for 4 initial months (December 2015 through March 2016). Results can be viewed here:

http://www.nco.ncep.noaa.gov/pmb/changes/downloads/CFSv2_Atlantic_cold_bias_problem.pdf

EMC found that there was no Atlantic cold bias in the initial states as compared to current operational CFSv2. The long lead forecasts evolved from the current El Nino event into neutral or La Nina conditions during the next 9 months in the Nino3.4 SST plumes.

Based on the positive results of the CFSv2 tests and the severity of the cold bias anomaly in the Atlantic Ocean, NCEP will implement this change outside of the normal upgrade process. Users do not need to take any action and will continue receiving data as they do today.

For questions regarding the correction to the CFS GODAS and resulting improvement to the Atlantic cold bias please contact:

 

Suranjana Saha

NCEP EMC/College Park, Maryland

301-683-3727

Suranjana.Saha@noaa.gov

 

NWS national TINS are online at: http://www.weather.gov/os/notif.htm

 

NNNN

52 thoughts on “NOAA tweaks their poorly performing Climate Forecast System model to remove ‘cold bias’

    • “””””….. This fix is expected to remove an erroneous cold anomaly …..”””””

      I thought an anomaly was already erroneous ??

      g

    • So they are saying that by August their range of erroneous information will be about 60% of the full scale range of possible data (six out of ten squares).

      Well I suppose in some lines of research it is ok if you get the order of magnitude correct.

      I seem to recall that in the derivation of the Planck black body radiation formula, that we did in school, or maybe it was Sir James Jeans calculation of the specific heats of solids at low temperatures there is a calculation where you are prevented from simplifying the expression by a pesky factor. You can complete the simplification process and proceed to the generally accepted answer if you just say that the value of that nuisance factor is near enough to one so you can just make it one, and it will go away so you can take the square root of what is left.

      So that is exactly what you do.

      The only problem is that the problem factor that you said is near enough to one, is actually Avogadro’s number; 6.023 E+23 or thereabouts.

      And yes in this case it is close enough to one, because the quantity that it is multiplying, turns out to be factorial Avogadro’s number.

      So their is almost, and then near enough, and then there is roughly. And in that case, Avogadro’s number is roughly one.

      G

    • Does anybody ever get fired for publishing these erroneous data in the first place so that somebody else has to come in and fix it after everybody has already used the information for what it was supposed to be published for ??

      It seems like there is no accountability for these screw-ups.

      Well it is the modern apologisation of politics, isn’t it. I screwed up and I accept full responsibility for the foul up, and the fact that I fully accept the blame for this snafu should satisfy you that we really care about what we do in this job.

      Um, er, no, I am not going to resign since I already accepted responsibility for the misteak.

      What a bunch of charlatans.

      G

    • At this day and age with the coverage we have, a blip leads to a 2-degree bias. I say it is time to expand the error bar of the past. Uncertainty shouldn’t encompass what we know or understand. Kudos, though, to the professionalism shown in finding and reporting this.

  1. Have they adjusted CO2 measurements to fit their theory one wonders, everyone but my granny has!

  2. Joe Bastardi mentioned that a week or so ago on his Atmospheric Avenger. He detailed the problems that non-existent cold pool caused, including how it stopped the CFS V2 from seeing a collapsing el Nino and a developing la Nina.

  3. Damm that NOAA poorly performing Climate Forecast System model – Why can’t we just have the power to say trust me, instead of having to publish pesky graphs that on 2nd look alway make us look like crooked adjusting fools?

  4. I have often wondered if the security agencies or the military realize how wrong NASS and all their products always are. I have wondered if the spooks know that it is all a delusion and that it prevents us from getting real projections that might be useful.

    Then I realized they are spying on China and Russia where real science is produced in this modern age and get their intel from outside the Western lunacy over CO2. Well, that is my speculation at least.

  5. I used to play board games with friends and there was always one who changed the rules in her favour, and I say ‘her’ because… he always got slapped… sorry girls, put your thought and complaints into a self addressed envelope please.

      • Love the new graphs showing all possible weather outcomes but wouldn’t my weather predicting rock be a lot cheaper and it accurately predicts the weather 100 percent no need for graphs or computers .
        If the rocks wet it’s raining
        If the rocks dry it’s not raining
        If the rocks covered in cold white stuff it’s snowing
        If the Rock is moving it’s very windy
        If the rocks gone it’s a hurricane etc etc

      • …Hey Sparks, get back in the net ! No wonder we’re always losing, sheesh ! ( sorry, bad Leafs joke )

      • Hey Robert it sounds just like the rock that I have; but I haven’t actually used mine to tell the weather or climate.

        I use mine mostly to make soup for practically nothing.

        You just boil the rock in a pot of ordinary water over a fire, and it turns into soup. It’s magical the stone acts like a catalyst, so you get your stone back intact after you eat the soup.

        You can use a selection of condiments to alter the taste a bit; but you really don’t need anything else, to make stone soup, besides a nice stone, and a pot of water.

        I’ll give mine a shot at telling the weather; but I suppose you have to put the rock outside to do that.

        g

  6. Climate science has become a (sp?) of SNL, circa Eddy Murphy, Gilda Radner, and Bill Murray. Caddy Shack comes to mind; this model revision reminds me of the gopher. One of the greatest golf movies ever. Hint: another is Legend of Bagger Vance.

  7. The question is: did they actually fix the model so that it properly predicts unknown events? Or did they just tweak it to fit the existing data, which means squat for prediction accuracy?

  8. The seasonal bounce is wholly because of the DMI Arctic temperature which has been 10C higher in winter but unchanged in summer (in another thread I pointed out that it would be worthwhile to audit the changing DMI algorithms). Last summer was warm because of the El Nino, so they’re projecting that that won’t be as prevalent next summer. Lots of fancy graphs and big budgets for simple inputs.

  9. 1) There are very reliable models in this order 1) ECMWF 2) JAMSTEC 3) UKMET 4) SCRIPPS They have done the best job at forecasting the enso trends, and those 4 have the La Nina that is coming due to the physical realities present ( one should be people simply looking at what has happened in the wake of the last super nino, and many other ninos) and two) the very warm water surrounding Australia which encourages the increased low level easterlies needed to bring the cold water up that is evident in the sub surface temps. three) Why do so many people run to a model that is notorious for its poor convective feedback in the tropics. I see both sides showing the CFSV2, the AGW people trying to stop a coming La Nina , our side for instance not understanding that it has fatal flaws known to most people in the forecasting industry. Close individual examination of each model, against previous enso events, which is grunt work, rather than displaying CFSv2 graphics, will help to come up with a better answer. In the meantime, the very fact that the model had such a cold bias in the tropical Altantic, which would affect the entire tropical feedback and lead to the model insisting the el nino would try to continue, should raise alarms about the prowess of this model. I see no problem with its initialization of air temps as we all see how warm they are on the model, at record levels, but its failure here and a needed fix IMO is putting a band aid on a fatal wound, which has to be the convective feedback in the model in the tropics, and that goes for the GEFS through the entire US model suite. Until the day that is fixed The euro overall will continue to rule the roost. BTW there is a huge day 10 fight between the US ensemble and the Euro If the past is any indicator, the Euro will win. again.

    • The Euro is prog’ing El Nino effects to be nearly nil for the rest of Spring, out here on the West Coast. From that one might indeed surmise rapid fall into La Nina.

  10. Why do so many people run to a model that is notorious for its poor convective feedback in the tropics.
    Because computers. “If you put tomfoolery into a computer, nothing comes out but tomfoolery. But this tomfoolery, having passed through a very expensive machine, is somehow ennobled and no-one dares criticize it.” – Pierre Gallois

    • In IMD Pune/India in early 70s has no computer. The long range rainfall forecast was issued by India Meteorological Department [IMD] I & D section of DDGC’s office – Pune — I was with it — with a premitive calculator [rotating handle driven numbers] and Indian Institute of Tropical Meteorology [IITM] used to have IBM 1600 computer wherein Dr. Sukla & his group used to work on modeling in collaboration with IMD-Delhi numerical weather prediction groups. Later IMD acquired Super Computer to transfer met data that was in the form of punched cards. The modelling group started using it with barrowed from USA models. These are the group that get funds and promotions and not the forecasters using ground based data and later linking it with satellite data — manually. These manually estimated forecasts were in fact accurate while the model based forecasts were vry poor. Even the long range rainfall forecasts failed all most all the years with models using sophisticated computers in Delhi, Bangalore. They changed the average rainfall from IMD forecast average. The modelling group started drawing isolines over Himalayan Mountains — MONEX data. In one of the seminar, Dr. Y. P. Rao [DDGA] from IMD asked them, do you know the basic meteorology and asked them stop such research but yet that group was powerful at government level even in planning commission.

      Dr. S. Jeevananda Reddy
      .

      • when you are granted the big grant for the big computer its kind of like getting tapped on the shoulders by the King’s sword.

      • I had always wondered how well they did before computers. Your experience reminds me of the saying, “To err is human, but to really foul things up, it takes a computer.” :)

  11. Just heard a “scientist” say that coral bleaching has only occurred since the industrial revolution .
    I’m not a scientist but have seen the ancient results of coral beds being forced up out of the water due to tectonic activity and it looked pretty white and bleached to me .

  12. Let’s just stop calling the crap science and move on. Hey I rolled double sixes then adjusted the dice to a six and a four. Ergo next year will be warmer than anything ever.

  13. I think you should always give credit were it is due , and although it is fair to say that with this area poor professional and personal practice is not just acceptable but rewarded and honoured , there may be little doubt that when it comes to the field of manipulating data to make it ‘sing the right song ‘ , climate science is a class apart . Its one area where they can truly be said to be world leaders .
    And even better given its a area full of people that seems to have all the scientific skill of road kill , such has Dr Doom and F Mann , its an area in which their ‘talents ‘ really do come in useful.

  14. Isn’t removing a cold bias from models like removing your unicorn protection clause from your home insurance policy?

  15. ‘Cold Bias’. Wow. I keep telling myself that there is no outrage left that would surprise me.
    And yet, I was still unprepared.
    Reality literally inverted.
    I guess the best liars lie to themselves best.

    • Try reading the technical aspect to it will you.
      Instead of answering the whistle-call of “cold” and “bias”.

      The bias is something in the model parameters that is causing the outcome to give continual local effect – – which happens in this case to be cooling.
      Would you do your “Pavlovian” act if it were a warming bias?
      And they happen too.
      You just wouldn’t get to here of it here.

      • Try working on your grammar, will you? Perhaps spelling?
        Your comment ‘Would you do your “Pavlovian” act if it were a warming bias?’ is laughable, because the warming bias is so prevalent (whether programed into the models themselves or simply existing within the grey matter of those pushing them) that it can be pretty much counted on – because, yes, you’re right about that, for sure, warming bias DOES happen all right – and there’s no real need for it to be emphasized here since it’s pretty much strewn all over the entire media. There’s no ‘Pavlovian-effect’ here – just simple observation – an almost total trend to adjust data to create warming. Whether that means retranslating the way past data is measured, restructuring current data, dividing data into statistically insignificant units of measure – or, most importantly, assigning alarmist consequences to minor effects. As far as this latest ‘adjustment’, it’s just another illustration that models are not reflecting reality.

      • “Try working on your grammar, will you? Perhaps spelling?”

        Oh really, pray tell?

        ” an almost total trend to adjust data to create warming.”

        yep QED – says more of you than anything in the (peer-reviewed) science, and by etension the scientsists involved.

        “There’s no ‘Pavlovian-effect’ here – just simple observation ”

        Don’t be obtuse – the whole of WUWT is dedicated to it.

        “.etc …etc….. As far as this latest ‘adjustment’, it’s just another illustration that models are not reflecting reality.”

        Spectacular hand-waving my friend, and again, par for the course.

        As I said READ the technical report.

  16. Current climate science seems to just “adjust data’ instead of finding out where the problem in the model or whatever is. As Joe Bastardi said, “putting a band aid on a fatal wound”. In the past, science would have demanded an explanation of why the data was so far off and what went wrong with the models,etc. Now, just adjust away. It’s a very scary thing to see science turned into what amounts to fudge factoring to get the answers you want or that match what is happening when your models fail. Why is happening is important, not just adjusting away the problem.

  17. I remember seeing the odd very cold anomaly in the Atlantic just east of Brazil last fall in the University of Maine Climate Change Institute maps of daily CFSV2 SSTA and was wondering what could have caused such an obvious error. This odd extreme anomaly does not show up on any other SSTA data sets and appears to be bogus, but I have no idea what error caused it. It was not showing up on their real-time maps based on OISST V2 and therefore I was surprised to see it in the final CFSV2 maps. Below is the last CFSV2 map they currently have available – for December 31, 2015.

    The bogus cold blob started appearing in September on the UM CCI CFSV2 SSTA maps. All the UM CCI final CFSV2 daily maps can be seen here:
    http://cci-reanalyzer.org/Reanalysis_daily/

    FYI, here is today’s OISST V2 SSTA map from UM CCI:

    • For comparison, below is the UM CCI CFSV2 SST map for December 31, 2015, which also shows the bogus cold patch off the eastern tip of Brazil.

  18. Pretty soon they’re just going to delete the historic climate record. With no point of reference for comparison ……. viola ……. human influenced climate change. Science based on opinion and conjecture will be much easier to manage than the archaic analog approach of requiring things to be observable, recordable, and reproducible.

  19. When your model doesn’t give you the results you want, it’s time to change the data.

    This is an affront to every true scientist on the globe.

Comments are closed.