Dr. Gavin Schmidt, a lead researcher with NASA’s Goddard Institute for Space Studies (GISS) that does leading climate change studies, replied to one of my posts and made an assertion that the USHCN and GHCN stations and station data being discussed here in my blog are not used in validating climate models. This is surprising to me.
Here is the full correspondence:
Schmidt’s first post:
> Don’t let me get in the way of your efforts here, but please stop saying that “This data is in fact used in climate modeling to predict our climate future”.
>
> This is simply not so.
>
> You’ve downloaded the GISS model – perhaps you’d like to show me where these station data are used? You won’t be able to because they aren’t.
>
> Observational data at large scale (not individual stations) are used to evaluate the models after they’ve been run – but again generally only at the continental scale and above. The evaluation is not just with trends but
> with patterns of variability (El Nino responses, NAO etc.) and obviously, the better the data the more reliable the evaluation.
>
> Note that the climate model hindcasts for this area are around 0.5 over the 20th Century – significantly less than this individual station. Should this record therefore be shown to contaminated, it would actually improve our confidence in the models, not lessen it!
I responded to this on June 21st 2007 as follows:
> Gavin,
>
> I thank you for commenting on my blog, Watts Up with That? I’m honored
> that you would take the time. Rather than reply immediately, I thought
> I’d give some thought and research to my response, hence the delay. I
> also thought you’d appreciate a direct reply rather than a blog post.
>
> You wrote on the blog:
>
> “You’ve downloaded the GISS model – perhaps you’d like to show me where
> these station data are used? You won’t be able to because they aren’t.”
>
> I did some looking at a paper you authored, I found Schmidt et al 2006,
> from BAMS, which is also posted on your website:
> http://pubs.giss.nasa.gov/docs/2006/2006_Schmidt_etal_1.pdf
>
> You wrote on page 168 of the BAMS article:
>
> “We endeavor to compare the model simulations to as many suitable
> datasets as possible. … . Where useful gridded datasets exist of
> selected in situ data we use those.”
>
> After reading through your paper, I agree that you did not show any
> comparisons to GISS gridded data and I will withdraw any implication
> that you used GISS station data. However, I must say that I’m surprised
> to learn that GISS gridded data did not meet the standards of Schmidt et
> al 2006 of being either “useful” or “suitable”. Thank you for drawing
> this to my attention.
>
> However, later in the article, on page 176, you show comparisons of
> model output to CRU surface temperature data on two occasions:
>
> “Surface air temperatures (SATs; Fig. 17) show a general warm
> continental bias in comparison to the updated Climate Research Unit
> (CRU) data (Jones et al. 1999).
>
> Figure 23 on page 187 shows Taylor diagram comparisons among the
> selected models for the December-February (DJF) and June-July (JJA)
> extratropical NH CRU surface air temperature (SAT)”
>
> It is my understanding that CRU uses GHCN station data, which includes
> the USHCN sites discussed here in my blog. So, my answer to your
> question is that Figures 17 and 23 of Schmidt 2006 et al use the station
> data discussed here via the CRU gridded data. It has always been my
> understanding that adjusted GHCN and USHCN surface station data (also
> listed on the GISS webpage) including the ones I show plots of, is
> applied to a gridded data scheme for use in the computer models, such as
> model E. If I am in error in that assumption, I welcome you pointing out
> that error.
>
> If you felt that I was speaking of a specific station data being “used
> to predict our climate future” that of course is not my intent. If that
> was the case, I’ll revise the wording to make it clearer.
>
> Regarding your mention that “contamination of station data would improve
> your confidence in your model”, I must say that I’m a bit surprised at
> this. I’m not really in a position to dispute this yet, but would
> appreciate some additional clarification as why you are so certain of
> this without even seeing the impact of contaminated data. I surmise the
> opposite to be true, but I welcome further understanding.
>
> Again I thank you for your comments, and I welcome any correspondence or
> suggestions you may have.
>
> Best regards,
> Anthony Watts
Dr. Gavin Schmidt replied on June 22nd, 2007 with:
My comments stand. The station data are not used *in* climate models, and
they are not used to predict future climate. So yes, the sentence you have
is just wrong. I’m not sure how you could edit it to make it correct.
We compare the models to the gridded products that deal with individual
station problems as best they can. We have used the GISTEMP and CRU
products to do so. (Semantic note, ‘compare to’ is not the same as
‘include in’). For the specific station you have highlighted, the grid
point trends in the products (~0.5 deg – eveballing it, since I’m on
travel) are significantly less than the trend you show (2 deg or so).
Climate model results for the 20th C are similar (i.e. 0.5 deg). Thus
reductions of the trend at this station would actually improve the match
to the model – always being clear that you shouldn’t really compare
model grid boxes to individual stations…
If you are of the opinion that this station is contaminated, then you have
to admit that the process designed to remove artefacts in the GISS or CRU
products has in fact done so – (i.e. that grid box in the product does not
have a 2 deg/Century trend).
Improvements to that process and the data are always welcome, but do not
ascribe consequences to your project that clearly do not follow.
Gavin
*——————————————————————–*
| Gavin Schmidt NASA/Goddard Institute for Space Studies |
| 2880 Broadway |
| Tel: (212) xxx-xxxx New York, NY 10025 |
…| |
*——————————————————————–*
[email address and tel# removed by Anthony for privacy/spam purposes]
So one has to wonder.
If Dr. Schmidt’s point is only the observation that they do not reconcile their models with every individual station (as opposed to gridcell composites calculated by GISS and CRU), then there is no misunderstanding.
However, it is very clear that the NASA GISS and CRU (Climate Research Unit) use this station data in arriving at their gridcell values which are what is presumably used in testing the models. From 53 USHCN site surveys done so far we know that a number of stations do not meet published WMO (World Meteorological or NOAA (National Oceanic and Atmospheric Administration published standards.
There is no evidence at present that NASA GISS or CRU have made any effort to verify quality control standards at these USHCN stations. Whether these quality control issues will have a significant impact on overall averages remains to be seen. The only way to tell for certain is by examining individual stations though the site survey process as is being done on www.surfacestations.org and then doing an assessment of how pervasive the quality control problems are and what the potential impact of these problems may be.
But, any problems in individual USHCN stations will affect gridcell values. For non meteorologists, a gridcell is a box on a map that has been divided up into a x-y lines and specific data applied to each box. This helps in computer modeling because with computer programs it is easier to divide into cells, then calculate and display. Below is an example map that may help you visualize gridcells:
Whether it’s a big problem or a little problem remains to be seen, but it’s odd for Dr. Schmidt to pretend that it’s not a problem because they use the gridded version of the data.
I am not the smartest of people, but it sounds like he’s saying that this data is not used in away way other than comparison with IPCC climate models and scenarios.
Is this data used *by* climate models at all?
It is interesting that he is distancing his and other’s modeling work from the surface temp record. I think that must mean he realizes how poor it is.
The data may not be “in” the model but the output of the model is compared to the data and the model parameters are “tweeked” until they match so they can say that the model has been back tested to agree with past data. To me, that’s equivalent being “in” the model.
Bob
I don’t believe the models in fact use the data, but the models are in fact tweaked and tuned to deliver a reasonable historical match. Of course, that’s part of my problem with the models…at many points in the modeling process the modeler has to weight a feedback coefficient based largely on a gut feel. And part of that gut feel is how accurately does the coefficient help the model predict the past.
Schmidt does not use the data to construct the models, he merely uses averaged “grid” data to help verify the models. I guess he means that if his model were to seriously deviate from the surface temperature record he would not use this fact to go back and rework his model. It makes you wonder why he bothered to look at the station data, even in its aggregate form, at all. Clearly, he would not allow it to influence his work.
So, we can conclude that if the final result of the station audits would be a lowering of the rate of 20th century temperature increase that Schmidt would be as indifferent to the surface station record as he claims to be right now. He will make no claims that his model was better than he thought because he never really used that data to validate his model.
Yeah, sure.
What about input parameters for the models? Don’t they need some “starting gridcell temperatures”? Wouldn’t these come from GHCN/CRU datasets?
Hi Anthony,
Very interesting exchange.
My take on his comments is that he claims not to use the surface temperature data to develop any of the model parameterizations, such that the results of the runs would then “match” the data (i.e. the data are used to “tune” the model). On the other hand, time-dependent forcings, such as GHG emissions, solar fluxes, etc. – which can only be known approximately – have likely been “adjusted” so that the runs do indeed match the averaged data.
But clearly the surface temperature data must be used to (1) initialize the models and (2) compare the results over some period of time (these kinds of comparisons are all over the IPCC report). So I find it very strange that people such as Gavin Schmidt (and others) would be so dismissive and belittling of your efforts. That just doesn’t sit right with me. At the very least, I would think that documenting the conditions of the stations by ordinary citizens (using their own time and equipment) would be greeted enthusiastically by NOAA, GISS, and others.
Not sure if its the context, but RE initializing of models with gridcell data, per K. Trenberth of NCAR, “None of the models used by IPCC are initialized to the observed state and none of the climate states in the models correspond even remotely to the current observed climate.”, as discussed at:
http://climatesci.colorado.edu/2007/06/18/comment-on-the-nature-weblog-by-kevin-trenberth-entitled-predictions-of-climate/
Anthony, one way to reword your sentence is: “This data is in fact used in validating climate modeling to predict our climate future”.
This gets around Dr Schmidt’s quibble about “in”. Yet surface station data clearly plays a large role in the effort to predict future climate. Bad data leads to bad results which no amount of ‘gridding’ can remove.
RE: Steve Geiger. Thats a very relevant posting you’ve provided. Thank you.
In addition, the GISS surface data are used in the GISS Surface Temperature Analysis which documents the present and historical state of the surface global temperature.
http://data.giss.nasa.gov/gistemp/
I believe that is also a fair statement.
Just as a precautionary note, the global warming is registered in the oceans as well as the land, a fact that should not be ignored–but definitely cross-verified.
With respect to contamination by one or more station, it could very well be that after additional scrubbing of the surface data, there could be better agreement between observed and model output. In this manner, the better data would help validate the model.
RE: Goinggreen
“as a precautionary note…”
It’s not ever been my contention that the surface temperature record is the sole indicator of climate change. Absolutely, there indications are seen in sea surface temperatures, radiosonde balloon data, and in satellite measurements done by John Christy, though they are less in magnitude than surface temperature trends.
But since the surface temperature record seems to be both the strongest trend indicator, and the one that has undergone the most adjustments to raw data, it is worthy of a detailed examination of the source of that data.
I submit that if the surface temperature record cannot withstand the scrutiny of a few professionals and laymen taking photos and doing surveys of data collection points, then something must be wrong with it on a large scale. When the survey of all 1221 USHCN stations is complete to the best of our ability, we’ll know if the problem is big or small. The fact that the this basic quality control using photography and hands on surveys has not been done callls the surface temperture record into question.
I am surprised at the response I’ve been getting to this effort, both positive and negative.
It remains to be seen what the true linkage is between the surface temperature record and climate models.
Anthony
Gavin stated:
“If you are of the opinion that this station is contaminated, then you have
to admit that the process designed to remove artefacts in the GISS or CRU
products has in fact done so – (i.e. that grid box in the product does not
have a 2 deg/Century trend).
Improvements to that process and the data are always welcome, but do not
ascribe consequences to your project that clearly do not follow.”
Gavin’s statement implies to me that station data and station selection process is adjusted to obtain individual grid cell trends of 2 deg/century, otherwise the grid cell temperature series is considered an artefact. Is that your impression? Is the 2 deg/century trend the goal for all the station adjustments?
Phil B.
Phil, I don’t think we should try to create scenarios from these comments without further study, and I want to lean more about the models and the data input, data adjustments, formulae, and begininng to end process before I do anything more in the way of comments about it.
The man is an expert on this subject, lets not lose sight of that.
Heroic work, exemplary respectful tone. Your slow, careful, conscientious approach deserves an extended write up from a major publication. Most of us were taught that science consists, in part, of meticulous data collection, and it is always great to see amateur data collection complementing, and improving upon, that of “professionals.”
Anthony, I appreciate your comment and the need for civility. Gavin’s statement seems very clear to me. The trends in the grid cell temperature series products from CRU and GISS can be checked.
Phil B.
Anthony, don’t let the negative responses get to you (I’m sure you are not). This temperature measurement problem is a huge one and your initiative is extremely important and long overdue.
Some suggestions….(1) an expanded GISS gallery (I could get to some of those faster than the USHCN stations due to travel distance), (2) An ability for people to comment, perhaps on the state or station level, (3) Post the official USHCN station history file (or its contents), so people can try to get information on the historical sites as well. I’m guessing your queue of things to do is pretty long.
The data may not be used in the climate models, but isn’t it used by NASA to compare the global temperature from one year to the next so that they can issue periodic press releases declaring a particular year to be “the warmest on record”? I think that is the real problem with poor quality control of the stations. Their data are used to generate these press releases that are then picked up by the mass media who can use them – because of their sensationalist nature – to sell more newspapers and magazines.
Gavin said: “Observational data at large scale (not individual stations) are used to evaluate the models after they’ve been run – but again generally only at the continental scale and above.”
I can’t remember the last time I saw a climate model compared to observations “only at the continental scale and above” (although modelers prefer to present these sort of results because they tend to make the models look much better than when the details of smaller scales are looked at).
This would be most interesting and nonsensical, since the models are often conversely used to predict future climate conditions at smaller, regional levels. How can a model not verified to work at regional levels be used to make predictions and dictate policies for regional levels?
It also just so happens that Hansen, Schmidt, et al, have a paper in press http://pubs.giss.nasa.gov/abstracts/inpress/Hansen_etal_1.html which compares modelE runs to observed data at levels smaller than “contentintal and above.”
Gavin’s insistence that, “My comments stand. The station data are not used *in* climate models, and they are not used to predict future climate,” is pretty amazing taken at full value. It suggests that climate models are developed completely independent of temperature observations and are in no way, shape, or form dependent on these observations to improve, calibrate, etc, the models. They are just used to “evaluate?” So after all the labor-intensive model development, if the model is pathetic at matching observations, they just either scrap it completely or don’t make any adjustments?
Seems like he was trying to get away with the technicality of pointing-out that station temperature data is not found anywhere in the code, etc (which isn’t what anyone has claimed), and hoping to minimize the model comparisons to observational data (which is where people think it does affect the models, with calibration/tuning, etc).
“If you are of the opinion that this station is contaminated, then you have
to admit that the process designed to remove artefacts in the GISS or CRU
products has in fact done so – ”
This statement makes no sense to me.
I do not see how agreeing that the data is contaminated means that we MUST also believe that Schmidt’s attempts to remove the contamination have been 100% successfull.
If I’m understanding correctly, the gridcells are averages of site data. And if there is significant site contamination or lack of quality control, then the averages are wrong to begin with. Aren’t they?
My read of the back and forth with G Schmidt is that he saw you find errors in a single station, perhaps not being too aware of your larger efforts, and wanted to make it clear that the entire edifice of climate modelling was not destroyed by the errors at that station.
As to his point that the model does not use station data, one has to concede that this is correct, at least on a direct basis.
Clearly, though, the surface temperature record is important in determining whether the model is working, hence the many comparisons in the papers describing model runs with historical temperature records.
Also, it appears that the GISS model is tweaked to try to reproduce hstorical temperature records. It is hard to read this quote as suggesting anything else “The main implication is that the 124-year warming in our model
with ‘all forcings’ is ~0.10ºC less than observed, rather than
0.05ºC less. Thus the need for less tropospheric aerosol amount
becomes clearer in the global mean temperature, as well as from
unrealistic cooling over Europe.” Troposhperic aerosol is input into the model for various periods that have no direct measurement, and so becomes a tuning method. (The paper quotes other reasons to change the aerosol inputs too.)
Also, many of the research papers that are used to develop mathematical equations used in the climate models are tested against the surface temperature record. These mathematical equations may be mis-specified if the surface temperature record is wrong.
The paper quoted is HANSEN ET AL.: CLIMATE SIMULATIONS FOR 1880-2003 WITH GISS MODEL E, electronic version
to good night
much interesting found for you on http://www.norcalblogs.com much scorched tem here spam all right I went away to meeting
key mesage for you
mature guy link for you
“>http://thai-sex.yourda4a.info/lesson-milf.html
Re “nanny_govt_sucks”
climate models from GISS used by the UN/IPCC aren’t intitialized, meaning that they don’t begin with todays climate. Hence they have no use of todays temperature. How they can claim them to be accurate without initialization is a bit of a mystery to me, but that is what they do.