Jump to content
  • Member Statistics

    17,583
    Total Members
    7,904
    Most Online
    23Yankee
    Newest Member
    23Yankee
    Joined

CFS Dataset- Why is it not a "robust" long term dataset?


blizzard1024

Recommended Posts

I am starting this thread because it can shed light on why CFS is different that the satellites and surface observations. The CFS is written off because it diverges but what if the methodology of this dataset is different and the others are the same? Then the differences are known and can be judged either good or poor.

 

The CFS is the cooler of the data sets so naturally there are a lot of folks routing for it to be wrong. But let's be objective. I know some folks who use this dataset and I will see what I can find. Please share.

 

Respectfully,

Blizzard

 

 

Link to comment
Share on other sites

On the other hand, why is it a robust dataset?  I simply point to the fact that it isn't meant to be used in that fashion and that wasn't its point.  What it is lacking that the other datasets incorporate or have I do not know.  However, given that all the other datasets are very different what is the reason we should use CFS in that manner?  

Link to comment
Share on other sites

The 2010 CFS Reanalysis paper has a comparison of reanalysis 1, CFSR, and gridded GHCN temperature trends over a 31 year period. But the authors don't really go into much detail about its global temperature performance. They did find the CFS trend vs. GHCN "compelling" because the CFSR did not assimilate 2 meter temperature observations over land. 

 

Also, "The interannual detrended anomalies in all three datasets are very similar (not shown). Both R1 and CFSR display upward trends over the oceans (not shown) of about 0.3 K over 1979–2009. This is much less than over land, a puzzle that is yet to be fully understood."

 

post-35-0-78163200-1405552432_thumb.png

Link to comment
Share on other sites

So the CFS R/A actually is well funded and peer reviewed?

Even if that's true, it is still a huge outlier relative to the other data sets, so I wouldn't put much stock in it

 

Yeah it's a peer-reviewed work of NOAA, but it wasn't intended to be a global temperature data set in the first place.

 

I would like to see a paper that does deeper comparisons and explores the issue, but I've seen no indication that the CFS researchers intend to do so. 

Link to comment
Share on other sites

Yeah it's a peer-reviewed work of NOAA, but it wasn't intended to be a global temperature data set in the first place.

 

I would like to see a paper that does deeper comparisons and explores the issue, but I've seen no indication that the CFS researchers intend to do so. 

 

Even though it was intended to be a global temperature data set (I'm a little confused by this, since what it shows is global temperature data), I'm curious what caused the trend change in 2005.

Link to comment
Share on other sites

 

Thanks. People trash this dataset because it does not fit the others...but why??

 

I don't know if they trash it only because it doesn't fit the others. The CFSR is an older frozen-code version of the GFS model coupled with oceans and sea ice. It has biases from bad physics assumptions, etc., like the operational GFS model we all love. Even though it's using real data (except for things like observed 2 meter temperatures), it isn't a "record" of temperatures. It's a dynamic model reanalysis of the state of the coupled ocean-atmosphere. 

 

It's really no different than a 3-hour forecast from the ECMWF. Even though the Euro is highly regarded and probably highly accurate at short range, I don't know anyone who would consider that model run to be a record of reality akin to thermometers recording temperatures. 

Link to comment
Share on other sites

Even though it was intended to be a global temperature data set (I'm a little confused by this, since what it shows is global temperature data), I'm curious what caused the trend change in 2005.

 

It obviously calculates and provides temperatures like any weather model, but the CFSR wasn't invented and run to act as a temperature record. 

 

As far as the trend change, unfortunately, the authors/developers don't delve into that type of thing. The literature I've seen mostly describes the model and modeling strategies and provides short "gee whiz" observations. 

Link to comment
Share on other sites

 

 

 

I don't know if they trash it only because it doesn't fit the others. The CFSR is an older frozen-code version of the GFS model coupled with oceans and sea ice. It has biases from bad physics assumptions, etc., like the operational GFS model we all love. Even though it's using real data (except for things like observed 2 meter temperatures), it isn't a "record" of temperatures. It's a dynamic model reanalysis of the state of the coupled ocean-atmosphere. 

 

It's really no different than a 3-hour forecast from the ECMWF. Even though the Euro is highly regarded and probably highly accurate at short range, I don't know anyone who would consider that model run to be a record of reality akin to thermometers recording temperatures. 

 

 

Then folks should not reanalysis data either for climate trends as well...you see this often. You are correct here. It is a record of climate fluctuations and not real temperature and hard to assess its long term track record. But just because it diverges somewhat from the other datasets doesn't mean it should be ruled out. In science you explore the reason why...good post. Thanks!

Link to comment
Share on other sites

It's not the reanalysis, but I wrote scripts awhile back that download and extract different parameters from the operational CFS to try to extend end-run global model forecast ideas a bit further into the future.

 

This graph is of forecast 2 meter max temperatures for San Antonio, TX (graph is mislabeled as just temperature) showing a clear cold bias for high temps...if the CFS were right, they'd only see a handful of high temperatures in the mid 90s through mid-August. If that verified, I'd tongue-kiss a badger. 

 

This thread has made me curious. I'll see if I can use the same scripts to extract reanalysis temperatures and compare them to specific temperature records just to see what happens. 

 

post-35-0-97369800-1405567100_thumb.png

Link to comment
Share on other sites

Cory thanks.

 

That is exactly what I thought would be the downfall of CFS.

 

It seems to have the same high temp bias as GFS.

 

Combine that with the global ssta flattening during the slow down of OHC uptake and walla.

 

When the land temps are properly accounted for you get a flatline until 2013 where sea and land temps are combining for near record or record global temps thru much of the last year+.

 

16-weekly-global-ssta.png?w=640&h=420

Link to comment
Share on other sites

Thanks. Insightful information. So basically the anomalies that we see on the weatherbell site and the graphs are misleading to some extent because they have a documented cold bias. If that is true, then it is best used to see very short term trends in climate but not the longer term. Ok. I can buy that.

Link to comment
Share on other sites

Cory thanks.

 

That is exactly what I thought would be the downfall of CFS.

 

It seems to have the same high temp bias as GFS.

 

Combine that with the global ssta flattening during the slow down of OHC uptake and walla.

 

When the land temps are properly accounted for you get a flatline until 2013 where sea and land temps are combining for near record or record global temps thru much of the last year+.

 

16-weekly-global-ssta.png?w=640&h=420

 

 

After further thought....let's say the CFS has a cold bias. Ok. But when it is compared to its OWN mean its the anomaly that matters not the actual temperature. So if it has station A with a long term mean temp of 45F and the real long term mean is 47F the anomalies both warm and cold will be the same for the CFS as the "real" long term mean.  So I don't see why the anomalies are colder other than the normal period for the CFS is the warmer 1981-2010.

Link to comment
Share on other sites

It doesn't appear that the CFS and CFSR are the same thing.

 

The CFSR appears to show just as much warming as GISS. 

 

Yeah, according to the paper, it did catch on to the trend. 

Link to comment
Share on other sites

 

 

 

I don't know if they trash it only because it doesn't fit the others. The CFSR is an older frozen-code version of the GFS model coupled with oceans and sea ice. It has biases from bad physics assumptions, etc., like the operational GFS model we all love. Even though it's using real data (except for things like observed 2 meter temperatures), it isn't a "record" of temperatures. It's a dynamic model reanalysis of the state of the coupled ocean-atmosphere. 

 

It's really no different than a 3-hour forecast from the ECMWF. Even though the Euro is highly regarded and probably highly accurate at short range, I don't know anyone who would consider that model run to be a record of reality akin to thermometers recording temperatures. 

 

First, the CFSR was motivated and designed to provide ICs for reforecasts in order to calibrate CFS forecasts.  It was never really intended to be a proper climate dataset (though that is what a reanalysis is).  The CFS isn't simply a frozen version of the GFS coupled to a dynamic ocean and sea ice model, as there were a variety of decisions/changes made specific to the CFS.  Some of this is discussed in the CFSR paper (and more in the actual CFS paper).  For example, a different thermodynamic variable was used, different radiation (including McICA to better capture cloud issues), and others.  It's also important to note that the CFSR isn't frozen like the CFS model itself is.  Most of the CFSR was done at T382 spectral resolution with a particular set of physics and an evolving observing system, but was upgraded to T574 (along with some physics changes, I believe)...I think in 2010 or 2011.  Again, that is only for the model that drives the analysis part....the actual CFS prediction model (T126) has always been frozen and will remain frozen so that the reforecasts and calibrations are applicable.

 

Also, in terms of anomalies, the are computed relative to its own climatology so systematic bias is taken care of.

Link to comment
Share on other sites

First, the CFSR was motivated and designed to provide ICs for reforecasts in order to calibrate CFS forecasts.  It was never really intended to be a proper climate dataset (though that is what a reanalysis is).  The CFS isn't simply a frozen version of the GFS coupled to a dynamic ocean and sea ice model, as there were a variety of decisions/changes made specific to the CFS.  Some of this is discussed in the CFSR paper (and more in the actual CFS paper).  For example, a different thermodynamic variable was used, different radiation (including McICA to better capture cloud issues), and others.  It's also important to note that the CFSR isn't frozen like the CFS model itself is.  Most of the CFSR was done at T382 spectral resolution with a particular set of physics and an evolving observing system, but was upgraded to T574 (along with some physics changes, I believe)...I think in 2010 or 2011.  Again, that is only for the model that drives the analysis part....the actual CFS prediction model (T126) has always been frozen and will remain frozen so that the reforecasts and calibrations are applicable.

 

Also, in terms of anomalies, the are computed relative to its own climatology so systematic bias is taken care of.

 

Interesting. Thanks for the info!

 

(and for correcting my bad info and misinterpretations ;) )

Link to comment
Share on other sites

After further thought....let's say the CFS has a cold bias. Ok. But when it is compared to its OWN mean its the anomaly that matters not the actual temperature. So if it has station A with a long term mean temp of 45F and the real long term mean is 47F the anomalies both warm and cold will be the same for the CFS as the "real" long term mean.  So I don't see why the anomalies are colder other than the normal period for the CFS is the warmer 1981-2010.

 

Good point. I hadn't considered that.

Link to comment
Share on other sites

Yes. So obviously the CFSR is not the same thing as the CFS. The CFSR shows rapid warming the last 34 years similar to GISS.

If I'm reading the file correctly, the Wxbell data is derived from the CFSR (reanalysis). The divergence seems to become evident in 2010...the graph above runs through 2009...for what reason this divergence occurs I don't know. But through 2009, the CFSR is pretty well in line with the other data

Link to comment
Share on other sites

Archived

This topic is now archived and is closed to further replies.

  • Recently Browsing   0 members

    • No registered users viewing this page.
×
×
  • Create New...