Thursday, February 4, 2016

Satellites, surface temperatures up in January

The satellite temperatures for January are out now, and also the Mohyu NCEP/NCAR index. All increased, quite significantly. UAH Ver 6 beta was up by 0.09°C, to 0.54°C. RSS-MSU was up from 0.543 to 0.663°C. In both indices, Jan 2016 was the highest ever, making four such months in a row. However, the jump from Dec to Jan was rather smaller than in 1997/8. I have an eye here on the implications for the RSS "pause". February will have to be a lot warmer (0.88) to erase the pause in that month. OTOH, the January warmth, if sustained, will ensure there are only positive trends after March data.

The NCEP/NCAR index rose from 0.621 to 0.665°C. Again, very warm, but not as big a rise as might have been expected. In fact, the first half of January was very warm indeed, but then came a relatively cool period. February is starting warm again.



51 comments:

  1. I suppose this could still be a bursting EL Nino.

    ReplyDelete
  2. Hi Nick,
    Have you compared your NCEP/NCAR index to the www.karstenhaustein.com/climate website? The "relatively cool period" you mentioned in your post was, if my memory serves me correctly, predicted in their 7day forecast of global temperatures. I think someone else a couple of months ago mentioned the same thing on one of your other posts. It would be interesting to see how accurate their forecasts are compared to the Moyhu NCEP/NCAR index.
    BTW, thanks for all of your time and effort you put in to this site!

    ReplyDelete
    Replies
    1. Jared,
      I've been looking at doing that directly from the GFS/CFSR data. The main problem with Karsten's information, as at CCI, is that AFAIK you can't digitally read the global numbers - they are presented in images. I see that he does his own forecast checking.

      Delete
    2. Isn't the number below and to the left of the image the anomaly forecast? For example, Karsten's forecast for noon on february 5th says, "Anomaly Global: 0.639K". It also shows the forecast for the Antarctic and the Arctic in the center and to the far right.

      Delete
    3. Jared,
      Yes, but it is in an image. You can copy it by eye, but not download automatically. I rely on automatic download to follow this kind of data.

      Delete
    4. Now I see what you mean. Sorry!

      Delete
    5. Nick,

      By the Way, what Values does the NCEP/NCAR index account for the oceans? Is it real SST-Values or the air temperature 2m above the oceans? I cant find a good source which would explain this, i think this should be a important factor if sometimes there is a small divergence between NCEP/NCAR and observations...

      Delete
    6. Nick, I can produce a text or NetCDF file (whatever you prefer) and send you a link where you can download it if you want to do some analysis on your own with my anomalies.

      One interesting bit though: My NCEP reanalysis anomalies shows quite a bit colder January conditions compared to your NCEP/NCAR index. Which version of the reanalysis are you using?

      @All: GFS is notorious for having too high an amplitude when it comes to strong cold or warm anomalies particularly over Eurasia. The mid-January cooling didn't show up nearly as strong in CFSv2 than in GFS. Yet the December 2015 drop was much more similar.

      @Olof: I didn't change anything in the way the anomalies are calculated. The upgrade was a server issue (if that's what you are referring to) and had nothing to do with the data analysis at all. All we see at the moment really is that GFS tends to have a higher variability. Which is why I made some initial adjustments in order to fit GISS better. I should be updating the adjustment, but I'm pretty sure a few issues will remain. For example, GISS is based on SSTs (ERSSTv4), whereas GFS uses 2m temperature across the globe. Nothing which can be perfectly reconciled with a simple adjustment procedure. More sophistication may help ;)

      Delete
    7. KarSteN,
      Thanks for the response. I'm using V1 of NCEP/NCAR. The page where I describe the index is here. I'm using NCEP/NCAR V1; specifically the sig995 data here. The README file is here.

      Yes, I'd be very glad to look at your anomalies. My email name is nastokes and the address at westnet plus com and au. I'm happy with text or NetCDF. I have dabbled with GFS data on NOMADS, described here and here. But I haven't followed up systematically. My main problem is that I have to download a huge amount of data there to calculate an anomaly base.

      Delete
    8. Anon,
      I use the NCEP/NCAR sig995 data, which is an air level - basically the bottom layer of grid cells. So yes, there is possibility of divergence from SST. The same issue affects comparing GCM output with land/ocean indices (C&W). I mention the issue here and here, for example.

      Delete
    9. Nick,
      Thanks! I'm using NCEP/NCAR 2m temperature rather than the sig995 temperature. Would be interesting to know why it seems to differ quite a bit at times. I'm probably uploading the NetCDF file of my anomalies so that you can use the un-adjusted data. I'll let you know as soon as I made it available.

      Delete
  3. The reanalysis by Weatherbell agrees with the Moyhu NCEP/NCAR index, January slightly up from December.
    However, karstenhaustein.com/climate has obviously got lost during the upgrade a month ago. Both his reanalysis indices report January down 0.15 C from December...

    OISST v2 1/4° (at KNMI climate explorer) is up 0.06 C from December, a great leap for an SST index, which indicates that it will be new records for the blended surface indices in January..

    I am not impressed by the TLT January indices. They are slightly up from January 1998, but actually the TMT-channels at NOAA/STAR and RSS are cooler compared to Jan 1998.
    TLT January 2015 wins because the upper troposphere (MSU3/AMSU7) was warmer in 1998, and that value is subtracted from TMT to contruct the TLT

    ReplyDelete
  4. Nick

    A comparison of NCEP/NCAR R1 and UM CCI CFS would be interesting. There was an update from CFS to CFS2 in early 2011 and the CFS series cools relative to other re-analysis and surface temperature series in the 2010/11 period. Clearly documenting the comparison would be worthwhile.

    Chubbs

    ReplyDelete
  5. Nick, well Monckton is preparing his acolytes for the "end of the pause" in his latest effort at WUWT. I made a very rare outing to WUWT to see what the fuss was about. I noticed you were pointing out just a few of the flaws in what passes for reasoning on the part of the Noble Viscount. I made a few comments myself. Interesting that none of the acolytes seems to make much of an effort to argue with what you or I actually say or indeed defend what Monckton claims.

    ReplyDelete
    Replies
    1. Sou has a superb response to Monkton up now.
      http://blog.hotwhopper.com/2016/02/christopher-moncktons-trend-of.html

      Delete
  6. Nick and Karsten don't find the same numbers for january but we should remind they don't use the same base. 1994-2013 for Nick, 1981-2010 for Karsten. With those different bases, you don't find the same ranking for wamest months.
    However that can't explain the big drop found in Karsten's temps in january. Both NCEP/NCAR and GFS show a weird decline in january 2016 on Karsten website. Usually, its data are consistent. My guess would be he had a technical problem in january.

    ReplyDelete
    Replies
    1. Well, TempLS (early) is now also showing a big drop in January. Oddly, it's missing US data, which is usually the first. But I don't think that will have a big warming effect.

      Delete
    2. No, US data will rather have a cooling effect. Finally, Karsten dataset may have matched your TempLS data better than NCEP/NCAR and NCEP CFSv2. We'll see GISS... Probably it will be closer to your TempLS as Karsten. Karsten is "GISS ajusted" so maybe that's what explaining the discrepancy. Really, that's a curious month... From the beginning of january, I thought Karsten had a technical problem. The only thing that made me doubt a little bit was temps in Europe (were I live) and US, cooler than in december. The key is certainly the NCEP/NCAR : your date is correct, I looked at in on ESRL website for "surface" ... But finally, Karsten's NCEP may be closer to TempLS (and probably GISS).
      Maybe you can both figure out what happened.

      Delete
    3. How far could GISS drop? Below November?

      Delete
    4. Currently TempLS mesh for January is just below its November 2015 level.

      Delete
    5. When your NCEP number for January landed at 1.243, I became somewhat hopeful 2016 could end up another warmest year. Just seems like there need to be big numbers out of the gate. Hard to imagine that a GISS number just over 1.0C is possibly not big enough.

      Delete
    6. It's interesting to look at the early breakdown. SST is up, most land down somewhat, except Arctic. Biggest drops in Europe and Antarctica.

      Delete
    7. Keri,
      the ranking might be different but the change should be similar (baseline shift only). After all, NH temperature in GFS went down from +1.2K to -0.2K within a matter of 10 days in mid-January. Eurasia turned very chilly once AO switched strongly negative, making way for a typical WACCy (Warm Arctic Cold Continent) situation. The subsequent blocking wasn't as long-lasting as it sometimes can be, hence Eurasia now back to warmer conditions, yet the NH-temperature drop was substantial as a result. As mentioned above, GFS tends to produce large amplitudes for these kinds of events which most reanalysis dataset don't support. Drop in CFSv2 was "only" -0.8K and NCEP/NCAR probably even less. GISS usually follows GFS at least to some extent, so I'd be surprised if it follows NCEP/NCAR sig995 temperature. I'm using NCEP/NCAR 2m temperature which went down from December in lockstep with GFS.

      Bottomline: You simply can't have a record month if the Arctic cold is flooding Eurasia. The land mass is just too massive to be compensated by whatever warm Arctic. January did see a legitimate Eurasian cold spell, hence GISS should be colder than December too. TempLS seems to support this notion. February so far with much more potential for another record month, tho it's far from set and done at that point.

      Delete
    8. Yes, I looked at Feb 2015 yesterday just to get a feel for the daily anomalies, and it's much stouter so far this year.

      Delete
    9. Following KasSteN's, comments, there is more detail here on where it was cold. Basically a belt through from Europe through Central Siberia to China.

      Delete
    10. Karsten,
      What you say makes sense but I'm still troubled by the difference for ncep/ncar. If you check surface, not sig 995on ersl, you'll see january as warmest on record. Maybe the difference is you are using forecast as data while Nick is using reanalyzed data. His updates come three days after while yours are on daily basis and even forecasted.

      Delete
    11. ... But that's not correct, you're not using forecast as data. That would only apply to GFS... No, I don't understand what's going on with that NCEP data. We'll see GISS.

      Delete
    12. Keri,

      Something I noticed a while ago, looking at past El Ninos in NCEP data, was that land average anomalies tended to be warmer and far less variable than seen in GISS around the peak months of December through May.

      JCH,

      A 1.0C anomaly would still be 0.2C warmer than the 2015 January-September average in GISS. A big drop from this point is needed to avoid a new record. This is my rough prediction for how things will proceed, with previous El Nino events marked for reference. That picture works out to a new record by about 0.05C.

      Delete
    13. Understood Paul S, but .87℃ is a high bar. ~.92℃ would be good for 2016. I think a big thing is going to be whether or not the PDO remains positive; have we seen the peak, or is another PDO surge coming? It's decaying now, but warm water is stuck along the coasts. Maybe it will persevere.

      Delete
    14. I believe that TempLSmesh will rise to level of December with US data. CONUS is probably neutral, but Alaska and some missing high Arctic stations in Canada will likely fill out the "blue hole" in the Arctic ocean with at least + 4 C anomalies.
      Likewise, Sudan data may affect the cold spot in Central Africa. There is also data missing for a majority of the Andean countries..

      Delete
    15. January 2016 is the warmest month ever, 0.72 C above the global average for 1981-2010, up 0.03 C from December, according to ERA-interim

      Delete
    16. Olof, Nice, thanks, it seems exclusive data is provided for jan 2016, not yet on ecmwf. Great arctic warming. Climatedataguide say there is maybe à warm bias for the arctic.

      Delete
  7. The latest Ratpac A data is out now. The troposphere temperature 850-300 mbar is skyrocketing, winter (2 months of 3) is up by 0.25 C from autumn.
    At the peak of the 1998 el Nino (spring season), Ratpac and UAH v6 were quite similar in 1981-2010 anomalies. Now,in the present winter season Ratpac leads by 0.4 C...

    ReplyDelete
    Replies
    1. Very interesting. A failure to register the magnitude of an el Nino doesn't look good for the satellite data. Interesting to see how UAH 5 copes. On the subject of satellites vs balloons I happened to notice on some d***er blog that John Christy has been claiming that UAH6 agrees very well with balloon based measurements. Does anyone know anything more about this?

      Delete
    2. Actually, can you give me the url for monthly Ratpac data. I can only find annual data on the NOAA website

      Delete
    3. Bill H, they don't have monthly resolution for the globally and zonally weighted indices, the finest is season and they start to report a season when two months of three are in..
      The satellite series agreed with radiosondes back in the MSU-only days, but with the introduction of AMSU in 1998 they started to drift away.
      UAH 5.6 behaves quite well, I don't know if the reason is that they don't use diurnal drift correction with AMSU:s and rely more on the non-drifting satellites..

      Delete
    4. Olof, fine: do you have the ultimate for seasonal resolution?

      Delete
    5. Sorry, that should be url,not "ultimate". (wretched tablets!)

      Delete
    6. Ratpac A seasonal layers 850-300 mbar is here
      I don't know the baseline, so it has to be rebaselined to 1981-2010 to be compared with satellites etc.

      If you step up in the directory you can find more..
      E. g. Ratpac v2 beta which seems to be an improvement. Same methods but based on the more comprehensive database IGRA v2 beta, resulting in fewer data gaps and loss of stations..

      Delete
  8. I’ve never seen an adequate explanation from Christy or Spencer, of exactly which balloon indices are shown in that graph they keep showing: https://science.house.gov/sites/republicans.science.house.gov/files/documents/HHRG-114-SY-WState-JChristy-20160202.pdf

    The graph says “NOAA,” but NOAA’s web site presents NOAA’s RATPAC-A as NOAA’s radiosonde index that is appropriate for looking at global trends: https://www.ncdc.noaa.gov/data-access/weather-balloon/radiosonde-atmospheric-temperature-products-accessing-climate

    And RATPAC-A well matches surface trends and not RSS or UAH: https://tamino.wordpress.com/2016/01/15/drift/

    The graph says it uses the “UKMet” balloon data from about 1978, but the UKMet’s web site says its global index contains data only from 1997 forward: http://catalogue.ceda.ac.uk/uuid/f2afaf808b61394b78bd342ff068c8cd

    So Christy’s graph must be using only the European UKMet index: http://badc.nerc.ac.uk/data/radiosonde/

    RAOBCORE goes up to only 2011, I don’t know if it’s global, and its authors warn that its homogenization can be biased by satellites. Obviously it’s inappropriate to validate the satellite indices with a balloon index that can be biased by those very same satellite data: http://www.univie.ac.at/theoret-met/research/raobcore/

    RICH uses only other radiosondes for homogenization, but I believe has the same limited time span and possibly limited geographic span as RAOBCORE: http://www.univie.ac.at/theoret-met/research/raobcore/

    ReplyDelete
    Replies
    1. Another rather serious flaw is that on this much cited graph of Christy is the abbreviation TMT, short for "temperature of the middle troposphere. However, the UAH series is for the Lower troposphere. What a mess. This graphic is starting to rival the notorious "schematic" from the first IPCC assessment.

      Delete
    2. Bill,
      I think Christy is moving away from TLT to TMT, and I think this will be a trend. TLT is just too uncertain. I've started tracking UAH TMT in the various data page displays. Also a NOAA Nesdis version. I'll write this up soon.

      I think Christy is OK there as long as he has been consistent.

      Delete
  9. Independent of the baseline used for RATPAC A, it's linear slope is 1.62 K/century for 1958-2016 (full range) and 1.72 K/century for 1979-2016 (full satellite range). NOAA is 1.55 K/century for 1979-2016) and RSS is 1.24 K/century for 1979-2016.

    ReplyDelete
  10. The RATPAC A slope is for 850-300 mbar data

    ReplyDelete
  11. Does RATPAC A incorporate RAOBCORE, RICH, and UKMet data in its global temperature? You would think that it must.

    ReplyDelete
  12. Owen, Ratpac is a separate dataset, and the only one that is freely available and continuously updated (the 6th every month, or so)

    HadAT (UKMO) ended in Dec 2012
    IUK v2 ended in Feb 2013
    RICH and RAOBCORE ended in Dec 2013

    There is seemingly little interest to maintain a radiosonde dataset. Maybe the reanalysis approach has taken over..? "Fresh" data for the troposphere layers from NCEP/NCAR and ERA-interim kan be downloaded at KNMI Climate Explorer.

    Here is a comparison of troposphere temperatures by satellites, radiosondes and reanalyses over a disputed time period, using some Christy tricks ;-)

    ReplyDelete
    Replies
    1. I was wrong about RAOBCORE and RICH, they are updated after each full year. The latest updates, from Feb 9 2016, including 2015 can be found in this directory:
      ftp://srvx7.img.univie.ac.at/pub/v1.5.1/
      (the directory is often unavailable)
      There are gridded datasets for Rich, Raobcore, and Raw. With a netcdf reader (like Panoply) and a spreadsheet, it is quite easy to make zonally weighted global temperature series.
      RICH is now included in my troposphere trend comparison, link in my comment above.
      RAOBCORE is essentially similar to ERA-interim ( which is used for its homogenisation). I recommend RAOBCORE for those who want to find the elusive tropical hotspot, the trend during the satellite-era at 300 mbar is about two times larger than that at 850 mbar.

      Delete
  13. This comment has been removed by the author.

    ReplyDelete