Tuesday, November 3, 2015

NCEP/NCAR index up 0.2°C in October

I posted earlier about a big spike in the Moyhu NCEP/NCAR index in early October. That index is one that I derive by integrating NCEP/NCAR reanalysis data, as explained here. The index came back from the peak, but only back to levels that would have been seen as very high in earlier months, and stayed high right to end month. So the average finished at 0.567°C, as compared to September 0.368°C. These numbers are relative to base years 1994-2013.

That makes October by far the highest monthly anomaly in the record; in fact, it beats the previous record (Jan 2007) by 0.15°C. That can be seen in the following graph of all monthly anomalies since 1994:

Relative to the 1951-80 base of GISS, October would be 1.18°C, and on the NOAA 20th Cen base, it would be 1.14°C. I wouldn't expect to see those indices rise so high, because they have been somewhat lagging the NCEP/NCAR index recently. In September, GISS was only 0.81°C. Still, there is clearly a possibility of GISS reaching 1°C, and a very strong probability of being the highest anomaly ever, in all indices.

In a related news item, Australia's October was the hottest month ever. Also very dry, where I am. We had a very unusual heat wave at the start of the month, and it continued mostly warm and sunny. It looks like a dangerous fire season coming.




25 comments:

  1. 1°C for GISS looks indeed a possibility in october. Maybe a little more...

    ReplyDelete
  2. Yes, October can really become something extraordinary. UAH v6 is up 0.18 and CFSR up 0.19 from September. The Arctic and Antarctica has been quite warm and will not hold back GISS like in the previous months. I would not be surprised if GISS approached 1.10 this month. An upgrade of September with a few hundreths is also likely.. JMA Sept just went up by 0.01, from its preliminary record high level.

    ReplyDelete
  3. Nick, please could you include the Moyhu NCEP/NCAR index into the active plotter for global temperature indices? So it would be easier to compare to the other indices.

    ReplyDelete
    Replies
    1. Anon,
      Yes, I'll do that. It only goes back to 1994. There is data before that, but it's patchy.

      Delete
    2. There is an active graph, and other graphs, in this post. Up to about a year ago.

      Delete
  4. RSS TLT is in for October and it's the warmest anomaly since 2010. Starting to see El Nino influence here, on schedule. It's particularly clear in the Tropics, as expected, more so in TMT.

    ReplyDelete
    Replies
    1. RSS TLT is "only" up by 0.08 from September, but the coverage over Antarctica is poor compared to UAH. This is slightly improved in RSS TTT which is up by 0.12 C

      Delete
    2. Which is why Karl Mears of RSS looks at a thermometer when he wants to know the temperature on a place called earth.

      Delete
    3. UAH rose by about 0.18°C, about the same amount as the NCEP/NCAR index. It started from a lower base than RSS, and of course NCEP, so while it is the hottest October in the record, it isn't such a big outlier.

      Delete
  5. The breakdown of contributions to the monthly anomalies gives a first look on TempLSmesh for October. I don't know how many stations are already in, but the 'All' is out of scale. If I add the contributions graphically I get approx 0.96 for TempLSmesh for October, mainly due to a large temperature rise in Antarctica and the Sea. Because TempLSmesh is close to GISS and they where 0.669 and 0.81 in September respectively, GISS October maybe over 1.0.

    ReplyDelete
    Replies
    1. Anon,
      That's actually a bug, well, sort of. Thanks for noticing. I don't let it report at all until at least 3200 stations in all are in - then I can check and write up when major countries have reported. But I see that the breakdown escapes that requirement. It gives be some time to get the axis right.

      I don't know how many stations are in, but in about 3 hours there will be another update.

      Delete
    2. Anon,
      I see the problem is inconsistent restrictions. I let it report if >3200 stations, but I don't let it report a mean unless there are 3500. The breakdown works independently of this. There were 3452.

      I fixed the axis issue. The current (premature) calc is a rise from .669 to .876, almost exactly like the NCEP index.

      Delete
    3. I wonder why the large discrepancy between my value of 0.96 and 0.876 comes from. I repeat the graphical adding for September (in the old figure) and get between 0.69 and 0.70, a slight overestimate, probably due to the line thickness. Even accounting for this, I get still approx 0.05 more (in the new figure) when adding the lines. Maybe my graphical method is to inexact.

      Delete
    4. What I also wonder why, that the Sea contribution rises so much, but the HADSST3 falls in October. Maybe it's the area over sea ice?
      Now there is the other update is there with 3938 stations, but still missing some large areas. The temperature so far is 0.889.

      Delete
    5. Anon, the changes in the SST indices from September are small, 1-2 hundreths. The mother of all SST:s, the ICOADS 2.5, is down 0.01 for instance.
      I believe that the sea contribution (in absolute numbers) rises because the the global anomaly rises, but the sea contribution in percent of the global anomaly is actually slightly less in Oct than in Sept.

      Delete
    6. The SST discrepancy is interesting. One thought I have is this. Arctic seas were relatively warm, in October, and ice advances rapidly. I treat iced points as NA, so their values get interpolated, and remain a warm anomaly. But ERSST enters them as -1.8°C, the freezing point of sea water. Now if you take that literally, those points go to zero anomaly (because they are usually -1.8 at that time).

      Delete
    7. Olof, yes, but in the breakdown of contributions to the monthly anomalies the contribution of the Sea to the anomaly is 0.1 greater in October than in September.

      I think it may be the Antarctic sea ice. It is still close to the southern winter maximum and the flip of the few Antarctic land stations from cold to warm gets interpolated over this sea ice. So the area (land+sea ice) of Antarctica is about twice as large than the continent itself.

      I think for a better global temperature index we really need more stations in Antarctica. Had not GISS used additional stations there?

      Delete
    8. GISS uses SCAR. But I don't think that helps. AFAIK, SCAR just has 10 stations on the continent, and only two or three are not in GISS. Those few have short records, and are no longer reporting.

      Delete
  6. ERSSTv4 now out for October. Another new record but only 0.01degC warmer than September anomaly. This seems to be consistent with NCEP/NCAR SST within ERSSTv4 coverage (roughly 60S - 70N), where there is overall essentially no change between September and October.

    The full global (90S - 90N) NCEP/NCAR SST shows an increase of 0.07 from September. It seems all of this difference is related to very warm Arctic temperatures North of 70N, which are largely not picked up by ERSSTv4. This means the "non-interpolated" datasets (NOAA, HadCRUT4) will mostly not include this warming which apparently accounts for a large proportion of the September to October change.

    Over to land, NCEP/NCAR has October global anomaly about 0.45degC warmer than September. Cutting out Antarctica this changes to 0.3degC warmer. Again, non-interpolated datasets will miss a big chunk of warming. I wouldn't be surprised if NOAA and HadCRUT4 report only a small increase, maybe 0.05, assuming the NCEP/NCAR warming is real.

    GISS is somewhat more unpredictable, it depends on what measurements are at the edges of uncovered regions for interpolation. On a statistical basis month-to-month global land SAT fluctuations in GISS and NCEP/NCAR show good correlation, and indicate a 90% chance that the October land anomaly will be at least 0.25degC warmer than September (the comparative figure for NOAA is 90% chance greater than 0.1degC warmer). Combining with the SST data already in (via 0.7/0.3 ratio), and assuming interpolation over uncovered ocean will not reduce the global average, then indicates a 90% chance of global land+ocean being greater than 0.95degC (though note that's based on using independent land and ocean numbers for September and combining these seems to over predict the reported LOTI anomaly by 0.04). Because NOAA starts from a higher position (0.9 in September) the figures work out to be similar for that dataset. Based largely on subjective judgement I think there is probably <50% chance of either reporting over 1degC. I consider there is pretty much no chance of >1degC anomaly in HadCRUT4.

    Well, that was longer than I expected.

    ReplyDelete
    Replies
    1. GISS loti has the rule "no SST within 100 km of a reporting land station". This month when land temps are warmer than SST, it might give a slight warming bias, maybe more than the +0.21 from September indicated by TempLSmesh right now.
      At least in practice, GISS seems to interpolate land temps over sea ice. I don't know if it is a rule behind this..

      Delete
    2. Paul,
      "Again, non-interpolated datasets will miss a big chunk of warming."
      I think you may be right. TempLSgrid shows quite a small increase, about 0.03°C (I'm not sure why it isn't showing yet). It's true that in Sept grid went up while mesh went down, and I had been expecting a possible compensating move. But the discrepancy here is large.

      The SST behaviour is a puzzle. There are lots of indicators of small SST rise - my latest post on weeklies suggests a modest change in October. And yet, the larger rise shown in the breakdown is a weighted sum of actual SST readings. It may be upweighting some Arctic data.

      Delete
    3. I've found something interesting, though I'm not sure it affects the issue you're seeing. On the HadSST3 landing page the anomaly map shows, contrary to my earlier point, reasonably good coverage in the upper Arctic. Looking through maps for past months suggests my initial comment was fair, because this level of coverage seems to be unprecedented in October.

      A couple of questions popped into my mind: How did they get these measurements - wouldn't that area be covered in sea ice? How have they produced anomaiesy when there's never been any October data collected there before?

      The latter question was easily answered by looking at the HadSST3 climatology file - measurements for all months in grids above 80N are compared to a fixed -1.8degC value. I think this is a problem for tracking temperature change over time in a particular grid cell.

      According to reanalyses the SST annual cycle above 80N is rather large. The drop from September to October is about 10degC, from -10C to -20C. This annual cycle won't be reflected in SST observations because measurements are only taken in open water, which pretty much guarantees >-1.8degC and it's not going to get far above 0C. If HadSST3 reports +1degC anomalies in a grid cell for September and October that actually means the area was 10degC more anomalously warm in October, but that won't be recorded. In other words, the way upper Arctic SST measurements are collected (i.e. only when there is open water) and processed into anomalies (in comparison to fixed -1.8C) makes it pretty much impossible to see the change indicated by NCEP/NCAR.

      Indeed, HadSST3 reports pretty mild October anomalies in these upper Arctic grid cells. But we know this essentially means these areas were perhaps 15-20degC above normal. This may be part of the reason why it was possible to collect so many measurements - unusual presence of open water areas due to warm temperatures. A case where the existence of measurements is more indicative than the measurements themselves?

      I think this was an issue covered by Cowtan et al. 2015, but interesting to go through a practical example.

      Delete
    4. The latter question was easily answered...

      I have to take that back. The question of how they produced "anomaiesy" is still open.

      Delete
  7. Paul: I asked John Kennedy this question because the determination of anomalies is a key factor in determining the ice-edge bias in our recent paper on comparing models to observations. In HadSST3, anomalies for newly melted cells are determined by fitting the SST observations for those cells to a pre-existing climatology which is described in the HadSST2 paper.

    This is rather opaque and hard to replicate with the model outputs. It also means you need a different procedure for each observational dataset, and that procedure depends on factors which may not even be documented because no-one realised they might be important. Hence one of the aims of our work was to try and trigger a rethink of the preparation of both model outputs and observational data to allow a more realistic comparison to be made. It's not an easy problem though.

    ReplyDelete