Sunday, May 31, 2015

Early note on May temperatures

There is only a few hours left of May here, and the NCAR reanalysis has three days of data to go. May has had fewer extremes than earlier months, but warmed toward the end. It is currently showing the same average anomaly as February, and I think that is about where it will end up. That is actually not quite as warm as last May, or as March 2015, but it continues a warm start to the year. The GISS anomaly for February was 0.8°C.

26 comments:

  1. Assuming the GISS May anomaly is at least .75C, that makes the 5-month mean for 2015 .77C, and the El Nino, which flagged earlier in the year, probably has a bunch more boost in it. What is reasonably possible for a 2015 spread? Is.90C within sight on the high end? It seems unlikely 2015 could fail to beat 2014, so .68C seems a reasonable number for the bottom.

    ReplyDelete
    Replies
    1. A naive projection (regressing annual mean T on Jan-May mean T, and assuming May will be 0.8) would put 2015 at 0.75 ... because years with higher (or lower) than normal temperatures in the early months tend to regress to the mean a bit later on. The 2 sigma range would be 0.62 to 0.88.

      Based on that, the 95% CI for 2015 would extend slightly below 2014, 2010, and 2005. So 2015 will *probably* set a new record, but there is a non-trivial chance of ending up in 2nd, 3rd, or even 4th place. At the other extreme, 2015 could realistically approach 0.90, but probably not reach it.

      But that's ignoring all the context (ENSO, and whatever other context you think might be relevant). So YMMV.

      Delete
    2. David Appell is predicting .63C. That seems a tad high to me.

      Delete
    3. I think you mean 0.83 for GISS? Yes, maybe hign. My best guess is, same as February, o.8..

      Delete
  2. Yes, you have to account for the fact that the AMO is about to go negative!! I remember the year when an early monthly anomaly topped .90C, and Tamino said anything could happen. The GMST nosedived at the end of the year. I won't get ahead of myself again until the end of next month.

    ReplyDelete
  3. Based on the NCAR May value, I would predict GISS May 2015 anomaly to 0.82 C (0.72 , 0.91) 95%CL by regression.
    And, assuming the GISS March and April values remain unchanged, GISS MAM value to 0.79 C (0.76 , 0.82).

    ReplyDelete
  4. The regression (using the 1997-2014) prediction for the full J-D GISS year 2015 is 0.72 C (0.64 , 0.80) without using enso data.
    Using the (up to now only weak positive) first 4 month of the MEI, I get 0.72 C (0.64 , 0.79).
    For the probability of a new record high, assuming normal distribution of the error, I got
    72% for May, 23% for MAM, and 85% for the J-D year 2015 (84% with enso data).

    ReplyDelete
  5. RSS May is up 0.135 compared to April, which indicates that GISS could reach 0.85.
    The first value for TempLS, based on ERSST and CONUS land only, is quite low (0.564) but will probably rise above 0.7 with more land data.
    With all the fuss about new ISTI, GHCNV4, ERSSTv4 and MLOST, I hope that NOAA doesn't reduce the service with daily updates etc. of the old data bases and temperature series..

    ReplyDelete
    Replies
    1. The GHCN situation is rather bizarre. The latest data file is dated 27 May. And it has no data except for US. Usually there is quite a bit of new month data by the 5th. And how that US data could actually be May data (if it is) is a mystery.

      So ignore TempLS for the moment - it is just SST data. I start publishing when ERSST comes out, because land data by then is not complete but substantial. I may have to review that.

      Delete
    2. Update - I see that GHCN is bringing out v3.3.3 in early June. I guess that is the reason for the delay.

      Delete
  6. Olof - .85C would be good as my evil plan is for June to hit .90C. And then the big El Nino numbers will start July through November.

    ReplyDelete
  7. I improved my regression and including data now from 1970-2014 (old 1997-2014), but excluding year followed large volcanos. I use now also the first 5 month of MEI. But the main difference in prediction for the year is due the longer time horizon.
    My results now: best estimate, (95% CL), probability of new record (old record,year).
    May 2015 : 0.81 C (0.73 , 0.89) 70% (0.79 C, 2014)
    MAM 2015 : 0.79 C (0.76 , 0.81) 15% (0.80 C, 2010)
    June 2015 : 0.79 C (0.64 , 0.95) 71% (0.75 C, 1998)
    Year 2015 : 0.79 C (0.71 , 0.87) 99.7% (0.68 C, 2014)

    ReplyDelete
  8. With the HADSST3 ocean surface numbers coming in at .593 for May 2015 (warmest ever for may) compared to .477 for May 2014, I would think that this May should be .80 to .84 for NASA LOTI. Do the ocean surface numbers used by NASA or NOAA really differ that much from HADSST3 numbers? Climate Reanalyzer averaged about .59 for May in case anyone is wondering. I have found if you add .16-.26 you get the monthly value for NASA. Also another question, does NOAA plan to incorporate the approaches from Karl 2015 anytime this year?

    -TC

    ReplyDelete
    Replies
    1. TC,
      Rather to my surprise, when GHCN resurfaced with V3.3, TempLS Mesh shows May cooler than April, despite warmer SST. I have a new post about that.

      On NOAA, I expect that they will start using V4 of ERSST soon, possibly this month. The first V4 results started coming out back in December. I'll change when they do.

      Delete
  9. If I repeat my regression for GISS May 2015 with TempLSmesh replacing NCAR, I get 0.71 C (0.63 , 0.79). This is lower, but the correlation of GISS and TempLSmesh is better, so it may be a better estimate. RSS.MSU and HADSST3 are not usefull in predicting monthly GISS temperatures.

    ReplyDelete
    Replies
    1. Something is odd with GHCN this month. The latest version has about 300 less station than the day before. I'm checking.

      Delete
  10. 0.71 for GISS May is way to pessimistic. ERSST is actually up 0.07 from April (found it at KNMI explorer)
    My standard recipe for early guesses is two thirds ERSST and one third RSS land ( +0.29 from april) which gives a global change of +0.14 from April, indicating 0.85 for GISS May.

    ReplyDelete
  11. GISS May data now out: It is 0.71 C. Equal to my TempLSmesh regression mean estimate.
    GISS uses also GHCN.
    If you look at the distribution it is very cold in the Arctic and Antarctica. The 250 km smoothing mean would be higher.

    ReplyDelete
  12. Interesting. June stared out very hot. Hit .400C on Nick's tracker, but it is cooling off now and the forecast looks cool for another week. .71C remains very warm.

    ReplyDelete
  13. Where is the announcement?

    ReplyDelete
  14. Whoops, refreshed the GISS page and it updated.

    ReplyDelete
  15. I still dont trust the new GHCN 3.3, it feels like some kind of beta test error. For contrast, JMA is also out today with the warmest May ever, up 0.06 compared to no 2, May 2014.

    ReplyDelete
  16. GISS is also very warm from 60S to 75N but very cold north or south of that. Maybe it is possible to compare the GISS with the NCAR result averaged over May?
    Also in July 2014 NCAR and GISS global mean differs a lot.

    ReplyDelete
    Replies
    1. I have compared TempLS Mesh with NCAR in my latest post. TempLS agrees with GISS, using much the same data. I think the key difference is in E Siberia.

      Delete
  17. I improved my regression by including the GISS May value.
    My results now: best estimate, (95% CL), probability of new record (old record,year).
    June 2015 : 0.77 C (0.61 , 0.92) 58% (0.75 C, 1998)
    JJA 2015 : 0.75 C (0.60 , 0.90) 74% (0.70 C, 1998)
    SON 2015 : 0.82 C (0.67 , 0.97) 82% (0.75 C, 2014)
    Year 2015 : 0.78 C (0.71 , 0.85) 99.8% (0.68 C, 2014)

    ReplyDelete