In my
previous post, I noted that comparing data on different anomaly bases brought in discrepancies because of vagaries of monthly averages over the base period. These are actually significant enough to detract from the use of the
NCEP/NCAR index (base 1994-2013) in predicting GISS (1951-80) and NOAA (1961-90) indices.
So I have added to that table, below the NCEP monthly averages, recent monthly values adjusted to the earlier basis periods by adding the 1994-2013 month averages of GISS and NOAA anomalies. This expresses NCEP on those scales, although strictly, it is unique to those indices. Adding HADCRUT averages would give slightly different 1961-90 values.
I have also shown the actual GISS/NOAA values in the right column. The current tables are here:
GISS adj
| NCEP | GISS
| Jun | 0.835 | NA
| May | 0.773 | 0.71
| Apr | 0.702 | 0.71
| Mar | 0.855 | 0.84
| Feb | 0.818 | 0.82
|
|
NOAA adj
| NCEP | NOAA
| Jun | 0.618 | NA
| May | 0.558 | NA
| Apr | 0.461 | 0.469
| Mar | 0.573 | 0.564
| Feb | 0.545 | 0.575
|
|
Except for GISS May, the correspondence is very good.
Update. I calculated a whole lot more values, in a table below the fold. It looks as if the last few months aren't typical. The correspondence is still reasonable, but discrepancies as observed April-May are not unusual.
HadCRUT adj
| NCEP | HadCRUT
| Jun | 0.743 | NA
| May | 0.671 | NA
| Apr | 0.608 | 0.655
| Mar | 0.726 | 0.68
| Feb | 0.719 | 0.66
| Jan | 0.616 | 0.69
| Dec | 0.595 | 0.63
| Nov | 0.546 | 0.487
| Oct | 0.721 | 0.62
| Sep | 0.655 | 0.592
| Aug | 0.669 | 0.666
| Jul | 0.581 | 0.544
| Jun | 0.553 | 0.62
| May | 0.711 | 0.596
| Apr | 0.669 | 0.658
| Mar | 0.652 | 0.548
| Feb | 0.438 | 0.305
|
|
GISSlo adj
| NCEP | GISSlo
| Jun | 0.835 | NA
| May | 0.773 | 0.71
| Apr | 0.702 | 0.71
| Mar | 0.855 | 0.84
| Feb | 0.818 | 0.82
| Jan | 0.716 | 0.75
| Dec | 0.715 | 0.73
| Nov | 0.686 | 0.63
| Oct | 0.829 | 0.78
| Sep | 0.769 | 0.82
| Aug | 0.745 | 0.74
| Jul | 0.649 | 0.5
| Jun | 0.645 | 0.61
| May | 0.813 | 0.79
| Apr | 0.763 | 0.72
| Mar | 0.781 | 0.71
| Feb | 0.537 | 0.44
|
|
NOAAlo adj
| NCEP | NOAAlo
| Jun | 0.618 | NA
| May | 0.558 | NA
| Apr | 0.461 | 0.469
| Mar | 0.573 | 0.564
| Feb | 0.545 | 0.575
| Jan | 0.454 | 0.5
| Dec | 0.451 | 0.5
| Nov | 0.442 | 0.424
| Oct | 0.611 | 0.514
| Sep | 0.546 | 0.495
| Aug | 0.534 | 0.511
| Jul | 0.449 | 0.416
| Jun | 0.428 | 0.475
| May | 0.598 | 0.491
| Apr | 0.522 | 0.497
| Mar | 0.499 | 0.447
| Feb | 0.264 | 0.149
|
|
TempLSgrid adj
| NCEP | TempLSgrid
| Jun | 0.715 | NA
| May | 0.655 | 0.667
| Apr | 0.578 | 0.618
| Mar | 0.681 | 0.669
| Feb | 0.675 | 0.668
| Jan | 0.565 | 0.641
| Dec | 0.576 | 0.643
| Nov | 0.535 | 0.512
| Oct | 0.703 | 0.627
| Sep | 0.628 | 0.598
| Aug | 0.623 | 0.606
| Jul | 0.531 | 0.51
| Jun | 0.525 | 0.581
| May | 0.695 | 0.61
| Apr | 0.639 | 0.606
| Mar | 0.607 | 0.546
| Feb | 0.394 | 0.318
|
|
TempLSmesh adj
| NCEP | TempLSmesh
| Jun | 0.713 | NA
| May | 0.655 | 0.614
| Apr | 0.605 | 0.628
| Mar | 0.7 | 0.709
| Feb | 0.698 | 0.699
| Jan | 0.592 | 0.663
| Dec | 0.605 | 0.658
| Nov | 0.582 | 0.574
| Oct | 0.725 | 0.659
| Sep | 0.64 | 0.669
| Aug | 0.634 | 0.62
| Jul | 0.527 | 0.464
| Jun | 0.523 | 0.58
| May | 0.695 | 0.67
| Apr | 0.666 | 0.643
| Mar | 0.626 | 0.582
| Feb | 0.417 | 0.368
|
|
I also compared NCEP and GISS temperatures and I think now there can be no doubt : each time you have cold anomalies in polar regions, the anomaly is lower for GISS that NCEP cfsv2. The opposite is true : look at may and september 2014 : these two months were warmer for GISS that for NCEP. Both months were quite hot in polar regions. July 2014 is quite extraordinary with low anomaly for GISS : it was cold in the Arctic an Antarctica. GISS is covering those regions and any discrepancy shows that something is going on there.
ReplyDeleteFortunately, GISS posts 2x2 grids of temperature anomaliess, and NCEP also posts 2x2 grids. I'll see if I can post difference maps (with global averages).
DeleteOops - memory lapse. NCEP is 2.5x2.5. Not so easy.
DeleteOk, different base periods explain some of the discrepancy. An other source of error could be that reanalyses calculate 2 m air temperatures, but the global observational datasets use 71% SST, and that is not always the same..
ReplyDeleteAnd then we have the polar regions that are poorly covered:
I encountered "strange" values when I ran NCEP/NCAR 2m for the Antarctic area 70S-90S at KNMI Climate Explorer. The May anomaly was +1.42 (1981-2010 base). That is not what the Gistemp map tells, the zonal anomaly for 70S-90 S is more like -1.7 for May (1981-2010 base).
I observed the difference over Antarctica too. The difference may be due the use of the sig995 level instead of 2m surface temperatures so missing inversions in the Antarctic winter?! See
ReplyDeletehttp://www.moyhu.blogspot.com.au/2014/11/a-new-surface-temperature-index.html
Olof, Anonymous, you might be right about the causes of discrepancy. When I check NCEP cfsv2 anomaly, it is to have an estimate of futur GISS data. Usually, you can have an estimate. But sometimes it doesn't work so I tried to understand why. Anomalies in polar regions (that you can see on NCEP month-to-date) seems to be associated with low giss global anomaly. So my question would be : can you use NCEP cfsv2 data to estimate Gistemp ? I would say it is possible to have an estimate if you make a correction, depending on anomalies in polar regions.
ReplyDelete