Guest Post By Walter Dnes
As a follow-up to my April 30th post attempting to project April global anomalies this post will attempt to project May anomalies for various global data sets, before they are released. But first, let’s look back at the April projections, and see how well they fared.
April 2016 Projections
The NCEP/NCAR re-analysis is based on the “sig995” data set. The 995 millibar level is close to (and in some cases theoretically below) sea level. One would expect that the NCEP/NCAR data would more closely match the surface-based global data sets, than the satellite-based lower-troposphere data sets. Indeed, GISS and NCEI come in within approximately +/-0.03 degree of the projections, while RSS and UAH are approximately 0.14 and 0.20 off, respectively. HadCRUT4 April data is not yet in as of this writing.
May 2016 Projections
NCEP/NCAR runs 2 days behind real-time, so data from the first 29 days of May is available. The last 2 days are assumed to be the same as the 29th, when calculating the monthly mean. There is an additional complication with the May projections. The NCEP/NCAR index jumped sharply from +0.368 in September 2015 to +0.567 in October 2015. For at least the satellite data sets, if not all 5, there was a sharp discontinuity. The data on each side of the divide appear to be different populations. The data associated with NCEP/NCAR >= +0.567 has a steeper slope (“the blade of the hockey stick”) than the data associated with NCEP/NCAR <= +0.368 (“the shaft of the hockey stick”). This month, NCEP/NCAR is in the transition zone between the two populations, so one has to choose one of 2 possible values.
To avoid confusion, I’ll list my picks for May 2016 first, and which data set I used. The rationale will follow afterwards, along with the 2 data tables. This month, I’m switching over to my own calculated version of the NCEP/NCAR monthly anomalies, instead of copying numbers from Nick Stokes’ web page. As noted in my previous post, my calculations are in very close agreement to his, but you may see the occasional difference in the last (thousandths of a degree) digit.
- HadCRUT4 +0.834 (shaft)
- GISS +0.96 (blade)
- UAHv6 +0.299 (blade)
- RSS +0.414 (shaft)
- NCEI +1.0005 (shaft)
Rationale for my choices
The lines of “the shaft” and “the blade” meet somewhere in the transition zone between NCEP/NCAR = 0.368 and NCEP/NCAR = 0.567. In the transition zone, I go with the value from “the blade”, unless it crosses below “the shaft”. In other words, I go with the higher of the 2 values. The crossing point need not be the same for each of the 5 data sets. First, here are the data and results from the 7 months of data from “the blade”. Note that there are minor changes from the values given the previous month. The data sets always seem to make minor changes in the past few months.
In the case of HadCRUT4, I had no choice, because April data is not yet in. I’m assuming that the “shaft” data is either the correct choice, or close enough. Note that this is only possible in a transitional month. I.e. the current month is slightly above +0.45, so I use the “shaft” data (below +0.45).
Here are the data and calculations for “the shaft of the hockey stick”. The 12 most recent months with NCEP/NCAR <= +0.4 are used, i.e. October 2014 through September 2015.