The US National Center for Environmental Prediction (NCEP) has produced a Climate Forecast System Reanalysis (CFSR) based on “all available conventional and satellite observations”. Most of these data were used to initialize real-time global weather forecast model runs four times each day since 1979. This reanalysis can also be used to estimate annual global temperatures and temperature anomalies. The University of Maine Climate Change Institute has compiled the CFSR data and provided an easy-to-use interface for viewing some of the data using maps and graphs with their Climate Reanalyzer web site.
I pulled CFSR annual global temperature anomaly data from the Climate Reanalyzer for 1979 through 2013 and added a compatible estimate for 2014 from the Weather Bell model temperature web page to complete the period 1979-2014. I then graphed this data against the US National Climatic Data Center (NCDC) estimates based on the Global Historical Climate Network (GHCN) for the same period, as shown in Figure 1 below. Both data sets have been normalized to the 1981-2010 period for comparison.
In general, the two approaches show a similar result, but there are some interesting differences. These differences help to indicate some of the uncertainty in trying to estimate a global temperature anomaly as discussed in my previous post. Of particular interest is the result for the most recent portion from 2001 through 2014. The “pause” in the NCDC estimates is actually more of a peak and decline in the CFSR estimates. The warmest years in the CFSR estimates are 2002-2003 and 2005-2007 with a peak in 2005. In contrast, the warmest year estimated by NCDC is 2014. In the CFSR data, 2014 ranks 12th for the 36-year period – far from being the “hottest year ever” as promoted by some.
Considering that the NCEP CFSR approach incorporates a much larger data set with much better spatial coverage for estimating global temperatures than the NCDC GHCN approach, I suspect the CFSR annual estimates are more accurate.