Yesterday, NCDC released Version 2.5 of the USHCN data set. For those who don’t know, this is the U.S. Historical Climatology Network (USHCN) which NCDC considers a “gold standard” for US temperature measurements. Problem is, the reality of the old network is that it is fraught with all sort of inconsistencies throughout its record such as multiple station moves, equipment changes, time of observation changes, encroachment by urbanization, and of course faulty station siting which I discovered that only 1 in 10 USHCN stations met the criteria of NOAA’s 100 foot rule, a charge backed up by an investigation done by the U.S. General Accounting Office (GAO). As a result of the most recent research, we found significant positive biases in the raw data:
Of course the line from NCDC is always that “none of this matters” and that such things can be solved by adjustments. After seeing the differences between the USHCN2.0 and USHCN 2.5 data set, I ask: “in what temperature measurement universe does a hockey stick like this occur“? See below.
Graph from Steve Goddard, source here.
How can adjustments that affect only one decade like this be justified? The answer is that they can’t.
Here’s what NCDC said about it yesterday in their monthly State of the Climate report:
USHCN Version 2.5 Transition
Since 1987, NCDC has used observations from the U.S. Historical Climatology Network (USHCN) to quantify national- and regional-scale temperature changes in the conterminous United States (CONUS). To that end, USHCN temperature records have been “corrected” to account for various historical changes in station location, instrumentation, and observing practice. The USHCN is a designated subset of the NOAA Cooperative Observer Program (COOP) Network. USHCN sites were selected according to their spatial coverage, record length, data completeness, and historical stability. The USHCN, therefore, consists primarily of long-term COOP stations whose temperature records have been adjusted for systematic, non-climatic changes that bias temperature trends.
Did you know — the National Climatic Data Center periodically improves the quality of the datasets maintained at the center and releases updated versions’ Beginning with the September 2012 processing, NCDC will use USHCN version 2.5 for national temperature calculations as well as in other products, including Climate at a Glance and the Climate Extremes Index.
For additional information on the improvements made to USHCN version 2.5, please see http://www.ncdc.noaa.gov/oa/climate/research/ushcn/.
The page at http://www.ncdc.noaa.gov/oa/climate/research/ushcn/ describes some the procedures, but makes heavy use of references on published papers, rather than provide an operational flowchart. This is an impediment to replication, and I suspect that given the mishmash of references they provided, few if any would be able to replicate the process fully.
Steve Goddard did an additional analysis and showed how the temperature changes over the different phases of the USHCN data:
Also in the monthly report from NCDC is this statement:
The average contiguous U.S. temperature during September was 67.0°F, 1.4°F above the 20th century average, tying September 1980 as the 23rd warmest such month on record. September 2012 marks the 16th consecutive month with above-average temperatures for the Lower 48.
According to the “platinum standard” state of the art US Climate Reference Network (USCRN) which has none of the problems and requires none of the adjustments of the problem plagued and aging USHCN network, the CONUS monthly average was 66.0°F.
That puts the temperature difference (above normal) at 0.4F, within the bounds of standard deviation. Ever wonder why NCDC never mentions the new state of the art USCRN data in their State of the Climate press releases but prefers to rely on the old network and its wonky hockey stick like adjustments? This is why.
I’ll have more on the official announcement about the map above soon.