Archive for May, 2015

Has UAH’s new method made v6.0 inferior?

May 3, 2015

I’m afraid the answer appears to be yes. Specifically the method appears to actually move UAH in the direction of RSS’s two major “problems.”

We previously discussed how RSS appears to have a serious drift problem associated with NOAA-12, and the period during which UAH uses AQUA as the “backbone” of it’s analysis, during which RSS arguably developed a cooling bias. UAH’s new method, currently in beta, is described here. I was enthusiastic about the possibility that UAH had in fact significantly improved upon their old methods in some way, but I was skeptical as well. I decided to do two things: first, to see whether I could still detect RSS’s discrepancies from UAH, with the causes I’d previously identified, and second, to examine more closely the differences between UAH v5.6 and the new v6.0. The first test, comparing v6.0 RSS, turned up something surprising: during the period from September of 1991 to April of 1995 (with UAH and RSS re-anomalized to their 1979-2014 means) UAH now warms relative to RSS-albeit very slowly. This was a major red flag, as it indicated that UAH probably now had a worse spurious drift problem during the NOAA-12 transition than RSS. Taking the differences between v6.0 and v5.6 confirmed that the new version warms relative to the old version during the NOAA-12 transition period we previously discussed. A drift of about .02 K per annum for about 44 months amounts to a jump of about .085 K. Previously attempting to use surface temperature data in this limited interval to adjudicate the dispute between UAH and RSS over this period suggested UAH had been correct and RSS wrong-such an analysis performed now would probably lead to the conclusion that RSS and UAH both have a severe warming bias in this particular interval. Worse still, during the AQUA interval, during which v5.6 should have been stabilized compared to RSS, v6.0 now has no trend difference at all! In other words, if-as I believed-UAH’s use of AQUA as the primary satellite during that interval (whereas RSS simply treated it as just another satellite after applying their drift corrections to them) then the drift of RSS cool during that period should have strongly indicated that RSS was flawed. But UAH now agrees with RSS over that interval, because over the AMSU period UAH v6.0 now has a significant cooling drift, too.

For the moment I’m going to be sticking with v5.6 unless Roy or John can reasonably convince me that v6.0 actually is an improvement. I’m going to leave a comment on Roy’s blog so hopefully they can address my concerns.