Archive for November, 2015

UAH version 6 may not be so bad after all

November 8, 2015

In my previous post, I questioned the reliability of UAH’s revision of its dataset for its most recent version of the Lower Tropospheric Temperature anomaly. I never did hear back from Roy about the issues that I raised that made me think the new adjustments were resulting in a flawed product that was in fact inferior for accurate trend assessment over the long term to the previous version of the dataset. But a recent assessment I have done suggests to me I was perhaps too hasty in concluding UAH was now biased for trend assessment, at least over the long term. Specifically, I revisited an earlier analysis I did.

Previously, I argued that UAH data over the continental US agreed well with NOAA’s USHCN data, especially when adjusting for the higher variance of the surface relative to the atmosphere (whereas the opposite is true globally. Recalling from the discussion around Klotzbach et al. this anti-amplification is actually about what is theoretically expected for a mid latitude land area) and that this implied that, if one believed the adjustments in the USHCN data, UAH should also be pretty accurate-either that, or the agreement was a rather unlikely coincidence. The US has one of the densest observing networks in the world, and, it would not be unreasonable to suppose, probably more accurate data than many other countries. But when I repeated my analysis, this time using UAH version 6 (currently in its third beta iteration) I find that there is still essentially no trend difference over the US after variance adjustment. USHCN warms around .027 K per decade relative to variance adjusted UAH v6 and if I perform the exact same procedure to estimate the variance adjustment with v5.6 it warms just .015 K per decade relative to that data. So UAH v6 has comparably good agreement with the USHCN record as v5.6 did, suggesting there is little reason to think UAH v6 should have a long term trend bias. I still have my misgivings about the behavior of v6 over the shorter term, where it appears to have some spurious discontinuities and drifts, but the analysis I’ve just done has convinced me that these effects probably mostly cancel out over the long term. There is not then that much reason to prefer UAH v5.6 to v6.