Satellite Temps Getting Closer
Posted by Jeff Id on October 28, 2009
Dr. Christy offered some explanations for the divergence in RSS vs UAH which is visually related to land vs sea, originally discovered in blogland by Chad at treesfortheforest – link on the right. Chad is doing some very interesting work beyond this minor issue. His plots lately could take a dozen posts up.
RSS computes diurnal corrections based on climate model diurnal signals for each grid. UAH calculates the effect empirically (from 3 AMSUs which observed the earth simultaneously for about 13 months but at different times of the day – so sampling a specific grid 6 times per day from which we could reconstruct a diurnal cycle for land or ocean for each month of the year.)
UAH focuses on large-scale precision, so at each latitude a single land diurnal correction is applied to all land grids in that band, and similarly for oceans. It turns out that the noise generated in trying to calculate the diurnal correction of a single grid is so great, that more harm than good comes of it. Our goal is to make the zonal and large scale averages as precise as possible, so we deal with large scale corrections since the noise is beaten down that way. In a pleasant outcome, our gridpoint anomalies more precisely match those of radiosondes at those grids than either RSS or ZOU-STAR
So for the pre-2002 years RSS using models to correct for diurnal cycle and UAH using a 3 satellite calulation from actual data came up with what visually seem to be the same answer. The next figure shows that the main difference between the two sets seems to be latitude based prior to 2002. Two completely different methods got what appears to be almost the same result.
Unfortunately for those of us who like things simple, that’s not the end of the story. I asked for more clarification of how the trends could be 2 C different suddenly in the record over Africa and South America.
The RSS v3.2 relies on NOAA-15 for most of the post 2002 period. NOAA-15 is backing up in time which assumes a warming over land (the earlier temps from 6 p.m. dominate the earlier temps from 6 a.m.). So to correct this over land, RSS “adds” a lot of cooling. The effect is different over the oceans for which RSS “adds” a little warming to compensate for assumed cooling. Note that the backing up of NOAA-15 is the first time a morning satellite has been used through such a long-term drift, so this is the first time such a difference could be identified – hence we wouldn’t have seen it with NOAA-10, and 12 for example. And, previously, the data from drifting by morning satellites, NOAA-10 or NOAA-12, were averaged with afternoon satellites. RSS is using only NOAA-15 post 2004 – so we are seeing the impact of (1) the largest adjustments for longest drift of a morning satellite and (2) a morning satellite that is NOT being averaged with an afternoon satellite, so its errors will be exposed more clearly.
This explains a lot of my questions, look again at the equatorial crossing time plot.
Basically he’s saying that the large drift of NOAA 15 is currently not being balanced by another satellite. I don’t know why NOAA16 isn’t being used but as you can imagine by the short lives of the individual satellites, the individual instruments often go out over time, also the correction factor is too strong for land/ocean relationship. Basically Dr. Christy claims it has been that way all along and it’s only now showing up because there’s no opposing saellite data to balance it. It’s good evidence that the RSS satellite temp curve needs some work.
What I’m not confident has been resolved entirely is the pre-AQUA UAH, the recent AQUA data has exposed a real issue with the datasets. Dr. Christy also said the following.
Our contention is that RSS’s adjustments are actually over-corrections which is best seen when the satellites need the largest adjustments. So, as NOAA-14 drifted to diurnally-cooler temps through 2002 or so, RSS over-adjusted the temps to make them too warm. Similarly, with NOAA-15 now drifting into warmer temps (diurnally), the RSS adjustment over-cools the temps – seen especially over land (i.e. since 2002 RSS would show cooling relative to UAH). There is a paper in press which compares RSS, UAH and the new ERA-Interim Reanalysis (since 1989 so as to correct a flaw in the old ERA-40 related to poor handling of HIRS 11 during Pinatubo). The new ERA-I matches UAH exceedingly closely and demonstrates that RSS possesses relative warming through 2002 then a relative cooling thereafter – just as I had described above. So, here we have independent evidence for our contention.
The NOAA 11 is he discusses is likely the step transition seen when taking a difference of the two datasets. Christy has written on this particular step in the dataset where he used radiosonde data to determine that RSS was the series which was out of alignment. I did a simpler analysis using GISS data and got similar results HERE. I wasn’t totally happy with the result b/c of some sensitivity in result to time windows chosen but it turned out that giss based trend corrections preferred UAH results over RSS. – See Christy et al, Tropospheric temperature change since 1979 from tropical radiosonde and satellite measurements Published in JOURNAL OF GEOPHYSICAL RESEARCH, VOL. 112, 2007 – It’s a truly nice paper, there’s a lot of information which is relatively easy to grasp.
With the variety of corrections and this new explanation, I feel we’re getting closer to an accurate and trustworthy lower troposphere trend. Personally I don’t trust the land metrics at all, the corrections are too large and too arbitrarily quantified. However, there is still a difference in annual signal in RSS vs UAH which can be shown through a (fast) fourier transform that shows a stronger annual signal for UAH in recent years. This that means the point is not entirely settled. I almost accidentally nailed the timeframe for the AQUA satellite transition to after 2003 in this post here. See Figure 1, where the 1 year spike is much greater than the 1 year spike in RSS. This means UAH has a different annual max/min average curve from before the AQUA switch and therefore means UAH likely wasn’t correcting for diurnal trend perfectly either. I say almost accidentally nailed, because the 1 year cycle was visible in the anomaly but I didn’t know why. In other words UAH probably ain’t perfect either.
Anyway, here’s a plot of the difference in trend between RSS and UAH global data. I took the global trend as averaged by the source scientists and re-anomalized them for post June 2002 data. The global UAH-RSS difference on average doesn’t reach significance compared to series noise, there isn’t much question that the land average alone or sea average alone don’t though.
Our contention is the same as yours, that the diurnal adjustments regionally, and globally are overdone in RSS. RSS is working on a new version and may have some of these problems solved soon.
Thanks again to John Christy for giving some insight into what’s happening behind the scenes. It should be interesting to see how these issues are resolved in the near future. The nice thing about the satellite data is that many of these issues while complex, are quantifiable. Science minded people who are skeptical of AGW claims are said to prefer satellite records over ground measurements simply because of the lower trends. In my opinion the preference is due to the cleaner and more sensible data. After all, it’s not like satellites don’t show statistically significant warming.