Apr 21 2008
Interannual Divergence in Satellite Temperature Records
Several months ago, our favorite four-legged animal looked at the annual variations between the RSS and UAH temperatures. He showed that “…there was a systematically increasing difference between RSS and UAH (shown by the trend line btw 1979 and 2005)…” That is, the UAH temperatures were warmer in the past and cooler in the present compared to the RSS values.
I decided to replicate his analysis using the monthly data. Actually, I did the analysis first, and then went searching the web to see if anyone else could explain what I was seeing. In any case, I took the RSS and UAH data from my climate metrics file. I then subtracted the RSS values from the UAH values for each month. Note: this is the opposite as what The BunnyTM did.
There are two main things that jumped out at me from this simple graph. I’ve highlighted them both below.
 |
First, the earlier years in the UAH data are warmer than the RSS data. This is the same finding as Eli found in January.
But what I found is that there doesn’t appear to be a relatively slow decrease in the difference, but instead there appears to be a jump at around 1992. I’ve highlighted this on my plot by using the colors red and blue. From eyeballing the graph, it appears that on average pre-1992 years were around 0.5C warmer than post-1992 years.
The main thing that shocked me was that for most of the post-1992 time, there is a definite interannual signal. Keep in mind that both of these datasets are supposed to be monthly anomalies. That is, they are anomalies from the monthly average. So there should be no interannual signal at all in either of the time series.
When I subtracted the RSS temperature from the UAH temperatures, this removed the covarying signal from the satellite data. What I have plotted is the noise. If the two time series were independent, I think that the noise would be either white or maybe red. The difference time series should not have a period of 1 year.
To be sure that the ‘noise’ represented a 1 year cycle, I plotted in grey a sinusoid with a period of exactly 1 year and an arbitrary phase to make the cycles line up. This is tough to see in the figure above, but easier in the larger, linked version. I also plotted the FFT power spectrum of the noise below.
 |
Before calculating the power spectrum, I first subtracted the long-term running mean to remove any extremely long cycles or trends. I then applied a Hanning window. This last step turned out to be for kicks and giggles since it didn’t actually influence the resulting power spectrum.
The plot clearly shows that the difference ‘noise’ clearly has a significant spike in the power spectrum at a frequency of 1 cycle per year. Neither the RSS or UAH data should have a significant peak in power for a period of 1 year. Therefore, the difference shouldn’t have a significant peak for a period of 1 year. Right? Or am I missing something fundamental here?
(No Ratings Yet)
Loading ...Related Posts:
12 Responses to “Interannual Divergence in Satellite Temperature Records”
To reduce spam, comments are automatically closed 30 days after the last comment. If you would like to comment on any closed thread, please use the contact form at the top of this page.
Have you thought of comparing each of them to GISS or HadCrut to see if you get something similar? I say do both if you can just to see if there is a problem with one or the other relative to the other data sets.
Some more power spectrums. The power at 1 year is greater than for uah-rss, but relative to surrounding frequencies, it is much less.
The first thing to look for if you think there is a sudden jump in 1992 is a new satellite. Also there really is a slow decrease btw 1978 and 1992 before the jump.
[Reply: From here:
That's about 3 months prior to the 'jump'. And there was overlap with NOAA-11.]
Looks like you got a change point there. good detective work.
It’s a fun game to try to guess where the one-year-period difference comes from.
As I understand it, the two satt networks “see” pretty much the same distribution of surface. (Do please wield the clue bat if I’m wrong on that.) So it shouldn’t be anything like the anomaly being different at the extreme north and south and only one satt network sees it.
It’s odd that the signal is growing in amplitude but not drifting, at least not much. That tends to suggest that something is being interpreted differently between the two networks, such as season dependent ground cover or surface of ice or some such. Then the larger amount of land in the norhtern hemisphere would be giving a periodic signal.
Hi Dan, they actually analyze the same data from the same satellites. Also FWIW replaced here is tricky, because a satellite can be on orbit for a while before it starts taking data.
[...] from Atmoz Atmoz comments: The plot clearly shows that the difference ‘noise’ clearly has a significant [...]
atmoz have a read of lucia’s take on this, and spence UK in the comments
Atmoz–
Spence said he was using older data. When he downloaded the new data, and processed he gets what you got.
So what you are seeing is definitely there. The question is why. I gave it my shot– which is pretty much “it’s what’s left over”. It’s interesting you always get the 1 year. That would be consistent with “it’s left over”– but it might be consistent with other theories too.
[...] a week ago, I looked at the apparent interannual divergence in satellite temperature records. Lucia also blogged about why the UAH vs RSS temperature differences might have energy at 1 year. [...]
[...] blogged before about the differences in the two satellite tropospheric temperature records. There are two major features of the difference time series: one is an apparent 1-year periodicity, [...]
[...] records. I first blogged about how there is a divergence in the two time series (RSS and UAH) with power at 1 year. In the same post, I noted that there appears to be a step in the difference time series, but at [...]