I had begun calculations for a post on urban warming, but as usual, I got sidetracked on something else. This is that sidetrack. There’s much interest in the differences in the RSS and UAH temperature data sets. We’ve seen plenty of graphs showing the difference in the spatial averages of both data sets. But I think we can learn more by looking at the differences between the gridded data. I took the difference between UAH and RSS from January 1979 – Sept 2009 and calculated the trend at each grid point.

uah.minus.rssAs you can see, there is strong differential warming of RSS relative to UAH in the tropics. This probably explains why Santer’s analysis of tropical tropospheric amplification went well with RSS and not so much with UAH. Outside of the tropics, the differential warming is in the opposite direction but dependent on hemisphere with UAH in the lead in the North. Approaching the North Pole, the differential warming becomes unmistakably larger and larger. However, approaching the South Pole, the warming rises, then at about 50° S, it starts to drop. I remember reading that RSS’s correction of diurnal drift is latitude dependent, but I couldn’t find the paper or website where I found it. If anyone knows if this is true, I’d appreciate a reference. Here’s the zonal trend.

Now let’s look at how each satellite record compares to the three surface temperature records. To make the comparison, I had to calculate the monthly mean anomaly and subtract it from the entire gridded data set. This is where I ran into some problems. GISS has the best consistent coverage of the three data sets. As you know, GISS produces two anomaly series. One that interpolates anomalies out as far as 250 km and another up to 1200 km. I used the 1200 km series for this analysis. HadCRUT and NCDC however don’t interpolate to fill in missing data. This presents something of a problem when trying to rebaseline the data. If I have a series of January anomalies corresponding to my new baseperiod, the easiest situation is where all the anomalies area real-valued. What does one do when not all of the anomalies area real valued? By computing a mean and ignoring missing values, it could bias the results in unpredictable ways. That wasn’t an issue I was interested in dealing with. So I resorted to only computing the mean when all values are defined. This resulted in a serious loss of coverage for HadCRU and NCDC.

The period of analysis is Jan 1979 – Dec 2008. I chose this period because a) it makes the programming easier and b) the NCDC gridded data only goes out to April 2009. Here’s RSS and UAH minus GISS.


Both data sets show relatively similar patterns of warming but RSS shows stronger more widespread warming in the tropics. Here’s the trend by latitude.

Oct 26- CORRECTION: John N-G discovered an error with the zonal trend calculation. See his comment below. I’ve updated the zonal trend graphs. You can see the three original versions here, here and here.)

Next up is HadCRUT.


It’s not apparent from these two plots, but there is a significant difference in warming in the tropics as the latitude trend shows. 26, UPDATE: I’ve truncated the x-limits of the graph because of extreme values at high latitudes in the South. These values shouldn’t be taken seriously because there are few grid boxes with real-valued trends in this region, so it’s not really a valid spatial average.

Moving on to NCDC,


The pattern of warming for both data sets is almost identical. So identical I thought I may have mistakenly processed the same data set twice. I checked and that wasn’t the case. The latitude trends confirm this.

This plot raises some questions. There is a strong difference between RSS and UAH anomalies. So then why does this difference almost completely disappear relative to NCDC? It’s almost like RSS and UAH suddenly became similar. Strange.

UPDATE! Oct 21, 2009

JeffID suggested that I repeat the trend calculation for the period before and after RSS/UAH were using the same satellites. I recalculated the trend map for the periods Jan 1979 – Dec 1991 and Jan 1992 – Sept 2009. Here are the results.

rss.minus.giss.end.1991rss.minus.giss.start.1992The differences are like night and day. Before 1992, the trends were similar as what I found previously. But after 1992, there is strong differential warming over South America, Africa (with some cold spots) and Australia. I’ll leave it to you discuss the specifics.

This entry was posted in Data comparisons, Satellite temperature record. Bookmark the permalink.

32 Responses to RSS vs. UAH

  1. Pingback: Gridded Comparison of Temperature Metrics « the Air Vent

  2. Jeff Id says:

    Chad, UAH has switched to a different satellite in recent times from RSS. The rest of the record used the same two satellites. People always want more from blogs and don’t understand the huge time this took.

    Still I wonder if you would consider taking a look at the portion of the record where RSS and UAH were using the same input data and perhaps a second windowed comparison to the portion of the record where UAH uses new satellite data. The reason is that to investigate whether the amazing spatial differences are caused by the switch.

    Interesting post.

  3. timetochooseagain says:

    RSS minus GISS has a weird El Nino signature. What’s up with that?

  4. Jeff Id says:

    TTCA left this on tAV. I agree with him so I’ll put it here.

    timetochooseagain said
    October 21, 2009 at 10:45 am e

    From what I’ve seen, UAH as a group has done a number of papers showing that UAH has the best agreement with balloon datasets at all locations.

    Anyway, this is definitely interesting. I wonder what it looks like before and after the “jump” that several groups have identified in RSS in about 1992.

    And RSS’s diurnal correction makes use of a climate model. I’m not sure how that works, but that is what they do, and I would imagine that would change with latitude. Of course, I don’t think that is the issue, since my understanding is that over the oceans, diurnal correction is hardly needed. There seems to be a lot of differences over land and ocean.

    • Chad says:

      I would expect a latitude dependence as well from such an approach. If you look at the annual cycle by latitude from model output, it changes very significantly by latitude so it wouldn’t surprise me that the diurnal cycle shows similar behavior.

  5. Layman Lurker says:

    Here is a document published by Carl Mears Feb. 2007 describing the changes between RSS version 2.1 and version 3:

    Intersatellite offsets now vary as a function of latitude. This leads to changes in the long-term trends when plotted as a function of latitude. These changes are fairly small for TLT, TMT, and TLS, but quite large for TTS (MSU3/AMSU7). The intersatellite offsets for MSU3 are strong functions of latitude, with the later satellites (NOAA-14 and NOAA-12) showing substantially different offsets when compared to the earlier satellites (NOAA-10 and NOAA-11). This coherence between the later satellites results in a large change in the long-term trends as a function of latitude. This difference is large enough that earlier versions of TTS should be considered to be wrong.

  6. BarryW says:

    It would be interesting to see the data mapped into a globe such as google earth so that you didn’t get the polar distortion.

    • Chad says:

      Do you mean onto a map like a Mollweide projection? I can do that. I’ve been working on a function to do that, but I can’t get the legend to show up.

      • BarryW says:

        Yeah, anything that is equal area. I was thinking of Google Earth because I’m familiar with it. With the projection you’re using it looks like there is massive areas at the poles that are different.

      • jorgekafkazar says:

        print it out and put the legend on with a ball point pen. then scan it.

  7. Pingback: Bias In Satellite Temperature Metrics « the Air Vent

  8. SineCos says:

    The picket fencing far south reminds me of when a shifting pattern is partially sampled. I wonder if the direction of satellite samples is partially synchronized with weather patterns in the southern hemisphere. That could cause oversampling or undersampling, causing what looks like clusters of similar data with sudden gaps.

    Is the picket fence in the region where the Southern Ocean is circulating west-to-east? If the measurement is done west-to-east so each sample is within the same weather formation, a blob of similar temperature would be detected with a sudden change when the measurement moves off the edge of the air/water mass.

    This should be visible in short-term measurements. Because we’re seeing the pattern in a long-term map, that implies that there are patterns which often repeat in about the same region. Perhaps South America’s tip tends to cause regular vortices?

    • Chad says:

      I don’t know if the picket fence is moving or not. I could look at some raw model temperature data to see if one could expect South American’s southern tip to shed vortices.

  9. John N-G says:

    A little “eyeball averaging” of your maps suggests that the difference along the equator for UAH-NCDC is about 0.2 C/decade while RSS-NCDC is about 0.3 C/decade. Your graphs show 0.05 C/decade and 0.07 C/decade respectively. Maybe you averaged in all the missing values as zeroes?

    • Chad says:

      I checked my function that calculates the zonal average and I make a mistake. The proper way of calculating the area-weighted average is to multiply the data by the weights, which happen to be the same because the area of a grid box is independent of longitude. The missing values are ignored and the weighted sum is divided by the sum of weights that correspond to real-valued data, thus ignoring the missing values. I accidentally divided by the entire area of the zone. I’ll post a correction. Thank you for brining this error to my attention.

  10. Chad says:

    A correction to the zonal averages has been posted.

  11. Geoff Sherrington says:

    This is excellent work. It overlaps with work I am doing on a much smaller scale in Australia.

    I chose about 15 land stations that Australia’s BOM decided were definitely rural and looked at Tmax and Tmin for the past 40 years, starting with daily data and then presenting annual data. I acknoweledge the fine work by the BOM in collecting the data, which I have in filled on rare days when missing.

    This is too small a subset to generalise, but in the final analysis there was a strong dependence on whether the station was on the coast or inland (usually more than 200km). The coastal trends were close to zero degrees C per century, but the inland ones were about 2 deg C per century if a linear fit was used.

    Thus, there is prima facie evidence that the inland behavior for 40 years at these rural stations has not been the same as coastal. This must be a transient effect, or in some decades before or after the 1968-2008 study period the inland stations would freeze or boil. The GISS and HADCRU versions might well be different through adjustment.

    This table shows the differences in deg C change per year.

    STATION Tmax Tmin Tmax Tmin
    Broome airport -0.0002 0.0033
    Carnarvon airport 0.0158 -0.0001
    Ceduna AMO 0.0176 0.0089
    Charleville airport 0.029 0.0185
    Cobar MO 0.0369 0.0161
    Esperance 0.0072 0.0136
    Forrest air/p 0.018 0.0295
    Giles 0.0192 0.0238
    Gove airport 0.005 -0.0019
    Halls Ck 0.0057 0.0111
    Learmonth airport 0.0125 -0.0048
    Longreach airport 0.026 0.0341
    Lord Howe Island 0.0121 0.011
    Macquarie Island -0.0032 0.0002
    Meekatharra air/p 0.0189 0.0073
    Tennant Creek MO 0.0144 0.0313
    Woomera airport 0.02933 0.014
    Deg C per year 0.0084 0.0038 0.0219 0.0206

    These figures might help in setting criteria for the validity of satellite temperatures, or satellite versus land/sea, in the sense that a difference between sea and land might indeed be present and real, not anomalous and therefore in need of correction.

    On some global maps above, the sea circling the coast of some continents shows as cooler blue, but not enough to generalise.

    I have no idea why there is a difference betwen coastal and inland in the subset I used. There is not much difference in other effects like latitude, or airport/non-airport, or rainfall.

    Lacking the ability to do this work quickly on a large machine, I would very much like to se it extended to other continents to see if I’m simply looking at an undersampled situation.

    Please email me if you’d like a detailed account with data. A short version was posted by David Stockwell at
    dated 4 May 09. Thank you, David.

  12. Geoff Sherrington says:

    Re 11, Sorry, the tabulation dropped out and you can’t tell which are coastal. Coastal are Broome, Carnarvon, Ceduna, Esperance, Gove, Learmonth, Lord Howe Is, Macquarie Is. Station elevations average about 200m for inland sites, max 600m at Giles. Elevation is hard to separate from inland categories.

  13. Geoff Sherrington says:

    Because I do not know of your backgound too well (my fault), I’ll risk stating what you probably already know, that there is extensive literature about conversion of point data to gridded data in the mining industry. I have not looked at this conversion in the current context, but if you are unsure about how robust (I love that word!) it is, there are several consulting firms and quite a few active mathematicians who might be able to suggest beneficial methodology if you felt you needed it.

    • Chad says:

      Sorry for not responding in a timely manner to your previous comments. I was dithering as to what to say in response. What do you mean converting point data to gridded? Do you mean working backwards from a spatial average back to the original data? Are you referring to the interpolation I implemented?

  14. Ian L. McQueen says:

    OT, a comment on your website. As it came in to my computer, the text was very small. As usual, I clicked on “Page” and selected 150%. The text was then readable. BUT, a wide black band remained on the RH side, and the figures “slid under” that, making that part of them invisible. It is not possible to remove the black band, and there is no way to scroll right and left to see all of the figure. The only way to see a figure is to return to 100%, thus making the text almost unreadable again.
    Is there any way to revise your website to allow scrolling R-L?


    • Chad says:

      Thanks for your comment Ian. I just received a similar comment. I’m going to look for a new wordpress theme that has a wider column to solve the problem.

  15. Dave says:

    Geoff Sherrington is probably referring to Kreiging (or Kriging), a method for extrapolating the subsurface contours of an ore body from a grid of core samples. The Canadian government was also looking into the technique for bathymetry.

  16. Robert Wood says:

    I don’t understand the full significance of all this but I do recognize that these deltas between “observations” are very much greater than the supposed global warming; how about the error bars?

    I just don’t see how … oh, forget it, rhetorical question :-)

  17. Geoff Sherrington says:

    For Chad, Above, I referred to the steps involved in taking a series of measurements at points (weather stations, satellite readings) and then interpolating between them to derive values to be assigned to cells of defined size and shape; then to see that edge effects did not happen at cell boundaries and a few other side effects.

    The analogue of this is one of the critical steps in the developmental stage of mining (more often done on 3D data between drill holes) and many mathematicians have investigated ways to do the math. Projects can profit or fail depending on how well it is done. It’s far from a one size fits all option; there are some quite different approaches and not all give the same results or error estimates.

    I do not know if RSS and UAH use the exact same interpolation methods (and I know little about CRU and GISS) but in theory until I was satisfied that the interpolation methods yielded the same results on the same training data I’d be reluctant to ascribe climate causes or input data variations to difference maps. The differences might arise from slight or gross differences in the interpolation packs. (Differences might also arise from the treatment of a cell that is part sea and part land).

    Personally, on the globe, I would not go for cells based on lat and long because that forces different sub-properties to Poles and the Tropics. I’d be consulting the G.P. for a referral to a Specialist. I know that Jeff Id and Ryan have tried other grid shapes.

    Before I retired, companies tended to keep methods to themselves for competitive edge, but there are now several well-tested commercial packages on the market. Others can name the good ones better than I because I’m too rusty.

    I do hope I’m not “teaching grandma to suck eggs” here – you are probably streets ahead of me.

    • Chad says:

      The interpolation scheme I use is pretty straight forward. Nothing terribly advanced. I may do some experimenting with other methods. Who knows how different methods could change the overall picture. It’ll certainly make for an interesting post.

  18. Geoff Sherrington says:

    I’ve started asking around old colleagues for a recent summary of the art, if there is one. Will be in touch. Geoff.

  19. bugs says:

    People always want more from blogs and don’t understand the huge time this took.

    The irony.

  20. Geoff Sherrington says:

    Hi Bugs,

    I hope that I’m not one of the unappreciative offenders. I know that it was a very long job to make these maps. Indeed, in the late 70’s we made several tonnes of them for a foreign country and it took a couple of years. Slightly diffrent data and scale, but I’d guess about the same number of numbers.

    It would be disappointing if there was a blank page reaction to the work above, so I’m trying to contribute another angle that might assist understanding. That’s how science works. Or used to.

  21. TonyS says:

    It would be interesting to see the zonal trend graphs for land only and for sea only.

Comments are closed.