Tisdale on ENSO step changes in RSS global temperature data

RSS MSU TLT Anomalies February 2011 Update and A Look At Version 3.3

FEBRUARY 2011 UPDATE

RSS TLT anomalies continue to drop in response to the 2010/2011 La Niña. RSS MSU TLT anomalies are now at 0.051 deg C, Figure 1.
http://i54.tinypic.com/qn4qvm.jpg
Figure 1

RECENT VERSION UPDATE

RSS recently updated their MSU Lower Troposphere Temperature (TLT) anomaly data with a new version, v3.3. This was discussed last month at Watts Up With That? in the post RSS global temp drops, version change adjusts cooler post 1998. At that time, RSS had not described the changes. They now have at their website on their data description webpage.
http://www.remss.com/msu/msu_data_description.html

Refer to the Version Notes. Here’s what RSS has to say:

##########
Version Notes
RSS Version 3.3 Channel TLT, TMT, TTS, and TLS – January, 2011
Change from 3.2 to 3.3:

* Additional satellites are now included in the merge. Version 3.2 only used data from one AMSU instrument, NOAA-15. For TLT, TMT, and TLS, Version 3.3 includes data from the AMSU instruments on NOAA-15, AQUA, NOAA-18, and METOP-A. AMSU channel 7 exhibits unexplained drifts in METOP-A, so for TTS, data from METOP-A is not used.

* Comparisons with other AMSU satellites are now used to detemine [sic] the AMSU merging coefficients.

* When merging MSU and AMSU together, the data for each generation of satellites is weighted by the number of satellites with valid data for that month. This has the effect of de-emphasizing MSU data after the advent of the AQUA satellite in June 2002. Since the 2002-2004 period is when there is an unexplained warming drift in MSU channel 2 data from NOAA-14 relative to AMSU data, this change has the effect of lowering the overall warming in TMT and TLT during the post 2002 period.

* The changes also result in a reduction of sampling noise and “orbital striping” for periods when data from more satellites is used.

* Data from NOAA-16 is not used because all 3 channels show unexplained drift throughout it’s [sic] lifetime. NOAA-17 was only operational for a short period of time, thus it’s [sic] data is of little use for climate studies. We plan to begin including data from NOAA-19 after 3 years of operation.
###########

VERSION COMPARISON

Figure 2 compares the anomaly data and linear trends of the new RSS TLT Version 3.3 to the obsolete Version 3.2. The update lowered the linear trend since 1979 from approximately 1.6 deg C to 1.5 deg C per Century, Figure 2.
http://i52.tinypic.com/16c8lcw.jpg
Figure 2

The difference between the two datasets is shown in Figure 3.
http://i53.tinypic.com/258uqzb.jpg
Figure 3

Figure 4 is a .gif animation that compares the2010 anomaly maps for the new and old versions when using 1979-1980 as the base years. Basically both maps are showing the change in TLT anomalies from the average of the years 1979 and 1980 to the year 2010. The patterns for both datasets are similar, but there are minor changes in the variations.
http://i54.tinypic.com/261bxja.jpg
Figure 4

COMPARISON TO UAH MSU TLT DATA

The linear trends of the RSS version 3.4 and the most recent version of UAH TLT anomaly data (v5.4) are basically the same: 1.47 versus 1.44 deg C per Century. Refer to Figure 5. Note that I’ve switched to KNMI climate Explorer as the source for both datasets, so that I could limit the UAH latitudes to those used by RSS, 70S-82.5N.
http://i51.tinypic.com/2d0xmds.jpg
Figure 5

Figure 6 shows the difference between the two datasets.
http://i52.tinypic.com/iz263m.jpg
Figure 6

And Figure 7 is a gif animation similar to Figure 4, but this compares RSS (v3.3) to UAH (v5.4) TLT anomaly data.
http://i56.tinypic.com/14w3s6w.jpg
Figure 7

THE ENSO-INDUCED STEP CHANGES

I illustrated and discussed the ENSO-induced rises in the RSS MSU TLT anomalies for the data north of 20N in the post RSS MSU TLT Time-Latitude Plots… Show Climate Responses That Cannot Be Easily Illustrated With Time-Series Graphs Alone. I further discussed the likely cause for the upward steps in the post The ENSO-Related Variations In Kuroshio-Oyashio Extension (KOE) SST Anomalies And Their Impact On Northern Hemisphere Temperatures.

Figure 8 illustrates Volcano-adjusted RSS TLT anomalies north of 20N in “raw” form and smoothed with a 13-month running-average filter. Also included are the period average temperature anomalies of -0.187 for 1979 to 1987, -0.016 for 1988 to 1997, and 0.268 for 1998 to present.
http://i54.tinypic.com/5c0svt.jpg
Figure 8

I adjusted the data for the linear effects of the two major volcanic eruptions, El Chichon and Mount Pinatubo. To determine the scaling factor for the volcanic aerosol proxy, I used a linear regression software tool (Analyse-it for Excel) with global RSS TLT anomalies (v3.3) as the dependent variable and GISS Stratospheric Aerosol Optical Thickness data (ASCII data) as the independent variable. The scaling factor determined was 2.9.

And in Figure 9 the “raw” data has been deleted to help show the ENSO-induced upward steps in this dataset. So the revisions have not changed these to any great extent, so I won’t go back and update the earlier posts.
http://i52.tinypic.com/15p4uia.jpg
Figure 9

SOURCES

The following are links to the data use to create Figures 1, 2, and 3.
RSS_Monthly_MSU_AMSU_Channel_TLT_Anomalies_Land_and_Ocean_v03_2.txt
RSS_Monthly_MSU_AMSU_Channel_TLT_Anomalies_Land_and_Ocean_v03_3.txt

All other data were downloaded, and the maps were created, using the KNMI Climate Explorer Monthly observations webpage.

(Many thanks to Dr. Geert Jan van Oldenborgh of KNMI for the quick update to RSS TLT version 3.3.)

Posted by Bob Tisdale at 12:01 PM

About these ads

43 thoughts on “Tisdale on ENSO step changes in RSS global temperature data

  1. What puzzles me about these type of “satellite” measurements is that I do not find the same trends back on the ground.
    e.g. I have not seen an increase in average temps. since 1974 in Pretoria, South Africa, see here

    http://www.letterdash.com/HenryP/assessment-of-global-warming-and-global-warming-caused-by-greenhouse-forcings-in-pretoria-south-africa

    There was an increase in max. temps. but this was somehow neutralized by a decrease in min.temps
    I double checked my results with La Paz, Bolivia during the dry months there and got almost the same results.

  2. The sensors are calibrated before flight and compared to a onboard reference. So why do the sensors not agree about surface temp straight out of the box?
    Is the uncertaintly in the orbit or a limit in our knowledge of modelling temp from the reading?
    At least with satellites the revision of the past are limited and documented.

  3. @Sean, the weighting functions of the various channels does not really correspond to the weighting function of any of the instruments. The weighting function is also dependent on the reflectivity of the surface it will vary over terrain and needs to be estimated. Because of all of this plus some other stuff, the system turns out to be under-constrained. It is not plug and play by any matter of means.

  4. Sean says: March 11, 2011 at 11:06 am
    …..
    I am not certain, but sensors may be affected by solar wind (proton and electron flux), since they are not protected by the Earth’s magnetic field. In order to pick up the IR etc radiation from the ocean surface, sensors can’t be totally shielded, unless they have a parallel sensor just for the solar radiation to provide compensation data.
    We need Dr. Svalgaard’s comment on this one.

  5. ShrNfr, with such uncertainties, adding measurements or finer calculation should just narrow the error bars not cause the drop after 1999 between V3.3 and V5.4 . A drop implies a systematic change or mis splice before or after.

  6. The step changes are even more apparent in the TLS channel, where the temperature level drops abruptly following the large spikes caused by the El Chichon and Mt. Pinatubo eruptions, but then is flat or even slightly upward. I find the use of a linear trend line unreasonable when there’s almost no trend since 1994 (which is over half the total period).

  7. While it’s [sic] clear from figure 8 that some of the 1998 peak stuck around and could be crudely represented as step increase followed by a generally flat period, your idea of seeing a step in 1988 is pretty tenuous to say the least.

    If any part of the whole period can be approximated as a straight line it is the period from 1985 to ’92. The straight line would be linear rise not a flat line and a step.

    It seems that you are doing what so many climate scientists have rightly been slated for: seeing what you expect or want to see at the outset.

    In fact the pre- 1998 period could best be characterised by a three saw tooth profiles, not a step increase.

    The relatively flat period >1998 represents a change in the overall behaviour, not a repetition of what happened in the preceding 20 or 25 years.

  8. Dear Bob,

    I respect and appreciate your fine work and analysis, and I am not qualified to criticize it. However, two questions jump out:

    1. Are there other equally valid smoothed lines that fit the data other than your 3-step summary (especially if other, earlier, starting points were considered)?

    2. How sensitive is your model to possible future states? That is, what if by the end of the year 2011 data shows a drop to -0.2°C anomaly? Would that change your conclusions?

  9. @Sean, putting new instruments on the end usually means putting a new inversion scheme on the end from the primary investigators. You then multiply the temperatures from your inversion scheme times the weighting function and sum them to get the brightness temperature. Going back later and perhaps applying a consistant inversion scheme over the datasets unifys them. But you are free dipute my assumptions. I am really not that familiar with what they do with the processing these days. Once upon a time, I shared an office with 4,000 reels of 7 track from one of these things. I suspect the state of the art has advanced quite a bit in the time since. What we would have given for one of the flash devices with 32 GB back then.

  10. P.Solar: While it’s [sic] clear from…

    Why the sic? In this case, it’s is used as an abbreviation for it is, so the apostrophe is correct. In the main post’s quotation from RSS, it’s is intended as the possessive form of it, so the apostrophe is incorrect.

  11. Mike D. says: “Are there other equally valid smoothed lines that fit the data other than your 3-step summary (especially if other, earlier, starting points were considered)?”

    The thought that TLT anomalies north of 20N were rising in steps in response to significant ENSO events came from the RSS Hovmoller plots. Refer to…

    http://bobtisdale.blogspot.com/2009/06/rss-msu-tlt-time-latitude-plots.html

    With respect to earlier TLT data, the satellite-based TLT anomaly data starts in 1979, so there’s no way to consider earlier starting points

    You asked, “How sensitive is your model to possible future states? That is, what if by the end of the year 2011 data shows a drop to -0.2°C anomaly? Would that change your conclusions?”

    IF (big if), after the 2009/10 El Nino & 2010/11 La Nina ENSO episode, the TLT anomalies north of 20N were to drop and remain at a lower average for a number of years, I, like many others, would be investigating the reason for the change.

  12. Max Hugoson says: “What are the ‘standard deviations’ of this data over this period.”

    For the global RSS TLT anomalies version 3.2, the standard deviation is 0.23 deg C and for version 3.3, it’s 0.22 deg C.

  13. It’s good to see RSS and UAH have finally been reconciled.

    Dr Spencer must be thrilled that after years of being ‘low man’ in global warming he’s got company now.

  14. @Sean: From the RSS site “The brightness temperature for each channel corresponds to an average temperature of the atmosphere averaged over that channel’s weighting function.” This appears to confirm that they have done an inversion step. Since frequency drift in the radiometers would shift the weighting functions either up or down in altitude, going back and using a time varying weighting function to re-invert the temperature data would be logical. Something that we ignored way back when, but something that causes a drift in the retrieved temperature profile with time. As I stated, you are trying to pull 15 temperatures from 3 or 4 channels of data and an estimated reflectivity. 15 numbers from 5 numbers leaves 10 degrees of freedom that you have to pull out of something. The adiabatic lapse rate of the atmosphere helps you under the assumption of no vertical mixing. Again, I do not know what inversion scheme they use these days. If anyone knows, I would be interested to hear what it is.

  15. Actually, I doubt that because its dipping so low now and comparing with 2008, at this rate 2011 looking very like 2008 or cooler, ie: no warming again

  16. I’ve got the daily UAH temp at -0.243C.

    The daily temperatures from the AMSU site are not the official UAH temps. There is an adjustment made of anywhere from -0.1C to +0.3C to these readings and the average is +0.123C.

    Here is the daily Global, SH, NH and Tropics UAH temps from 2010 to Mar. 2011.

    An amazing reduction of -0.9C from the El Nino-induced peaks in the spring of 2010 to today’s La Nina-induced temps. I don’t think it is going to fall any more from these levels. They will stay around -0.1C to -0.2C until at least July, 2011.

  17. “The linear trends of the RSS version 3.4 and the most recent version of UAH TLT anomaly data (v5.4)…”

    3.4?

  18. It would sure be nice to get a little bit of regional warming here in Western Nevada. We could use a C or two as this winter drags on and on and on. Did I mention the snow on the surrounding hills

  19. Bob – you are dead wrong with your volcano adjustments. Neither Pinatubo nor El Chichon had any global cooling effect on tropospheric temperature. It did have both warming and cooling effects in the stratosphere but this never descended to ground level. The tropospheric cooling attributed to Pinatubo is nothing more than a normal La Nina cooling that follows an El Nino. Pinatubo just happened to erupt at the exact time that an El Nino peaked, Self et al. appropriated the deep La Nina valley that followed to Pinatubo cooling, and everybody else just blindly copied him. They go on and babble about the cooling effect of Pinatubo overriding the warming effect of an El Nino which is utter nonsense because the El Nino that did coincide with the eruption is there as big as life. As to El Chichon, it erupted when a La Nina had just bottomed out and an El Nino was starting. If a volcano can override an El Nino this was a golden opportunity for El Chichon to demonstrate that but it never happened. This volcano had no cooling effect whatsoever at ground level and even Self et al. admit that as an unexplained mystery. It is not a mystery if you understand that tyis volcanic cooling effect is a myth. El Chichon aerosols did their stuff in the stratosphere and that is where its influence remained. For this reason there is no justification whatsoever to use any kind of volcanic adjustment because there is no way you can show any ground level cooling from it. And this goes for all volcanoes of the twentieth century. As I have pointed out in my book there is just one step change in the entire satellite record and that is the one that raises global temperature by a third of a degree between 1998 and 2002. There was no temperature rise before or after it. It started with the super El Nino of 1998 which itself does not belong to ENSO and should not be included in any sort of averaging like you have done in figures 2 and 5. You need to redraw these figures and show two horizontal lines on both sides of the step rise initiated by the super El Nino and its aftermath. It is best accomplished by using the full monthly resolution temperature chart and applying the magic marker method to fully articulate the temperature trend. Check out my figure 15 in the book to see how it is done. Better yet, spring for my book to avoid future embarrassments like this.

  20. Bob Tisdale says…
    > The linear trends of the RSS version 3.4 and the most recent version
    > of UAH TLT anomaly data (v5.4) are basically the same: 1.47 versus
    > 1.44 deg C per Century. Refer to Figure 5. Note that I’ve switched to
    > KNMI climate Explorer as the source for both datasets, so that I could
    > limit the UAH latitudes to those used by RSS, 70S-82.5N.

    That explains why my personal spreadsheet version differs from yours. Like it or not, the far north is part of the planet. I don’t agree with Hansen’s statistical gymnastics, but the far north does contribute to global temps. Using a spreadsheet at home, I get 1.4673 degrees / century for RSS and 1.3886 for UAH, from day 1 through February 2001. So UAH is slightly pulled down, overall, by including more of the far north, i.e. the “Global” column from http://vortex.nsstc.uah.edu/public/msu/t2lt/tltglhmam_5.4

    Recent trends show the opposite. RSS has a negative slope linear trend using data from June 1997 to Feb 2011. The furthest back I can get negative slope for UAH is May 2009 to Feb 2011. Obviously there is a difference. Since UAH covers more of the globe, I consider the data more complete.

  21. Sorry Bob, I just don’t understand. It appears to me that the data is being adjusted to fit a pre-ordained result. Kind of like adding UHI corrections to cool rural stations. The results even look fake. But then I am just a dirt farmer. pg

  22. vukcevic says:
    March 11, 2011 at 12:19 pm
    I am not certain, but sensors may be affected by solar wind (proton and electron flux) [...] We need Dr. Svalgaard’s comment on this one.
    I don’t comment on things I don’t know anything about.

  23. Leif Svalgaard says: March 11, 2011 at 11:30 pm
    …..
    Not knowing, even more reason to find out; essentially drive for the scientific progress!

  24. Walter Dnes says: “I don’t agree with Hansen’s statistical gymnastics, but the far north does contribute to global temps. ”

    RSS reaches as far north as 82.5N. Isn’t it the coverage of the Antarctic where RSS and UAH differ the most, with RSS only reaching as far south as 70S?

  25. Arno Arrak says: “Bob – you are dead wrong with your volcano adjustments. Neither Pinatubo nor El Chichon had any global cooling effect on tropospheric temperature. It did have both warming and cooling effects in the stratosphere but this never descended to ground level. The tropospheric cooling attributed to Pinatubo is nothing more than a normal La Nina cooling that follows an El Nino. Pinatubo just happened to erupt at the exact time that an El Nino peaked, Self et al. appropriated the deep La Nina valley that followed to Pinatubo cooling, and everybody else just blindly copied him.”

    Arno: Are you at this again? There was no La Nina after the 1991/92 El Nino. The following is a graph of detrended Global RSS TLT anomalies versus detrended and scaled NINO3.4 SST anomalies. Show me the La Nina in 1992/93 that was capable of explaining the divergence between Global TLT anomalies and the scaled NINO3.4 SST anomalies from 1991 to 1995.

    Again, there was no La Nina at that time. In fact, there are papers that describe the period from 1991 through 1995 as one long El Nino–Trenberth and Hoar (1996) for example:

    http://www.agu.org/journals/ABS/1996/95GL03602.shtml

    Over the past six months (year, whatever it was), you have repeatedly claimed that volcanic aerosols have no impact on global surface temperatures, yet you have never backed your claims with data. Please provide an ENSO index that shows a La Nina that was capable of explaining the divergence between the NINO3.4 SST anomalies and Global TLT anomalies from 1991 to 1995.

  26. Bob, is there any instance over the recorded period of the reverse – a down-step in global temperature caused by (for instance) a La Nina? What are the chances of such a down-step if we move from an el Nino dominated to a La Nina dominated phase of the ENSO?

  27. The ’88 step is a figment of your imagination. This period could be simplified as a linear ramp but not a step.

    You chose not to respond to my earlier comment to that effect, so I guess you accept the point.

    The ’85 to ’98 ramp is what got everyone panicking. The post ’98 flat is what has got the warmists panicking.

    I don’t think you’ll get very far with trying to persuade anyone that there was a step change in climate in 1988.

  28. P. Solar says: “You chose not to respond to my earlier comment to that effect, so I guess you accept the point.”

    I didn’t see your earlier comment because I was searched only for my name on this thread. I’ll read it and reply later.

  29. phlogiston says: “Bob, is there any instance over the recorded period of the reverse – a down-step in global temperature caused by (for instance) a La Nina?”

    I have not run across an instance.

  30. Bob Tisdale says:
    March 11, 2011 at 3:19 pm

    IF (big if), after the 2009/10 El Nino & 2010/11 La Nina ENSO episode, the TLT anomalies north of 20N were to drop and remain at a lower average for a number of years, I, like many others, would be investigating the reason for the change.

    Hi Bob, I would think that if OHC continues to fall, this is what we’ll see.

    phlogiston says: “Bob, is there any instance over the recorded period of the reverse – a down-step in global temperature caused by (for instance) a La Nina?”

    I have not run across an instance.

    If my cumulative solar energy model is near the mark, I thnk you’ll see a drop to a lower baseline starting this year. By 2015 it will have settled around 0.2C lower than now I would expect from my model Unless the Sun does something spectacular. Always possible.

  31. Try removing ENSO and the longterm trend from the temperature record and see what’s left, looks very much like an 11 year cycle. An oscillation over a linear upward trend will resemble a series of plateaus and step changes upward. And of course those step changes upward will be most eye-catching at the point of the first significant el nino in the upward phase of the solar cycle.

  32. As both AQUA and MSU estimate the same “real” temp and errors in AQUA and MSU should be distributed around the same mean “real” temp. If changing the weighting between AQUA and MSU between versions changes smouthed temp as suggested, then blending is unsatisfactory. If MSU has a spurious trend, we need to adjust it out over the whole MSU measurement period before blending. If that had been done, you would expect slope over the 1979 to 1999. Of course the problem can be in AQUA. Either way splicing sensors showing diffrent trends during the overlap period is unsatisfactory.

  33. P. Solar and onion2: Regarding P. Solar’s disagreement about the upward steps in the Volcano-adjusted TLT anomalies north of 20N, and onion2’s solar suggestion, refer first to the following comparison of the that dataset to NINO3.4 SST anomalies (Kaplan, scaled (0.25), lagged 9 months, and a shift of -0.2 deg C). The scaling factor for the GISS Aerosol Optical Thickness data is 2.9. The adjusted TLT anomalies mimic NINO3.4 SST anomalies from the start of the dataset until the peak of the 1986/87/88 El Niño. Then they fail to respond to the La Niña that follows. That’s step one:

    The next graph is of the same two datasets with the NINO3.4 SST anomalies shifted +0.07. The adjusted TLT anomalies mimic NINO3.4 SST anomalies from the start 1995 until the peak of the 1977/98 El Niño. And then they don’t respond to the La Niña. That’s step two:

  34. Tallbloke says: “Hi Bob, I would think that if OHC continues to fall, this is what we’ll see.”

    Hi Tallbloke. The drop in OHC is only in a few basins: North Atlantic, South Pacific, Arctic… I’ll be updating the NODC OHC graphs per basin through Dec 2010 hopefully on Monday.

    You wrote, “If my cumulative solar energy model is near the mark, I thnk you’ll see a drop to a lower baseline starting this year. By 2015 it will have settled around 0.2C lower than now I would expect from my model Unless the Sun does something spectacular. Always possible.”

    I’ll hope for the success of your model. I would like the opportunity to explain a 0.2 deg C downward shift.

    Regards

Comments are closed.