New RSS and Balloons

A while ago I compared satellite temperature data to that from balloon-borne instruments. I used the RSS satellite data (the deniers’ favorite), but since then they have revised their data. Several readers have requested an updated comparison, using the new RSS release.


I can’t use the TLT (lower-troposphere temperature) for comparison because RSS hasn’t yet revised that product. But they have updated the TTT (total-troposphere temperature) data, so let’s see how that compares to radiosonde (balloon-borne) data, before and after the new improved RSS version.

Here are all three: radiosonde data from RATPAC (Radiosonde Atmospheric Temperature Product for Assessing Climate), the prior version of RSS (v3.3) in blue, and the improved version (4) in red:

rat_rss

The new version warms faster than the old, and by doing so its trend matches that of the balloon data better. This is more easily seen in the differences between RSS data and RATPAC, the left graph for v3.3, the right for v4:

diff

Indeed the updated data match the balloon data better, and their difference no longer shows the large downward trend it showed before. But there’s still a significant difference, with RSSv4 warming a bit faster than balloon data early, a bit more slowly later.


If you like what you see, join fellow readers by donating at Peaseblossom’s Closet.

8 responses to “New RSS and Balloons

  1. Tamino,

    Its not that easy, the comparison you made (like month before) is not correct at all, becuase, RSS TTT is a Product which is calculated by the TMT-Level, which itself has spurios cooling of the lower stratosphere, so if you wish the make a more better comparison do use the weighting which is given in RSS, which means, you use first the TMT weighting for your Layers in RATPAC, then caculated direct from RSS formula: TTT= 1.1*TMT-0.1*TLS.

    • Don’t radiosondes measure temperatures at multiple altitudes? Those measurements could be weighted to match the TTT profile (or TMT). You could also subsample the RSS data to match the location of the radiosonde measurements. Not sure how easy any of this is.

  2. If I understand Christian correctly, Gavin Schmidt’s reasonable assumptions version of the Christy graph uses the same approach.

  3. JHC,

    Ne, first of all, Gavin Schmidt does not use Balloons, he is only using UAH/RSS/STAR and the Values given by Po-Chedley et al. The other point, he weigths the atomsperic layers in CIMP5 to make them compatible to TMT by UAH/RSS.

    If understood Tamino correctly, he makes the mean betweem 850mb to 300mb for RATPAC, while TTT from RSS is calculated by values of TMT and lower statosphere (see first comment here). This could make trouble, it would be better, if first use RSS TMT-weighting on RATPAC and then calculate to TTT.

  4. So did my own analysis:

    RATPAC-A TMT vs. RSS TMT v3.3 and v4.0

    RSS TMT v3.3 minus RATPAC-A TMT

    RSS TMT v4.0 minus RATPAC-A TMT

    Used weighting for RATPAC-A
    surf 0.097575157
    850 0.110341195
    700 0.22173345
    500 0.1193423
    400 0.147669078
    300 0.068631995
    250 0.068197936
    200 0.058799792
    150 0.054296926
    100 0.023684779
    70 0.013698017
    50 0.016816335

  5. PS: I also reproduce (with small errors because i dont know how Gavin weigth the Layers) Gavins plot, just only on RCP-mean not with Ensemble.

  6. Nice work Christian,
    However, you shouldn’t doubt the standard index of the free troposphere, the 850-300 hPa layer, it is very similar with the satellite TTT just as described by its inventors Fu et al (2004, 2005).
    Here is a comparison of 850-300, TTT and TLT ( a la UAH v6) weightings, notice the similarity:
    https://drive.google.com/file/d/0B_dL1shkWewaZXlwMlB6bjVYelU/view?usp=docslist_api

    Cce also raised the issue of subsampling of Ratpac station areas. This can not be done directly with Ratpac A without “destroying” its homogenisation and nice global weighting (ie using the raw ratpac B data). However, I have done it the other way around, subsampled UAH data from Ratpac station locations and made a global index with the Ratpac A method:
    https://drive.google.com/file/d/0B_dL1shkWewaVUdobkJYcjM5VmM/view?usp=docslist_api

    I have mainly worked with UAH v6 ( actually, I can’t read the RSS netcdf, there is some error), but I think that this chart serves to demonstrate that the Ratpac A method gives a fair global representation, and that subsampling and detailed weighting do not change the distinct trend break between satellites and radiosondes after year 2000.

    As Tamino demonstrates above, TTT v4 is an improvement compared to the old version. Half the trend divergence vs Ratpac is gone. The other half is IMO caused by errors in AMSU 5 onboard NOAA-15. In version 4 RSS doesn’t use NOAA-15 data after 2011, and I think it is quite obvious that RSS TTT v4 follows Ratpac much better than v3.3 after 2011..

  7. Olof,

    Very Thanks

    “I have mainly worked with UAH v6 ( actually, I can’t read the RSS netcdf, there is some error), but I think that this chart serves to demonstrate that the Ratpac A method gives a fair global representation, and that subsampling and detailed weighting do not change the distinct trend break between satellites and radiosondes after year 2000.”

    Yes, but the Update of RSS does it, just look auf my figure RSS TMT v3.3 minus RATPAC-A TMT and your subsampled UAH. They are very similar, after the correction to RSS 4.0 the distinct “cooling” trend breaks.

    To all other Points:
    I think, we must be carefull to what we compare with what. Since the RSS and UAH calculate the Temperature, they are unable to gives us results that easy to compare, because all is constructed by the TMT, which is always influenced by some of strosphere cooling, since TTT or TLT is calculated by TMT, its would gives us always this: dT(TLT) > dT (TTT) > dT (TMT), based on this, UAH/RSS will never confirm that there is a troposhere Hotspot, because on their calulation, the trosphere in lower altitudes will always warm/cool more then die upper altitudes.

    That means, if we want to compare balloons to UAH/RSS we should alway do the same way, first construct a TMT-Layer, then construct a TLS-Layer and then use the Formula: TTT= 1.1*TMT-0.1*TLS.