CSALT Volcanic Aerosols

The volcanic aerosol factor of the CSALT model is an example of a perfectly interlocking piece in the larger global surface temperature puzzle.  I thought I would present a more detailed description in response to the absolutely hapless recent volcano posts at the WUWT blog (here and here).  The usual deniers in the WUWTang Clan can't seem to get much right in their quest to intelligently spell out ABCD (Anything But Carbon Dioxide).

Fig 1 :  CSALT model using the GISS Stratospheric Aerosol forcing model.

The addition of the volcanic aerosol factor is no different than the other components of the CSALT model. Two flavors of volcanic aerosol forcings are provided. The standard forcing table is the GISS stratospheric aerosol optical thickness model maintained by Sato [1] and I use this table as is (see Figure 1).   The more experimental model that I generated is a sparse table that features only the volcanoes of Volcanic Explosivity Index (VEI) of 5 or higher.

The VEI scale is logarithmic so that a VEI of 6 contains 10 times as much ejected particulates by volume than a VEI of 5 (which recursively is 10 times as much as VEI=4, and so on). This means that by modeling VEI of 5 or 6 we should capture most of the particulates generated as discrete events.

Figure 2 shows the residual of a fit given that the volcanic aerosol forcing is excluded from the model. The majority of the sharp negative excursions map to the major eruptions of VEI scale 5 or 6 over the past 130+ years.

Fig 3 :  Identification of eruptions of VEI scale 5 or 6 with major cold spikes identified by the CSALT model residual.

To create a sparse table, we look up the volcanic eruptions from the Smithsonian Institute Global Volcanism Program database and attach an intensity scaled to compensate the negative excursion of the residual.

eruption year month month# intensity VEI
krakatoa 1883 8 42 25 6
santamaria 1902 10 275 19 6
ksudach 1907 4 328 6 5
novarupta 1912 6 390 14 6
cerroazul 1916 4 432 7 5
agung 1963 2 999 21 5
elchichon 1982 4 1228 20 5
pinatubo 1991 4 1336 16 6
cerrohudson 1991 8 1340 8 5

Some volcanic eruptions of VEI=5  such as Mt. St. Helens in 1980 and Bezymianny in 1955 are not included in the table because they do not appear to have generated  cooling of any magnitude or duration.

A scaled response function is attached to each event so that a sparse forcing profile is generated as in Figure 4.

Fig 4 : Sparse forcing profile for volcanic eruptions of VEI = 5 and VEI=6.

The dates of the eruptions are accurately matched to the Smithsonian records. The table below shows the historic Month start since 1980 and the corresponding Model start.  These are coincident apart from a few cases where Delay = Model - Month is set to a few months after the eruption occurs. The last column shows the Lag defined as the time the Minimum residual cooling dip is observed after the Model start month.  In general, the observed cooling spike minimum is consistently seen about 1 to 1.5 years after the initial eruption is recorded. The CSALT model has resolution in terms of months so that these observed lags have significance.

VEI Month Model Minimum Delay Lag
Krakatau 6 42 42 57 0 15
Santa María 6 275 275 286 0 11
Ksudach 5 328 328 342 0 14
Novarupta 6 390 390 398 0 8
Azul, Cerro 5 432 434 450 2 16
Agung 5 999 1003 1014 4 11
Chichón, El 5 1228 1228 1239 0 11
Pinatubo 6 1336 1339 1350 3 11
Hudson, Cerro 5 1340 1341 1357 1 16

After this sparse model is included in the CSALT model, the residual error is much reduced as shown in Figure 5.  The sparse model works just as well in improving the model fit as the more sophisticated GISS stratospheric forcing model, indicating how important the major volcanic events are in determining climate response.

Fig 5 :  Fit using sparse volcanic forcing model. The major negative excursions are removed and the correlation is much improved.

The likelihood that the cooling excursions shown in the original residual are volcanic eruption related is extremely high. This kind of agreement in timing and in scale does not happen by accident. By the same token, the planetary cooling of 0.1 C for the largest events matches the historical observations of the transient impact that eruptions have on the climate.  The CSALT model is able to extract the contribution of volcanoes to the natural variability with skill. There are still cooling spikes that occur which do not appear to be associated with major volcanic eruptions (or other obvious natural variability factors captured in CSALT) such as the dip in 1976-1977 or the Cold Sunday of January 1982, but these are the exception.

Compare this work to the posts at WUWT.  In the first post, the author "Wondering" Willis Eschenbach makes the claim that :

"In particular, despite widespread skepticism, I have persisted in saying that volcanoes basically don’t do jack in the way of affecting the global temperature."

The reason the wondering one had so much difficulty was likely because he was dealing with a temperature profile that had not been sufficiently defluctuated of other natural variability factors, such as SOI and TSI.   Same issue with the second WUWT post, which did have a good start by concentrating on the high VEI eruptions, but likewise was not able to find evidence of volcanic aerosols in the global temperature anomaly.  So a similar conclusion was found:

"The effect of aerosol emissions on global temperatures from volcanic eruptions appears very small and may not be discernable from natural variation."

This is very representative of the poor quality of research that climate skeptics and deniers consistently demonstrate.  It often appears that they have no interest in pursuing the science but instead make only an effort to create scenarios that supports their underlying agenda.  The only good that comes out of WUWT posts is that their "motivated reasoning" provides me with my own motivation to demonstrate just how wrong they are.  Yes, climate science can be complex, but there are always simplifications that can reveal a more concise path to follow.  Whether the skeptics and deniers even bother or care to learn is another story.


[1] J. E. Hansen and M. Sato, “Paleoclimate implications for human-made climate change,” Climate Change, pp. 21–47, 2012.
[2] T. J. Crowley and M. B. Unterman, “Technical details concerning development of a 1200-yr proxy index for global volcanism,” Earth System Science Data Discussions, vol. 5, no. 1, pp. 1–28, 2012.












4 thoughts on “CSALT Volcanic Aerosols

  1. In Figure 2, the two strongest unattributed cooling dips in the residual occur in 1925 and 1976.
    From a comment I received at Climate Etc today:

    maksimovich | March 13, 2014 at 9:57 pm |
    The modelling of volcanic excursions is sensitive to location and timing,and also sensitive to parameters ie which data set you use.there is a new micro physics based data set with better chemical extinction coefficients (or at least better then previous )

    This may be an "own goal" by commenter maks, as from that paper's Table 1, eruptions identified in ice core data from 1925 and 1976 coincide perfectly with the 2 cooling dips

    The ice core evidence in 1925 is from some unknown eruption while the 1976 evidence may be from Fuego, a VEI=4 eruption.

    Like I asserted, climate science is like a jigsaw puzzle where the pieces fit snugly together. That's why the skeptics keep scoring own goals, as the discovery of new evidence almost always supports the consensus models of climate science.

    • This is an update to the residual figure using the ice core data from Gao identified by Arfeuille et al paper
      linked to by Maks

      Gao update

      What is very interesting in the CSALT model is that the vast majority of the cold spikes in the residual match datewise and intensity to the volcanic eruptions identified in the Arfeuille paper.
      Of the top 10 CSALT cold excursions, 9 map to the top 10 significant aerosol mass deposits in ice core data in the Arfeuille paper. The one that CSALT identifies not in the Arfeuille table is CerroAzul in 1916 (during WWI), and the one that CSALT does not catch is an unidentified eruption in 1943 (during WWII). War year temperature data has greater uncertainty in general.

  2. Pingback: CSALT re-analysis | context/Earth

Leave a Reply