Introduction


Following the injection of 150 Tg of black carbon over the United States and Russia on May 15th, there is warming simulated over Eurasia in the CESM-WACCM4 and GISS ModelE climate models (see Fig. 1). As this basic response was simulated in both a complex coupled global climate model and a less complex, lower resolution model, I sought out to understand what is driving this response.




Fig. 1 – Temperature anomaly during the first winter after the injection of 150 Tg of black carbon over the US and Russia for CESM-WACCM4 and GISS ModelE. Warming is present in both simulations.

As this anomaly resembled the response following volcanic eruptions, it seemed probable that shifts in the Arctic Oscillation (AO) circulation could be prompting these anomalies, which act in the opposite direction of the significant global cooling response. It became apparent that the injection of soot, which heats the stratosphere and cools the troposphere, shifts the AO into an entirely new regime, characterized by strongly positive AO winters and strong negatively AO summers. Fig. 2 (a) shows the effect of the lack of solar radiation over the pole on the thermodynamic circulation surrounding the northern pole during the winter. Geopotential heights are anomalously low compared to high geopotential heights in the midlatitudes, because the soot is being heated up everywhere but the pole. This increases the pole-to-equator temperature gradient in the stratosphere and drives a strengthening in the zonal mean winds. Fig 2. (b) shows the regime change in the AO after the injection of black carbon. This effect lasts for nearly 12 years, until most of the aerosols have fallen out of the stratosphere and the thermal wind response is no longer possible.


Fig. 2 – (a) 250 mb geopotential height anomalies during the first 6 winters. (b) Arctic Oscillation index before and after the black carbon injection.






Fig. 3 shows the zonal mean winds at 10 hPa and 60N, a common metric for tracking the strength of the Stratospheric Polar Vortex (hereafter SPV). It has all six nuclear winter simulations conducted to date, and shows the impact of the black carbon on the winds in the stratosphere during the winter. They are highly anomalous and well outside of the +/-2 standard deviation envelope on either side of the climatological mean. In fact, they are nearly +5 standard deviations from the climatological mean value for the most extreme 150 Tg case. Even in the smaller 5 Tg case, the zonal mean winds at this latitude and altitude are approaching 80 m/s for significant periods of time.


Fig. 3 - Daily zonal mean winds at 10 hPa and 60N for all nuclear winter cases.




There is clearly a shift in the circulation that is statistically significant for all cases in the stratosphere. There are questions about the mechanism, and how it propagates down, which will be addressed later and with additional experiments.


Now, the question is, does the simulated temperature signal in our cases match the expected temperature signal from this type of circulation change? I perform a simple linear regression of the Arctic Oscillation index onto surface temperature anomalies, using a control climatology to train the data. Fig. 4 shows the simulated temperature anomaly in the 150 Tg case compared to the regressed (or expected) temperature anomalies. Warm surface temperatures are to be expected across most of Eurasia with the anomalously positive AO index during the nuclear war simulation. However, there is warming closer to the pole over the Arctic Ocean that is an aberration from the typical highly positive AO pattern. This may be explained by an increase in longwave radiation downward from the relatively warm soot in the upper troposphere and lower stratosphere. This region is sensitive to longwave radiation during the winter because there is no shortwave radiation.


Fig. 4 - Comparison of simulated temperature anomaly and regressed temperature anomaly as a function of AO index for the 150 Tg case.



I take a look at this relationship through all of the nuclear war simulations. Fig. 5 shows the first winter for all of the nuclear winter simulations. For the first winter, warming is restricted to the landmasses only in the 50 Tg case, but cooling obscures any circulation-related warming in the areas that typically experience above average temperature anomalies when the AO is as positive as it is during the 50 Tg case. Fig. 6 shows the second winter, where there is still strong warming in the 150 Tg, 37 Tg, 16 Tg, and 5 Tg cases. The spatial pattern of warming is clearly not only a function of the AO index. A very strong El Nino is simultaneously occurring, so I will also examine the signal of El Nino on Arctic surface temperatures.


Fig. 5 – Comparison of simulated temperature anomaly and simple linear regression of temperature anomaly based on the AO index for all nuclear war simulations during the first winter.




Fig. 6 Comparison of simulated temperature anomaly and simple linear regression of temperature anomaly based on the AO index for all nuclear war simulations during the second winter.




Fig. 7 – Comparison of simulated temperature anomaly and simple linear regression of surface temperatures with the SOI for all nuclear war cases during the first winter.



Fig. 8 - Comparison of simulated temperature anomalies and a simple linear regression of SOI on surface temperature for the second winter after injection.



The signal from the strong El Nino during the winter is clearly for a cooling of the polar regions along the northern coast of Eurasia (see Fig. 7 and Fig. 8). Based on this analysis, the El Nino signal in CESM-WACCM4 is working to mitigate any warming at the pole. There is no signal for the 5 Tg case. To confirm, I perform a multiple linear regression using AO and SOI as variables, to account for any interaction between the two that is not reflected in two separate simple linear regressions.



Fig. 9 – Comparison of simulated temperature anomalies and a multivariable linear regression of SOI and the AO on surface temperature for the first winter after the injection.



Fig. 10Comparison of simulated temperature anomalies and a multivariable linear regression of SOI and the AO on surface temperature for the second winter after the injection.






Examining the Physical Mechanism

The change in wind circulation is controlled through (a) changes to the stratospheric temperature gradient or (b) changes to the tropospheric temperature gradient (then changes in upward propagating waves influence the NH winter circulation. We begin by quantifying the changes to the stratospheric temperature gradient.


Polvani et al. (2019) investigated changes to the stratospheric temperature gradient by examining the temperature at 50 hPa over the pole (60N-90N) and the equatorial regions (30S-30N), where T = POLAR – TROPICAL. To measure the change in strength of the stratospheric polar vortex, we use zonally averaged zonal winds at 10 hPa and 60N (U1060). Each winter after the soot injection, dT grows more and more negative while U1060 accelerates substantially. There is a clear relationship, which can be determined physically through the thermal wind relationship and statistically through a number of regression techniques. A comparison of the physical relationship and the statistical relationship will be highly instructive, pointing to processes affecting the zonal winds in addition to a basic thermal wind relationship.


Fig. 11 shows a scatterplot of T at 50 hPa and U1060, demonstrating the relationship between pole-to-equator temperature gradient and stratospheric polar vortex strength for the control run in CESM-WACCM4. When the tropics are far warmer than the poles at 50 hPa, there is an acceleration of the zonal winds at this level. I apply this relationship to the T for the nuclear winter case and try to reconstruct the actual zonal mean winds in Fig. 12.


Fig. 11 - Relationship between T and U1060 for the control run in CESM-WACCM4, and fitted 3rd degree polynomial.






We compare these two 3rd degree polynomial fits on the same graph to demonstrate the effect of the 150 Tg data in bounding the fit to a far more realistic value. A regression of simply the control run is not sufficient to replicate the actual zonal wind anomalies for the 150 Tg case due to a lack of data at the extreme ends of the temperature gradient (see Fig. 12). Fig. 13 shows a regression using the 150 Tg case only


Fig. 12 - 3rd degree polynomial fit of the relationship between T and U1060 for the control run in CESM-WACCM4.



Fig. 13 – Zonal mean winds at 10 hPa and along 60N during the 150 Tg case and regression using T during the 150 Tg case. The second panel is a regression trained using the 150 Tg case data, but shows the 46.8 Tg case. The regression does a good job.






Using the 150 Tg run to construct this relationship produces satisfactory results, even if it is not independent of the training data. Polvani et al. (2019) found no meaningful relationship between the strength of the polar vortex (using U1060) and surface temperature anomalies in Eurasia. We will conduct the same analysis using a simple linear regression of U1060 onto surface temperatures using CESM-WACCM4. We will then compare the surface temperatures in the nuclear winter simulations with what is expected from the vortex strength.

First, Polvani et al. (2019) found an r2 value of 0.89 between T and U1060, and we find an r2 value of 0.85. Polvani et al. (2019) averaged surface temperatures over Eurasia, but the latitude/longitude coordinates he used for this calculation is unclear. This becomes important because the relationship between the SPV strength and Eurasian surface temperatures has a sharp cutoff along the northern edge of Eurasia closer to the Arctic circle (see Fig. 14). A regression of surface temperatures using the 150 Tg case U1060 for the first winter shows this clearly, and averaging over a large area would remove this. It is interesting to note that a linear regression of the surface AO and surface temperatures does not reveal warming along the northern pole, but a linear regression of the strength of the SPV and surface temperatures does show this relationship. Because of this, we will perform a multivariable regression for the surface AO and the SPV strength using U1060, to see if this gets us closer to the actual signal. But what is the link between U1060 and surface temperature if not the AO? What about a strong SPV causes warming of the northern pole if it not acting through the AO?


Fig. 14 – Regression of surface temperatures using relationship between control U1060 and surface temperatures.







All of these factors for the first winter on one plot:




In conclusion, I’ve shown that in CESM-WACCM4, there is a significant link between T and U1060 which is both statistical and physical. There is a link between U1060 and surface temperatures which is slightly different than the link between the surface AO and surface temperatures, indicating that stratospheric-tropospheric coupling varies and is more complicated. However, the relationship between U1060 and surface temperatures is very similar to what occurs following the black carbon injection.