Andy Lacis has graciously continued the dialog that was started in our weblog posts
Today, I am posting a new constructive contribution. I will respond (and invite Roy Spencer to respond) for next week.
Guest Post by Andy Lacis – Atmospheric CO2 Thermostat: Continued Dialog (Part I)
The interaction here has been both useful and informative, as opposed to what so frequently happens in the more typical climate blog interactions where only predictable and otherwise immutable opinions get tossed about with no real exchange of information or ideas.
It seems to me that the root cause of the varied differences in interpretation regarding our Science paper conclusions may well originate from the perceived understanding (or misunderstanding) of what exactly the GISS ModelE climate model is capable of simulating, and what specifically is or is not being assumed in the ModelE climate experiment simulations that we describe.
This point is particularly relevant to Roy Spencer’s remark that “our assumptions determine our conclusions”. Or more specifically, that “after assuming clouds and water vapor are no more than feedbacks upon temperature”, they then “prove their paradigm that CO2 drives climate by forcing the model with a CO2 change.” And, further along these lines that, “if they had forced the model with a water vapor change, it would have done the same thing.”
Basically, Roy’s comments would have been applicable had we used a 1D radiative/convective model (as in Hansen et al., 1981, Science, 213, 957–966) for our Science paper calculations. As you well know, in 1D RCMs, there is no real capability for including model ‘physics’. Instead, all of the cloud and water vapor feedback effects are either implicitly or explicitly prescribed, and thus by definition, ‘assumed’. However, with respect to our ModelE climate experiment simulations, Roy’s comments are completely off target because there is really nothing that is being assumed about cloud and water vapor feedbacks, other than clouds and water vapor behave according to established physics. Climate feedbacks are simply the end result of model physics.
Roy’s last point about forcing the model with water vapor brings up an interesting point. If water vapor is a feedback, can it also be a forcing? The answer is “Yes, absolutely!” any externally imposed water vapor change beyond the ‘feedback equilibrium’ distribution of water vapor will constitute a radiative forcing. To illustrate this point, we performed two GCM runs – one with instantaneously doubled water vapor, the other with instantaneously zeroed water vapor (actually reduced by a factor of 1000 to avoid a divide check in a diagnostic routine that expects finite column water vapor). Initial model temperatures (and model physics) are the same as in the control run. The instantaneous net TOA forcing for doubled water vapor is 12 W/m2 warming, and –60 W/m2 cooling for zeroed water vapor (see Schmidt et al., 2010, 115, D20106).
As expected, for the doubled water vapor experiment, there is enhanced rapid rainout, while for the zeroed water vapor experiment there is rapid evaporation into a very dry atmosphere. This is because the condensation and evaporation are significantly faster acting processes than the changes in atmospheric and ocean temperature in response to the applied radiative forcings (which diminish rapidly as atmospheric water vapor returns to its control-run equilibrium distribution). Within a year, atmospheric water vapor distribution is back to being virtually indistinguishable from the control-run climate with no significant long term impact.
As you can see, Roy’s point about water vapor forcing is actually an excellent example to illustrate the feedback role of water vapor, even more directly and emphatically than by zeroing out the non-condensing GHGs in our Science paper. From all this, it is abundantly clear that it is the non-condensing GHGs that control the terrestrial greenhouse effect (and thus the global equilibrium temperature of Earth). Water vapor and clouds, although accounting for 75% of the total greenhouse effect, participate only to the extent of feedback amplification. This then leads to the conclusion that atmospheric CO2 (accounting for 80% of the non-condensing GHG forcing) acts as a thermostat in controlling the temperature of Earth by regulating the strength of the terrestrial greenhouse effect. Since the non-condensing GHGs (CO2 in particular) are all being accurately measured and monitored, and since humans are directly linked to these GHG increases, it then follows that the global warming aspect (increase in terrestrial greenhouse effect strength) of global climate change is directly the result of human industrial activity.
NOAA lists the current level of atmospheric CO2 at Mona Loa at 387.18 ppmv. This is indicative of the precision with which the non-condensing GHG forcing is known, both for recent trends and going back into the geological ice core record. Some uncertainty (possibly substantial) is associated with the magnitude of the water vapor and cloud feedback amplification. Similarly, there is uncertainty with the climate response time of the ocean heat capacity. The GISS ModelE produces an overall feedback amplification (climate sensitivity) of about 3 °C for doubled CO2 (or 0.75 °C/Wm–2). This climate feedback sensitivity is corroborated by the 400,000 year Antarctic ice core record (Hansen et al., 2008, Open Atmos. Sci. J., 2, 217–231). Hansen et al. also show that the global climate response time takes about 5 years to achieve 40% of the equilibrium warming, 100 years to reach 60%, and about 1500 years to approach the 100% level of the eventual global equilibrium warming for a given radiative forcing.
So much for global warming by the anthropogenic GHGs. We are in basic agreement with you that global warming fueled by increasing GHGs is not the only thing that is causing global climate change. Our best estimate of climate forcings for the period 1750–2000 (Hansen et al., 2005, J. Geophys. Res., 110, D18104), is that GHG increases account for 2.9 W/m2 (of which CO2 contributes 1.5 W/m2). Aerosols are the really big uncertainty in global climate forcing with black carbon type aerosols causing 0.8 W/m2 warming, non-absorbing aerosols –1.1 W/m2 cooling, and indirect aerosol effect producing about –1.0 W/m2 cooling. Smaller radiative forcings are attributed to land use change (–0.15 W/m2) and to solar irradiance (0.30 W/m2).
Aerosol and land use forcings are particularly troublesome because the observational constraints are so poor for these forcings. This is largely because current satellite measurements are only capable of making spectral intensity-only measurements, which makes it is impossible to attribute intensity contributions as coming unambiguously from the land surface, aerosols, or from undetected cloud contamination. Hopefully, polarimetric measurements on the upcoming NASA Glory mission will greatly improve on the aerosol forcing uncertainty. Because of their short atmospheric life time, aerosols (especially black carbon) are the more attractive targets for mitigating global warming than GHGs with their long atmospheric life times.
Natural (unforced) climate variability (e.g., El Nino, La Nina, decadal temperature fluctuations in the Pacific and Atlantic oceans), is another factor that is an important part of the ongoing global climate change. But all these are fluctuations about the global equilibrium temperature and do not by themselves produce a long term temperature trend. Still, they do confuse trend analysis of the existing climate record (which is too short to establish a statistical certainty).
Atmospheric CO2 Thermostat: Continued Dialog (Part II)
The GISS ModelE is specifically designed to be a ‘physical’ model, so that Roy Spencer’s water vapor and cloud feedback ‘assumptions’ never actually need to be made. There is of course no guarantee that the model physics actually operate without flaw or bias. In particular, given the nature of atmospheric turbulence, a ‘first principles’ formulation for water vapor and cloud processes is not possible. Because of his, there are a number of adjustable coefficients that have to be ‘tuned’ to ensure that the formulation of evaporation, transport, and condensation of water vapor into clouds, and its dependence on wind speed, temperature, relative humidity, etc., will be in close agreement with current climate distributions. However, once these coefficients have been set, they become part of the model physics, and are not subject to further change. As a result, the model clouds and water vapor are free to change in response to local meteorological conditions. Cloud and water vapor feedbacks are the result of model physics and are thus in no way “assumed”, or arbitrarily prescribed. A basic description of ModelE physics and of ModelE performance is given by Schmidt et al. (2006, J. Climate, 19, 153–192).
Of the different physical processes in ModelE, radiation is the closest to being ‘first principles’ based. This is the part of model physics that I am most familiar with, having worked for many years to design and develop the GISS GCM radiation modeling capability. The only significant assumption being made for radiation modeling is that the GCM cloud and absorber distributions are defined in terms of plane parallel geometry. We use the correlated k-distribution approach (Lacis and Oinas, 1991, J. Geophys. Res., 96, 9027–9063) to transform the HITRAN database of atmospheric line information into absorption coefficient tables, and we use the vector doubling adding method as the basis and standard of reference for GCM multiple scattering treatment.
Direct comparison of the upwelling and downwelling LW radiative fluxes, cooling rates, and flux differences between line-by-line calculations and the GISS ModelE radiation model results for the Standard Mid-latitude atmosphere is shown in Figure 1 below.
As you can see, the GCM radiation model can reproduce the line-by-line calculated fluxes to better than 1 W/m2. This level of accuracy is representative for the full range of temperature and water vapor profiles that are encountered in the atmosphere for current climate as well as for excursions to substantially colder and warmer climate conditions. The radiation model also accounts in full for the overlapping absorption by the different atmospheric gases, including absorption by aerosols and clouds. In my early days of climate modeling when computer speed and memory were strong constraints, the objective was to develop simple parameterizations for weather GCM applications (e.g., Lacis and Hansen, 1974, J. Atmos. Sci., 31, 118–133). Soon after, when the science focus shifted to real climate modeling, it became clear that an explicit radiative model responds accurately to any and all changes that might take place in ground surface properties, atmospheric structure, and solar illumination. Thus the logarithmic behavior of radiative forcings for CO2 and for other GHGs is behavior that has been derived from the GCM radiation model’s radiative response (e.g., the radiative forcing formulas in Hansen et al., 1988, J. Geophys. Res., 93, 9341–9364) rather than being some kind of a constraint that is placed on the GCM radiation model.
Climate is primarily a boundary value problem in physics, and the key boundary value is at the top of the atmosphere being defined entirely by the incoming (absorbed) solar radiation and the outgoing LW thermal radiation. The global mean upwelling LW flux at the ground surface is about 390 W/m2 (for 288 K), and the outgoing LW flux at TOA is about 240 W/m2 (or 255 K equivalent). The LW flux difference that exists between the ground and TOA of 150 W/m2 (or 33 K equivalent) is a measure of the terrestrial greenhouse effect strength. We should note that the transformation of the LW flux that is emitted upward by the ground, to the LW flux that eventually leaves the top of the atmosphere, is entirely by radiative transfer means. Atmospheric dynamical processes participate in this LW flux transformation only to the extent of helping define the atmospheric temperature profile, and in establishing the local atmospheric profiles of water vapor and cloud distributions that are used in the radiative calculations.
Armed with a capable radiative transfer model, it is then straightforward to take apart and reconstruct the entire atmospheric structure, constituent by constituent, or in any particular grouping, to attribute what fraction of the total terrestrial greenhouse effect each atmospheric constituent is responsible for. That is where the 50% water vapor, 25% cloud, and 20% CO2 attribution in the Science paper (for the atmosphere as a whole) came from. “Follow the money!” is the recommended strategy to get to the bottom of murky political innuendos. A similar approach, using “Follow the energy!” as the guideline, is an effective means for fathoming the working behavior of the terrestrial climate system. By using globally averaged radiative fluxes in the analysis, the complexities of advective energy transports get averaged out. The climate energy problem is thereby reduced to a more straightforward global energy balance problem between incoming (absorbed) SW solar energy and outgoing LW thermal energy, which is fully amenable to radiative transfer modeling analysis. The working pieces in the analysis are the absorbed solar energy input, the atmospheric temperature profile, surface temperature, including the atmospheric distribution of water vapor, clouds, aerosols, and the minor greenhouse gases, all of which can be taken apart and re-assembled at will in order to quantitatively characterize and attribute the relative importance of each radiative contributor.
Validation of the GCM climate modeling performance is in terms of how well the model generated temperature, water vapor, and cloud fields resemble observational data of these quantities, including their spatial and seasonal variability. It would appear that ModelE does a generally credible job in reproducing most aspects of the terrestrial climate system. However, direct observational validation of the GCM radiation model performance to a useful precision is not really feasible since the atmospheric temperature profile and absorber distributions cannot all be measured simultaneously with available instrumentation to the required precision that would lead to a meaningful closure experiment. As a result, validation of the GCM radiation model performance must necessarily rely on the established theoretical foundation of radiative transfer, and on comparisons to more precise radiative transfer bench marks such as line-by-line and vector doubling calculations that utilize laboratory measurements for cloud and aerosol refractive indices and absorption line radiative property information.
Atmospheric CO2 Thermostat: Continued Dialog (Part III)
As you commented earlier, attribution of the greenhouse effect would be of greater interest if performed for climate forcing perturbations relative to current climate (such as doubled CO2). Long ago, we did describe such attribution for cases of doubled CO2 and 2% solar irradiance increase (Hansen et al., 1984, AGU Geophysical Monograph, 29, 130–163), both of which produced a global equilibrium warming of about 4 °C. The question of how large is the cloud feedback sensitivity in current climate is the one question that is most acute. While clouds may account for 25% of the total atmospheric greenhouse strength, a strongly positive cloud feedback response is not representative of current climate modeling analyses, which suggest a near-zero cloud feedback sensitivity. This point can be investigated by performing a detailed flux change attribution study for doubled CO2. The radiative responses for both CO2 and water vapor changes relative to current climate amounts are basically logarithmic, i.e., indicative of strong saturation. It would appear that the radiative response of cloud changes relative to the current climate cloud distribution is even more strongly saturated than the CO2 and water vapor responses.
We performed a zonal feedback analysis (described in part by Lacis and Mishchenko, 1995, in Aerosol Forcing of Climate, Dahlem Workshop Reports, 17, 11–42) of the Hansen et al. (1984) climate sensitivity experiments. This analysis showed cloud feedback to be rather complicated, comprised of changes in cloud cover, cloud height, and column optical depth with a latitudinal dependence that produced positive cloud feedback in low to middle latitudes, but negative cloud feedback at high latitudes, as shown (by the orange curve) in Figure 2.
Water vapor feedback (blue curve, positive at all latitudes) is also a complicated feedback in that there is latitudinal dependence, a significant amount of the feedback response due to the vertical redistribution of water vapor, a strong negative feedback due to moist adiabatic lapse rate change at low latitudes, but which becomes a positive feedback in the polar regions. As expected, the snow/ice feedback (green curve) is also a strong feedback, but confined to polar latitudes. In the above, all of the feedback responses have been expressed in terms of temperature equivalents, i.e., the fraction of the zonal temperature change attributable to that particular feedback effect.
In Figure 2, the solid black curve is the zonal mean equilibrium surface temperature change for doubled CO2 (4.2 °C global mean). The dotted black line is DTo = 1.2 °C the (near-constant with latitude) no-feedback equilibrium temperature change due directly to doubled CO2 that would be climate system temperature response produced in the absence of feedbacks. The red curve is the ‘advective’ feedback, depicting the net effect (and near cancellation) of changes in the advective transport of latent heat, sensible heat, and geopotential energy (each one of which is an order of magnitude larger than the radiative water vapor and cloud feedback contributions). However when averaged globally, the advective feedbacks must educe to zero.
As has been pointed out by Aires and Rossow (2003, Q. J. Royal Meteorol. Soc., 129, 239–275), the feedback interactions of the climate system are non-linear, and the feedback sensitivities are state-dependent and therefore variable in time, and thus are not ‘constants’ of the system. This means that inferring climate feedbacks from linear regressions of cloud or water vapor changes with respect to applied forcings is not likely to be successful. Nevertheless, by performing a detailed radiative flux change attribution for all contributing radiative components between two different climate equilibrium states (say, control vs doubled CO2), it is possible to infer on the basis of the flux change attribution how much forcing was provided by the changes in non-condensing radiative forcing agents, compared to the flux changes attributable to the feedback contributing components. From this radiative flux change comparison, we can infer the relative magnitude of the feedback sensitivity for that particular radiative forcing experiment. From analysis of many such radiative forcing experiments, we can get a better idea of the general characteristics of the climate feedback response, and how the feedback response may depend on the nature of the applied forcing, as described in the radiative forcing and climate response analyses by Hansen et al., 1997, J. Geophys. Res., 102, 6831–6864.
We are in the process of doing a feedback attribution analysis (as in Figure 2) for doubled CO2 with the GISS ModelE. The analysis is straightforward, but tedious, in having to swap water vapor, cloud, and temperature profile fields between the control and double CO2 equilibrium run results and evaluating the instantaneous TOA radiative flux changes. A feedback sensitivity of 3°C for doubled CO2 is strongly supported by the geological record, suggesting that this analysis will provide realistic feedback sensitivities for climate perturbations relative to current climate.
The real uncertainties in climate modeling lie in the area of understanding the natural (unforced) climate fluctuations that occur on inter-annual and decadal time scales, and on regional spatial scales. This variability occurs because the local climate system responses to energy imbalances strongly overshoot the imbalance, achieving energy balance only in a global and time averaged sense. Fortunately, this natural (unforced) climate variability produces fluctuations about the equilibrium climate state, and therefore does not contribute to the long term climate trend.
Perspective and Overview
The complexity of the physical processes that constitute the terrestrial climate system is undeniable. Clearly, full understanding of climate is not likely to be achieved in the foreseeable future since everything from microscopic to cosmic makes some contribution to climate, even if that contribution is miniscule. On the other hand, an adequate understanding of how the climate system works and operates is within reach.
Things that we know well
Terrestrial climate is established as the result of energy balance between SW solar radiation absorbed by the Earth and the LW thermal radiation emitted by the Earth.
Atmospheric absorption of LW radiation by water vapor, clouds, CO2, and other trace gases produces a greenhouse effect that keeps the surface temperature of Earth about 33 °C warmer than it otherwise would be without the atmospheric greenhouse absorbers.
Of the 33 °C terrestrial greenhouse effect, water vapor is responsible for about 50% of the effect, 25% is due to clouds, 20% is due to CO2, and the remaining 5% is contributed by CH4, N2O, O3, CFCs, and other lesser constituents.
The atmospheric distribution of water vapor and clouds is the result of feedback processes, hence the water vapor and cloud amounts are determined by the prevailing meteorological conditions.
The non-condensing greenhouse gases (CO2, CH4, N2O, O3, and CFCs) provide the ultimate support structure for the terrestrial greenhouse effect, even though by themselves they account only for 25% of the total atmospheric greenhouse effect.
Accurate measurement and monitoring of the non-condensing GHGs shows unrelenting increase in atmospheric GHG concentrations, with an accumulated radiative forcing of about 3 W/m2 since 1880.
Since the non-condensing GHG increase is due almost entirely to human industrial activity, primarily the burning of fossil fuel, humans are fully responsible for the global warming.
Accurate measurements of solar irradiance over three solar cycles since the late 1970s show solar cycle variability to be of roughly 1 W/m2 amplitude, but with no significant trend.
Aerosols are important contributors of climate forcing, with non-absorbing aerosols and associated cloud-aerosol indirect effect contributing about –1 W/m2 apiece, and black carbon aerosols contributing about 0.8 W/m2 of warming.
The current climate model sensitivity (for doubled CO2) of 3 °C per 4 W/m2 forcing is in good agreement with the geological (400 K-year) ice core record.
The climate system also undergoes natural (unforced) variability about its global equilibrium state with regional shifts in climate patterns on inter-annual and decadal time scales.
Things that we know less well
We know that aerosols are significant contributors to global climate change, but aerosol radiative properties, cloud-aerosol indirect effect, and the trend in aerosol changes are poorly constrained by intensity-only measurements that have great difficulty separating aerosol radiative properties from sub-pixel cloud contamination and from changes in spectral surface reflectivity.
The long term trend in solar irradiance change must be inferred indirectly based on sunspot cycle changes and proxy information.
While changes in cloud distribution between two equilibrium climate states can be interpreted as ‘cloud feedback’, cloud response to changing meteorological conditions can impact multiple cloud characteristics (cloud cover, cloud height, cloud life time, water/ice phase, optical depth, particle size, diurnal phase), all of which have radiative consequences, some affecting the SW more, others the LW, but which are not readily confirmable with available observational data.
While climate models do exhibit natural variability on inter-annual and decadal time scales that is qualitatively comparable to the real world, climate models have limited skill in modeling the regional and inter-annual climate fluctuations that take place in the climate system even in the absence of external forcing.
The bulk of the problems related to the realistic modeling of regional climate patterns and unforced variability are undoubtedly attributable to the still primitive state of ocean circulation and heat transport, which is decades behind the advances made in atmospheric modeling.
The state of modeling of ice sheet dynamics, in particular the rate at which ice sheets will disintegrate in the face of continued global warming, is in an even more primitive state than that of the ocean climate response modeling.
There is anticipation, perhaps even support from observational evidence but with considerable uncertainty, that the frequency and magnitude of extreme weather events may be increasing as the strength of the hydrological cycle intensifies in step with global warming.
Beside the geological evidence that the Earth could not support polar ice caps when atmospheric CO2 was greater than about 450 ppm (and sea level was more than 200 ft higher than present), there is little clear indication of a ‘tipping point’ beyond which recovery from polar ice cap meltdown might become problematic.
We currently seem to be operating under the ‘no regrets’ climate policy first formulated under the first Bush administration, which basically states that if anything undesirable should happen because of global climate change, we will then deal with that problem after the fact.
This approach appears to have saved about $200 million in not upgrading New Orleans levees. Unfortunately, the cost of dealing with Katrina after the fact was about $200 billion.
Hundreds of billions will be saved by not combating global warming. But the eventual cost may be hundreds of trillions to relocate major cities to higher ground ahead of rising sea levels.