And the series continues, poking another hole in the models, with bigger holes to come.
See the larger version in the post below
What if CO2 caused more greenery, which produced more volatile organic gases, which increased rainfall and changed cloud cover? The models would be blind to it. They’re “supercomputer-complicated”, but miss many of the feedbacks on Earth. The only feedbacks the models consider are ones that occur because of changes in temperature. And worse, it’s not just changes in temperature, but specifically, changes in surface temperature.
If, say, cosmic rays caused a change in cloud cover, or the Sun influenced ozone which in turn caused the jet streams to shift closer to the equator, there are no feedbacks worth mentioning according to the large GCM models. The conventional basic model assumes, is built on the idea that nothing causes changes to Earth’s climate unless it works through surface heating — and the GCMs have the same architecture. Cloud cover does not change ice cover. Ocean currents don’t change cloud cover. Changes in biology don’t change clouds. Only changes in surface temperature changes cloud cover.
It’s a good place to start looking for missing negative feedbacks (though, technically, “feedback” means a “feedback to surface warming” in much of the climate literature). Funny how the language matters isn’t it?
This architectural feature is inherited by the GCMs. Here David shows how the conventional model could be structured if feedbacks were introduced systematically. Baby steps…
— Jo
—————————————————————-
5. Error 2: Omitting Feedbacks that are not Temperature-Dependent
Dr David Evans, 27 September 2015, David Evans’ Basic Climate Models Home, Intro, Previous, Next, Nomenclature.
The second error in the conventional basic climate model is an architectural error, a systematic error in structure: it omits all feedbacks that are not responses to surface warming.
In a general sense, a “feedback” is a response to a change that affects whatever caused the change in the first place. For example, surface warming causes more evaporation from the oceans and thus more water vapor, but water vapor is the main greenhouse gas so this might in turn cause more surface warming — so water vapor is a feedback (and the biggest feedback to surface warming; see the components of the total feedback f before Eq. (10) of post 3).
The architecture of the conventional model is a radiation balance. For each climate driver and the surface temperature, it computes the forcing (or radiation imbalance, the increase in net TOA downward flux) — and of course these forcings sum to zero in steady state.
This arrangement is symmetric in the climate drivers and the surface temperature (that is, in each element of the set {climate driver 1, …, climate driver n, surface temperature}). (There is a computational asymmetry in the status of these variables, because the surface warming is unknown while the forcings for the other variables are known — as illustrated by the arrows in the diagram of the conventional model (Fig. 2 of post 3), which indicate what is computed from what. That is, the forcing from surface warming must be exactly opposite and equal to the combined forcings from the other climate influences.)
When feedbacks were introduced to the conventional model (see post 3), they are applied to the surface temperature but not the climate drivers: all the conventional feedbacks are in response to surface warming. This makes the architecture fundamentally asymmetric.
The symmetric but unconventional introduction of feedbacks is shown in Fig. 1 below, where each driver, not just the surface temperature, has its own feedbacks (which always go in the opposite direction to existing information flow, because they feed back).
Figure 1: The symmetric application of feedbacks to the symmetric radiation–balance architecture: each driver as well as the surface temperature has its own specific feedbacks. The conventional model (Fig. 2 of post 3) only has feedbacks in response to surface warming — it omits the dashed feedbacks. (The no-feedbacks-ASR has no feedbacks, so fA is zero — it is just shown for completeness.)
In the conventional model all “feedbacks” are in response to surface warming: they are directly dependent on the surface temperature, but not on the climate drivers or other feedbacks.
If there exist feedbacks that respond weakly to surface warming but respond strongly to changes in a climate driver, or if there are feedbacks to climate drivers that are not also feedbacks in response to surface warming, then the conventional model overlooks them.
One might argue that any climate driver will affect the surface temperature, at least indirectly, so modeling all feedbacks as responses to surface warming is adequate. But this implies the feedbacks to any type of influence are the same — that feedbacks do not “know” which driver caused them: for example, a surface warming of 0.2 °C could be due to extra CO2 (which leaves outgoing longwave radiation (OLR) constant, ignoring albedo feedbacks) or an increase in TSI (which increases OLR), but in the conventional model the feedbacks are identical in both cases. Physically, this risks serious over-simplification — surely the feedbacks to different climate influences might be different.
These omissions would be remedied by adding feedbacks to each driver, as in Fig. 1 above. A driver’s feedbacks might then significantly change the ultimate forcing due to the driver. This would make the conventional basic model much more difficult to solve (there would be a lot of extra terms in Eq. (3) of post 3, which could become an intractable mess).
A feedback that responds much more strongly to changes in CO2 than to surface warming is proposed in the post after next, corresponding to a negative value of fC in Fig. 1.