Here's a comment from "Taking climate model evaluation to the next level", Veronika Eyring et al, Nature Climate Change (2019):
An important issue that remains to be fully addressed is the extent to which model errors affect the quality of climate projections and subsequent impact assessments. Traditionally, many climate projections are shown as multimodel averages in the peer-reviewed literature and IPCC reports, with the spread across models presented as a measure of projection uncertainty.
There is now emerging evidence that weighting based on model performance may improve projections for specific applications.
A further complication in devising model weighting approaches is that many CMIP models share components, or are variants of another model in the ensemble, and hence are not truly independent.
This has the potential to bias the multimodel results in ways that are only beginning to be explored.
The lack of independence challenges the notion of a ‘model democracy’, in which each model is weighted equally.
Here's AR6 (IPCC 6th assessment report) making a valuable comment re the rainfall example I raised in the article. From ch 3, p449:
"A fact hindering detection and attribution studies in precipitation and other hydrological variables is the large internal variability of these fields relative to the anthropogenic signal."
Steve: In contrast to the work you describe, I find it very easy to constrain the change in precipitation due to climate change to a narrow range - as long as we are talking about total GLOBAL precipitation. If we hypothesize that climate sensitivity is between 1.8 K and 3.6 K and that the forcing from doubled CO2 is 3.6 W/m2, then our planet emits and reflects an additional 1-2 W/m2 radiation per 1 degK of surface warming (a climate feedback factor of 1.8 to 3.6 W/m2/K. (These values are chosen to simplify calculations and can be changed modestly later to agree with estimates from energy balance models, for instance.)
Now let's assume that the planet's albedo remains constant with rising surface temperature (an assumption we can revisit later). If the flux of heat leaving the planet for space increases 1-2 W/m2/K, then the increased flux of heat leaving the surface must increase at the same rate at steady state. Modtran shows that the increased upward flux of radiation from a warmer surface is almost perfectly balanced by the increase in DLR, especially if you chose the constant relative humidity option. That means that convection of heat from the surface (80 W/m2 of mostly latent heat) must increase about 1-2 W/m2/K or about a 1.25%/K to 2.5%/K increase in evaporation and precipitation. That is a fairly tight constraint on how much global rainfall can increase as a result of global warming.
If saturation vapor pressure increases at 7%/K, how can the increase in convection of latent heat be much smaller? The overturning of the atmosphere must slow, raising relative humidity at the surface and thereby suppressing evaporation. With 80% relative humidity over the ocean, a 1% increase in relative humidity is about a 5% decrease in undersaturation and therefore evaporation.
Now, how much can changes in the planet's albedo with temperature change this 1.25-2.5%/K increase in precipitation? The planet reflects about 100 W/m2 back to space, so a 1 W/m2/K increase in albedo with rising surface temperature (a negative feedback of -1 W/m2/K) would allow precipitation to increase 2.25-3.5%/K. If the planet's albedo decreases with rising temperature by 1 W/m2 (a positive feedback of +1 W/m2/K), then precipitation will only increase 0.25-1.5%/K. Most climate models think cloud feedback is positive, but less than +1 W/m2.
Could cloud feedback be bigger in magnitude than +/- 1 W/m2/K? Actually, the small values reported by climate models do make some sense. In most places, the sky is cloud where the air is rising and clear where the air is sinking, and there is no reason that the area devoted to these fluxes should change. That would make SWR cloud feedback zero. The exceptions to this rule are marine boundary layer (MBL) clouds, which are produced at the top of the boundary layer where air is subsiding over an ocean cooled by upwelling or ocean currents from the poles. Climate models simulate far too few MBL clouds. Most climate models think MBL will decrease modestly with warming, but my intuition says that a boundary layer with 1% higher humidity should have more clouds at the top.
Consequently, the large changes in precipitation with global warming seen in some locations must be balanced by the opposite changes in other locations. Maps showing those changes globally are often badly distorted in the polar regions of flat projections. If extreme changes in precipitation (2 standard deviations) are about +/-50% of average ANNUAL precipitation, climate models that properly simulate this variability need to chaotically shift large amounts of precipitation from one place to another in different years. Variability in temperature is much smaller on a percentage basis. Modeling precipitation is much harder than temperature for this reason.
Will, I've moved your comment over to the entertaining comments section of this article:
Digression #3 - The "Greenhouse effect"
https://scienceofdoom.substack.com/p/digression-3-the-greenhouse-effect
I guess that’s easier than addressing the cyclical nature of “natural variability”. Just bury my comment in a dead thread. Well done!
Here's a comment from "Taking climate model evaluation to the next level", Veronika Eyring et al, Nature Climate Change (2019):
An important issue that remains to be fully addressed is the extent to which model errors affect the quality of climate projections and subsequent impact assessments. Traditionally, many climate projections are shown as multimodel averages in the peer-reviewed literature and IPCC reports, with the spread across models presented as a measure of projection uncertainty.
There is now emerging evidence that weighting based on model performance may improve projections for specific applications.
A further complication in devising model weighting approaches is that many CMIP models share components, or are variants of another model in the ensemble, and hence are not truly independent.
This has the potential to bias the multimodel results in ways that are only beginning to be explored.
The lack of independence challenges the notion of a ‘model democracy’, in which each model is weighted equally.
Here's AR6 (IPCC 6th assessment report) making a valuable comment re the rainfall example I raised in the article. From ch 3, p449:
"A fact hindering detection and attribution studies in precipitation and other hydrological variables is the large internal variability of these fields relative to the anthropogenic signal."
Steve: In contrast to the work you describe, I find it very easy to constrain the change in precipitation due to climate change to a narrow range - as long as we are talking about total GLOBAL precipitation. If we hypothesize that climate sensitivity is between 1.8 K and 3.6 K and that the forcing from doubled CO2 is 3.6 W/m2, then our planet emits and reflects an additional 1-2 W/m2 radiation per 1 degK of surface warming (a climate feedback factor of 1.8 to 3.6 W/m2/K. (These values are chosen to simplify calculations and can be changed modestly later to agree with estimates from energy balance models, for instance.)
Now let's assume that the planet's albedo remains constant with rising surface temperature (an assumption we can revisit later). If the flux of heat leaving the planet for space increases 1-2 W/m2/K, then the increased flux of heat leaving the surface must increase at the same rate at steady state. Modtran shows that the increased upward flux of radiation from a warmer surface is almost perfectly balanced by the increase in DLR, especially if you chose the constant relative humidity option. That means that convection of heat from the surface (80 W/m2 of mostly latent heat) must increase about 1-2 W/m2/K or about a 1.25%/K to 2.5%/K increase in evaporation and precipitation. That is a fairly tight constraint on how much global rainfall can increase as a result of global warming.
If saturation vapor pressure increases at 7%/K, how can the increase in convection of latent heat be much smaller? The overturning of the atmosphere must slow, raising relative humidity at the surface and thereby suppressing evaporation. With 80% relative humidity over the ocean, a 1% increase in relative humidity is about a 5% decrease in undersaturation and therefore evaporation.
Now, how much can changes in the planet's albedo with temperature change this 1.25-2.5%/K increase in precipitation? The planet reflects about 100 W/m2 back to space, so a 1 W/m2/K increase in albedo with rising surface temperature (a negative feedback of -1 W/m2/K) would allow precipitation to increase 2.25-3.5%/K. If the planet's albedo decreases with rising temperature by 1 W/m2 (a positive feedback of +1 W/m2/K), then precipitation will only increase 0.25-1.5%/K. Most climate models think cloud feedback is positive, but less than +1 W/m2.
Could cloud feedback be bigger in magnitude than +/- 1 W/m2/K? Actually, the small values reported by climate models do make some sense. In most places, the sky is cloud where the air is rising and clear where the air is sinking, and there is no reason that the area devoted to these fluxes should change. That would make SWR cloud feedback zero. The exceptions to this rule are marine boundary layer (MBL) clouds, which are produced at the top of the boundary layer where air is subsiding over an ocean cooled by upwelling or ocean currents from the poles. Climate models simulate far too few MBL clouds. Most climate models think MBL will decrease modestly with warming, but my intuition says that a boundary layer with 1% higher humidity should have more clouds at the top.
Consequently, the large changes in precipitation with global warming seen in some locations must be balanced by the opposite changes in other locations. Maps showing those changes globally are often badly distorted in the polar regions of flat projections. If extreme changes in precipitation (2 standard deviations) are about +/-50% of average ANNUAL precipitation, climate models that properly simulate this variability need to chaotically shift large amounts of precipitation from one place to another in different years. Variability in temperature is much smaller on a percentage basis. Modeling precipitation is much harder than temperature for this reason.