@MM: I don’t think [there are adequate models], but I welcome good reports if you find any.
Whether predicting next week’s weather, next month’s flu season, or next century’s climate, three ingredients are needed for any reasonable degree of predictive skill.
1. A qualitative model of the relevant processes.
2. Historical data quantifying those processes.
3. Statistical reconciliation of the qualitative and the quantitative.
We can all quarrel about the incompleteness and imprecision of the data, and the lies and damned lies of the statisticians. But these pale into insignificance compared to today’s climate models.
If wading through the 50-odd chapters of WG1 and WG2, each summarizing the research of several hundred peer-reviewed publications, seems like a tall order, try reading the millions of lines of code of the thirty or more CMIP5 climate models that have grown like Topsy over several decades starting from very successful numerical weather prediction models.
In the division of labor needed to accomplish so much, is there even one person who has a clear picture of how these models bear on expected climate in 2100?
Ask yourself. Do you care whether your first great-great-great-granddaughter’s fifth birthday party in the nearest park to her house will be rained out? Or would you be satisfied with knowing the global mean surface temperature of the planet as averaged over the period 2070-2130?
Surely the former question would require a more complex model than the latter.
Which raises the question, how does the complexity of a climate model depend on the question it was designed to answer with an acceptable degree of predictive skill?
On the afternoon of December 17 I’ll be addressing that question at the annual Fall Meeting of the American Geophysical Union. The bottom line will be that the questions most often raised about climate in 2100 can be answered with a very high degree of predictive skill based on mind-bogglingly simpler models than those of the massive CMIP5 suite. Check in that time.