Technical post on Nordhaus' new DICE model

I'm using this post to document a comment I made about Dr. Nordhaus' revised DICE dynamic model. This model is used to predict world wide CO2 emissions, and has been used to create "business as usual" economic runs which in turn are used to set USA, EU, and other countries' energy and emissions policies. Below I explain the model is flawed, requires a significant overhaul, and explain briefly the reasons why. 

Nordhaus' description of the work he did with DICE is here: 

http://cowles.yale.edu/sites/default/files/files/pub/d20/d2057.pdf

If we refer to page 12, in the paragraph headed "decarbonization" we can read "the model has a highly condensed version of the energy sector". I don't have access to Dr Nordhaus, but I would like to point out that a dynamic model expected to forecast CO2 emissions ought to treat the energy sector in detail. Emissions reductions are dictated in part by government policies, population growth, changes in the size of economies, the cost of fossil fuels (ignoring externalities) and the cost of replacement energy sources. 

Please note that emissions have almost stabilized over the period 2014-15-16, even though oil and coal prices were in the lower part of the price cycle. This was driven in part by policy, but it was mostly due to market conditions. The key item to focus on is the significant drop in the cost of photovoltaic solar power, a slight improvement in energy storage costs, slight improvement in wind turbine costs, etc. 

Now we are entering a period of rebounding oil and natural gas prices, and we do know there's a limit to reserves (this seems to be ignored by the dice model). The forces which "bent" the CO2 emissions to a much lower growth rate are still working. Therefore I must conclude the DICE model needs to be revised and gave a much more detailed and properly documented energy sector treatment. 

This flaw renders its results fairly useless for policy setting, and this means it's fairly easy to overturn any policy decisions made using this particular model's results.