EnergICE: a Simple but more Realistic Energy Sector for the DICE model (with an application to solar geoengineering)

The stylised Integrated Assessment Model DICE by Nobel laureate William Nordhaus has been criticised for its overly simplifying assumptions, yet is still widely used as a testbed model (e.g. for investigating policy effects of climate tipping [Cai, Lenton and Lontzek, 2015] or solar geoengineering [Helwegen et al., 2019]) and, occasionally, policy advice. Surprisingly, much of the past criticism was focusing on the extremely difficult issue of modelling climate-induced damage, while the equally problematic formulation of mitigation costs [Grubb, Wieners and Yang, 2021] remained relatively unchallenged, despite being a more tractable problem. In particular, DICE’s mitigation costs at any time t only depend on the fraction of emissions avoided at time t, ignoring the fact that past mitigation investment affects future costs (“if you build a wind park this year, it will still save carbon next year”).

In the current study, I introduce EnergICE, a DICE version with a still simple, but dynamically more consistent energy sector. Rather than picking a fraction of emissions (w.r.t. a baseline) to be avoided by unspecified means, the social planner now makes investment decisions: To fulfil the system’s energy demand, the planner can choose from “brown” (fuel-using) and “green” (renewable) power plants. As green energy cannot always be generated (dark, windless days), storage facilities can also be built. Green plants have initially only slightly higher lifetime costs than brown ones, but storage is very expensive. Learning-by-doing effects reduce the price of both green plants and storages, while brown plants are a mature technology with little learning potential. On the other hand, fuel becomes more expensive with cumulative use, as harder-to-extract reservoirs must be mined once the easy-to-extract ones are exhausted. Therefore even without climate change, some green transition will eventually occur. However, the transition can remain incomplete for decades, with only enough green plants for the “sunny” periods, but fuel-based energy being used in “dark” times.

A simplified version of EnergICE without storage also exists; in that version, the green plants implicitly contain a storage facility and are thus initially expensive. While capturing the investment-like character of mitigation, the simplified version is hardly more complex than the original DICE model; in particular, it does not add a decision variable.

As a test case, the EnergICE model is used to study the desirability (or undesireability) of solar geoengineering under uncertain climate sensitivity. The choice of the energy model (DICE vs EnergICE) can alter the “optimal” level of solar geoengineering by up to a factor of 6, which illustrates that the treatment of mitigation deserves more attention when using DICE-like models as testbed for new concepts.