Adaptive stochastic integrated assessment modeling of optimal greenhouse gas emission reductions
- 2.3k Downloads
We develop a method for finding optimal greenhouse gas reduction rates under ongoing uncertainty and re-evaluation of climate parameters over future decades. Uncertainty about climate change includes both overall climate sensitivity and the risk of extreme tipping point events. We incorporate both types of uncertainty into a stochastic model of climate and the economy that has the objective of reducing global greenhouse gas emissions at lowest overall cost over time. Solving this problem is computationally challenging; we introduce a two-step-ahead approximate dynamic programming algorithm to solve the finite time horizon stochastic problem. The uncertainty in climate sensitivity may narrow in the future as the behavior of the climate continues to be observed and as climate science progresses. To incorporate this future knowledge, we use a Bayesian framework to update the two correlated uncertainties over time. The method is illustrated with the DICE integrated assessment model, adding in current estimates of climate sensitivity uncertainty and tipping point risk with an endogenous updating of climate sensitivity based on the occurrence of tipping point events; the method could also be applied to other integrated assessment models with different characterizations of uncertainties and risks.
Keywordsstochastic dynamic programing approximate dynamic programing Bayesian inference tipping point climate sensitivity
The authors would like to extend their deepest appreciation to Dr. Hayriye Ayhan, Dr. Alexander Shapiro, and Dr. Roshan Joseph Vengazhiyil from the School of Industrial & Systems Engineering, as well as Dr. Athanasios Nenes from the School of Earth & Atmospheric Sciences at Georgia Institute of Technology for providing advice and intellectual insight into the research question.
- Cai Y, Judd K, Lontzek T (2012) Dsice: A dynamic stochastic integrated model of climate and economy. SSRN Scholarly Paper ID 1992674, Social Science Research NetworkGoogle Scholar
- Cai Y, Judd K, Lontzek T (2013) The social cost of stochastic and irreversible climate change. Working Paper 18704, National Bureau of Economic ResearchGoogle Scholar
- Geist M, Pietquin O (2011) Parametric value function approximation: A unified view. In: 2011 IEEE Symposium on Adaptive Dynamic Programming And Reinforcement Learning (ADPRL), pp 9–16Google Scholar
- IPCC (2013) Climate change 2013: The physical science basis. Technical ReportGoogle Scholar
- Murphy KP (2000) A survey of pomdp solution techniques. Technical report, UC BerkeleyGoogle Scholar
- Nordhaus WD (2008) A Question of Balance: Weighing the Options on Global Warming Policies. illustrated edition edn. Yale University PressGoogle Scholar
- Powell WB (2011) Approximate Dynamic Programming: Solving the Curses of Dimensionality, WileyGoogle Scholar