Characterizing and responding to uncertainty in climate change

The development and analysis of climate policy proposals intertwine with the structure of knowledge and the possibility for changing it. Key questions concern the long-term interaction between policy, technology, infrastructure, and the earth system, but each of these components is deeply uncertain....

Full description

Bibliographic Details
Main Author: Lemoine, Derek Mark
Other Authors: Kammen, Daniel M
Format: Other/Unknown Material
Language:English
Published: eScholarship, University of California 2011
Subjects:
Online Access:https://escholarship.org/uc/item/5t82v5j3
Description
Summary:The development and analysis of climate policy proposals intertwine with the structure of knowledge and the possibility for changing it. Key questions concern the long-term interaction between policy, technology, infrastructure, and the earth system, but each of these components is deeply uncertain. This dissertation advances the description of knowledge about the climate system, the assessment of economic responses to climatic possibilities, and the development of policy that positions society to achieve long-term climate goals. It offers new paths to describing understanding of complex systems and to modeling optimal management under structural uncertainty.The first chapter formalizes uncertainty about equilibrium climate change. Its hierarchical Bayes framework allows climate models to be incomplete and to share biases, and it shows how prior beliefs about models' completeness and independence interact with models' estimates of feedback strength to determine distributions for temperature change. When models might share biases, the results of additional models might tell us more about models' common structure than about the real-world processes they aim to represent. The most valuable information would then come not from related models but from alternate estimates that should carry a different set of unobservable biases. The possibility that models are wrong in common ways limits the degree to which models' estimates can narrow the probability distribution for feedback strength, which also limits our ability to rule out extreme climatic outcomes.The second chapter empirically estimates a feedback that is especially difficult to model. Climate-carbon feedbacks (or carbon cycle feedbacks) describe the effect of temperature on carbon dioxide (CO2). If they are positive, then not only does anthropogenic CO2 cause warming via the greenhouse effect and earth system feedbacks, but this warming itself increases CO2 and so causes further warming. Previous empirical work estimated a stronger feedback than did coupled climate-carbon cycle models. However, those empirical estimates were probably biased upwards while coupled models' estimates were primarily driven by a few ill-constrained parameters. This chapter attempts to obtain an unbiased estimate of climate-carbon feedback strength by using variations in summer radiation in the Arctic (i.e., variations in orbital forcing) to identify the effect of temperature on CO2 in 800 ky ice core records. It finds a range for climate-carbon feedbacks that is closer to coupled models' estimates than to previous empirical work. Since climate-carbon feedbacks are probably positive, temperature change projections tend to underestimate an emission path's consequences if they do not allow the carbon cycle to respond to changing temperatures.The next three chapters assess economic responses to climate change in a policy-optimizing integrated assessment model, in games with long-lived investments into abatement capital, and in a cost-effectiveness model with multiple policy options stretching over long time horizons. The first of these chapters extends a well-known integrated assessment model to include the possibility of abrupt shifts in the climate system. It also changes the model's structure to make the decision-maker aware of uncertainty and of the possibility for learning over time, and it generalizes the welfare evaluation to reflect that uncertainty about temperature change is qualitatively unlike uncertainty about climate thresholds. It finds that tipping points can increase the near-term social cost of carbon by more than 50% when they raise climate sensitivity or make damages more convex. They have less of an effect when they increase the atmospheric lifetime of CO2 or the quantity of non-CO2 greenhouse gases. Allowing the policymaker to be differentially averse to consumption fluctuations over time and over risk increases the near-term social cost of carbon by 150%, with tipping point possibilities then increasing it by another 50%. The possibility of tipping points is more important for the social cost of carbon than is the ambiguity attitude the decision-maker uses in evaluating them.The second of these climate economics chapters models the optimal emission tax when firms can adopt low-pollution technology that reduces abatement cost. The regulator anticipates this adoption but must set the tax before firms invest. In many cases, a linear emission tax cannot obtain both socially optimal investment and socially optimal emissions because the regulator either will set it inefficiently high to stimulate investment or will set it at an ex post optimal level that obtains inefficiently low investment. The difficulty is that an emission tax fixes both the incentive to invest and the incentive to abate, but these two goals rarely align perfectly when investment is lumpy. In contrast, tradable permits policies do not suffer this tension because the permit price responds automatically to realized investment. A numerical model then considers the ability of the regulator to select not only the level but also the duration of the tax. It shows that outcomes are still often socially inefficient. Further, the regulator will occasionally use a longer tax to obtain investment when firms expect their investments to lower the tax in the next period, but the cost of not being able to adjust the next period's tax limits the parameter space in which the longer tax is employed.The fifth chapter constructs cost-effective dynamic policy portfolios of abatement, research and development (R&D), and negative emission technology deployment in order to achieve 21st century climate targets. It includes two types of stochastic technological change in a stylized numerical model and allows each type of technology to respond both to public R&D and to abatement policies. It compares worlds where negative emission technologies are and are not available, and it compares a world where the century's cumulative net emissions are constrained with a world in which threshold possibilities lead policy to constrain cumulative net emissions in each year during the century. It finds that R&D options are valuable and exercised but do not substitute for near-term abatement. The type of R&D undertaken depends on long-term emission goals because those determine the magnitude of future abatement. When the cumulative emission constraint is stringent, negative emission technologies substitute for near-term abatement and affect the type of R&D undertaken, but if threshold considerations eliminate the freedom to temporarily overshoot emission targets, negative emission technologies become less valuable. The availability of negative emission technologies provides a valuable option to partially undo previous emissions, but abatement also gains option value from increasing future flexibility to forgo reliance on negative emission technologies if the technology or climate prove problematic in the interim.The concluding chapter directly connects uncertainty about climate change to uncertainty about the cost of achieving CO2 targets. It shows how beliefs about technology, temperature, and damages interact to affect the cost-effectiveness of climate targets. It finds that the speed with which damages increase at higher temperatures is the most important of these factors. Both 450 parts per million (ppm) and 550 ppm CO2 targets provide net benefits for quadratic damage functions that reduce annual output by less than the 1-2% estimated for 2.5°C of warming. Cubic damage functions support both CO2 targets even if 2.5°C of warming only reduces output by 0.2% or less. More convex damage functions also reduce the importance of abatement cost uncertainty. significantly increase the range of damage functions that support these targets and decrease the importance of abatement cost uncertainty. In addition, because extreme feedback outcomes have little effect over the next decades, a thinner-tailed temperature distribution (resulting from optimistic prior beliefs about climate models' independence and biases) supports CO2 targets under slightly less severe damages than does the thicker-tailed distribution (resulting from skepticism about climate models' independence and biases). Emission reductions hedge against greater societal sensitivity to temperature increases while exposing society to the upside of positive technology surprises.The epistemology of complex systems in an out-of-sample world is a key motif. This dissertation advances knowledge of climate change and understanding of policy design in settings with limited ability to predict future changes or responses. Further work should seek a more unified framework for describing and acting on knowledge of evolving complex systems.