Developing a Theoretical Foundation for Information for Climate-Risk Management: Assessing the Value of Information in a Stylized Example

Many actors and institutions — from insurers to local governments to utilities to individuals — are looking to integrate information about climate into their decision-making and risk management. However, the wide variety of adaptation decisions, across many locations, dependent on different weather variables and timescales, has made a generalized understanding of the kinds of climate information required to support adaptation decisions elusive. In this research we work towards such a generalized understanding, and a broad theoretical foundation to underpin the production and use of climate risk information into adaptation decisions.

We apply a “value of information” (VOI) approach to begin addressing this problem. VOI is appropriate for understanding climate change adaptation because it speaks to settings where an actor must make a decision under uncertainty and may have the option of acquiring new information. Actors use the information available to them to evaluate the set of available options and select the one that seems to perform best based on the actors’ criteria. New information (for instance information on the anthropogenic forcing effect from General Circulation Models ) allows the decision-maker to better optimize their decision and therefore has value, which can be formally quantified as the difference in the expected value of the decision with and without the new information. The VOI framework can flexibly be used to examine how different information about weather and climate risks provides value for the very different types of adaptation decisions that various actors across society face. These decisions can be characterized by a payoff function (combining vulnerability and exposure in a single function) that can be modified by the actor, based on their evaluation of relevant weather and climate risks using the information they have available.

In this talk we will present results from an idealized example. The climate system will be represented by the Lorenz (1963) model, with a slow time-dependent variation in one of the parameters to represent climate change. A variety of idealized loss functions will be considered, designed to represent decisions that might be relevant to agriculture, infrastructure, and insurance. Different types of uncertainties --- epistemic, aleatoric, etc. --- will be represented in this “perfect model” framework. We will consider the trade-offs between them, for example, considering the strengths and weaknesses of “models’ (copies of the Lorenz system with some uncertainty in the parameters, representing climate models with structural uncertainty, and possibly limited ensemble size) vs. synthetic “observed” time series (short samples of output from the actual system, which may or may not resolve either rare events or climate change trends) and describe how the VOI of different information states for adaptation differs across decision-settings.