Learning about climate change uncertainty enables flexible water infrastructure planning

Sarah Fletcher,Megan Lickley,Kenneth Strzepek

doi:10.1038/s41467-019-09677-x

Abstract

Water resources planning requires decision-making about infrastructure development under uncertainty in future regional climate conditions. However, uncertainty in climate change projections will evolve over the 100-year lifetime of a dam as new climate observations become available. Flexible strategies in which infrastructure is proactively designed to be changed in the future have the potential to meet water supply needs without expensive over-building. Evaluating tradeoffs between flexible and traditional static planning approaches requires extension of current paradigms for planning under climate change uncertainty which do not assess opportunities to reduce uncertainty in the future. We develop a new planning framework that assesses the potential to learn about regional climate change over time and therefore evaluates the appropriateness of flexible approaches today. We demonstrate it on a reservoir planning problem in Mombasa, Kenya. This approach identifies opportunities to reliably use incremental approaches, enabling adaptation investments to reach more vulnerable communities with fewer resources.

Highlights

Water resources planning requires decision-making about infrastructure development under uncertainty in future regional climate conditions
Note that while this study focuses on water supply infrastructure, the challenge of characterizing learning about climate uncertainty to enable adaptive planning has been highlighted in a range of other disciplines
We develop a method that integrates iterative Bayesian learning about climate uncertainty into a multi-stage stochastic infrastructure planning model in order to address a critical limitation of adaptive infrastructure planning in both water supply and other domains: estimating upfront how much planners can expect to learn about climate change in the future and whether adaptive approaches are likely to be reliable and cost effective

Summary

Introduction

Water resources planning requires decision-making about infrastructure development under uncertainty in future regional climate conditions. We develop a new planning framework that assesses the potential to learn about regional climate change over time and evaluates the appropriateness of flexible approaches today. We demonstrate it on a reservoir planning problem in Mombasa, Kenya. These updated uncertainty estimates characterize the transition probabilities in a non-stationary stochastic dynamic program (SDP); each possible change in SDP climate state is equivalent to a virtual climate observation This SDP planning formulation takes into account all the potential new information that may be learned in the future as it develops optimal planning policies. We use these polices to evaluate flexible infrastructure planning approaches and compare them to static approaches

Methods

Results

Conclusion