Climate Risk Management (Jan 2023)
Climate learning scenarios for adaptation decision analyses: Review and classification
Abstract
Economic decision analysis is an important tool for developing cost-efficient adaptation pathways in sectors that involve costly adaptation options, such as flood risk management. Standard economic approaches, however, do not consider learning about future changes in climate variables even though a large literature on adaptive planning emphasises the key role of learning over time, because uncertainties about climate change are substantial. An emerging, diverse and fragmented set of economic adaptive decision making approaches, coming under labels such as real-option analysis or optimal control, have started to address this challenge by including the economic valuation of learning in the economic appraisal of adaptation options through making use of so-called climate learning scenarios. We synthesise this literature and classify the climate learning scenarios applied with respect to which climate variable is learned about, which learning sources are employed, how the learning is modelled, which climate data is used for calibrating learning scenarios, which goodness of fit information is provided and how deep uncertainty is handled. Our results show that publications consider learning through observations or do not explicitly state the source of learning. Most authors generate climate learning scenarios through stochastic processes or Bayesian approaches and use climate model output from the IPCC or the UK Met Office to calibrate the learning scenarios. The reviewed literature rarely provides information on the goodness of fit of learning scenarios to the underlying climate data. We conclude that most of the methods used to generate climate learning scenarios are not well-grounded in climate science and are inadequate to represent climate uncertainty. One avenue to improve climate learning scenarios would be to combine a Bayesian approach with emulators that mimic climate model runs based on observations from future moments in time.
