By Erin Owain and Richard Bater
A recently published paper calls upon climatologists to build models on decision-relevant timescales to inform shorter-term, local decision making by policy makers.
Over the past few decades, significant scientific advancements in global climate models have revolutionised our understanding and perception of climate change By mimicking the dynamics of the global climate system, climate models have enrichened our knowledge and understanding of the knock-on impacts that changes to the climate system can have on the wider Earth system. Climate models, produced by over 20 centres around the world, have played a critical role in providing scientific evidence for decision makers to act to reduce emissions and adapt to projected impacts.
However, in recent years demand has been placed on climate science by policy makers to produce increasingly high-resolution climate projections to inform shorter-term, local decisions. The authors of a recently published paper argues that this is partly attributable to an over-estimation, on the part of decision makers, of the level precision with which the current set of models are able to project future change.
As the authors note, “… the adaptation community should be aware that widely available climate change projections are overconfident and are advised to avoid seductive promises of information about future climate conditions at local scales and particular future dates”. Additionally, ‘optimising’ decisions using such data, in the absence of rigorous contextualisation and evaluation, can represent poor adaptation practice, especially where inadequate, expensive, or inflexible adaptation measures become ‘locked-in’.
Decision-making timescales across the public and private sectors are often relatively short-term, relative to the timescales of climate projections. The paper draws attention to an over-reliance by decision makers on high-resolution climate projections derived from downscaled climate models. Additionally, it questions whether the demands placed on climate services to produce high resolution climate projections is warranted given that decision-making does not always require such granularity, noting that: “The predominant focus on end-of-century projections neglects more pressing development concerns, which relate to the management of shorter-term risks and climate variability.” A shorter time horizon is often more relevant in lower income countries, which can be more vulnerable to climate shocks due to higher sensitivity and lower adaptive capacity.
A common approach to meet the demand for climate projections at the local level has been to downscale General Circulation Models (GCMs). Downscaling is a process of generating higher spatial and temporal-resolution data from lower-resolution data and is used to derive local-scale data able to inform short-term decision-making.
However, the various methods of downscaling have limitations:
- Uncertainties regarding the underlying GCM projection data can be compounded, as additional assumptions and approximations are introduced during model selection and processing.
- Dynamic downscaling can give a false sense of spatial precision whilst relying on fewer models, whereas temporal downscaling can risk mis-portraying shorter-term projections (3 to 10 years) as being akin to forecasts.
- The error between observed and projected climate change for some parameters can be considerable at local scales, with observed change often being more severe than that projected.
- Models can struggle to reliably represent seasonality, extreme values, and tipping points.
These factors mean that it is important to consider both ‘outlier’ models and future values that could exceed those projected by any of the climate models, whilst bearing in mind that some models are known to perform better in some region better than others.
While a common reflex has been to request such high-resolution climate data, in other areas decision makers across sectors have been accustomed to acting in a context of uncertainty, whether related to cyber-attacks, political instability, fluctuation in oil prices and exchange rates, disease epidemics or natural disasters. It is well understood that such eventualities cannot be predicted with high levels of certainty beyond the short-term: as the paper also note, “Often…detailed planning is possible without detailed climate change projections”.
On the other hand, integrating historical climate data with analysis of real-time data and short-term forecasting can be an effective, high-confidence guide to making robust decisions related to climate adaptation. As noted by the authors, there should be a focus by climatologists on building models on decision-relevant timescales, encouraging further dialogue or intermediation between climate science and end users.
Climate projection data remain an indispensable and scientifically sound guide to how climate is likely to change in the future. This paper, however, is a timely corrective to a tendency to overstate the precision of climate model outputs and to make resilience building efforts contingent on the ever-finer optimisation of climate models. A broad understanding of the direction and magnitude of change in given climate parameters, and their likely impacts for given users, can be adequate to identify and prioritise adaptation strategies and measures today. Decisions can be taken today that are robust to a range of climate scenarios, and low-regret, low-cost measures can be implemented that can be easily reversed in light of experience and new information.
In future, as the paper concludes, it is important that the climate services community refocuses attention on better assessing and translating the significance of projected change versus observed variability and trends. Moreover, whilst noting the resource implications it can carry, they could improve the evaluation of climate models selected for use in climate risk analysis. In representing future climate change, it remains, as ever, imperative to consider and translate model reliability and uncertainty, and convey the range of plausible future change.