Uncertainty in Climate Science The Constitutive Elements of Uncertainties in Climate Model Projections
Few fields in human life benefit from what we call ‘climate change’. The consequences of climate change become apparent and are obstructing–sometimes taking–our lives. This is reason enough to anticipate them and implement mitigation and adaptation strategies through policymaking. However, climate model outputs are plagued with uncertainty; this fact evokes controversy about how they should be interpreted, communicated, and used in decision-making. What do these models tell us and should we base any policy decisions on their outcomes if these are highly uncertain? To warrant our trust in the model outcomes regarding decision-making, the sources of and reasons for the uncertainty should be well-understood.
In this thesis, I investigate the emergence of uncertainty in climate model projections. More precisely, I examine what elements throughout the modelling process are the reason for the uncertainty in the outcomes and how this should be interpreted. Is trust in the outcomes warranted? I evaluate this question as a philosopher of science, conducting a philosophical analysis of the climate modelling process and the concepts related to it.
I cast light on the question by analysing climate modelling in three main ways: through a conceptual analysis of ‘model’, ‘climate’, and ‘system’, through an analysis of the experience of ‘climate’, and from a technical perspective by examining the construction of single-model components, coupled global earth system models, and model ensembles.
When considering how the climate is experienced, I find that it is not clear how it could be experienced since it is a scientifically constructed concept. On the other hand, knowledge of this scientific concept may influence our interpretation of weather phenomena. Furthermore, there are many phases in the model-building process; each phase introduces errors and therefore uncertainty that is carried along to the subsequent phases. Observational data is used to estimate some free values in the model. Choosing which values to align with the data on the one hand and with the physical theory on the other hand requires expert opinion. It is a matter of debate whether the outcomes are trustworthy when they result from alignment with observational data, at the expense of adequate representation of the physical processes. The choices made form a source of subjective uncertainty, while the technical constraints of the chosen approach introduce errors. The concept of ‘adequacy-for-purpose’ means that a model is never adequate relative to the entire climate system, but more apt to answer some questions and less apt to answer others.
Overall, there are three important elements to remember: first, a climate model is built to gain insight into climate phenomena that are highly uncertain. Although the model includes uncertainty too, it reduces the uncertainty in the climate phenomena per se. Secondly, climate models are never ‘true’, but can be adequate for their intended purpose. Using climate model outcomes for answering questions they were not built for will result in meaningless responses. Lastly, including more models and more plausible scenarios will result in more different outcomes; if these outcomes are, on average, closer to the ‘truth’, this should not be a problem. Converging outcomes does not necessarily mean they are more correct. Obtaining more correct average answers, by evaluating more models and more scenarios, is a good practice.
Meer lezen