An important epistemic issue in climate modelling concerns structural uncertainty: uncertainty about whether the mathematical structure of a model accurately represents its target. How does structural uncertainty affect our knowledge and predictions about the climate? How can we identify sources of structural uncertainty? Can we manage the effect of structural uncertainty on our knowledge claims? These are some of the questions that an epistemology of structural uncertainty faces, and these questions are also important for climate scientists and policymakers. I develop (...) three desiderata for an epistemological account of structural uncertainty. In my view, an account of structural uncertainty should identify sources of structural uncertainty, explain how these sources limit the applicability of a model, and show how the severity of structural uncertainty depends on the questions that can be asked of a model. I argue that analyzing structural uncertainty by paying attention to the details of model building can satisfy these desiderata. I focus on parametrizations, which are representations of important processes occurring at scales that are not resolved by climate models. Parametrizations are often thought to be ad-hoc, but I show that some important parametrizations are theoretically justified by explicit or implicit scale separation assumptions. These assumptions can also be supported empirically. Analyzing these theoretical and empirical justificatory roles of the scale separation assumptions can provide insights into how parametrizations contribute to structural uncertainty. I conclude by sketching how my approach can satisfy the desiderata I set out at the beginning, highlighting its importance for policy-relevant scientific statements about the climate. (shrink)

For computer simulation models to usefully inform climate risk management, uncertainties in model projections must be explored and characterized. Because doing so requires running the model many ti...

This paper deals with the question of whether uncertainty regarding model structure, especially in climate modeling, exhibits a kind of “chaos.” Do small changes in model structure, in other words, lead to large variations in ensemble predictions? More specifically, does model error destroy forecast skill faster than the ordinary or “classical” chaos inherent in the real-world attractor? In some cases, the answer to this question seems to be “yes.” But how common is this state of affairs? Are there precise mathematical (...) results that can help us answer this question? And is dependence on model structure “sensitive” in that arbitrarily small errors can destroy forecast skill? We examine some efforts in the literature to answer this last question in the affirmative and find them to be unconvincing. (shrink)

This article critically examines a recent philosophical debate on the role of values in climate change forecasts, such as those found in assessment reports of the Intergovernmental Panel on Climate Change. On one side, several philosophers insist that the argument from inductive risk, as developed by Rudner and Douglas among others, applies to this case. AIR aims to show that ethical value judgments should influence decisions about what is sufficient evidence for accepting scientific hypotheses that have implications for policy issues. (...) Advocates of extending AIR to climate science claim that values are deeply... (shrink)

To study climate change, scientists employ computer models, which approximate target systems with various levels of skill. Given the imperfection of climate models, how do scientists use simulations to generate knowledge about the causes of observed climate change? Addressing a similar question in the context of biological modelling, Levins (1966) proposed an account grounded in robustness analysis. Recent philosophical discussions dispute the confirmatory power of robustness, raising the question of how the results of computer modelling studies contribute to the body (...) of evidence supporting hypotheses about climate change. Expanding on Staley’s (2004) distinction between evidential strength and security, and Lloyd’s (2015) argument connecting variety-of-evidence inferences and robustness analysis, I address this question with respect to recent challenges to the epistemology robustness analysis. Applying this epistemology to case studies of climate change, I argue that, despite imperfections in climate models, and epistemic constraints on variety-of-evidence reasoning and robustness analysis, this framework accounts for the strength and security of evidence supporting climatological inferences, including the finding that global warming is occurring and its primary causes are anthropogenic. (shrink)

Climate change adaptation is largely a local matter, and adaptation planning can benefit from local climate change projections. Such projections are typically generated by accepting climate model outputs in a relatively uncritical way. We argue, based on the IPCC’s treatment of model outputs from the CMIP5 ensemble, that this approach is unwarranted and that subjective expert judgment should play a central role in the provision of local climate change projections intended to support decision-making.

This is the second of three parts of an introduction to the philosophy of climate science. In this second part about modelling climate change, the topics of climate modelling, confirmation of climate models, the limits of climate projections, uncertainty and finally model ensembles will be discussed.

Projections of future climate change cannot rely on a single model. It has become common to rely on multiple simulations generated by Multi-Model Ensembles (MMEs), especially to quantify the uncertainty about what would constitute an adequate model structure. But, as Parker points out (2018), one of the remaining philosophically interesting questions is: “How can ensemble studies be designed so that they probe uncertainty in desired ways?” This paper offers two interpretations of what General Circulation Models (GCMs) are and how MMEs (...) made of GCMs should be designed. In the first interpretation, models are combinations of modules and parameterisations; an MME is obtained by “plugging and playing” with interchangeable modules and parameterisations. In the second interpretation, models are aggregations of expert judgements that result from a history of epistemic decisions made by scientists about the choice of representations; an MME is a sampling of expert judgements from modelling teams. We argue that, while the two interpretations involve distinct domains from philosophy of science and social epistemology, they both could be used in a complementary manner in order to explore ways of designing better MMEs. (shrink)

The paper discusses modelling uncertainties in climate models and how they can be addressed based on physical principles as well as based on how the models perform in light of empirical data. We ar...

In recent publications we have drawn attention to the fact that if the dynamics of a model is structurally unstable, then the presence of structural model error places in-principle limits on the model’s ability to generate decision-relevant probability forecasts. Writing with a varying array of co-authors, Eric Winsberg has now produced at least four publications in which he dismisses our points as unfounded; the most recent of these appeared in this journal. In this paper we respond to the arguments of (...) Winsberg and his co-workers, and we point out that their criticisms fail. We take this as an opportunity to restate and explain our arguments, and to point to fruitful directions for future research. (shrink)

. Reproducible results and repeatable measurements at the same location are fundamental to science, yet of grave concern to scientists. Involvement in biological re-surveys under MVZ-Berkeley, Harvard-LTER and Hamaarag elucidated “replication” and “location” and untangled “incommensurability” from “no fact of the matter” and “indeterminacy.” All cases revealed incommensurability without indeterminacy on the smallest scale and indeterminacy without incommensurability on higher scales, with communication failure in the former and successful workarounds in the latter. I argue that an involved philosophy helps clarify (...) fundamental concepts in this case, and since it could also help other cases it therefore should be examined in every case before being ruled out. (shrink)

Climate modelling plays a crucial role for understanding and addressing the climate challenge, in terms of both mitigation and adaptation. It is therefore of central importance to understand to what extent climate models are adequate for relevant purposes, such as providing certain kinds of climate change projections in view of decision-making. In this perspective, the issue of the stability of climate models under small relevant perturbations in their structure seems particularly important. Within this framework, a debate has emerged in the (...) philosophy of science literature about the relevance for climate modelling of the mathematical notion of structural stability. This paper adresses several important foundational and epistemological questions that arise in this context, in particular about the the role of abstract mathematical considerations of a qualitative nature for concrete modelling projects with mainly quantitative purposes. (shrink)

Roman Frigg and others have developed a general epistemological argument designed to cast doubt on the capacity of a broad range of mathematical models to generate “decision relevant predictions.” In this article, we lay out the structure of their argument—an argument by analogy—with an eye to identifying points at which certain epistemically significant distinctions might limit the force of the analogy. Finally, some of these epistemically significant distinctions are introduced and defended as relevant to a great many of the predictive (...) mathematical modeling projects employed in contemporary climate science. (shrink)

Recent work by Frigg et. al. and Mayo-Wilson have called attention to a particular sort of error associated with attempts to model certain complex systems: structural modeling error. The assessment of the degree of SME in a model presupposes agreement between modelers about the best way to individuate natural systems, an agreement which can be more problematic than it appears. This problem, which we dub “the system individuation problem” arises in many of the same contexts as SME, and the two (...) often compound one another. This paper explores the common roots of the two problems in concerns about the precision of predictions generated by scientific models, and discusses how both concerns bear on the study of complex natural systems, particularly the global climate. (shrink)

The sensitive dependence on initial conditions (SDIC) associated with nonlinear models imposes limitations on the models’ predictive power. We draw attention to an additional limitation than has been underappreciated, namely, structural model error (SME). A model has SME if the model dynamics differ from the dynamics in the target system. If a nonlinear model has only the slightest SME, then its ability to generate decision-relevant predictions is compromised. Given a perfect model, we can take the effects of SDIC into account (...) by substituting probabilistic predictions for point predictions. This route is foreclosed in the case of SME, which puts us in a worse epistemic situation than SDIC. (shrink)

Scientific models consist of fictitious elements and assumptions. Various attempts have been made to answer the question of how a model, which is sometimes viewed as a fiction, can explain or predict the target phenomenon adequately. I examine two accounts of models-as-fictions which are aiming at disentangling the myth of representing the reality by fictional models. I argue that both views have their own weaknesses in spite of many virtues. I propose to re-evaluate the problems of representation from a novel (...) perspective in which some of the model representations can be regarded as fictional representations. I argue that this type of model representation is credible despite being a fictional representation of the reality. (shrink)

Climate models don’t give us probabilistic forecasts. To interpret their results, alternatively, as serious possibilities seems problematic inasmuch as climate models rely on contrary-to-fact assumptions: why should we consider their implications as possible if their assumptions are known to be false? The paper explores a way to address this possibilistic challenge. It introduces the concepts of a perfect and of an imperfect credible world, and discusses whether climate models can be interpreted as imperfect credible worlds. That would allow one to (...) use models for possibilistic prediction and salvage widespread scientific practice. (shrink)

The goal of this dissertation is to contribute to the epistemology of science by addressing a set of related questions arising from current discussions in the philosophy and science of climate change: (1) Given the imperfection of computer models, how do they provide information about large and complex target systems? (2) What is the relationship between consilient reasoning and robust evidential support in the production of scientific knowledge? (3) Does taking the mean of a set of model outputs provide epistemic (...) advantages over using the output of a single ‘best model’? Synthesizing research in philosophy and science, the thesis analyzes connections among consilient inductions, robustness analysis, and the aggregation of various sources of evidence, including computer simulations, by investigating case studies of climate change that exemplify the strength of consilient reasoning and the security of robust evidential support. It also explains the rationale and epistemic conditions for improving estimates by averaging multiple estimates, comparing a simple case of averaging estimates to practices in multi-model ensemble studies. I argue: (A) the concepts of consilience and robustness account for the strength and security of inferences that rely on imperfect computer modelling methods, (B) consilient reasoning is conducive to attaining robust evidential support, and (C) an analogy can explain why averaging the outputs of multiple models can improve estimates of a target system, given that conditions of model independence, skill and unequal weighting are taken into account. (shrink)