Climate Change Attribution: When Is It Appropriate to Accept New Methods?

Abstract

The most common approaches to detection and attribution (D&A) of extreme weather events using fraction of attributable risk or risk ratio answer a particular form of research question, namely “What is the probability of a certain class of weather events, given global climate change, relative to a world without?” In a set of recent papers, Trenberth et al. (2015, https://doi.org/10.1038/nclimate2657) and Shepherd (2016, https://doi.org/10.1007/s40641-016-0033-y) have argued that this is not always the best tool for analyzing causes, or for communicating with the public about climate events and extremes. Instead, they promote the idea of a “storyline” approach, which asks complementary questions, such as “How much did climate change affect the severity of a given storm?” From the vantage of history and philosophy of science, a proposal to introduce a new approach or to answer different research questions—especially those of public interest—does not appear particularly controversial. However, the proposal proved highly controversial, with the majority of D&A scientists reacting in a very negative and even personal manner. Some suggested the proposed alternatives amount to a weakening of standards, or an abandonment of scientific method. Here, we address the question: Why is this such a controversial proposition? We argue that there is no “right” or “wrong” approach to D&A in any absolute sense, but rather that in different contexts, society may have a greater or lesser concern with errors of a particular type. How we view the relative risk of overestimation versus underestimation of harm is context‐dependent.

Tropical Cyclones and Climate Change Assessment: Part I: Detection and Attribution

“…Confidence in anthropogenic climate change detection depends on a number of factors: confidence in the estimates of anthropogenic signal and natural variability contributions, whether the observed series being considered is free from significant spurious trends due to data inhomogeneity, and whether the observed time series spans a long-enough period to reliably distinguish an anthropogenic signal from natural variability. Lloyd and Oreskes (2018) discuss the issue of type I and type II errors for detection and attribution. Here, we interpret a type I attribution error as concluding that anthropogenic forcing had contributed nontrivially in a certain direction to an observed change when it had not done so, while a type II error means not concluding that anthropogenic forcing had contributed to some observed change or event in a certain direction when it had done so to a nontrivial extent.…”

“…As discussed by Lloyd and Oreskes (2018), whether a type I or type II error is more important to avoid is context and audience dependent. If the goal is to advance scientific understanding, an emphasis on avoiding type I errors seems logical.…”

Resolving Future Arctic/Midlatitude Weather Connections

“…Thus, part of the controversy in studying Arctic/midlatitude weather linkages stems from the inherent complexity and intermittency in underlying atmospheric physics (Lloyd & Oreskes, 2018). Models point both to the difficulty in resolving linkage physics relative to large-scale atmospheric dynamics, and also demonstrate the inherent variability of such events.…”

“…Recent warm Arctic events and cold midlatitude events show a large role of internal variability and interannual differences in atmospheric jet stream behavior. Thus, part of the controversy in studying Arctic/midlatitude weather linkages stems from the inherent complexity and intermittency in underlying atmospheric physics (Lloyd & Oreskes, 2018).…”

Science Policy Training for a New Physician Leader: Description and Framework of a Novel Climate and Health Science Policy Fellowship

“…However, the health care costs and needs of those impacted by climate‐related events are rarely factored into health policy decisions. Attribution of mortality and morbidity from climate‐related events is hindered by a lack of aggregated health data in many regions of the United States (and globally) as well as the challenge in assigning causality of a specific environmental event to climate change . Additionally, there is a dearth of interdisciplinary collaboration and scientific investigation into the direct impacts of climate change on health care cost and utilization, and analyses are almost always retrospective in nature.…”