This talk has been automatically
captioned. You can remove these by pressing CC on the video toolbar.
Name: Daniel Cooley
Talk Title: Climatic extremes: current statistical challenges
Abstract: Extreme value analysis has a long history of describing the intensity of extreme weather events. Much of this historical work was done under an assumption of a stationary climate. A changing climate introduces challenges for the extremes statistics community. A fundamental question is the projection of future extremes. For example, one might need to estimate the magnitude of a 1-in-100 annual exceedance probability event for a future time, perhaps in 50 or 100 years. Climate projections are typically done via numerical climate models, but any projection estimate would possess not only statistical uncertainty from the data, but also uncertainty from the climate projection itself. Additionally, working with climate model output can introduce a calibration (or downscaling) issue to convert output from the model's spatial resolution to the posed question’s resolution, which often is a point corresponding to a monitoring station. Additionally, there are statistical challenges associated with detection and event attribution of climate change to extreme events. Detection is the process of statistically showing that climate has changed, which is more difficult for extremes than for mean changes. Event attribution is the process of assigning an amount of risk of an observed extreme event to the changed climate. This talk aims to introduce and explore these ideas, as well as to illustrate possible statistical approaches from our own work. In particular, we will look at calibration of future river flow extremes; we will look at an approach of performing climate change detection via a principal component decomposition of extreme precipitation data, and we will look at performing event attribution of recent wildfire weather conditions for the Western US.
This
talk is a plenary lecture at EVA 2021.