Tuesday, Aug 6: 2:00 PM - 3:50 PM
1670
Topic-Contributed Paper Session
Oregon Convention Center
Room: CC-G132
Policymakers and politicians have become increasingly aware of the impacts of anthropogenic induced climate change over the past three decades and are taking measures to reduce its progression and impact. This is most evident by various climate agreements, accords, and panels, such as the Paris Climate Accords, the Conference of the Parties, and the Intergovernmental Panel on Climate Change. Increasingly, the meetings surrounding these agreements, accords, and panels focus on methods to limit the rise in mean global temperature and toward understanding how weather and climate events will be altered in our new climate. However, the policymakers and politicians rely heavily on academics to produce new methodology, findings, and actionable results to help meet these goals. Our session focuses on statistical advances in methodology to help make inference on how our climate will respond under new regimes and the impacts of these new regimes on various physical processes.
Applied
Yes
Main Sponsor
Section on Statistics and the Environment
Co Sponsors
Government Statistics Section
Section on Statistical Learning and Data Science
Presentations
Many climatological and environmental dynamic processes exhibit Gaussian behavior. However, for various reasons, it is often the case that some clusters of locations in space and periods of time exhibit lower or upper tail extremes. We propose a bulk and tail spatio-temporal dynamic model (DSTM) that can accommodate such behavior, with either Gaussian or heavy tail behavior varying across the spatial domain and across time. This is accomplished by a regime switching model with stable and Gaussian distributions on the conditional innovation process. The non-stationarity is facilitated by a reduced rank basis function representation and estimated efficiently in a Bayesian inferential paradigm. The model is demonstrated through simulation and application to environmental data.
Studying the impacts of climate on the distribution and composition of fish communities is a recent priority of monitoring programs throughout the Northeast. Climate-mediated shifts in seasonal activities such as spawning have the potential to greatly influence a monitoring program's ability to distinguish changes in abundance or occupancy from shifts in phenology due to climate change. For fish, seasonal shifts in streamflow have the potential to affect growth, reproductive success, recruitment, abundance, and population size structure. In this work we develop a spatio-temporal model to quantify the spatial and temporal change in streamflow across the Northeast. Our approach captures spatial dependence and temporal dynamics through spatially and temporally-varying coefficients. We apply the model to streamflow data collected across 858 gauges in the northeast United States for the years 1965 to 2022. Formal model inference enables the identification of regions experiencing significant changes in seasonal cycles of streamflow. Further, it will inform supplemental sampling designs necessary for monitoring programs to track climate change impacts on fish populations.
Structural equation models (SEM) allow scientists to hypothesize system linkages in multivariate analyses, and coefficients in a sparse "path matrix" are estimated based on the sample covariance among variables. Conveniently, the path matrix yields a sparse precision matrix and SEM can be fitted as a Gaussian Markov random field (GMRF). I first discuss how SEM can be extended using R-package dsem to represent a nonseparable process including simultaneous and lagged interactions among variables and over time, where the joint path matrix is constructed via the sum of separable path matrices across time-lags. To illustrate, I use dsem to represent ecosystem interactions in the eastern Bering Sea, including linkages from ocean physics through phyto- and zooplankton to fishes, seabirds, and seals. I also introduce a spatial extension using R-package tinyVAST that also includes a separable spatial process. To illustrate, I use tinyVAST to estimate associations between flatfishes and rockfishes and corals and sponges in the Gulf of Alaska and Aleutian Islands. Throughout, I emphasize that SEM provides a natural extension for GMRFs to multivariate settings.
The effects of solar climate intervention (SCI) efforts are confounded with climate change and natural variability. The June 1991 Mt. Pinatubo eruption resulted in a massive increase of aerosols (in the form of sulfur dioxide) in the atmosphere, absorbing radiation and thus serving as a natural analog for SAI which presents a chance to develop tools for climate attribution of regionalized sources. Our goal is to characterize the multivariate and dynamic nature of the climate impacts following the Mt Pinatubo eruption. To achieve this goal, we introduce a space-time multivariate model that captures correlations between climate effects following an event. Specifically, spatial variation is modeled using a flexible set of basis functions for which the basis coefficients are allowed to vary in time thru a vector autoregressive (VAR) structure. We show how this novel model can be casted in the Dynamic Linear Model framework and estimated via a customized MCMC approach. Furthermore, we demonstrate how the model characterizes the observed pathways following the Mt. Pinatubo eruption with reanalysis data from MERRA-2.
Determining the probability and severity of a low-likelihood high-impact weather event from the historical record is difficult due to their relatively rare occurrence. Instead, we shift our focus to the drivers of the climatology surrounding weather events. Specifically, we represent the climate system as the sum of two parts, the climatological forcing and internal variability, and model the drivers of these two processes. We model the climatological forcing as changes in the system due to anthropogenic induced climate change using a set of measurable variables. The internal variability represents the variation in the system due to its natural cycle, which we model using Bayesian singular value decomposition where the basis functions in the decomposition capture the spatial and temporal modes of variability. By decomposing the climate system in terms of its climate forcing and internal variability, we can determine which combination of the drivers result high-impact weather events and the probability of these events occurring. We apply our framework to two-meter air temperature in the Pacific Northwest, providing additional insight into the 2021 heatwave.