Characterizing climate pathways using echo state networks and feature importance

Climate change due to human activity is considered a serious threat to our future. Intervention methods that aim to reduce the negative impacts of climate change, such as stratospheric aerosol injections, are being discussed by policy makers. However, the downstream effects of such actions on the climate system are not well understood. The development of algorithmic methods for understanding the impacts of such events could help inform decision makers. Echo state network (ESN) models are a data driven option for modeling climate impacts. ESN are known for modeling nonlinear dynamic systems and provide increased computational efficiency over other statistical methods. However, ESN parameters are not interpretable. We develop a methodology for computing ESN temporal variable importance in the context of spatio-temporal data. The method is used to quantify relationships between climate variables associated with the 1991 volcanic eruption of Mount Pinatubo, which acts as a natural stratospheric aerosol injection. We forecast temperatures given lagged values of other relevant climate variables such as aerosol optical depth, and variable importance is used to characterize the temporally evolving relationships between temperature and the predictor variables. SNL is managed and operated by NTESS under DOE NNSA contract DE-NA0003525.

Speaker bio: Katherine Goode is a statistician at Sandia National Laboratories. She has a PhD in statistics from Iowa State University and a masters in statistics from the University of Wisconsin – Madison. Her research focuses on explainable machine learning, trustworthy in AI, model assessment, and data visualization. Currently, Katherine works on projects in the areas of climate security and cyber security.