A Bayesian Sparse Generalized Linear Model With an Application to Multiscale Covariate Discovery for Observed Rainfall Extremes Over the United States

Abstract

Predictive insights on extreme and rare events are important across multiple disciplines ranging from hydrology, climate, and remote sensing to finance and security. Characterizing the dependence of extremes on covariates can help in identification of plausible causal drivers and may even inform predictive modeling. However, despite progress in the incorporation of covariates in the statistical theory of extremes and in sparse covariate discovery algorithms, progress has been limited for high-dimensional data where the number of covariates is large. In this paper, we propose a general-purpose sparse Bayesian framework for covariate discovery based on a Poisson description of extremes frequency and a hierarchical Bayesian description of a sparse regression model. We obtain posteriors over regression coefficients, which indicate dependence of extremes on the corresponding covariates, using a variational Bayes approximation. Experiments with synthetic data demonstrate the ability of the approach to accurately characterize dependence structures. The method is applied to discover the covariates affecting the frequency of precipitation extremes obtained from station-level observations over nine climatologically homogeneous regions within the continental U.S. The candidate covariates at multiple spatial scales represent station-level as well as regional and seasonal atmospheric condition, indices that attempt to capture large-scale ocean-based climate oscillators and hence natural climate variability, as well as global warming. Our results confirm the dependence structures that may be expected from known precipitation physics and generate novel insights, which can inform physical understanding and perhaps even predictive modeling.Index Terms-Bayesian model, climate change, covariate discovery, sparse regression, statistical downscaling.

ML-based regionalization of climate variables to forecast seasonal precipitation for water resources management

“…Since hydrological variables are mostly in the range of [0, ∞], gamma distribution has been popularly used in the literature (Segond et al 2006, Ambrosino et al 2011, Das et al 2015, Rashid et al 2017. The generalized linear models (GLMs) provide a flexible framework including different marginal distribution such as gamma and exponential family, so this model was applied in the current study for regression.…”

Statistical downscaling of global climate models with image super-resolution and uncertainty quantification

“…Rather than compressing all the features, some of which may not have relevant dependencies with the predictor, one can select a subset of these features using dependency measures. A few methods for features selection include correlation analysis [122], sparse regression [110,43], and Bayesian models [24]. A combination of feature selection and transformation can also be applied.…”

Evaluation of surface runoff from CMIP6 model projections and their implications on population at risk in major river basins of the world

“…They explored how using information outside the typical ENSO region integrated with AI helps in prediction of hydrology. Another study [44] used observation of precipitation extreme for estimating the dependence of extremes on covariates which helps in identifying the causal drivers and ultimately inform predictive modeling. So for future works, along with understanding the physics and biogeochemistry for the earth system models, we can quantify informed risk and improve our predictive understanding by integrating Artificial intelligence with earth system model projections [38].…”