Abstract by Daniel Sheanshang
Using a Heteroscdastic Spatially Varying Coefficients Regression Model to Estimate Below-Ground Temperatures
Understanding how snow density changes as a function of depth gives us valuable information about local climate history. To accurately model snow densification and, thus, snow density, we need the temperature below the surface, generally 10-m below. Unfortunately, below-surface temperatures are cannot accurately be estimated using climate models and can only be acquired from field measurement. Climate models are available for 2-m air temperature. Our primary goal is using climate models for 2-m air temperature to estimate temperature 10-m below the ground, accounting for heteroscedastic errors and spatially correlated parameters between measurements throughout the region. To accurately predict below-ground temperatures, we pose a heteroscedastic spatially varying coefficients regression model. Because significant prior information is available and accurate uncertainty quantification in temperature predictions, we fit using a Bayesian framework. This model will then be incorporated into our model for snow density.