IMAGe Brown Bag - Geostatistical Approach to Modeling Teleconnections

09/30/2016 - 12:00pm to 1:00pm

Josh Hewitt
Colorado State University

We present a geostatistical method for studying teleconnections at regional scales by extending hierarchical Bayesian spatial models to incorporate predictor variables that are spatially remote from response variables. Our method uses remote large scale predictors, such as sea surface temperatures (SSTs), to predict local variables, like precipitation. The method uses spatially varying coefficients so that teleconnection effects are localized, allowing the model to capture the varying impacts that large scale predictors have on different regions. Hierarchical Bayesian frameworks also facilitate testing the significance of teleconnection effects and for quantifying forecast uncertainties. We present encouraging preliminary results of using this method with seasonal sea surface temperatures to study regional teleconnections and to predict seasonal precipitation in Colorado and Oklahoma. This is joint work with Jennifer A. Hoeting (CSU), James Done (NCAR), and Erin Towler (NCAR).

Friday, September 30, 2016
12:00-1:00 pm
Mesa Lab, Chapman Room
(Bring your lunch)