Using Covariates to Model Dependence in Non-Stationary, High-Frequency Meteorological Processes

• Activity Number: 87
• Type: Invited
• Date/Time: Sunday, August 3, 2014 : 8:30 PM to 10:30 PM
• Sponsor: Section on Statistics and the Environment
• Author(s): Andrew Poppick*+ and Michael Stein
• Companies: University of Chicago and University of Chicago
• Keywords: high-frequency meteorology ; nonstationary process ; bivariate process ; temperature ; dew point ; spectral domain

Abstract

Meteorological processes that are measured at high temporal frequencies require non-standard statistical models to adequately characterize their observed behavior. We examine this problem through the specific example of a nonstationary, bivariate process -- the high-frequency changes in surface temperature and dew point -- conditionally on the average hourly level of relative humidity, magnitude of minute-to-minute changes in wind direction, and presence of sunlight. The data analyzed are from early May from the year 2003 through 2012 at the Atmospheric Radiation Measurement (ARM) Program's Southern Great Plains (SGP) site in Northern Oklahoma, at the central facility near Lamont, OK. Our model gives a parametric description of how the spectral matrix of the process varies with covariates over blocks of time. The spectral approach allows for convenient and interpretable models of bivariate processes in time and our model captures many of the observed changes in the behavior of the process.