Tuesday, Apr 11, 2017 - 3:30pm - Allen 14
A statistical analysis of snow depth trends in North America
Dr. Jon Woody, Statistics, Msstate
Title: A statistical analysis of snow depth trends in North America
Abstract: Several attempts to assess regional snow depth trends across various portions of North America have been made. Previous studies estimated trends by applying various statistical approaches to snow depth data, snow fall data, or their climatological proxies such as snow water equivalents. In most of these studies, inhomogeneities (changepoints) have not been taken into account on a region-wide basis.
This talk begins with considerations of how changepoints effect statistical inference in environmental data, with particular consideration applied towards snow observations. A detailed statistical methodology to estimate trends in daily snow depths from a given data set that accounts for changepoints is considered. Changepoint times are estimated by applying a genetic algorithm to a minimum description length penalized likelihood score. A storage model balance equation with periodic features that allows for changepoints is used to extract standard errors of the estimated trends. The methods are demonstrated on a scientifically accepted gridded data set covering parts of United States and Canada. Results indicate that over half of the grid cells are estimated to contain at least one changepoint and that the average daily snow depth is increasing without changepoints and decreasing with changepoints included in the model.