Online Program Home
  My Program

Abstract Details

Activity Number: 22 - Statistical Methods for Heterogeneous and Massive Remote Sensing Data
Type: Topic Contributed
Date/Time: Sunday, July 30, 2017 : 2:00 PM to 3:50 PM
Sponsor: Section on Statistics and the Environment
Abstract #323314 View Presentation
Title: Evaluation of Climate Models Using the Wild Scale-Enhanced Bootstrap
Author(s): Megan Heyman* and Ansu Chatterjee and Amy Braverman
Companies: Rose-Hulman Institute of Technology and University of Minnesota and Jet Propulsion Laboratory
Keywords: bootstrap ; wavelets ; time series
Abstract:

Understanding earth's changing climate is vital to informed decision-making. Thus, numerous physical models exist which predict climate at various resolutions across the globe. With so many models, it is of interest to find which models produce "accurate" climate predictions. An added complication associated with these physical models is that the predictions may change based upon the initial conditions assumed. Remote sensing instruments provide global measurements of quantities associated with climate, such as temperature, specific humidity, and precipitation, at a high resolution. Although these measurements are variable due to discrepancies in atmospheric composition, they comprise a dataset that may be considered as observed climate. We use the Wild Scale-Enhanced (WiSE) bootstrap to evaluate existing physical model predictions with respect to the observational remote sensing data. This method allows for a comparison of two time series (i.e. within a grid-cell, the observational data and a physical model) where each series has intrinsic heteroscedastic variability. The developing methodology also addresses spatial relationships between grid cells.


Authors who are presenting talks have a * after their name.

Back to the full JSM 2017 program

 
 
Copyright © American Statistical Association