|
Activity Number:
|
422
|
|
Type:
|
Contributed
|
|
Date/Time:
|
Wednesday, August 5, 2009 : 8:30 AM to 10:20 AM
|
|
Sponsor:
|
Section on Statistics and the Environment
|
| Abstract - #305489 |
|
Title:
|
Comparison of Deterministic and Stochastic Spatiotemporal Interpolation Methods for Ozone in the Conterminous US
|
|
Author(s):
|
Lixin Li*+ and Xingyou Zhang and Reinhard Piltner
|
|
Companies:
|
Georgia Southern University and CDC and Georgia Southern University
|
|
Address:
|
Department of Computer Sciences, Statesboro, GA, 30460-7997,
|
|
Keywords:
|
spatiotemporal interpolation ; shape functions ; kriging ; cross-validation
|
|
Abstract:
|
This paper discusses and compares two spatiotemporal interpolation methods for a set of data with annual ozone concentration measurements in the conterminous United States: i) 3D shape function-based spatiotemporal interpolation method - a deterministic method, and ii) 3D kriging-based spatiotemporal interpolation method - a stochastic method. The comparison results show that the kriging-based method yields a better MAPE (Mean Absolute Percentage Error) value than the shape function-based method based on the leave-one-out cross-validation, especially in the beginning and ending years. However, the 3D shape function-based method has other advantages over the kriging-based method such as its invariance to coordinate scales and efficiency.
|
