A class of nonseparable and nonstationary spatial temporal covariance functions

Area of Science:

• Environmental Science
• Statistics
• Geospatial Analysis

Background:

• Standard spectral and geostatistical methods often assume stationarity and separability in covariance functions, which may not reflect real-world spatial-temporal processes.
• Unrealistic assumptions can limit the accuracy and applicability of models for complex environmental phenomena.

Purpose of the Study:

• To develop a general and flexible parametric class of spatial-temporal covariance models.
• To overcome limitations of stationarity and separability in existing models.
• To introduce a unique parameter for quantifying spatial-temporal interaction.

Main Methods:

• Utilizing a spectral representation of spatial-temporal processes.
• Developing a new parametric class of covariance models.
• Incorporating a unique parameter to control spatial-temporal interaction strength.

Main Results:

• Introduced a novel class of spatial-temporal covariance models.
• The new models allow for non-stationarity and non-separability.
• The separable covariance model is a special case within this new framework.
• A unique parameter quantifies the interaction between spatial and temporal components.

Conclusions:

• The proposed covariance models offer enhanced flexibility for analyzing spatial-temporal data.
• These models provide a more realistic approach to understanding complex environmental processes.
• Demonstrated applicability using ambient ozone air pollution data from the U.S. EPA.