On the change of support problem for spatio-temporal data.

Area of Science:

• Geospatial statistics
• Environmental science
• Spatio-temporal modeling

Background:

• Spatial data are collected as point-referenced or block data.
• The change of support problem involves inferring values at unobserved locations or supports.
• Existing methods often lack a unified framework for diverse spatial data types.

Purpose of the Study:

• To propose a unifying Bayesian kriging approach for spatial data prediction.
• To enable inference across different data supports: point-to-point, point-to-block, block-to-point, and block-to-block.
• To extend the framework for spatio-temporal data analysis.

Main Methods:

• Developed a fully Bayesian kriging methodology.
• Integrated a unifying approach for change of support problems.
• Incorporated spatio-temporal association for dynamic data analysis.

Main Results:

• Demonstrated a unified prediction framework applicable to various spatial data supports.
• Successfully applied the static spatial model to point-level ozone measurements in Atlanta.
• Illustrated the dynamic spatial case using a temporally extended ozone dataset.

Conclusions:

• The proposed Bayesian kriging approach provides a flexible and unified solution for change of support problems.
• The methodology effectively handles both static spatial and dynamic spatio-temporal data.
• This framework offers improved accuracy and consistency in spatial data prediction across different supports.