Date of Award

Summer 8-2026

Document Type

Dissertation

Degree Name

Doctor of Philosophy (PhD)

Department

Engineering Management & Systems Engineering

Program/Concentration

Engineering Management and Systems Engineering

Committee Director

Hyoshin Park

Committee Member

Eric Larour

Committee Member

Mecit Cetin

Committee Member

Pilar Pazos

Abstract

Many physical and networked systems evolve under continuously changing spatial and temporal conditions. Transportation networks respond to fluctuating demand, atmospheric fields reorganize as storms intensify, and coastal response depends on localized forcing pathways. Modeling such systems requires learning formulations that adapt to evolving structure, operate on irregular geometries, and provide interpretable measures of predictive uncertainty. This dissertation develops a physics-guided spatiotemporal learning framework designed for structured dynamical systems whose governing interactions are neither static nor Euclidean. The central premise is that spatial relationships in these systems are dynamic and geometry-dependent. To represent this behavior, system states are modeled on time-varying graphs whose connectivity encodes localized correlations and directional influence. This dynamic graph formulation enables learning directly on non-Euclidean domains, preserving physically meaningful structure while allowing interactions to evolve with system conditions. Temporal evolution is integrated within this graph representation so that changing spatial organization and sequential dynamics are learned jointly. When subsystems interact through directional forcing–response relationships, localized coupling operators are constructed to reflect physically consistent pathways of influence. Uncertainty is treated as an intrinsic component of the modeling process. Domain-specific probabilistic mechanisms are incorporated to quantify epistemic variability and to assess calibration through residual diagnostics and empirical coverage analysis. The framework is evaluated in multiple structured dynamical contexts, including traffic state estimation, ice-sheet model emulation, and hurricane-driven storm surge prediction. Across these settings, the proposed approach demonstrates stable performance under evolving conditions, spatially coherent field predictions, and uncertainty estimates that scale systematically with dynamical intensity. This work establishes a general methodology for constructing uncertainty-aware predictive digital twins on network-constrained, non-Euclidean domains. By embedding dynamic spatial structure and physically guided coupling within spatiotemporal learning, the dissertation advances the development of adaptable, interpretable forecasting systems for complex environmental and engineered processes.

Rights

In Copyright. URI: http://rightsstatements.org/vocab/InC/1.0/ This Item is protected by copyright and/or related rights. You are free to use this Item in any way that is permitted by the copyright and related rights legislation that applies to your use. For other uses you need to obtain permission from the rights-holder(s).

DOI

10.25777/tvrp-jc04

ISBN

9798193214441

Download

Share

COinS