PhD Student: Kazi Shahrukh Omar

Committee Members:
Dr. Fabio Miranda, Chair and Advisor
Dr. Michael Papka
Dr. Khairi Reda
Dr. Saeed Boorboor
Dr. Lane Harrison, Worcester Polytechnic Institute

Date/Time: Tuesday, February 10, 2026, 4 PM
Location: CDRLC 1219

Zoom Details:
https://uic.zoom.us/j/9284268608
Meeting ID: 928 426 8608

Abstract:
As cities continue to grow rapidly, urban environments are producing increasingly large, diverse, and interconnected datasets. While this abundance of data creates new opportunities for urban analytics, it also imposes significant computational challenges when analyzing large-scale urban phenomena. For practical use, computational models should efficiently process heterogeneous datasets and produce faster model outputs. A representative example is accumulated shadow or sunlight access modeling, which requires computation over dense spatial structures and lengthy temporal scales, posing significant challenges for efficient execution at large scales, such as city-level analysis. Another key challenge lies in how these urban data and computational models are presented to stakeholders for effective decision-making. Beyond understanding what is happening, stakeholders must be able to explore what to do next, which calls for scenario-based representations through visual analytics (VA) that structure urban data and models around alternative design and planning outcomes.

However, the current landscape of urban visual analytics (VA) tools remains largely centered on exploratory data analysis, emphasizing descriptive questions about existing urban conditions. In practice, these tools often take the form of domain-specific dashboards focused on exploratory analysis of static data through visualizations, rather than generating or comparing alternative scenarios. As a result, they offer limited support for the forward-looking reasoning required in urban decision-making, where stakeholders must compare alternative design choices and assess their potential outcomes. Although more recent modular urban VA toolkits improve flexibility in composing data and visualizations, they either lack explicit support for scenario-oriented workflows or require substantial expertise to assemble them. Addressing this gap calls for a shift from an exploration-centric VA to a scenario-oriented VA that explicitly supports reasoning over alternative urban futures.

A scenario-oriented modular VA stack poses several key challenges across urban data and modeling pipeline. These challenges include: (C1) how to appropriately model complex urban phenomena, (C2) how to enable domain experts to externalize scenarios, (C3) how to support the comparison and analysis of alternative scenarios, and (C4) how to facilitate iterative exploration of counterfactual inputs that would lead to a scenario. This dissertation addresses these challenges by (C1) introducing Deep Umbra, an efficient computational framework for large-scale accumulated shadow modeling; (C2–C3) presenting SCOUT, a modular, scenario-oriented urban visual analytics toolkit that enables domain experts to externalize, analyze, and compare alternative urban scenarios across domains; and (C4) extending SCOUT with ongoing work on CF-Planner, that supports interactive exploration of counterfactual inputs that leads to a scenario outcome.