QUORUM: A SYSTEMS ENGINEERING FRAMEWORK FOR MIGRATION RISK ASSESSMENT

Abstract

This thesis presents QUORUM (Quantitative Understanding of Regional Outmigration Using Multivariate data), a modular, systems-engineered framework designed to assess development indicators with migration trends in a case example location of Central America. The work addresses a documented gap in research literature that no existing Water-Energy-Food (WEF) nexus model incorporates migration as a thematic focus, leaving policymakers and affected communities without the analytical tools needed to anticipate displacement pressures and plan adaptive responses.This system is architected as three constituent modules. The Information Module handles heterogeneous data ingestion and harmonization. The Analytics Module performs statistical and machine learning analysis. The Design Module supports human-centered visualization and stakeholder communications. The system is developed using the V-Model of systems engineering, with functional requirements decomposed across each of the modules and governed by formal Interface Control Documents specifying data schemas, transformation rules, and acceptance criteria at each module boundary. The Analytics Module was validated using a proof-of-concept implementation that integrates World Bank World Development Indicators with Meta mobility-derived migration data for seven Central American nations over a 2019 to 2022 study period. A four-stage analytical pipeline was applied containing descriptive migration characterization, lagged cross-correlation analysis, fixed-effects panel regression, and Random Forest modeling with Shapley Additive Explanations (SHAP) for feature importance attribution. As a proof-of-concept demonstration, the analytics pipeline produced outputs broadly aligned with theoretical expectations, suggesting that publicly available environmental and socioeconomic indicators carry information relevant to migration pressures. Nine of 45 tested correlations reached statistical significance (p < .05), and SHAP analysis identified that metrics around a nation’s carbon intensity and freshwater availability as the highest-ranked indicators within the exploratory analysis. Fixed-effects panel regression identified Forest Cover (%) as the most robust within-country migration predictor (beta = -11.53, p = .029). The thesis contributes a reproducible, transparent, and grounded analytical architecture that explores how open-source data, when processed through a structured systems engineering pipeline, can surface preliminary quantitative relationships relevant to migration risk indicators with existing and available datasets. Limitations related to panel size (N = 14 country-year observations) and cross-country generalization are discussed, along with future directions for expanding temporal coverage, integrating higher-resolution monthly environmental data, and deploying bespoke Design Module elements for community-facing applications.