A Sustainable Building Retrofit Decision-Making Framework Under Uncertainty

Thesis Abstract

This study addresses the pressing challenge of achieving sustainable building performance through retrofit decisions under uncertainty, recognizing that occupant behavior, energy prices, material availability, and regulatory changes introduce significant risk to retrofit outcomes. The aim is to develop and validate a decision-making framework that integrates multi-criteria analytics, probabilistic risk assessment, and adaptive governance to guide retrofit choices that optimize life-cycle environmental and economic performance while maintaining occupant comfort. Specific objectives include (i) to identify and quantify key drivers of uncertainty in building retrofit projects; (ii) to develop a framework that combines stochastic optimization, Bayesian updating, and scenario analysis to support robust retrofit selections; (iii) to evaluate the framework on a representative sample of existing commercial office buildings; (iv) to assess the framework’s decision quality relative to conventional deterministic approaches; and (v) to derive actionable guidelines for practitioners and policymakers. The methodology adopts a mixed-methods, multi-site case-study design. The population comprises commercially leased office buildings in a metropolitan region with retrofit activity in the last decade. A stratified sample of 60 buildings is selected to capture variations in age, envelope performance, HVAC systems, and occupancy patterns. Data collection employs (a) archival performance data (energy use, retrofit costs, and maintenance records), (b) semi-structured interviews with facility managers (n=20) and retrofitting contractors (n=12), (c) sensor-based energy monitoring over a 12-month period for a subset of buildings (n=15) to capture operational variability, and (d) policy and market data on energy prices, incentives, and material supply chains. Instruments include validated energy-use dashboards, structured interview guides aligned with the framework constructs, and a pilot-tested data fusion template. Validity and reliability are ensured through triangulation, inter-rater reliability checks for qualitative coding (Cohen’s kappa > 0.7), and test-retest reliability for energy metrics. The analytical approach integrates (i) Bayesian hierarchical modeling to update probability distributions of uncertain parameters (e.g., energy savings, retrofit costs, and material lead times) as new information becomes available; (ii) stochastic multi-criteria decision analysis (MCDA) employing the PROMETHEE and VIKOR methods to rank retrofit options under uncertainty; (iii) stochastic optimization (two-stage and robust optimization) to derive retrofit strategies that minimize life-cycle cost (LCC) and carbon emissions subject to performance constraints; (iv) scenario analysis to explore regulatory, price, and climate-related stressors; and (v) sensitivity analysis to identify influential parameters. A hybrid analytical framework is developed in Python and R, with results validated against observed operational performance from the 12-month monitoring period. Key expected findings include (i) a quantified uncertainty profile for retrofit decision factors, (ii) a set of retrofit options ranked by robustness and sustainability performance across scenarios, (iii) demonstrable improvements in expected net present value and CO2e reductions when using the framework versus traditional deterministic methods, and (iv) practical decision guidance on trade-offs between capital cost, energy performance, and adaptability to future policy changes. The study anticipates that integrating Bayesian updating with stochastic MCDA will reduce target deviations between projected and actual retrofit outcomes by 15–25% and improve decision acceptance among facilities managers by enhancing transparency and traceability. Contributions to knowledge include (i) a novel, integrative decision-making framework for sustainable building retrofits under deep uncertainty; (ii) an empirically grounded calibration of probabilistic models for energy savings, costs, and material supply risk specific to retrofit contexts; (iii) methodological advancement by combining Bayesian inference, robust optimization, and MCDA in a single decision-support tool; and (iv) transferable guidelines for practitioners and policymakers that link technical feasibility with financial viability and regulatory foresight. The study concludes that retrofit decisions grounded in probabilistic, multi-criteria, and scenario-aware reasoning yield more resilient outcomes, enabling performance targets to be met with lower risk and greater stakeholder confidence. Recommendations include adopting the framework as an ongoing governance mechanism within retrofit programs, investing in standardized data protocols for building performance and supply-chain data, and aligning incentives to reward robust, long-term sustainability rather than upfront cost minimization.