Abstract
Standard building performance simulation tools, including EnergyPlus, use ASHRAE 90.1 residential occupancy schedules that assume nighttime window-opening rates of 50–60%. These rates were measured in low-security, temperate Western housing and have never been validated for urban public housing in Sub-Saharan Africa. In security-constrained multi-story housing, where documented nighttime security risks make window closure a near-universal behavior, this assumption generates a large, systematic, and uncorrected simulation error that distorts every retrofit investment informed by it. This study defines, quantifies, and corrects the Security-Closure Thermal Penalty: the systematic overestimation of the cooling benefit from natural ventilation when applying security-neutral occupancy schedules to security-constrained residential buildings. Security-Closure Thermal Penalty is the hot-climate, Global South counterpart of the European prebound effect. Where income constraints suppress heating below cold-climate model predictions, security constraints suppress ventilation below hot-climate model predictions. No prior study has named, measured, or provided a validated correction instrument for this error class. A Longitudinal Structured Digital Diary protocol was deployed in n=48 stratified households within Addis Ababa’s Condominium Housing Program (CHP), yielding N=2,688 twice-daily observations across hot (March–April 2024) and cool (July–August 2024) seasons. Households were stratified by typology (G+4; G+10), façade orientation (N/S/E/W), and floor level (low/mid/high). EnergyPlus v9.6 models were calibrated through a validated four-stage protocol; at Stage 3, ASHRAE default schedules were replaced with LSDD-derived probabilistic window-operation inputs formatted directly for EnergyPlus IDF integration. Multilevel modelling (MLM; n=2,688 observations; n=48 households; 5,000 bootstrap replications) and the novel Thermal Trap Index, the proportion of diary entries in which TSV ≥+2 and no adaptive action was available, completed the analytical framework. Observed night-time window opening rates were 8–12% (95% CI: 7–13%), against the 50–60% ASHRAE default, a 44 percentage-point structural divergence that was statistically identical across both seasons (χ² (1) =1.8, p=0.18). This divergence results in a mean peak indoor temperature penalty of 1.8–2.6°C and a 25–38% overestimation of the ventilation cooling benefit in the standard simulation. Security behaviour explains 41% of the variance between predicted and observed thermal outcomes (95% CI: 37–45%). Replacing the ASHRAE default schedule with LSDD-derived inputs reduced Mean Bias Error from -32% (SD=4.2%) to -7.8%, a 24.2 percentage-point improvement that meets ASHRAE Guideline 14 accuracy thresholds achieved through schedule correction alone, with no changes to physical building parameters. The Thermal Trap Index revealed a structural equity gradient: the lowest-control households had TTI=0.41, meaning 41% of their discomfort hours lacked an available adaptive response, a threefold burden relative to the highest-control group (TTI=0.14). Security context is not a qualitative annotation to BPS workflows; it is a mandatory boundary condition. The LSDD protocol operable at USD 0–15/month via standard SMS infrastructure and producing outputs directly compatible with EnergyPlus IDF files provides a generalizable, low-cost calibration pathway for data-scarce urban environments. A documented three