Novel data assimilation-assisted calibrations on high-rise multizone airflow analysis by ensemble Kalman filter and Sobol sensitivity method

Accurate airflow modeling is essential for understanding airflow patterns and for designing effective ventilation systems. However, calibration of airflow models remains challenging due to complex inter-zonal interactions, limited availability of airflow measurements, and uncertainty in input parameters. Additionally, few studies have focused on multizone airflow calibration. In this study, a novel calibration framework is proposed that combines global sensitivity analysis (GSA) using the Sobol method with the Ensemble Kalman Filter (EnKF) for multizone airflow modeling by CONTAM. A Python tool is developed to perform Sobol analysis, and a Java interface is applied to automate the coupling between simulation and calibration workflows. The framework is applied to a 16-story institutional building where CO₂ tracer gas experiments were conducted to characterize both inter-zone and inter-floor airflow behavior. A detailed CONTAM model representing three floors is developed and calibrated using the EnKF approach. GSA is applied to identify the most influential parameters affecting peak CO₂ concentrations, enabling a reduction in calibration variables. Results indicate that in the source room, when the air-conditioning (AC) system is turned off, the initial indoor CO₂ concentration and generation rate account for 94% of the variance in peak CO₂ levels. When the AC is enabled, their contribution decreases to 62%, while the influence of door opening and undercut increases to 25% and 11%, respectively. In adjacent rooms, initial concentration and generation rate also dominate. The EnKF-based calibration substantially improved the prediction accuracy with up to 6% CVRMSE of multizone airflow modeling, achieving reliable agreement with measurements for both room-to-room and floor-to-floor tests.