This study characterizes the indoor airflow and occupants’ thermal sensations in a cross-ventilated building model sheltered by generic cube arrays based on large-eddy simulations (LESs). Four ventilation models, which comprise different cross-ventilating openings, streamwise (STR) and lateral (LAT) windows, and block arrangements, lattice-type square (SQ) and staggered (ST) patterns, were examined to understand the following geometry-oriented features: i) the temporal and spatial deviations of wind speed at openings and inside the ventilation models, ii) effects of time and space resolutions for the velocity data on the estimation accuracy of the ventilation rate, and iii) predicted mean vote (PMV) and predicted percentage of dissatisfied (PPD) indices calculated with elaborately simulated velocity data. The difference in distribution of fluctuating normal velocity at openings was more significant when varying the conditions of the opening locations than that observed when varying the building arrangements. Therefore, the ventilation rates in the STR conditions were reasonably estimated using only the time-averaged flow rate at the center position of the windward opening; meanwhile, when the contributions of reverse flow were ignored at the openings, the ventilation rates in the LAT conditions were drastically underestimated using highly resolved velocity data at openings. Based on the thermal comfort assessment at an air temperature of 26°C, the discrepancies of area-averaged PMV values between STR and LAT cases were within 0.7 and 0.9 at the lower and middle heights of naturally ventilated buildings, resulting in a 5% difference in the PPD values.
All Science Journal Classification (ASJC) codes