Predictive modelling of multi-domain factors on window, door, and fan status in naturally ventilated school classrooms

Most studies regarding the investigation of occupant behaviour (OB) in school classrooms addressed the environmental influence on window operation solely and were conducted in oceanic climates. This study aimed to identify and quantify the influence of multi-domain factors (including thermal, indoor air quality, contextual and multi-behaviour domains) on window, door, and fan status in naturally ventilated school classrooms in a humid subtropical climate, in order to predict OB. Environmental variables, manual operation of windows, doors and fans, and occupancy rate were monitored and questionnaires were applied in a set of classrooms of three public school buildings in the state of Sao Paulo, Brazil, during four rounds at two-month intervals, resulting in a comprehensive year-long study. During part of the physical monitoring, restrictive occupancy measures due to the COVID-19 pandemic were observed. Generalized Linear Mixed Models were applied to assess the influence of the recorded parameters on the window, door, and fan status and to generate OB predictive models. Results showed that indoor environmental variables influenced window, door, and fan status in school classrooms, with few exceptions. Yet, the models including school routines, social norms and teachers' behaviour as predictors led to the highest accuracy. This suggests that, while a more complex model with additional predictors can provide more accurate predictions of OB, it also becomes more context-dependent and less generalizable. The trade-off between model complexity and generalizability is an important consideration in this research study, and it highlights the nuanced relationship between multi-domain factors affecting occupant behaviour in school buildings.