We conducted comprehensive sensitivity model simulations to investigate the key factors that could affect the performance of satellite-based PM2.5 exposure in data-poor regions including temporal sampling frequency, number of ground monitors, accuracy of the chemical transport model simulation, and different combinations of the additional predictors.