Metro systems play a critical role in contemporary urban development. With proper planning and deployment, metro systems can often lead to enormous socioeconomic and environmental benefits. As such, a bulk of studies have been conducted to investigate metro usage determinates for more efficiently harvesting its benefits. However, past studies are found with three inherent drawbacks, including incapability of capturing multiscale features of different covariates among commonly adopted models, overlooking socioeconomic ingredients (e.g., housing prices and fine-grained population density), and neglecting temporal variations, thus likely leading to biased findings. This study aims to understand metro usage determinants in four divided time slots, i.e., morning peak, off-peak (day), evening peak, and off-peak (night) in Shenzhen, China by applying multiple datasets into a novel model of multiple geographically weighted regression (MGWR). The findings indicate that (1) the built environment factors, socioeconomic elements, and intermodal connection factors influence metro usage in different manners; (2) compared with the other four traditional models, MGWR demonstrates the best performance in explaining metro usage determinants; (3) different important variables affecting metro ridership are identified with the Random Forest technique for all the four time periods in Shenzhen and (4) the correlation between explanatory variables and metro usage is found varying with different spatial patterns. This study is not only of methodological value to firstly apply MGWR to the domain of metro usage but also of practical value to inform decision-makers on how to better promote and predict metro travel demand and ultimately achieve low-carbon and sustainable urban development visions.