We present a very high-resolution (400 m) operational air quality forecasting system developed to alert citizens of Delhi and the National Capital Region (NCR) about acute air pollution episodes. Such a high-resolution system has been developed for the first time and is evaluated during October 2019-February 2020. The system assimilates near real time aerosol observations from in situ and space-borne observations in the WRF-Chem model to produce a 72-h forecast every day in a dynamical downscaling framework. The assimilation of aerosol optical depth and surface PM 2.5 observations improves the initial condition for surface PM 2.5 by about 45 µg/m 3 (about 50%). The accuracy of the forecast degrades slightly with time as mean bias increases from +2.5 µg/m 3 on the first day to-17 µg/m 3 on the third day of forecast. Our forecasts are found to be very capable both for PM 2.5 concentration and unhealthy/ very unhealthy air quality indices categories. 2