The widely used Weather Research and Forecasting (WRF) model provides various options for the land surface models (LSM) to compute turbulent fluxes of heat and moisture. The LSMs differ in the physical description of soil and vegetation processes and in the input data depending on the soil texture and vegetation classifications employed. In order to evaluate their role in the air quality simulations in Paris area, we performed WRF simulations for four of the available WRF LSMs. We focused on the summer 2009 and the winter 2008/2009. The model simulation results were compared with the SIRTA atmospheric observatory and Meteo-France network measurements for meteorological variables and the Air Parif network for pollutant concentrations. The results show that while modeling uncertainty due to the LSMs can prevent accurate pollution threshold forecasts, one can still identify, in the limit of this study conditions, the most suitable LSM for PM and summer ozone maximum forecasting in Paris area.