A new box model is proposed to describe the dynamic trend of the spatially averaged concentrations of pollutants over a large urban area such as metropolitan Seoul. Being averaged temporally and spatially over a thresh-hold scales, the dynamics of the pollutant concentration becomes simple enough that the governing equation can be expressed in an explicit algebraic form as a function of several meteorological factors and the pollutant emission rate. The single most important meteorological factor is the wind speed dominating the daily variations of the pollutant concentrations. Given the meteorological data from the surface station in the metropolitan Seoul, the model concentration shows excellent agreement with observations from January 1, 1990 to December 31, 2000: the modeling uncertainty, for example, of $NO_2$ concentrations, defined as mean differences between the model concentrations and observations is $16\%$ of the model concentrations. Even for $PM_{10}$ of which secondary sources are considered to be very important and simple box model is irrelevant to, the model performance turns out good, modeling uncertainty being about $32\%$.