We present here a study about the limitations found when trying to develop an accurate atmospheric particulate matter forecasting model based on real data, and evidence that the time series of fine particulate matter concentration exhibit deterministic chaotic behavior. We have calculated the Lyapunov exponents of PM2.5 time series obtained from measurements from four monitoring stations located in the city of Santiago, Chile, in recent years. Values obtained for the largest Lyapunov exponents turned out to be positive and ranging between 0.3 and 0.5 which, according to the theory of chaos, is a condition for the presence of deterministic chaos and random behavior in time series. Given the shape of decay of autocorrelation functions and values of correlation dimension and Hurst exponents, random behavior can be discarded: we therefore conclude that the series are chaotic and very sensitive to initial conditions. The study presented here can be replicated in other mid-sized cities that present similar situations to the city of Santiago, where complexity of topography, meteorology and seasonal trends favor the generation of high concentration episodes of atmospheric particulate matter and where a reliable air quality forecasting model may be important for environmental management.