Routine weekday forecasts of fine particulate matter (PM_2.5) on the metropolitan scale began in the mid-Atlantic region in October 2003. A large number of metropolitan areas provided forecasts beginning at that time (see, http://www.epa.gov/airnow) and this paper reports preliminary results for the Philadelphia metropolitan area, including southern New Jersey and northern Delaware, for the period October 2003 to August 2004. Fundamental forecast questions are posed by recent efforts to include PM_2.5 forecasts as part of a standard suite of air quality forecast products. Because there is currently no operational PM_2.5 numerical forecast model, objective forecast guidance is limited to persistence coupled with statistical forecast methods. In this case, Classification and Regression Tree (CART) techniques were utilized. The skill of statistical methods are limited by a sparse historical database. The current network of Federal Reference Method (FRM) PM_2.5 monitors was initialized in 1999-2000 and is overwhelmingly composed of observations made every third day. Even persistence methods are limited by the difference, particularly during the winter season, between continuous monitoring methods, available in near-real time, and the FRM monitors which require extensive analysis and are reported after a lag of several months. Methods used by air quality forecasters to adapt to this challenging forecast environment are reported here along with an initial benchmark measure of forecast skill.