This study investigated the peak shaving and pollutant removal performance of an underground stormwater detention facility. The facility was located beneath the parking lot of a retail shopping mall in Prince William County, Virginia. The study methodology incorporated automatic inflow and outflow monitoring stations containing flow metering and automatic sampling equipment. The flow metering equipment continuously monitored flow rates, thereby making it possible to describe the inflow and outflow hydrographs. The automatic sampling equipment collected equal volume - variable time composite samples of runoff, making it possible to determine the inflow and outflow event mean concentrations of various pollutant constituents. Twenty runoff events were monitored between April, 1993 and January, 1994. The inflow and outflow hydrographs were compared to determine peak shaving performance. Due to the large volume of groundwater infiltration into the facility, which altered the inflow and outflow behavior, only 15 events were used in the peak shaving analysis. The inflow peak discharges for the 15 events were substantially less than the two year design storm used for the facility. To determine peak shaving performance at higher flow rates, design storms with greater peak flow rates were routed though the facility using the inventory method. The resultant peak shaving performance was similar to performance estimates in the facility design specifications. In fact, the facility actually reduced peak flows somewhat more than anticipated due to an outlet orifice that was constructed smaller than the design specified. To determine pollutant removal performance at the facility, inflow and outflow event mean concentrations for the various pollutant constituents were compared. A subset of 10 events was used in this analysis because of sampling equipment failures that resulted in insufficient distribution of sample aliquats over a runoff event. Seven different methods of quantification were used to estimate pollutant removal performance. Results indicated that, generally, the facility design and operational protocol provided little or no pollutant removal for most constituents monitored. The trace metals lead and copper were removed more consistently than any other constituents. Zinc was removed somewhat less than either lead or copper, probably because the storage pipe, being made of galvanized steel, was a source of zinc. / Master of Science