Fate of benzene in a Stratified Lake receiving contaminated groundwater discharges from a Superfund site

Predicting the fate of benzene in aquatic environments, and estimating corr esponding human exposures, is critically dependent on knowledge of this car cinogen's biodegradation rate under the site-specific conditions. We used t hree approaches for quantifying this key fate process: (1) shortterm (hours ) observations of benzene loss in laboratory incubations of representative water samples, (2) whole-lake benzene mass balance studies, and (3) modelin g of the temporal evolution of benzene vertical profiles in the lake. Our f ield site, the Hairs Brook Holding Area (HBHA), continuously receives benze ne input (about 20 muM or 1.5 ppm) into its anoxic hypolimnion via discharg e of saline groundwater from an adjacent Superfund site (IndustriPlex in Ea stern Massachusetts.) Using summertime, lake water samples in the laborator y, we found benzene was degraded in three metalimnion samples at rates betw een 1 and 2.5 d(-1). An epilimnion sample yielded a similar result, but no degradation was observed in another epilimnion sample. Losses were less tha n or equal to0.04 d(-1) in a sulfate-rich hypolimnion sample. Since benzene loss could be inhibited by filtration or with a mixture of poisons and ant ibiotics, it was apparently being biodegraded. In the whole-lake mass balan ce studies of benzene, it was found that approximately 80% of the benzene e ntering the lake was degraded during the water's residence in the lake. Ver tical distributions of benzene in the HBHA water column indicated that the chief sink of benzene was located in the metalimnion. A two-month progressi on of summertime profiles of benzene concentration vs depth was fitted well using a dynamic model, CHEMSEE, and assuming that the only sinks were epil imnetic flushing, water-to-air exchange, and biodegradation in a 0.4 m-thic k metalimnetic layer at 2 d(-1). The biodegradation rate derived from such whole-system study appears more dependable than rates deduced from grab sam ples, and we suggest that we must learn to predict these intact-system rate s.