Bacterial strains (SL1: Rhonococcus rhodochrous; WT1: Acinetobacter junii) capable of biodegrading atrazine and simazine in surface water were inoculated into pilot-scale fixed beds of granular activated carbon (GAG, Im depth, 15 min empty bed contact time, EBCT). River water was screened, clarified, ozonated and spiked with atrazine and simazine (post-ozonation concentration 0.5-15 mu g l(-1) of each s-triazine) before entering GAC columns that were monitored for 302 d. Elution of inoculated isolates from GAC columns peaked 15 min (WT1) and 30 (SL1) min after inoculation, but had fallen to 3.8 +/- 0.13 x 10(3) ml(-1) (WT1) and 3.6 +/- 3.37 x 10(2) ml(-1) (SL1) by 120 min. During routine operation, elution of SL1 was frequently <10 ml(-1) and less than WTI, and from day 0 to day 119 SL1 was retained on GAC in greater numbers (minimum 8.8 +/- 0.32 x 10(4) g(-1) dry wt GAG) than indigenous WTI (1.1 +/- 0.16 x 10(4) g(-1) dry wt GAG) or introduced WTI (9.1 +/- 0.00 x 10(3) g(-1) dry wt GAG). On days 49, 84, 119, 265 and 302 inoculated GAC (SL1 or WTI) had less adsorbed s-triazine per column than non-inoculated GAG. Inoculation of GAC with s-triazine-degrading isolate SL1 reduced transient breakthrough concentrations of atrazine and simazine compared to non-inoculated and WT1-inoculated column effluents on days 21, 84, 140 and 170. However, from day 189 onwards consistent breakthrough occurred in all effluents exceeding the EU maximum admissible concentration (MAC, 0.1 mu g l(-1)).