Two Hyphomicrobium VS inoculation protocols were compared for start-up of a biotrickling filter removing dimethyl sulfide (DMS). One biotrickling filter (HBF 1) was filled with rings that were submerged in a nutrient medium containing hyphomicrobium VS fed with DMS, another biotrickling filter (HBF 2) was similarly filled with rings that were submerged in nutrient medium, but continuously supplied with actively growing Hyphomicrobium VS and fed with methanol. Initially, about 40 times more Hyphomicrobium VS cells were attached to the rings in HBF 2. During the experiment, two to three times more Hyphomicrobium VS cells were still found to be present on the rings in HBF 2 compared to HBF 1. The maximal DMS elimination capacity at 90% removal efficiency of HBF 1 was 7.2 g m(-3) h(-1) (200 ppmv inlet concentration). operation. The elimination capacity decreased, however, when the inlet loading rate exceeded 15 a m(-3) The performance of HBF 2 was much better, with an elimination capacity of 8.3 g m(-3) h(-1) (90% removal efficiency) after 2 days of operation, increasing to a maximum of 57 g m(-3) h(-1) at 92% removal efficiency. Microbial community analysis with denaturing gradient gel electrophoresis showed very different microbial communities in both biotrickling filters (Pearson correlation coefficient of 0%). Moreover, the decreased DMS elimination capacity of HBF 1 at higher influent loading rate corresponded with a drastic change of the m microbial community on the rings. The latter observations suggest that the functional efficiency of the microbial community in a biotrickling filter can be related to its composition. (c) 2005 Elsevier B.V. All rights reserved.