We studied the persistence of Bacillus thuringiensis serovar israelensis (Bti) in a typical breeding site of the mosquito Ochlerotatus caspius in a particularly sensitive salt marsh ecosystem following two Bti-based larvicidal applications (Vectobac 12AS, 1.95 L/ha). The treated area was composed of four larval biotopes that differed in terms of the most representative plant species (Sarcocornia fruticosa, Bolboschoenus maritimus, Phragmites australis, and Juncus maritimus) and the physical and chemical characteristics of the soil. We sampled water, soil, and plants at various times before and after the applications (from spring to autumn, 2001) and quantified the spores of B. thuringiensis (Bt) and Bacillus species. The B. cereus group accounted for between 0% and 20% of all Bacillus spp. before application depending on the larval biotope. No Bti were found before application. The variation in the quantity of bacilli during the mosquito breeding season depended more on the larval biotope than on the season or the larvicidal application. More bacilli were found in soil (10(4)-10(6) spores/g) than on plant samples (10(2)-10(4) spores/g). The abundance in water (10(5) to 10(7) spores/L) appeared to be correlated to the water level of the breeding site. The number of Bti spores increased just after application, after declining; no spores were detected in soil or water 3 months after application. However, low numbers of Bti spores were present on foliage from three of the four studied plant strata. In conclusion, the larvicidal application has very little impact on Bacillus spp. flora after one breeding season (two applications).