Water body and riparian buffer strip characteristics in a vineyard area to support aquatic pesticide exposure assessment Original Research Article

The implementation of a geodata-based probabilistic pesticide exposure assessment for surface waters in Germany offers the opportunity to base the exposure estimation on more differentiated assumptions including detailed landscape characteristics. Since these characteristics can only be estimated using field surveys, water body width and depth, hydrology, riparian buffer strip width, ground vegetation cover, existence of concentrated flow paths, and riparian vegetation were characterised at 104 water body segments in the vineyard region Palatinate (south-west Germany). Water body segments classified as permanent (n = 43) had median values of water body width and depth of 0.9 m and 0.06 m, respectively, and the determined median width:depth ratio was 15. Thus, the deterministic water body model (width = 1 m; depth = 0.3 m) assumed in regulatory exposure assessment seems unsuitable for small water bodies in the study area. Only 25% of investigated buffer strips had a dense vegetation cover (> 70%) and allow a laminar sheet flow as required to include them as an effective pesticide runoff reduction landscape characteristic. At 77 buffer strips, bordering field paths and erosion rills leading into the water body were present, concentrating pesticide runoff and consequently decreasing buffer strip efficiency. The vegetation type shrubbery (height > 1.5 m) was present at 57 (29%) investigated riparian buffer strips. According to their median optical vegetation density of 75%, shrubberies may provide a spray drift reduction of 72 ± 29%. Implementing detailed knowledge in an overall assessment revealed that exposure via drift might be 2.4 and via runoff up to 1.6 fold higher than assumed by the deterministic approach. Furthermore, considering vegetated buffer strips only by their width leads to an underestimation of exposure by a factor of as much as four. Our data highlight that the deterministic model assumptions neither represent worst-case nor median values and therefore cannot simply be adopted in a probabilistic approach.