Future modelling should therefore find standard procedures to apply probabilistic approaches. Generally, deterministic models have to incorporate newly gained knowledge by empirical investigations continually, in order to improve the model performances with regard to sorption behaviour, effects of diversely structured vegetation patches, and small-scale geomorphological structures on pesticide transport. For the same reasons, pesticide models not only have to be improved qualitatively, but they also have to gain higher temporal and spatial resolutions, as we demonstrated that some hydrological models overestimated pesticide loads due to low temporal resolution. However, the resolution of pesticide models is additionally limited by the resolution of available input data and by computer hardware, the latter of which is supposed to experience further rapid improvements.
In summary, the modelling of pesticide runoff has experienced great progress since the early stages in the 1970s, but until now there is no ultima ratio that enables to predict pesticide emissions accurately from small to large scales.