In the present project AZ - 35919 (duration: 01.01.2021 to 31.03.2022) the pesticide pollution of waters in protected areas in Germany was investigated, which, despite presumably high relevance, has not been considered in this way before. The pesticide load in waters in nature reserves (NSGs) was assessed primarily on the basis of extensive application and monitoring data for the federal state of Saxony. Due to considerable delays in the provision of data, a detailed situation analysis was carried out using Saxony as an example, in contrast to the original project proposal. This showed that the pollution of waters in NSGs by pesticides is lower than that of waters located in areas without a special protection status (NNSGs). Both the detection rate (proportion of samples with pesticide detections to total samples) of pesticides and the detected pesticide concentrations were lower in NSGs than in NNSGs. In contrast, however, the pesticide detections yield very similar ecotoxicological risk profiles in NSG and NNSG. These results indicate that a consideration of the ecotoxicological risks of the pesticides used seems very reasonable. Thus, potential environmental degradation from pesticides occurs to a similar degree in NSG and NNSG, despite differences in the intensity of protection efforts. 

Multivariate analyses of land-use parameters in watersheds revealed that agricultural use is the primary source of pesticide inputs to NSGs and most strongly determines ecotoxicological risk. In contrast, urban and industrial sources of inputs played only a minor role. Further geostatistical analyses indicated that agricultural use (e.g., cropping, orcharding) in NSGs, both in Saxony and throughout Germany, occurs only to a very small extent and, moreover, the uses occurring there are among the less pesticide-intensive crops (e.g., grasslands). Thus, the sources of pesticide inputs to NSGs are probably often not located in the protected areas themselves, and inputs via inflowing waters or atmospheric transport probably play an important role. This is consistent with the multivariate spatial analyses previously conducted.

The machine-learning based evaluation of protected area ordinance texts was conducted differently than originally planned. On the one hand, which was not known before, comprehensive workups of protected area ordinances were carried out at the same time as the present project on behalf of the UBA (Mühlenberg et al. 2021) and thus the topic had already been dealt with; on the other hand, the low prevalence of direct pesticide applications in protected areas presented above suggests that an extensive analysis of ordinance texts related to the topic of "regulations on pesticide use" is not target-oriented. Finally, a semi-automated text analysis demonstrated a high similarity in the regulation texts related to protected areas, which is a poor prerequisite for a machine learning approach. 

First own field analyses (biosphere reserve Pfälzerwald) supported the assumption of a transport of pesticides into protected areas, suggesting that especially the atmospheric transport of pesticides could be a major input pathway. Thus, complex exposure dynamics arise for protected areas, which often find their origin in adjacent or more distant areas and should be urgently investigated in more detail. 

Overall, the present project has already developed relevant insight into pesticide exposure to surface waters in protected areas. The project also provided initial estimates of the sources or dynamics of these loads. On the basis of the information provided so far, especially drift of pesticides into protected areas, atmospheric transport, fluvial transport via flowing waters and the ecotoxicological risk profile of the pesticides used are to be mentioned as concrete points where risk reduction can start. The methodology developed provides a framework, as it were, that can be used in the future for further evaluations in other federal states and with a focus on other protected area categories.