Our study revealed that the main factors determining the present distribution of J. aquatica are abundant summer rainfall and wet soils. With a mean average summer precipitation (June–August) over 400 mm, J. aquatica grows on a broad range of soil types including histosols, gleysols, stagnosols, luvisols, pelosols, and cambisols. When the average summer precipitation drops below this threshold, the occurrence of the species becomes increasingly restricted to wet and moist soil types mainly occurring in peatlands, floodplains, or other wetlands. When summer precipitation falls below 200 mm, occurrence of J. aquatica becomes unlikely. This preference for moist soils agrees with Forbes [10] who described severe J. aquatica infestation problems on the Orkney Islands and found that “soil surface wetness” was the most significant predictor for the occurrence of the target species. The minor influence of precipitation there may be due to climatic differences. Mean annual temperatures 9 K below those in the Alpine foreland and the wide distribution of peat soils [49] suggest that summer precipitation does not limit growth of J. aquatica on the Orkney islands due to a low evapotranspiration. In our study region, temperatures had little influence on the occurrence of J. aquatica. Only in the higher elevated areas of the Alps the number of frost and ice days limit the distribution despite the high summer precipitation there.
Modeling the development of species in dependence on environmental trends has become an important tool to understand ecosystem changes and to devise sustainable management strategies [50], [51]. Many of these models have been designed to predict the spread of harmful organisms and to limit actual and future damages [52]. Most of these target organisms were invasive species with a negative impact on native ecosystems. Jacobaea aquatica can be considered as a ‘native invader’ [53] which also shows a considerable spread at least in parts of its distribution range [54], [55], [16]). For the territory of Bavaria, however, our study could not confirm a significant spread of this problematic species. In contrast, the overall area at risk considerably decreased since 1988–1997, and the probability of occurrence in the affected areas declined. The reason for this is a decrease in summer precipitation, especially in the lower Alpine forelands. A soil type particularly affected by this development is gleysol. Today, the risk areas are mainly restricted to two habitat types: (i) a narrow, sharply delineated area along the northern edge of the Alps where summer rainfall decreased only slightly or even increased, and (ii) wet soils of wetlands and riverine lowlands where the previous water regime remained unchanged.
Our models show a clear decrease in the species frequency since the 1990s accompanied by a distinct shift in the regional occurrence mainly due to a change in rainfall distribution. Here, our findings confirm an earlier study by Suttner et al. [16], who also observed a decline at most of the monitoring points and a shift in the distribution patterns. However, our risk shifts modeled are not fully congruent to the changes observed by Suttner et al. [16] as these authors only used data from the ‘Bavarian Biotope Mapping Database’, which mainly includes protected areas and not agricultural land. Thus, both analyses come to the conclusion that the ‘increase’ of J. aquatica which has been reported by practitioners is more a local phenomenon and does not represent an overall direction in the spatial distribution of the target species.
The trends detected for the recent past will continue in the future: Both climate scenarios rcp4.5 and rcp8.5 predict decreasing summer rainfalls and increasing temperatures until 2037. Most likely, decreasing summer rainfall will move the potential risk areas to higher areas with sufficient rainfall, where growth is actually limited by low temperatures. In the lowlands, the risk areas along streams and rivers and also the general probability of occurrence will decrease. This effect is even more pronounced with the rcp8.5 scenario where the expected decrease of summer rainfall is higher. As a result of climate change, such shifts and the retreat to higher altitudes are observed or expected for many species [56], [57], [58].
Due to higher rainfalls and temperatures and the considerable decrease of frost and ice days predicted by both climate scenarios an expansion of J. aquatica to higher altitudes in the Alps can be expected creating new potential risk areas. In the rcp8.5 scenario, however, this is mitigated by a stronger decrease in precipitation also in higher regions while the decrease of risk in the Alpine forelands is somewhat less due to slightly higher precipitation there. However, as these new risk areas are comparatively small, they will not compensate for the areas with reduced risk. For agricultural practice, problems due to the spread of J. aquatica to higher altitudes can be estimated as low due to the minor importance of agriculture there. For the riverine grasslands on wet histosols, stagnosols, and gleysols in the extra-Alpine lowlands, our models also predict a decreasing infestation risk which indicates a substantial reduction of future management problems by J. aquatica there.
Management practice did not show significant effects on the occurrence of J. aquatica in our study. Neither the cultivation systems, i.e., organic vs. conventional farming, nor land-use intensity in the form of stocking rates, the implication of conservation schemes, the type and amount of fertilization or the frequency of mowing and grazing showed an impact on J. aquatica occurrence.
While these results well agree with Forbes [10] who also found little or no impact of fertilization, cutting frequencies or stocking rates on the occurrence of J. aquatica on 96 farms in Scotland, it contradicts various studies that report significant effects of different management practices from field experiments (e.g. [16], [17], [25], [23], [21]). A major reason for this contradiction may be that the management methods applied in the experiments were specifically targeted at reducing J. aquatica populations. In contrast, our study and the analysis of Forbes [10] reflect the actual farming practice where management decisions are rather determined by the cost-effective achievement of fodder yields than on targeted weed control. Furthermore, J. aquatica does not always respond linearly to control measures and complex interactions of different measures also play a role [23], [21]. In the study area, J. aquatica mainly occurs in wet meadows with a low to intermediate management intensity [21]. At this level of land-use intensity, conventional and organic grassland farming have much of their management practices in common, and the most noteworthy difference between the systems is in the application of organic or mineral fertilizers. As both types of fertilizers similarly stimulate the growth of J. aquatica, no significant difference between the effects of the two systems could be detected.
Also, interactions between the schedule of management measures and the population development can play a decisive role in the establishment of the species. Although J. aquatica may principally find favorable site conditions within a risk area, it needs dispersal and suitable germination conditions to colonize and infest potential areas.
Generally, the production of large numbers of wind-dispersed seeds enables J. aquatica to rapidly occupy areas in the surrounding of existing populations. Due to the light requirement of establishing seedlings, gaps in the grassland sward essentially facilitate the colonization of so far unoccupied areas [22]. Therefore, the maintenance of a close and vigorous plant canopy is an important tool to prevent infestation [19]. In our study, however, no such correlation between J. aquatica infestation and the occurrence of vegetation gaps was observed. Several reasons could have caused this result. Hence, when phases of seed dispersal and availability of suitable gaps do not overlap, the risk of seed predation, mortality or false germination is substantially increased. Considering the short dispersal distances of Jacobaea spp. seeds ( [10], [34], [19] and the impact of wind direction, gaps suitable for germination may have been not close enough to seed producing J. aquatica plants in our study.
Due to the relevance for the spread of the poisonous plant J. aquatica the spatial and temporal interactions between seed production, seed dispersal, and the accessibility of sites with suitable germination conditions should become an important issue of further experiments and models. For practical farmers, maintenance of a close grassland sward should become an essential precautionary measure to avoid future infestation problems.