Farm-scale development and sustainability outcomes
Farmer interviews revealed considerable changes in farm structure and management over the past twenty years (Fig. 2a). The average farm size increased from 24 to 38 ha (Wilcoxon test, p < 0.01) (Table 2). Concomitantly, the average total livestock units per farm increased from 42 to 69 (p < 0.01). The most important crops by area were corn, wheat, carrots, and potatoes. While 14 different crops were reported, 33% of the crop area was corn, and corn was the most important crop by area for 75% of farmers. There was no significant trend in crop diversity over time (Table 2). All but two farmers had livestock, mostly dairy cows, pigs, beef cattle, and poultry. Eight farms were classified as dairy farms, ten as mixed (arable with dairy, cattle, and/or pig), and two were purely arable with no livestock. Three farmers reported quitting dairy during the period under study. Livestock diversity decreased from on average 5.1 different livestock types per farm in 2000 to 3.7 in 2020 (p = 0.02), as farms specialized more on certain livestock types and products. The reported share of animal feed purchased from retailers (feed import) was on average 20% in 2000 and did not change significantly over time (Supplementary Fig. 1a).
Farmer satisfaction was stable, with about the same number of farmers reporting a decrease in satisfaction as an increase (Supplementary Fig. 1b). However, 61% of respondents reported that perceived societal valuation had decreased, which farmers explained in conversation was related to growing political and societal pressure to reduce fertilizer and pesticide use. With only 35% of farmers over 50 years old (compared to the national average of 56%, Erdin et al. 2017), the sample does not show an over-aging of farmers in the study region. Of the farmers interviewed who were over 55 years old, 67% had a successor. The fraction of own land as opposed to leased land decreased during the period of study (p = 0.04) (Table 2).
Overall, there was no clear trend in the perceived economic situation (Supplementary Fig. 1b), but there are more pronounced trends for individual farm types: while 50% of dairy farmers reported being worse off today than 20 years ago, this value was only 25% for non-dairy farmers. Deteriorating economic situations for dairy farms was strongly related to a falling milk price, which was reported to be 25% lower today than 20 years ago. Non-dairy farmers reported no or only small changes in prices received for their main products (Table 2). All farmers either increased (82%) or maintained (18%) production volumes of their most important product (Supplementary Fig. 1a). On average reported production per farm more than doubled (Table 2). The high gains in productivity were mostly the result of farm growth and specialization (e.g., increasing animal numbers while outsourcing parts of their lifecycle to other farms), and only to a lesser degree due to gains in efficiency (e.g., higher milk output per input). Most farmers worked full time on the farm. While off-farm employment tended to increase, there was a large spread in the data and thus no significant effect (Table 2).
The average share of ecological focus areas for biodiversity promotion increased from 12 to 18% of agricultural land per farm (p < 0.01). The average N intensity on the main crop was around 130 kg N ha−1 and did not change during the study period. There was a large scatter in the number of pesticide applications related to crop types. While farmers whose main crop was corn reported applying just one herbicide application, farmers whose main crop was rapeseed or carrots had up to 10 pesticide applications per growing season. However, there was no trend in the number of pesticide applications over time (Fig. 2a). Since the average farm area increased in parallel to livestock units, livestock density also did not change significantly over time (Table 2).
Landscape-scale agricultural development
Landscape mapping showed a slight decrease in the total agricultural area by − 2.5% from 1781 to 1737 ha (Table 2). The decrease in the total agricultural area was mostly due to the conversion of intensive grassland and cropland to settlement (Fig. 3). The proportion of intensively used agricultural land also decreased as the share of extensive grassland grew, confirming the rise in ecological focus area reported in interviews. Meanwhile, the average field size increased from 1.35 to 1.81 ha (p < 0.0001) (Table 2), with a noticeable decrease in small (< 1 ha) crop fields (Supplementary Fig. 2). The total number of field trees decreased from 2229 to 1818 (− 18%), with large trees (− 23%) declining more than small trees (− 6%). These developments collectively describe land use rationalization through consolidation of larger fields and removal of field trees to facilitate management with large machines (Fig. 4).
Total semi-natural habitat area increased. While the area of high-stem orchards decreased (− 7.5 ha), extensive grasslands (+ 60.6 ha) and wetlands (+ 8.0 ha) increased in total area. Also, the total length of linear habitats increased by 11%, which was mostly due to the planting of new tree rows (+ 35%), while hedgerow length remained more or less stable (+ 2%). While old trees in high-stem orchards, relics of the traditional silvopastoral system, were scattered in the landscape, new trees were planted in lines mostly on field edges to facilitate agricultural management.
Variables that were determined both at the farm scale through interviews and at the landscape scale through mapping could be used for cross-validation of the two approaches. The 20 farms interviewed managed 724 ha, equivalent to 38% of the agriculturally used area in the case study landscape. Hence, the sum of farm-scale changes should be representative and reflect changes at the landscape scale. Indeed, the reported increase in ecological focus area was confirmed at the landscape scale, where we noticed an increase in flower strips, field margin vegetation, and other extensively used areas that are likely to qualify for agri-environmental scheme direct payments (Figs. 3 and 4). Many farmers also mentioned that they installed new semi-natural habitats in cooperation with a regional biodiversity conservation non-governmental organization due to financial incentives. Similarly, the sum of land uses determined by farmer interviews correlated well with mapping results (Supplementary Fig. 3). The fraction of agricultural area was 49% crop, 34% intensive grasslands, and 17% extensive grasslands according to interviews, and 45% crop, 37% intensive grasslands, and 18% extensive grasslands according to landscape mapping. Hence, the two approaches seem to be remarkably consistent for variables where consistency could be compared.
A second opportunity for validating observed changes is by comparing our results to a farmer survey carried out roughly 20 years ago in the same study area (Herzog et al. 2006). While this earlier farm survey had a different focus, questions relating to the farm area, crop diversity, livestock units, livestock density, N intensity on the main crop, and pesticide use on the main crop were similar enough to allow direct comparison. If we compare the median result from Herzog et al. (2006) with farmers reported values for 2000 and 2020 in this study, we see that median values of 2000 are closer to Herzog et al. (2006) than 2020 for farm area, crop diversity, livestock units, and pesticide use supporting the trends we report for those indicators (Supplementary Fig. 4). Livestock density had higher variability in our study than in Herzog et al. (2006), which can be explained by the fact that we interviewed a broader selection of farmers, including farmers with no livestock. Nitrogen intensity, on the other hand, was reported to be higher and more variable in Herzog et al. (2006) than by farmers interviewed in this study, which may be an indication that farmers in our study underestimated the former N intensity.
Our interpretation of the 1998 aerial photograph can be validated against habitat mapping in the field performed in parallel to the farmer survey mentioned above (Bailey et al. 2007). The 156.6 ha classified as wetlands in this study for 1998 is almost the same as 157.1 ha of base-rich fens, littoral zone of inland surface water bodies, seasonally wet and wet grasslands, and sedge and reed beds mapped in the field (data by Bailey et al. (2007), reanalyzed). Also, 32.8 ha of intensive orchards and high-stem orchards in this study are very similar to 34.3 ha fruit and nut orchards reported in the earlier study (Bailey et al. 2007). Comparison of these two land uses suggests high accuracy of aerial photograph interpretation, though we expect lower accuracy for the identification of extensive grassland, which was often difficult to differentiate from intensive grassland. For grasslands and other habitat types, the earlier study followed a different mapping protocol, so that validation is not possible.
Desired change according to the three visions
All three stakeholder organizations want to improve agriculture to make it fitter for the future, and they all claim that their vision is sustainable. However, desired changes vary strongly between the three stakeholders and are oftentimes contradicting. There is not a single indicator for which all three visions agree on the desired direction of change (Table 3).
AS represents market-liberal interests, desires, an increase in the farm area and livestock units, and a decrease in crop diversity and livestock diversity (Table 3). A key aspect of Avenir Suisse’s vision is promoting free trade so that agricultural products can be produced in countries/regions where they can be produced most efficiently. This implies higher feed imports since feed concentrates can be produced more cheaply abroad. In general, AS wants less land used for agriculture in Switzerland, since valorization by agriculture is much lower compared to other sectors. Other key elements of Avenir Suisse’s vision are decreasing prices for agricultural products and increasing production per farm, as well as decreasing direct payments to farmers. The focus is thus more on decreasing consumer prices and taxes in general. The vision does not formulate clear social goals for the farmer other than to reduce regulation of farmers to promote entrepreneurship and accelerate structural change towards fewer, larger, more specialized farms. We interpreted this to have ambivalent effects on farmer satisfaction and farm economic situations: while some farms may profit from reduced regulation, others would go out of business (Table 3).
The focus of the vision promoted by the Swiss Farmers Association (SBV) is on improving farmer economic and social situations (Monin et al. 2018). Key demands of the SBV are to improve the income and well-being of farmers, secure the family farm model for future generations, and improve the image of farmers in society (Table 3). For the most part, the SBV wants to slow down change and maintain current farm structures, agricultural landscapes, and levels of environmental protection. However, the SBV wants to increase the total area used for agriculture.
The agroecological movement (LmZ) is in many ways diametrically opposed to Avenir Suisse. LmZ wants smaller, more diversified farms operating in local value chains. Other key elements of their vision are a reduction in livestock units and livestock density, reduction in feed import, and improved biodiversity conservation (Table 3). This translates into more ecological focus areas, more and better connected semi-natural habitats, and reduced fertilizer and pesticide use (Kehnel et al. 2018). In terms of farmer satisfaction and farm economic situation, the desired changes of LmZ were interpreted to be ambivalent (akin to AS). While small farms may profit from these changes, larger farms, especially those that invested in intensive livestock production, would face severe difficulties.
To check the consistency of the desired change matrix, we compared the sum of weights given to indicators from each sustainability dimension (Fig. 5). According to these weights, the Swiss Farmer’s Association (SBV) prioritizes social sustainability aspects such as farmer well-being, Avenir Suisse (AS) prioritizes economic aspects, and the agroecological movement (LmZ) prioritizes environmental aspects relative to the other sustainability dimensions. The focus of the visions on different aspects of sustainability is consistent with the political behavior of these groups. For example, the weights reflect the decreasing environmental focus from LmZ > AS > SBV (Metz et al. 2020) (Fig. 5c).
Agreement between observed and desired development
The observed developments agreed most with the vision of AS. While the median agreement with AS was 72%, with SBV it was 67%, and with LmZ 52% (Fig. 6a). Agreement was significantly affected by the vision (Kruskal–Wallis test, p < 0.001), with significant differences between AS-LmZ (Dunn test, p < 0.001), SBV-LmZ (Dunn test, p < 0.001), but not between AS-SBV (p = 0.05). Although the median agreement between observed and desired development was highest for AS, there was considerable variability between the farms, reflecting their individual development trajectories. When looking at individual farm trajectories, small subsets of farms were more in line with the visions of SBV (25% of farms) and LmZ (5% of farms) as compared to AS (70% of farms). This suggests that while most farms developed in line with desired changes by AS, a few farms followed different trajectories. These farms not fitting into the main trend either had high levels of persistence and or focused more on shifting towards environmentally friendlier forms of production.
Breaking down total agreement into agreement by category revealed that agreement with AS was particularly high for farm structure and economic indicators (Fig. 6b). Most farms in the area increased in size and number of livestock units while becoming more specialized (Table 2), which reflects key elements of the AS vision. Also, productivity increased on almost all farms while prices received for farm products decreased for many farms (Table 2), so, for economic aspects, median agreement with AS was almost 90%. Contrary to the demands of AS, the overall budget for direct payments in Switzerland stayed constant over the period of study (Metz et al. 2020). Due to the contradictive nature of the LmZ and AS visions, LmZ had the lowest agreement in the categories where AS had the highest agreement (Fig. 6b). However, agreement was in general high with the environmental goals of LmZ, especially on some farms. This can be explained by the increase in biodiversity conservation efforts both at the farm and the landscape scale as a result of increased direct payments for such activities (Table 2) (Metz et al. 2020). However, other farms had little change in environmental indicators, translating to a high agreement with the SBV vision.
A sensitivity analysis revealed that the above-described agreement levels are fairly robust (Supplementary Fig. 5). In general, agreement between observed and desired change did not vary more than a few percentage points by varying the “no change” threshold from 0 to 10%. This is because most farm-scale changes were significantly stronger (Table 2, Supplementary Fig. 1). Similarly, the weights assigned in the desired change matrix (Table 3) also had little impact on the results. Setting all weights equally decreased agreement with AS (− 2.6%), increase agreement with SBV (+ 1.8%), and LmZ (+ 1.9%). However, neither changing the “no change” threshold, nor the weights, nor a combination of the above affected the rank of agreement between the visions (Supplementary Fig. 5).
In their review of progress in sustainability science, Sala et al. (2013) concluded that future development of sustainability assessment methodologies should focus on “holistic and system-wide approaches, the shift from multidisciplinarity toward transdisciplinarity; multiscale (temporal and geographical) perspectives, and better involvement and participation of stakeholders” (Sala et al. 2013). Progress has been made, and several new approaches adopted more multiscale and holistic sustainability framings. For example, Chopin et al. (2017) assessed cropping system sustainability in Guadeloupe from 2004 to 2010 by integrating indicators at the field, farm, and regional scale, while Barron et al. (2021) propose a system-wide and multiscale assessment of the sustainability of pasture-based dairy sheep systems including sustainability issues often omitted in earlier such assessments (Barron et al. 2021). However, as pointed out in a more recent review, there is still a gap when it comes to the involvement of stakeholders in the design of indicators and assessment of sustainability in agriculture (Chopin et al. 2021).
The main novelty of our approach is that it can accommodate multiple, even contradictory interpretations of sustainability, addressing the challenge of dealing with the normative dimension of sustainability assessment (Miller et al. 2014; Pascual et al. 2017; Schlaile et al. 2017). A recent study conducted workshops and interviews with stakeholders across Europe to define five socioeconomic pathways (SSPs) for agricultural development in Europe, including one sustainable pathway (Mitter et al. 2020). Such “what could be” scenarios implicitly represent the “what should be” visions of specific stakeholder groups. However, more often than not, stakeholders have opposing definitions of sustainable futures (Robinson et al. 2011; Zorondo-Rodríguez et al. 2014). By focusing on “what should be,” the approach presented here is aiming directly at stakeholder agreement/disagreement and not bothered by probabilities and internal logic of specific scenarios. In our Swiss case study, the three societal visions had almost diametrically opposing desires for future development: Avenir Suisse promoted “land sparing” and economic efficiency while the agroecological movement promoted more of a “land sharing” approach, focusing on different sustainability dimensions (Table 3, Fig. 5). This underlines that a legitimate sustainability assessment needs to be able to accommodate different visions as benchmarks for agricultural development rather than treating sustainability as an absolute value.
In addition to accommodating multiple interpretations of sustainability, our approach also meets the requirements of being holistic, transdisciplinary, and multiscale (see Sala et al. 2013). The approach (Fig. 2) is holistic since it covered cropping and livestock systems together, with indicators that are applicable to a wide set of farming contexts. Furthermore, by including questions on farmer satisfaction and societal valuation, we paid special attention to cover also social dimension of sustainability, which is chronically underrepresented in the agricultural sustainability debate (Janker and Mann 2018). Transdisciplinarity can be defined as crossing disciplinary and scientific/academic boundaries to develop integrated knowledge and theory among science and society (Tress et al. 2005). Our approach is a step in this direction, as it integrates scientific disciplines with non-academic knowledge contained in societal visions. Finally, our approach covers both the farm and the landscape spatial scales, and tracks the development over two decades, thus also including a temporal dimension. We learned that the temporal dimension was especially useful for comparing widely different farming systems, since focusing on the direction of change rather than absolute values smoothed out context variability between farms producing combinations of different crops and livestock. Also, analyzing change rather than one-time measurement removes the need of having (subjective) reference values inherent in most traditional sustainability assessment tools. The added value of analyzing both the landscape and the farm scale was that it allowed cross-validation (Sect. 3.3) and made it possible to detect that there is considerable variation in individual farm trajectories within the landscape.
While we screened visions from Swiss political interest groups, future work could utilize repositories such as the visions described by the Global Scenario Group (Electris et al. 2009; GSG 2021), which are applicable anywhere and resonate strongly with visions presented here. Alternatively, in future work, visions and indicators could be defined specifically for the case study region in stakeholder workshops (Mitter et al. 2019) or crowd sourced from citizens directly (Metzger et al. 2018). While being more resource and time intensive, such an approach would further strengthen the participatory element and increase the relevance of the vision for the case study area. Future studies should also consider that, like in all indicator-based assessments, the choice of indicators will influence the results (Kienast and Helfenstein 2016). In our case, the choice to focus on a broad set of indicators representing three sustainability dimensions was at the cost of not being able to analyze specific topics in more detail (such as farm economics of biodiversity loss). For indicators based on survey answers, an additional issue is that reported answers are biased by perception, which may lead to a distorted view of observed change. This bias can be accounted for by validating survey answers (where possible) with other data sources. Finally, it is important to be mindful of the possible consequences of outsourcing the normative aspect of sustainability assessments to societal visions. Visions may be heavily biased by partisan interests or may contain objectively unsustainable components. In our case, the AS vision only had a poorly developed social dimension in terms of the farmer, focusing more on lower prices and taxes for consumers, elements that the other visions did not prioritize (Fig. 5a). The fact that individual visions tend to focus on different aspects of sustainability, or for different stakeholder groups needs to be considered with required caution and transparency in such a comparison.