Sown field margins can improve the conservation of biodiversity in rural areas and can contribute to the aesthetics of rural landscapes, thereby potentially increasing public support for agri-environmental measures. However, these two functions do not necessarily coincide. This raises the question whether field margins that are appreciated for their contribution to landscape aesthetics also deliver on the conservation of biodiversity. We conducted choice experiments with different groups of citizens and collected biodiversity data in the Netherlands, to investigate if the number of colors and vegetation cover in field margins increased respondents’ appreciation for them, and how these visual cues correlated with taxonomic diversity and abundance of plants and invertebrates in those field margins. Using manipulated photos, we also assessed whether the presence of colorful field margins in a range of different rural landscapes increased respondents’ appreciation of those landscapes. Respondents preferred colorful margins with high vegetation cover and showed a preference for green rural landscapes with colorful field margins. The presence of colorful field margins increased landscape aesthetics most in the least appreciated landscapes. The number of colors correlated positively with the diversity of sown and spontaneous plant species, and overall invertebrate abundance and abundance of predatory invertebrates, but was not related to invertebrate diversity. Our results show for the first time that colorful field margins support both public appreciation and diversity of plants and abundance of ground-dwelling invertebrates, with potential advantages to farmers in terms of natural pest control, at least in intensively used agricultural landscapes. However, management practices to maintain a high number of colors over time may be detrimental for invertebrate diversity. To optimize the different functions, we recommend that field margin layouts should consist of a perennial part that is allowed to develop over time, in combination with a part that is managed for its colorfulness.

1.1.      General set up

Field work was carried out in Zeeland in 2006, a province in the southwest of the Netherlands, which is dominated by intensive arable agriculture on marine clay soils. The province is made up by five areas of open, flat landscapes in the marine clay district separated by strands of the Scheldt River estuary. By selecting farms only in this province, the influence of differences in soil or landscape context were minimized (Noordijk et al. 2010). Main crops in the rotation are winter wheat, potatoes, onions and sugar beet (Lokhorst et al. 2009). An arable farm had on average an area of 0.32 km² crop land in 2006 (CBS 2022). Parcels have no natural boundaries, but are typically bounded by ditches, roads, hedges, or dykes.

In the Netherlands, subsidised schemes for promoting agri-environmental management have been in place since 1975 (De Snoo et al. 2016). Non-crop field margins became part of those schemes later (around 2000) and were a popular option in Zeeland. The field margins in our research were targeted at fauna conservation, specifically focusing on birds and insects. Subsidy prescriptions require that these margins are 6-12 m wide, at least 50 long, and border cropland (LNV 2006). All margins were sown with either grasses, wildflowers, or a mixture of these when they were created, but sowing was not annually. The use of synthetic pesticides and fertilizer is not allowed, and mowing is permitted once a year between July 15th and September 14th. Exceptions are made for locally combating very persistent weeds. All farmers involved in field margin management were member of a local agri-environmental farmer collective (De Snoo et al. 2016). For our research we randomly selected 36 arable farms with field margins throughout the whole province. On these farms, 54 field margins were visited in July 2006 to collect data about plants and insects. During these visits, digital pictures of the field margins were taken. One with a landscape perspective and one close-up, showing a representative part of the margin vegetation from above. Pictures were standardized as much as possible in terms of distance and angle and were taken at least 10 m away from field corners and disturbances such as field access or machinery tracks. To standardize the number of colors, all colors observed on the 54 pictures couples were assigned to the closest matching color in the 40 colors standard MS Office 2003 color chart. To avoid variation in interpretation of colors by different people, this was done by the same person (WV).

A complete overview of all the variables that were available on the field margins in the study is given in the Supplementary Material, Table S1. All plants species and groups of invertebrates found are in Table S1 and S2 . None of the farms were organic. The amount of semi-natural area on the farm did not correlate to the number of colors, cover, or the biodiversity variables (results not shown). Distance of margins to semi-natural area was not assessed.

1.2.      Aesthetic appreciation of field margins

We tested the effect of two visual indicators on the aesthetic perception of field margins by the respondents: the number of colors and the vegetation cover. To assess the influence of the two visual indicators on the aesthetic appreciation of respondents, we assigned all close-up pictures of the field margins to one of three color classes (high: ≥ 7 colors, medium: 4-6 colors and low: between 2-3 colors) and one of two vegetation cover classes (high: >90% cover and low: 40%-90% cover) based on the actual distribution of the number of colors and vegetation cover (Fig. S1a and b in Supplementary Material). Field margin photos were then combined in a series of six photos that provided a full factorial representation of color and vegetation cover. To account for potential effects of variability between photos within classes, we created two such ‘margin photo series’ to present to the respondents (Fig. S2).

To assess which type of field margin added the most to landscape aesthetics as perceived by the respondents, artificial landscape pictures were created with Adobe Photoshop by combining a picture of one of our field margins with three types of rural landscape contexts, the ‘landscape series’: a landscape with only green elements present, a landscape with both green elements and buildings (houses and farmhouses) present, and a landscape with green elements and the presence of a larger road with traffic. These landscape contexts were all photographed in the same province as where the field margin study was conducted. To limit the number of pictures the respondents had to classify, we only used two color classes (high or low) and only used pictures of margins with full vegetation cover. The photo of the crop next to the field margin that was used to create the images was kept constant in each picture. Again, we created full factorial combinations of these three types of landscape context and the two-color classes of the field margin, resulting in a series of six pictures. Again, to account for potential effects of variability between pictures within classes, we created two photo series to present to the respondents (Fig. S3).

To assess the aesthetic appreciation of the field margins, a total of 108 respondents ranked the six pictures in each series from the highest to the lowest attractiveness. We asked three citizen groups of respondents, i.e., urban inhabitants, inhabitants of rural villages, and farmers, to rank the same 4 series of pictures to assess differences in appreciation between the three groups. We conducted the study in different provinces (the Dutch provinces of Noord-Holland and Zuid-Holland) than where the pictures were taken to avoid local bias. Urban inhabitants were interviewed in the main shopping street in the city of Leiden (Zuid-Holland), rural inhabitants were interviewed outside a mall in the rural village of Nieuw-Vennep (Noord-Holland). Farmers were interviewed in the Haarlemmermeerpolder (Noord-Holland), an area with arable farms that are generally comparable with those in the province of Zeeland in terms of soil type, size, farming intensity and crops grown. Farmers were first contacted by phone and later visited on their farms to do the ranking.

The 108 respondents were distributed over the three citizen groups as follows: 40 urban inhabitants, 38 rural inhabitants and 30 farmers. For each respondent, we also registered gender, age (in years), highest level of education (primary school, preparatory vocational secondary education, senior general secondary and university preparatory education, vocational education and training, and higher education). Interviews took place in May and July 2007. 

1.3.      Biodiversity assessment of field margins

In the 54 field margins with seeded species, we measured richness and abundance of plant species and invertebrate groups. All biodiversity assessments were made at the locations in the field margin where the pictures were taken.

Vegetation composition, relative cover of each plant species, and total cover were recorded in 25 m long and 1 m wide transects in the middle of each margin in June and July 2006, using an adapted Braun-Blanquet method (Barkman et al. 1964). For assessing the number of plant species, native species were identified using the local flora of Van der Meijden (1990). Plants that could not be identified in the field were collected and compared with herbarium material or identified by experts from the Dutch Foundation for Floristic Research (FLORON). Sown cultivars were identified using Brickell (1999). Vegetation data were processed using Turboveg (Hennekens and Schaminée 2001). All plant species that were found, but of which no seeds were sown during the establishment of the field margins, are indicated as ‘spontaneous plant species’ in the rest of this paper. For studying the relationship between number of colors and the abundance of plants, we added the cover of each individual plant species to a total sum per margin as a proxy for the total abundance of plants in the margins. This was also done with the spontaneous plant species.

As mentioned above, the field margins in our research were targeted at fauna conservation. Their agricultural function was the stimulation of natural pest control. Pollination was not of interest for farmers given their cropping systems. For this reason, we only studied the taxonomic richness and abundance of soil dwelling invertebrates. These were sampled at the end of June and the beginning of July 2006 (weeks 26–27) using 4 pitfalls traps in each field margin (fixation liquid: 50% ethylene glycol) placed 10 m apart. The traps had a diameter of 11 cm, were 7 cm deep, had an elevated plastic cover to keep out rainwater, and were open for 7 days. Catches of the 4 traps were pooled to represent one sample for each field margin. Invertebrates were classified to family level if possible and otherwise to order level by J. Noordijk (Noordijk et al. 2010). Invertebrates were classified into four functional groups based on their main food source. Chilopoda, Araneae, Coccinellidae (including their larvae), carnivorous Carabidae, and Staphylinidae were considered to be predators. Isopoda, Diplopoda, and Collembola were considered as detritivores. Gastropoda, Curculionidae, Orthoptera, Cicadellidae, Heteroptera, and Aphidoidea were considered to be herbivores. And all other species groups were classified as omnivores.

1.4.      Data analysis

All statistical analyses were performed in R software version 4.0.3 (R Core Team 2020).

We used conjoint analysis (Green et al. 2001) to assess the preference of respondents for the pictures in each series. The two series of pictures of margins, as well as the two series of landscapes, represent two observations within a single test and, therefore, cannot be considered independent observations. Rankings of pictures with the same combination of color class and vegetation cover class of the margin series were averaged for each respondent before analysis, as was done with the rankings of the landscape series. The full factorial setup of each photo series allowed us to assess the contribution of each variable (the color and vegetation cover classes in the margin series and colorfulness of the margin and landscape context classes in landscape series) to the appreciation of the respondents of each photograph. All these analyses were performed using the conjoint package in R (Bak and Bartlomowicz 2012).

For testing the differences between the three groups of citizens (urban inhabitants, rural inhabitants, farmers) and the effect of gender, age, level of education on their preferences, we applied the method of Lee and Yu (2013) for testing the difference between marginal matrices, but used Fisher exact probability, the function fisher.test() of R, instead of the Chi-square test.

Spatial autocorrelation in the number of colors in the margins and the main biodiversity response variables was checked by calculation Moran’s I with the function moranI() of the package lctools of R, using a weight that selects the 5 nearest margins, which usually includes the second margin of the same farm, as well as the margins of two neighboring farms (Kalogirou 2020). In the results, we have given the P-values of the randomized z-score, which in all cases were almost equal to the P-values of the resampling.

We used Linear Mixed Models (LMM) to test whether the various measures of biodiversity of the field margins were related to the color and cover classes, as well as to age of the field margin, i.e., the time since sowing. These analyses were performed using the lmer() function of the lme4 package in R (Bates et al. 2015). The different measures of biodiversity were the response variables, with either color or cover as fixed effect variable and farm as random effect variable. For testing the confounding effect of age on the relationship between the measure of biodiversity and color, we used the LMMs for color, but extended them by including age, as well as the interaction between age and color as fixed effect variables. Residuals were checked in all cases. Graphs were made with the scatterplot() function of the car package in R (Fox and Weisberg 2019).