The costs of diversity: higher prices for more diverse grassland seed mixtures

We collected detailed information about seed mixtures and their prices from online price catalogs for shops in Germany and Switzerland during 2019 and 2020 (see Text S1 for further details). For the data collection, we systematically searched for price catalogs on Google using various combinations of German search terms (table S1 (available online at stacks.iop.org/ERL/16/094011/mmedia)). We only used data of seed mixtures if both price and species information were available and only if products were sold to farmers and other land managers for grassland production and/or restoration. Thus, we focused on online shops for a professional target audience, i.e. we excluded online shops that only offered very small quantities for private gardening ⁴ . Based on these criteria, we used data from six online shops (see table S2), which sell within the respective country and do not restrict their sales to specific regions. Moreover, while we cannot assess the representativeness of the six online shops in our study compared to German and Swiss shops without online available price and species information due to unavailability of market information, we can assume that these six online shops i) offer a similarly large variety in seed mixtures as other shops and ii) that their prices are comparable. This assumption is supported by the large seed mixture variety in our dataset and the similarity of online and offline prices (Cavallo 2017).

We collected information on several characteristics of the seed mixture (table 1). First, we collected data about price per kg and recommended seed density (kg per ha) ⁵ . From these two variables, we calculated the seed mixture costs per ha. Second, we collected data on the proportions of individual species in seed mixtures. Based on this information, we computed two different plant diversity measures per seed mixture: (1) the number of species and (2) the Shannon index. The latter was calculated using information about the proportions of species in seed mixtures: $Shannnon\,index = - \mathop \sum \limits_n^N {S_n}{\text{ln}}\left( {{{\text{S}}_n}} \right)$, where ${{\text{S}}_n}$ is the share of the species $n$ ($n$, ..., N) in a seed mixture (Krebs 1999). For 5% of the seed mixtures, species identity but not their proportions were available. As a robustness check, we conducted an additional analysis for which we included these observations when using the Shannon index and assumed equal species distribution. As we focused our diversity assessment on the identity and the number of plant species, we did not distinguish between different genotypes or cultivars of the same species in a mixture. Third, we identified the provenance, i.e. native ecotype vs. cultivar, of the species in the seed mixture according to the production standards and origin of the seeds used for production. We categorized seed mixtures as 'seed mixtures of native ecotypes' when seed mixtures exclusively contained ecotypes of regional provenance, i.e. when seeds originated from and were propagated in a distinct region in Germany or in Switzerland as a whole (see Bucharova et al 2017, Durka et al 2017 for Germany, SKEW (Schweizerische Kommission für die Erhaltung von Wildpflanzen) 2009 for Switzerland) ⁶ . The alternative category 'cultivar seed mixtures' included all seed mixtures with native or non-native species that had been modified by plant breeding, i.e. cultivars, and which were frequently (though not in all cases) propagated outside of the specific regions or countries where they were used. Seed mixtures containing both cultivars and native ecotypes were also categorized as 'cultivar seed mixtures', as such seed mixtures are not suitable for ecological restoration in a strict sense. Fourth, we classified seed mixtures designed specifically for permanent grassland as 'permanent grassland seed mixtures' when the mixture was recommended to be used for more than three years (Suter et al 2017). All other seed mixtures, e.g. grass-clover mixtures specifically for temporary grassland, were categorized as 'other (temporary grass and other non-permanent forage mixtures)'. Seed mixtures with the declared purpose of introducing grass and/or clover species into existing swards through over-sowing were also categorized as 'permanent grassland seed mixtures'. Finally, we classified if seed mixtures had an organic certification.

In total, our data consisted of 262 different seed mixtures (table S2) for which information was available about seed mixture name, number of species, individual species proportions, price per ha and kg, recommended seeding density, seed mixtures containing only native ecotypes, recommended use for permanent vs. other grasslands, organic certification, designation for over-sowing, presence of legumes, the share of legume plus herb species seeds, the share of grass species seeds, online shop, and country (see tables S2 and S3 and figures S1 to S2 for descriptive statistics) ⁷ .

For simultaneously exploring a large number of different species combinations in seed mixtures, their frequency, species frequency, and price ranges, we used the UpSet approach (implemented in the R package 'UpSetR' (Gehlenborg 2019)). The UpSet approach is a matrix-based layout approach, which can be used to explore and visualize seed mixtures (Conway et al 2017). Hereby, we focused on (1) the 20 most frequent plant species and assigned all remaining species to the category 'Other Species', and (2) specific species combinations that occurred at least twice in the dataset.

For identifying the relationship between the price per ha and the plant diversity of seed mixtures, i.e. either number of species or Shannon index, we employed two different linear models ⁸ . First, a simple model that focuses on the correlation between seed mixture price and plant diversity controlling only for systematic price differences among shops:

$$\begin{equation}log\left( {{p_i}} \right) = \,{\alpha _0} + {\alpha _{Div}}Di{v_{ij}} + {\alpha _{Shops}}Shop{s_i} + {e_{1i}}\end{equation} \tag{ 1 }$$

where ${p_i}$ is the price per ha of seed mixture $i$. $Di{v_{ij}}$ is the plant diversity of seed mixture $i$, expressed either as the number of species or the Shannon index depending on the index $j$.$\,Shop{s_i}$ is a vector of dummies for the different online shops.$\,{e_{1i}}$ is the error term. Second, we estimated a model that controls for online shops and other important price determinants that were likely linked to plant diversity:

$$\begin{align} log\left( {{p_i}} \right) & = {\beta _0} + {\beta _{Div}}Di{v_{ij}} + {\beta _{Eco}}Ec{o_i} + {\beta _{PermG}}Perm{G_i}\nonumber\\ & \quad +\! {\beta _{Org}}Or{g_i} \!+\! {\beta _{Over}}Ove{r_i} \!+\! {\beta _{Shops}}Shop{s_i} \!+\! {e_{2i}}\end{align} \tag{ 2 }$$

where $Ec{o_i}$ is a dummy variable indicating if seed mixture $i$ consists of native ecotypes (=1) or not (=0), $Perm{G_i}$ is a dummy variable indicating if seed mixture $i$ is designed for permanent grasslands (yes = 1, no = 0),$\,Or{g_i}$ is a dummy variable indicating if seed mixture $i$ is organic (yes = 1, no = 0), and $Ove{r_i}$ is a dummy variable indicating if seed mixture $i$ is designed to be used for over-sowing (yes = 1, no = 0).$\,{e_{2i}}$ is the error term. We reported standard errors that are corrected (i) for the clustered nature of the collected data, i.e. online shops, and (ii) for heteroscedasticity, e.g. due to changing variance with increasing plant diversity. To this end, we used wild cluster bootstrapped standard errors (Cameron et al 2008), implemented in the R package 'multiwayvcov' (Graham et al 2016).

In the main analysis, we used a linear specification of plant diversity to understand the relationship between seed mixture prices and plant diversity. In a sensitivity analysis, we also used a quadratic specification of plant diversity as it allows for diminishing and increasing marginal relationships between plant diversity and seed mixture price ⁹ . The marginal relationship between plant diversity and seed mixture price is $\Delta {p_1}/\Delta Di{v_{ij}}$.

The number of different seed mixtures was 262 in Germany and Switzerland (55 in Germany and 207 in Switzerland) and comprised in total 181 different species (159 in Germany and 81 in Switzerland, respectively; table S2). The three most commonly used species within all seed mixtures were Lolium perenne (180 times; 69%), Poa pratensis (142 times; 54%), and Trifolium repens (141 times; 54%; figure 1). In Germany, the three most commonly used species were Lolium perenne (30 times; 55%), Poa pratensis (29 times; 53%), and Phleum pratense (26 times; 47%), while in Switzerland, they were Lolium perenne (150 times; 72%), Trifolium repens (123 times; 59%), and Trifolium pratense (120 times; 58%; figures S3 and S4).

Zoom In Zoom Out Reset image size

Some rather species-poor mixtures were identified as the most commonly offered products with a distinct species combination. The three most common seed mixtures consisted of: (1) Lolium perenne, Poa pratensis, Trifolium repens, Trifolium pratense, Phleum pratense, and Festuca rubra (14 times; 5%), (2) Lolium perenne (10 times; 4%; note that these seed mixtures usually contained several different genotypes/cultivars of one species), and (3) Lolium perenne, Poa pratensis, Trifolium repens, Trifolium pratense, Phleum pratense, Festuca rubra, and Dactylis glomerata (8 times; 3%; figure 1) ¹⁰ . Of all 262 seed mixtures, 32 (12%) seed mixtures did not include any legumes (table S4). These mixtures were often offered for over-sowing of existing swards and establishing new swards such as clover-free pastures. In Germany, 25% of the seed mixtures did not include legumes, compared to only 9% in Switzerland. We generally found higher per ha prices for seed mixtures that included less commonly used species (summarized in the category 'Other Species') than for seed mixtures that included one or more of the 20 most commonly used species. Moreover, out of the 20 most commonly used species Achillea millefolium, Anthoxanthum odoratum, Centaurea jacea, Cynosurus cristatus, Festuca rubra, and Lotus corniculatus were characteristics of the four most expensive seed mixtures (figure 1).

Seed mixtures with native ecotypes had considerably higher plant diversity (figure 2). Moreover, seed mixtures with native ecotypes were always designed for permanent grassland and not organic (figure S1). In contrast, seed mixtures with cultivars (i.e. not only native ecotypes) were designed for both permanent grasslands and other use (i.e. temporary grass and other non-permanent forage mixtures) as well as were for both organic and conventional production systems.

Zoom In Zoom Out Reset image size

Seed mixture price per ha significantly increased with the number of species in the mixture (figure 3(A)). The relationship of plant diversity with the absolute price per ha showed on average a +6% (95%-confidence interval (95%-CI): +4% to +7%) increase in price per plant species in the simple model that only controlled for online shops (equation (1), table 2). Thus, the price per ha for seed mixtures with, for example, 10 species were predicted to be on average +142 (95%-CI: +134 to +150) Euro per ha higher than the price for seed products containing one species only, which had a predicted absolute price of +223 (95%-CI: +206 to +242) Euro per ha. This represents a price increase of +63%. Along these lines, prices for seed mixtures with 30 and 60 species were predicted to be on average +864 (95%-CI: +683 to +1087) and +5366 (95%-CI: +3404 to +8409) Euro per ha, respectively, higher than seed products containing only one species. These increases represent average price markups of +387% and +2402% for a 30- and 60-species mixture, respectively.

Zoom In Zoom Out Reset image size

In the model that controlled for shops and other important price determinants likely linked to plant diversity (i.e. native ecotypes, permanent grassland use, organic, and designed for over-sowing; equation (2)), we found again a positive, though weaker correlation between plant diversity and seed mixture price (+3% (95%-CI: +2% to +5%); table 2). Seed mixtures containing only native ecotypes had a substantially higher price per ha, namely +75% (95%-CI: +53% to +99%), than those containing cultivars. Seed mixtures designed for permanent grassland use and organic seed mixtures exhibited +30% (95%-CI: +9% to +54%) and +20% (95%-CI: +4% to +38%), respectively, higher prices per ha compared to their counterparts. Seed mixtures designed for over-sowing of permanent grasslands had −29% (95%-CI: −44% to −25%) lower prices per ha, linked to their lower recommended seeding density.

Considering the Shannon index, which is based on the number of species and their species proportions in the seed mixture, as a measure for plant diversity, we also found a positive relationship between plant diversity and seed mixture price per ha when considering price differences among online shops (figure 3(B), table 2). On average, an increase of the Shannon index by 1 was associated with a rise of +59% (95%-CI: +40% to +81%) in the per ha price (the range of the Shannon index in our sample was 0–3.70). Again, we found a weaker, but still significant positive relationship between the Shannon index and the seed mixture price when also controlling for other price determinants than only online shops (i.e. +30% (95%-CI: +23% to +37%)).

Sowing and over-sowing grasslands with seed mixtures are important tools for sward renewal as well as increasing and maintaining grassland plant diversity (Walker et al 2004, Kiehl et al 2010, Merritt and Dixon 2011, Ladouceur et al 2020). The three most frequent species in seed mixtures in our study were Lolium perenne, Poa pratensis, and Trifolium repens which are rather low-growing, valuable forage species, indicative of intensive management, and commonly found in temperate European grasslands (Suter et al 2017, Arbeitsverband der norddeutschen Landwirtschafskammern 2020). All three species are typical dominant species of grazed swards, the backbone of European grassland-based dairy and beef production, and of high nutritive value for ruminants (Mielke and Wohlert 2019, Van Den Pol-Van Dasselaar et al 2019). The most abundant plant species typical of mown grasslands was Trifolium pratense. Most seed mixtures included legumes, in particular clover species. Legumes can have several positive benefits, such as substituting mineral fertilization due to symbiotic N₂ fixation, reducing greenhouse gas emissions by reducing fertilization in general, mitigating drought-induced yield losses, as well as increasing forage nutritive quality and quantity (Temperton et al 2007, Nyfeler et al 2011, Lüscher et al 2014, Fuchs et al 2018). Other frequently found species were mainly grass species in mixtures for high to mid productive (mown) swards, such as Phleum pratense, Festuca rubra, and Dactylis glomerata. All these frequent species were competitive and relatively fast-growing, following acquisitive strategies such as a high specific leaf area that makes them winners of land-use intensification in grasslands (Busch et al 2019). Thus, the most frequent seed mixtures were designed for intensive grassland management and production of forage with high nutritive value for ruminants.

Non-legume herbs were absent in the most frequently found seed mixtures, which indicates that even the most frequent herb species were not elements of seed mixture that were designed for intensively managed grasslands. The most frequent herbs observed in seed mixtures, such as Achillea millefolium, Plantago lanceolata, and Centaurea jacea indicate mid-intensive grassland use. While Achillea millefolium and Plantago lanceolata can benefit from intensive grazing of permanent grasslands, Centaurea jacea is often found in extensively managed lowland hay meadows and negatively affected by land-use intensification, primarily by mowing and fertilization (Busch et al 2019).

The number of species of the most frequently offered seed mixtures ranged from one to nine species, which makes them suitable for agricultural production purposes, but not for restoration efforts as semi-natural grasslands are considerably more diverse (Buchmann et al 2018). Monocultures have generally been strongly discouraged for both intensive and extensively managed grasslands because of their poor performance and low sustainability (Nyfeler et al 2011, Isbell et al 2017). However, most of these one-species seed products were designed for the over-sowing of established grasslands. On the other end of the diversity gradient, high-diversity seed mixtures consisting of native ecotypes are used mostly for nature conservation purposes to restore semi-natural grasslands, often on arable land or after topsoil removal (Baasch et al 2016, Resch et al 2021). However, the availability of these seed mixtures can be limited by regional and production constraints (Nevill et al 2016, Bucharova et al 2019). The limited availability of native seed mixtures led in the past to large-scale restoration projects using species-poor mixtures of cultivars even for nature conservation purposes (Török et al 2010). Furthermore, farmers can use native seed mixtures to establish key plant species needed for result-oriented agri-environmental schemes, e.g. in extensively managed permanent grasslands (e.g. Mack et al 2020). In Switzerland, specific low-intensity grasslands, so-called biodiversity priority areas, are eligible for additional subsidies when specific key species are present. If these key species are lacking, they can be introduced by re- and over-sowing of existing species-poor grasslands with appropriate seed mixture (FOAG (Federal Office for Agriculture) 2021). Assessing the characteristics of the most frequently available seed mixtures for grasslands revealed two distinct types of seed mixtures: first, most mixtures contained 1–10 species, designed for mostly intensive grassland swards; second, very diverse mixtures with 30 to more than 60 species, and a Shannon index >2 for grassland restoration.

We found strong positive relationships between plant diversity and the price of seed mixtures. For example, seed mixtures with 10 species were on average +63% (95%-CI: +62% to +65%) and with 30 species +387% (95%-CI: +331% to +449%) more expensive than seed products with one species. Besides seed costs, also other costs, such as opportunity and direct costs, can increase with plant diversity, with considerable differences between sites (Török et al 2011). For example, maintaining certain species, especially valuable forage species, over time in more intensively used grasslands can require regular over-sowing (Van Den Pol-Van Dasselaar et al 2019), causing costs, such as labor and fuel costs, to farmers. While plant diversity increases some costs, it can also reduce others, including mineral fertilizer costs due to N₂ fixation by legumes (Nyfeler et al 2011). Therefore, seed costs and other costs are important when evaluating farmers' potential economic benefits of plant diversity (e.g. Koellner and Schmitz 2006, Binder et al 2018, Schaub et al 2020a, 2020b).

Furthermore, the relationship of plant diversity and seed mixture prices per ha as well as the price per ha overall were related to other seed mixture characteristics such as provenance and organic certification. Our results showed that especially seed mixtures of native ecotypes were substantially more expensive (⩾+75% (95%-CI: +53% to +99%)) than those not classified as such. This price increase is mainly relevant for the restoration of semi-natural grasslands, for which these seed mixtures are primarily used, and not for intensively used grasslands that are regularly renewed by re- or over-sowing. Many grassland restoration projects are financed or at least supported by public entities, such as the European Commission (Török et al 2011). Therefore, public entities are often indirectly paying for seed mixtures containing native ecotypes. Moreover, subsidies that farmers receive for low-intensity and species-rich permanent grasslands ¹³ can be a motivation to use species-rich seed mixtures of native ecotypes to introduce missing key species, as discussed above.

The higher costs for seed mixtures of native ecotypes can be linked to production processes, including time-consuming seed collection from local populations, limited number of propagations, hand harvest, complicated cleaning of harvested seeds, and quality control standards (Rieger et al 2014, Bucharova et al 2017, De Vitis et al 2017), and can thus limit the use of native ecotypes on larger spatial scales. Previous studies highlighted the critical issue of the shortage of native seed supply (Nevill et al 2016), which can lead to higher prices of native seed mixtures under an increasing restoration demand in the future. Moreover, although the restoration of species-rich grasslands is crucial and the main focus of our study, we note that conserving the few existing semi-natural grasslands should still be a major priority for policymakers (Temperton et al 2019).

Furthermore, price markups for organic certification, regularly required in Germany and Switzerland for organic farming (Bio Suisse 2020, FiBl 2021, FMFA (Federal Ministry of Food and Agriculture) 2021) ¹⁴ , mainly affect seed mixtures for organic forage production, while mixtures consisting of native ecotypes never had an organic certification in our study. Under current policy trajectories, which often promote organic farming, e.g. the EU 'Farm to Fork Strategy', the demand for organic seed mixtures might increase in the future.

Aside from the seed mixture price and the other characteristics considered in our analysis, also other characteristics, such as the species' palatability and drought resistance, can influence farmers' and other land managers' selection of seed mixtures. Those characteristics can also be connected to the recommended use of seed mixtures by seed producers to farmers, such as for permanent or temporary grassland use. The analysis of these other characteristics of available seed mixtures should be addressed by researchers as it remains an important area for future research. Moreover, while it is plausible to assume that online shops represent the available seed mixtures and prices, extending our research by including data from offline shops and sales data might offer interesting future research avenues.

In conclusion, it is important to consider plant diversity costs together with private and public benefits of grassland plant diversity when farmers and other land managers, policymakers, and other stakeholders decide about land management and grassland plant diversity. Neglecting these aspects can lead to a divergence between plant diversity levels recommended to farmers by, for example, extension services, and the actual optimal levels for farmers as well as the public. We showed that plant diversity, and related seed mixture characteristics, especially provenance, are considerable drivers of seed mixture costs, which carries implications beyond the focus of our analysis: reducing farmers' costs of plant diversity could be an effective option for policymakers to support the increase in grassland plant diversity. Furthermore, high costs for plant diversity could also have significant financial implications for restoration activities and might therefore inhibit the realization of farming practices connected to specific agri-environmental schemes. However, reducing seed costs will require considerable efforts, such as competition among seed providers as well as research to develop more cost-effective production and operations (Merritt and Dixon 2011, Pedrini et al 2020).