Species Composition and Ecological Characteristics of Native Seed Mixes in the Midwest (USA)

Database Construction and Sampling Native Seed Mixes

We first screened a database compiled by Zinnen and Matthews (2022) to identify plant vendors that sold native seed mixes. The database was composed of 557 vendors in seven study states in the Midwest (Illinois, Indiana, Iowa, Michigan, Minnesota, Missouri, and Wisconsin). Once we identified plant vendors with native seed mixes, we recorded the listed species in the mixes directly from the vendors' online catalogs; we recorded these species using taxonomy from the USDA Plants Database (USDA NRCS 2020). We also recorded each species' proportion of weight in the seed mixture, if available, by directly recording a vendor's listed proportion of a species, or by calculating the percentage through seeding weights per unit area. We recorded all seed mix compositional data from June-August 2021. We only recorded seed mixes for which at least half of the included species were native to the study region. We compiled a potential commercial pool of native vascular plant species by combining state vascular floras, primarily from the Universal Floristic Quality Assessment (FQA) Database (Freyman et al. 2016). We used a regional approach to designate native status. Specifically, a species was included in the potential commercial pool if it was designated as native on at least one of the seven state lists. Native plant cultivars in seed mixes were recorded as their parental species. We only recorded data when specific seed mix contents were provided by vendors. Our overall database included 1,031 seed mixes that were recorded from 32 plant vendors. There were 868 seed mixes with data on species' proportional weights in the seed mixes.

Cluster Analysis

We used a cluster analysis to identify the major types of native seed mixes being sold. Our total matrix consisted of species presence/absence data (0 or 1) across all species in seed mixes. We made two alterations to this matrix to calculate the dissimilarity matrix for the cluster analysis. First, we excluded nonnative species because they formed a marginal component of the database, because they did not have readily available ecological information for subsequent analyses, and because we were focused on species native to the region. Furthermore, many prevalent nonnative species (e.g., Lolium spp. [ryegrasses], Avena sativa [oats]) were short-lived cover crops and were not intended to persist in the resulting vegetation. A total of 181 seed mixes (17.6% of the total) included nonnative species; records of nonnative species were removed from those seed mixes. Second, when conducting the cluster analysis, we removed unique species found in only a single seed mix, since these species can add noise and obscure interpretation of resulting clusters. We then calculated the dissimilarity matrix using Simpson's distance metric using the R package “recluster” (Koleff et al. 2003, Dapporto et al. 2020). We used Ward's minimum variance clustering from the “hclust” function (R Core Team 2020). We then characterized the resulting clusters using word clouds of the names of the seed mixes to validate and interpret the clusters. Word clouds are weighted word lists which highlight prominent keywords of texts and can be used rapidly to perceive and outline their contents. We created the word clouds using the R packages “tm” (Feinerer and Hornik 2020) and “wordcloud” (Fellows 2018).

Ecological Comparisons of the Seed Mix Types

We quantified four ecological characteristics of the seed mixes after excluding all nonnative species: species richness, percentage of graminoids in seed mixes by weight, relative fidelity to wetland areas, and blooming phenology for the forbs.

Native species richness was log-transformed to moderate a right-skewing effect of highly diverse seed mixes. After comparing species richness across seed mixes, we also conducted a post hoc analysis comparing log-transformed species richness between CRP and non-CRP mixes. We did this after observing the frequency of CRP mixes in the dataset, which could have influenced the species richness comparisons. In fact, 431 seed mixes (42% of the total) were explicitly tailored to CRP practices. We compared log-transformed species richness between CRP and non-CRP seed mixes across the entire dataset using a Wilcoxon rank sum test, followed by additional Wilcoxon rank sum tests within the seed mix clusters.

Fidelity to wetlands was characterized using wetland indicator values, which were used to explore and validate clusters. Wetland indicator values were developed for wetland delineation; they reflect an estimated probability that the species occurs in a wetland habitat across its entire range (Reed 1988). Five categories are used, which indicate a species' fidelity to wetland habitats: obligate (OBL), facultative wetland (FACW), facultative (FAC), facultative upland (FACU), and upland (UPL). These data were downloaded from the National Wetland Plant List (U.S. Army Corps of Engineers 2020). We calculated the mean wetland indicator value of the seed mixes by recoding the indicators as integers from −2 to 2, reflecting OBL to UPL. The percentage of graminoids was calculated by summing the proportion of the seed mix by weight of species made up of sedges (Cyperaceae) and grasses (Poaceae). We determined the percentage of graminoids because it reflects functional differences among seed mixes and influences community-wide physiognomy of the resulting restoration. We also performed a subset of the graminoid analysis only considering sedges and allies (Cyperaceae) because some studies (e.g., Kindscher and Tieszen 1998, Sivicek and Taft 2011) suggest sedges are substantially underrepresented in prairie restorations. Due to a lack of normality in the data, we used Kruskal-Wallis tests to compare transformed native richness, average wetland values, and percentages of graminoids among seed mix types. These were followed by post hoc Dunn tests with Bonferroni p-value adjustments to determine which seed mix types were significantly different.

We acquired blooming phenology data from Wilhelm and Rericha (2017); these authors reported the range of flowering dates for vascular species in the Chicago region, approximately the center of the midwestern study region. We limited our dataset to “forbs” (non-graminoid, typically wildflower species) because graminoids are wind-pollinated and generally have low relevance to provisioning resources to pollinators. Second, we converted species' blooming days to blooming status by week number (e.g., a hypothetical species blooming between Julian days 10−14 would be recorded as flowering in week 2). We limited our analysis to the core growing season from weeks 14−43, which corresponds to the beginning of April through the end of October. For seed mixes with forbs, we determined the percentage of species that would be flowering during each week. We compared the percentage of weekly flowering across all seed mixtures, and then among the six seed mixture types.

Identifying Species Common among Seed Mixes

Previous literature has emphasized popular “workhorse” species as being easily available and used in high densities by restoration practitioners (Broadhurst et al. 2016, Zinnen et al. 2021). We identified the most common species by accounting for species' occurrence frequency across all seed mixes, as well as their seed weights and densities in the seed mixes. Thus, we calculated a “popularity value” for each species by summing 1) the percentage of seed mixes in which a species was found, across all 1,031 seed mixes, and 2) the mean percentage of total weight the species occupied in a seed mix for the 868 seed mixes with proportional weight data available, and then dividing the sum by two. Consequently, this “popularity value” is analogous to importance values. We opted to incorporate weight into the popularity values instead of seed number density per unit area because the resulting popularity values were highly correlated (R² > 0.95), and because using seed density inflated the popularity of some small-seeded (e.g., Juncus ssp.) species (data not shown). Using the same approach, we also identified the most popular species specific to each seed mix type because some seed mix types were more frequent than others. Furthermore, one trait commonly associated with popular species is that their seed is inexpensive, likely reflecting their high supply due to ease of cultivation and/or harvest, and favor among both growers and restoration practitioners. To further characterize and verify our approach to identify popular species, we assessed whether the species' popularity values were associated with their prices per ounce and pound. Native plant vendors in the Midwest typically sell seed of individual species by ounce (28.3 g) or pound (453.6 g) rather than metric units. We collected per ounce and per pound data separately because prices per pound are often cheaper than purchasing the equivalent weight in ounces. We did not extrapolate price per ounce or pound from smaller fractions if those were the only quantities available (e.g., we did not assume that the price of a seed per ounce was four times that of price per 0.25 ounces). We collected mean price data from a subset of six vendors with detailed price data for a wide variety of species (Supplementary Material Table S1).

Major Seed Mix Types

For the cluster analysis, we examined the resulting dendrogram and chose k = 6 clusters, which formed interpretable groups of seed mixes in the dataset.

The resulting six clusters had clear compositional differences and intended end-uses (Table 1, Figure 1). We interpreted cluster 1 as pollinator mixes. These commonly encountered seed mixes were primarily associated with showy tallgrass prairie forbs. Cluster 2 consisted of basic tallgrass prairie mixes; these seed mixes were similar to pollinator mixes but had greater emphasis on typical tallgrass prairie matrix species, namely popular graminoids (e.g., Andropogon gerardii [big bluestem], Sorghastrum nutans [Indian grass], and Panicum virgatum [switch grass]). Clusters 3 and 4 were seed mixes for wetter soils. However, cluster 3 tended to consist of emergent marsh and wetland species, whereas cluster 4 consisted of many wettolerant prairie species (e.g., Vernonia fasciculata [common ironweed], Zizia aurea [golden Alexanders]). Cluster 5 consisted of medium to shortgrass prairie species, often for dry areas. Cluster 6 was the rarest observed seed mix type but also distinct because there were many unique species in these mixes not found elsewhere. This cluster commonly consisted of a mix of woodland forbs, including some spring ephemerals, and shade tolerant Elymus (wild rye) species. Our results were also robust after removing CRP mixes and choosing k = 6 clusters for the analysis (data not shown).

Native species richness was typically modest across all seed mix types; richness was between 15−30 in over half (51%) of the seed mixes. Across all seed mixes, median native richness was 21 and mean native richness was 22.5. Native richness in seed mixes ranged from 2−102. There were significant differences in log-transformed native richness among the seed mix types (Kruskal-Wallis χ² = 48.1, p < 0.001). Emergent wetland mixes had the highest richness, though they were not significantly different than pollinator, wet prairie, or woodland/savanna mixes. In contrast, tallgrass and dry mixes were the most speciespoor (Figure 2A).

Across all seed mixes, CRP-specific mixes had significantly lower log-transformed species richness than non-CRP mixes (W = 112269, p < 0.001), though the effect size of this was small (r = 0.114). However, when we compared CRP and non-CRP mixes within the clusters, the pattern was not universal (Figure S1). Non-CRP mixes had significantly greater log-transformed species richness in wet prairie mixes (W = 1252, p < 0.001). However, there were no significant differences in log-transformed species richness found between non-CRP and CRP mixes for the other four tested seed mix types (p > 0.29).

• Download figure
• Open in new tab
• Download powerpoint

Figure 1.

Word clouds of seed mix names recorded from the vendor and grouped by seed mix cluster. Larger words are more common in the names of seed mixes within its respective cluster. Above each word cloud is the cluster number 1-6 corresponding to pollinator, tallgrass prairie, wetland, wet prairie, dry prairie, and woodland/ savanna seed mixes, respectively. Note that some Conservation Reserve Program practices in the word clouds are written as words rather than numerals because we used code that removed numbers from seed mix names to clean texts. Some notable CRP practices herein include CP42 (“cpfortytwo”), CP2 (“cptwo”), CP38 (“cpthirtyeight”), and CP25 (“cptwentyfive”). See Box 1 for explanations of these practices.

There were also significant differences of percent graminoids by weight types (Kruskal-Wallis χ² = 146.4, p < 0.001), with pollinator mixes having the lowest, and tallgrass prairie mixes the highest, percentage of graminoid weights (Figure 2B). When we subdivided graminoids into sedges and allies (Cyperaceae), we found that wetland and wet mesic mixes had the highest percentage of Cyperaceae weight (Figure S2). Nonetheless, despite the emphasis on Cyperaceae in wetland mixes, they were still typically less than half of the graminoid weight in seed mixes (Figure 2B and Figure S2). Cyperaceae seed in tallgrass prairie, pollinator, and dry prairie mixes was almost completely absent (Figure S2). Mean wetland indicator values were significantly different (Kruskal-Wallis χ² = 676.1, p < 0.001). The post hoc Dunn test indicated that all six seed mix types had significantly different mean wetland indicator values, illustrating a progressive gradient between wetland and dry prairie mixes (Figure 2C).

We used weekly phenology data for 362 forb species. The distribution of woodland/savanna seed mix blooming percentages was distinct from those of the other five seed mix types (Figure 3F). This seed mix type had a more platykurtic blooming percentage density curve, resulting from the abundance of spring-blooming forbs which flower from April−May. Across the non-woodland seed mix types, the percentage of species blooming during early to late spring was low (Figure 3A–E). The percentage of species flowering tended to peak from mid-July to the end of August; these mixes often had over 75% of the forb species potentially flowering during this period. Blooming percentages sharply declined from mid-September through the end of October for all types of seed mixes.

In general, we interpreted species with lower popularity values (< 7.5) as rare species (~84% of species), whereas species with higher values (> 7.5) were interpreted as common (~16% of species). Species with the highest values (> 20) could be considered the most popular components of seed mixes in the Midwest (Table 2). Across all seed mixes, these most popular species were prairie grasses and showy forbs that had low fidelities to specific seed mix clusters. Many species could be considered rare in seed mixes. For example, 45% of observed native species were found in less than 1% of all seed mixes; furthermore, nearly 35% of recorded species averaged less than 1% of the weight the total seed mix. There was a positive relationship between log-transformed percentages of frequency of occurrence and arcsine-transformed average percent weight (Figure S3), suggesting that species frequently included in seed mixes are also included in greater proportional weights. Unsurprisingly, several popular species across all seed mixes were also repeatedly found as being the most popular species in specific seed mix types (Tables 1 and 2). There were modest (R²_adj > 0.19) but significant negative correlations between species' popularity values and seed prices at both the ounce and pound scales (Figures 4 and S4).

• Download figure
• Open in new tab
• Download powerpoint

Figure 2.

Violin plots comparing the ecological characteristics of different seed mixture types. There were significant differences among the mixes for log-transformed species richness (A), percent of mixes composed of graminoids by weight (B), and mean wetland indicator values (C). For each subpanel, seed mix types are ordered from lowest to highest mean characteristic. Uncapitalized letters represent significant (p < 0.05) post hoc differences between groups following Bonferroni p-value adjustment. Each point represents a seed mix (i.e., raw data point), and the bars represent group means. Bands (box-like objects centered around mean) represent mean inference; colored density curves (“beans”) reflect data distribution. Random noise was added to the data to facilitate data visualization.

• Download figure
• Open in new tab
• Download powerpoint

Figure 3.

Blooming phenology density of forbs in the different seed mix types (A−F) during the core growing season (weeks 14–43). Each data point represents the percentage of forb species flowering in a single seed mix (“% Sp. flowering”) on a given week. The black lines are density curves reflecting the percentage of per given week. To ease interpretation of the data, we added specific dates on the x-axis and added random noise to raw data points to prevent overplotting.

• Download figure
• Open in new tab
• Download powerpoint

Figure 4.

The most popular species in seed mixes are associated with individually inexpensive seed prices. Cost data for 328 species are displayed on a logarithmic scale, allowing for improved data visualization. “Popularity value” reflects a species' frequency of occurrence across all seed mixes, and the mean percentage of a species' proportional weight in seed mixes; the value was calculated by averaging the percentages and ranges from 0−100.

Seed Mixes and the Midwestern Native Seed Industry

There are many native seed mixes available to practitioners in the Midwest from a modest number of native plant vendors. The vendors with the largest number of surveyed mixes were often tailored to providing CRP mixes in several midwestern states; these vendors also often separated seed mixes by slight alterations of the species or proportion of weight to cater to different site conditions (e.g., “wet mesic to wet” vs. “wet mesic”), or by CRP practice. Consequently, our results are influenced by CRP practices and the vendors that specialize in assembling a variety CRP mixes to meet the market demand (Box 1). However, this may also be a broad reflection of the relative demand of these seed mixes in the Midwest.

Major Seed Mix Types and Their Ecological Characteristics

Seed mixes can be divided into six interpretable types. These seed mix types were identifiable based on their characteristic species, the ecological characteristics of their component species, and even their names.

Unsurprisingly, most seed mixes in the Midwest could be associated with tallgrass prairie restorations, often with low- to medium-diversity CRP plantings. However, we found that prairie mixes can be further subdivided into four types: pollinator, tallgrass prairie, wet, and dry prairie. Pollinator mixes and tallgrass prairie mixes were highly similar in terms of species composition. However, pollinator mixes emphasized showy prairie forbs and prairie grass weight was generally subdued, whereas tallgrass prairie mixes had higher proportions of grasses by weight (see also Meissen et al. 2020). Wet and dry prairie mixes also shared many species with pollinator and tallgrass mixes, but more mesic prairie species were replaced with more lowland and upland species, respectively. Therefore, these four types broadly reflect the differing goals of the resulting prairie restorations (e.g., pollinator conservation-based vs. general habitat creation), as well as the diversity of site hydrology throughout the Midwest.

Both wetland and woodland/savanna mix types had unique species compositions, and given their rarity, were clearly for more specialized restoration projects. There are projects to restore savannas or forest understories in the Midwest; these projects utilize selective tree cutting and reintroduction of fire, and seed mix addition (Ladwig et al. 2020). However, relative to grassland restoration, wetland or woodland restorations are less commonly associated with CRP practices. Some of the demand for wetland and wetter woodland/savanna mixes may be driven by Section 404 of the Clean Water Act, which requires mitigation of wetland impacts through compensatory wetland restorations (Hough and Robertson 2009). Yet, with respect to habitat coverage across the midwestern landscape, and total market value for native plant vendors, such projects are clearly overshadowed by prairie restorations.

Ecological Characteristics

The ecological characteristics of the seed mix types significantly varied and reinforced their differences. Most simply, we found a sharp, distinct gradient of species' fidelity to wetlands in the different seed mix types, which reflected the soil moisture conditions of the intended restorations. Species destined for specialized habitats like wetlands and dry prairies strongly matched those habitats' respective hydrology. In contrast, woodland/savanna mixes could be structured by shade more than hydrology; consequently, these mixes included hydrological generalists.

Secondly, species richness was highest in wetland mixes, whereas tallgrass and dry prairie mixes tended to be species-poor to species-moderate, often consisting of fewer than 20 species. Perhaps species diversity was higher among wetland mixes because they were not associated with species-poor CRP practices like those observed in many prairie mixes. Furthermore, users of wetland seed mixes (e.g., non-governmental or compensatory restoration practitioners) could be more consistently interested in introducing greater species diversity to restored sites.

Pollinator mixes had the lowest proportion of graminoids by weight. Pollinator mixes trade popular graminoid species seed for more showy forbs to facilitate higher floral density (i.e., pollinator resource availability). In contrast, tallgrass prairie mixes had the highest proportion of graminoids. The proportion of graminoids can have substantial impacts on the quality of resulting restorations. Commonly, prairie restorations consist of high densities of C₄ grasses, often at the expense of other functional types (McCain et al. 2010, Grman et al. 2020). We also found that Cyperaceae species were either entirely absent or only marginally present in most prairie mixes. Sivicek and Taft (2011) also described the inadequate amount of Carex species in prairie mixes; these authors found the lack of sedges to be one of the most profound incongruities between restorations and high-quality remnants. The poor representation of Carex species in prairie seed mixes precludes their functional benefits to restoration. Specifically, Carex species can account for a substantial amount of tallgrass prairie biomass, particularly during the early growing season (~25% of biomass, Coppedge et al. 1998), and they occupy physical and ecological space to curtail invasion of nonnative, cool-season grasses (Houseal and Smith 2010). Thus, we strongly encourage greater inclusion of Cyperaceae species (particularly Carex spp.) in native seed mixes (see Boeck Crew et al. 2020).

Lastly, we found differences in blooming phenology among the types. Most notably, woodland/savanna mixes had a greater proportion of earlier-blooming species, which led to a differential blooming distribution. To some extent, it is expected that prairie habitats will have differing phenological peaks compared to woodlands and savannas. However, prairie remnants have a wide variety of early blooming species (Wilhelm and Rericha 2017) that are underrepresented or absent in seed mixes (Havens and Vitt 2016). Correspondingly, we found that most prairie mixes had few—if any—early blooming forbs, which may limit pollinator resources. Unfortunately, many recommended early blooming species for prairie restorations (see Havens and Vitt 2016) are rare in seed mixes (see section below), as well as expensive (e.g., Dodecatheon meadia, Geum triflorum, and Phlox pilosa are > $70 / oz). Restoration managers have reported they want more early blooming species in their seed mixes—including hemiparasitic forbs (e.g., Pedicularis canadensis)—but these species have limited availability or establishment (Barak et al. 2021). The poor availability and steep costs of early blooming forbs could be because they are difficult to cultivate and harvest. These species are often small and ecologically conservative, which might also complicate their establishment in both cultivated or restoration settings. Nonetheless, more efforts are needed to include early blooming species in seed mixes to provision pollinators during the earliest phases of the growing season. Perhaps this could be accomplished by improving cultivation techniques, or by identifying alternative species that are more amenable to agronomic cultivation.

It becomes clear from these basic ecological characteristics that restorations resulting from seed mixes will not compositionally match many high-quality natural areas. However, partial rather than full recovery of native community composition is the goal for many practitioners and restoration projects (Gann et al. 2019). Seed mix design is difficult because there must be balance among costs, the appropriateness of species, supply and feasibility, and diversity, which influences the eventual composition and quality of the restoration (Meissen et al. 2020). In some situations, restorations resulting from simple seed mixes can achieve functional similarity to natural areas, such as carbon storage in nondiverse CRP plantings (Baer et al. 2002). Although more diverse seed mixes can increase diversity of the resulting restoration (e.g., Piper et al. 2007, Meissen et al. 2020), some species included in diverse mixes may fail to establish (Grman et al. 2015), and some managers emphasize repeat seeding events rather than single plantings (Barak et al. 2021). Nevertheless, when over 100 native plant species can be found in some midwestern prairie relicts (e.g., Auffenorde and Wistendahl 1985, McClain et al. 2014), few seed mixes in the region approach the species diversity of the best remaining natural areas. In other words, the native biodiversity promoted by most commercially available seed mixes is quite limited. Perhaps new CRP practices or similar initiatives could be developed that incentivize the fuller recovery of native species pools, thereby championing higher-quality restorations and simultaneously stimulating the native seed industry (Zinnen et al. 2021).

The Most Popular Species

Our work clarifies the characteristics of the most popular “workhorse” species in seed mixes, which have both positive and negative implications for restoration. Unsurprisingly, we found that popular species overlap with some of the most dominant species by cover in natural prairies, namely C₄ grasses. C₄ grasses are considered some of the most important functional species in a prairie community. For example, they can drive community productivity (i.e., carbon capture) and suppress invasive species (e.g., Smith and Knapp 2003, Wilsey 2010; Grman et al. 2020). However, the overutilization of C₄ species sourced from commercially produced seed could be generating homogenous restorations across the Midwest (Zinnen et al. 2021).

The popular forb species are often early successional, so they quickly enhance the aesthetics of the resulting planting for stakeholders. They also provision pollinators. Indeed, several of the popular forbs have been previously highlighted as strategic resources for pollinators. Using a database of native bee visitation of Iowan prairies and comparing them to available seed mixes, Harmon-Threatt and Hendrix (2015) identified Dalea purpurea (purple prairie clover), Ratibida pinnata (yellow coneflower), and Zizia aurea (golden Alexanders) as keystone species for supporting native bee diversity in prairie restorations. Similarly, by accounting for floral availability, Simanonok et al. (2021) found that native bees were biased toward foraging for Oligoneuron rigidum (stiff goldenrod) and Monarda fistulosa (wild bergamot), although they selected against Rudbeckia hirta (black-eyed Susan). Monarda fistulosa (wild bergamot) was also highlighted as a preferred species for bees in Michigan CRP plantings (Quinlan et al. 2021). Together, these studies suggest that the popular forb species in seed mixes can be uniquely desired by pollinators.

We also found that popular species have inexpensive seed, which is beneficial in two ways. First, these species limit costs of ecological restoration. Cost of seed is widely regarded as one of the most limiting and vexing aspects of grassland restorations (e.g., Peppin et al. 2010, Rowe 2010). The relevance of seed mix cost is even greater for CRP plantings, for which agricultural stakeholders receive modest financial compensation for land retirement. Secondly, because these species are affordable, they can be added at greater weight to increase cover. Nonetheless, inexpensive seed may reinforce the omnipresence of a rehashed shortlist of generalist, easy-to-establish prairie species.

Another attribute of popular species is that they were commonly constituents of multiple seed mix types. In other words, these species display adaptability with respect to restoration end-use. Perhaps the best example is Elymus virginicus (Virginia wild rye). Virginia wild rye can thrive in both shade and open areas, and while it is mostly found in wetland areas, it can be found in mesic habitats (Shadow 2009). In seed mixes, it was frequently found in wetland, wet prairie, and woodland/savanna mixes. Other popular species like Schizachyrium scoparium (little bluestem), Zizia aurea (golden Alexanders), Rudbeckia hirta (blackeyed Susan), and Andropogon gerardii (big bluestem) might be considered typical mesic prairie species. However, these species inhabit several broad vegetation communities, including wetland, woodland, and cultural (i.e., early successional human-impacted) plant communities (Iverson et al. 1997).

We have highlighted a small number of species that are the most popular, but clearly many species found in seed mixes are uncommon. These rare or underutilized species could be infrequent and in low weights in seed mixes due to their high costs and likewise sporadic supply, obscurity to consumers, or redundancy compared to more commercially viable species (Zinnen et al. 2021, Zinnen and Matthews 2022). It is unclear and warrants further study if hitherto underutilized or uncommon species in seed mixes could—or should—be targeted for greater inclusion into seed mixes. Some native species are undesirable components of restoration settings, whereas others may be too specialized to be a target species in most contexts (Zinnen and Matthews 2022). Still, we suspect that there is a wide variety of species that would be valuable additions to seed mixes, but which are unknown or underappreciated by producers or practitioners. Future studies could identify these species by identifying their functional characteristics (e.g., value to pollinators, Ladouceur et al. 2022). Increasing a species' representation in seed mixes will also necessitate sourcing seed for its cultivation, increasing seed production, and consistent demand from consumers (Jones 2019, Pedrini et al. 2020).

Synthesis

The midwestern United States has an advanced native seed industry underpinned by large-scale ecological restorations that generate market demand. We identified six seed mix types available from suppliers, types which fundamentally reflect the different target restorations in the region. Although some seed mixes in the market are species rich, most are poor or moderate in diversity with fewer than 30 species, owing to the prevalence of simplified native plantings on the landscape. A small number of species could be considered popular or common in midwestern seed mixes. These are prairie species that can be found in a variety of conditions, with readily available and inexpensive seed. In contrast, most species found in midwestern seed mixes are rare, being found infrequently and present in small amounts by weight. These findings represent a detailed overview of the native seed mixes for restorations in the Midwest. We suggest regionalized studies of the native plant trade can improve knowledge about ecological restorations.