Context matters: the landscape matrix determines the population genetic structure of temperate forest herbs across Europe

Context Plant populations in agricultural landscapes are mostly fragmented and their functional connectivity often depends on seed and pollen dispersal by animals. However, little is known about how the interactions of seed and pollen dispersers with the agricultural matrix translate into gene flow among plant populations. Objectives We aimed to identify effects of the landscape structure on the genetic diversity within, and the genetic differentiation among, spatially isolated populations of three temperate forest herbs. We asked, whether different arable crops have different effects, and whether the orientation of linear landscape elements relative to the gene dispersal direction matters. Methods We analysed the species’ population genetic structures in seven agricultural landscapes across temperate Europe using microsatellite markers. These were modelled as a function of landscape composition and configuration, which we quantified in buffer zones around, and in rectangular landscape strips between, plant populations. Results Landscape effects were diverse and often contrasting between species, reflecting their association with different pollen- or seed dispersal vectors. Differentiating crop types rather than lumping them together yielded higher proportions of explained variation. Some linear landscape elements had both a channelling and hampering effect on gene flow, depending on their orientation. Conclusions Landscape structure is a more important determinant of the species’ population genetic structure than habitat loss and fragmentation per se. Landscape planning with the aim to enhance the functional connectivity among spatially isolated plant populations should consider that even species of the same ecological guild might show distinct responses to the landscape structure. Supplementary Information The online version contains supplementary material available at 10.1007/s10980-021-01376-7.

Besides the land-use types commonly used in other studies, such as forest, grassland and arable land, we also mapped some land-use types that we considered important as foraging habitat for pollinators, i.e. traditional grassland orchards, semi-natural grassland and unsealed green settlement areas (incl. gardens) ( Table S1). As linear landscape elements, we mapped hedgerows, tree lines, water courses, roads as well as broad herbaceous fringes (width > 3 m), which might serve as foraging or nesting habitat for pollinators. The following mapping standards were used:  Minimum size for polygons: 0.2-0.5 ha; minimum width: 15 m  Minimum length for lines: 50 m; maximum width: 15 m The mapping of green areas was based on structural and color differences, as well as the context in comparison with the surrounding areas. Arable land is characterized by a very homogeneous and evenly cultivated growth. In particular, a specific width of lanes and the often large surface area indicate this type of use. The growth can appear comparatively high. Flower-rich areas show higher heterogeneity in color and texture, which indicates a less intensive management. Sometimes small patch size, as well as slight woody growth can indicate an extensive management. In contrast, intensively managed grassland is often characterized by narrow lanes, a homogeneous structure and/or animal tracks. These areas are usually easily accessible and located within intensively used agricultural areas. They are often very large in comparison with extensive grasslands. Grasslands were indicated as "not definable" if they neither appeared to be intensively managed, nor showed a high structural diversity or floweriness. Wet meadows are often located near rivers or lakes and have a darker color, as well as a high structural diversity. Also numerous ditches can indicate high moisture content. The following land-use maps each comprise the 5 × 5 km² landscape window plus a 2 km buffer. Table S2 Minimum (Min.), median and maximum (Max.) of basic population genetic determinants (population size, connectivity and geographic distance among pairs of populations), within-population genetic diversity (allelic richness (Ar), expected (He) and observed heterozygosity (Ho) and inbreeding coefficient (F)) and amongpopulation genetic differentiation (G'' ST

S3 Microsatellite markers used for genotyping and PCR protocols
Anemone nemorosa  75) that did not result in countable banding patterns after a single PCR. Here, we conducted the first PCR as described above and a second PCR with the same conditions but with 0.5 µl of the PCR product of the first run as template.
For all loci, we applied the following standard PCR program: Step Initial denaturation For all loci, PCRs were performed in a final reaction volume of 12.5 µl, containing 1 µl of DNA (ca. 50-100 ng/µl), 6.25 µl QIAGEN Multiplex PCR Plus Kit (100), 4 µl of H2O and 1.25 µl of primer mix. The primer mix for a singleplex PCR contained 2 μl of forward primer (stock solution concentration 100 pmol), 0.2 μl of reverse primer (with an oligonucleotide tail at its 5' end), 2 µl of fluorescent-labelled oligonucleotide (identical to the 5' tail of the reverse primer) and 95.8 μl of H2O per 100 µl.
The oligonucleotide tails used were the universal sequences M13 (GGA AAC AGC TAT GAC CAT), CAG (CAG TCG GGC GTC ATC), and T3 (AAT TAA CCC TCA CTA AAG GG). The three oligonucleotides were labelled with the HEX dye, the FAM dye, and the TAMRA dye, respectively. During the first cycles of the PCR, the reverse primer with the tail is incorporated into the accumulating PCR products. When this primer is used up, the annealing temperature is lowered, so the fluorescently labelled M13, CAG, or T3 oligonucleotide can anneal and start acting as a primer.
For all loci, we applied the following standard PCR program: Step Initial denaturation For all loci except Pc17 and Pt9, we applied the following standard PCR program: Step Initial denaturation

S4 Crop dominance
For all land-use parcels mapped as arable field, we determined the dominance of three different crop types over the preceding decade (2008 -2017): oilseed rape, maize and other cereals. This distinction was based on land-use data generated within the European Integrated Administration and Control System (IACS) (European Commission 2020) and made available by the respective co-authors in each region. These data provide vector geometries of all agricultural land-use parcels and information on the grown crop types for each year. To aggregate the data across years, we converted them to aligned raster data with a cell size of 10 m. For each cell and crop type, we calculated a dominance value between 0 (crop type present in none of the years) and 1 (crop type present in each year). In some regions, data for the years 2008 (West Germany, Central Sweden) and 2009 (Estonia) were not available. Also, some datasets did not cover all arable fields. In general, a dominance value was calculated for a cell, when ≥ 6 layers were available for that cell. To calculate the area of a crop type in a given buffer zone or landscape strip, we multiplied the dominance value of each cell with its cell size (10 × 10 m²). The area of raster cells, for which no dominance value could be calculated, were subtracted from the total buffer zone or landscape strip area in order to not bias the calculation of percent cover values. Table 1 in the main text  (02)  Shannon diversity of land-use types + diversity effect a + diversity effect a + diversity effect a Edge density +-edge density effect b +-edge density effect b +-edge density effect b +-edge density effect b a Diversity effect: high diversity of land-use types increases chance to find suitable habitats.

S5 Additional information on the expected landscape effects shown in
b Edge density effect: high density of edges (due to small land-use patch sizes or complex shapes) increases richness and abundance of pollinators and birds (nesting and foraging habitat), but also restricts animal movements across the landscape (barrier effect)  The landscape metric FOREST refers to the percent cover of deciduous forest. LFRINGE and LROAD refer to the relative length of herbaceous fringes and roads, respectively. Numbers or ratios added to the variable names correspond to the most influential buffer distance in meters or the most influential width-to-length radio of the landscape strips, respectively  Tables S6.1 and S6.2, which are not presented in the main text. Shown are the partial slopes and residuals as well as the 95% confidence band. All variables are scaled in standard deviation units. Colors represent the different landscape windows: France (Fra), West Germany (GeW), East Germany (GeE), South Sweden (SwS), Central Sweden (SwC), and Estonia (Est). Population genetic variables are allelic richness (Ar), expected (He) and observed heterozygosity (Ho), inbreeding index (F), and genetic differentiation (G''ST and DPS). The landscape metrics SEMNATGRASS and SEMNATVEG refer to the percent cover of semi-natural grassland and other semi-natural vegetation, respectively. LFRINGE, LROAD, and LWATER refer to the relative length of herbaceous fringes, roads and water courses, respectively. EDGEDEN is the land-use parcel edge density. SHANNON is the diversity of land-use types. pcLWOODGRASS is a principal component reflecting the percent cover of grassland and the relative length of hedgerows/treelines (cf. Table 2 in the main text). Numbers or ratios added to the variable names correspond to the most influential buffer distance in meters or the most influential width-to-length radio of the landscape strips, respectively The landscape metrics FOREST, ORCHARD, and SEMNATGRASS refer to the percent cover of deciduous forest, traditional grassland orchards, and semi-natural grassland, respectively. LROAD, LWATER, and LWOOD refer to the relative length of roads, water courses, and hedgerows/treelines, respectively. SHANNON is the diversity of land-use types. pcARABvsGRASS is a principal component reflecting the trade-off between arable land (cereals and oilseed rape) on the one hand and grassland on the other hand (cf. Table 2). Numbers or ratios added to the variable names correspond to the most influential buffer distance in meters or the most influential width-to-length radio of the landscape strips, respectively