Background: Plant communities of fragmented agricultural landscapes, are subject to patch isolation and scale-dependent effects. Variation in configuration, composition, and distance from one another affect biological processes of disturbance, productivity, and the movement ecology of species. However, connectivity and spatial structuring among these diverse communities are rarely considered together in the investigation of biological processes. Spatially optimised predictor variables that are based on informed measures of connectivity among communities, offer a solution to untangling multiple processes that drive biodiversity.

Results: To address the gap between theory and practice, a novel spatial optimisation method that incorporates hypotheses of community connectivity, was used to estimate the scale of effect of biotic and abiotic factors that distinguish plant communities. We tested: i) whether different hypotheses of connectivity among sites was important to measuring diversity and environmental variation among plant communities; and ii) whether spatially optimised variables of species relative abundance and the abiotic environment among communities were consistent with diversity parameters in distinguishing four habitat types; namely Crop, Edge, Oak, and Wasteland. The global estimates of spatial autocorrelation, which did not consider environmental variation among sites, indicated significant positive autocorrelation under four hypotheses of landscape connectivity. The spatially optimised approach indicated significant positive and negative autocorrelation of species relative abundance at fine and broad scales, which depended on the measure of connectivity and environmental variation among sites.

Conclusions: These findings showed that variation in community diversity parameters does not necessarily correspond to underlying spatial structuring of species relative abundance. The technique used to generate is extendible to incorporate multiple variables of interest along with a priori hypotheses of landscape connectivity. spatially-optimised variables with appropriate definitions of connectivity might be better than diversity parameters in explaining functional differences among communities.

We performed this study between July 2015 and June 2017 in the Vega del Tajo-Tajuña agricultural region of the Tagus River Basin, in the South-Central Plateau of the Iberian Peninsula (Fig. 1). We conducted 78 individual collections that included 23 sampling sites. Four sites each of Oak (n = 4 sites x 4 re-samples each) and Wasteland (n = 4 sites x 4 re-samples each) were visited with collections made in autumn and spring over two growing seasons. Edge (n = 2 sites x 6 re-samples + 2 sites x 5 re-samples) and Crop (n = 7 sites x 2 re-samples + 3 sites x 3 re-samples + 1 site x 1 re-sample) with four and eleven sites respectively, were visited in spring, summer, autumn. Eleven sites were chosen to better characterise the variation expected from the Crop communities as cultivated fields are subject to crop rotation and fallow periods. Crop collections comprised 4 fields of Cucumis melo (melon), 2 of Zea mays (maize), 2 of Brassica oleracea (cabbage and cauliflower), and 3 of Hordeum vulgare (barley), i.e., the major summer or winter crops in the area. In Oak and Wasteland sites, 25 m x 25 m quadrats were marked out and 150 samples per site systematically collected at each resampling. In Edge and Crop, 50 samples from a 25 m x 2 m area at each site were collected at each resampling. A boustrophedonic transect method (a line taken alternately from right to left and from left to right, and so on) was used in all instances except for Edge that have highly linear configurations. Depending on the habit of the species, a number of leaves from different parts of the individual were collected, each collection of leaves from the individual representing a single sample. The samples were harvested at fixed points along the transect. Individuals of each plant species were preserved at each collection and specimens assigned a provisional species, genus, or family rank prior to consultation with an herbarium for taxonomic assignments. The identifications were undertaken by Dr. Rosario Gavilán.

Structuring of plant communities across agricultural landscape mosaics: the importance of connectivity and the scale of effect

Michael McLeish^1*, Adrián Peláez¹, Israel Pagán¹, Rosario Gavilán², Aurora Fraile¹, Fernando García-Arenal¹.

¹Centro de Biotecnología y Genómica de Plantas (CBGP), Universidad Politécnica de Madrid (UPM) and Instituto nacional de Investigación y Tecnología Agraria y Alimentaria (INIA) and E.T.S.I. Agronómica, Alimentaria y de Biosistemas, Campus de Montegancedo, UPM, 28223 Pozuelo de Alarcón, Madrid, Spain.

²Unidad de Botánica. Departamento de Farmacología, Farmacognosia y Botánica, Facultad de Farmacia, Universidad Complutense, Madrid, Spain.

Principal investigator:

Email: fernando.garciaarenal@upm.es

Date and geographic location of data collection:

We performed this study between July 2015 and June 2017 in the Vega del Tajo-Tajuña agricultural region of the Tagus River Basin, in the South-Central Plateau of the Iberian Peninsula. We conducted 78 individual collections that included 23 sampling sites, which comprised communities of four habitats with distinct cover types. The four habitat categories were nominated a priori to represent dominant land-cover types in the ecosystem and were distinguished by expert knowledge of community composition gained over twenty years of research in the region. We chose plant communities at sites present in forest (Oak), successional scrubland (Wasteland), and at the borders (Edge) between crops (Crop) to represent habitat categories.

Description of methods for data collection:

Four sites each of Oak (n = 4 sites x 4 re-samples each) and Wasteland (n = 4 sites x 4 re-samples each) were visited with collections made in autumn and spring over two growing seasons. Edge (n = 2 sites x 6 re-samples + 2 sites x 5 re-samples) and Crop (n = 7 sites x 2 re-samples + 3 sites x 3 re-samples + 1 site x 1 re-sample) with four and eleven sites respectively, were visited in spring, summer, autumn. Eleven sites were chosen to better characterise the variation expected from the Crop communities as cultivated fields are subject to crop rotation and fallow periods. Oak sites supported expansive assemblies and required a relatively large sample size to account for patch heterogeneity and rare (low frequency) species. Similarly, Wasteland sites were also subject to patch heterogeneity but in a smaller area than Oak. Crop collections comprised 4 fields of Cucumis melo (melon), 2 of Zea mays (maize), 2 of Brassica oleracea (cabbage and cauliflower), and 3 of Hordeum vulgare (barley), i.e., the major summer or winter crops in the area. In Oak and Wasteland sites, 25 m x 25 m quadrats were marked out and 150 samples per site systematically collected at each resampling. In Edge and Crop, 50 samples from a 25 m x 2 m area at each site were collected at each resampling. A boustrophedonic transect method (a line taken alternately from right to left and from left to right, and so on) was used in all instances except for Edge that have highly linear configurations. Depending on the habit of the species, a number of leaves from different parts of the individual were collected, each collection of leaves from the individual representing a single sample. The samples were harvested at fixed points along the transect. Individuals of each plant species were preserved at each collection and specimens assigned a provisional species, genus, or family rank prior to consultation with an herbarium for taxonomic assignments. The identifications were undertaken by Dr. Rosario Gavilán. The voucher specimens are available on request with permission from the authors.

Description of methods used for data processing:

Filename: McLeish_etal_Abundance_Dryad.csv (Created: 22 October 2019).

These data were collected during the study and comprise species relative abundance counts for each site and habitat of the study. The data were used in the detrended correspondence analysis (DCA) to visualise the homogeneity of relative species abundance estimates among the collections (n = 78) made in each habitat category. The data were also used in the computation of the extrapolated estimate of Sannon’s diverisity (D_AE) and Tsallis entropy diversity (S_q).

Filename: McLeish_etal_LDA_Dryad.csv (Created: 22 October 2019).

These data were generated during the study and from the species relative abundance counts and comprise columns for: 1) a factor for each site of the study (“site”); 2) a factor for each habitat category (“habitat”); 3) a scaled (not centred) continuous variable for asymptotic estimator (D_AE) of Shannon diversity (“DAE.sca); 4) a scaled (not centred) continuous variable for Tsallis entropy (S_q) estimate of diversity (“S_q.sca”); and 5) three successive scaled (not centred) continuous variables for Moran’s eigenvector map scores (“mem.x.sca”). The dataset was used in the Linear discriminant analysis (LDA) that was used to compare the variables in distinguishing the four habitat categories. Three sets of LDAs were used that comprised either the two diversity estimates and the spatially optimised predictors (MEMs), the two diversity estimates, or the three spatially optimised predictors (MEMs).

Filename: McLeish_etal_Spatial_Dryad.csv (Created: 22 October 2019).

The spatial autocorrelation approaches were computed using the geographic coordinates georeferenced during the study (Filename: Virome_Host_Ecology_Coordinates_Dryad.csv, Created: 22 October 2019) and site-by-species relative abundance matrix generated from the counts described above. These data were used to compute the MEM eigenfunctions that correspond to the (n = 23 - 1) study sites. Subsets of the data table were used in the principal component analysis (PCA) of the site-by-species matrix (columns P to MF), a redundancy analysis (RDA) of the site-by-species relationships given environmental variation (columns D to O), and a partial residual analysis (PRA) used to model the effects (a filter of the orthogonal instrumental variable) of the environment on the residuals of species relative abundances. The continuous environmental variables comprise columns for “elevation”, “slope”, and “aspect” for each site. These topographic variables were generated from a raster layer of “elevation” (available from: https://www.europeandataportal.eu). Aspect and slope were calculated from the elevation layer.

The data also include a factor for “landcover” type derive from a spatial polygon object (available from: http://centrodedescargas.cnig.es) as well as a factor for soil type (“fao9soil”) [1, 2], also derived from a spatial polygon object (permissions granted to the first author ID 22640). The polygon objects were rasterised and the variable factors extracted using the minimum value function.

A generalised linear model was used to select from the 19 WorldClim climate variables (available from: https://www.worldclim.org/bioclim). Six WorldClim climate variables were used: Annual Mean Temperature (“bio_1”), Mean Temperature Diurnal Range (“bio_2”), Temperature Seasonality (“bio_4”), Minimum Temperature of Coldest Month (“bio_6”), Precipitation Seasonality (“bio_15”), and Precipitation of Coldest Quarter (“bio_19”).

Funding:

This work was funded by Plan Estatal de I+D+I, MINECO, Spain (grant RTI2018-094302-B-I00) and a Formación de Personal Investigador contract awarded to Adrián Peláez, reference number BFU2015-64018-R (BES-2016-077810). Plan Estatal de I+D+I provided funding for a metagenomics study of plant virus ecology and evolution. The latter funder supported AP’s doctorate. The funding bodies played no role in the design of the study and collection, analysis, and interpretation of data and in writing the manuscript.

References:

1. The European soil database (2006) GEO: connexion, 5 (7), pp. 32-33.

2. The European Soil Database distribution version 2.0, European Commission and the European Soil Bureau Network, CD-ROM, EUR 19945 EN, 2004.