Detecting assembly rules in farmland bird communities: a challenge in highly dynamic and heterogeneous landscapes Laura Henckel1,4, Nicolas Mouquet2, Vincent Devictor2, Christine Meynard3, Vincent Bretagnolle1 1
CNRS, Centre d'Etudes Biologiques de Chizé (CEBC), FRANCE 2 CNRS, Institut des Sciences de l'Evolution de Montpellier (ISEM), FRANCE 3 INRA, Centre de Biologie et de Gestion des Populations (CBGP), FRANCE 4 Funded by ANR, in the framework of the European project FarmLand (ERA-Net BiodivERsA)
Objective Ecological communities can be structured by different processes: environmental filtering, biotic interactions or stochastic processes (e.g colonization/extinction). The objective of this study is to compare different methods to assess the relative importance of environment vs geographical distance (dispersal) in the structuration of farmland bird communities in an intensive agro-ecosystem. Materials and Methods
Fig. 3 RDA axis
0,3
2nd axis Built area
Fig.1 Study site and location of the point counts
House Sparrow
0,2
- Intensive cereal agro-system - Area : 450 km2, 19 000 fields - Location: central western France Duration : 2009 to 2012 (results presented only for 2012)
1st axis "Landscape complexity gradient"
Skylark
Stone curlew Yellow wagtail
-1,2
-1
-0,8
0,1
European Greenfinch
Intensive crops -0,6
-0,4
European goldfinch Chaffinch
0
-0,2
Red-legged Partridge
-0,1
Melodious Eurasian Wren European robin warbler 0 0,2 0,4 0,6 0,8 Great tit Cirl Bunting European blackcaps
Common Linnet
- Study on breeding birds (62 species) - An average of 341 five minutes point counts/year (in 200 m buffers) - Environmental variables computed in various buffers : from 200 to 1400 m - Analysis with R-2.15.1 (2012)
Tree Pipit -0,2
Forest and hedgerows
v
Nightingale
1
The 1st axis represents a gradient of landscape complexity opposing intensive agrosystems (annual crops only) and heterogeneous landscapes (with forest and hedgerows). Complex landscapes contain more species, but different species are found in each habitat type.
-0,3
Alfalfa -0,4
Corn Bunting -0,5
Variation partitioning based on Redundancy Analysis (RDA) Objective Analysis of the variation of community composition Matrices Community composition (site by species matrix) (Hellinger transformed) Environmental composition by site Geographical matrix (Principal Coordinates of Neighbor matrices (PCNM) or Trend Surface Analysis (TSA)) R Packages PCNM, packfor, vegan ("varpart" and Ref. function) (Legendre et al. 2005)
Multiple Regression on Distance Matrices (MRDM) Analysis of the variation of betadiversity between pairs of sites Compositional dissimilarity matrix among sites =Hellinger Distance Environmental distance (Euclidean distance) Geographical distance (Euclidean distance). ecodist ("MRM" function) (see Lichstein, 2007)
Results (1) – Variation of community composition (RDA) 20%
18%
18%
16%
16%
% of explained variation
% of explained variation
Fig2.b Variation partitioning with PCNM
20%
12% 10% 8% 6% 4%
14%
10% 8% 6% 4% 2%
0%
0%
Fig 2.c Significant environmental variables (buffer of 200m) 3% 3% 8% 9% 9%
Hedgerows Alfalfa
68%
Built area Grassland
Forest Maize
0,08
0,2 0,18 0,16 0,14 0,12 0,1 0,08 0,06 0,04 0,02 0
0,07 0,06 0,05 0,04 0,03 0,02 0,01 0 200 m 400 m 600 m 800 m 1000 1200 1400 m m m
Hedgerows
Built area
Forest
Maize
R2 total
The beta-diversity variation between pairs of sites is mainly due to environmental distance (geographical distance is not significant). Differences in hedgerows, built area and forest composition explain a part of the variation of beta diversity, mostly at low spatial scale (200 m).
Discussion and perspectives components at fine spatial scale (200 m) for both composition and beta-diversity variation. Bird communities are strongly structured by landscape composition (habitat selection). Little impact of dispersion at these spatial scales (very mobile taxa and small size of the study site or low spatial structuration of habitats at this scale).
12%
2% 200m 400m 600m 800m 1000m 1200m 1400m
Fig 3.a Coefficient values in the Multiple Regression on Distance Matrix (MRDM)
- Greater effect of environment than space due to perenial
Fig2.a Variation partitioning with Trend Surface Analysis
14%
Results (2) – Variation of beta-diversity (MRDM)
R2 for the full model
Table.1 Comparison of the methods of analysis
10 km
Coefficient values for environmental variables
0
200m
400m
600m
800m 1000m 1200m 1400m
- The total explained variation is mostly due to environmental factors (explaining between 7 and 15% of the total variation, depending on the year) for both methods but TSA gives more importance to the environmental part and less to the geographical part than PCNM. Among the environmental factors, the amount of hedgerows has the strongest influence, followed by built area and forests. These factors appear to globally have an influence at low spatial scale (200m or less).
- Results are globally coherent between methods even if the relative part of environment vs geographical varies slightly between the two methods. Our results illustrate the statistical limitations of each method and potential bias (Gilbert and Bennett, 2010). - Futhers analysis (glm or gam models) are needed to test more accurately the environmental effect (non-linear or interaction effect) and differences between species or to identify pure geographical effect at a particular spatial scale.
References Legendre P., Borcard D., and Peres-Neto P. (2005), Analyzing beta diversity: partitioning the spatial variation of community composition data. Ecological Monographs, Vol. 75, No. 4, pp. 435-450. Lichstein J. (2007), Multiple regression on distance matrices: a multivariate spatial analysis tool. Plant Ecology, Vol. 188, No. 2, pp. 117-131 Gilbert B. and Bennett J. R. (2010) Partitioning variation in ecological communities: do the numbers add up? Journal of Applied Ecology, Vol. 47, No. 5., pp. 1071-1082
Contact : Laura HENCKEL,
[email protected]