Appendix S2: calibration of the relationship between the composition of the birds community and agricultural intensity
We based our analysis on the findings of Teillard et al. (2015), who explored the relationship between the composition of farmland bird communities and agricultural intensity using bird data collected by the French Breeding Bird Survey (FBBS, Jiguet etal. 2011).
We focused on a community of 22 species classified as farmland birds (Table S2.1). The three bird community descriptors were the community trophic index (CTI), the community specialisation index (CSI) and the community specialisation for grassland index (CSIg). They were computed as follows:
CTI=i=1nNiNtotSTIi (S2.1)
CSI=i=1nNiNtotSSIi (S2.2)
CSIg=i=1nNiNtotSSIgi (S2.3)
where the indices at species level (STI: species trophic index, SSI: species specialization index, SSIg: species specialization index for grassland, see Table S2) for each species i were weighted by the species’ abundance Ni and divided by the summed abundances of all 22 species, Ntot .
The three indices were computed as weighted averages of the species trophic index (STI), the species specialization index (SSI) and the species specialization index for grassland (SSIg), by the relative abundance of the different species, as shown in Eqs. (S2.1) to (S2.3). The value of the STI, SSI and SSIg for the 22 species of the farmland community are provided in Table S2.1.
Species / Specialisation index (SSI)(Julliard et al. 2006) / Trophic index (STI)
(Jiguet et al. 2011) / Grassland specialisation index (SSIg)
(Teillard et al. 2014)
Perdix perdix / 1.31 / 1.1 / 1.25
Motacilla flava / 1.19 / 2 / 1.33
Miliaria calandra / 1.08 / 1.28 / 1.56
Vanellus vanellus / 1.55 / 1.9 / 1.56
Carduelis chloris / 0.86 / 1.05 / 1.58
Coturnix coturnix / 1.21 / 1.22 / 1.59
Alauda arvensis / 1.13 / 1.25 / 1.6
Carduelis carduelis / 0.67 / 1.05 / 1.66
Alectoris rufa / 0.69 / 1.1 / 1.84
Carduelis cannabina / 0.62 / 1.05 / 1.85
Corvus frugilegus / 0.92 / 1.63 / 1.94
Anthus pratensis / 1.33 / 1.75 / 2
Sylvia communis / 0.63 / 1.6 / 2.04
Falco tinnunculus / 0.48 / 2.85 / 2.12
Emberiza citrinella / 0.54 / 1.3 / 2.26
Saxicola torquatus / 0.66 / 2 / 2.29
Emberiza cirlus / 0.39 / 1.3 / 2.37
Buteo buteo / 0.39 / 2.9 / 2.42
Saxicola rubetra / 1.23 / 2 / 2.44
Upupa epops / 0.29 / 2 / 2.53
Lanius collurio / 0.87 / 2.15 / 2.58
Lullula arborea / 0.58 / 1.5 / 2.61
Table S2.1: Species in the farmland bird community and their habitat descriptors used in this study.
Relationships between the CTI, CSI, CSIg, and agricultural intensity (IC/ha indicator, Teillard et al. 2012) were determined at the national scale and across SARs using generalized additive models. Several degrees of freedom were tested and quadratic relationships were selected based on AIC. Relationships were highly significant (p < 0.001) for all three community descriptors. In addition to the IC/ha indicator, other continuous explanatory variables were also included in the models. They consisted of climate, specifically mean temperature and precipitation, and land-use variables (relative area of 15 land-cover categories). Climate data were averaged from 2006 through 2008, and were acquired from Météo France, the French meteorological institute. Land-use data were supplied from CORINE land cover (CLC2006) for the 2006 value. Climate and land-use variables vary importantly along the large geographical gradient of the calibrations and they are expected to have an effect on farmland bird communities; therefore, they were included in the models in order to isolate the effect of intensity. However, we assumed that these variables remained constant in all the simulations of intensity modifications (Section 2.3). Due to the large number of explanatory variables, model selection (based on AIC) was performed to keep only those with highest explanatory power and to avoid model overparametrization.
Teillard et al. (2015) also found a significant interaction between agricultural intensity and its spatial aggregation on the CSIg. This effect was not significant on the other two indices. In their study, the sample of SARs was divided into aggregated and non-aggregated SARs using the AI index described in Eq. 1. Aggregated SARs had aggregation value lower than the national average value. An interaction effect between agricultural intensity and spatial aggregation (binary variable: aggregated or not) was added to the models. The models computed the difference in intercept and slopes of the relationships between intensity and CSIg in aggregated vs. non-aggregated SARs. We integrated these differences in the function used to predict the CSIg from national agricultural intensity allocations (Eq. 5).
The details of the statistical relationships between agricultural intensity and the bird community descriptors are provided in Table S2.2. For details on the other variables included in the models (climate and land use), please refer to Teillard et al. (2015).
Bird community descriptor / F / p-value / Explained deviance / Leave-one-out cross validation errorCSI / 18.928 / < 0.001 / 67% / 11%
CTI / 12.23 / < 0.001 / 31.7% / 9%
CSIg / 24.09 / < 0.001 / 69.5% / 5%
Table S2.2: Performance summary of the generalized additive models linking IC/ha agricultural intensity indicator the three bird community descriptors. n = 332 points for all models. The explained deviance reflects the goodness of fit of the models. The leave-one-out cross validation error reflects predictive ability (the predictive error) of the models (Geisser 1975). CSI: community specialisation index, CTI: community trophic index, CSIg = grassland specialisation index of the community.
References
Geisser, S. (1975). The predictive sample reuse method with applications. Journal of the American Statistical Association, 70, 320-328.
Jiguet, F., Devictor, V., Julliard, R. & Couvet, D. (2011). French citizens monitoring ordinary birds provide tools for conservation and ecological sciences. Acta Oecologica, pp. 1–9.
Julliard, R., Clavel, J., Devictor, V., Jiguet, F., Couvet, D. (2006). Spatial segregation of specialists and generalists in bird communities. Ecology Letters, 9, 1237–1244.
Teillard, F., Allaire, G., Cahuzac, E., Léger, F., Maigné, E. & Tichit, M. (2012). A novel method for mapping agricultural intensity reveals its spatial aggregation : Implications for conservation policies. Agriculture, Ecosystems and Environment, 149, 135–143.
Teillard, F., Antoniucci, D., Jiguet, F., Tichit, M. (2014). Contrasting distributions of grassland and arable birds in heterogeneous farmlands: Implications for conservation. Biological Conservation,176, 243–251.
Teillard, F., Jiguet, F. & Tichit, M. (2015). The response of farmland bird communities to agricultural intensity as influenced by aggregation. PLoS ONE, 10(3), e0119674.
XXX