Editor András Báldi
Mixed Effects of Long-Term Conservation Investment in Natura 2000 Farmland
Version of Record online: 5 DEC 2013
©2013 Wiley Periodicals, Inc.
Volume 7, Issue 5, pages 467–477, September/October 2014
How to Cite
Santana, J., Reino, L., Stoate, C., Borralho, R., Carvalho, C. R., Schindler, S., Moreira, F., Bugalho, M. N., Ribeiro, P. F., Santos, J. L., Vaz, A., Morgado, R., Porto, M. and Beja, P. (2014), Mixed Effects of Long-Term Conservation Investment in Natura 2000 Farmland. Conservation Letters, 7: 467–477. doi: 10.1111/conl.12077
- Issue online: 27 OCT 2014
- Version of Record online: 5 DEC 2013
- Accepted manuscript online: 31 OCT 2013 02:53AM EST
- Manuscript Accepted: 23 OCT 2013
- Manuscript Received: 15 JUL 2013
- Portuguese Foundation for Science and Technology (FCT) through project PTDC/AGR-AAM/102300/2008
Table S1 Summary of key conservation investments made in the Castro Verde Special Protection Area (southern Portugal) between 1993 and 2012.
Table S2 Summary of the land-use changes during the study, using the Portuguese Agricultural Census from 1999 (1995–1997) and 2009 (2010–2012) for the main municipalities of the study area (see Figure 1): Castro Verde (SPA) and Ferreira do Alentejo (Control) (INE 1999, 2009; http://ra09.ine.pt/xportal/xmain?xpid=RA2009&xpgid=ine_ra_publicacoes&xlang=en).
Table S3 Distribution of bird sampling effort (number of transects) and observers across farming type (SPA and Control) and period (1995–1997 and 2010–2012).
Table S4 Mean count per transect ± standard error (minimal and maximum) and percentage of occurrence (Occ) of birds recorded in 78 plots in the Castro Verde Special Protection Area (SPA) and in a control area (Control) (southern Portugal). Species are categorized in terms of habitat specialization (Habitat) and conservation status (SPEC). For each species we indicate the conservation status in Europe (SPEC). Abbreviation (Abbr) is provided for species used in the Principal Components Analysis shown in Figure 5. Flagship species are underlined.
Table S5 Mean richness (number of species per transect) and abundance (number of birds per transect) ± standard error (minimum and maximum) and percentage of occurrence (Occ) of bird categories from 78 plots sampled in the Castro Verde Special Protection Area (SPA) and in a control area (Control) (southern Portugal).
Table S6 Model averaged coefficients (95% confidence intervals) from the five candidate models (Table 1), using a negative binomial family and zero inflation correction (“glmmadmb” function), relating bird species richness and abundance to farmland type (SC; Castro Verde SPA vs. control area), sampling period (BA; 1995–1997 vs. 2010–2012), and an interaction term (SC:BA). Model probabilities (wi) for each full model are also given.
Table S7 Summary results of permutations tests (10,000 permutations) comparing results obtained with focal and random groups of species. In each case we report the percentile of the interaction coefficient estimated for the focal group in relation to the frequency distribution of coefficients estimated for random groups. Large percentiles (close to 100%) indicate that the coefficient was larger (i.e., more positive or less negative) than it might be expected by chance, whereas small percentiles (close to 0%) indicate that the coefficient was smaller (i.e., more negative or less positive) than it might be expected by chance. Finally, medium percentiles (close to 50%) indicate that coefficient was not different than expected by chance. Random groups were obtained by random sampling (without replacement) of species from a larger species pool, while maintaining the same species richness of the focal group. As groups were built hierarchically (e.g., farmland species were a subset of all species, whereas ground-nesting species were a subset of farmland species), the species pool used in each random sampling respected the same hierarchy. In some cases, random sampling produced sets of species that could not be analyzed using zero inflation models with negative binomial errors (fitted using “glmmadmb” function, Neg. binomial) due to lack of convergence, and so these sets were discarded from analysis. The impact of this option was negligible, because similar analysis with Poisson errors and without zero inflation correction (fitted using “glmer” function, Poisson) produced basically the same results.
Table S8 Model averaged coefficients (95% confidence intervals) of models relating site scores along the first two axis (PC1 and PC2) extracted from a Principal Component Analysis, to farmland type (SC; Castro Verde SPA vs. control area), sampling period (BA; 1995–1997 vs. 2010–2012), and an interaction term (SC:BA). Model probabilities (wi) for each full model (full model) are also given (see Table 1).
Please note: Wiley Blackwell is not responsible for the content or functionality of any supporting information supplied by the authors. Any queries (other than missing content) should be directed to the corresponding author for the article.