Comment:

Using generalized linear mixed models (GLMMs), we looked at the effects of restoration age, site size, and surrounding land cover on three response variables: wild bee abundance, species richness, and richness rarefied to the lowest sample size among our sites to control for differences in abundance.

Comment:

We decomposed the changes in bee community composition over time, as measured above, into their two components, species replacement and richness effects. To accomplish this, we used the beta diversity partitioning analysis described by Carvalho et al. (2012). This analysis is based on presence–absence data and uses the Jaccard index of dissimilarity rather than presence–absence Bray–Curtis. Richness effects (βrich) can be defined as: βrich = |b − c| ∕ (a + b + c) , in which |b−c| represents absolute difference in richness between sites. Thus, βrepl and βrich are additive and sum to βtot. A more detailed explanation of these equations can be found in Carvalho et al. (2012). We used each of these three measures to create pairwise dissimilarity matrices, which we plotted against a matrix of pairwise age differences between our sites using code from Ensing (2011). For each of our three plots (βtot, βrepl, and βrich vs. age difference), we fitted least square regressions and examined the y-intercept and slope of each. These regressions were used to compare intercepts and slopes in a heuristic way only, because the plotted points were calculated from all possible pairwise comparisons within the dataset and thus not independent from each other. To statistically assess the significance of the relationship between dissimilarity and age difference for each of the three dissimilarity metrics, we used a Mantel test (Lichstein 2006; Carvalho et al. 2012) in the package ecodist (Goslee & Urban 2007).

Comment:

We decomposed the changes in bee community composition over time, as measured above, into their two components, species replacement and richness effects. To accomplish this, we used the beta diversity partitioning analysis described by Carvalho et al. (2012). This analysis is based on presence–absence data and uses the Jaccard index of dissimilarity rather than presence–absence Bray–Curtis. The species replacement component (βrepl) is defined as: βrepl = 2 (min (bc) ∕ (a + b + c)) , in which the number of substitutions between sites is the minimum number of unique species min(b,c), multiplied by two because substitution involves two species. We used each of these three measures to create pairwise dissimilarity matrices, which we plotted against a matrix of pairwise age differences between our sites using code from Ensing (2011). For each of our three plots (βtot, βrepl, and βrich vs. age difference), we fitted least square regressions and examined the y-intercept and slope of each. These regressions were used to compare intercepts and slopes in a heuristic way only, because the plotted points were calculated from all possible pairwise comparisons within the dataset and thus not independent from each other. To statistically assess the significance of the relationship between dissimilarity and age difference for each of the three dissimilarity metrics, we used a Mantel test (Lichstein 2006; Carvalho et al. 2012) in the package ecodist (Goslee & Urban 2007).

Comment:

We decomposed the changes in bee community composition over time, as measured above, into their two components, species replacement and richness effects. To accomplish this, we used the beta diversity partitioning analysis described by Carvalho et al. (2012). This analysis is based on presence–absence data and uses the Jaccard index of dissimilarity rather than presence–absence Bray–Curtis. Total Jaccard dissimilarity (βtot) is defined for each pairwise comparison of restorations of different ages as the proportion of species that are not shared between the two sites: βtot = (b + c) ∕ (a + b + c) , where a is the number of species found in both sites, b is the number of species unique to the first site, and c is the number of species unique to the second site. We used each of these three measures to create pairwise dissimilarity matrices, which we plotted against a matrix of pairwise age differences between our sites using code from Ensing (2011). For each of our three plots (βtot, βrepl, and βrich vs. age difference), we fitted least square regressions and examined the y-intercept and slope of each. These regressions were used to compare intercepts and slopes in a heuristic way only, because the plotted points were calculated from all possible pairwise comparisons within the dataset and thus not independent from each other. To statistically assess the significance of the relationship between dissimilarity and age difference for each of the three dissimilarity metrics, we used a Mantel test (Lichstein 2006; Carvalho et al. 2012) in the package ecodist (Goslee & Urban 2007).

Comment:

We decomposed the changes in bee community composition over time, as measured above, into their two components, species replacement and richness effects. To accomplish this, we used the beta diversity partitioning analysis described by Carvalho et al. (2012). This analysis is based on presence–absence data and uses the Jaccard index of dissimilarity rather than presence–absence Bray–Curtis. Total Jaccard dissimilarity (βtot) is defined for each pairwise comparison of restorations of different ages as the proportion of species that are not shared between the two sites: βtot = (b + c) ∕ (a + b + c) , where a is the number of species found in both sites, b is the number of species unique to the first site, and c is the number of species unique to the second site. We used each of these three measures to create pairwise dissimilarity matrices, which we plotted against a matrix of pairwise age differences between our sites using code from Ensing (2011). For each of our three plots (βtot, βrepl, and βrich vs. age difference), we fitted least square regressions and examined the y-intercept and slope of each. These regressions were used to compare intercepts and slopes in a heuristic way only, because the plotted points were calculated from all possible pairwise comparisons within the dataset and thus not independent from each other. To statistically assess the significance of the relationship between dissimilarity and age difference for each of the three dissimilarity metrics, we used a Mantel test (Lichstein 2006; Carvalho et al. 2012) in the package ecodist (Goslee & Urban 2007).