Landscape structure, bird, and vegetation recruitment in restoration plantations

Reversing large-scale habitat degradation and deforestation goes beyond what can be achieved by site-level ecological restoration and a landscape ecology perspective is fundamental. Here we assess the relative importance of tree cover and its configuration on forest-dependent birds and late-successional tree seedlings in restoration sites in southern Costa Rica. The abundance and species richness of birds increased in landscapes with more corridors, higher tree cover, and lower levels of fragmentation, highlighting the importance of riparian corridors for connectivity, and continuous tree cover as suitable habitat. Landscape variables affected abundance and species richness of seedlings similarly, but effects were weaker, possibly because seedlings face establishment limitation in addition to dispersal limitation. Moreover, the scale of landscape effects on seedlings was small, likely because proximal individual trees can significantly influence recruitment in restoration plots. Results underscore the importance of incorporating landscape-level metrics to restoration projects, as knowing the extent, and how the landscape may affect restoration outcomes can help to infer what kind of species will arrive to restoration plots.Funding provided by: National Science FoundationCrossref Funder Registry ID: http://dx.doi.org/10.13039/501100008982Award Number: NSF-DEB 14-56520For the purposes of this study, we used bird and tree seedling data collected within three years (2011–2017) of the high-resolution aerial imagery from 2014 on which tree cover classifications were determined. We used these years to match with the date of landscape tree cover image, as we assumed that the landscape in the region did not change substantially in that period. Bird composition. Each of the 12 plantation plots were surveyed three times annually from 2011 to 2017 by a single observer, Juan Abel Rosales. Each plot at each site was actively searched randomly for 20 min per observation, and all birds detected visiting the plot (by sound or sight) were recorded. All surveys were conducted from sunrise (~ 05:30) until 09:00. Birds flying over plots were excluded from analyses. For detailed information on the field survey methodology see Reid et al.2014. From all detections recorded, we extracted observations of species categorized as highly forest-dependent by BirdLife International. These species have high forest-dependence that is characteristic of the interior of undisturbed forest and almost invariably breed within the forest. While they may persist in secondary forest and forest patches if particular ecological requirements are met, they are usually less frequently encountered in such situations. We chose to only include forest-dependent birds in this study because generalist frugivores are known to be less sensitive to landscape variables, and a preliminary analysis including all frugivores weakened explained deviance in all models. We used the cumulative number of species recorded over the seven survey years to determine species richness observed at each plantation plot and the average number of detections per year for each species as a proxy for abundance (hereafter 'abundance', see Reid et al. 2014) as our two forest-dependent birds response variables. Person to contact with questions about bird data: Francis Joyce, fjoyce@ucsc.edu Tree seedling recruitment. We measured tree seedling recruitment (hereafter seedlings) at four sampling transects located within each plantation (n = 52 transects in total). All tree recruits ≥ 0.2 and 20 years, herbarium specimens at Las Cruces, and literature resources. Late-successional species are those mainly found in the interior of old-growth forests; they are generally hard wooded, slow growing, shadetolerant, and long-lived species. During each annual survey, new recruits were permanently tagged, identified to species, and their height measured. For this study, we considered the sum of new recruits for each species in the period 2011–2017, and the total species richness accumulated within that period as seedling response variables. Person to contact with questions about tree recruitment data: Karen Holl, kholl@ucsc.edu Landscape structure. We used a finely detailed tree cover map created by hand digitizing photographs taken in 2014 by commercial QuickBird satellites with 0.61-m spatial resolution in Google Earth for the canton of Coto Brus (for additional details see Mendenhall et al.40). Tree cover included the percentage of the landscape covered by old-growth and secondary forest fragments of all sizes, single or groups of remnant trees, live-fences,hedgerows, as well as non-native timber and fruit tree plantations. Tree cover represents any source of fruits and seeds within the landscape and more extensive tree cover indicates larger forest patches, which provide habitat to forest specialist birds. Total tree cover in the landscape surrounding the restoration plots (radius 900 m, ~ 255 ha) ranged from 30 to 79%. We classified all tree cover elements in the landscapes according to their area (A), perimeter (P), and compactness (SI) where the compactness ratio of the polygons was calculated using the shape index formula SI = P/2 √(A*Pi). A criterion to distinguish live-fences, remnant trees, fragments, and corridors was to assign threshold shape index values according to the minimal observed values for well detected live-fences (n = 45) and corridors (n = 15) within the study region. Categories include: (1) patches—A ≥ 0.25 ha, SI 0.25 ha and SI > 2; (3) remnant trees—A 1.6. For all landscape classification we used Quantum GIS v3.10.1245. Then, we obtained the percentage of the landscape covered by corridors (riparian and non-riparian) and livefences as both are important connectivity elements in the landscape, they are widely used by birds and other animals, may promote long-distance dispersal, and can also be seed sources for many species. Fragmentation was measured as the number of tree patches divided by landscape area (n/ha). This classical fragmentation metric is inversely related to mean patch size. Thus, for a given land cover area, fragmentation increases the edge-to-core ratio at the landscape scale, potentially increasing edge effects (i.e., small-scale process). Yet, as it is also positively related to landscape connectivity and to the number of (sub)populations (seed sources) in the landscape, it can also have stronger effects on seed dispersal over larger spatial scales. Remnant tree density was calculated as the number of remnant trees divided by landscape area (n/ha). Small clusters of trees and isolated trees function as stepping stones in the landscape, promoting animal movement, and hence, landscape connectivity. The patch aggregation index is a percentage of like-adjacencies (nearness of the same land cover) among forest patches, where a single compact patch has a maximum aggregation. Aggregation refers to the degree of clumping of patches, where more aggregated patches facilitate inter-patch movement, promoting dispersal across the landscape. The mean inter-patch distance is the Euclidean or straight-line (edge-to-edge) distance among all forest patches in the landscape and it is a simple measure of patch context and has been used extensively to quantify degree of landscape isolation and act as a proxy for structural connectivity at the landscape scale. Finally, the largest patch index represents the percentage of the landscape comprised by the largest patch which can provide important habitat for forest-dependent birds and late-successional trees. We determined landscape metrics for 17 progressively larger circular areas around each plantation plot (radius from plot edge = 20, 40, 60, 80, 100, 150, 200, 250, 300, 350, 400, 450, 500, 600, 700, 800, and 900 m). All landscape metrics (i.e., tree cover, corridor cover, live-fence cover, fragmentation, remnant tree density, patch aggregation index, mean inter-patch distance, largest patch index) were obtained using the FRAGSTATS software56. Person to contact with questions about landscape structure data: Miriam San José – sanjosemiriam@gmail.co