Large-scale reforestation is promoted as an important strategy to mitigate climate change and biodiversity loss. A persistent challenge for efforts to restore ecosystems at scale is how to accelerate ecological processes, particularly natural regeneration. Yet, despite being recognized as an important barrier to the recovery of diverse plant communities in tropical agricultural landscapes, the impacts of dispersal limitation on natural regeneration in secondary forests – and especially how this changes as these forests grow older – is still poorly studied. In a region where animals have been shown to be the dominant seed dispersers, we evaluate the impacts of proximity to a connected network of narrow streamside strips of forest (SSF) on recruitment in 1–40-year-old secondary forests. We used eight years of annual census data from 45 sites with paired plots, one directly adjoining a SSF and the other further uphill (henceforth ‘landscape context’), and a null model approach to test the effects of proximity to SSFs and basal area, while accounting for variation in soil, topography, distance between plots and stand structure. In general, we found that landscape context affects multiple aspects of recruitment, including species diversity and the proportion of rarer and less-widely distributed species among the recruits. Unexpectedly, this effect did not weaken over time, despite a fast increase in stand basal area and diversity. This suggests that forest development over the first decades of succession may not be sufficient to attract the animals that disperse rarer tree species. Our results provide empirical evidence to guide restoration initiatives in agricultural landscapes in tropical regions, principally prioritizing the restoration of forest corridor networks along streams, while also highlighting the knowledge gap about restoring animal dispersers in secondary forests.

Study site

Our study was conducted as part of the Agua Salud Project (https://stri.si.edu/facility/agua-salud) that seeks to explain how different land-uses and forest restoration approaches impact ecosystem services in rural tropical landscapes. The study area covers over 750ha of land within a 15 km² area in the central Panama Canal watershed (9°13' N, 79°47’ W, 330 m amsl), adjacent to the 195.5 km² Soberania National Park, which consists of a mosaic of ≥ 80 year old secondary and old-growth forests and which is part of a much larger buffer zone that runs from the Atlantic to the Pacific coast along the Panama Canal. The forest in the region is classified as lowland tropical moist forest ecosystem. Animal dispersal is considered to be the dominant dispersal mode, with one study estimating that over 85% of the >400 tree species encountered in their study plots were animal dispersal in old growth and different-aged secondary forests. Annual precipitation averages 2700 mm per year, with a dry season from mid-December to early May. The area is characterized by an undulating topography, with short, steep slopes intersected by a dense network of narrow headwater streams. Soils are classified predominantly as Inceptisols, and are strongly weathered, infertile, and well drained. Topsoil texture (silty clays to clays) and soil nutrient concentrations vary little across the study area. Most hill slopes exhibit a downward concave shape, with mean angles of 19.6 ± 0.85 SE and 25.6 ± 0.70 SE degrees in the upper and lower portions of the hills, respectively. While there are in general no distinct riparian zones along the streams, across-site means of soil fertility and dry season soil water were slightly but significantly higher at downslope locations than at locations towards the top of the hill slopes, but with strong variation among sites. Within-site differences in forest structure and light availability did not differ significantly from zero. The landscape is dominated by active and abandoned cattle pastures and secondary forest of different ages.  Over at least the last 40 years (age of oldest plot in this study at the end of the study), farmers have typically maintained narrow strips of secondary forest alongside streams that cross or border their active pastures (henceforth, SSF). In pastures within the study area, these SSFs cover on average 14.3% (± 6.6 sd) of the total pasture area, are on average 12m wide and presumably act as corridors that connect many parts of the landscape to the SNP

Data collection

In 2009, 54 sites were randomly selected within the study area and information on land-use history and time since abandonment was acquired from interviews with former landowners and local residents and corroborated with satellite images. Sites were defined as a single slope within a secondary forest on an abandoned pasture where the landowners had removed cattle, and which were no longer grazed or cleared from vegetation. Two plots of 20 m x 50 m were established in each site; one plot near the bottom of the slope adjacent to the streamside strip of secondary forest (SSF) and another further from the SSF, near the top of the slope. Distances between plots varied between 28m and 149m in all but one site, in which the distance between plots was 315m (mean distance ± sd: 85m ± 41m). Because topography was complex, the shortest distance of the upper slope to the SSF was generally shorter than the distance between plots (mean ± sd: 69m ± 26m) (Suppl. Info. S4). The age of these SSFs was unknown but was always existing pre-abandonment and generally at least a decade older than the regrowing secondary forest on the hill slope (and thus the study plots). The age of the five oldest sites was not known (but > 50 years) and in four sites the two plots turned out to be of different ages. These nine sites were excluded from our main analyses, leaving us with 45 study sites with initial age ranging from one to 32 years.

In each plot, all stems of trees, shrubs, and palms with a diameter-at-breast-height (DBH) equal to- or more than five cm and all stems of woody climbing plants with a diameter of one cm or more were tagged, identified and their DBH was measured. In one-half of each plot, stems of trees, shrubs, and palms with a DBH of 1 to 4.99 cm DBH were also included. From 2009 until 2017, the growth and mortality of established plants and the recruitment of new plants, i.e., individuals that surpassed the size threshold since the previous inventory, was monitored annually, apart from 2013. 

Data on the floristic composition of streamside SSFs adjoining the study sites was collected in 2010. An inventory of trees ≥ 20cm DBH was performed in two-meter wide transects perpendicular to the stream, from the waterside to the border of the SSF at 5 m intervals over a total distance of 150m, at alternating sides of the stream. The average length of these transects was 11.8m (± 0.19 SE). the average width of the SSF was thus 23.6m. The mean combined sample area of the transects in the SSFs was 737m² (± 26 SE).

Seed dry mass data (g) was collected for 88 common species in the Agua Salud study area. For another 140 species, seed mass data was collected from trees in forest plots across the Panama Canal Watershed (Wright et al., 2010). Seeds were collected from 3 to 6 trees per species. Per species, seed mass data was calculated from 2 to 6 fruit replicates, with each replicate consisting of between one and >20 seeds, depending on seed size. Seed dry mass was measured without testa, unless seeds were too small to separate the testa from the seed. We acquired seed dry mass data for an additional 51 species from published databases . In the case of multiple data sources, priority of seed dry mass was given to measurements taken locally, followed by data collected in the region, to data collected at other sites. Seed mass data covered on average 80.7% (± 6.4 SD) of the species and 89.6% (± 7.1 SD) of individuals in the study plots.