1. In degraded tropical landscapes lack of seed dispersal can strongly limit recovery, and restoration interventions can overcome this barrier by attracting dispersers. However, seed dispersal patterns are typically studied over short time periods, thus the influences of temporal and spatial variability on seed arrival cannot be teased apart.

2. Choice of management approach can have important implications for restoration-mediated community reassembly. Accordingly, we used a 3.5-year record of seed deposition in premontane tropical wet forest in southern Costa Rica to examine how seed arrival differed between passive (natural regeneration) and active (applied nucleation, plantation) restoration after a decade of recovery, compared to remnant forest. We investigated: (1) how restoration treatments affected seed deposition rates and community composition; (2) how within-plot heterogeneity of animal-dispersed seed deposition varied by intervention; and (3) how interannual variation influenced animal-dispersed seed arrival across treatments.

3. Overall seed rain composition and diversity in restoration treatments is converging towards, but still differs substantially from, remnant forest (89.7, 86.6, and 76.3% Shannon diversity recovered in applied nucleation, plantation, and natural regeneration, respectively).

4. Within-plot animal-dispersed seed heterogeneity was similar in applied nucleation and remnant forest, 27.0% more heterogeneous in applied nucleation than plantation, and equivalent when comparing natural regeneration to individual active restoration treatments.

5. In contrast to active interventions, animal-dispersed tree and shrub communities did not differ year to year in natural regeneration, which may promote the assembly of relatively homogeneous plant communities at this successional stage.

6. Synthesis and applications: Compared to natural regeneration, active restoration interventions: (i) catalyzed recovery of seed diversity (overall Shannon diversity 17.5 and 13.4% higher in applied nucleation and plantation, respectively), (ii) shifted seed community composition towards remnant forest more rapidly (overall Shannon diversity 13.4 and 10.2% closer), (iii) almost doubled the proportion of later-successional tree species arriving, and (iv) had seed communities that differed year to year – a pattern not observed in natural regeneration. Finally, applied nucleation was the only intervention where seed arrival was as spatially heterogeneous as remnant forest, highlighting that this approach may facilitate the recovery of specific natural dispersal processes.

Description of data collected:

Site and study description

We conducted this study at five ~1-ha sites in southern Costa Rica, a subset of the 13 replicate restoration sites of this long-term experiment, located between the Las Cruces Biological Station (8° 47' 7" N, 82° 57' 32" W) and the town of Agua Buena (8° 44' 52" N, 82° 56' 39" W). At the start of the study, from 2004-2006, we established three 0.25-ha (50?50 m) plots separated by a ≥5 m buffer at each site. Three restoration treatments: natural regeneration, applied nucleation, or plantation, were applied to one of the plots at each of the sites. At three of the five sites selected for this study, equivalent sized plots were established in adjacent remnant forests to serve as references

Seed rain data

We measured seed rain twice monthly from May 2013 - November 2016 (3.5 yr) in four subplots (separated by ≥10 m) within each treatment plot, referred to as sampling ‘stations.’ At each station three 0.25-m2 seed traps with 0.69-mm mesh were placed at randomly-selected ≥1-m intervals (e.g., at 1.0, 2.5, and 4.5 m) along transects. In total, 216 traps were monitored: each restoration treatment had 60 seed traps (5 sites X 4 stations X 3 traps) and reference forests had 36 traps (3 sites X 4 stations X 3 traps).     

For each sampling period, seeds were collected from each trap, stored separately in paper envelopes, and then dried at 65°C to preserve until processing. Seeds were counted and identified to species using a reference collection at Las Cruces Biological Station. We excluded grass seeds (Poaceae), as they are not a major component of mature forest flora. Less than 0.01% of seeds were identified to only family or morphospecies. Using local knowledge and literature reviews, we classified species’ dispersal syndrome (ectozoochorous, endozoochorous, anemochorous, explosive, or gravity), categorized species as early or later-successional, and grouped species into ten growth forms: two herbaceous (forb, vine), eight woody groups (liana, shrub, understory palm, understory tree, canopy tree, emergent tree), and epiphytes (both herbaceous and woody species).

Preparation for analysis
We summed species-level seed(s) observations for each each sampling station for each year (rows in the data set). 
 

Species-level seed rain data summed annually at each sampling station - WerdenEtAl2021_JAE_seedRainDat.csv 

For more details see the attached ReadMe file - WerdenEtAl2021_JAE_seedRain_ReadMe.txt

Missing values: The seed length column (seedLength) has missing values (represented with NAs) as seed length was not recorded for all species.