Conservation Evidence strives to be as useful to conservationists as possible. Please take our survey to help the team improve our resource.

Providing evidence to improve practice

Individual study: Bird perches increase forest seeds dispersal on landslides at the Luquillo Experimental Forest, Puerto Rico

Published source details

Shiels A.B. & Walker L.R. (2003) Bird perches increase forest seeds on Puerto Rican landslides. Restoration Ecology, 11, 457-465


In hilly regions of Puerto Rico, landslides are part of the natural forest disturbance regime but about half are associated with roads as, during their construction, areas are denuded of vegetation and thus become more vulnerable. Landslides result in further vegetative loss, loss of soil and the soil seed bank. These losses make it difficult to alleviate soil erosion and to restore rainforest communities. In this study bird perches were added to six Puerto Rican landslide sites in an effort to facilitate inputs of forest seeds through bird dispersal and to accelerate plant succession.

Study site: This study took place on six landslide sites (450 - 650 m a.s.l.) in the 110 km² Luquillo Experimental Forest (LEF), northeastern Puerto Rico. Annual precipitation ranges from 3,000 to 4,000 mm with high year-to-year variation and little seasonality.

The natural vegetation at the study sites is subtropical lower montane wet and rain forest. Common landslide-colonizing vegetation include several grasses (e.g. Andropogon, Paspalum), thicket-forming climbing ferns (Dicranopteris, Gleichenia), other ferns (e.g. Cyathea, Nephrolepis), herbs (e.g. Nepsera, Sauvagesia) and trees such as Cecropia and Tabebuia.

The six chosen landslides comprised two from each of the following three dominant ground cover types: Gleicheniaceae climbing fern, grass, and bare (>80% exposed soil). The following were recorded: Dominant vegetation surrounding each, elevation, area of the landslide, aspect, slope, and light intensity.

Perch addition: On each landslide four circular plots (2 m diameter; 3.14 m²) were established, at least 5 m from the forest edge to the plot centre. Two plots were randomly assigned perches and two served as controls. Perches were made from tree saplings cut from near each landslide. The saplings were stripped of any leaves, flowers and fruits. Perches (4-4.5 m tall) were held in place by burying approximately 0.7 m of the lower section in the ground and stabilizing it by attaching guide wires. All perches were thinned to have about equal branching (17–18 branches >1 cm diameter at the main stem). Three tall naturally occurring saplings were trimmed to 1.3 m height to avoid problems which might confound data collection.

Seeds: Within each perch and each control plot, ten 33 × 23.5 cm subplots (0.0776 m²) were randomly assigned to either trap seeds or record established seedlings (see below). Five seed baskets were placed on the ground in each plot, secured with a stake, and lined with nylon cloth to trap seeds. These were covered with 1.3 cm diameter metal mesh screening to exclude rodents. Only about 5% of all seeds in the study area were more than 1.3 cm in diameter and were thereby excluded from the analysis. study. Seed baskets were visited every 3 weeks between 14 June 2000 and 17 August 2001. All seed at least 2 mm was collected and identified, and each categorized as bird or wind dispersed. Because of the large quantity of seeds less than 1 mm diameter in Miconia fruits, these fruits were counted as a single bird-dispersed seed if found more than 50% intact.
Seedlings: Each of the five seedling subplots, was visited in July 2000 (≤45 days after perch addition), August 2000, January 2001, June 2001 and August 2001, and the number of seedlings of bird-dispersed seeds were counted. All bird-dispersed seeds and seedlings of bird-dispersed seeds in this study represent forest species.

Bird observations: Bird observations were undertaken from 1 June 2001 to 15 August 2001 at the six landslide sites. Observations were conducted from either 5:30 to 8:00 am or 12:00 to 2:00 pm. The morning period was the main sampling period as this was when there was most bird activity. However, less frequent mid-day sampling was conducted to account for potential species and/or visitation differences. For each observation, the species and total visitation time were recorded. Each was assigned a location category: in the landslide area on a provided perch, on a natural perch, on vegetation at the forest edge ( ≤15 m into the forest), or flying over the landslide. The 3-month period of bird observation was considered likely to be representative, or an underestimate, of average daily bird visitation given that all the birds observed were non-migratory. Additionally, a past bird study suggested that the resident bird populations do not change seasonally except in severe hurricane conditions. Over the elevational range covered, the avian species diversity was known to be similar.

Seeds: Over the 14 month study, 21,507 seeds (≥2 mm) were collected in plots at the six landslides. Approximately 1% (222 seeds) were bird dispersed, of which the highest proportion attributable to a single species (Schefflera morototoni) was 89 seeds. The average number of seedlings of bird-dispersed seeds that colonized was 1.5 (± 1.3) individuals/m2 in controls and 5.2 (± 2.1) individuals/m² in perch plots. Averaged across all six landslides, perch plots had significantly more bird-dispersed forest seeds than controls, with only two bird-dispersed seeds found in the control plots over the whole of the 14 months. Perches also increased species richness as 14 of the 28 total species collected were bird dispersed and only found in perch plots, whereas 12 species were wind dispersed and found in both perch and control plots. However, variation among landslides was high. Landslides with grass-dominated ground cover tended to have more bird-dispersed seeds below perches compared with both climbing fern-covered and bare landslides.
Most wind-dispersed seeds were from two grasses, Andropogon bicornis and Paspalum conjugatum. Inputs of bird-dispersed seeds changed seasonally, with most seeds appearing during the wet season. Landslides varied widely with respect to wind-dispersed seeds but with no significant difference among vegetation cover types.

Birds: During the study, 22 bird species, 16 of which consume fruits and/or seeds were observed. Bananaquit Coereba flaveola, black-faced grassquit Tiaris bicolor (the most common visitor to perches at four of the six landslides), scaly-napped pigeon Columba squamosa (all potential seed-dispersers) and Puerto Rican emerald hummingbird Chlorostilbon maugaeus were found at all six landslides. Seven (six of which consume fruits and/or seeds as part of their diet) of the 22 species were observed using the perches, but overall birds were observed most commonly perching at the forest edges.

Morning observations of birds on perches (1.46 birds/hr) were similar to afternoon observations (1.67 birds/hr). The average number of perch visitations varied but was not significantly different among landslides although bare ones tended to have lowest visitation rates (with a single stripe-headed tanager Spindalis zena the only observed visitor to perches on the bare landslide). The average time birds occupied perches was 1.1 (± 0.3) minutes per visit, this did not differ among landslides or their cover types.

Conclusions: Numbers of bird-dispersed forest seeds were significantly higher in plots beneath perches than in controls. However, provision of bird perches did not increase forest seedling densities compared with the control plots but species richness was enhanced. Although the perches had the effect of increasing seed inputs from forest species which might accelerate revegetation of landslides, the authors suggest that supplemental restoration techniques should be applied in addition to bird perches to promote forest recovery.

Note: If using or referring to this published study, please read and quote the original paper. Please do not quote as a case as this is for previously unpublished work only. The original paper can be viewed at: