The (Forest) Island Effect

Could planting trees in islands, rather than in rows, improve the rate of tropical forest recovery?

A study plot awaits restoration in Costa Rica

By Dr. Karen D. Holl, University of California, Santa Cruz

When one thinks about forest restoration likely the first thing that comes to mind is planting trees. This isn’t the only way to restore deforested land, but it is one of the most common. When we plant trees, we usually plant them in rows. But think about what a naturally recovering forest looks like in an area that was previously used for agriculture or livestock grazing. After a few years this land will typically sport scattered patches of young trees and shrubs in a highly disorganized, yet more organic pattern. As a restoration ecologist I have thought a great deal about how to design land recovery strategies that are both cost effective and as similar to natural processes as possible. Would planting trees in patches or islands rather than in systematic rows produce a more heterogeneous, resilient system? Could we reduce restoration costs by requiring fewer seedlings to grow, plant, and maintain?

Over a decade ago my colleague Zak Zahawi and I, along with the help of many students and field assistants, began an experiment in a premontane forest in southern Costa Rica to answer just these questions. We established 16 restoration sites and tested three restoration approaches at each site: “Passive recovery” where the forest was left to recover without any human intervention; “plantation” style tree planting where the entire area was planted with tree seedlings in even rows; and “island” style planting where just over a quarter of the area was planted with three different sizes of tree seedling clumps or “islands”. To evaluate the effect on recovery rates of having intact forests nearby, we established our experimental sites in locations with different degrees of surrounding forest cover. Some sites were situated in the middle of agricultural lands while others were adjacent to the largest forest remnants in the region. We hoped to learn not just whether different planting regimes can affect forest recovery, but also the effect of different locations for planting. In other words, is it more important how you restore, or where?

We have been following the recovery of our different test sites for ten years by collecting data, sometimes on a monthly basis, on seed dispersal, soil nutrients, seedling establishment, understory and canopy cover, bird and bat presence, mycorrhizae, and leaf litter insects. Tropical forests are highly diverse, and rates of recovery can vary greatly depending on the type of recovering forest, the intensity of past land-uses, and whether there are seed sources nearby (in other words trees and adjacent forests). In this study our aim was not to plant the full diversity of trees found in the forest but rather to provide a diverse vertical structure (e.g., vegetation of different heights) that could attract a diversity of birds and bats and shade out the tall pasture grasses that perpetually compete with tree seedlings for light and nutrients. Birds and bats are key to the vitality of a tropic forest, as they disperse the majority of tree seeds. Only through such faunal seed dispersal can a restoration site establish a real diversity of forest plants.

What we have found so far is striking: planting tree islands is cheaper than and as effective as traditional plantation-style restoration in accelerating the recovery of deforested land. Moreover, recovery in tree island treatments appears to better simulate natural processes.

What did we find exactly? First, we found that plantation and tree island restoration outperform passive recovery treatments with more than double the amount of seed dispersal and four times the number of seedlings established, two key indicators of restoration success. Second, we found that there is a critical minimum size (~100 m2) for tree islands to outperform passive recovery, particularly for them to effectively enhance bird activity, dispersal of forest tree seeds, and tree seedling establishment. Third, we found that planted islands expand in size over time due to both tree growth and new seedling recruitment at their edges. We also found that islands are expanding and coalescing at some of our sites. Finally, we found that planting tree islands results in greater canopy heterogeneity than the plantation-style approach.

Results to date also suggest that the amount of surrounding forest cover does not strongly influence animal-dispersed seed rain and native woody species establishment during the first 10 years of succession. For now it seem seems that restoration strategy is more important than site location within a landscape. But we anticipate that surrounding forest cover will have a progressively stronger effect on forest recovery as succession proceeds and a greater diversity of plants and animals colonize the sites.

We now think that the tree island restoration approach holds promise as an innovative restoration strategy for some systems, particularly where natural recovery rates are slow, large areas need to be restored, or restoration resources are limited. The irregular, tree island planting strategy does offer some logistical challenges over plantation planting, as most post-planting site maintenance (e.g., controlling competitive grasses) is done along even lines in planted forests. The farmers in our region were certainly more accustomed to this approach and this resulted in some trees being damaged in the island test sites during routine clearing of grasses. The most appropriate restoration approach at any site will depend on the goals and constraints of a project. Nevertheless, our findings show promising results and decreased cost from this new strategy – something that should be hard to ignore in most places attempting large restoration without large budgets.

Our next step will be to test this approach at additional locations and through large-scale restoration projects. We are currently planning a study comparing tree island plantings and planting in strips of trees alternating with open unplanted areas in the Atlantic forest of Brazil. We would be very interested in learning the results of others who have implemented similar forest restoration strategies. You can visit our website ( for selected publications and a broader discussion behind the science of this study. You can find the

This is a blog of the IUCN Science and Practice of Forest Landscape Restoration blog series. If you have new analysis or research you would like to share with a wider audience, please send us at note at

Work area: 
Forest Landscape Restoration
Go to top