Tree islands for tropical forest restoration: the outlook is rosy after 10 years

Planting tree islands has many of the benefits of larger plantations, but entails significantly less cost. Karen Holl (University of California, Santa Cruz), Leighton Reid (Missouri Botanical Garden), and Zak Zahawi (American University of Beirut) describe recent findings on tree seedling recruitment in a long-term experiment in southern Costa Rica.

Over the past few years there have been a growing number of commitments at the global, national and regional scale to restore forests because of their importance to conserve biodiversity, sequester carbon, reduce erosion, and provide goods and services to people. For example, Initiative 20×20, led by the International Union for the Conservation of Nature, aims to restore 20 million hectares of tropical forest by 2020, an area roughly equivalent to the size of Uruguay or Nebraska.

A common strategy to restore forests is to plant trees. But, the big question is: where will the money come from to plant billions of trees when there are so many pressing needs? As restoration ecologists, we started thinking about how we could most efficiently allocate resources to get the best bang for the buck and restore the largest area of forest.


Trade-offs in forest restoration strategies. Planting fewer trees leaves more to chance and can require more time, but tree plantations are more expensive and leave a bigger ecological footprint. Our study tests an intermediate option, and after 10 years it appears to provide a good balance. Figure modified from Corbin & Holl (2012).

Starting over 10 years ago, we set up a large-scale tropical forest restoration experiment in southern Costa Rica to test two ideas.

First, we tried planting tree “islands”. The idea is to plant groups of trees that attract birds and bats, which disperse most tropical forest tree seeds. The tree canopy also shades out light-demanding grasses that can outcompete tree seedlings. In one experimental treatment, we planted tree islands that covered about 20% of 50 × 50 m plot of former cattle pasture. We compared that to plots where no trees were planted (natural recovery) and to the more intensive (and more typical) restoration strategy of planting trees in rows throughout the plot (plantation).

Second, we asked: is it only possible to restore forest near remnant forests or can you restore forest anywhere in the landscape? This is important information to help guide forest restoration efforts. To do this we set up our entire experiment at 13 sites, some of which were mostly surrounded by agricultural land and some of which were adjacent to the largest remaining forests in the region.

Then we monitored establishment of new tree seedlings in our research plots over a decade. We compared the number of seedlings, number of species, and types of species in the restoration plots with those found in the nearby forest to evaluate how well the forest is recovering.

The tree island planting method not only saves money on buying, planting, and maintaining seedlings, but it also results in a more heterogeneous distribution of trees, so it looks more like a natural forest.


Profuse tree seedling and sapling recruitment in the understory between two tree islands in southern Costa Rica.

We counted over 6000 tree seedlings, 88% of which have seeds that are dispersed by animals. On average there were many more tree seedlings in the tree island and plantation treatments than in the natural recovery plots. These results suggests that some tree planting helps the forest to recover faster, but that it is not necessary to plant the whole area with trees. The tree island planting method not only saves money on buying, planting, and maintaining seedlings, but it also results in a more heterogeneous distribution of trees, so it looks more like a natural forest.

Even though there were many tree seedlings in the island and plantation plots, on average there were less seedlings of tree species that have big seeds (>0.5 cm/0.2 inches across) compared to mature, reference forests. It seems that the larger-seeded species that are common in mature forests are much slower to colonize restored sites, likely because they are eaten and dispersed by a small number of larger animals, such trogons and agoutis. Many of those dispersers are less likely to visit early successional forest.


Small frugivores, small seeds. Most of the birds we see in these experimental plots are small-gaped omnivores (e.g., Yellow-bellied Elaenia, Elaenia flavogaster, left), but it usually takes large-gaped species to disperse larger seeds1. The figure at right shows the maximum fruit size that a bird species with a given gape size was able to consume in a cloud forest in central Costa Rica (modified from Wheelwright (1985)). In our experiment, small seeds were ubiquitous, but large seeds were mostly absent.

We were surprised that the amount of forest cover around the experimental plots had a weak effect on the number of seedlings establishing. In other words, isolated plots had just as many tree seedlings as plots right next to old-growth forests. We think that this is likely due to the fact that there are many trees in the agricultural landscape surrounding our plots; these trees include remnant trees, living fence rows, and riparian corridors. Trees in the landscape can serve an important role in both providing sources of seeds and stepping stones for the movement of seed-dispersing fauna. We anticipate that having forest nearby will be more important in future years as these forests build up greater diversity of rare, large-seeded species. Nonetheless, our results suggest that there are good prospects for restoring forests in many locations in this landscape.

Our key finding is that planting tree islands can be a cost-effective way to restore tropical forests at our study site in Costa Rica, but we hasten to note that the strategy should be tested in other locations, particularly areas with fewer forest elements in the surrounding countryside. Our study also demonstrates that tropical forests can recover some species quickly but it will take many decades, if ever, for forests to fully recover. So, preserving existing rain forests is critical to conserve biodiversity and the services they provide to people.


Diverse tree cover in an agricultural landscape in southern Costa Rica. Remnant trees in pastures, trees along fence rows, and riparian forests provide important sources of flora and fauna to speed up forest recovery.

1See Melo et al. (2009) for an example to the contrary: small-gaped animals dispersing fairly large fruits and seeds.

Ecological Restoration in a Changing Biosphere

If you were at the MBG Fall Symposium, we want to hear from you! How did the symposium change your perception of restoration? Send us an email at

On October 8th, Missouri Botanical Garden hosted its 63rd annual Fall Symposium. This year’s theme was Ecological Restoration in a Changing Biosphere. Author and journalist Paddy Woodworth moderated the day, and seven speakers presented contemporary perspectives on a core challenge in modern restoration ecology. Namely: in the post-COP21 world, when all three UN conventions call for scaling up and mainstreaming of restoration, it is clear that restoration will affect hundreds of millions of hectares – and as many people – over the coming decade. At the same time, we find ourselves in an era of unprecedented change where climate, ecological baselines, and future land-use changes are highly uncertain. This raises the question: What should large-scale restoration look like in the remainder of the 21st century?


2016 Fall Symposium speakers. From left to right: Peter Wyse Jackson, Curt Meine, Robin Chazdon, James Aronson, Leighton Reid, Pedro Brancalion, Karen Holl, Don Falk, Paddy Woodworth, and Jim Miller. Photo by Andrea Androuais.

Talks during the morning focused on tropical forests, where much of the international restoration dialogue is focused.

  • Leighton Reid (Missouri Botanical Garden) opened with a presentation on restoration longevity – the idea that some restoration projects create ecosystems that persist for more than a century (e.g., Floresta da Tijuca), while other projects fail quickly. Dr. Reid argued that how long restored ecosystems persist is quantifiable, predictable, and manipulable, opening the possibility for more ambitious restoration planning.
  • Robin Chazdon (University of Connecticut and beyond) then spoke about forest landscape restoration, an approach that aims to regain ecological integrity and enhance human well-being in deforested, human-impacted, or degraded forest landscapes. Drawing on a wealth of large-scale studies, Dr. Chazdon made the case that natural forest regeneration is the most ecologically effective and economically feasible approach to forest restoration globally.
  • Karen Holl (University of California Santa Cruz) presented her take on research priorities for forest restoration in the Neotropics. She highlighted that researchers could make an impact by studying forest restoration at larger spatial scales, at longer temporal scales, and in collaboration with stakeholders. Improving information exchange and standardizing monitoring protocols were also among her top priorities. (Graduate students, take note!)
  • Dr. Pedro Brancalion (University of São Paulo) completed the morning session with a TED talk-style discussion of the linkages between science, technology, policy, and best practice in Brazilian Atlantic Forest restoration. Using Thomas Kuhn’s structure of scientific revolutions, Dr. Brancalion argued that restoration ecology is in a crisis period, in part because disciplinary research has predominantly created solutions at smaller spatial scales than the (growing) problems the discipline seeks to address. Perhaps restoration is ripe for a paradigm shift?

Dr. Pedro Brancalion (right) asks whether restoration ecology is ready for a new paradigm shift, as Paddy Woodworth (left) moderates. Photo by Robin Chazdon.

After lunch, the conversation turned towards a major academic debate in restoration ecology. Has global change outpaced the restoration approach? And is a new approach needed?

  • Curt Meine (The Aldo Leopold Foundation) drew on his long experience in the upper Midwest, and, in particular, his studies of author and environmentalist Aldo Leopold (1887-1948). He argued that Leopold avoided the simple polarities through which some contemporary restoration debates are framed. He viewed nature in a relative way, neither entirely wild, nor entirely domesticated in any given landscape. Although he practiced ecological restoration in some contexts, he also advocated soil conservation and sustainable agriculture – activities motivated by his core values, as expressed in The Land Ethic (1949).
  • James Aronson (Missouri Botanical Garden) followed with an elucidation of the reference ecosystem concept. Reference ecosystems, he noted, help determine the social and ecological vision for a restoration project or program – a critical issue for restoring historic continuity in degraded landscapes. Dr. Aronson described a family of restorative actions for achieving progress towards the reference system, drawing on examples from Jordan and South Africa. He argued we need to look deeper into the past and ponder our choices from many angles as we decide how to do more effective restoration at the landscape and larger scales.
  • Donald Falk (University of Arizona) delivered the keynote address. He painted a disturbing portrait: rapid climate change is driving a massive forest-to-non-forest transition in the southwestern United States. In particular, many ponderosa pine forests will not be able to persist in the future where they have been in the recent past and present. Perhaps restoration ecologists should transition too. Rather than “chasing the ambulance”, maybe we could get out ahead of disasters and ease transitions between stable ecosystem states. Anticipating ecosystem transitions could mitigate the loss of ecosystem functioning that accompanies major climate-driven forest fires, but it would require a shift in restoration thinking. Importantly, Dr. Falk noted that ecosystems do not care what words we use – ecosystems respond to actions.

With moderator Paddy Woodworth’s help, we finished the day with a panel discussion, inviting questions from the audience. Among the thoughts and questions that we were left with:

  • Is ecological restoration more difficult in places with greater population density?
  • Should restoration focus on policy, economic, or cultural motivations for engaging people?
  • Are values a better guide for land management than ecological history? Are the two complementary?
  • How can the reference ecosystem concept accommodate rapid biome changes, as we are seeing in the Southwestern USA?
  • What is the way forward to mainstream serious, multisectorial monitoring and evaluation with all these new factors to consider? Who will fund it?
  • To what extent can we move from restoring degraded ecosystems to avoiding degradation in the first place?
  • Can forest landscape restoration and natural forest regeneration bridge the gap between small-scale, past restoration experience and present, large-scale restoration needs?

PhD candidates Ricardo Cesar (University of São Paulo) and Leland Werdan (University of Minnesota) compare notes on seedling functional traits in dry tropical forest restoration. Leland was the recipient of the annual Delzie Demaree award. Photo by Robin Chazdon.


More than 150 people registered for the symposium. They came from three continents, five countries, and seven US states.

Environmental determinants of plant community change during restoration at Shaw Nature Reserve

Olivia Hajek spent 10 weeks this summer studying woodland restoration at Shaw Nature Reserve with CCSD scientist Leighton Reid. She participated in MBG’s NSF-funded Research Experience for Undergraduates (REU) program.


Wildflowers in the restored Dana Brown Woods: purple milkweed (Asclepias purpurescens; left) and buffalo clover (Trifolium reflexum; right).

During my ten weeks in Missouri, I completed a research project evaluating the role environmental conditions play in restoration at Shaw Nature Reserve.  Specifically, I worked in the Dana Brown Woods management unit, a part of the Missouri Ozark foothills that features diverse plant communities across its heterogeneous landscape.  Sixteen years ago, the Dana Brown Woods was a closed-canopy woodland highly invaded by eastern red cedar.  However, restoration practices including reintroduction of fire and mechanical removal of woody shrubs like eastern red cedar have dramatically changed plant communities since 2000.  I was very fortunate coming into this project because there was extensive data about the plant communities in the Dana Brown Woods from 2001-2012 while restoration was occurring.  A local botanist, Nels Holmberg, monitored understory plants beginning a year before the first fire, creating complete information about the plant community before restoration and as it changed over time.

We wanted to see how different environmental conditions affect how plant communities change over time in response to restoration.  To answer this question, we visited 300 points across the woodland and measured several environmental parameters, including aspect, slope, rockiness, elevation, and juniper stump density (juniper stumps decay slowly, so many of the trees cut in 2006 were still visible).


Fieldwork in Dana Brown Woods. Olivia makes friends with a hog peanut (Amphicarpaea bracteata).

Just from field observations, we could see noticeable differences in the environment and plant community composition across the woodland.  Higher slopes were rockier, covered in old juniper stumps, and rich in sunflowers, whereas the lower regions near the Meramec River floodplain had deeper soil and more mesic plant species, like spicebush.

Data analysis confirmed that environmental gradients moderated plant community change over time. Higher, rockier areas experienced greater plant species turnover and greater increases species richness and abundance from 2001-2012, whereas shaded valleys changed relatively little.


Plant composition change from 2001-2012 increased with elevation, particularly during spring surveys. BC = Bray-Curtis dissimilarity, which measures the difference in plant species composition between a plot in 2001 and itself in 2012. Juniper, red oak, and white oak were subjectively determined habitat classifications at the outset of the study.

Our observations were likely driven by differential fire behavior across the woodland. Historically, fires were a frequent disturbance in the Ozark foothills. Four prescribed fires from 2001-2012 probably had larger impacts on the drier upland areas than in the wet lowlands, which would not have burned as well.

Quantifying how ecological restoration practices, like prescribed fire, vary across environmental gradients is important for land management planning, especially in the Ozark foothills where the landscape is so heterogeneous.


Leighton stood by while Olivia presented her research to the public at Sensational Summer Nights.

Fig Stakes: Shoreline Restoration for a Costa más Rica

Andres Santana is the graduate program coordinator at the Organization for Tropical Studies. During a recent fieldtrip in southern Costa Rica, he and CCSD restoration ecologist Leighton Reid compared notes on using fig stakes for ecological restoration.

Tropical beaches are many things to many people. To plants, beaches are hot, sandy, and salty – complicating their restoration.

Costa Rica has 1228 km (763 mi) of coast line – including 1016 km on the Pacific side and 212 km on the Caribbean. Along Costa Rica’s northern Pacific coast, the beach forms the natural edge of the dry forest. Farther south the adjacent forest is more humid. Giant trees, 40 m or more in height, grow right up to the high tide mark, particularly along the Caribbean.

But as with so many tropical ecosystems, Costa Rica’s coastal forests have been subject to human impacts. Many shoreline forests were cleared for cattle ranching, and exotic grasses were introduced as forage. Some of these grasses are fierce competitors and prevent tree seedlings from establishing, even long after the pastures have been abandoned.

Playa Hermosa Antes y Despues

Playa Hermosa, before (left) and after (right) planting 2-m long cuttings of a coastal fig species (Ficus goldmannii).

In 2009, a small non-profit organization, Costas Verdes, was formed to restore coastal forests along degraded shorelines, particularly wildlife refuges. The restoration work was initially challenging; tree seedlings were hard to establish along the coast because of the harsh environment – high temperatures and salinity and lack of freshwater were among the most significant obstacles. Not to mention the invasive cattle forage grasses.


Coastal restoration at Playa Hermosa

Playa Hermosa, a surfing destination on the Central Pacific coast, was among the most heavily deforested project sites. This area, part of a wetland and river estuary, was declared a national wildlife refuge in 1998. By 2009, very little forest had naturally regenerated. This led Costas Verdes to implement a restoration project at this beach. Planting plots were established where invasive grass was removed. In other areas, grasses left intact, as a comparison. It quickly became evident that tree seedlings were outcompeted by the grass. Those in the cleared plots grew better, but they still faced the other coastal habitat challenges.

Some native trees are resistant to hot substrates and high salinity, but these species were not available in tree nurseries, most of which focused on ornamental species. This meant that seedlings needed to come from locally collected and germinated seeds. We realized that this would take time to get going. Tree seedlings under 50 cm rarely survive, even if they have the proper coastal adaptations.

To accelerate the restoration, we decided to use tree cuttings rather than growing seedlings from seed. A colleague suggested Ficus goldmannii as a candidate species, so in 2011 we conducted a planting trial. We planted 225 2-m long cuttings. Of these, 195 (87%) survived their first year. By the second year all 195 survivors had become established and were quickly providing canopy cover and lowering the temperature of the sand.


An established fig stake with a dense canopy. Note the weak, patchy grass below it.

Once fig stakes created some canopy cover, we brought in other tree species – mostly from the coastal tree nursery that we created. Shade from the fig canopy also began to inhibit the invasive grasses, which require high sunlight to photosynthesize efficiently. Reduced competition with these grasses allowed other tree seedling species to survive.

In this instance Ficus cuttings turned out to be useful in promoting restoration. We have since used cuttings for other plots with similar success.


Coastal trees and shrubs growing below established fig cuttings at Playa Hermosa.

Can the Ozark chinquapin successfully re-colonize interior highland forests?

Jenn Rosen is an undergraduate at the University of Missouri – Saint Louis. This summer, she participated in the Missouri Botanical Garden’s Research Experience for Undergraduates (REU) program. She spent eight weeks working with scientists in the Center for Conservation and Sustainable Development. Here, Jenn writes about her field experiment at Shaw Nature Reserve.

Conservationists and ecologists have recently sought to restore an ecologically important tree, the Ozark chinquapin (Castanea ozarkensis), that became threatened not by the effects of clear-cutting, but instead due to an incurable parasitic fungus called chestnut blight (Chyphonectria parasitica). After devastating its cousin (American chestnut) in eastern forests, the fungal disease moved west and began infecting chinquapin trees in the Ozark and Ouachita Mountains. The blight is presumed to have been brought into the United States by accident in the early 1900s when Chinese chestnuts (C. mollissima) were imported into the country. What makes the Ozark chinquapin and the American chestnut so susceptible to the disease is that the fungus’ wind-borne ascospores can easily enter through small cavities in their bark where they then grow and spread to neighboring trees. The blight causes Ozark chinquapin trees to die back to the roots, from which multiple stems resprout to form a large shrub-like growth form. Once a widely dispersed canopy tree in upland Interior Highland forests, chinquapins declined in abundance and are now often found as multi-stemmed, blight-infected subcanopy shrubs.

Castanea ozarkensis seeds from two maternal lines. One group's seeds are twice the size of the others. Large seeds hold more resources, which can give seedlings a head start. But larger seeds also face a higher risk of being preyed upon by rodents.

Castanea ozarkensis seeds, with white radicles, from two maternal lines. One group’s seeds are twice the size of the others. Large seeds hold more resources, which can give seedlings a head start. But larger seeds also face a higher risk of being preyed upon by rodents.

Chinquapins were prized by many folks of the Ozarks for the nutrient-rich nuts they produced, and its rot-resistant woods were used to make furniture, railroad ties, and fence posts, among other products. Additionally, large mammals such as black bears were known to forage in the Ozarks in search for the nutritious nuts to help replenish their fat reserves for the upcoming breeding season. There is a collaborative effort among groups to restore the chinquapin throughout its’ former range once blight-resistant seeds become widely available, estimated to take approximately 20 – 30 years.

Once blight-resistant seed becomes available, understanding how chinquapin trees successfully regenerate in the wild will be the key to successfully restoring this species. Unfortunately, little is known about the ecology of the tree prior to blight infection. Like other nut-bearing trees in Ozark woodlands (e.g., oaks), we suspect that there may be several limiting factors to chinquapin seedling recruitment, such as seed predation, poor soil quality and/or poor light availability. In order to test these predictions, we planted 320 chinquapin seeds (from two distinct maternal origins from the wild) across ten experimental replicates at the Shaw Nature Reserve (Gray Summit, Missouri). The criteria for each replicate was that there had to be a shrub microhabitat, which was dominated by the common understory tree, eastern redbud (Cercis canadensis), and an open habitat, separated by at least three meters. Like other Ozark woodlands, prescribed fire is currently being used to restore woodlands at the Shaw Nature Reserve and will likely be a key component to successfully reintroducing chinquapins back into the wild. After enduring many scrapes and pricks from constructing mammal exclusion cages for half of the seeds, roughly two weeks after planting the seeds we had a bounty of little chinquapin seedlings emerge.

Castanea ozarkensis seedling, protected from marauding rodents by a wire cage.

Castanea ozarkensis seedling, protected from marauding rodents by a wire cage.

We found that consumer treatment and microhabitat structure, as expected, influenced the rates of Ozark chinquapin seed emergence. Nearly all of the seeds that were exposed to small mammals were eaten or removed, even though seeds were buried a few centimeters in the soil. Interestingly, small mammals consumed seed at greater rates in shrub than open microhabitats. These results imply that understory vegetation structure determines where chinquapin seedlings can successfully recruit through its influence on small mammal behavior. Also, small mammals removed larger, and presumably more nutritious, seed at greater rates than smaller seed. Environmental factors, like light availability, did not affect seedling growth, but the short duration of the study may not have been adequate for the potential influence of these factors to become apparent.

Our work contributes to a larger ongoing project by the Ozark Chinquapin Foundation to restore and conserve Ozark chinquapins. Given the high rates of seed removal, future restorationists will have to transplant chinquapin seedlings as opposed to seeds to successfully reestablish this species in the wild. However, once reintroduced seedlings grow to maturity and produce seed, our study suggests that microhabitat structure in Ozark woodlands will play a key role in determining the recruitment and growth rates of restored populations.