Tropical forests are highly diverse ecosystems located within 23.5º N or S of the equator in Asia, Oceania, Africa, and Central and South America. They are found in areas with relatively warm and constant temperature from sea level up to 3000 m elevation. Rain forests receive >1500 mm of rainfall annually, whereas seasonal or dry tropical forests receive less rain and have a distinct dry season. Large areas of tropical forests have been cleared due to a complex set of causes which vary by region; these include clearing forest for pasture to graze cattle, logging for wood exports, collection of firewood, commercial and subsistence agriculture, growing crops for biofuels, and human-caused fire (reviewed in Geist & Lambin 2002). Over half the tropical moist forest cover worldwide has been cleared (Asner et al. 2009) and much of the remaining forest is affected by fragmentation, selective logging, and overhunting of large fauna. Together, these have contributed to extensive loss of biodiversity and result in greater than 12% of annual global carbon dioxide emissions (van der Werf et al. 2009).

Although large areas of tropical forest have been cleared during the last 20 years and extensive deforestation continues in some regions, there has been an increase in tropical secondary forest due to primarily passive restoration (i.e., natural regeneration), but also to active restoration (Lamb et al. 2005, Wright & Muller-Landau 2006, Chazdon 2008). The interest in restoring tropical forest has grown substantially in the past few years with increasing efforts to reduce anthropogenic carbon emissions caused by deforestation and forest degradation, as well as increase carbon stocks by restoring and better managing forests (Elias & Lininger 2010). Tropical forest conservation efforts should first focus on conserving relatively intact tropical forests to maintain the ecosystem services (biodiversity, carbon sequestration, erosion control) provided by these forests, as restoring forest is not a substitute for conservation. But in areas where large areas of forest have already been cleared, restoring degraded areas can help to restore both ecosystem services and biodiversity.

Ecology of Forest Recovery

Selecting a restoration strategy, for tropical forests or any ecosystem, must be based on a thorough understanding of the ecology of the system. The pace and direction of ecosystem recovery following human disturbance varies greatly across tropical forests systems. Tropical forest plants regenerate either from sources from within the site (seed bank, resprouting from roots and stems, or remnant vegetation) or colonization by seeds dispersed from outside the site. Recovery is much slower in sites where seeds must colonize from outside the site, as seed dispersal by animals is the primary dispersal mechanism for tropical trees and most seed-dispersing animals (in particular birds) are unlikely to move from forest into open agricultural lands (Holl 2007). Once seeds arrive at the site they must be able to germinate, survive and grow. Several factors limit seedling establishment, including aggressive pasture grasses, stressful microclimatic conditions (high temperature and light, and lower humidity), limited soil nutrients, and high seed predation/seedling herbivory (reviewed by Holl 2002).

The importance of these different processes in limiting forest recovery in former agricultural lands in the tropics varies greatly from site to site and depends on several factors, including the natural resilience and adaptations of the system to disturbance, the past land-use history, and the surrounding landscape matrix (Figure 1, see Holl 2007, 2012). Some tropical forests systems naturally recover more quickly. For example, systems in which many species can resprout from roots and species are predominantly wind dispersed tend to recover more quickly than those that rely heavily on animal-dispersal of seeds from sources outside the site (Figure 2, Holl 2007). Moreover, seedling growth tends to be faster in lower elevation sites that are warmer and have higher rainfall.

Past land-use history strongly affects the rate of recovery. Areas that have been used for pasture or industrial scale agriculture for many years recover more slowly than those used for shifting agriculture and shorter periods of time. In more intensively used sites, few or no forest seeds remain in the soil and resprouting is generally lower, both of which limit regeneration from within the site. Moreover, compacted and nutrient poor soils, and stressful microclimatic conditions limit seedling establishment and growth in heavily used sites. In former pasture sites, aggressive grasses often severely inhibit survival and growth of forest seedlings (Figure 2, Holl 2007).

Finally, the surrounding land-use matrix influences the availability of seeds. Not only is proximity to remnant forest important to recovery as a source of plants and animals, but maintaining some tree cover in agricultural lands through agroforestry systems or hedgerows provides seeds of some trees species and facilitates the movement of animals (Harvey et al. 2008). Therefore, management of the agricultural lands surrounding a site strongly influences the rate of recovery within a given site. To a lesser degree, surrounding land uses may affect species establishment through the movement of nutrients and agrochemicals into the site, as well as herbivores and pathogens.

Selecting a Restoration Strategy

Given the large areas of tropical forest that have been cleared and the limited resources available for restoration, it is critical to document the rate and direction of forest recovery and assess factors limiting recovery in order to select the most efficient restoration strategy (Holl & Aide 2011). Moreover, it is important to consider the goals of the restoration project, which might range from maximizing carbon sequestration to restoring the full composition of species to providing habitat for a specific faunal species of concern.

In sites where natural regeneration is rapid, passive restoration (i.e., simply allowing the system to regenerate naturally) may be sufficient to restore the majority of species present prior to disturbance (Janzen 2002, Letcher & Chazdon 2009). In such sites, it is common that large-seeded and later successional species are the slowest to colonize, so restoration efforts should focus on planting seedlings or seeds of such species, particularly when they are of conservation concern (Martínez-Garza & Howe 2003).

In sites where recovery is initially slow, another approach is to identify whether there are barriers to establishment that can be removed with comparatively low effort, an approach often referred to as assisted natural regeneration (Shono et al. 2007). One example of this approach is controlling fire during the dry season, which strongly inhibits seedling establishment in tropical forests (Janzen 2002). Another is marking all naturally regenerating woody seedlings and then clearing surrounding grasses to reduce competition and fire risk (Shono et al. 2007). Not only are these efforts to remove obstacles to natural regeneration without actively planting or seeding much cheaper, they usually leave less of a human imprint on the long-term species composition of the resulting forest (Lamb et al. 2005).

In areas that recover more slowly and are dominated by pasture grasses or other agricultural weeds, the most common restoration strategy is planting tree seedlings (Lamb 2011). Once established, these trees serve to attract seed dispersers, shade out pasture grasses, and improve soil and microclimatic conditions, thereby facilitating recovery (Figure 3, Holl 2002). Historically, most reforestation in the tropics utilized a few exotic tree genera (e.g., pine, eucalyptus, teak), but in the past decade or two there has been extensive research on both screening and developing propagation methods for native species suitable for including in restoration efforts. Strategies for planting native tree species vary greatly with respect to species diversity and composition, method of introduction (e.g., seedling, cutting, or stakes), and spatial distribution of plantings.

Many restoration efforts plant a small number (<10) of tropical tree species to facilitate colonization and establishment of the highly diverse native flora and fauna (Figure 3), whereas others plant more (20–30) tree species that represent a range of growth rates and dispersal guilds (Lamb 2011). It is much less common to plant >30 species given the necessary resources and knowledge for propagation, although some restoration efforts strive for diverse plantings (e.g., 60–80 species, Figure 4, Rodrigues et al. 2009) that include vines and shrubs. Most commonly trees are planted as seedlings, but some species can be propagated by cuttings or by direct seeding, which are less expensive and do not require greenhouse facilities (Zahawi & Holl 2009, Cole et al. 2011). Seedlings are generally planted in rows of 2–4 m at the outset, often interplanting trees with different growth rates, and then thinned as trees mature. Current research is exploring the idea of planting patches of trees, which is cheaper, better simulates the natural recovery process, and creates more heterogeneity (Cole et al. 2010).

Assisting natural regeneration by removing stresses or by planting native tree seedlings are the most common strategies for restoring tropical forests. Several other approaches to restoration have been tested in scientific experiments, but these approaches rarely have been applied at a larger scale and have met with mixed success. Some of these include installing perches to encourage seed dispersal by birds, using bat boxes or certain odors to attract bats, and placing piles of woody debris to facilitate seedling establishment and provide cover for small mammals (Holl 2012).

Broader Considerations

It is promising that some tropical forests recover rapidly from disturbance, and that a range of restoration approaches have been demonstrated to facilitate tropical forest recovery. Nonetheless, in most cases these secondary tropical forests still have lower biomass (above- and below-ground) and species richness than intact tropical forest. Therefore, the first priority should be to conserve relatively intact tropical forests. When tropical forest has been destroyed, clearly identifying restoration goals and understanding the ecology of the system will help in selecting the most appropriate and cost effective restoration strategy, which may be simply allowing for natural recovery.

It is important to note that tropical forest restoration efforts to date have focused on restoring a semblance of the predisturbance condition. But human-induced climate change, including altered patterns of temperature, precipitation, and cloud cover, will certainly strongly affect tropical forests. This means that in the tropics, as elsewhere, restorationists will have to select species to plant, considering both the physiological tolerances of plants and animals to a changing climate and how these tolerances, along with limited capacity for dispersal, will affect ecosystem processes, such as carbon sequestration, and biotic interactions in the future.