Fragmentation and habitat destruction of tropical forests is nowhere more apparent than in the seasonally dry tropical forests (SDTFs) of Central America (Janzen 1988b; chap. 1). In Central America, old-growth tropical dry forest had been reduced to less than 20 percent of its original extent by the mid 1980s (Trejo and Dirzo 2000), largely as a result of disproportionately high human population density and intensive agricultural activity within this habitat zone (Murphy and Lugo 1986a). Although rates of deforestation in Central America peaked in the twentieth century, palynology data indicate that humans have been using fre to manipulate forest cover in Central American SDTF for thousands of years (Janzen 1988b; Piperno 2006).

Although preserving the few remaining stands of old-growth SDTF in Central America is of critical importance to biodiversity conservation in the region, the remaining stands of intact, mature forest may be too small and isolated from one another to preserve this system (Bierregaard et al. 1992; Laurance, Lovejoy et al. 2002; Laurance et al. 2006; chap. 12). The conservation biology of Central American SDTF, therefore, must be in large part restoration biology (Janzen 1987).

The challenges facing restoration efforts in SDTF are manifold. Much of the tropical conservation literature has focused on the status and importance of tropical rain forest (Quesada, Sánchez-Azofeifa et al. 2009). This may be due to the fact that old-growth SDTF is becoming rare in the Neotropics (chap. 3). The very rarity of SDTF hinders scientifc work on forest regeneration, as there is very little high-quality SDTF to study. Without an understanding of how ecological processes relevant to regeneration operate in intact forests, it is challenging to determine how they are altered by fragmentation. Another challenge is that SDTFs are dominated by secondary growth (Quesada, Sánchez-Azofeifa et al. 2009). Abandoned agricultural felds face particular restoration challenges, such as the presence of exotic ungulates or a history of management by intense and frequent fire. Historical land use has resulted in great variation in species composition and forest characteristics of current SDTF, from closed-canopy evergreen forests to scrub savannas dominated by exotic pasture grasses. Variation in climate and soil properties also contributes to structural variation in SDTF. Even relatively undisturbed SDTF will show a high degree of variation across sites within regions due to the amount and pattern of rainfall and soil properties (Powers et al. 2009).

Despite the rarity and continued loss of SDTF, restoration of SDTF has rarely been attempted on a large scale. Worldwide, most restoration efforts have been passive and undocumented in the scientific literature (e.g., Fajardo et al. 2005). However, several regions in Latin America are well described. Colón and Lugo (2006) demonstrated that the outcome of 50 years of passive restoration of SDTF in Puerto Rico was highly dependent on past land use patterns. Calvo-Alvarado et al. (2009) found evidence for major recovery of forested area since the 1980s in Guanacaste province, Costa Rica, although they concluded that this was mostly a result of economic and social changes in the region and had little to do with conservation policy. Nevertheless, deforestation of SDTF continues outside park boundaries. Sánchez-Azofeifa, Daily et al. (2003) showed that outside of the network of protected areas in Costa Rica, deforestation has not declined. Similarly, substantial forest regrowth has occurred within the Chamela-Cuixmala Biosphere Reserve, but beyond the boundaries a high degree of fragmentation continues (Sánchez-Azofeifa et al. 2009).

Although few studies document restoration of SDTF, many studies have identified factors that limit forest regeneration These include ungulate exclusion (e.g., Cabin et al. 2002), a complex issue because ungulates may either facilitate regeneration (by reducing fuel loads in fire-prone areas; Janzen 1986a) or retard it (by grazing or trampling establishing seedlings; Cabin et al. 2002). The use of “framework” or “nuclear” trees to reduce grass cover and provide habitat for seed dispersers has also been advocated, but species selection requires an understanding of both the growth of the framework species and the seedling community that establishes beneath it (Elliot et al. 2003). The role of seed rain and the distribution and germination of seeds outside of fragments (Holl 1999) as well as appropriate timing of reseeding efforts (e.g., Mascia Vieira et al. 2008) have also been identified as critical factors in regeneration of SDTF. One common thread among all of these studies, however, is that the parameters for restoration discovered from this bottom-up approach depend largely on local conditions. As a result, it is unlikely that general solutions to SDTF restoration will be discovered. In an instructive example, Griscom et al. (2009) showed that while removal of cattle was beneficial to restoration, herbicide application to control exotic grasses only improved outcomes in sites that were far from remnant forest fragments. Inappropriate restoration techniques may even have a negative impact on “natural” regeneration. Sampaio, Holl et al. (2007) showed that in Brazilian cerrado, passive restoration (preventing disturbance) is superior to some active restoration techniques (those that involve mowing or plowing before planting nursery-grown seedlings). As a result, they argued that reseeding efforts should be targeted and limited to species not naturally regenerating in the area of interest.

SDTFs in Guanacaste and the Area de Conservación Guanacaste

The Area de Conservación Guanacaste (ACG) is a 110,000-hectare conservation area located in northwestern Costa Rica. About 50,000 hectares of its area are SDTFs, averaging about 1500 millimeters of rain per year. It experiences a 6-month dry season running from December to May of each year. Unpredictable weather events such as hurricanes cause a high degree of year-to-year variation in rainfall, although it is unusual for more than a few millimeters of rain to fall during the dry season.

Throughout Guanacaste there has been widespread conversion of SDTF to pastureland for the grazing of cattle and selective logging for valuable species such as lignum vitae (Guayacan sanctum) and mahogany (Swietenia humilis). Tropical dry forest in the ACG now exists as a complex mosaic of SDTF and pastures in varying states of regeneration. True old-growth SDTF is restricted to a few very small patches, which have been selectively logged. There also exist extensive stands of SDTF that are more than 100 years old that support high tree and animal diversity.

The core area of the ACG, Parque Nacional Santa Rosa, was declared a national park in 1971 to protect high-quality tropical dry forest habitat (Janzen 1986b). Throughout the 1970s, 80s, and 90s, the main external threat to this conserved wildland and the remnant forests surrounding it was anthropogenic fire used to maintain cattle pastures. These fires were devastating to remnant forest patches because of high fuel load from the introduced pasture grass species jaragua (Hyparrhenia rufa). Following the establishment of the national park, fires set by neighboring landowners encroached on the protected area, causing persistent degradation of remnant forest stands.

Since the mid 1980s a fire control program within the ACG has reduced the impact of fire on SDTF regeneration tremendously, limiting the area burned to fewer than 2000 hectares per year. Since active fire suppression began in the 1980s, rapid recolonization of abandoned pastures by woody vegetation has been observed in much of the seasonal ACG (Janzen 1986a, 1988d).

Efforts to actively manage regeneration of SDTF in the ACG have been minimal. However, Guanacaste SDTF appears to be amenable to passive restoration. SDTF soils are more fertile than many tropical soils, and many species grow well even in soils taken from beneath pasture vegetation (Klemens 2003). Also, the major introduced species, jaragua (Hyparrhenia rufa), does not outcompete woody vegetation directly but persists because it tolerates fire better than woody species (Daubenmire 1972a). Many SDTF species are therefore able to establish in the middle of a jaragua pasture, including many “old-growth” species (Janzen 1988b; Gerhardt 1993), although subsequent survival in the pasture environment seems to be highly species specific (Gerhardt 1993) and has not been monitored in a comprehensive way.

Quercus oleoides in Guanacaste

Our work focuses on the regeneration of the tropical live oak, Quercus ole-oides, in the ACG. Although the ACG is one of the “success stories” of SDTF restoration, the regeneration of Q. oleoides forests is thought to be severely “limited” under the current regime of passive regeneration, compared with other taxa in the regional species pool. These mono-dominant forests formed a unique SDTF ecosystem that once covered tens of thousands of hectares of Guanacaste Province prior to the severe forest fragmentation of the last century (Boucher 1981; Janzen 1987). Structurally and ecologically important species that do not respond to a particular regeneration regime are a major challenge to restoration in SDTF. Despite being formerly widespread, Q. oleoides is something of a biological enigma here: the local population is geographically disjunct, physiologically differentiated, and genetically distinct from conspecifics and similar species (J. Cavender-Bares, A. Pahlich, A. Gonzalez-Rodriguez, and N. Deacon, unpublished data; Cavender-Bares 2007). Unlike many other dry forest species, Q. oleoides is geographically quite restricted within Costa Rica. Moreover, Q. oleoides is ectomycorrhizal in a habitat dominated by vesicular arbuscular mycorrhizal associations, possesses an atypical developmental process with regard to germination and emergence system (a fruit type and associated dispersal syndrome that is extremely rare in the tropics), is wind pollinated in a habitat dominated by insect-pollinated species, and is evergreen in a habitat where most species are deciduous or semideciduous. Finally, a large proportion of trees in the population show a reproductive phenology that seems largely mismatched to the seasonally dry environment of Guanacaste (Cavender-Bares et al. 2010), producing crops of dessication-sensitive acorns at the beginning of a lengthy dry season that is more severe than anywhere else in its range.

Despite this seeming mismatch between traits and environment, Q. oleoides is the most common large tree wherever it occurs and may represent 80 percent or more of individuals present at a site (Boucher 1983). The persistence of leaves in the dry season is probably explained by drought adaptations common to many oaks (e.g., Abrams 1990; Pallardy and Rhoads 1993; Goulden 1996; Cavender-Bares et al. 2004). The evergreen leaves are very resistant to wilting, compared with co-occurring species (Brodribb et al. 2003; Brodribb and Holbrook 2005; Cavender-Bares et al. 2007). Continued gas-exchange rates and high predawn water potentials of mature trees during the dry season suggest that they are deep rooted as adults, although seedlings are considerably more vulnerable (Cavender-Bares, unpublished data). The abundance of Q. oleoides in the region and its unusual combination of functional traits make it an extremely important species in Guanacaste dry forest communities. The seeds are consumed by a wide range of mammalian and avian seed predators, and its evergreen habit affects the abiotic environment experienced by many dry forest organisms.

Scale of the Study

In the ACG, Q. oleoides extends from the dry forest of sectors Santa Rosa and Santa Elena, at approximately 280 meters elevation, to the much wetter forests at approximately 800 meters on the Pacific slope of Volcán Rincón de la Vieja.

Because our study was designed to examine the range of habitat types present within the ACG, six study sites were chosen that exhibited an abrupt ecotone between mono-dominant oak forest and abandoned pasture: three low-elevation (approximately 280 meters) dry forest sites in sector Santa Rosa and three high-elevation (approximately 800 meters) wet forest sites on the Pacifc slope of Volcán Rincón de la Vieja. The particular sites were selected for the presence of a clear forest-pasture ecotone across which to establish transects and to be generally representative of the biophysical environments present at the two elevations (fig. 13-1).

Figure 13-1
figure 1

Abiotic differences in adjacent forest and pasture ecosystems. (A) Daily integrated photosynthetic active radiation (PAR) across the forest-pasture ecotone during the wet and dry seasons averaged for six sites. The x-axis indicates the distance from the edge (0) into the forest or the pasture. (B) Estimated values of volumetric soil water content θ to 30 centimeters depth) averaged for 10 meters and 20 meters into the forest or the pasture for the three upland (800 meters) and three lowland (300 meters) sites. θ measurements were made using time domain reflec-tometry and calibrated with the equation θ = (67.345*x) − 149.74 for highlands and θ = (32.643*x) − 67.678 for lowlands, where x is flight time of the electromagnetic pulse in arbitrary units, following Cavender-Bares and Holbrook (2001). (C) Seasonal high and low daily means for temperature. (D) Relative humidity at 20 meters into the forest or pasture for the same sites. For B–D, significant differences between means in the forest and the pasture, based on two-way ANOVA with habitat and site as main effects, are indicated as follows: ***P < 0.001, **P < 0.01, *P < 0.05.

Full size image

Our study focused on the immediate forest-pasture transition; if there were factors limiting regeneration into the pasture, we expected them to operate near the patch edge. At each site we established three 40-meter transects across the ecotone. All transects at a site were within 100 meters of one another, but no two were closer than 20 meters. Each transect consisted of seven points: one at the patch edge, defined as the farthest extent of oak canopy into the pasture, and 5, 10, and 20 meters from the edge in both directions.

Recolonization Failure

It has long been believed among workers in the ACG that the Q. oleoides forest is not regenerating, compared with other forest types, despite widely dispersed remnant populations that set large seed crops on a regular basis (Boucher 1981). Visually this is apparent as a clear boundary between the oak forest and the adjacent pasture.

We attempted to quantify this by surveying naturally established juvenile oaks along the transects described above. Plants were sampled in a band 2 meters wide along the transects; seedlings and juveniles less than 250 centimeters tall were included in the sample. Across all six sites, 819 seedlings and juveniles occurred on the transects.

At all sites, seedling density dropped to nearly zero at 0–5 meters beyond the forest edge, with only the occasional seedling establishling beyond the oak canopy (fig. 13-2). These results indicate that seedlings are naturally establishing beneath oak canopy but are absent in pastures. Hence, any explanation for limitation must operate at this scale.

Figure 13-2
figure 2

Distribution of naturally occurring seedlings across the forest-pasture transition occurring at six oak forest sites (see inset). Solid lines and closed symbols from three lowland dry forest oak patches at 300 meters elevation, dashed lines with open symbols from three oak patches near the dry-forest–wet-forest transition at 800 meters. Values plotted are means of seedling density on each of three 2-meter-wide transects stretching between the distances indicated. A two-way nested ANOVA comparing the density of naturally occurring seedlings across the forest-pasture transition shows a significant effect of distance from the midpoint of the transition (df = 5, MS = 5.16, F = 32.2, P < 0.0001) and a significant effect of site (nested within forest type: upland or lowland; df = 4, MS = 1.05, F = 11.9, P = 0.035). Interactions of forest type and distance from the edge (df = 5, MS = 1.6, F = 10, P < 0.0001) as well as site by distance (df = 20, MS = 0.525, F = 3.28, P = 0.011) were also significant. Forest type, transect (site), and the interaction of transect by distance from the edge were not significant.

Full size image

Explanations for Limits to Regeneration

Three major explanations for regeneration or recolonization failure in Q. oleoides have been proposed. The first is that for small stands, rates of seed predation are so high that remnant forest patches fail to satiate the local predator community (Boucher 1981). Second, Hallwachs (1994) suggested that Q. oleoides seedlings are unable to tolerate the physical environment of the pastures or are outcompeted there by jaragua. Finally, because of the atypical dependence of Q. oleoides on ectomycorrhizal fungal symbionts relative to the vesicular arbuscular mycorrhizal dependence of most dry forest species, Boucher (1983) suggested that perhaps too little ectomycorrhizal mycelia remained following conversion of oak forest to grassland to inoculate seedlings. Another possibility is that dispersal limitation prevents recolonization, as has been demonstrated for other animal- dispersed species (e.g., Zimmerman et al. 2000; Dalling et al. 2002; Denslow et al. 2006; Dosch et al. 2007).

Seed Predation

The only previous empirical work focused on regeneration of Q. oleoides is the work of Boucher (1981). Boucher concluded that for small stands of oaks, at least, seed production was insufficient to satiate the local predator community and suggested that this explained the failure of Q. oleoides forests to regenerate on the landscape. A large number of vertebrate seed predators do consume acorns in large numbers (table 13-1). There are, however, several reasons to doubt whether seed predation is sufficient as a general explanation for regeneration failure in Guanacaste. First, as shown above, oak seedlings exist at relatively high density within oak patches and on the forest pasture ecotone (fig. 13-2). Second, viable acorns can be collected from oak stands throughout the wet season and can be observed germinating in place. Third, Boucher assumed a long time interval during which acorns were exposed to predators, and this interval was used to project seedling germination and survival rates based on seed predation rates observed over the short term. However, in all cases where we observed mature acorns falling from trees, they germinated and extended the primary radicle within days of landing on the soil surface given even moderate levels of soil moisture. An experiment conducted on seeds from a number of different parental trees indicated that 78 plus or minus 23 percent of seeds collected from the ground would germinate immediately if placed in a damp environment. However, even mild desiccation (i.e., air drying during the humid wet season in the absence of direct sunlight) reduced germination rates to 18 plus or minus 19 percent after 5 days of desiccation. Germination was reduced to 0 percent after 15 or 29 days of desiccation (data not shown).

Table 13-1. Vertebrate seed predators known to consume Quercus oleoides acorns and details of the spatial patterns of the interaction
Full size table

Finally, direct measurements of seed removal rates indicated that seed predators are not removing seeds at a high enough rate to prevent seedling emergence, given the rapidity of germination. Furthermore, at sites where predation rates varied between the forest and the pasture, seed survival was higher in the pasture than the forest (fig. 13-3).

Figure 13-3
figure 3

Pooled seed removal rates in the forest in the pasture habitats for the six study sites. Values are the averages of the per cage removal rates in forest and pasture points on each transect. A randomized two-way nested ANOVA of shortterm removal rate shows a trend in forest type over site (df = 1, MS = 1.777, F = 7.20, P = 0.0586). There was a significant effect of site nested within forest (df = 4, MS = 0.247, F = 3.78, P = 0.0172) and the interaction of habitat by site (df = 4, MS = 0.257, F = 7.86, P = 0.0010). Habitat and the interaction of habitat by site were not significant.

Full size image

These results, taken together with the observations of naturally occurring seedling establishment, indicate that while seed predation undoubtedly plays a role in local regeneration dynamics, it is unlikely to serve as a general explanation for regeneration failure.

Survival and Growth of Seedlings

The pasture is significantly hotter than the forest understory, and during the dry season it is significantly drier, in terms of both atmospheric relative humidity and soil moisture (fig. 13-1). It is not well understood to what extent the seedlings tolerate the dry season, particularly in the harsh and variable conditions of the pasture environment. In general, oak seedlings are less resistant to drought than mature trees because of shallower rooting and less desiccation-tolerant leaves (Cavender-Bares and Bazzaz 2000). Potential competitive interactions with the exotic pasture grass, jaragua, which already has established root systems, provide further reason to consider that seedlings may be particularly vulnerable in the pasture.

To test seedling performance in pastures relative to the forest understory, seeds were planted into the predator exclusion cages on the transects at each site in early July 2005. Seeds were monitored monthly or semimonthly over the next year for emergence and survival. A subset of these seedlings was harvested after 6 months to assess the degree of mycorrhizal infection on the roots. All remaining seedlings were harvested after one year and root systems excavated to assess the rate of mycorrhizal infection.

Emergence

Overall emergence rates varied among sites and among transect positions; however, there was only a moderate drop in emergence rate in the pasture habitat (fig. 13-4A).

Figure 13-4
figure 4figure 4

(A) Effect of transect position on emergence rate per cage. A two-way nested ANOVA comparing emergence across the forest-pasture transition shows significant effects of distance (df = 6, MS = 0.230, F = 8.018, P < 0.0001), site within forest type (df = 2, MS = 3.036, F = 177, P < 0.0001), and the interaction of site by distance (df = 12, MS = 0.181, F = 6.336, P < 0.0001). Forest type, transect nested within site, the forest by distance interaction, and the transect by distance interaction were not significant. (B) Survival to greater than 1 year of emerged seedlings as a function of transect position. A two-way nested ANOVA comparing survival across the forest-pasture transition shows a significant effect of site nested within forest type (df = 2, MS = 0.735, F = 7.470, P = 0.001). Distance, forest type, the interactions of forest by distance, and site by distance were not significant. For all box plots, heavy lines are medians, boxes represent middle quartiles of the distribution, whiskers are set to 1.5 times box length, and open circles are outliers. (C) Dry mass in grams of surviving seedlings as a function of transect position at final harvest in July 2006. A two-way nested ANOVA comparing final seedling size across the forest-pasture transition shows a significant effect of distance (df = 6, MS = 0.675, F = 8.426, P = 0.001). There is no significance of forest type or site nested within forest type. The interactions of forest by distance and site by distance were also not significant.

Full size image

Survival and Growth

Survival of emerged seedlings was uniformly high, with the median survival higher than 80 percent for all distances (fig. 13-4B). There was no significant effect of transect position on survival. For growth, measured as total aboveground dry mass for each harvested plant, there was a significant effect of transect position. Plant growth was highest at 10 meters into the pasture (fig. 13-4C) and was generally higher in the pastures.

While additional experiments are ongoing, these data lead us to conclude that seeds and young seedlings of Q. oleoides are robust in the face of the biotic and abiotic conditions of the pasture environment. Competition with grasses does not appear to inhibit seedling performance once seedlings have emerged and may even facilitate it. Gerhardt (1993) observed that the dry season is the time at which most seedling mortality occurs in SDTF. However, in our study seedling mortality often occurred in the wet season, perhaps due to submersion by flooding during the wet season, fungal infection, or light limitation in forested areas.

Ectomycorrhizal Infection

In December 2005 (5 months postplanting) a small sample of plants was harvested from each position on one transect at one wet forest and two dry forest sites in order to assess mycorrhizal infection. For each plant, a target sample of 300 to 400 root tips was examined under a dissecting microscope. For many plants fewer root tips were recovered, as oak root tips are very delicate and have a high tendency to shear when removed from clay soils, despite soaking of the soil with water and cutting through the soil with a sharp knife before extraction. As a result, complete samples could not always be collected, and the wet forest site examined is missing the data from the edge transect position. Fungal morphotypes were identified, photographed, and counted. From these data, the number of morphotypes and the percentage of root tips that were infected with mycorrhizae were calculated following Kennedy et al. (2003).

Among our samples, infection was uniformly high (fig. 13-5A). ANOVAs of both morphotype diversity and percent colonization indicated that there was a significant or marginally significant effect of site on both measures of fungal colonization but no effect of transect position on either variable. Also notable is that some of the highest ectomycorrhizal diversity observed among all samples occurred in the pasture environment (fig. 13-5B). Species richness of morphotypes was plotted against the number of tips successfully collected from each sample in order to see if differences in richness might simply be a function of the number of tips recovered. There was no relationship between number of tips collected and number of morphotypes encountered (R 2 = 0.04, p = 0.38) and sites tended to cluster in terms of both morphotype diversity and tips collected, meaning that relative comparisons of transect positions within sites are appropriate.

Figure 13-5
figure 5

Mycorrhizal infection of plants grown from seed along three transects, each at a different site, at the early harvest (December 2005). (A) Number of distinct morphotypes present at each transect position. (B) Proportion of root tips colonized for the same samples. Dotted lines with open circles are for site SRF, dashed lines with open triangles for FJ, and solid circles with unbroken line for VJ.

Full size image

These data, taken together with the data presented earlier on robust plant growth in the pastures, suggest that a lack of fungal symbionts is not likely to be important in limiting regeneration in this species. This is surprising, given the paucity of ectomycorrhizal species in this system and the important role that ectomycorrhizal fungi may play in maintaining mono-dominance in tropical forest systems (McGuire 2006). Studies in tropical vesicular arbuscular mycorrhizae systems have shown, however, that pasture soils often contain as much or more innoculum potential as forest soils (Fischer et al. 1994), and a study in SDTF showed that innoculum from early successional stages (2 years postburning) promoted seedling growth more effectively than fungal innoculum drawn from later successional stages (Allen et al. 2003).

It also contrasts somewhat with recent studies of forest-grassland systems in the temperate zone (Dickie and Reich 2005; Dickie et al. 2007). It is possible that at some point beyond our sampling scale (20 meters into pasture) infection would decrease, and at that scale it may limit regeneration.

Dispersal Limitation

Growth and survival of seedlings seem to be robust in the pasture environment over the timescale of our study. Therefore, at least in favorable years, it seems clear that Q. oleoides seedlings will not be prevented from establishing in the pasture environment. Furthermore, there does not appear to be a generalized reproductive failure, based on the abundant recruitment of seedlings within oak patches (fig. 13-2).

The remaining plausible explanation is that acorns do not arrive in the pasture environment in the frst place. One major difference in the oak forests of Guanacaste compared with the oak forests of temperate America is the absence of squirrels that exhibit caching behavior. North American squirrels are behaviorally flexible and highly mobile dispersers that serve as effective dispersers of acorns in the temperate zone (Steele and Koprowski 2001). The variegated squirrel, the only squirrel native to most of the range of Q. oleoides in Guanacaste, is almost completely arboreal and is not known to exhibit caching behavior (Harris 1937; Boucher 1981).

Most vertebrates that interact with Q. oleoides in Guanacaste probably act exclusively as seed predators (Boucher 1983) (table 13-1). The mammal most likely to play a large role in oak seed dispersal is the Central American agouti (Dasyprocta punctata), which has been shown to play a disproportionate role in the dispersal of other large-seeded SDTF species (Hallwachs 1986). Although agoutis will cross open pasture, they do not cache seeds in pastures (Hallwachs 1994) or even in large forest gaps (Hallwachs 1986). It has been shown that in fragmented oak forests in the United States, fragmentation-tolerant squirrel species (Sciurus niger) can make up for the absence of more fragmentation-sensitive species (S. carolinensis) in maintaining substantial dispersal (Moore and Swihart 2007). However, no such secondary disperser occurs in Costa Rican SDTF.

Preliminary data on seed dispersal in this system (Klemens and M.A. Steele, unpublished data) indicate that most seeds are killed within a few meters of the presentation point, and we have so far failed to detect any movement of seeds from the forest environment into the pastures. In fact, when seeds are moved, they seem to be moved deeper into oak patches, away from the pasture. Many unanswered questions remain, however, and this is an important area for future study.

Conclusions and Implications for Restoration

Our data indicate that acorns germinate and survive in abandoned pastures. Once they are established, growth appears to be promoted by the highlight environment of the pastures. However, given the lack of effective dispersal agents, recolonization is likely to occur only very gradually outward from the edge of the forest. The combined results of these experiments help provide guidelines for active restoration of tropical live oak forest ecosystems. Given the plausible role of dispersal limitation and the high performance of established oak seedlings in the pasture, a direct-planting strategy is likely to work well. Seeds or seedlings can be planted into abandoned pastures during the wet season to give their roots time to establish before the dry season. As seedling nurseries to facilitate such efforts can be readily established at low cost and with minimal maintenance, our find-ings thus far are optimistic with respect to the regeneration potential of the live oaks in Guanacaste. Given the complexity of the tropical dry forest agroscape, this study demonstrates the utility of basic research in deciding among alternative restoration strategies. Our understanding of this system remains preliminary, but it reinforces the notion that any understanding of SDTF regeneration will depend on understanding how highly local, and even species-specific, factors, both “natural” and anthropogenically caused, interact to promote or retard regeneration. This may be the major challenge to restoration ecologists working within a particular system. As there is no reason to expect that many general solutions to the problem of SDTF restoration will emerge beyond “stop the disturbance,” for any restoration effort there is probably no substitute for immersion in the natural history of the organisms involved.