Tropical forests harbor at least two-thirds of the floristic diversity of the planet (Kier et al., 2005). This is a vital natural resource as it provides ecosystem services such as food, fresh water, timber, fuel, climate and flood regulation, cultural values, spiritual, educational and recreational resources (Butler et al., 2005).

Much of old-growth tropical forests have been declared as forest concessions; these are areas of permanent timber production. The management in forest concessions is not clear-felling, but is a selective extraction that preserves much of the forest cover after cutting. This fact creates some opportunities for biodiversity conservation because many forest concessions, which were formerly designed for timber production, are now the main restricted-use areas to give shelter to plant and animal species whose habitat, outside concessions, has been destroyed (Roche, 1997; Arets, 2005). The selectively logged forest is a kind of “intermediate situation” between total protection and deforestation, which deserves more attention to conservation management because it preserves much of the original diversity and its associated ecosystem services (Putz et al., 2012; Edwards et al., 2014).

Selective harvesting of timber generates impacts that may lead to an alteration or reduction of floristic composition and diversity (Hamilton, 1988). Johnson & Cabarle (1995) indicate that high harvest rates have generated threat levels in species such as Dalbergia nigra, Pericopsis elata, Swietenia mahogani and Guaiacum officinale. Similarly, ter Steege et al. (2002) point out that (after 75 years) logging of Chlorocardium rodiei in Guyana decreased 63 percent in its abundance and it reduces the financial feasibility of its current yield. However, Gourlet-Fleury et al. (2013) suggest that many of the reports, which describe a negative effect of harvesting on diversity, come from a high variety of situations (frequency and intensity of the felling, the initial state of the forest, interaction with other types of disturbances) that could hide a possible increase in species richness.

The process of plant succession that takes place after a disturbance is closely related to the theory of gaps (Hartshorn, 1980). In the early stages of succession (before 30 years) conditions are propitious for pioneer species that develop throughout their life cycle in a full light incidence (Felton et al., 2006). That circumstance could decrease diversity initially, but in a few years, the nomadic species will arrive and finally the tolerant species, typical of mature forest, will do the same (Kariuki et al., 2006).

The theory of “climax” indicates that the process of plant succession means a restoration that appropriately recovers the original or pristine condition (Odum, 1972). The results of some studies in tropical forests seem to support this assumption (Faber-Lagendoen, 1992; Magnusson et al., 1999; Guariguata & Ostertag, 2001). “Climax” is one of those terms that has generated more controversy in ecology, but is still used by eminent authors like Dajoz (2002), Strasburger et al. (2004), Molles (2006), Odum & Warret (2006), Smith & Smith (2007). The dispute comes from the idea of the climax like a stable state. However, along the succession, there is a recovery of structure, floristic composition and ecosystem services that arrive at a situation very similar to the original one. Along this work we use the “consensus” term old-growth forest instead climax forest.

According to Connell (1978), the “Intermediate Disturbance Hypothesis” (IDH) implies an increase of ecosystem diversity after small and frequent disturbances that generate greater habitat heterogeneity. This hypothesis has also been the subject of intense debate, has acquired support from the results of Burslem & Whitmore (1999), Sagar et al. (2003) and the compilation made by Kershaw & Mallik (2013). Obviously, there is no consensus and the controversy continues. For example, Fox (2013) pointed out that the IDH must be abandoned. Sheil & Burslem (2013) replied that, although there could be misunderstandings, the IDH still was valid and useful.

Anyway, some studies show a satisfactory recovery of richness, of floristic diversity, or even an increase in these parameters after selective logging. Situations like these have been reported in Uganda (Plumptre, 1996), Guyana (Van Andel, 2001), Borneo (Verburg & van Eijk-Bos, 2003; Berry et al., 2008), Central Africa (Gourlet-Fleury et al., 2013; Decocq et al., 2014) and Tanzania (Rutten et al., 2015).

Considering the above arguments, we hypothesize that after several years of a selective logging perturbation, some tropical forest plots recovers most of its diversity, and the secondary forest is quite similar to the original. Our aims are to evaluate the changes that occur in the tree and palm diversity throughout the secondary plant succession taking place in a forest, after a selective timber harvest with different levels of intensity. Our results could be also useful in order to improve the conservation strategies that different countries must implement in tropical forest concessions.

The high dominance of the palm A. butyracea has already been reported in the western plains of Venezuela by Kammesheidt (1998) and Guevara (2001). Similarly, the pioneer temperament of genus Cecropia, Guazuma, Inga, Triplaris and Ochroma has been mentioned in reports of Kammesheidt (2000) and Castro-Luna et al. (2011).

Changes in species composition are supported not only by the increased importance of pioneer species but also by the decrease of nomadic species (such as P. quinata, P. reticulata, according to the definition of Vásquez-Yanes & Guevara, 1985) especially in the treatments of medium and high impact. These results are consistent with the abundant literature on the dynamics of tropical forests (Hartshorn, 1980; Moret et al., 2008; Ruschel, 2008; Lohbeck, 2014). Nevertheless, the decline of C. peltata from year 15 is not coincident with some reports that indicate that pioneers such as this species have a life cycle up to 50 years and that at year 20 still maintains a high proportion of the basal area community (de Oliveira, 2005).

In forest ecosystems there is no an absolute stability, but a kind of dynamic balance caused by growth, recruitment and mortality processes that leads to changes in the presence, density and species dominance, mainly those with low abundance (de Oliveira et al., 2014). That is why when making long-term assessments in forests with no anthropogenic interventions, small changes in floristic composition are reflected in changes in the IVI of species (Table 2).

The IVI decline of nomadic species with high commercial value is worrisome. In conventional logging these species have the highest pressure; if extraction continues in subsequent cutting cycles, to the long term they could be threatened by a significant decrease in population or its genetic diversity (Zimmermann, 1992; Plonczak, 1993; Kammesheidt, 1994). Therefore, harvesting cycles should be extended to at least 60 years in order to achieve sustainable yields (Kammesheidt et al., 2001b).

Sapium stylare shows an interesting behavior. It is a high value nomadic species, but is not harmed by the logging intensity and appears with elevated IVI values after 25 years in plots with medium and high impact. A similar performance was reported by Kammesheidt (1998) who reported that this is due to some ecological characteristics of this species: clumped distribution, well adapted to poorly drained soil, use of light availability derived from logging, the presence of sufficient remaining mature trees that allow regeneration and increase of individuals in the lower diameter classes.

The plant species diversity values found in this study generally approach the lower limits of the scope reported for South American tropical forests. The values of species richness are less than 80 spp/ha, which correspond to low diversity places that are mainly on flooded sites (Knab-Vispo et al., 1999). The values of the Shannon-Wiener index are intermediates in the range (2.1-3.6) reported by Campbell et al. (1992) and Damasceno-Junior et al. (2005). Furthermore, areas with prolonged drought periods are reported to show low diversity indexes (ter Steege et al., 2003; Saatchi et al., 2008). In Caparo a combination of the limiting factors mentioned above are present: alternating of drought and flooding. Consequently, small decreases in diversity which were detected in this study as the result of logging have a low significant ecological importance. In other words, none of the parameters assessed changed from a condition of high to medium or low diversity.

Diversity indexes show very slight oscillations in the control plots. In logged forests this fluctuation is more accentuated and it seems to increase with the impact level (Fig. 1). However, the changes are not, in general, statistically significant (Table 3), except in one case. This could mean that small losses of diversity that might occur immediately are recovered in few years. Only the species richness (N0) in the high impact level is lower in year 2, but recovers at year 5. Similar results were found by ter Steege et al. (2002) in Guyana; but, these authors also suggest that the diversity indices commonly used (Fisher´s Alpha and H ‘) are not sensitive to detect changes in an area with many individuals and species. Meanwhile, Bongers et al. (2009) found that the disturbance is less important to explain the patterns of diversity in humid ecosystems; in dry forests disturbance explain more variation of diversity.

Under strong disturbances, large gaps produce immediate microclimatic changes that promote regeneration, development and dominance of pioneer species, which could decrease the ecosystem diversity (Denslow, 1980; Cannon et al., 1998). This argument could explain the decrease observed at the high impact level in species richness. However, Guariguata & Ostertag (2001) indicate that the process of plant succession retrieves microclimatic conditions which, together with nearby seed source, generates a restoration of diversity in a few decades. Our results show that restitution occurs after 5 years (Table 3) and seems to be favored by our experimental design since the treated plots (1 ha) are surrounded by strips of an undisturbed forest of at least 100 m wide.

Space heterogeneity is a conspicuous property of tropical forest. In Caparo this is mainly caused by a geomorphological dynamics that produce a textural variability in the soils and its consequent diversity of floristic composition (Lozada et al., 2010). These changes are noticeable in the IVI values for year 0 (Table 2) and Chao-Sørensen pre-treatment indexes (Table 4). But it is remarkable to note that post-treatment similarities (year 20) are higher than pre-treatment ones. These values prove that, for the treatments and period assessed, spatial changes are more important than logging induced changes, as has been previously reported by Bongers et al (2009).

With respect to the recovery of floristic composition, Ferreira & Prance (1999) determined that a secondary 40-year-old forest in the Amazon recovered 95 percent of the original biomass, but the floristic similarity only reached 12 percent (Jaccard Index). Guariguata & Ostertag (2001) note that in some cases, even after 80-100 years, floristic composition is not similar to mature forest. In Panama, DeWalt et al. (2003) found that at the age of 70 years secondary forests reach structural features similar to those of primary forests, but with only a 26 percent of similarity (Sørensen Index). Therefore, the minimum level of similarity that we found in forests, after 25 years of selective logging, is quite adequate (91%, Table 5). The divergence on the figures of similarity, between other studies and ours, are probably related with the object of measurements (trees, palms, vascular plants…), but also with the type of similarity index used. Jaccard and Sørensen indexes are the most commonly used, but they are likely to underestimate the similarity between plots because they use presence/absence data and are very sensitive to rare species, which include the so-called “singleton” and “doubleton”, i.e., those that have an abundance of 1 and 2 individuals respectively. The Chao-Sørensen index (Chao et al., 2005) calculated in our work takes into account not only presence but also abundance of species and in this sense reflects the IVI values better.

Long-term research on the recovery of the Neotropical forests after logging is scarce (Verburg & Van Eijk-Bos, 2003). However, it is broadly stated that forests, even without human intervention, perform slight changes in floristic composition and diversity. Thus, we can conclude that selective harvesting of timber, with low and medium intensity, is a disturbance that works in a similar way to natural disturbances of similar intensity. For all the diversity indexes assessed a recovery to the pre-harvest level was found, but not an increase in such parameters.