Generic Composition, Structure and Diversity 01 Secondary Forests at Amisconde, the PacificSlope 01 the Cordillera de Talamanca, Costa Rica

Most Costa Rican forests have been intensively studied in recent years. One exception is the transition zone from lowland wet forest to the high elevatíon Quercus forest belt al Ihe pacific slopes of the Cordillera de Talamanca. An inventory of secondary forest composition, structure and diversity was done on a specific slope (l1502300-m elevation) in the conservation and development project Amisconde. Thirteen plots of 500 m2 were evenly spread a10ng an elevation gradient. Specimens were collected of a11 woody individuals (> 3 cm DBH), dried, placed in a herbarium of morphospecies and afterwards identified. In total 90 genera within 49 families were found. The vegetation was separated in three forest types using TWINSPAN c1assification. Forest types were elevatíon based. Elevation and forest age showed (overall) no correlatíon with diversity using ANOVA, with the single exceptíon of a positíve correlatíon of the number of genera and elevatíon. This was opposite to the negative correlations mostly found on elevatíon gradients. The main factors for this positíve correlation were the level of recent disturbance and the distance to primary forest, in combination wíth forest age.

Costa Rican forests are intensively studied; complete inventories exist out of most many parts of the country.Forest structure descriptions can be found, among others, of La Selva biological station (Mcdade et al. 1994 and more recently e.g.: Clark & Clark 1996, Guariguata et al. 1997, Laska 1997 with a nice overview of the forest understory structure), the slopes of volcano Barva (Heaney & Proctor 1990) and the Monteverde cloud forest reserve on the higher Tilaran ridge (Lawton &  Dryer 1980, Nadkarni et al. 1995, Matelson et  al. 1995).A study on the lower part of the Tilaran ridge is currently is progress (Hooftman, Bongers & Watson, unpublished  data).Recently the high elevation Quercus forest belt of the Cordillera de Talamanca has been under intensive research (Jiménez et al.  1988, van Velsen et al. 1993, KappeUe 1996).
One of the forests not described recently is the transition zone from lowland wet forest to this Quercus forest belt on the westem Talamanca ridge, a highly interesting zone with influences from both lowland and high elevation forests.
To fill this gap I present in this paper the generic vegetation structure, composition and diversity of the secondary forest in this transition zone (1000-2300 m).Here, human pressure is rather high, resulting in high deforestation rates of primary forest leading to severe soil erosion and land degradation (Sader & Joyce 1988).In protecting and conserving the soil and overall vegetation on the steep irregular slopes on the pacific site of the Cordillera de Talamanca, secondary forest play a very important role (Calvo Alvarado et al. 1993).For this reason, it is necessary to know its floristic composition, structure and diversity.Knowledge of forest structure is a necessary tool.for understanding forest dynamics (Nadkarni et al. 1995) and thus possible protection functions.The Centro Cientifico Tropical in Costa Rica (V.Watson, pers com.) has recently stressed the need for such vital knowledge in .this area.
Next to forest type zonation .along an elevation gradient 1 will correlaie the changing diversity, plot based, with an elevation gradient.Diversity along a elevation gradient is mostly assumed to decrease with elevation, .as stressed by different authors (Cleef et al. 1984, Gentry 1988, Kappelle et al. 1995a) although contrasting results are reported in forest edge situations on high altitudes in Costa Rica (van Velsen et al. 1993). 1 will discuss the found correlation for this speCific secondary forest studied.

Site:
The study area was located on the south�west facing Pacific slopes of the Cordillera de Talamanca in southern Costa Rica, in the bufferzone of the La Amistad Biosphere Reserve and Chirripó National Park.The forest inventory took place in an ± 8 km2 range (515-517 East and 364-368 North, Lambert coordinates Costa Rica), in the conservation and developlll ent project Amisconde (La AMIStad CONservation and DEvelopment), situated around the village San Jerobimo de San Pedro (Peréz Zeledón canton, San José province).The Centro Cientifi co Tropical in San José coordinates the project.Because of the steep and irregular slopes and the large amount of yearly rain (2000-3000 mm), the .area is vulnerable to severe soil erosion (Calvo Alvafado et al. 1993).Forests on this slope weie mainly secondar)' forests aging lO to 40 years (oral informatlon from farmers); elevation of the studied plots was between 1100-2300 meters.Forests regétier�ted mainly at frequently burned pasturegrounds used for cattle grazing with only few remnartts oí prior vegetation (inainly Q�ercus, Ocot?a and Ficus) , " ;: .'Floristic sampling: 1 distributed plots to the best possible extent evenly along an elevation gradient, using aerial photograph interpretation (photographs taken in 1992) and conversations with local farmers and Amisconde staff.Plots contained a buffer zone of at least �O Ol.eters to all signs of recent human forest disturbance.
Plots size w �s500 m 2 (50*10 m), perpendicular to the exposition of the slope, with the shorter site following the con tour lineo Within the plot all woody individuals > 3 cm DBH (Diameter Breast Height) were measured in two diameter classes: 3 cm í DBH í 10 cm and ;;:: 10 cm DBH.The larger class was measured at the whole plot, the smaller at half the plot (stem-density and basal area doubled afterwards).The diameter distinction is modifled from Clark (1994, see also Hooftman 1995).
Of each woody individual a specimen was collected.Those were dried and gathered in a herbarium of morphospeCies.Main criteria for .classification were the general position of the leaves on the branch, raw leaf morphology and other specific leaf features.Morphospecies were identified with the help of taxonomists at .INBio in Heredia (Q.Jiménez and J.P. Morales); in addition the Amisconde staff provided help as well as local people with knowledge of vernacular species names.The herbarium was afterwards donated to a local school for education purposes.Several morphospecies could not be identified to generic level; in the TWINSPAN classiflCation those morphospecies were not omitted.Taxonomic nomenclature.follows mainly Kappelle (1996) with additions from manual de la flora Costa Rica (lNBio, Costa Rica).
Vegetation classification: A rough vegetation table was made using basal area and stem density per genus.Subsequently by using box-plots, assumed remnant individuals from previous vegetation were omitted from the data set.Excluded were several large Ficus ati d Quercus trees (minimal cut-off: DBH > 70 cm).Subsequently .I discovered that other researchers had used the same cutoff (Lugo & Brown 1992, Clark & Clark 1996), finding it a useful diameter criterion (Clark & Clark 1996).The raw vegetation table was treated using the multivariate cluster analysis program TWINSPAN (Jongman et al. 1987) in order to identify possible secondary forest communities (called types in this study).In addition 1 used the cluster analysis program DENDRO, with a Sorensen dissimilarity indexo Forerunning TWINSPAN the vegetation table was converted into eight TWINSPAN cut level classes.This conversion followed a logarithmic octave scale proposed by Gauch (1982) and given in Kappelle et al. (1994).A TWINSPAN analysis was conducted for both basal area and stem density data per genus, using socío ecological groups of genera.Nomenclature of vegetation types is provisional and analogue to Kappelle et al. (1994).
The two different TWINSPAN classifications showed no major" differences.Several minor changes occurred in socio ecological group structure.In this paper 1 will use further the classification based on basal area per genus, shown in Appendix l.Differences between the forest types were tested with the use of a student-t test.Trends (p <0.1) were seen as differentiating between forest types.

Gradiellt anaiysis:
The successional stage of the forest and elevatíon were correlated with diversity usíng ANOVA and regression techniques.Total basal area per pIot was used as a parameter for estimating the secondary forest stage.A positive relation between forest age and the inerease of basal area in young secondary forest was assumed.This relation has been observed among others in the Talamanca area by Kappelle et al. (1995b).
Generic diversity of secondary forest plots was ealculated in three ways.Firstly, this was done by using the number of individual genera per plot and subsequently by calculating the Shannon-Wiener index of species diversity per plot on basis of both basal area and the number of individual stems per genus.The ShannDn-Wiener index of species diversity (Hs) algorithm used follows Magnurran (1988) and will be further referred to as Shannon-index.

RESULTS
General species composition: In the study area 13 secondary forest plots were made (pIot names in order of measurement: A M). Forests were between 8 and 15 m tall, standing on slopes ranging from 38% to 95%. 90 genera within 49 families were identified.Most common genera on the slope of the ridge were Cornus and Mollinedia.Other important genera in the area are Quercus (in higher areas), Ocotea, Sorocea (in lower areas), Protium, Toxicodendron and the shrub Palicourea.However, dominance of any genus rarely exceeded 20% of local basal area within plots.A complete overview of an genera with their abundance can be.found in Appendix 1. Diameter distribution showed a clear J-shape pattern for all forests (data not shown) suggesting a good regeneration potential.
TWINSPAN classification revealed three forest types withinfue secondary forest (Table 1).The second clustering technique, using a Sorensen dissimilarity analysis, obtained fue same resulto wiili only minor differences in similarity 1evels due to difference in algorithm (Fig. 1).Elevation was most likely to be the most important determining factor in combination with forest age and local disturbance.An outlier was pIot M that, according lo the dendrogram, had a very low similarity with other plots.For elevation rea'iOns, this plot was pul in the lowest forest type.

Secondary forest types description:
1: Quercus-Cleyera forest The elevation range of this secondary forest type Ís ca 2000-2300 m.It was characterized by a cIea!' presence of Quercus and Cleyera, covering 10-15% of the total basal area in this zone.Other distinctive genera in this forest type were Rauwolfia, Styrax, Picramnia and Citharexylum.The Quercus-Cleyera forest was very dense with a compact undergrowth layer of several bamboo and shrub species.The forest floor contained a thick layer of organic material, ínhabiting large amounts of herbs.This suggests compiled with a higher basal area (ns.) the presence of a rather older secondary forest compared to the Cecropia Inga forest at lower elevation.Stem density (DBH > 3 cm) was higher than in the Cecropia-Inga forest type (p<0.029) and the Clusia-Cornus forest type (p< 0.085).Slope and exposition of all forest types showed no significant differences.

2: Clusia-Cornus ¡orest
The Clusia-Cornus forest contains elements of both the Quercus-Cleyera type and the Cecropia-Inga type but did not inelude the most distinctive species of these groups.It can be seen as a transition zone.between the other two forest types with its own characteristics.Most genera reported within this belt were seen in low but more even distributions (4.6 stems per genus/500 m 2 ; Quercus-Cleyera type: 5.6 stems per genus/500 m 2 ) The Cecropia-Inga forest plots contained fewer genera (p<O.O ll), individuals (p<0.029) and basal area (ns.) than the Quercus-Cleyera forest type.In additíon, no abundant undergrowth was found, the litter layer was thin and contained mostly fresh leafs.This, and the presence of the distinctive pioneer Cecropia indicates a younger, recently more disturbed secondary foresto Diversity: 1 found in the Quercus-Cleyera forest type 67 genera of woody species within 45 families, in the Clusia-Cornus forest 45 (31 families) and in the Cecropia-Inga forest 53 (35 families).However, this was measured with an uneven distribution of plots for the different forest types (Table 1), therefore comparing the forest types was done by using averages per plot.Fig. 2 gives an overview of the number of genera and the average diversity of the three forest types.As already stated the forest types differed most in quantity of genera and less in calibrated diversity (Shannon index).The Quercus-Cleyera forest contained more genera (p<O.085) and a higher diversity (p<0.012,basal area based) than the transition Clusia-Cornus forest and the Cecropia-Inga forest (respectively p<0.011 and p<0.066).No differences between diversity in the Quercus Cleyera forest and the Cecropia-Inga• forest type were found.This was mainly caused by the already observed more even distribution of genera over the stem density in the Cecropia Inga forest compared to in the Quercus Cleyera foresto The Quercus-Cleyera forest contained a more c1umped distribution of genera over the area.---------- Diversity along a successional and elevation gradient: In this paper the successional stage of the forest plots was estimated usingthe basal area of the separate plots.No significant correlation between basal area and number of genera and diversity was found.The same accounts for Shannon-index yalues along the elevation gradient.The amount of genera appeared to increase significant1y with altitude (p<0.01,Fig. 3).The Quercus-Cleyera forest type contained a significantly higher number of genera compared to the Clusia-Comus and tbe Cecropia-Inga foresto

I;>ISCUSSION
1 Forestzona�on: Recent studies on the cordillera de Talamanca (Jiménez et aL 1988, Orozco 199 1, Kappelle et al. 1994, Kappelle et al. 1995b) overviewed mainly vegetation composition aboye 2000 -m elevation, referred to in general as Ocotea-Quercus forest (Kappelle et al. 1994, Kappelle et al. 1995b).Comparing especially .theQuercus-Cleyera forest type described in tbis paper to Kappelle's studies, including Van Velsen et al. ,(1993), many important genera appear to be in common in a similar abundance: Quercus, Cleyera; Vibumum, Clusia, Comus and several genera in tbe Lauraceae (Ocot ea, Persea, Nectandra).The clear similarities betweeilthe upper Quercus-Cleyera forest in this study and tbe lower Freziera-Quercus (KappeUeetal.1994) forest zone suggest that the altítudínal zonation found in the Amisconde area corresponds to the lower part of the spectrum of secondary fores!zonation of the Ocotea-Quercus forest.Gathering more data, preferably on species level, is needed to pro vide us a better.insight on forest belt zonation within this area.This for example by means of using the, same cluster analysis for the whole elevation gradient on the Talamanca ridge on species levels (1100-3000 m elevation). 1 checked only woody índividuals in this study, consequently.the quantity of genera ís difficuIt to compare with the earlier mentioned studies on Talamanca, those include mostly all vascular species (see van Velsen et al. 1993, Kappelle 1996).Similar studíes on protected montane forest in Northern Costa Rica, revealed a similar number of genera in the Monteverde cloud forest reserve (Nadkarni . et al. 1995) and clearly les s genera on volcano Barva (Heaney & Proctor 1990).This states the value of this, mostly not protected, transition zone from lowland forest to true montane forest of high eleyation on the Talamanca ridge, although the mentioned studied used less elevation differences but a larger sampling areas.

Fadors influencmg generic diversity:
Recent studies in the Cordillera de Talamanca showed a decrease in species richness with increasing altitude (Kappelle el al. 1995a).A relation also showed in the Colombian Andes (Cleef et al. 1984).In this study one notes the opposite, a positive correlatíon between number of genera and elevation.This opposite relation was also found by Van Velsen et al.
(1993) but they included all vascular species and credited this to migrating paramo.herbs and ferns species• into herb-and forblands at approx.2800 meters, this situation is not similar to this study.Measuring only woody species as done here would also exc1ude the herbs immigrating from the pasture grounds into the forest edges, a process clearly visible in these Talamancan forests (M.ten Hoopen & M. Kappelle, pers com.).To expiain the positive correJatíon several(c ooperating) explanations are possible, including forest age, the level of disturbance in recent history and most likely the distance to primary forest.Those exp1anatíons are discussed below in this.
Unfortunately local farmers did not provide reliable information about land use history.No written recordings were made and several families were only for a short period of time in the area.No correlation was found between basal area and floristic composition, in which basal area is supposed to be a good indication for age in younger successional forest.However, by using other indications 1 think that the Cecropia-Inga forest type is possibly younger than the Quercus-Cleyera forest.With younger 1 mean a smaller tíme span of secondary growth after abandonment of a specific area after utilization (assumed to be cattle grazing).Younger secondary forests mainly inelude fewer species than oIder forest stands.A importantindicator is the presence of the genus Cecropia which is a very light demanding short living pioneers, able to grow only in full Iight conditions and very rarely seen in clase forests (Lamprecht 1989).The genus was only found within lower elevatíon plots but it was reported up to approximately 2000 m in the Amisconde area on other slopes (internal Amisconde report, S.T;).Furthermore, Kappelle (1996) díd report Cecropia in young secondary forest up to 2300 meters elevation on the Cordillera de Talamanca.The absence of Cecropia in higher elevation plots (1500-2300 m) may be a clear indícation of an oIder successional age of those forests.
A further factor is the type and leve} of disturbance in the secondary forests.The research area is a bufferzone in which tree cutting is not allowed.However until 1993, the start of Ami sconde project (meaning a more intensive controllíng in the forest) illegal tree fellings were coromon practice inside those forest (pers obs.and pers como with Amisconde staff).In addition small-scale.fires, s e t for agricu1tur�1 .purposes, repeatedly affected the forest.The remnants of those disturbances are likely to be higher in the lower elevation forest patches, near the villages, and not in the inaccessible higher situated patches near the edge of the Biosphere reserve Amsistad (walking distance 2-3 hours of the village with very unclear trails).A possible mechanism is that in low altitude patches, shade tolerating late successional tree species were constantly outcompeted, if not cut, by light demanding pioneer species due to unusual large gaps, resulting in a possible retrogressive and delayed succession, a process clearly shown on the lower slopes of Tilaran (Hooftman, Bongers & Watson, unpublished data).
A more ecological factor influencing higher levels of diversity in high elevation plots in this area is the geographicaI distance to the primary forest and thus possible isolation from seed flux.Primary forests in the area occur at higher altitudes (ca.2500 m, pers obs.) and are montane Quercus forest.As stated by Purata (1986)  Furthermore those primary Quercus forest contain presumably other species than would establish in those Iower forest, which would mean that seed and flux from other genera than present in the forest is very scarce.The further distance from seed sources compiled with the earlier mentioned factors would be a clear explanation of the positive elevation-diversity correlation found within this study.Plot: Altitude: Genera (#):

Fig. l .
Fig.l.Dendrogram of dissimilarity between plots, using a Sbrensen (presentlnon-present) index based on.basal area data.

Fig. 2 .
Fig. 2. Av erage Corest type descriptives (per 500 m'): a: nu m ber oC genera; b: Shann on index based on b asis stem density per genus; e: Shann on index based on basal area per genus.
and KappeIle et al. (1996) succession is speeded up in conditions close to a primary forest.Further away from the mature stands succession may develop much slower due to reduced seed dispersal by birds (among others: Guevara et al. 1986, Robinson & Handel 1993, Heitz Siefert et al. 1996, Wilms & KappelIe, unpublished data).
secondary forest types in the Amisconde area.Costa Rica.with genera in socio-

TABLE 1
Forest types revealed by TWINSPAN classification and plots (exact location of plpts on the slope is not given in this paper but can be given upon request by the aUlhor) Elevation was 1500-1750 m.The genus Clusia was abundantly present.Cornus reached its optimum distribution in this forest, as does Sorocea.Other important present genera were Ocotea, Toxicodendron, Ilex, Mollinedia and the edible dwarfpalm Geonoma.Stands were more open with less individual stems (p<0.085), a less thick undergrowth and herb layer, compared to the Quercus-Cleyera forest type.3: Cecropia-Inga ¡orest Cecropia-Inga forest (1150-1500 m elevation) ís dístinguished by the present of the pantropícal pioneer genus Cecropia in association with the distínctive genus Inga.Other prominent genera in this belt were Sorocea, Cornus, Mollinedia and Protium.