Value of Small Patches in the Conservation of Plant-Species Diversity in Highly Fragmented Rainforest

Contributed Paper

Value of Small Patches in the Conservation ofPlant-Species Diversity in Highly FragmentedRainforestVICTOR ARROYO-RODRIGUEZ,∗†‡ EDUARDO PINEDA,∗ FEDERICO ESCOBAR,∗

AND JULIETA BENITEZ-MALVIDO†∗Departamento de Biodiversidad y Ecologıa Animal, Instituto de Ecologıa A.C., Xalapa, Veracruz, Mexico†Centro de Investigaciones en Ecosistemas, Universidad Nacional Autonoma de Mexico, Morelia, Michoacan, Mexico

Abstract: We evaluated the importance of small (<5 ha) forest patches for the conservation of regional plant

diversity in the tropical rainforest of Los Tuxtlas, Mexico. We analyzed the density of plant species (number of

species per 0.1 ha) in 45 forest patches of different sizes (1–700 ha) in 3 landscapes with different deforestation

levels (4, 11, and 24% forest cover). Most of the 364 species sampled (360 species, 99%) were native to the

region, and only 4 (1%) were human-introduced species. Species density in the smallest patches was high

and variable; the highest (84 species) and lowest (23 species) number of species were recorded in patches

of up to 1.8 ha. Despite the small size of these patches, they contained diverse communities of native plants,

including endangered and economically important species. The relationship between species density and area

was significantly different among the landscapes, with a significant positive slope only in the landscape with

the highest deforestation level. This indicates that species density in a patch of a given size may vary among

landscapes that have different deforestation levels. Therefore, the conservation value of a patch depends on

the total forest cover remaining in the landscape. Our findings revealed, however, that a great portion of

regional plant diversity was located in very small forest patches (<5 ha), most of the species were restricted

to only a few patches (41% of the species sampled were distributed in only 1–2 patches, and almost 70% were

distributed in 5 patches) and each landscape conserved a unique plant assemblage. The conservation and

restoration of small patches is therefore necessary to effectively preserve the plant diversity of this strongly

deforested and unique Neotropical region.

Keywords: biodiversity, deforestation, fragmentation, habitat loss, Los Tuxtlas, Mexico, plant diversity, tropicalrainforest

El Valor de Parches Pequenos para la Conservacion de la Diversidad de Especies de Plantas en un Bosque LluviosoSumamente Fragmentado

Resumen: Evaluamos la importancia de parches de bosque pequenos (<5 ha) para la conservacion de la

diversidad regional de plantas en el bosque lluvioso tropical de Los Tuxtlas, Mexico. Analizamos la densidad

de especies de plantas (numero de especies por 0.1 ha) en 45 parches de bosque de diferentes tamanos (1–700

ha) en tres paisajes con diferentes niveles de deforestacion (4, 11 y 24% de cobertura forestal). La mayorıa de

las 364 especies muestreadas (360 especies, 99%) fueron nativas de la region, y solo 4 (1%) fueron especies

introducidas por el hombre. La densidad de especies en los parches mas pequenos fue alta y muy variable; el

mayor (84 especies) y el menor (23) numero de especies se registro en parches de hasta 1.8 ha. No obstante el

tamano pequeno de estos parches, contenıan comunidades diversas de plantas nativas, incluyendo especies

en peligro de extincion y economicamente importantes. La relacion entre la densidad de especies y el area

‡Current address: Centro de Investigaciones en Ecosistemas, Universidad Nacional Autonoma de Mexico (UNAM), Antigua Carretera a Patzcuaro

No. 8701, Ex Hacienda de San Jose de la Huerta, 58190 Morelia, Michoacan, Mexico, email [email protected] submitted March 30, 2008; revised manuscript accepted August 6, 2008.

729Conservation Biology, Volume 23, No. 3, 729–739C©2008 Society for Conservation BiologyDOI: 10.1111/j.1523-1739.2008.01120.x

730 Conservation Value of Small Patches

fue significativamente diferente entre los paisajes, con una pendiente significativamente positiva solo en el

paisaje con el mayor nivel de deforestacion. Esto indica que la densidad de especies en un parche de tamano

determinado puede variar entre paisajes que tienen diferentes niveles de deforestacion. Por lo tanto, el valor

de conservacion de un parche depende de la proporcion total de cobertura forestal en el paisaje. Sin embargo,

nuestros resultados revelaron que una gran porcion de la diversidad regional de plantas se localizo en parches

de bosque muy pequenos (<5 ha), la mayorıa de las especies se restringieron a unos cuantos parches (41% de

las especies muestreadas se distribuyeron en solo 1–2 parches, y casi 70% se distribuyeron en cinco parches)

y cada paisaje conservo un ensamble de plantas unico. La conservacion y restauracion de parches pequenos

es por lo tanto necesaria para la preservacion efectiva de la diversidad de esta region Neotropical unica y

sumamente deforestada.

Palabras Clave: biodiversidad, bosque lluvioso tropical, deforestacion, diversidad de plantas, fragmentacion,Los Tuxtlas, Mexico, perdida de habitat

Introduction

Biodiversity conservation in highly fragmented land-scapes requires appropriate management of undevel-oped land and the surrounding modified land matrix.Two well-known ecological theories—island biogeogra-phy theory and metapopulation theory—highlight theimportance of large habitat patches to maintain biodi-versity because species richness tends to increase withpatch area and connectivity (MacArthur & Wilson 1967;Hanski 1999). Current empirical evidence shows thatconservation of many animal and plant species is onlypossible within very large reserves (Laurance 2005) be-cause larger patches can support larger populations (e.g.,Arroyo-Rodrıguez et al. 2007, 2008) and are less affectedby edge effects (Laurance et al. 1998; Santos et al. 2008),hunting pressure, and logging (Peres 2001; Tabarelli et al.2004).

In recent years, however, the results of a growing num-ber of studies have shown that the conservation value ofsmall patches should not be ignored. Small patches can in-crease landscape connectivity (Dunning et al. 1992) andcan provide important ecological services (e.g., crop pol-lination and seed dispersal: Bodin et al. 2006) and habitatfor numerous animal species (e.g., birds, Fischer & Lin-denmayer 2002; dung beetles, Estrada & Coates-Estrada2002; butterflies, Benedick et al. 2006). Nevertheless,very few researchers have analyzed the importance ofsmall patches to plant diversity (but see dos Santos et al.2007), and almost all the studies have been conductedat the patch scale. Landscape-scale fragmentation studiesare necessary because the response of species to fragmen-tation may vary depending on the level of deforestation(e.g., Andren 1994; Hill & Curran 2005).

Globally, many forest habitats have been logged or per-manently converted to other land uses; therefore, in manyregions, establishing large forest reserves is not possible(see Schwartz 1999). For example, in Los Tuxtlas Bio-sphere Reserve, Mexico, approximately 90% of the origi-nal forest cover has been lost, and the largest patches aremainly found above 700 m asl. Many small patches, how-

ever, are dispersed across the entire Biosphere Reserve(Dirzo & Garcıa 1992; Guevara et al. 2004). Furthermore,as elevation increases (>700 m asl) the rainforest givesway to less disturbed cloud and scrub forests (Castillo-Campos & Laborde 2004), and the largest areas of thereserve do not protect lowland rainforests. Under thissituation, it is critical to assess the conservation valueof remaining small and intermediate lowland rainforestpatches to regional species diversity because this infor-mation could be used to establish efficient managementstrategies for biodiversity preservation at a regional scale.

Although there are studies of the floristic and struc-tural characteristics of the vegetation in the region (e.g.,Bongers et al. 1988; Ibarra-Manrıquez et al. 1997), mosthave been carried out at sites where the vegetation is wellpreserved (i.e., Los Tuxtlas Field Station). The few studieson changes that arise from anthropogenic disturbance ofthe rainforest (e.g., Arroyo-Rodrıguez & Mandujano 2006)are focused on a single landscape unit. Here we adopteda regional approach, in which landscapes with differentlevels of deforestation were analyzed comparatively. Todetermine the changes in species richness produced bythe loss and fragmentation of the Los Tuxtlas rainforestand to evaluate the importance of small forest patches forthe conservation of regional plant diversity, we analyzedplant species density within patches located in 3 land-scapes with different levels of deforestation. Specifically,we examined the contributions of small patches to re-gional diversity; relationship of species density to patcharea; differences in the relationship of species densityto patch area among landscapes; and pattern of speciescomposition within and among landscapes.

Methods

Study Area

The Los Tuxtlas region is in southeastern VeracruzState, Mexico (18◦8′–18◦45′N; 94◦37′–95◦22′W; Fig. 1).The climate is warm and humid, with a mean annual

Conservation Biology

Volume 23, No. 3, 2009

Arroyo-Rodrıguez et al. 731

Figure 1. Location of the 3

landscapes studied in the Los

Tuxtlas Biosphere Reserve,

southeastern Veracruz, Mexico.

Black polygons represent studied

patches (LDL, lowest level of

deforestation, 24% of remaining

forest cover; IDL, intermediate

deforestation level, 11%; HDL,

highest deforestation level, 4%).

temperature of 25 ◦C and annual rainfall between 3000and 4600 mm (Soto 2004). The elevation ranges from 0to 1780 m asl. Los Tuxtlas was decreed a Biosphere Re-serve in 1998 because of its high and unique biodiversity(CONABIO 2000). The reserve is 155,122 ha and repre-sents the northern limit of tropical rainforest distributionin the Neotropics (Dirzo & Garcıa 1992). The originaldominant vegetation type (below 700 m asl) was tropicalrainforest, but the area has been heavily deforested andfragmented since the 1970s, and the remaining rainfor-est is surrounded by a matrix of pastures and croplands(Castillo-Campos & Laborde 2004).

The 3 fragmented landscapes we examined repre-sented a gradient of deforestation, were situated between0 and 400 m asl, were of a similar overall land area (about5000 ha), and were delimited by the coast and large rivers(Fig. 1). We digitized the landscapes in ArcView 3.2 onthe basis of aerial photographs (1:20,000), orthophotos,and field information. Although the 3 landscapes havebeen highly deforested, there were notable differencesin the degree of deforestation, which we used to nameeach landscape. The landscape with the lowest deforesta-tion level (LDL) was 5356 ha, 24% of which was rainforestdistributed among 75 patches ranging from 0.5 to 700 ha(median size 3.2 ha); the landscape with intermediate de-forestation level (IDL) was 4965 ha, 11% of which wasrainforest distributed among 88 patches (median size 2.6ha, range 0.5–76 ha); and the landscape with the highestdeforestation level (HDL) covered 5046 ha, 4% of whichwas rainforest distributed among 46 patches (median size

2.0 ha, range 0.5–68 ha). Mean patch size and shape com-plexity (mean shape index [SD]: LDL = 2.2 [1.0]; IDL =1.9 [0.7]; HDL = 1.8 [0.6]) was significantly greater in LDLthan in the other 2 landscapes; whereas mean distance tothe nearest patch (LDL = 102.7 m [172.0]; IDL = 112.7 m[99.9]; HDL = 288.7 m [299.8]) was significantly greaterin the HDL (for further details, see Arroyo-Rodrıguezet al. 2007, 2008) than in the other 2 landscapes.

Experimental Design and Vegetation Sampling

We used Gentry’s (1982) protocol to sample vegetation in45 randomly selected forest patches (15 per landscape).We chose this method because it is logistically simple,economical in terms of both time and money, and ap-propriate for the analysis of species diversity in tropicalforests (Gentry 1982). Within each patch, we randomlyselected 10, 50 × 2 m transects (0.1 ha of area sam-pled per patch). We recorded all species of trees, shrubs,lianas, palms, and herbs (i.e., Heliconiaceae) with a diam-eter at breast height (dbh) ≥2.5 cm. Species were classi-fied into 2 groups according to their light requirements—shade-tolerant (old-growth forest species) and shade-intolerant species (early colonizers or pioneers)—according to Martınez-Ramos et al. (1989) and Flora de Ve-racruz fascicles (available from www.ecologia.edu.mx/publicaciones/p_periodicas.htm). The dominance of pi-oneer species is indicative of forest disturbance (Hill &Curran 2003; Santos et al. 2008). Species not identifiedin the field were collected for subsequent identification

Conservation Biology

Volume 23, No. 3, 2009

732 Conservation Value of Small Patches

Figure 2. Relationship between plant species density

and patch area in 3 landscapes with different

deforestation levels (LDL, lowest level of deforestation;

IDL, intermediate deforestation; HDL, highest

deforestation). The number of species was determined

in 0.1 ha (10, 50 × 2 m sample plots per patch).

in the MEXU (Institute of Biology, UNAM, Mexico City)and XAL herbaria (Institute of Ecology A.C., Xalapa, Ver-acruz; see further details in Arroyo-Rodrıguez & Mandu-jano [2006] and Arroyo-Rodrıguez et al. [2007]).

Data Analyses

To control for differential spatial sampling across patchesof different sizes, we based our analyses on the plantspecies density (i.e., number of species per 0.1 ha) foundwithin each sampled patch. We pooled all patches (45)and tested the relationship of species density to patcharea with linear regression. A scatterplot (Fig. 2) of thedata seemed to show that the relationship changed overand below a patch-area threshold (i.e., a negative rela-tionship was evident over this patch-area threshold anda positive relationship was evident below this thresh-old). Therefore, we used a piecewise regression modelto identify this breakpoint because this statistical tech-nique identifies ecological thresholds (Toms & Lesper-ance 2003). To determine whether the proportion offunctional species groups depended on the amount of for-est cover in a landscape, we used a G test for contingencytables (Statsoft 1999). Differences in species density–arearelationships among landscapes were tested with an anal-ysis of covariance (ANCOVA) through a generalized linearmodel (species density = landscape + patch size + land-scape × patch size). As suggested for count-dependentvariables (i.e., species density), we selected a Poissondistribution with a log- link function and corrected foroverdispersion (Crawley 2002).

To assess the contribution that small patches madeto regional diversity, we used cumulative species–areacurves. First, we ordered patch area from small to largeand then calculated values for cumulative area and cumu-lative number of species. For comparison, we repeatedthe analysis with patches ordered from large to small(Quinn & Harrison 1988; Fischer & Lindenmayer 2002).Furthermore, following Fischer and Lindenmayer (2002),we created graphs that were not dependent on the distri-bution of patch sizes and plotted the cumulative numberof species against the number of patches added (ratherthan the cumulative area) again, sorting the patches fromsmallest to largest, and largest to smallest.

To evaluate patterns of species turnover within and be-tween landscapes, we constructed a triangular Q-modematrix of similarity taking into account the 20 most abun-dant species in each landscape with a Bray–Curtis index(or percentage of similarity), which is an effective coeffi-cient of similarity to use with crude values of abundanceof the species. The index ranged from 0 (no species incommon between samples) to 1 (identical species com-position and abundance). We used nonmetric multidi-mensional scaling (MDS) to define the overall differencesin species composition within and among landscapes.We chose MDS because neither normality nor linearity ofdata is required (Kruskal & Wish 1978). To evaluate howmany dimensions are needed to reproduce similarity be-tween sites, we calculated a stress value. The smaller thestress value, the better the fit of the reproduced similaritymatrix to the observed similarity matrix.

Results

We recorded 9451 plants belonging to 75 families, 207genera, and 364 species. The average species density perpatch was 58 (ranging from 23 to 84). Analyzing thefull set of patches (45), there was no significant relation-ship between species density and patch area (R2 = 0.04,p > 0.1). A high variability in species density occurredin the smallest patches; both the highest (84) and lowest(23) number of species was found in patches of up to1.8 ha (Table 1; Fig. 2). The piecewise regression modelexplained 63% of the variance and indicated that speciesdensity increased with area (S = 44.6 + 4.77 log10 area)in patches of up to approximately 5 ha (the breakpoint).In patches over 5 ha, species density decreased with area(S = 58.4–4.50 log10 area). Given the 5-ha breakpoint,most of the species sampled (294 species, 81%) occurredin patches of up to 5 ha (n = 26), whereas 245 species(67%) occurred in patches between 5 and 20 ha (n = 10),and 222 species (61%) occurred in patches of >20 ha(n = 9). Certainly, this could be due to the large numberof small patches sampled in the 3 landscapes (a samplingeffect). Thus, when we randomly selected 9 small patches

Conservation Biology

Volume 23, No. 3, 2009

Arroyo-Rodrıguez et al. 733

Table 1. Plant species density (number of species/0.1 ha) within 45 rainforest patches in 3 landscapes of Los Tuxtlas, Mexico, with differentdeforestation levels.∗

LDL IDL HDL

Patch patch size species density patch size species density patch size species density

1 700.0 52 75.6 53 13.3 572 265.5 60 57.2 62 8.0 743 88.0 60 32.6 54 3.8 494 81.1 71 30.0 71 3.8 585 22.2 73 13.0 70 3.0 676 9.4 65 11.8 60 2.2 567 6.4 65 7.4 71 2.1 588 5.3 69 6.7 67 2.0 569 4.9 67 6.6 65 1.9 38

10 4.9 62 3.8 58 1.8 2811 4.1 36 3.0 67 1.3 4812 3.0 33 2.7 61 1.3 3313 1.7 82 1.3 76 1.1 4914 1.6 57 1.1 84 0.9 4215 1.1 65 1.0 55 0.7 23

Total 242 161 181Mean (SD) 61 ± (13) a 65 ± (9) a 49 ± (14) b

∗Abbreviations: LDL, lowest deforestation level, 24% of remaining forest cover; IDL, intermediate deforestation, 11%; HDL, highest deforestation,

4%. For mean, different letters indicate significant differences (p < 0.001 in all cases) among landscape units (with contrast tests; Crawley

2002).

(<5 ha), we found 221 species, a trend almost identical tothat in the largest patches (>20 ha, 222 species). Overall,most of the species were restricted to only a few patches(e.g., 41% of the species sampled were distributed onlyin 1 or 2 patches, and almost 70% were distributed in 5patches; Fig. 3).

Eight of the species sampled (2.5%) are classifiedas endangered by the Mexican government (Calophyl-

Figure 3. Frequency distribution of plant species in

different numbers of patches. Total frequency is

indicated for each landscape (LDL, lowest level of

deforestation; IDL, intermediate deforestation; HDL,

highest deforestation).

lum brasiliense, Chamaedorea alternans, Geonoma

oxycarpa, Mortoniodendron guatemalense, Spondias

radlkoferi, Talauma mexicana, Tetrorchidium rotun-

datum, and Vatairea lundellii) (SEMARNAT 2002). Ahigh proportion of these plants (n = 110 individuals ofthese species, 48%) were found in patches of up to 5 ha.Only 61 endangered plants (27%) were found in patchesbetween 5 and 20 ha, and 58 endangered plants (25%)were found in patches of >20 ha. Again, the high propor-tion of endangered species in the smallest patches couldbe a consequence of the great number of small patchesin the study region. Most of the total species sampled(360 species, 99%) are native to the region, and only 4(1%) are human-introduced species (Citrus sp., Coffea

arabica, Theobroma cacao, and Manguifera indica),which represents only 0.4% of the individuals sampled.All individuals of C. arabica (19) and T. cacao (5) oc-curred in a 4.9-ha and a 266-ha patch of LDL, but M.

indica (3 individuals), and Citrus sp. (10 individuals) oc-curred more frequently in small patches of HDL (5 of 15sampled patches in this landscape were occupied by atleast one of these species). Finally, most of the speciessampled were old growth (203 species, 56%), and only67 species (18%) were pioneers. The proportion of func-tional groups was independent of the remaining forestcover in the landscape units (p > 0.05 in all cases). Never-theless, a slightly lower proportion of old-growth specieswas observed in both the smallest patches and in HDL,particularly for animal-dispersed species, whereas the op-posite trend was observed in animal-dispersed pioneerspecies (Fig. 4).

Conservation Biology

Volume 23, No. 3, 2009

734 Conservation Value of Small Patches

Figure 4. Percentage and absolute number (in boxes)

of plant species belonging to different functional

groups (old-growth and pioneer species) with different

dispersal syndromes (animal dispersed and other)

within (a) 3 patch size ranges (n = 9 randomly

selected patches for each range) and (b) 3 landscapes

with different deforestation levels (LDL, lowest level of

deforestation; IDL, intermediate deforestation; HDL,

highest deforestation). Other dispersal syndromes

include wind dispersed, gravity dispersed, and

unknown.

Differences among Landscapes

Species density differed significantly between landscapes(ANCOVA, χ2 = 20.1, df = 2, p < 0.001; Table 1). Thecovariate (patch size) affected species density (χ2 = 5.1,df = 1, p < 0.05), but the between-species density andpatch area significantly differed among landscapes (χ2 =10.9, df = 2, p < 0.01), with a significant positive slopeonly in HDL (Fig. 2).

In the landscapes we examined, the cumulative speciesrichness for a given cumulative area was consistentlyhigher when small patches were added first (Fig. 5). Nev-ertheless, when the cumulative number of species wasplotted against the number of patches added, in LDL andIDL the species accumulation curves were similar regard-less of whether small or large patches were added first(Figs. 5a & 5b). In HDL the cumulative species richnesswas notably higher when large patches were added first(Fig. 5c).

The MDS ordination (stress value = 0.075) of the 20most abundant species in each landscape revealed 3 dif-ferent groups; thus, each landscape hosted plant com-munities that for the most part had distinct compositions(Fig. 6). There were certain species that were dominant insmall patches in highly fragmented landscapes (e.g., thepioneer shrub Stemmadenia donnell-smithii), whereasother species were dominant in larger patches in land-scapes of lower fragmentation (e.g., the old-growth palmAstrocaryum mexicanum). Nevertheless, no changes inthe total proportion of species within functional groupswere observed because old-growth species dominated all3 landscapes.

Discussion

Importance of Small-Sized Patches

Overall, our results suggest that small patches are ex-tremely valuable for maintaining regional plant diversity.First, despite the small size of these patches, they aremade up of diverse native plant communities, includingendangered and economically important forest species.Altogether, the 364 species sampled represented 58% ofthe dicotyledonous species reported for the Los Tuxt-las Field Station (627 spp.; Ibarra-Manrıquez et al. 1995,1996a, 1996b), and most of these species (81%) werefound in patches smaller than 5 ha. Although the patcheshad a high percentage of pioneer species, the old-growthforest species were dominant. Furthermore, of the 148useful plant species reported for the Los Tuxtlas rain-forest (not including epiphytes; Ibarra-Manrıquez et al.1997), 108 (73%) were sampled in our study patches. Sec-ond, because most of the remaining rainforest patches inthe region are very small and are still threatened by defor-estation (75% of the 208 digitized patches were of up to 5ha; Arroyo-Rodrıguez et al. 2007, 2008), biodiversity con-servation will be possible if small, intermediate, and largepatches are protected and managed in a complementarymanner.

Considering our standardized sample size (0.1 ha), thespecies density in several very small patches was higheror similar to the larger patches. Actually, the sharp in-crease in cumulative species richness (Fig. 5) suggests

Conservation Biology

Volume 23, No. 3, 2009

Arroyo-Rodrıguez et al. 735

Figure 5. Cumulative number of

plant species sampled in 0.1 ha

per patch versus both cumulative

patch area and number of

patches added in landscapes with

different degrees of deforestation:

(a) lowest deforestation, (b)

intermediate deforestation, and

(c) highest deforestation. In all

graphs, patches were added from

small to large or large to small,

respectively, and then

corresponding cumulative species

counts were obtained.

that small patches are occupied by many plant species.Similar findings have been reported for other organisms(e.g., birds, Fischer & Lindenmayer 2002; insects, Estrada& Coates-Estrada 2002; Benedick et al. 2006) includingplants (e.g., dos Santos et al. 2007). Finally, we foundthat the distribution of most species was restricted tofew patches, and each landscape had a set of patcheswith markedly distinct communities. This finding is con-sistent with previous analyses in which we demonstratedthat in Los Tuxtlas there is a high species turnover (β di-versity) among forest patches and landscapes (F. Escobar,V. Arroyo-Rodrıguez E. Pineda, and J. Benıtez-Malvido,unpublished data). This result also highlights the impor-tance of conserving as many patches as possible, includ-ing the smallest (Fischer & Lindenmayer 2002; dos Santoset al. 2007).

Factors Influencing Plant-Species Density within Patches

Variations in species density within patches could be aconsequence of several processes acting at different spa-tial scales. At the landscape scale, the relationship be-tween species density and patch area was positive andsignificant only in the most deforested landscape (HDL),whereas in the other 2 landscapes the relationship wasnot significant. The results of several studies support theidea that below a certain proportion of suitable habi-tat (e.g., <10–30%: Andren 1994; <20%: Fahrig 1998)landscape attributes such as patch size and isolation be-come very important factors in explaining the distribu-tion, abundance, and richness of species. Thus, becauseforest cover in HDL was much lower than this range ofvalues (only 4%), we expected landscape attributes, such

Conservation Biology

Volume 23, No. 3, 2009

736 Conservation Value of Small Patches

Figure 6. Multidimensional scaling ordination of the

20 most abundant plant species across each landscape

(LDL, lowest level of deforestation; IDL, intermediate

deforestation; HDL, highest deforestation).

as patch area, to strongly influence plant species richnessin this landscape. Patch isolation was significantly higherin HDL; therefore, increased isolation distances in thislandscape can modify the dispersal and colonization ca-pacity of numerous plant species, which will affect floris-tic composition and species richness within the patches(Arroyo-Rodrıguez & Mandujano 2006; Arroyo-Rodrıguezet al. 2007).

On the other hand, extinction cascades are particularlylikely to occur in landscapes with low native vegetationcover, especially if keystone species or entire functionalgroups of species are lost (Terborgh 1992; Fischer &Lindenmayer 2007). The Los Tuxtlas region has sufferedfrom a severe defaunation process (i.e., several species ofmammal dispersers have disappeared; Dirzo & Miranda1991), which may be greater in HDL due to its loweramount of forest cover. For example, in this region asforest cover decreases, the distribution and abundanceof howler monkeys (Alouatta palliata) decreases too(Arroyo-Rodrıguez et al. 2008). This primate species isan important seed disperser for many native trees (e.g.,Estrada & Coates-Estrada 1991; Andresen 2002), so the re-duction of howler populations can affect the distributionand structure of many tree species (e.g., Julliot 1997).Thus, the abrupt reduction in patch size and increasedisolation, along with a severe defaunation process, couldexplain the overall low species richness we found in HDL.

At the patch scale, species density was highly vari-able and surprisingly higher in the smallest patches thanin the largest ones. This result is consistent with thosereported by dos Santos et al. (2007) for several semide-ciduous forest patches in southeastern Brazil, and maybe related to factors that contribute to species extinctionand increase or maintain species richness. As we statedpreviously, several relatively well-understood factors pre-

dict that species richness will be low in small patches.For example, both island biogeography and metapopu-lation theories predict higher extinction rates in smallerpatches (e.g., MacArthur & Wilson 1967; Hanski 1999).Smaller patches may sustain smaller populations, whichincreases the probability of extinction resulting fromenvironmental, demographic, and genetic stochasticity(Ellstrand & Elam 1993; Fischer & Lindenmayer 2007).Species density can also decrease because habitat diver-sity may be lower in small patches (Connor & McCoy1979). Furthermore, considering that edge effect exertsits greatest influence within 100 m of the forest edge (Lau-rance et al. 1998), most of the patches sampled (38 vs. 7patches) do not have relatively unaltered interior habitat,and edge effects are expected to be important in mostpatches. For instance, tree mortality may increase nearthe edges (Laurance et al. 1998), with the most damagesuffered by emergent old-growth trees (Benıtez-Malvido1998; Hill & Curran 2003), which could be the reason wedetected slightly lower proportions of old-growth speciesin smaller patches.

Large patches have a higher probability of beingrandomly occupied by a given individual or speciesthan smaller patches (i.e., passive sampling) (Connor &McCoy 1979). Thus, we expected species richness to belower in small patches than in large ones. In HDL cumula-tive species richness increased faster when larger patcheswere added first, and we detected a positive a species–area relationship. For LDL and IDL, however, we did notfind any significant relationship between species densityand patch area and the species accumulation curves weresimilar regardless of whether small or large patches wereadded first. This result suggests that for LDL and IDLseveral plant species have the same probability of beingfound in large or small patches, emphasizing the value ofsmall patches to biodiversity conservation, particularly inlandscapes with lower deforestation level.

Several factors may explain the maintenance of plantbiodiversity in small patches. First, deforestation and frag-mentation in Los Tuxtlas has been most severe in theproductive areas of the lowlands (Guevara et al. 2004).Therefore, some small patches are located in high-qualityenvironments that can maintain a high number of species.Second, although edge effects may decrease the numberof old-growth plant species, the increment in gap open-ings and light incidence near edges can favor the estab-lishment and growth of many pioneer species (Benıtez-Malvido 1998; Hill & Curran 2003; Arroyo-Rodrıguez &Mandujano 2006). This process can be particularly im-portant in younger patches (Santos et al. 2008), likethe ones we studied (30–40 years old; Guevara et al.2004). Furthermore, many ecological processes, such aslocal species extinctions, take some time following habi-tat loss and fragmentation (Tilman et al. 2002). Becausemany tropical trees are long lived (Martınez-Ramos &Alvarez-Buylla 1998), extinction of many species may not

Conservation Biology

Volume 23, No. 3, 2009

Arroyo-Rodrıguez et al. 737

have occurred yet. We found that most species weredistributed in a few patches, so population isolationcould negatively affect the reproductive success of manyspecies (Jansen 1986), potentially leading to future plantspecies extinctions in the region. A study including vege-tation with dbh <2.5 cm could show greater differencesgiven the young age of the forest patches.

Implications for Biodiversity Conservation

On the basis of our results, we believe that in Los Tuxtlas,it is critical to conserve as many rainforest patches as pos-sible because the rainforest is one of the most deforestedecosystems in the region (5% of the original cover remain-ing; Castillo-Campos & Laborde 2004) and because smallpatches of <5 ha contain high plant diversity. Further-more, small patches can be used as stepping stones thatincrease the landscape’s connectivity and facilitate inter-patch movements, and they may have complementary orsupplementary resources that may favor survival of manyanimal species in fragmented landscapes (Dunning et al.1992; Asensio et al. 2008). Finally, small patches can pro-vide important services, such as crop pollination by wildinsects (e.g. bees) and seed dispersal by primates andother animals that live in these small patches (Bodin et al.2006).

Small patches by themselves, however, cannot providesufficient habitat for viable populations of many species(Zuidema et al. 1996; Arroyo-Rodrıguez et al. 2007). Smallpatches can be less productive, potentially acting as sinkhabitats (Pulliam 1988; Dunning et al. 1992). When sinkhabitats increase in a landscape, population decline is ex-pected to occur if source habitats do not increase (Dun-ning et al. 1992). Small patches can also act as “ecologicaltraps” to which individuals of a species are attracted butin which they cannot reproduce (Battin 2004).

In contrast to our previous study, in which we fo-cused on a single landscape unit in the region (Arroyo-Rodrıguez & Mandujano 2006), our results here suggestthat species density in a patch of a given size may varyin landscapes with different deforestation levels. There-fore, the conservation value of a patch depends on thetotal forest cover remaining in the landscape. The conser-vation value of these small patches might also decreaseover time because extinctions may occur generations af-ter fragmentation (Tilman et al. 2002). The small sizes ofthe remaining populations, the absence of source habi-tats, and the absence of animal dispersers in the region,among other factors, can affect ecological processes suchas pollination and seed dispersal, making tree recruitmentand establishment difficult within the patches. Therefore,to maintain plant biodiversity in Los Tuxtlas, it is criticalto conserve as many patches as possible and to increasehabitat amount and decrease patch isolation by restoringsmall and intermediate patches.

Although the urgent need to conserve the most exten-sive areas of well-protected rainforest cannot be forgot-ten, in highly deforested and fragmented regions, such asLos Tuxtlas, biodiversity conservation will only be possi-ble by preserving and restoring small- and intermediate-sized remnants concurrently (Hill & Curran 2003; Arroyo-Rodrıguez & Mandujano 2006; dos Santos et al. 2007).Nevertheless, under current circumstances, rather thanfocus on the creation of several small- or medium-sizedisolated reserves, a regional-scale planning approach isrecommended to rescue plant biodiversity from local ex-tinctions (da Silva & Tabarelli 2000). The people fromseveral villages of the region are creating small private re-serves, and if their management efforts are coordinated,they may be able to conserve forest patches at a landscapescale, which could have a greater impact on conservationof the rainforest.

Acknowledgments

We thank B. Gomez-Chagala and R. Mateo-Gutierrez andfamily for their invaluable help. M. Peredo-Nava (XALherbarium) and G. Castillo-Campos helped with taxon-omy and the identification of specimens. This studywas funded by the Departamento de Biodiversidad yEcologıa Animal (Instituto de Ecologıa A.C.), SEMARNAT-CONACYT (project 2004-56-A1), and the CONANP. J. L.Hill, J. Fern, and E. Andresen made valuable suggestionson the manuscript.

Literature Cited

Andren, H. 1994. Effects of habitat fragmentation on birds and mammalsin landscapes with different proportion of suitable habitat: a review.Oikos 71:340–346.

Andresen, E. 2002. Primary seed dispersal by red howler monkeys andthe effect of defecation patterns on the fate of dispersed seeds.Biotropica 34:261–272.

Arroyo-Rodrıguez, V., A. Aguirre, J. Benıtez-Malvido, and S. Mandujano.2007. Impact of rain forest fragmentation on a structurally impor-tant palm species: Astrocaryum mexicanum Liebm. at Los Tuxtlas,Mexico. Biological Conservation 138:198–206.

Arroyo-Rodrıguez, V., and S. Mandujano. 2006. The importance of trop-ical rain forest fragments to the conservation of plant species diver-sity in Los Tuxtlas, Mexico. Biodiversity and Conservation 15:4159–4179.

Arroyo-Rodrıguez, V., S. Mandujano, and J. Benıtez-Malvido. 2008. Land-scape attributes affecting patch occupancy by howler monkeys(Alouatta palliata mexicana) at Los Tuxtlas, Mexico. AmericanJournal of Primatology 70:69–77

Asensio, N., V. Arroyo-Rodrıguez, J. C. Dunn, and J. Cristobal-Azkarate.2008. Conservation value of landscape supplementation for howlermonkeys living in forest patches. Biotropica (in press).

Battin, J. 2004. When good animals love bad habitats: ecological trapsand the conservation of animal populations. Conservation Biology18:1482–1491.

Benedick, S., J. K. Hill, N. Mustaffa, V. K. Chey, M. Maryati, J. B. Searle,M. Schilthuizen, and K. C. Hamers. 2006. Impacts of rain forest

Conservation Biology

Volume 23, No. 3, 2009

738 Conservation Value of Small Patches

fragmentation on butterflies in northern Borneo: species richness,turnover and the value of small fragments. Journal of Applied Ecol-ogy 43:967–977.

Benıtez-Malvido, J. 1998. Impact of forest fragmentation on seedlingabundance in a tropical rain forest. Conservation Biology 2:380–389.

Bodin, O., M. Tengo, A. Norman, J. Lundberg, and T. Elmqvist. 2006.The value of small size: loss of forest patches and ecologicalthresholds in southern Madagascar. Ecological Applications 16:440–451.

Bongers, F., J. Popma, J. Meave, and J. Carabias. 1988. Structure andfloristic composition of the lowland rain forest of Los Tuxtlas,Mexico. Vegetatio 74:55–88.

Castillo-Campos, G., and J. Laborde. 2004. La vegetacion. Pages 231–265in S. Guevara, J. Laborde, and G. Sanchez-Rıos, editors. Los Tuxtlas.El paisaje de la sierra. Instituto de Ecologıa, A. C., y Union Europea,Xalapa, Mexico, D.F.

CONABIO (Comision Nacional para el Conocimiento y Uso de la Bio-diversidad). 2000. Regiones terrestres prioritarias para la conser-vacion. CONABIO, Mexico, D.F.

Connor, E. F., and E. D. McCoy. 1979. The statistics and biology of thespecies-area relationship. The American Naturalist 113:791–833.

Crawley, M. J. 2002. Statistical computing. An introduction to dataanalysis using S-Plus. John Wiley & Sons, London.

da Silva, J. M. C., and M. Tabarelli. 2000. Tree species impoverishmentand the future flora of the Atlantic forest of northeast Brazil. Nature404:73–74.

Dirzo, R., and M. C. Garcıa. 1992. Rates of deforestation in Los Tuxtlas,a Neotropical area in Veracruz, Mexico. Conservation Biology 6:84–90.

Dirzo, R., and A. Miranda. 1991. Altered patterns of herbivory anddiversity in the forest understory: a case study of the possible con-sequences of contemporary defaunation. Pages 273–287 in P. W.Price, T. M. Lewinsohn, G. W. Fernandes, and W. W. Benson, edi-tors. Plant–animal interactions: evolutionary ecology in tropical andtemperate regions. John Wiley & Sons, New York.

dos Santos, K., L. S. Kinoshita, and F. A. M. dos Santos. 2007. Tree speciescomposition and similarity in semideciduous forest fragments ofsoutheastern Brazil. Biological Conservation 135:268–277.

Dunning, J. B., B. J. Danielson, and H. R. Pulliam. 1992. Ecological pro-cesses that affect populations in complex landscapes. Oikos 65:169–175.

Ellstrand, N. C., and D. R. Elam. 1993. Population genetic consequencesof small population size: implications for plant conservation. AnnualReview of Ecology and Systematics 24:217–242.

Estrada, A., and R. Coates-Estrada. 1991. Howler monkeys (Alouatta

palliata), dung beetles (Scarabaeidae) and seed dispersal: ecologicalinteractions in the tropical rain forest of Los Tuxtlas, Mexico. Journalof Tropical Ecology 7:459–474.

Estrada, A., and R. Coates-Estrada. 2002. Dung beetles in continuousforest, forest fragments and in an agricultural mosaic habitat is-land at Los Tuxtlas, Mexico. Biodiversity and Conservation 11:1903–1918.

Fahrig, L. 1998. When does fragmentation of breeding habitat affectpopulation survival? Ecological Modelling 105:273–292.

Fischer, J., and D. B. Lindenmayer. 2002. Small patches can be valuablefor biodiversity conservation: two case studies on birds in south-eastern Australia. Biological Conservation 106:129–136.

Fischer, J., and D. B. Lindenmayer. 2007. Landscape modification andhabitat fragmentation: a synthesis. Global Ecology and Biogeography16:265–280.

Gentry, A. H. 1982. Patterns of Neotropical plant species diversity.Evolutionary Biology 15:1–85.

Guevara, S., J. Laborde, and G. Sanchez-Rıos. 2004. Los Tuxtlas. Elpaisaje de la sierra. Instituto de Ecologıa, A. C., y Union Europea,Xalapa, Mexico, D.F.

Hanski, I. 1999. Metapopulation ecology. Oxford University Press, Ox-ford, United Kingdom.

Hill, J. L., and P. J. Curran. 2003. Area, shape and isolation of trop-ical forest fragments: effects on tree species diversity and im-plications for conservation. Journal of Biogeography 30:1391–1403.

Hill, J. L., and P. J. Curran. 2005. Fragment shape and tree species com-position in tropical forests: a landscape level investigation. AfricanJournal of Ecology 43:35–43.

Ibarra-Manrıquez, G., and S. Sinaca-Colın. 1995. Lista florıstica comen-tada de la Estacion de Biologıa Tropical “Los Tuxtlas,” Veracruz,Mexico. Revista de Biologıa Tropical 43:75–115.

Ibarra-Manrıquez, G., and S. Sinaca-Colın. 1996a. Estacion de BiologıaTropical “Los Tuxtlas,” Veracruz, Mexico: lista florıstica comentada(Mimosaceae a Verbenaceae). Revista de Biologıa Tropical 44:41–60.

Ibarra-Manrıquez, G., and S. Sinaca-Colın. 1996b. Lista florıstica comen-tada de plantas de la Estacion de Biologıa Tropical “Los Tuxtlas,”Veracruz, Mexico: (Violaceae-Zingiberaceae). Revista de BiologıaTropical 44:427–447.

Ibarra-Manrıquez, G., M. Ricker, G. Angeles, S. Sinaca-Colın, and M. A.Sinaca-Colın. 1997. Useful plants of the Los Tuxtlas rain forest (Ver-acruz, Mexico): considerations of their market potential. EconomicBotany 51:362–376.

Jansen, D. H. 1986. The future of tropical biology. Annual Review ofEcology and Systematic 17:305–324.

Julliot, C. 1997. Impact of seed dispersal by red howler monkeysAlouatta seniculus on the seedling population in the under-storey of tropical rain forest. The Journal of Ecology 85:431–440.

Kruskal, J. B., and M. Wish. 1978. Multidimensional scaling. Sage Publi-cations, Beverly Hills, California.

Laurance, W. F. 2005. When bigger is better: the need for Ama-zonian megareserves. Trends in Ecology & Evolution 20:645–648.

Laurance, W. F., L. V. Ferreira, J. M. Rankin-de Merona, and S. G. Lau-rance. 1998. Rain forest fragmentation and the dynamics of Amazo-nian tree communities. Ecology 79:2032–2040.

MacArthur, R. H., and E. O. Wilson. 1967. The theory of island biogeog-raphy. Princeton University Press, Princeton, New Jersey.

Martınez-Ramos, M., and E. R. Alvarez-Buylla. 1998. How old are tropicalrain forest trees? Trends in Plant Science 3:400–405.

Martınez-Ramos, M., E. R. Alvarez-Buylla, and J. Sarukhan. 1989. Treedemography and gap dynamics in a tropical rain forest. Ecology70:555–558.

Peres, C. A. 2001. Synergistic effects of subsistence hunting and habi-tat fragmentation on Amazonian forest vertebrates. ConservationBiology 15:1490–1505.

Pulliam, H. R. 1988. Sources, sinks, and population regulation. TheAmerican Naturalist 132:652–661.

Quinn, J. F., and S. P. Harrison. 1988. Effects of habitat fragmentationand isolation on species richness: evidence from biogeographic pat-terns. Oecologia 75:132–140.

Santos, B. A., C. A. Peres, M. A. Oliveira, A. Grillo, C. P. Alves-Costa,and M. Tabarelli. 2008. Drastic erosion in functional attributes oftree assemblages in Atlantic forest fragments of northeastern Brazil.Biological Conservation 141:249–260.

Schwartz, M. W. 1999. Choosing the appropriate scale of reserves forconservation. Annual Review of Ecology and Systematics 30:83–108.

Secretaria del Medio Ambiente y Recursos Naturales (SEMARNAT).2002. Norma oficial Mexicana NOM-059-SEMARNAT-2001, lista deespecies en riesgo. Diario Oficial de la Federacion, Mexico, D.F.

Soto, A. 2004. El clima. Pages 195–198 in S. Guevara, J. Laborde, and G.Sanchez-Rıos, editors. Los Tuxtlas. El paisaje de la sierra. Institutode Ecologıa A.C., and Union Europea, Xalapa, Mexico.

Conservation Biology

Volume 23, No. 3, 2009

Arroyo-Rodrıguez et al. 739

StatSoft Inc. 1999. Statistica. Computer manual for Windows. Statsoft,Tulsa, Oklahoma.

Tabarelli, M., J. M. C. da Silva, and C. Gascon. 2004. Forest fragmen-tation, synergisms and the impoverishment of Neotropical forests.Biodiversity and Conservation 13:1419–1425.

Terborgh, J. 1992. Maintenance of diversity in tropical forests. Biotrop-ica 24:283–292.

Tilman, D., R. M. May, C. L. Lehman, and M. A. Nowak. 2002. Habitatdestruction and the extinction debt. Nature 371:65–66.

Toms, J. D., and M. L. Lesperance. 2003. Piecewise regression: a toolfor identifying ecological thresholds. Ecology 84:2034–2041.

Zuidema, P. A., J. A. Sayer, and W. Dijkman. 1996. Forest fragmentationand biodiversity: the case for intermediate-sized conservation areas.Environmental Conservation 23:290–297.

Conservation Biology

Volume 23, No. 3, 2009