Free-Feeding Insect Herbivores Along Environmental Gradients in Serra Do Cipo: ´ Basis for a Management Plan

Journal of Insect Conservation, 2, 107–118 (1998)

Free-feeding insect herbivores along environmental gradients in Serra do Cipo: ´ basis for a management plan

S´ervio P. Ribeiro*, Marco A.A. Carneiro and G. Wilson Fernandes

Evolutionary Ecology of Tropical Herbivores/DBG, ICB/Universidade Federal de Minas Gerais, C.P. 486, 30160–970, Belo Horizonte, MG, Brazil

Received 13 February 1998; revised and accepted 17 April 1998

The distribution of free-feeding insect herbivores in Brazilian savanna was studied in the National Park of Serra do Cipó. Insect samples were obtained with sweep nets across cerrado (savanna), rupestrian field and altitudinal grassland vegetation from 800 to 1500 m above sea level. We found a low species richness in xeric and mesic habitats during both wet and dry seasons. Sap- sucking insects were the most abundant guild (53.4%) with Cicadellidae the most abundant family (27.2%). The hypothesis that taxon richness of free-feeding insects decreases with increasing altitude was supported in xeric habitats during the wet season only, mainly as a function of mountain summit effect. There was a decrease of 65% in the number of families occurring at 1400 and 1500 m compared with lower elevations. The exclusion of sites of rupestrian vegetation at mid-elevations from the analysis increased significantly the proportion of variance explained by the model. An examination of taxon distribution using canonical variate analysis supported this result. The hypothesis that mesic habitats are richer in species of free-feeding insect herbivores than are xeric habitats was not supported. The data indicate that plant sclerophylly may exert a strong negative influence on insect species richness, and that variation due to particular characteristics of each site strongly affected the studied guilds. The present results should inform conservation strategies for the National Park Management Plan, which is currently being developed.

Keywords: altitudinal gradient; cerrado; free-feeding insect herbivores; high altitude grassland vegetation; sclerophylly.

Introduction processes in natural ecotones and human expanded habitat edges (review of landscape ecology and insect Patterns of insect species distribution in the tropics conservation in Samways, 1995). have not been sufficiently explored, despite many Whilst most studies point to a negative correlation recent studies (e.g. Leakey and Proctor, 1987; Fer- between species richness and altitude, mid-elevation nandes and Price, 1988, 1991; McCoy, 1990; Olson, 1994; Lopatin, 1996). In montane ecosystems, where climate peaks of insect richness have been reported for tropical variation and natural habitat fragmentation are com- mountains (Janzen, 1973; Young, 1982; McCoy, 1990). mon traits (Olson, 1994; Boggs and Murphy, 1997; Fer- Severe physical conditions associated with high alti- nandes et al., 1997; Haslett, 1997), it is important to tudes have been used to explain the decreased species understand the variation in species distribution within number in mountain ecosystems (Mani, 1962; Smith and between habitats, its ecological meaning and its and Young, 1987; Wolda, 1987; Boggs and Murphy, relevance to species conservation (Janzen, 1967). Spe- 1997). However, the importance of spatial complexity cies endemisms and insect assemblages are influenced of mountain ecosystems and its effects on invertebrate by landscape patches, and thus easily affected by communities have been recognized recently (Haslett, impacts like anthropogenic fragmentation or change in 1997), and may nullify a gradual decrease in species forest composition (Dempster, 1991; Samways, 1995; number along gradients (Ribeiro et al., 1994; Ribeiro Brown, 1997). Distribution of insect species richness and Fernandes, 1998). along environmental gradients corresponds, to some Significant changes in insect species richness are extent, to the patches in the environmental matrix. greater on mountain tops or on rough slopes due to a Therefore, consideration of specific habitat character- conspicuous rise in climatic severity or the harshness of istics in studies of gradients may elucidate ecological the habitat (Sarmiento, 1986; but see Leakey and Proc-

* To whom correspondence should be addressed at: CABI Bioscience UK Centre and Imperial College at Silwood Park, Buckhurst Road, Ascot, Berks SL5 7TA, UK.

tor, 1987; Fernandes and Price, 1988). However, since gallery forests in comparison with the surrounding most studies on species richness along altitudinal gra- sclerophyllous cerrado vegetation. They expected sim- dients do not include replicate samples at any given ilarities in physical conditions among gallery forests altitude (e.g. Janzen, 1973; Janzen et al., 1976; Krysan et along the altitudinal gradient, due to relatively consis- al., 1984; Wolda, 1987; McCoy, 1990), they do not tent structure and physical conditions in these mesic account for the relative importance of horizontal hab- habitats. itat variation. Recent reviews of spatial heterogeneity The aim of this work was to study the variation in in montane ecosystems are found in Boggs and Mur- species richness of free-feeding insect herbivores along phy (1997) and Haslett (1997). an altitudinal gradient in the Serra do Cipó. Two The Serra do Cipó is a high altitudinal area within hypotheses were tested: (i) that insect species richness the Espinhaço Mountain Range, in the south-eastern decreases with increasing altitude (e.g. Pianka, 1966), Brazilian Central Plateau. The most pristine and com- tested in xeric and mesic habitats independently (see plex mountains were declared a National Park of Fernandes and Price, 1988), both in wet and dry sea- 33800 ha in 1984. However, a strategy for use and man- sons; and (ii) that mesic habitats are more favourable, agement of this Park has yet to be developed, and eco- constant and therefore, richer in free-feeding insect her- logical data are needed. The region is covered by bivores than xeric habitats in both dry and wet seasons cerrado (savanna) vegetation which changes in struc- (Fernandes and Price, 1988). Implications for the elabo- ture gradually with altitude. There is a predominance ration of a management plan for the National Park are of xeric habitats, and most plant species are sclerophyl- discussed in the conclusion section. lous. The landscape is composed of patches of different plant species, though there are various levels of overlap Methods: study sites within any environmental gradient (Eiten, 1972, 1978; The study sites were established in the region of the Giulleti and Pirani, 1988). Serra do Cipó National Park, in Minas Gerais State, The arboreal cerrado occurs at lower altitudes (800 Brazil (between 19° 15' and 19° 30' S, and 43° 30' and and 900 m). Between 900 and 1000 m the cerrado is 43° 55' W). The altitudinal range in the study area present only as isolated patches on deeper lateritic varies from 800 to 1500 m. The study region is in the soils. On quartzitic rocks, a rupestrian vegetation dom- Brazilian Central Plateau which is, on average, 800 m inates, i.e. a combination of sclerophyllous shrubs high, with an area of over 1500000 km2, almost all growing in cracks or gaps between rocks, and frag- covered with cerrado vegetation (Goodland and Ferri, ments of high altitude grassland vegetation. It occurs 1979). Three sample sites widely distant from each from 900 m up to the summit (1500–1600 m) and con- other (Ͼ 1 km) were established for every 100 m of alti- tains several endemic plant species. Most belong to tude from 800 to 1500 m. Each sample site comprised genera which are highly diverse within, or unique to, one xeric habitat sample point and the closest mesic the Neotropical region, such as Paepalanthus (Eriocaula- habitat sample point. Cerrado was sampled at 800 and ceae), Vochysia (Vochysiaceae), Xyris (Xyridaceae) and 900 m, rupestrian vegetation between 900 and 1200 m, Tibouchina (Melastomataceae). High altitude grassland and high altitude grassland between 1200 and 1500 m. predominates on sand quartzitic soils from 1200 m up The mesic habitats (gallery forests) were sampled from to the summit, and is more frequent and continuous 800 to 1400 m. Only two sample sites could be estab- than the rupestrian vegetation in the study area. In lished at 1400 m because there were few springs at this companion studies using the same sampling sites, a altitude surrounded with gallery forests. For a detailed negative effect of altitude on gall-forming and mining site description see Ribeiro and Fernandes (1998) and insects and ants was found in these xeric habitats (Lara Meguro et al. (1996a,b). and Fernandes, 1996; Fernandes et al., 1997). Neverthe- less, it appears that this pattern does not apply to free- Sampling of insect herbivores feeding herbivorous insects (Ribeiro et al., 1994; Carneiro et al., 1995; Fernandes et al., 1997). Insects were sampled in 1991, in the wet (January– Narrow corridors of moist gallery forest predomi- February) and dry (July–August) seasons. In the dry nate along streams and rivers across the altitudinal season, we began sampling four months after rain. range, up to 1400 m (Meguro et al., 1996a,b). Fernandes Samples were taken with a sweep-sample net of 38 cm and Price (1988, 1991, 1992) argued that the species diameter. The sweeps were done by the same person- richness of free-feeding insects should be favoured by nel in all sample sites, always between 09:00 and microclimate and physical conditions associated with 16:00 h, though never while there was rain or strong

108 Tropical insect herbivores along environmental gradients

wind. Both seasonal samples were completed in signed-rank test (Zar, 1984; Wilkinson, 1989) was approximately three days. Sites from high, middle and used. low altitudes were randomly chosen to be sampled Ecological communities in mountain ecosystems are within each day. Both herbaceous and arboreal (when inadequately characterized by species richness esti- present) strata were sampled. In rupestrian vegetation, mates alone (Haslett, 1997). The distribution of insect samples were taken mainly in fragments of grassland, abundance by taxa was analysed using canonical vari- on sedimentary sandy soil between the rocks. Five sub- ate analysis, or discriminant analysis, hereafter, CVA samples of 30 sweeps were taken walking along one (Digby and Kempton, 1987; Chatfield and Collins, line of approximately 30 m, at each sample point. 1996). This analysis was performed only for data from Altogether, the 150 sweeps by point, 300 by sample site xeric habitats during the wet season, when insect spe- (two sample points per sample site: xeric and mesic), cies richness varied with altitude and when the differ- provided a total of 6600 sweeps per season. ential effects of habitat types were identified (see The cumulative sum of morphospecies from each set results below). Insects were grouped by recognizable of 150 sweeps of a sample point was used to estimate higher taxa level, namely superfamily, family and sub- species richness. All adult insect herbivores were con- family, or by feeding guild when individuals were not sidered, with the exception of flower feeders (Canthar- identified, since there were no biological data or reli- idae, Dasytidae and some Brachypnoea spp. able taxonomy available for most species. However, it (Chrysomelidae), known flower eating species), is possible to infer from family level some biological because flowers are typically a non-sclerophyllous information about the individuals. Further analyses of resource (Ribeiro et al., 1994). However, it is probable taxa distribution along altitude for the xeric versus that there were not many species among these exclu- mesic habitats are detailed in Carneiro et al. (1995). ded taxa, since the two sampling seasons were not dur- Canonical correspondence analysis (CCA) has been ing the flowering period for most plant species in the considered the most accurate technique to identify region. community gradients along multi-dimensional Insects were identified to morphospecies and esti- environment data (Ter Braak, 1986; Digby and Kemp- mates of species richness may be conservative, owing ton, 1987). CVA is the variant to CCA when just one p- to the possibility of the presence of cryptic species. For dimensional environment variable is used to separate instance, cryptic species are known to be frequent in species abundance (Chatfield and Collins, 1996). In this the Chrysomelidae (Young, 1982; Lopatin, 1996), an study, the altitude is an axis that defines gradual hab- abundant family in the region (Ribeiro et al., 1994; Car- itat and, consequently, community changes, fitting the neiro et al., 1995). Conversely, similarity between spe- assumptions of the CVA model. Therefore, an altitude- cies, due to common adaptations to vegetation, altitude by-taxa group matrix was computed, using the abun- and high solar radiation exposure might be confused dance of the 19 most common taxa (listed in Table 2), with high colour and size variability of some species of according to the procedures in Ter Braak (1986), Basset Chrysomelidae, which could lead to incorrect assigna- (1992) and Basset and Samuelson (1996). Each altitude tion of morphospecies (Basset and Samuelson, 1996; was considered as a one dimensional variate and sites Knoll et al., 1996). Therefore, the present study concen- were replicates of these variates, grouped by insect trated on distribution patterns and not on general taxa. diversity estimates. Since most families were distributed equally along sites (see discriminant analysis Results: general patterns below), different bias in species richness estimates A total of 4351 specimens of free-feeding insect herbi- along altitudes or between xeric and mesic habitats was vore, belonging to 38 families, were sampled during not anticipated. two seasons on the altitudinal gradient, in both xeric Data analysis and mesic habitats. Hemiptera (adopted as Homoptera and Heteroptera united) was the most abundant order The altitudinal gradient effect on insect species richness (52.4%), while the Cicadellidae was the most abundant was analysed using simple linear regression models. family (27.2% out of all families). Orthoptera represen- Although altitudinal data per habitat were distributed ted 23.8% of sampled insects, followed by Coleoptera normally, there was some pattern in the residues of the (20%), Blattodea (3.4%) and Phasmodea (0.2%). overall data set. Therefore, in order to compare xeric When separated into feeding guilds, sap-sucking and mesic habitats, and wet and dry season, Wilcoxon composed 56.2% of identified morphospecies against

109 S.P. Ribeiro et al.

43.8% of chewing insects. During the wet season, the test z ϭ 2.78, p Ͻ 0.01, Fig. 1A). However, in mesic hab- proportions of sap-sucking and chewing insects were itats the number of insect species did not change approximately the same in both xeric (51.8% and between the dry and wet season (Wilcoxon signed-rank 48.2%, respectively) and mesic (53.2% and 46.8%, test z ϭ 1.9, p Ͼ 0.05, Fig. 1B). respectively) habitats. A similar trend was found in the dry season, however the proportion of sap-sucking was Hypothesis of altitudinal gradient slightly higher than that of chewing in xeric (60.6% and There was no consistent relationship between species 39.4%, respectively) and mesic (60.3% and 38.8%, richness and altitude. In the wet season, insect species respectively) habitats. richness decreased with increasing altitude in xeric The overall number of morphospecies was low per 2 habitats (r ϭ 0.33, y ϭ 29.96 Ϫ 0.013x, F1,22 ϭ 11.002, site (mean ϭ 6.0; range: 0–20) and per altitude p Ͻ 0.01 Fig. 2A), but not in mesic habitats (r2 ϭ 0.01, (mean ϭ 31.0; range: 11–42). In xeric habitats in the dry p Ͼ 0.05). In the dry season, altitude did not affect season, the number of species sampled was low across insect richness in xeric habitats (r2 ϭ 0.03, p Ͼ 0.05). all altitudinal gradients, and was also signiﬁcantly However, contrary to expectation, species richness lower than in the wet season (Wilcoxon signed-rank increased with increasing altitude in the mesic habitats

Table 1. CVA (taxa abundance by altitude) for the main functions: eigenvalues, % variance explained, canonical discriminant functions, and pooled within-groups correlation coefﬁcients

Functions

Variable Func. 1 Func. 2 Func. 3 Func. 4 Eigenvalues 31.0299 4.4550 1.1038 0.6626 Variance explained % 81.13 11.65 3.12 1.73

Standardized canonical discriminant function coefﬁcients X800 1.27215 Ϫ0.01775 0.13489 Ϫ0.44002 X900 0.31016 0.13622 0.20110 0.64176 X1000 Ϫ0.01283 0.48901 0.12913 0.65293 X1100 0.52854 Ϫ0.19769 Ϫ0.22703 0.43502 X1200 0.20217 0.93693 Ϫ0.58065 Ϫ0.33621 X1300 1.06738 Ϫ0.98502 Ϫ0.23250 Ϫ0.32765 X1400 0.41947 Ϫ0.03181 0.93537 0.05418 X1500 0.03688 1.12150 0.07643 Ϫ0.19564

Pooled within-groups correlations between discriminating variables and canonical discriminant functions X1400 0.20292 0.01978 0.65530 Ϫ0.27143

X1200 0.19891 0.35282 Ϫ0.38326 Ϫ0.24143 X800 0.21479 0.17216 0.12543 0.03306

X1300 0.38336 Ϫ0.23922 Ϫ0.17775 Ϫ0.21240 X900 0.10211 0.10808 0.08591 0.48597 X1100 0.29764 Ϫ0.15627 Ϫ0.37710 0.39997

X1500 0.14653 0.24197 0.12982 Ϫ0.21470 X1000 0.16794 0.24640 Ϫ0.05068 0.53698

110 Tropical insect herbivores along environmental gradients

in the dry season (r2 ϭ 0.39, y ϭϪ1.74 ϩ 0.014x, tation physiognomy on species richness of free-feeding

F1,18 ϭ 11.756, p Ͻ 0.01, Fig. 2B). insects were performed. It was found that the lowest To test the effect of altitude further and search for a values in species richness observed at intermediate alti- more ecological explanation for species distribution tudes came from sites of rupestrian vegetation. This patterns, a second set of linear regression models with- vegetation increased the variability in species number out the 1400 and 1500 m sample sites were made. No between sites. A new model omitting values from relationship between species richness of free-feeding rupestrian vegetation sites explained 59% of variance insect herbivores and altitude was found when these in species richness (r2 ϭ 0.59, y ϭ 35.87 Ϫ 0.017x, data were excluded from the analysis. Even the pos- F1,14 ϭ 7.921, p Ͻ 0.001) in comparison with 33.3% in the itive relationship found in the xeric habitats in the wet previous model, but did not change the slope of the 2 season was not supported (r ϭ 0.03, p Ͼ 0.05). This model (t ϭ 0.732, p Ͼ 0.05), indicating that there is an analysis suggests that the decline in richness with alti- expressive variance in the number of species within a tude was an artefact of severe reduction in species rich- certain altitudinal range. ness at the highest sites within the altitudinal gradient and not due to a steady decline in species richness with Change in taxa abundance distribution altitude. with altitude A new set of analyses examining the effects of vege- Discriminant analysis functions were computed for xeric habitats in the wet season. The abundance of the 19 studied taxa varied within each altitude (U-statistics, p Ͻ 0.05), except at 900 m (U-statistics, p Ͻ 0.13) where most taxa occurred, but in low densities. The first and second discriminant functions explained 92.78% of data variation (function 1 ϭ 81.13% and function 2 ϭ 11.65%). Table 1 summarizes the eigenvalues and correlation coefficients of this analysis. The first function was highly correlated with altitudes 800 m and 1300 m whilst the second function was correlated with altitudes 1200 m, 1300 m and 1500 m. Canonical discriminant function by group centroids showed Cicadellidae, Acrididae, and Tettigoniidae as having the highest correlation coefficients with function 1 (Table 2). Basically, this means that these families were distinctive for the number of individuals in 800 m and 1300 m, especially Cicadellidae in the former altitude. Also Cicadellidae, Acrididae and Curculionidae had the highest correlation coefficients with function 2, showing high abundance of these families (and, therefore, total high abundance) for 800, 1200 and 1300 m (Table 2). The CVA diagram of function 1 versus function 2 (Fig. 3) distinguished Cicadellidae, Acrididae and Curculionidae from other taxa, highly clustered in the co-ordinates of intermediate sites. Also, Tettigonii- dae appeared slightly separated from the other taxa. The occurrence of families reflected the pattern found for species richness: there was an average of 20.2 taxa per altitude between 800 and 1300 m (range: Figure 1. Species richness of free-feeding insects in wet and dry 18–23), and just 13 taxa at 1400 and 1500 m, which season in Serra do Cipó. The number of species decreased in the means a loss of 27% of taxa from 1300 to 1400 m and an dry season in (A) xeric (Wilcoxon signed-rank, p Ͻ 0.05), but not in overall loss of 65% of identified taxa at the highest (B) mesic habitats (Wilcoxon signed-rank, p Ͼ 0.05). elevation.

111 S.P. Ribeiro et al.

Cicadellidae were abundant throughout all altitudes, tribution of most insect taxa at low and intermediate with highest values at 800 m (77 individuals), and at altitudes; (ii) a high abundance of few taxa at some one specific site at 1500 m (21 individuals). Acrididae specific sites, however not clearly correlated with the were most abundant at 800 and 1300 m (each altitude altitudinal gradient; and finally (iii) a sharp decrease in with 38 individuals), Curculionidae at 1200 m (18 indi- the occurrence of taxa at the highest sites. Therefore, viduals), while Tettigoniidae had 24 individuals sam- the discriminant analysis reinforces the influence of pled at 1300 m. mountain tops and habitat effects on species richness of The discriminant analysis showed: (i) a uniform dis- free-feeding insect herbivores.

Figure 2. Species richness of free-feeding insects in Serra do Cip´o (A) in xeric habitats, in wet season, along the altitudinal gradient (blanket points delimited are rupestrian vegetation sites). Model estimated for all sites: y ϭ 29.96 Ϫ0.013x. (B) In mesic habitats, in dry season, along the altitudinal gradient. Model estimated from every site: y ϭϪ1.74 ϩ 0.014x.

112 Tropical insect herbivores along environmental gradients

Hypothesis of favourable conditions in munity, may have influenced the low free-feeding mesic habitats insect species richness across the whole studied area, contrasting with other tropical mountains (Janzen, Although comparisons of wet versus dry season sug- 1973; Krysan et al., 1984; Wolda, 1987; Fernandes and gest that mesic habitats are more constant in species Price, 1988; McCoy, 1990). The Serra do Cipó is located richness than xeric habitats (Fig. 1), no statistical differ- in the Espinhaço mountain range, which has an old ence was found between the number of insect species and eroded relief, formed by the rise of Pre-Cambrian in both seasons when comparing mesic with xeric habitats (Wilcoxon signed-rank test: wet season z ϭ 0.13, and Cretaceous soils in the Tertiary period (Freitas, p Ͼ 0.05; dry season z ϭ 0.87, p Ͼ 0.05). Therefore, these 1951). Still in the Tertiary, the hot and sunny climate data did not support the hypothesis that mesic habitats and the already poor soils influenced the spread of a are more favourable to free-feeding insect herbivores, sclerophyllous plant community (Rizzini, 1979). Most even if they had been constant in species richness studies on insect diversity in tropical mountains were between seasons. made on higher and more recent areas, covered by richer soils (e.g. Janzen, 1973; Simpson, 1975; Wolda, Discussion 1987). The present basic assumption is that sclerophylly would have been favourable to the specialization of The prevalence of harsh environmental conditions, free-feeding herbivores, but not so favourable to their such as poor soils and a nutritionally poor plant com- speciation, owing to nutritional and chemical restric- tions and phylogenetic constraints on this guild (Coley Table 2. Canonical discriminant functions by group cent- et al., 1985; Salatino, 1993; Ribeiro et al., 1994; Turner, roids for functions 1, 2, 3 and 4 of the CVA analysis 1994). The pattern of distribution of free-feeding insect her- Functions bivores along altitudinal gradients was supported, Variable Func. 1 Func. 2 Func. 3 Func. 4 albeit weakly, by our data. Species declined with altitude only in the wet season in xeric habitats, but this a 10.77064 Ϫ4.57568 Ϫ1.20498 Ϫ0.19230 was primarily influenced by the harsh conditions on b Ϫ0.62629 0.19728 Ϫ0.16385 Ϫ0.02811 mountain tops (see Young, 1982; Lawton et al., 1987; c 0.55005 3.16429 Ϫ2.32477 Ϫ0.50633 Boggs and Murphy, 1997). Stochastic decreases in tem- d 3.74399 Ϫ2.16840 0.01684 0.05775 perature may result in a sudden reduction in insect e Ϫ2.13216 0.26665 Ϫ0.44412 Ϫ0.26066 abundance on the mountain tops during the early wet f 13.85264 4.22790 1.19086 0.10769 season (Galvão and Nimer, 1965), while intermediate and low altitudes are buffered by a more stable and g Ϫ2.85171 Ϫ0.11921 0.26027 Ϫ0.13679 warmed air mass, protected by a constant high solar h Ϫ3,22941 0.11634 Ϫ0.07917 Ϫ0.42307 radiation and physical wind barrier. Moreover, the pri- Ϫ i 1.93750 0.45621 0.25710 1.31818 mary cause of decline of insect species on mountain j Ϫ1.68641 0.13879 Ϫ0.56443 0.64445 tops seems to be influenced by increasingly harsh con- k Ϫ2.74289 Ϫ0.22312 0.11269 0.21098 ditions (Kingsolver and Watt, 1983, 1984; Lawton et al., l 0.15008 Ϫ0.77553 0.91700 Ϫ0.90424 1987; Boggs and Murphy, 1997), rather than reduction m Ϫ1.55791 Ϫ0.91138 2.17311 Ϫ0.29751 in plant species number or smaller overlap between n Ϫ2.81465 0.70237 Ϫ0.39083 0.37351 plant patches on the top (Eiten, 1972; Giulietti and Pir- o Ϫ1.63651 0.17405 Ϫ0.23406 Ϫ0.97121 ani, 1988; Ribeiro et al., 1994). The combination of these p Ϫ0.34404 Ϫ0.49955 Ϫ0.22846 1.58578 characteristics can severely reduce species diversification (Knoll et al., 1996), and benefit a few resource spe- q Ϫ1.82460 Ϫ0.26247 0.45149 0.62313 cialist species, as described for species of Brachypnoea in r Ϫ2.31491 0.02354 0.34156 Ϫ0.66272 Serra do Cipó (Ribeiro et al., 1994). s Ϫ3.36840 0.46249 Ϫ0.08625 Ϫ0.53853 The effect of habitat diversification at low and inter- Group centroids code: a, Acrididae; b, Alticinae; c, Curculionidae; d, mediate altitudes reduced the explanatory capability of Tettigoniidae; e, Membracidae; f, Cicadellidae; g, Scuttelleridae; h, the regression model for xeric habitats in the wet sea- Megascelinae; i, Blattodea; j, Pentatomidae; k, Priorinae; 1, Fulgor- son. This result indicates that the complexity of habitats idae; m, Dictyopharidae; n, Pseudococcinae; o, Cercopidae; p, Cryp- across any geographical range, along with different tocephalinae; q, Tingidae; r, Delphacidae; s, Cicadidae. levels of habitat fragmentation, may be an important

113 S.P. Ribeiro et al.

Figure 3. CVA diagram for the 19 taxa examined, with plot of canonical functions 1 and 2. Taxa code letter explained in Table 2. Some taxa codes were not presented for sake of clarity. determinant of species richness (Ribeiro and Fer- hypothesis considered mesic habitat homogeneous nandes, 1998). For instance, some highly diverse sites at compared with xeric habitats, the coefficients of varia- 1200 m and 1300 m could be interpreted as intermedi- tion of species number among sites for mesic (0.27) and ate peaks, but this was not a consistent pattern among xeric (0.25) habitats were both low and extremely sim- all intermediate sites and, in addition, the few peaks ilar in the wet season. Therefore, the pattern of higher were defined primarily by Cicadellidae. free-feeding insect species richness in mesic than in The maintenance of higher richness and abundance xeric habitats, obtained by Fernandes and Price (1988, of sap-sucking, especially Cicadellidae, than of other 1992), may have been influenced by some locally rich insect groups along altitudes was one important com- mesic sites, though the really favourable conditions for ponent of the variability at intermediate sites. Cicadelli- these insects in gallery forests are not understood com- dae species did not decrease between seasons in xeric pletely and do not exist in all mesic habitats. All gallery habitats (Carneiro et al., 1995), and the present results forests are narrow (between 2 and 20 m) and are indicated that sap-sucking insects can perform better restricted to immediate borders of streams and washes than chewing insects in both xeric and mesic habitats in (Meguro et al., 1996a,b). The surrounding soils (xeric the dry season. These data are in accordance with the habitats) are poor, and any autochthonous nutrients hypothesis that sclerophylly has negative effects on would be leached by the tropical storms. Therefore, free-feeding insects, while sap-sucking may avoid these habitats may be composed, similarly to xeric hab- some of the sclerophyllous traits by feeding directly itats, of nutrient-poor and chemically deterrent plant from the plant’s vascular system. species. The maintenance of similar numbers of insect species in mesic habitats between both seasons contrasts with Conclusions: insect species richness and the decline of species number in xeric habitats and, in a habitat conservation first analysis, corroborates Fernandes and Price’s (1988, 1992) hypothesis. However, there was high variability Preservation of pristine, unique habitats and endemic among the mesic sites. Although Fernandes and Price’s species, scientific research, conservation associated

114 Tropical insect herbivores along environmental gradients

with education and sustainability of surrounding to be balanced against potential threats to local hab- agricultural communities, and public leisure are prior- itats. ity objectives of the National Park of Serra do Cipó The data indicate that gallery forests in Serra do Cipó (G.W. Fernandes, unpublished technical report). may represent a less buffered system than previously Reliable ecological data are required in order to define thought, as they may have an intense interaction with where a certain activity may occur, or which manage- the surrounding xeric habitats. These may have resul- ment technique should be applied. Knowledge of ted in similar richness of free-feeding insects and taxa insect assemblages has been used to evaluate habitats composition to those of the adjacent xeric habitats. in a biogeographic and landscape ecology perspective Conversely, mesic habitats along streams and rivers (Dempster, 1991; Samways, 1995; Stork and Samways, may represent important sites for speciation and main- 1995; McGeoch and Chown, 1998). In the case of the tenance of diversity in the savannas (Smith et al., 1997). National Park of Serra do Cipó, further surveys and a Despite general similarities, the richest sampled sites more precise definition of species composition of were in mesic habitats connected with patches of forest. assemblages (and their ecological functionality) are To what extent this high species richness could be asso- required for accurate habitat labelling and evaluation ciated with the patches’ natural condition or human of conservation status (Dufrêne and Legendre, 1997). disturbance needs to be further explored. Due to past However, the present results suggest aspects of insect agricultural activities, patches of high altitude forest assemblages and their habitats which should be and cerrado have been impacted by logging and fire. explored further, and provide valuable baseline Consequently, these patches, as well as several areas of information for the setting of conservation pri- gallery forest, might require rehabilitation projects and careful protection against fire, since they might support orities. a less fire resistant fauna than xeric habitats (Louzada Our study emphasizes that the development of con- et al., 1996; Naves, 1996). As for rupestrian vegetation servation strategies in this region must consider the and mountain tops, natural fragmentation of mesic landscape as a mosaic of habitat types, rather than just habitats and their small sizes may exacerbate their a set of altitudinal zones. The pattern of species rich- vulnerability. ness described here underlines two contrasting compo- The same adverse conditions that influenced the nents of the landscape that may require special high diversification of gall-forming insects (see Fer- attention: the extremely low species richness of xeric nandes and Price, 1991) may have restricted the diver- sites in the mountain tops and in the rupestrian vegeta- sification of free-feeding insects in the Serra do Cipó. tion, and the unexpectedly low species richness in The need for specialization, plus the complexity of the mesic habitats. landscape, may have created in the Serra do Cipó an Beyond a simplified analysis based on comparison of ecosystem diverse in terms of insect survival strategies, diversity among sites, there is evidence that the paucity despite the apparent low number of observed species. of species on the mountain tops and in the rupestrian The point of concern is the vulnerability of the special- vegetation may result from special characteristics of ist, endemic and rare species, and the resistance or these sites, such as natural fragmentation, small sizes resilience of this community as a whole. and environmental harshness (Janzen, 1967; Leakey Provided that basic ecological information is avail- and Proctor, 1987; Olson, 1994; Boggs and Murphy, able, targeted experimental work can optimize efforts 1997). These particular habitats may support a distinc- to establish criteria for land management plans or con- tive fauna, perhaps more vulnerable as a consequence servation priorities. Insect herbivore assemblages in of its spatial constraints (Young, 1982). For instance, a Serra do Cipó have been well studied, and ecological companion study on Brachypnoea distribution showed and evolutionary hypotheses to explain patterns of that this genus is extremely abundant on top sites, and species distribution in areas of cerrado have been its occurrence is associated with several flowering developed and tested over the last 10 years (Fernandes montane plants (Ribeiro et al., 1994). The likely fragility and Price, 1988; Giulietti and Pirani, 1988; Lara and of this habitat is compounded by pressure of human Fernandes, 1996; Meguro et al., 1996a; Ribeiro and Fer- disturbance. Opening of roads at the top of mountains nandes, 1998). Knowledge of local insect and plant to facilitate tourist access is common practice. Holistic communities has identified habitat mosaics as one of management of these sites will require public access the key components of this ecosystem, and provides (approval of road building – or denial in favour of foot- the foundation for further research devoted to identify- paths – and road/footpath design on approved routes) ing specific management requirements.

115 S.P. Ribeiro et al.

Acknowledgements of insect populations. In The conservation of insects and their habitats (N.M. Collins and J.A. Thomas, eds) pp. We thank L. Fox, Y. Basset, G. Masters, O. Cheesman 143–53. London: Academic Press. and N. Springate for their criticism of earlier drafts of Digby, P.G.N. and Kempton, R.A. (1987) Multivariate analysis this paper, Y. Basset for helpful advice about multivari- of ecological communities. London: Chapman and Hall. ate analysis, and T. G. Cornelissen for drawing the final Dufrêne, M. and Legendre, P. (1997) Species assemblages and figures. The Conselho Nacional de Desenvolvimento indicator species: the need for a flexible asymmetrical Cient´ıfico e Tecnológico and US-Fish and Wildlife Serv- approach. Ecol. Monogr. 67, 345–66. ice/Fundação Biodiversitas provided a scholarship to S.P. Ribeiro. Financial support was provided by Con- Eiten, G. (1972) The cerrado vegetation of Brazil. Bot. Rev. selho Nacional de Desenvolvimento Cient´ıfico e Tecno- 38, 201–341. lógico (CNPq – 52.1772/95–8), the Fundação de Eiten, G. (1978) Delimitation of the cerrado concept. Vege- Amparo a` Pesquisa de Minas Gerais (FAPEMIG – tatio 36, 169–78. 821/90, 1950/95), and Fundação Biodiversitas to G.W. Fernandes, G.W. and Price, P.W. (1988) Biogeographical gra- Fernandes. The Parque Nacional da Serra do Cipó- dients in galling species richness: tests of hypotheses. IBAMA provided logistic support. This study was Oecologia 76, 161–7. in partial fulfilment of requirements for the MSc degree Fernandes, G.W. and Price, P.W. (1991) Comparison of trop- in Ecology, Conservation and Wildlife Management of ical and temperate galling species richness: the role of S.P. Ribeiro at the Universidade Federal de Minas environmental harshness and plant nutrient status. In Gerais. Plant–animal interactions: evolutionary ecology in tropical and temperate regions (P.W. Price, T. Lewinsohn, G.W. Fer- nandes and W.W. Benson, eds) pp. 91–115. New York: References John Wiley. Fernandes, G.W. and Price, P.W. (1992) The adaptive sig- Basset, Y. (1992) Influence of leaf traits on the spatial dis- nificance of insect gall distribution: survivorship of spe- tribution of arboreal arthropods within an overstorey cies in xeric and mesic habitats. Oecologia 90, 14–20. rainforest tree. Ecol. Entomol. 17, 8–16. Fernandes, G.W., Aráujo, L.M., Carneiro, M.A.A., Cornelis- Basset, Y. and Samuelson, G.A. (1996) Ecological character- sen, T.G., Barcelos-Greco, M.C., Lara, A.C.F. and Ribeiro, istics of an arboreal community of Chrysomelidae in S.P. (1997) Padrões de riqueza de insetos em gradientes Papua New Guinea. In Chrysomelidae biology vol 2: Eco- altitudinais na Serra do Cipó, Minas Gerais. In Contribui- logical studies (P.H.A. Jolivet and M.L. Cox, eds) pp. 243–62. Amsterdam: SPB Academic Publishing. ção ao conhecimento ecológico do cerrado – trabalhos selecio- Bernays, E., Driver, G.C. and Bilgener, M. (1989) Herbivores nados do 3o Congresso de Ecologia do Brasil (L.L. Leite and and plant tannins. Adv. Ecol. Res. 19, 263–302. C.H. Saito, eds) pp. 191–5. Bras´ılia: Depto Ecologia, Uni- Boggs, C.L. and Murphy, D.D. (1997) Community composi- versidade de Bras´ılia. tion in mountain ecosystems: climatic determinants of Freitas, R.O. (1951) Ensaio sôbre o relêvo tectônico do Bra- montane butterfly distributions. Glbl Ecol. Biogeogr. Lett. 6, sil. Revta. Brasil. Geogr. 2, 171–220. 39–48. Galvão, M.V. and Nimer, E. (1965) Clima. In Geografia do Brown Jr, K.S. (1997) Diversity, disturbance, and sustainable Brasil – grande região leste (IBGE) pp. 71–117. Rio de use of Neotropical forests: insects as indicators for con- Janeiro: IBGE. servation monitoring. J. Insect Conserv. 1, 25–42. Giulietti, A.M. and Pirani, J.R. (1988) Patterns of geographic Carneiro, M.A.A., Ribeiro, S.P. and Fernandes, G.W. (1995) distribution of some plant species from the Espinhaço Artrópodos de um gradiente altitudinal na Serra do range, Minas Gerais and Bahia, Brazil. In Proceedings of a Cipó, Minas Gerais, Brazil. Revta. Brasil Entomol. 39, workshop on neotropical distribution patterns (P.E. Vanzolini 597–604. and W.R. Heyer, eds) pp. 36–69. Rio de Janeiro: Academia Chatfield, C. and Collins, A.J. (1996) Introduction to multivari- Brasileira de Ciência. ate analysis. London: Chapman and Hall. Goodland, R. and Ferri, M.G. (1979) Ecologia do cerrado. Belo Coley, P.D., Bryant, P. and Chapin, F.S. (1985) Resource avail- Horizonte: Itatiaia. ability and plant antiherbivore defense. Science 230, Haslett, J.R. (1997) Insect communities and the spatial com- 895–9. plexity of mountain habitats. Glbl Ecol. Biogeogr. Lett. 6, Dempster, J.P. (1991) Fragmentation, isolation and mobility 49–56.

116 Tropical insect herbivores along environmental gradients

Janzen, D.H. (1967) Why mountain passes are higher in the McCoy, E.D. (1990) The distribution of insects along eleva- tropics. Am. Nat. 101, 233–49. tional gradients. Oikos 58, 313–22. Janzen, D.H. (1973) Sweep samples of tropical foliage insects: McGeoch, M.A. and Chown, S.L. (1998) Scaling up the value effects of season, vegetation types, elevation, time of day, of bioindicators. TREE 13, 46–7. and insularity. Ecology 54, 687–708. Meguro, M., Pirani, J.R., Mello-Silva, R. and Giulietti, A.M. Janzen, D.H., Ataroff, M., Farinas, M., Reyes, S., Rincon, N., (1996a) Establishment of riparian forests and forest island Soler, A., Soriano, P. and Vera, M. (1976) Changes in the – ‘capões’ – in the grassland ecosystems of Espinhaço arthropod community along elevational transect in the Range, Minas Gerais. Bolet. Bot. Univ. São Paulo 15, Venezuelan Andes. Biotropica 8, 193–203. 1–12. Kingsolver, J.G. and Watt, W.B. (1983) Thermoregulatory Meguro, M., Pirani, J.R., Mello-Silva, R. and Giulietti, A.M. strategies in Colias butterflies: thermal stress and the lim- (1996b) Floristic and structural characterization of ripar- its to adaptation in temporally varying environments. Am. ian forest island of Serra do Cipó, Minas Gerais. Bolet. Nat. 121, 32–55. Bot. Univ. São Paulo 15, 13–30. Kingsolver, J.G. and Watt, W.B. (1984) Mechanistic con- Naves, M.A. (1996) Efeito do fogo na população de formigas straints and optimality models: thermoregulatory strate- (Hymenoptera: Formicidae) em cerrado do Distrito Fed- gies in Colias butterflies. Ecology 65, 1835–39. eral. In Impactos de queimadas em areas´ de cerrado e rest- Knoll, S., Rowell-Rahier, M., Mardulyn, P. and Pasteels, J.M. inga (H.S. Miranda, C.H. Saito and B.F.S. Dias, eds) pp. (1996) Spatial genetic structure of leaf beetle species 161–177. Bras´ılia: UNB, ECL. with special emphasis on alpine populations. In Chrysome- Olson, D.M. (1994) The distribution of leaf litter inverte- lidae biology vol 1: The classification, phylogeny and genetics brates along a neotropical altitudinal gradient. J. Trop. Ecol. (P.H.A. Jolivet and M.L. Cox, eds) pp. 379–88. Amster- 10, 129–50. dam: SPB Academic Publishing. Pianka, E.R. (1966) Latitudinal gradients in species diversity: a Krysan, J.L., Branson, T.F., Schroeder, R.F.W. and Steiner Jr, review of concepts. Am. Nat. 100, 33–46. W.E. (1984) Elevation of Diabrotica sicuanica (Coleoptera: Price, P.W., Fernandes, G.W., Lara, A.C.F., Brawn, J., Gerling, Chrysomelidae) to the species level with notes on the D., Barrios, H., Wright, M., Ribeiro, S.P. and Rothcliff, N. altitudinal distribution of Diabrotica species in the Cuzco (1998) Global patterns in local number of insect galling Department of Peru. Entomol. News 95, 91–8. species. J. Biogeogr. In press. Lara, A.C.F. and Fernandes, G.W. (1996) The highest diversity Ribeiro, S.P. and Fernandes, G.W. (1998) Riqueza e abun- of galling insects: Serra do Cipó, Brazil. Biodivers. Lett. 3, dância de insetos maastigadores na Serra do Cipó. In 111–14. Lawton, J.H., MacGarvin, M. and Heads, P.A. (1987) Effects of Serra do Cipó: ecologia e evolução (G.W. Fernandes, ed). In altitude on the abundance and species richness of insect press. herbivores on bracken. J. Anim. Ecol. 56, 147–60. Ribeiro, S.P., Carneiro, M.A.A. and Fernandes, G.W. (1994) Leakey, R.J.G. and Proctor, J. (1987) Invertebrates in the litter Distribution of Brachypnoea (Coleoptera: Chrysomeli- and soil at a range of altitudes on Gunung Silam, a small dae) in an altitudinal gradient in a Brazilian savanna vege- ultrabasic mountain in Sabah. J. Trop. Ecol. 3, 119–29. tation. Phytophaga 6, 29–33. Lopatin, I.K. (1996) High altitude fauna of the Chrysomelidae Rizzini, C.T. (1979) Tratado de fitogeografia do Brasil. Vol 2. São of central Asia: biology and biogeography. In Chrysomeli- Paulo: Hucitec. dae biology vol 3: General studies (P.H.A. Jolivet and M.L. Salatino, A. (1993) Chemical ecology and the theory of oligo- Cox, eds) pp. 3–12. Amsterdam: SPB Academic Publish- trophic scleromorphism. An. Acad. Bras. Ciên. 65, 1–13. ing. Samways, M.J. (1995) Insect conservation biology. London: Louzada, J.N.C., Schiffler, G. and Vaz de Mello, F.Z. (1996) Chapman & Hall. Efeitos do fogo sobre a estrutura da comunidade de Sarmiento, G. (1986) Ecological features of climate in high Scarabaeidae (Insect, Coleoptera) na restinga de Ilha de tropical mountains. In High altitude tropical biogeography (F. Guriri-ES. In Impactos de queimadas em areas´ de cerrado e Vuilleumier and M. Monasterio, eds) pp. 11–45. New restinga (H.S. Miranda, C.H. Saito and B.F.S. Dias, eds) pp. York: Oxford University Press. 148–160. Bras´ılia: UNB, ECL. Simpson, B.B. (1975) Pleistocene changes in the flora of the Mani, M.S. (1962) Introduction to high altitude entomology. Insect high tropical Andes. Paleobiology 1, 273–94. life above the timber-line in the North-west Himalaya. Lon- Smith, A.P. and Young, T.P. (1987) Tropical alpine plant ecology. don: Methuen. Ann. Rev. Ecol. Syst. 18, 137–58.

117 S.P. Ribeiro et al.

Smith, T.B., Wayne, R.K., Girman, D.J. and Bruford, M.W. Turner, I.M. (1994) Sclerophylly: primarily protective? Func. (1997) A role for ecotones in generating rainforest bio- Ecol. 8, 669–75. diversity. Science 276, 1855–57. Wilkinson, L. (1989) SYSTAT: The System for Statistics. Systat Stork, N.E. and Samways, M.J. (1995) Inventorying and mon- inc, Evanston, Wyoming. itoring of biodiversity. In Global Biodiversity Assessment Wolda, H. (1987) Altitude, habitat and tropical insect diver- (V.H. Heywood, ed.) pp. 453–543. Cambridge: Cam- sity. Biol. J. Linn. Soc. 30, 313–23. bridge University Press. Young, A.M. (1982) Population biology of tropical insects. New Ter Braak, C.J.F. (1986) Canonical correspondence analysis: a York: Plenum. new eigenvector technique for multivariate direct gra- Zar, J.H. (1984) Biostatistical Ecology. New Jersey: Prentice- dient analysis. Ecology 67, 1167–79. Hall.

118