Edge Effects and Nature Reserves: A Study of Protected Birds in Grasslands Adjacent to Agricultural Landscapes in a Central Brazilian Cerrado

Study Area

The Cerrado is a vegetation province that occupies about 2,000,000 km2 in Brazil, Paraguay and Bolivia, being the second largest ecosystem in extension in South America [41, 42]. It dominates central Brazil and extends through peninsulas and isolated patches to other vegetation provinces, such as the Atlantic Forest [43]. Most landscapes are dominated by cerrado sensu lato - a gradient of xeromorphic savanna vegetation including grasslands and woodlands. This gradient covers most uplands, where vegetation such as dry forests and rocky grasslands also occur. Valleys are mainly occupied by gallery forests, marshes and floodplain grasslands [44].

This study was conducted at Parque Nacional das Emas, a 132,000 ha nature reserve located in the state of Goias [17o49´- 18o28´ S and 52o39´- 53o10´ W], western Cerrado, southwestern Brazil (Figure 1). This park harbours most physiognomies typical of Cerrado´s landscapes [34]. Shrubby grasslands are the main matrix type. Other physiognomies within the park are savanna woodlands, riverine forests and flood plain grasslands with palms [34, 45].

Click to enlarge

Shrubby grasslands [campo sujo] are physiognomies that occur on strongly drained sites with deep water table and low soil fertility [44]. They have a dense grass layer, numerous shrubs and scattered trees [43, 46]. A high richness of plant species can be found in shrubby grasslands [47, 48].

Elevations range between 720 and 900 m, and most of the park consists of flat tableland covered by grasslands [45]. The regional climate is strongly seasonal and marked by two well defined periods: wet and dry seasons. Annual rainfall ranges between 1,200 and 2,000 mm. Most of the annual precipitation falls between October and April. The dry season occurs between May and September. Temperatures

may descend to 0 oC in the Winter [July] and reach 40 oC in the Summer [January, February] [49].

The park is surrounded by agricultural fields, pastures and remnants of native vegetation. Agriculture in this area is mostly intensive and includes a rotation system that involves the cultivation of soybean and corn plantations between October and March, and cotton and other cultures in other periods of the year. The most external portions of the park are firebreaks, that are bands of grassland and woodland vegetation managed for fire control and usually 50 m wide [34, 45]. An unpaved road separates firebreaks and non-managed vegetation within the park. A 10 m wide road with low car traffic lies between the park and adjacent agricultural fields and is mainly used for the transportation of grains produced in this Cerrado region [34]. In this study, I considered as boundary line the band of modified land located between agricultural fields and the non-managed grasslands that were surveyed. Thus, the boundary line included unpaved roads, firebreaks and the paved road.

Study Sites

A total of 36 study sites were selected. They were plots [500 m x 140 m] of non-managed shrubby grassland vegetation [hereafter called “grassland”] located at Parque Nacional das Emas. They were distributed along nine stripes, each including a plot in the four distance intervals [bands] defined from the inner size of the boundary line: 0-500 m, 500-1,000 m, 1,000-1,500 m and interior [ranging between 2,000 m and 3,000 m]. These stripes were distant at least 500 m from each other. With the exception of those of the interior band, sites were situated with their longest axis perpendicular to the boundary line.

Bird Surveys

Bird counts were carried out during three sampling periods between late October 2006 and early March 2007, with a 15-20 days interval between consecutive periods. This period corresponds to the breeding season of most grassland and woodland species and coincides with the wet season in the study area [50]. Seasonal biases were controlled through sampling each study site once within the three sampling periods. Thus, 108 samples were performed during this study.

Birds were surveyed through transect counts [51]. A sample consisted in walking along a 500 m transect through the middle of a site over 20 min and recording all birds heard and/or seen within its limits. Birds flying over the study sites without using the vegetation or other resources were not recorded. Three to four study sites were surveyed each morning, between sunrise and 08:00 h. To avoid time biases, the schedule of sampling of each study site was different in each of the sampling periods.

The taxonomy of bird species followed the Brazilian Committee of Ornithological Records - CBRO [52]. Bird species were grouped into seven trophic guilds following relevant literature [50]: carnivores, granivores, frugivores, insectivores, omnivores, necrophagous and nectarivores. Classification of species as endemic followed major publications on birds of the Cerrado region [53, 54, 55, 56].

Patterns of Species Responses

A total of 59 bird species belonging to 23 families were recorded in the 36 study sites. The GLMMs evaluated variation in the abundance of 36 selected species along the four distance bands [see graphs, figure caption and table for significance and p values]. Six species showed a significant negative slope, thus being less abundant in grasslands located at deeper portions of the park (Figure 2a & Table 2). They included four granivores [Volatinia jacarina, Sporophila plumbea, Saltatricula atricollis and Ammodramus humeralis], Neothraupis fasciata and Eupsittula aurea. The edge effect [variation in abundance] was detected at least until the 1,000-1,500 m band in N. fasciata abundance, until the 500- 1,000 m band in the case of S. plumbea, V. jacarina and E. aurea, while only the first band was different to interior for A. humeralis and S. plumbea. However, this last species was not detected in the interior band and the contrasts in its model included the 1,000-1,500 m band as the reference level. Three other species showed a marginally significant negative slope across the gradient of distance from the boundary line (Figure 2b & Table 2). Among these species, Cypsnagra hirundinacea had fairly constant abundances in the three more external distance intervals and decreased significantly in the interior. There is also some evidence of edge effects in Synallaxis albescens since its abundance in the more external band was close to significantly different from the interior, but in the case of Caracara plancus no differences were detected (Figure 2b & Table 2).

Click to enlarge

Figure 2: Mean abundance [± 1 SE] of bird species for which variation along four distance intervals in grasslands protected at Parque Nacional das Emas [SW Brazil] was detected using GLMM. Left column: species with a negative slope. Right column: species with positive slope. Graphs at the top include species belonging to the prevailing guild in each slope group: granivores [negative] and insectivores [positive]. Middle graphs: species with significant slopes belonging to other guilds. Bottom graphs: species with marginally significant slopes. Distance intervals are coded from the boundary line to the park core area as: 1 [0-500 m], 2 [500-1,000 m], 3 [1,000-1,500 m] and [4] interior. The significance of the contrasts comparing each of the first three distance bands with the interior band is shown close to significant mean symbols. In some cases, species codes are written close to significance symbols to avoid ambiguities. *: P < 0.05; : P < 0.01; *: P < 0.001. For marginally significant comparisons the p value is shown. Note the different y-axis scales for each graph. See Table 2 for species codes.

Table 2: Mean abundance [individuals/sample] of bird species estimated from three visits between October 2006 and February 2007 in four distance bands in protected grasslands at Emas National Park, southwestern Brazil. Distance bands are in relation to the boundary line between the park and adjacent agriculture: 0-500 m [P1], 500-1,000 m [P2], 1,000-1,500 m [P3] and interior [P4]. Code: four letters code used to identify species in the figures. Trophic guilds: insectivores [INS], omnivores [OMN], granivores [GRA], frugivores [FRU], carnivores [CAR], nectarivores [NEC] and necrophagous [NCR]. The column “P-value” gives values of significance for the slope of the GLMM relating abundance to distance band. The sequence and taxonomy of species followed the Brazilian Committee of Ornithological Records - CBRO [52].

On the other hand, the slope was positive and significant for five species, thus being more abundant in deeper portions of the park (Figure 2c & Table 2). These included Coryphaspiza melanotis, and four insectivores [Cistothorus plantesis and the tyrants Alectrurus tricolor, Culicivora caudacuta and Tyrannus savana]. Similarly, variation in the abundance of Colaptes campestris had a marginally significant positive slope, since this species presented a low abundance close to the boundary line but increased noticeably in the next distance intervals (Figure 2d & Table 2). Like in the negative slope species group, there was variability in minimum distance at which edge effects were detected, that reached in some species at least until 1,000-1,500 m band [C. caudacuta and C. melanotis], in others until 500-1,000 m [C. plantensis and A. tricolor] or just the first band [T. savana and Colaptes campestris].

The abundance of other 21 species did not vary significantly among the four distance bands. They could be broadly divided into two groups of species. First, were those that occurred with very similar numbers in the four distance categories, such as Rhynchotus rufescens, Nothura maculosa, Emberizoides herbicola and Sporophila bouvreuil (Table 2). The other group included species that occurred in higher or lower abundances in one or two of the intermediate distance bands [500-1,000 m or 1,000-1,500 m]. Thus, the slope across the sequence of the four distance bands was not positive or negative. Among them, some species had significant lower abundance in middle distances, such as Alipiopsitta xanthops in band 2, and Cariama cristata, Elaenia chiriquensis and Falco femoralis in band 3 (Table 2).

Most families with more than one species analyzed included at least one whose abundance was affected by the distance to edge, the exception being Accipitridae and Tinamidae. Species endemic to the Cerrado region tended not to vary in abundance with distance to the boundary line. Among them, two species [Cyanocorax cristatellus and Charitospiza eucosma] had too low numbers. The abundances of Euscarthmus rufomarginatus and A. xanthops did not show a positive or negative response to adjacent modified land. On the other hand, two species with a large number of records [C. hirundinacea, S. atricollis and N. fasciata] were more numerous in some of the more external distance intervals than in the other two bands.

Trophic Guilds

The GLMMs for trophic guilds showed that variation in abundance across the range of distances studied was significant for three guilds [see also graphs, figure caption and table for significance and p values]. Insectivores had a significant positive slope [slope = 0.183, z = 9.46, P < 0.001], thus increasing in abundance with increasing distances from the boundary line (Figure 3). On the other hand, abundance of omnivores and granivores decreased farther from the boundary [slope = -0.108, z = 4.05, P < 0.001; slope = -0.311, z = 12.29, P < 0.001, respectively]. Despite this similar pattern, the decrease in abundance between the 0-500 m and 500- 1,000 m distances bands seemed to be stronger for granivores than for omnivores. In these three guilds, the edge effect was detected at least until 500-1,000 m, but in omnivores and insectivores it seems to be acting also in the 1,000-1,500 m band since their abundance showed marginally significant differences with the interior. Slopes of the abundances of nectarivorous, frugivorous and necrophagous birds along the four distance bands were not significant (Figure 3).

Click to enlarge

Figure 3: Mean total abundance [±1 SE] of six trophic guilds along four distance intervals to the boundary line between non-managed vegetation in the Parque Nacional das Emas [SW Brazil] and adjacent agriculture. A] guilds for which the GLMM showed a significant relationship. B] guilds for which GLMM were not significant. Distance intervals are coded from the edge to the interior as: 1 [0-500 m], 2 [500-1,000 m], 3 [1,000-1,500 m] and 4 [interior]. The significance of the contrasts comparing each of the first three distance bands with the interior band is shown close to significant mean symbols. *: P < 0.05; : P < 0.01; *: P < 0.001. For marginally significant comparisons the p value is shown. Note the different y-axis scales for each graph.

The observed richness [Mao Tau estimator] considering the whole set of species was very similar in all distance bands and their confidence intervals overlapped broadly. This pattern of similar abundances across the preserved portion of the boundary between the park and agricultural fields also occurred for granivorous, omnivorous and insectivorous birds (Table 1).

Table 1: Species richness and 95% confidence intervals obtained from individual-based Mao Tau rarefaction curves, run for the whole set of species and the most abundant trophic guilds found in the four distance bands at Parque Nacional das Emas, southwestern Brazil. Species richness and confidence intervals are the values estimated for a number of individuals sampled equal to the number found in the distance band where less individuals were counted [marked with *].

Community Composition

Community composition was summarized by the Correspondence Analysis, which identified two major axes that altogether explained 42.5% of the total variation in the data set (Figure 4). Axis 1, which captured 29.9% of the variation, clearly described an interior-to-edge gradient in community composition. Grasslands in the 0-500 m distance band were the most distinct group. Sites in this distance band had low overlap with those of the 500-1,000 m, and were even more distinct from those of the interior and 1,000- 1,500 m distance bands. Sites in the 500-1,000 m band were in the middle of the gradient. Their overlap was higher with the interior and 1,000-1,500 m distance bands than with the 0-500 m band. Sites located in the interior and 1,000- 1,500 m distance bands overlapped widely and their species composition was less variable than the two more external bands, as shown by the concentrated distributions of these sites (Figure 4).

Click to enlarge

Figure 4: Results of the Correspondence Analysis performed with the abundance of 36 bird species detected in grasslands at four distance bands in relation to the boundary line between non-managed vegetation of Parque Nacional das Emas and adjacent agricultural fields. A] ordination of the 36 study sites. The nine replicates of each distance band are identified by specific symbols. B] ordination of the bird species recorded in the 36 study sites, and that followed the two criteria [ten or more individuals, recorded in five or more samples]. The symbols of species groups are defined by the type and significance of slope in the GLMM fitted to each species. See Table 2 for species codes.

Species more associated with the 0-500 m distance band were V. jacarina, S. plumbea and C. plancus (Figure 4). On other hand, the 500-1,000 m distance band was more related to species such as N. fasciata, S. atricollis, S. albescens, A. humeralis and E. aurea. Species more associated with the interior and 1,000-1,500 m distance bands were C. caudacuta, A. tricolor and C. melanotis (Figure 4).

Part of the larger variation of sites located in the two most external distance bands was due to their distribution along Axis 2 [Figure 4]. Species responsible for this variation were A. xanthops, E. aurea and S. plumbea, all of them species that form flocks that often move long distances and select particular sites in function of food availability. Of these three species, only A. xanthops did not have its abundance related to the distance to the boundary line (Table 2).

Changes in Community Composition and Abundance of Species and Guilds

Substantial variation across the range of distances to the boundary line studied was noted for the abundance of 15 species [42% of the total evaluated], the abundance of half of the trophic guilds examined and the overall community composition. These results are in agreement with my hypothesis that adjacent agriculture would lead to edge effects on bird communities inhabiting grasslands at Parque Nacional das Emas. Thus, impacts of intensive agriculture in Cerrado exceed the mere conversion of native vegetation [30, 31] and include also edge effects on its biodiversity inside reserves.

About one fifth of the species were detected to be less numerous at shorter than at greater distances from adjacent modified land. Thus, they would be the species most affected negatively if agribusiness settlement be juxtaposed to nature reserves in Cerrado, especially in cases of narrow and/or small conservation units, due to ratios between edge and core areas [4, 25, 62]. Nine other species showed an opposite pattern of abundance and thus are species more positively related to influences of adjacent modified land. These species are likely to compete with or even replace typical grasslands bird species along the more external portions of the park.

A relatively large group contained species whose abundance did not change across the boundary. This result is contrary to the relatively high proportion of bird species sensitive to edges commonly found in tropical forests [11, 22, 63]. This might reflect the fact that juxtaposed modified land has no influence on their abundances, likely the case of species with numerous individuals detected in several samples, such as E. herbicola, M. torquata and N. maculosa. On the other hand, no significant variation in the abundance of some species may reflect limitation of the sampling method or sample size. For example, large birds such as raptors and R. americana, and flocks of psitacids, were usually seen during the counts, but often >70 m from the observer, beyond the surveyed plots. In the case of low- density species, few individuals were detected and patterns of abundance variation may reflect a greater influence of chance. This latter fact also might explain the lack of variation among distance intervals in the abundance of frugivores, nectarivores and carnivores, as they occurred in low numbers.

On the other hand, the increase of omnivores near the park border might reflect their high flexibility in the acquisition of food items. Three tinamous [N. maculosa, R. rufescens and C. parvirostris] adapt well to disturbed vegetation in Cerrado [50, 64] and were commonly found in the more external distance bands, likely due to their flexibility to habitat modification. Three large species [R. americana, C. cristata and C. plancus] often feed opportunistically on roads, firebreaks and agricultural fields [pers. obs.], thus using frequently nearby protected grasslands.

Granivorous birds also had a dramatic increase in grasslands located close to adjacent agricultural land. This increase was mainly due to high numbers of V. jacarina, which was closely associated to exotic grasses. Other finches well adapted to pastures [64], such as S. plumbea and A. humeralis, also were found using exotic grasses that occurred mainly < 150 m from the adjacent firebreaks. However, this overall pattern of edge affinity showed by granivores was opposite to that shown by C. melanotis which increased in abundance in more interior grasslands. Requirement of seeds of native grasses and competition with other finch species may be involved in this strong and negative response to adjacent modified land.

Insectivores appeared to be severely affected by adjacent agricultural landscape, as their abundance decreased by approaching to the park border. This high sensitivity to adjacent modified land also has been found for guilds of insectivorous birds in tropical forests [11, 22]. I suspect that this pattern found at Parque Nacional das Emas might reflect, in part, low food availability near adjacent agriculture, as there is an influx of agrochemicals used in plantations that might kill invertebrates in the park.

The species richness of the whole communities and that of the most abundant trophic guilds did not vary with distance to the boundary line. This pattern of equal species richness across the range of grassland studied might result mainly of two factors. First, some species were ubiquitous in relation to distance to adjacent modified land and others were recorded only rarely. Second, some species and some trophic guilds showed opposite patterns of response to the proximity to adjacent modified landscape. Thus, a substitution of species tended to occur along the strips of grasslands examined at Parque Nacional das Emas, producing a change of community composition but keeping species richness fairly constant.

Strength of the Edge Effects in Cerrado and Other Regions

Knowledge on the magnitude and extent of edge effects is important to understand how biodiversity can be affected by adjacent matrix in human-modified landscapes [62]. In my study, I have been able to estimate the extent of edge effects penetration on grasslands at Parque Nacional das Emas. Analysis at the level of the community composition revealed that the influence of edge effects on grasslands leads to substantial changes in bird communities up to 1,000 m into the park. The analyses of trophic guilds showed that the penetration of edge effects can reach up to 1,500 m into the park, being very strong at less than 1,000 m from the boundary line. Considering all the analyses of this study, I conclude that strong edge effects penetrate up to 500 m into protected grasslands, becoming less intense but still intense up to 1,000 m from the boundary line, and moderately affecting some taxa, such as the threatened C. caudacuta and C. melanotis, up to 1,500 m into the Parque Nacional das Emas.

My study is the first to examine variation in bird communities across boundaries between modified land and remnant vegetation in Cerrado. Previously, two studies investigated predation rates of artificial and natural bird nests in forests remnants [<300 ha] surrounded by cleared areas in southeastern Cerrado, and found that predation rates in forest interiors and edges were comparable [40]. Similarly, no relationship between predation of artificial nests and distance to the modified matrix was found across a 2 km range of savanna in a reserve in central Brazil [65]. They also surveyed jays C. cristatellus [a nest predator] but no variation in abundance with distance to the border was detected.

Further studies of edge effects conducted in Cerrado´s nature reserves involved grasses and mammals. The African grass Brachyaria decumbens was highly abundant along a 10 m wide strip bordering the Cerrado Pé-de-Gigante reserve, while Melinis minutiflora was found through the entire extension of a small reserve in southeastern Brazil [37, 38]. Domestic dogs were considered to be an edge effect penetrating 400 m into Parque Nacional de Brasília [39].

In relation to birds in regions other than Cerrado, a few studies examined edge effects along great distances across boundaries between modified land and continuous vegetation in protected or intact landscapes. For example, main changes in the community of terrestrial insectivorous birds occurred up to 0.3 km into continuous forests adjacent to petroleum operations in Ecuador [22]. Also, for most guilds of studied understory birds, strongest edge effects occurred up to 70 m from an unpaved road crossing a continuous forest tract in the Brazilian Amazon [11]. Substantial changes in bird diversity and in the abundance of terrestrial insectivores occurred up to 200 and 600 m into a forest reserve surrounded by developed areas in Malaysia [24].

Therefore, the extent of edge effects detected in my study is similar to or greater than those observed in Cerrado´s nature reserves or for birds in extensive portions of continuous forests in other tropical regions. My study, conducted across a landscape dominated by non-forest native vegetation, reinforces the fact that edge effects can occur over large distances, as documented by scarce studies of plants in tropical forests [26, 62]. Due to the dramatic edge effects observed at Parque Nacional das Emas, I suggest that more attention should be devoted to grasslands in Cerrado, agreeing with studies that assessed their biodiversity [37, 47, 64, 66, 67].

Studies conducted worldwide have brought evidence that matrix types more appropriate for biodiversity conservation often are those with vegetation structure similar to that of remnants [68, 69]. Although I have not compared matrix types, my study suggests that dramatic edge effects on remnants can occur even when there is some similarity between the structure of the native vegetation and the exotic matrix, due in this case to the height of soybean plants being similar to most grassland plants [despite the complex structure and floral diversity of grasslands]. Thus, the degree and intensity of human manipulation of the matrix is of paramount importance.

Factors Involved in the Edge Effects

Several factors might contribute to the edge effects found in my study. First, roads established between the agricultural fields and non-managed vegetation in the park might contribute to variation in the abundance of some species near the park limits, in a way similar to that noted in other regions [70]. Second, exotic grasses [B. decumbens and M. minutiflora] are abundant along margins of roads and non-managed protected grasslands in the Parque Nacional das Emas, as occurs in other Cerrado reserves [34, 37, 38]. Both species replace native grasses and provide habitat and food resources for species such as V. jacarina [64], especially close to the park border.

The use of substantial amounts of agrochemicals in the cultivation of soybean and other plantations Embrapa [71] also might contribute to negative responses of some species to nearby agricultural land. As noted by smell on more than five occasions, chemicals can be brought by wind up to 1-2 km inside the park during their application on exotic fields. This income of chemicals might kill invertebrates, and contaminate plants and vertebrates in grasslands. The invasion by dust from bare ground, containing or not chemicals, between periods of cultivation also can influence communities. The potential contribution of firebreaks on edge effects also should not be neglected. Besides accidental burning of nearby grasslands during their maintenance operations, firebreaks are marked by the abundance of exotic grasses and plant resources that attract some bird species [72].

Therefore, several factors might be involved in these edge effects at Parque Nacional das Emas. This accumulation of effects is strongly tied to the establishment of large-scale agriculture focusing on soybean production in this Cerrado region. The paved road of the boundary was created with the main purpose of transporting the production of soybean grains. Adjacent unpaved roads were created for the traffic of vehicles within farms and the park, and firebreaks were settled to protect the park from fire originated in adjacent properties. All these land modifications result of the establishment of such agricultural system [agribusiness] that leads to these strong edge effects at Parque Nacional das Emas.