Skip to main content
Advertisement
Browse Subject Areas
?

Click through the PLOS taxonomy to find articles in your field.

For more information about PLOS Subject Areas, click here.

  • Loading metrics

Edge Effects on Foliar Stable Isotope Values in a Madagascan Tropical Dry Forest

Abstract

Edge effects represent an inevitable and important consequence of habitat loss and fragmentation. These effects include changes in microclimate, solar radiation, or temperature. Such abiotic effects can, in turn, impact biotic factors. They can have a substantial impact on species, communities, and ecosystems. Here we examine clinal variations in stable carbon and nitrogen isotope values for trees along an edge-interior gradient in the dry deciduous forest at Ankarafantsika National Park. We predicted that soil respiration and differences in solar irradiance would result in stratified δ13C values where leaves collected close to the forest floor would have lower δ13C values than those growing higher up in the canopy. We also anticipated that plants growing at the savannah-forest boundary would have higher δ13C and δ15N values than plants growing in the forest interior. As expected, we detected a small but significant canopy effect. Leaves growing below 2 m from the forest floor exhibit δ13C values that are, on average, 1.1‰ lower than those growing above this threshold. We did not, however, find any relationship between foliar δ13C and distance from the edge. Unpredictably, we detected a striking positive relationship between foliar δ15N values and increasing distance into the forest interior. Variability in physiology among species, anthropogenic influence, organic input, and rooting depth cannot adequately explain this trend. Instead, this unexpected relationship most likely reflects decreasing nutrient or water availability, or a shift in N-sources with increasing distance from the savannah. Unlike most forest communities, the trees at Ampijoroa are growing in nutrient-limited sands. In addition to being nutrient poor, these well-drained soils likely decrease the amount of soil water available to forest vegetation. Continued research on plant responses to edge effects will improve our understanding of the conservation biology of forest ecosystems in Madagascar.

Introduction

A fundamental issue in ecology and conservation biology is determining how edge effects influence plant communities [1]. Numerous studies have established that the structure and function of plant communities at forest edges differ from those in the relatively more homogenous interior forest and in the surrounding matrix [2], [3], [4]. This ecological relationship has been defined as edge effects or edge influence [5], [6]. Researchers have focused considerable effort on studying edge effects because forests are becoming increasingly fragmented in many parts of the world [7], which results in edge habitats comprising a greater proportion of each fragment [8], [9]. Creation of edge habitats results in feedback loops for forest regeneration and fragment viability [10]. Consequently, landscape changes in forest cover and fragmentation can create complex patterns in the distance and intensity of edge penetration for many abiotic and biotic variables, such as ultraviolet light, organismal abundance, and life history patterns [11], [12], [13]. For example, Gehlhausen and colleagues [14] documented edge-related gradients in relative humidity, light, and soil moisture, each of which varied depending on edge aspect and matrix composition. Although much is known about edge effects in temperate forests [15], [16], there are fewer comparable data for tropical forests and even less is known about tropical dry forests.

Edge effects are particularly relevant to the forested regions of Madagascar, where anthropogenic disturbance over the last 50 years has resulted in the loss of half of the original forest cover [17]. Forest conversion is due primarily to slash and burn agriculture and anthropogenic fires used to create cattle pasture [18], [19]. The remaining forest is highly fragmented and prone to extreme edge effects [17]. Of the few in situ studies of edge effects conducted to date on the island, most have documented how plant and animal abundance covary with edge proximity in forest fragments [20], [21], [22], [23]. For example, trees in edge habitats are significantly shorter, have smaller diameters, and lower stem frequencies than trees in interior habitats in eastern humid forests in Madagascar, and these dendrometric variations have been linked to edge avoidance in some lemur species, such as fat-tailed dwarf lemurs [24]. However, there are few data on the biological processes resulting from edge creation, particularly in the island' rare tropical dry forests.

Malagasy dry forests are particularly prone to edge effects due to natural and anthropogenic fires along forest edges [17]. This has resulted in multiple fragments of varying sizes throughout the landscape and along the perimeter of continuous forest tracts. Most significantly, frequent anthropogenic fires suppress forest regeneration, leading to the encroachment of savannah into the fragmented forest and increasing the amount of forest edge that abuts savannah [18], [25]. Moreover, existing data suggest that biotic and abiotic variables are affected up to 400 and 625 meters into the dry forest interior, respectively [26]. These edge penetration distances are much higher than those measured at most other tropical forest sites, where the distance of edge effects rarely exceed 100–125 m [6]. Despite invaluable information on the institutional characteristics influencing patterns of loss and regeneration in Madagascan dry forests [27], there are few data on how edge creation leads to changes in plant biology, such as clinal variations in isotope fractionation for trees along edge-interior gradients.

Although most researchers measure abiotic and biotic conditions as separate processes, it is increasingly clear that these processes are highly related. Stable isotope values in leaves integrate both abiotic and biotic factors [28], [29]. They may, therefore, quantify edge effects better than conventionally measured variables (e.g., basal area, species richness, wind speed, light levels, air temperature). Previous isotopic applications to forest edge research have been limited to stable carbon isotopes in tropical moist forests [30]. Edge dynamics in dry forests and the influence of forest edges on foliar nitrogen isotope values have yet to be explored. Accordingly, we conducted the first survey of stable carbon and nitrogen isotope values in tree leaves growing along a savannah-dry forest edge. Leaves were collected at Ampijoroa Forest Station, Jardin Botanique A (JBA) Ankarafantsika National Park, northwestern Madagascar (Figs. 1, 2). We tested three hypotheses:

thumbnail
Figure 1. Map of study area at Ampijoroa.

Map includes location of Madagascar (A), Ampijoroa Forest Station in Ankarafantsika National Park (B), and Transects and sample plots in our study area along the western border of Ampijoroa (C).

https://doi.org/10.1371/journal.pone.0044538.g001

thumbnail
Figure 2. Photographs of the forest interior and examples of soils at Ampijoroa.

Images include the dry deciduous forest at Ampijoroa Forest Station, Jardin Botanique A (A), White, course soil (B), speckled, course soil (C), and red, fine soil (D). Scale bar = 1 cm.

https://doi.org/10.1371/journal.pone.0044538.g002

(H1) Vertical stratification in light levels, temperature, and the isotopic composition of CO2 available to plants will result in lower carbon isotope values in understory foliage growing close to the forest floor.

(H2) Increasing relative humidity, and decreasing temperature and light levels will result in lower carbon isotope values with increasing distance from the forest edge.

(H3) Increasing soil moisture and decreasing temperature will result in lower nitrogen isotope values with increasing distance from the forest edge.

Background on isotopes

The natural abundance of stable isotope values in biological materials, such as leaves, is the product of partitioning of isotopes of the same element with differing masses (e.g., 15N versus 14N; 13C versus 12C) during chemical and physical reactions. Because the absolute ratio of heavy isotope (15N, 13C) to light isotope (14N, 12C) tends to be very small, isotopic abundance is generally referred to using a standardized “δ” notation, where δ = ((Rsample/Rstandard)−1)*1000, and R = 13C/12C or 15N/14N. Values are reported as parts per thousand (per-mil; ‰). The standards for carbon and nitrogen are Pee Dee Belemnite and Air, respectively.

Foliar δ15N values are positively correlated with mean annual temperature and negatively correlated with annual rainfall, relative humidity, soil moisture, and soil nutrient content [28], [29], [31], [32], [33], [34], [35]. Foliar δ13C values are influenced by temperature, rainfall, relative humidity, sun exposure, soil moisture and water use efficiency [29], [36], [37], [38]. Additionally, understory leaves can exhibit lower δ13C values than those in the canopy. This tendency, called the canopy effect, is caused by low light intensities and uptake of soil-respired CO2 by understory plants growing close to the forest floor [39], [40].

Based on these trends, we anticipated finding vertical stratification in foliar δ13C values at Ampijoroa. We also expected that both δ13C and δ15N values of understory leaves would be highest close to forest edges, where trees and soils are exposed to more sun, drier air and higher temperatures. We anticipated that both carbon and nitrogen isotope values would decrease within 400–600 m of the savannah border, and that they would remain low and relatively uniform with increasing distance into the forest interior [26]. We predicted that nitrogen isotope values for canopy leaves would show a similar trend. However, because canopy leaves are exposed to abiotic conditions resembling those in the savannah, we expected that δ13C values for canopy leaves would be less affected by edges.

Results

Isotope data, distance from edge, species, tree height, sample height, diameter at breast height for each leaf sample, soil color, and average soil H2O% for each plot are presented in Table S1.

Foliar δ13C and Sample Height (H1)

Foliar δ13C values were significantly higher in “high" leaves collected above 2 m from the ground compared to “low” leaves sampled below this height threshold (Fig. 3). On average, δ13C values for leaves collected below 2 m (−29.2‰) were 1.1‰ lower than leaves collected above 2 m (−30.3‰). This pattern was driven largely by height threshold variations in Chrysobalanaceae and Loganiaceae; none of the other plant families exhibited a significant relationship between height and foliar δ13C values (Fig. 4).

thumbnail
Figure 3. Comparison of foliar δ13C levels below and above 2 meters from the ground.

“N” values are presented in parentheses. Foliar δ13C levels were significantly higher for leaves collected above 2 meters from the ground (F1,223 = 36.56, p<0.0001). Removal of outliers did not alter the statistical significance of the test (F1,217 = 37.93, p<0.0001).

https://doi.org/10.1371/journal.pone.0044538.g003

thumbnail
Figure 4. Comparison of foliar δ13C values above and below 2 m for by plant family.

“N” values are presented in parentheses. Leaves from Chrysobalanaceae (F1,32 = 14.33, p = 0.001) and Loganiaceae (F1,63 = 16.09, p<0.0001) collected above 2 m had significantly higher δ13C values than leaves collected below 2 m. This pattern is marginally significant for Euphorbiaceae (F1,34 = 3.88, p = 0.057). There was no significant height effect on foliar δ13C values for Meliaceae (F1,47 = 1.72, p = 0.196), Myrtaceae (F1,12 = 1.34, p = 0.269), or Oleaceae (F1,25 = 3.08, p = 0.091). Equality of variances was assessed using Levene's tests. None of the comparisons displayed significant unequal variances (all tests>0.05).

https://doi.org/10.1371/journal.pone.0044538.g004

Edge Proximity and Foliar δ13C and δ15N (H2 & H3)

Foliar δ13C and δ15N values differed in their response to edge effects. Foliar δ13C did not covary with distance from the forest edge. Conversely, δ15N was positively correlated with edge proximity (explaining 46% of the variation in foliar δ15N; Fig. 5, Table 1). Of the six plant families from which we collected leaves (Fig. 6), only samples from Loganiaceae and Meliaceae displayed significant correlations between edge distance and δ15N.

thumbnail
Figure 5. Spatial variability in foliar δ13C and δ15N values with increasing distance from the forest edge.

Carbon isotope values are solid squares and nitrogen isotope values are open circles. The solid regression line is for δ13C (F1,229 = 0.32, r2 = 0.001, p = 0.568) and the dashed line for δ15N (F1,229 = 194.45, r2 = 0.459, p = 0.0001).

https://doi.org/10.1371/journal.pone.0044538.g005

thumbnail
Figure 6. Foliar δ15N values and distance from the forest edge by plant family.

There is a significant positive relationship between δ15N values and distance from the forest edge for Loganiaceae (F1,65 = 62.51, r2 = 0.489, p<0.0001) and Meliaceae (F1,50 = 74.20, r2 = 0.597, p<0.0001). Foliar δ15N does not covary with edge proximity in Chrysobalanaceae (F1,33 = 0.11, r2 = 0.003, p = 0.739), Euphorbiaceae (F1,34 = 1.16, r2 = 0.033, p = 0.288), Myrtaceae (F1,12 = 3.83, r2 = 0.245, p = 0.072), or Oleaceae (F1,25 = 0.34, r2 = 0.013, p = 0.564).

https://doi.org/10.1371/journal.pone.0044538.g006

thumbnail
Table 1. Predictors of variability in foliar δ13C and δ15N values in Generalized Linear Models.

https://doi.org/10.1371/journal.pone.0044538.t001

Predictors of Foliar δ13C and δ15N Variability (H2 & H3)

Three variables (plant family, soil H2O%, and 2 m height threshold) were the main predictors of variability in foliar δ13C (Table 1). Edge proximity was not a significant determinant of foliar δ13C values. The full model explained 44.0% of the total variation in the response variable (F33,191 = 14.55, p<0.0001). For δ15N, distance from the forest edge, plant family, and soil color were the main drivers of spatial variability (Table 1). Full model performance was relatively stronger than δ13C, explaining 72.9% of the variation in δ15N (F33,191 = 47.28, p<0.0001). Only one predictor variable, plant family, was a significant determinant of spatial variation in both foliar δ13C and δ15N values.

All multivariate predictors of δ13C were statistically significant (Table 2). However, goodness of fit statistics indicate that the best fit model is predicted by plant family and 2 m height threshold, explaining 36.8% of the spatial variation in δ13C values. Although the AIC values for the best-fit 2-way model (plant family and height threshold) and 3-way model (plant family, 2 m height threshold, and edge distance) are very similar (65.77 vs. 67.68), the relevant AIC value for the two models, which estimates relative support for a model, indicates that the 3-way model is only 0.384 times as probable as the 2-way model to minimize the information loss (Table 2). For δ15N, the clear best-fit model is given by three predictor values (plant family, soil color, and edge distance), which explain 60.8% of the spatial variation in δ15N (Table 2).

thumbnail
Table 2. Selected multivariate predictors of variability in foliar δ13C and δ15N values in Generalized Linear Models.

https://doi.org/10.1371/journal.pone.0044538.t002

Discussion

We predicted vertical stratification in foliar δ13C values and decreasing δ13C and δ15N values with increasing distance from the forest-savannah border. As expected, we observed a canopy effect in δ13C values. However, we did not detect the predicted negative relationship between distance from edge and foliar carbon or nitrogen isotope values. Instead, we found no relationship between δ13C values and distance from the edge. Quite unexpectedly, we detected a strong positive correlation between distance from edge and δ15N values. We address each of these points below.

H1: Vertical stratification in light levels, temperature, and the isotopic composition of CO2 available to plants will result in lower carbon isotope values in understory foliage

On average, δ13C values were 1.1‰ lower for leaves collected below 2 meters from the ground. Although small, this difference is significant (Fig. 3). This 1‰ difference is smaller than the canopy effect measured in other dry or deciduous forests [40], [41], a probable reflection of the relatively open nature and short stature of the forest at Ampijoroa (Fig. 2). Additionally, respiration of 13C-depleted CO2 by soil microbes may be minor at this arid locality.

When broken down by family, the positive relationship between foliar δ13C and sample height is only significant for Chrysobalanaceae and Loganiaceae (Fig. 4). However, Euphorbiaceae, Meliaceae, Myrtaceae and Oloeaceae all exhibited a trend towards higher foliar δ13C values above 2 meters from the ground (Fig. 4). The varying significance of these trends may simply reflect small sample sizes among the other four families, particularly for samples below 2 m from the ground. Locating leaves close to the ground was difficult for all families except Loganiaceae. Noronhia boiensis tends to be an understory shrub. The other five species sampled are all emergent canopy species.

Prediction 2: Increasing relative humidity, and decreasing temperature and light levels will result in lower carbon isotope values with increasing distance from the forest edge

We detected no relationship between foliar δ13C values and distance from edge, even after accounting for sample height (Tables 1, 2). This is contra the findings of Kapos and colleagues [30], who found a distinctive decrease in the δ13C values of understory shrubs with increasing distance from the forest edge. Unlike the Amazonian tropical moist forest where these researchers conducted their study, the forest at Ampijoroa is quite open and the canopy near the edge is relatively discontinuous (Fig. 2). Whereas previous work has detected decreasing wind, light and temperature with increasing distance from the edge [26], these variables may not be consistent enough or large enough to impact foliar δ13C values.

Prediction 3: Increasing soil moisture and decreasing temperature will result in lower nitrogen isotope values with increasing distance from the forest edge

Instead of detecting a negative relationship between δ15N and distance from the forest edge, we found a striking positive relationship. We explore five possible explanations for this unexpected trend. First, plant family is a significant predictor of foliar δ15N values (Table 2). When broken down by family, only two taxa exhibit a significantly positive relationship, Loganiaceae and Meliaceae (Fig. 6). Despite our efforts to collect leaves from all six families in each of our plots, these two are the only families that are well represented along the entire edge-interior distance. The lack of a relationship between δ15N and distance from the forest edge for Chrysobalanaceae, Euphorbiaceae, Myrtaceae and Oleaceae may, therefore, reflect their spatial distribution. None of these families are adequately represented both close to and far from the forest-savannah border.

It is conceivable that interspecific physiological differences are driving the detected relationship between foliar δ15N and distance from the edge. Species may differ in their ability to exploit nitrogen sources. Differences in rooting depth, the extent of mycorrhizal associations, and transport, transformation, and reallocation of nutrients within the plant can all influence foliar δ15N values [42], [43], [44]. However, such physiological differences should only have small influences on foliar δ15N [44], especially considering our attempts to only sample taxa with vesicular-arbuscular mycorrhizas and no N-fixing microbes. It is, therefore, very unlikely that physiological differences among families are responsible for the observed 4.5‰ increase in mean δ15N values.

We explore four additional explanations for the unexpected positive relationship between foliar δ15N values and distance from the forest edge: (1) anthropogenic activity has disturbed the forest edge, (2) there is less organic input in the forest interior, (3) trees have deeper roots in the forest interior, and (4) soil moisture or nutrient content is lower in the forest interior.

Anthropogenic disturbance.

The forest at Ampijoroa has a history of human influence [18]. Despite its protected status, there is ample evidence of human disturbance within the forest. While conducting our fieldwork we encountered cows as well as numerous meter-deep pits from locals digging up Dioscorea tubers. Both of these activities disturb the soil and may influence nutrient cycling within the forest [45], but the isotopic ramifications of these impacts are unclear. Based on previous research, we would expect that such disturbances would increase the rate of gaseous 14N loss via nitrate oxidation or ammonia volatilization [28], [34], [44]. This should, in turn increase the δ15N values of soils and the plants that grow on them. The frequency of cattle and tuber foraging was highest close to the savannah border (Travis Steffens, personal communication), yet this is where we recorded the lowest δ15N values. It is possible that this level of human impacts is too low to have an isotopic effect. An alternative possibility is that the entire forest at Ampijoroa can be considered relatively disturbed [46]. Although the forest is currently protected, it is subjected to a steady stream of tourists and scientists. Cattle are also present throughout the reserve. The isotopic effects of human activities near the forest edge may, therefore, be indiscernible.

The forest has less organic input than the forest edge.

Reduced organic input from leaf litter and decomposition can result in increased soil nitrogen isotope values [35]. Variation in organic inputs might, therefore, be able to explain the observed pattern in δ15N. We find this explanation unsatisfactory. Although we have not conducted leaf litter measurements, it seems unlikely that trees growing in the forest interior receive less organic input than those growing closer to the savannah. Ground cover was minimal and leaf litter was thin to absent along the entirety of both transects [47]. Termites living in the savannah could increase plant available-N in the soils. However, their influence should be limited to within 40 m from their mound centers [48]. It is, therefore, highly unlikely that termite-derived nitrogen would be available to forest trees or responsible for the difference in δ15N values we have detected between edge and interior leaves.

Trees in the forest interior have deeper roots than those growing near the forest edge.

If trees growing in the forest interior have deeper roots than those growing closer to the savannah border, they may be accessing a 15N-enriched source [28], [42], [49]. Previous research in Ampijoroa has found that trees are shorter and smaller (as measured by diameter at breast height) closer to the savannah [26]. Presumably, smaller plants would have shallower root systems. However, neither tree height nor sample height affects foliar δ15N values (Table 2). Thus, while it is certainly possible that interior trees have deep roots, the likelihood that this is driving the observed positive trend between distance from the edge and foliar δ15N values is unlikely.

Soil moisture or nutrient content in lower in the forest interior.

In addition to distance from the edge, δ15N covaried with soil color (Table 2). The capacity for soil and plant δ15N to vary over small distances is not unexpected. “Microrelief” across seemingly homogenous landscapes can lead to variability both in soil moisture and nutrient retention [31], [50], [51]. The majority of Ampijoroa's precipitation falls during a brief rainy season, which typically occurs from November to April [52]. Red soils near the forest edge appear to retain moisture longer than the more porous white sandy “soils” in the forest interior (Fig. 7). Although all soil samples collected contained very little moisture (<4% by weight), even slight variations in soil moisture could cause trees in the forest interior to experience a more pronounced dry season than those growing closer to the edge. A recent study of island-wide isotopic variability suggests that soil moisture does, indeed, influence plant δ15N values [53]. For example, trees growing along the shores of Lake Ravelobe, (<1 km from JBA), exhibit δ15N values that are 7‰ lower than those growing in JBA.

thumbnail
Figure 7. Comparison of water content among soil color categories.

https://doi.org/10.1371/journal.pone.0044538.g007

Very little research has been conducted on the soils at Ampijoroa. Soil nutrient content has only been measured in three samples of white sandy soils [47]. Nothing is known about the nutrient content of the red soils. Nevertheless, if the soils are as nutrient-poor as suggested by Lourenço and Goodman [47], then the vegetation at Ampijoroa is operating on a very tight nutrient budget. It is, therefore, reasonable to believe that any organic matter that makes its way into the soil is quickly recycled back into the vegetation. During nutrient uptake, plants fractionate against 15N. When nutrients are limited, however, plants cannot discriminate as much against 15N and they incorporate more of this heavy isotope into their tissues [32]. Additionally, the absence of fine-grained clay minerals in the white sands may make it harder for them to retain labile ions such as nitrate and ammonium [32]. Episodic wetting of the forest soils could lead to the preferential loss of labile 14NO3 and 14NH4+ via volatilization or leaching [44]. If nutrients are more easily lost from the porous white sands, then interior trees growing on these soils may be recycling their nutrients more completely than those on red soils, a process that would result in increasingly elevated δ15N values for plants growing on white sands over time.

Although we were not able to measure nutrient content in the red soils, both their color and finer texture (Fig. 2) suggest that they have more nutrients than the nearly pure quartz white sands. Additionally, in the absence of labile nutrients, trees growing on white sands in the forest interior may be directly accessing nitrogen from organic matter. Reliance on organic nitrogen, particularly free amino acids, is widespread among plants from many ecosystems [54]. Organic N source may differ isotopically from other sources [34]. This is an intriguing possibility that we will investigate in the future.

The positive relationship we have detected between foliar δ15N and soil color is opposite to that observed for tropical ‘white sand’ moist forests growing on sandy soils in other parts of the world [49], [50], [55]. Previously, researchers have reported a negative relationship between percent sand and δ15N in soils. This difference may reflect the extremely dry nature of Ampijoroa. In addition to growing on nutrient-poor sands, this forest endures very water limited conditions. Mean H2O% is only 0.9±1.0%. This idea agrees with previous research in mainland Africa and Australia, where arid, sandy habitats exhibit high δ15N values, particularly during the dry season [32], [35], [56], [57].

Antiquity of the forest-savannah mosaic at Ampijoroa

It is unclear how stable the savannah-forest landscape is in northwestern Madagascar. This landscape may be a stable mosaic ecosystem of relative antiquity [58], [59], [60]. Alternatively, the patchy nature of forest and grasses in and around Ankarafantsika may reflect recently established vegetation responding to frequent anthropogenic disturbance. Arial photographs and vegetation maps indicate that the savannah bordering the forest at Ampijoroa has been present for at least 50 years. However, whether or not it persisted into the deeper past is unknown. Future examination of soil isotope values and radiocarbon ages both in the forest and the savannah will help to clarify the antiquity of the savannahs [59].

The forest at Ampijoroa is a fragile and delicately balanced system. Plants are growing on an extremely limited water and nutrient budget. Humans are clearly affecting the forest here, but the long-term ramifications of their actions are unknown. The savannahs surrounding Ampijoroa have been described as fire-maintained ‘biological deserts’ void of endemic plant species [46], but this may not be entirely accurate. Although frequently burned savannah is dominated by introduced grasses [18], less disturbed savannah contains endemic species [58]. Woody vegetation is generally found on white soils and savannah is generally found on red soils [18]. However, dry forest does grow on red sandy soils (personal observation). Fast-growing savannah grasses likely outcompete trees on these soils following disturbance, such as a brush fire [18]. Conversely, deep-rooted forest vegetation may have a competitive edge over grasses on the white sandy soils. Disentangling anthropogenic and naturally occurring vegetation will have profound implications for conservation. The striking isotopic patterns that we have detected will be useful for understanding foraging patterns in threatened fauna. Continued research on plant responses to edge effects will improve our understanding of the conservation biology of forest ecosystems in Madagascar.

Materials and Methods

Site Description

We conducted our fieldwork at Jardin Botanique A (JBA), Ampijoroa Forest Station, Ankarafantsika National Park, northwest Madagascar (Fig. 1). The vegetation at JBA is characterized as semi deciduous dry forest [47]. Canopy height ranges from 5–15 meters and understory and herbaceous vegetation is poorly developed (Fig. 2; [26], [47]). The study area is situated on a sandstone plateau ca. 250 m above see level [46], [61]. The topography is flat, the soil is sandy, acidic and nutrient-poor, and leaf litter is relatively thin and poorly decomposed [47]. The climate at Ankarafantsika is strongly seasonal. Rainfall during most years ranges from 1100–1600 mm per year [52]. The vast majority of this precipitation falls during January and February. Between May and October there is negligible rainfall. Average daily temperatures range from ca. 37°C in the rainy season to 16°C in the dry season [52].

Sample Collection and Preparation

We carried out our project along the western edge of JBA, where a sharp ecotone divides the forest from the surrounding savannah (Fig. 1). We collected leaf samples along two pre-existing transects which run perpendicular to the savannah edge into the forest interior. Along each transect, we set up 5 m×20 m plots (length perpendicular to transect) at 0, 5, 100, 150, 400, 450, 500, 900, 950, and 1000 m from the savannah-forest boundary (Fig. 1). We recorded geographic coordinates at the center of each plot using a handheld GPS. Using a soil corer, we collected three shallow soil samples (<25 cm) in each plot. We divided soil samples into “top” and “bottom” halves, and recorded their wet weights immediately after collection.

Within each plot, we attempted to collect foliage samples from six tree species representing six plant families (Table 3). These species were chosen because they are food sources for resident Propithecus coquereli groups, and because they are relatively common throughout the study site. All of these families rely on vesicular-arbuscular mycorrhizal (VAM) symbiotic associations [62], [63]. None have symbiotic nitrogen fixing bacteria. We collected 5–10 mature leaves from up to four individuals of each species in each plot. We measured diameter at breast height (DBH) and estimated the height of each individual. Leaves were collected using pruning shears and handheld loppers on a 5-meter extendable pole. In order to control for potential isotopic stratification within the forest, we collected foliage at variable distances above the forest floor (ranging from 0.5–7 m). Whenever possible, we collected both a “low” sample (<2 m) and a “high” sample (>2 m) from the same tree. All specimens were collected between 9 am and 1 pm, July 3 through July 10, 2010.

thumbnail
Table 3. Tree species and sample numbers included in this study.

https://doi.org/10.1371/journal.pone.0044538.t003

After collection, leaf and soil samples were thoroughly air-dried. Soil samples were photographed and their dry mass was recorded. Dried leaves were homogenized using a mortar and pestle. Approximately 5 mg of each homogenized leaf sample were added to tin capsules and combusted on a Finnigan ThermoElectron Delta+XP continuous flow system (Bremen, Germany) connected to a Carlo Erba Elemental Analyzer (Milan, Italy) at the University of California, Santa Cruz Stable Isotope Laboratory. Samples were analyzed for δ13C and δ15N values, and weight percent carbon, nitrogen, and C/N ratios (here referred to as %C and %N, and C/N). Carbon and nitrogen international standards were Pee Dee Belemnite and Air, respectively. The analytical precision (±1 SD) for carbon was 0.2‰ and 0.08‰ based on 24 IAEA Acetanilide and 23 UCSC oak leaf standard replicates, respectively. Analytical precision for nitrogen was 0.2‰ and 0.1‰ based on Acetanilide and oak replicates. C/N was more variable. Whereas precision based on 24 Acetanilide replicates was 0.1%, precision based on 23 oak replicates was 0.5%. The average difference between 17 leaf samples run in duplicate was 0.1‰ for both δ13C and δ15N values, and 0.9% for C/N. Raw data are provided in Table S1.

Analytical Methods

Explanatory Variables.

We explored how eight explanatory variables covaried with δ13C and δ15N: (1) distance to edge, (2) tree height, (3) DBH, (4) plant family, (5) soil color, (6) soil grain size, (7) soil moisture content (H2O%), and (8) sample height (using a 2 m height threshold for “high” samples above 2 m versus “low” samples below 2 m). Initial comparisons of δ13C and distance to edge indicated no significant difference in slopes (t = 1.78, p = 0.09) or y-intercepts (t = 1.86, p = 0.08) between the two transects. There were also no significant inter-transect differences in slopes (t = 0.25, p = 0.80) or y-intercepts (t = 0.72, p = 0.47) for δ15N and distance to edge. Therefore, we pooled isotopic data for the two transects. Soil color and grain size were assessed in the field using a relative visual scale. We used two groups for grain size (fine or course) and three groups for color (red, white, or speckled) (Fig. 2). Soil H2O% was determined by dividing the dry weight by the wet weight of each sample. There were significant differences in soil H2O% among color categories (Kruskal-Wallis X2 = 38.03, df = 2, p<0.0001). White sands had the lowest H2O% and red sands had the highest (Fig. 7). We arcsine square root transformed proportional data on H2O% and log10 transformed tree height and tree DBH. Significance for all tests was set at α = 0.05.

Data Analysis.

We used an analysis of variance (ANOVA) to test Hypothesis 1 (vertical stratification of biotic and abiotic variables will result in low δ13C levels close to the forest floor). Although Levene's test indicated that the population variances were equal for the model (W = 0.017; df = 1,223; p = 0.897), we ran a second test after removing any extreme δ13C outliers. We further explored the effects of collection height on δ13C levels for each plant family.

We used ordinary least squares regression estimates to test Hypotheses 2 and 3 (that edge proximity is negatively correlated with foliar δ13C and δ15N values, respectively). One sample Kolmogorov-Smirnov tests indicated that both dependent variables were normally distributed (δ13C: K–S = 0.041, df = 231, p = 0.200; δ15N: K–S = 0.055, df = 231, p = 0.093). The homoscedasticity of the data further precluded data transformation, for analyses. We also analyzed how nitrogen isotope values varied by distance from the forest edge for each of the six plant families.

We used a model building process with General Linear Models (GLM) to explore the effects of edge proximity and the seven other explanatory variables on spatial variations in δ13C and δ15N. Multivariate models were based on ecological knowledge of the edge system at Ampijoroa and Akaike Information Criterion (AIC) values for model selection. Soil color, soil grain size, and plant family were recoded as dummy variables. For each dependent variable in full models, we used four factors (soil color, soil grain size, plant family, and 2 m height threshold) and four fixed covariates (distance to edge, tree height, tree DBH, and H2O%). The fixed covariates were checked for cross-correlations before being used in the GLMs. Of the six cross-correlations, the strongest correlate was for log10 tree height and log10 tree DBH (r = 0.777, n = 233, p<0.0001). Although the other cross-correlations were statistically significant, none of the Pearson correlation coefficients were greater than 0.08.

Supporting Information

Table S1.

Plant data. Isotope data, distance from edge, species, tree height, sample height, diameter at breast height for each leaf sample, and soil color, average soil H2O% for each plot.

https://doi.org/10.1371/journal.pone.0044538.s001

(XLS)

Acknowledgments

Research was conducted under Repoblikani'i Madagasikara Ministère de l′Environment et des Forêts Madagascar permit No. 77/10/MEF/SG/DGF/DCB.SAP/SLRSE. Dried plant samples were exported from Madagascar under permit No. 167N-EV08/MG10 and imported into Canada using Canadian Food Inspection Agency permit No. P-2010-02474. We thank Kena Fox-Dobbs and an anonymous reviewer for advice; Lydia Ahamada, Emma Cancelliere, Marlena Cravens, Emily Hofstetter, Christine Ishu, Heather Kristjanson, Isabel Martins, Tommy Oliver, Andry Rajaonson and Tsihahifika for assistance with sample collection and preparation; Lalao Andriamahefarivo, Dyke Andreason, Steven Goodman, Sharon Kessler, Hong Li, Jacqueline Razaiarimanana, Abigail Ross, Travis Steffens, Uli Wortmann, Madagascar Institut pour la Conservations des Ecosystèmes Tropicaux and the Missouri Botanical Garden for technical and logistical assistance; and Durrell Wildlife Trust for providing unpublished climate data.

Author Contributions

Conceived and designed the experiments: BEC KM SML. Performed the experiments: BEC KM SML. Analyzed the data: BEC SML. Wrote the paper: BEC KM SML.

References

  1. 1. Murcia C (1995) Edge effects in fragmented forests: Implications for conservation. Trends in Ecology and Evolution 10: 58–62.
  2. 2. Saunders DA, Hobbs RJ, Margules CR (1991) Biological consequences of ecosystem fragmentation: A review. Conservation Biology 5: 18–32.
  3. 3. Fischer J, Lindenmayer DB (2007) Landscape modification and habitat fragmentation: A synthesis. Global Ecology and Biogeography 16: 265–280.
  4. 4. Turner I (1996) Species loss in fragments of tropical rain forest: A review of the evidence. Journal of Applied Ecology 33: 200–209.
  5. 5. Leopold A (1933) Game management. Madison: The University of Wisconsin.
  6. 6. Harper KA, MacDonald SE, Burton PJ, Chen J, Brosofske KD, et al. (2005) Edge influence on forest structure and composition in fragmented landscapes. Conservation Biology 19: 1–15.
  7. 7. Ritters K, Wickham J, O′Neill R, Jones B, Smith E (2000) Global-scale patterns of forest fragmentation. Conservation Ecology 4: 3.
  8. 8. Fahrig L (2003) Effects of habitat fragmentation on biodiversity. Annual Review of Ecology, Evolution, and Systematics 34: 487–515.
  9. 9. Gascon C, Williamson GB, da Fonseca GAB (2000) Receding forest edges and vanishing reserves. Science 288: 1356–1358.
  10. 10. Laurance MF (1991) Edge effects in tropical forest fragments: Applications of a model for the design of nature reserves. Biological Conservation 57: 205–219.
  11. 11. Ewers RM, Didham RK (2006) Confounding factors in the detection of species responses to habitat fragmentation. Biological Reviews 81: 117–142.
  12. 12. Dignan P, Bren L (2003) Modelling light penetration edge effects for stream buffer design in mountain ash forest in southeastern Australia. Forest Ecology and Management 179: 95–106.
  13. 13. Li Q, Chen J, Song B, LaCroix JJ, Bresee MK, et al. (2007) Areas influenced by multiple edges and their implications in fragmented landscapes. Forest Ecology and Management 242: 99–107.
  14. 14. Gehlhausen SM, Schwartz MS, Augspurger CK (2000) Vegetation and microclimatic edge effects in two mixed-mesophytic forest fragments. Plant Ecology 147: 21–35.
  15. 15. Wilcove D, Mclellan C, Dobson A (1997) Habitat fragmentation in the temperate zone. In: Goudie A, editor. The human impact reader: Readings and case studies. Malden, MA: Wiley-Blackwell. pp 237–256.
  16. 16. Hartley MJ, Hunter ML (1998) A meta-analysis of forest cover, edge effects, and artificial nest predation rates. Conservation Biology 12: 465–469.
  17. 17. Harper GJ, Steininger MK, Tucker CJ, Juhn D, Hawkins F (2007) Fifty years of deforestation and forest fragmentation in Madagascar. Environmental Conservation 34: 325–333.
  18. 18. Bloesch U (1999) Fire as a tool in the management of savanna/dry forest reserve in Madagascar. Applied Vegetation Science 2: 117–124.
  19. 19. Gade DW (1996) Deforestation and its effects in Highland Madagascar. Mountain Research and Development 16: 101–116.
  20. 20. Ramanamanjato JB, Ganzhorn JU (2001) Effects of forest fragmentation, introduced Rattus rattus and the role of exotic tree plantations and secondary vegetation for the conservation of an endemic rodent and a small lemur in littoral forests of southeastern Madagascar. Animal Conservation 4: 175–183.
  21. 21. Lehman SM, Rajoanson A, Day S (2006) Edge effects and their influence on lemur distribution and density in southeast Madagascar. American Journal of Physical Anthropology 129: 232–241.
  22. 22. Lehtinen RM, Ramanamanjato J-B, Raveloarison JG (2003) Edge effects and extinction proneness in a herpetofauna from Madagascar. Biodiversity and Conservation 12: 1357–1370.
  23. 23. Watson JEM, Whittaker RJ, Dawson TP (2004) Habitat structure and proximity to forest edge affect the abundance and distribution of forest-dependent birds in tropical coastal forests of southeastern Madagascar. Biological Conservation 120: 311–327.
  24. 24. Lehman SM, Rajoanson A, Day S (2006) Edge effects on the density of Cheirogaleus major. International Journal of Primatology 27: 1569–1588.
  25. 25. Pons P, Rakotobearison G, Wendenburg C (2003) Immediate effects of a fire on birds and vegetation at Ankarafantsika Strict Nature Reserve, NW Madagascar. Ostrich 74: 146–148.
  26. 26. McGoogan K (2011) Edge effects on the behaviour and ecology of Propithecus coquereli in Northwest Madagascar. Toronto: University of Toronto. 210 p.
  27. 27. Elmqvist T, Pyykönen M, Tengö M, Rakotondrasoa F, Rabakonandrianina E, et al. (2007) Patterns of loss and regeneration of tropical dry forest in Madagascar: the social institutional context. PLoS ONE 2: e402.
  28. 28. Evans DR (2007) Soil nitrogen isotope composition. In: Lajtha K, Michener, B., editor. Stable Isotopes in Ecology and Environmental Science. Boston, Massachusetts: Blackwell Scientific Publication.
  29. 29. Swap R, Aranibar J, Bowty P, Gilhooly IW, Macko SA (2004) Natural abundance of 13C and 15N in C3 and C4 vegetation of southern Africa: Patterns and implications. Global Change Biology 10: 350–358.
  30. 30. Kapos V, Ganade G, Matsui E, Victoria RL (1993) δ13C as an indicator of edge effects in tropical rainforest reserves. Journal of Ecology 81: 425–432.
  31. 31. Handley LL, Austin AT, Robinson D, Serimgeour CM, Raven JA, et al. (1999) The 15N natural abundance of δ15N of ecosystem samples reflects measures of water availability. Australian Journal of Plant Physiology 26: 185–199.
  32. 32. Muzuka ANN (1999) Isotopic compositions of tropical East African flora and their potential as source indicators of organic matter in coastal marine sediments. Journal of African Earth Sciences 28: 757–766.
  33. 33. Amundson R, Austin AT, Schuur AG, Yoo K, Matzek V, et al. (2003) Global patterns of the isotopic composition of soil and plant nitrogen. Global Biogeochemical Cycles 17: 1031.
  34. 34. Aranibar JN, Anderson IC, Epstein HE, Feral CJW, Swap RJ, et al. (2008) Nitrogen isotope composition of soils, C3 and C4 plants along land use gradients in southern Africa. Journal of Arid Environments 72: 326–337.
  35. 35. Aranibar JN, Otter L, Macko SA, Feral CJW, Epstein HE, et al. (2004) Nitrogen cycling in the soil-plant system along a precipitation gradient in the Kalahari sands. Global Change Biology 10: 359–373.
  36. 36. Farquhar GD, Ehleringer JR, Hubick KT (1989) Carbon isotope discrimination and photosynthesis. Annual Review of Plant Physiology and Molecular Biology 40: 503–537.
  37. 37. Kohn MJ (2010) Carbon isotope compositions of terrestrial C3 plants as indicators of (paleo)ecology and (paleo)climate. Proceedings of the National Academy of Sciences 107: 19691–19695.
  38. 38. Ehleringer JR, Field CB, Lin ZF, Kyu CY (1986) Leaf carbon isotope and mineral composition in subtropical plants along an irradiance cline. Oecologia 70: 520–526.
  39. 39. van der Merwe NJ, Medina E (1989) Photosynthesis and 13C/12C ratios in Amazonian rain forests. Geochimca et Cosmochimica Acta 53: 1091–1094.
  40. 40. Broadmeadow MSJ, Griffiths H, Maxwell C, Borland AM (1992) The carbon isotope ratio of plant organic material reflects temporal and spatial variations in CO2 within the tropical forest formations in Trinidad. Oecologia 89: 435–441.
  41. 41. Garten CT Jr, Taylor GE Jr (1992) Foliar δ13C within a temperate deciduous forest: Spatial, temporal, and species sources of variation. Oecologia 90: 1–7.
  42. 42. Evans RD, Ehleringer JR (1994) Water and nitrogen dynamics in an arid woodland. Oecologia 99: 233–242.
  43. 43. Kolb KJ, Evans RD (2002) Implications of leaf nitrogen recycling on the nitrogen isotope composition of deciduous plant tissues. New Phytologist 156: 57–64.
  44. 44. Pardo LH, Nadelhoffer KJ (2010) Using nitrogen isotope ratios to assess terrestrial ecosystems at regional and global scales. In: West JB, Bowen GJ, Dawson TE, Tu PU, editors. Isoscapes: Understanding Movement, Pattern, and Process on Earth Through Isotope Mapping. London: Springer Science+Business Media B.V. pp 221–249.
  45. 45. Williams-Linera G, Lorea F (2009) Tree species diversity driven by environmental and anthropogenic factors in tropical dry forest fragments of central Veracruz, Mexico. Biodiversity and Conservation 18: 3269–3293.
  46. 46. Alonso LA, Hannah L (2002) Introduction to the Réserve Naturelle Intégrale and Réserve Forestière d′Ankarafantsika and to the Rapid Assessment Program. In: Alonso LA, Schulenberg TS, Radilofe S, Missa O, editors. Une evaluation biologique de la Réserve Naturelle Intégrale d′Ankarafantsika, Madagascar RAP bulletin of biological assessment v 23. Washington DC: Conservation International. pp 28–32.
  47. 47. Lourenço WR, Goodman SM (2006) Notes on the postembryonic development and ecology of Grosphus hirtus Kraeplin, 1901 (Scorpiones, Buthidae) from the Parc National d′Ankarafantsika, northwest Madagascar. Zoologischer Anzeiger 244: 181–185.
  48. 48. Fox-Dobbs K, Doak DF, Brody AK, Palmer TM (2010) Termites create spatial structure and govern ecosystem function by affecting N2 fixation in an East African savanna. Ecology 91: 1296–1307.
  49. 49. Martinelli LA, Piccolo MC, Townsend AR, Vitousek PM, Cuevas E, et al. (1999) Nitrogen stable isotopic composition of leaves and soil: Tropical versus temperate forests. Biogeochemistry 46: 45–65.
  50. 50. Selles F, Karamanos RE, Kachanoski RG (1986) The spatial variability of nitrogen-15 and its relation to the variability of other soil properties. Soil Science Society of America Journal 50: 105–109.
  51. 51. Wooller MJ, Johnson BJ, Wilkie A, Fogel ML (2005) Stable isotope characteristics across narrow savanna/woodland ecotones in Wolfe Creek Meteorite Crater, Western Australia. Oecologia 145: 100–112.
  52. 52. Rendigs A, Radespiel U, Wrogemann D, Zimmermann E (2003) Relationship between microhabitat structure and distribution of mouse lemurs (Microcebus spp.) in northwestern Madagascar. International Journal of Primatology 24: 47–64.
  53. 53. Crowley BE, Thorén S, Rasoazanabary E, Vogel ER, Barrett MA, et al. (2011) Explaining geographical variation in the isotope composition of mouse lemurs (Microcebus). Journal of Biogeography 38: 2106–2121.
  54. 54. Lipson D, Näsholm T (2001) The unexpected versatility of plants: organic nitrogen use and availability in terrestrial ecosystems. Oecologia 128: 305–316.
  55. 55. Shearer G, Kohl DH, Chien S–H (1978) The nitrogen-15 abundance in a wide variety of soils. Soil Science Society of America Journal 42: 899–902.
  56. 56. Codron J, Codron D, Lee-Thorp JA, Sponheimer M, Bond WJ, et al. (2005) Taxonomic, anatomical, and spatio-temporal variations in the stable carbon and nitrogen isotopic compositions of plants from an African savanna. Journal of Archaeological Science 32: 1757–1772.
  57. 57. Schulze E–D, Williams RJ, Farquhar GD, Schulze W, Langridge J, et al. (1998) Carbon and nitrogen isotope discrimination and nitrogen nutrition of trees along a rainfall gradient in northern Australia. Australian Journal of Plant Physiology 25: 413–425.
  58. 58. Bond W, Silander JA, Ranoiavonasy J, Ratsirarson J (2008) The antiquity of Madagascar's grasslands and the rise of C4 grassy biomes. Journal of Biogeography 35: 1743–1758.
  59. 59. Wiedemeier DB, Bloesch U, Hagedorn F (2012) Stable forest-savanna mosaic in northwestern Tanzania: local-scale evidence from δ13C signatures and 14C ages of soil fractions. Journal of Biogeography 39: 247–257.
  60. 60. Koechlin J (1972) Flora and vegetation of Madagascar. In: Battistini R, Richard-Vindard G, editors. Biogeography and Ecology in Madagascar. The Hague: Dr. W. Junk B.V. pp 145–190.
  61. 61. Du Puy DJ, Moat J (1996) A refined classification of the primary vegetation of Madagascar based on the underlying geology: Using GIS to map its distribution and assess its conservation status. In: Lourenço WR, editor. Proceedings of the international symposium on the biogeography of Madagascar. Paris: Editions de l′Orstrom. pp 205–218.
  62. 62. Vincelette M, Rabenantoandro J, Randrihasipara L, Randriatafika F, Ganzhorn JU (2007) Results from ten years of restoration experiments in the southeastern littoral forests of Madagascar. In: Ganzhorn JU, Goodman, S.M, Vincellet, M., editor. Biodiversity, ecology and conservation of littoral ecosystems in southeastern Madagascar, Tolagnaro (Fort Dauphin). Washington, DC: Smithsonian Institution. pp 337–354.
  63. 63. Brundrett MC (2009) Mycorrhizal associations and other means of nutrition of vascular plants: understanding the global diversity of host plants by resolving conflicting information and developing reliable means of diagnosis. Plant Soil 320: 37–77.