Topographic position, but not slope aspect, drives the dominance of functional strategies of tropical dry forest trees

Located towards the lower part of the Balsas river basin in south-central Mexico (19° 01' 30.1'' N, 101° 58' 58.1'' W), the study area presents a landscape covered by tropical dry forest (see site description details in Méndez-Toribio et al 2016). The forest grows on low hills with altitudes ranging from 300 to 800 m a.s.l. where lithosols are the most abundant soil type (Ferrusquía-Villafranca 1993, Nava et al 1998). The climate is dry warm and highly seasonal, with an annual average temperature of 27.1 °C (30.2 °C monthly maximum during the dry season) and an annual average rainfall of 755.4 mm, of which 80% falls in the period June to September. A marked dry season occurs from October to May (data from the Tziritzícuaro meteorological station, 1951–2010). The structure and diversity of the area are similar to those documented for seasonal tropical forests of Mexico and the Neotropics (Méndez-Toribio et al 2014).

Vegetation sampling plots were located on the slopes of three mountains. To classify slope aspect, an ArcGIS 9.3 (ESRI 2009) geographic information system was used. North-facing slopes (315° − 45°) had an average steepness value of 40% (range 15%−63%) while south-facing slopes (135° − 225°) was 45% (range 14%−47%). Four 100 m² (10 × 10 m) plots, two facing north and two facing south, were established on three relative positions along the slope: the lower (380–500 m), middle (480−540 m) and upper (560–780 m) portions of each mountain slope, giving a total of 36 plots (0.36 ha; Méndez-Toribio et al 2016). Average steepness values at the lower, middle and upper portions of the slopes were 25% (14–47), 53% (40–68) and 48% (39–58), respectively. At each one of the 36 plots, rooted trees with a diameter at breast height (DBH) ≥ 2.5 cm were recorded. Plots were established in areas with no apparent evidence of disturbance, although the extraction of firewood and cattle breeding are common in the study area.

Environmental gradients generated by the interaction of the two topographic factors were characterized during two critical periods for plant growth and survival in the tropical dry forest regions: end of the dry season (March–April) and middle of the rainy season (June–July). To do so, we obtained values per plot of mean daily air temperature (°C), total incident radiation (MJ m⁻²) and potential evapotranspiration (mm) for the two mentioned periods. At each plot, temperature was registered hourly 2 m above the soil surface using one temperature sensor and data-logger (HoBo proV2, Onset, EUA). Incident solar radiation values per plot were derived by using ArcGis 9.3 ESRI (2009), based on a 20 m digital elevation model obtained for the studied landscape (see details in Méndez-Toribio et al 2016). Monthly potential evapotranspiration (a critical indicator of water demand, according to van der Maarel 2005) was then calculated based on air temperature and radiation, applying:

$$\begin{equation} ET_0 = 0.0135(KT)(R_a )(TD)(TC + 17.8) \end{equation}$$

where TD = T_max − T_min (°C); TC = average air daily temperature (°C); KT is an empirical coefficient and R_a = solar radiation × 0.408 (Hargreaves 1994, Samani 2000). Finally, values of topographic wetness index (TWI; given a hydrological model, a steady state indicator of topographic induced soil water flux and content, Beven and Kirkby 1979), for each plot were taken from Méndez-Toribio et al (2016). Thus for each plot we obtained mean daily temperatures (T), integrated incident radiation (RAD) and potential evapotranspiration (ET₀) for the two mentioned periods, and one single value of TWI, as an indicator of the ranking of each plot in terms of soil water content after rain events. In the studied landscape it is common to observe portions of the surface with rock exposed or covered by a thin layer of soil, particularly on the middle slopes and near the top hills. For this reason, we characterized variation in soil depth by measuring the maximum depth of mineral soil (MSD; dm) and % of coarse fragments on superficial horizon by excavating 1 m² trenches, following Siebe et al (2006). Additionally, maximum main rooting depth was taken at each trench (MRD, Siebe et al 2006). Due to restrictions on permission to do excavations in the area, MSD, CF and MRD were measured only for nine plots; one low, one middle and one upper hill positions, located on the north-facing slope of the three studied hills.

A total of 63 species of trees with DBH ≥ 2.5 cm were recorded at the 36 sample plots (Méndez-Toribio et al 2016). For each species, 12 functional traits of the leaves and branches were measured (table 1). These traits are informative of the strategies of water use, gas exchange and resistance to drought, which are fundamental aspects for the persistence of trees in dry forests (Tyree et al 2003).

During the rainy season, five individuals with a DBH ≥ 2.5 cm were selected from each species, with five leaves collected from each individual. The selected leaves were exposed to solar radiation, with no apparent damage by herbivores, and located at an equivalent height among the individuals. The leaves were stored in darkness inside a cooler in an airtight humid environment until subsequent processing. Fresh leaves were weighed and scanned, and petiole length measured with a digital caliper to 0.01 mm. Leaf area was calculated from digital images by using the ImageJ program (Abràmoff et al 2004). Finally, leaves were oven-dried for 48 h at 60 °C and leaf mass per area was calculated by dividing leaf dry mass by leaf area, while leaf dry matter content was obtained from the ratio of dry leaf mass: fresh leaf mass. In the case of composite leaves, the leaflets were used to obtain these attributes. Leaf compoundness was assigned the value of 1 for simple leaves, 2 for compound leaves and 3 for double-compound leaves. Leaf pubescence was a binary variable to which, the values of 0 and 1 were assigned, respectively, in the absence and presence of trichomes on the upper or lower leaf surfaces. Leaves without pulvini were assigned a value of 0, while those with pulvini on the first order leaflets had a value of 1 and those with pulvini on the second order leaflets had a value of 2. To calculate the leaf retention time for each species, a leaf area loss curve was generated following the procedures described in Méndez-Alonzo et al (2012), and the time taken by plants to lose 50% of their leaves was calculated.

Stem attributes (wood density, wood water content, bark thickness and bark water content) were measured on second order branches (1.5 to 3 cm in diameter) fully exposed to sunlight, taken from five trees at a similar height (2 m). On each branch, bark thickness was measured with a digital caliper to 0.01 mm, while wood density was calculated by obtaining the volume of a branch section without bark using the water displacement method, and drying the tissue in a conventional oven at 90 °C for 72 h, following Chave et al (2006). Maximum water content in the xylem (Xwc) or bark (Bwc) was estimated as ([fresh weight − dry weight] / dry weight) × 100. The fresh weight of the branches was obtained after hydration for 24 h in distilled water.

Variation of physical variables (T, RAD, ET₀), in relation to slope aspect and topographic position were examined by using two-way ANOVA, one for each season. In order to reduce any potential effect of spatial autocorrelation between the sampling units, the ANOVA was constructed using a re-sampling algorithm with 9999 permutations (Anderson and Legendre 1999). When one of the main factors (slope aspect or topographic position) was significant, the different factor levels were compared with a post-hoc test. When the slope aspect x slope position was significant, mean values among the six resulting groups were compared. Adjustment of P values were carried out following Holm (1979). Similar ANOVA was applied on values of TWI, while a one-way ANOVA was used for the case of MSD, % CF and MRD, to test only for topographic position on the N-facing slope.

The response of the functional dominance of vegetation to slope aspect and topographic position was evaluated for the 12 individual attributes and functional strategies, using multivariate and univariate approaches. The dominance of each functional attribute in each plot was obtained by calculating a community weighted mean (CWM), following the equation:

$$\begin{equation} \sum\limits_{i = 1}^S {p_i \times a_i} \end{equation}$$

where S is the total number of species, p_i is the relative abundance of species (i) in the community, and a is the attribute value of the ith species. This index indicates the average functional value of a given plant in the community, is very sensitive to environmental variation and has been correlated with certain ecosystem properties in various regions (Diaz et al 2004, Díaz et al 2007, Vandewalle et al 2010, Roscher et al 2012, Zhang et al 2014). CWM values of each attribute were calculated with the 'functcomp' function available in the 'FD' library of the R-CRAN (R Development Core Team 2016). The effects of the slope aspect and topographic position on the CWM of the individual attributes were assessed with a two-way analysis of variance. In order to reduce any potential effect of spatial autocorrelation between the sampling units, the ANOVA was constructed using a re-sampling algorithm, as the one previously described for the physical variables.

Exploration of co-variation among the attributes in the community, and determination of functional strategy continuums were both performed with principal component analysis (PCA), using the CWM values for each attribute in each plot. The effect of topographic factors on the dominance of functional strategies was assessed by analyzing the change in the average value of PCA scores along the main axes 1 and 2 using a two-way ANOVA. All data analyses were carried out with the open source program R 3.0.2 (R Development Core Team 2016); 'ade4' was used to perform the PCA with the function 'dudi.pca' (Dray and Dufour 2007), and the two-way analysis of variance with permutations was performed.

Physical variables were affected by slope-aspect and topographic position but less clearly by their interaction (table 2). During the dry season south-facing slopes presented higher values of RAD, T and ET₀ than N-facing slopes, and these results reversed during the rainy season (table 3). In addition, values of T and ET₀ decreased from the lower positions to the top hills, independently of the aspect and season (tables 2 and 3). During both seasons, values of RAD were lower in the middle slope position, mostly on the north-face slope (tables 2 and 3). The topographic wetness index exhibited similar values between slope aspects, with values increasing from top hills to valleys, while % of coarse fragments and main root depth exhibited similar, but marginal trends along topographic position, together indicating increasing of soil depth towards the valleys (tables 2 and 3).

The statistical tests applied to the CWM values of the individual functional traits, indicated that the response was more strongly affected by topographic position (eight attributes) than by slope aspect (two attributes) or the interaction of these two factors (two attributes; table 4). Bark water content, bark thickness, petiole length and leaf pubescence increased significantly towards the upper parts of the slopes, while wood density, leaf pulvini, leaf dry matter content and leaf retention time were higher on the lower parts of the slopes (figure 1).

Zoom In Zoom Out Reset image size

CWM for wood density and leaf retention time were significantly higher on north-facing slopes compared with those south-facing slopes (table 4, figure 1). Bark thickness and leaf retention time were the only attributes affected by the interaction of both topographic factors (table 4). In particular, thick barks were more dominant in the south than in the north-facing aspect at both the lower and upper parts of the slopes, while in the middle portions the opposite trend was presented (figure 1). Similarly, the dominance of readily-deciduous species (species with low leaf retention time) was higher on the southern aspect, although only on the upper parts of the slopes, while aspect did not affect dominance in the other two positions (figure 1).

The PCA showed that axes 1 and 2 explained 45.7% and 15.4% of trait co-variation, respectively (figure 2; table 5). The first axis of the PCA was negatively correlated with bark thickness, petiole length, bark water content, leaf area and leaf pubescence, and positively with wood density, leaf retention time, leaf dry matter content and leaf pulvini, partly suggesting a trade-off between two drought-coping strategies: (i) avoidance via the use of stored water with a rapid reduction in leaf area, and (ii) tolerance, via dense tissues and small leaves. Attributes negatively associated with axis 2 were leaf compoundness and leaf pubescence (two traits related to mechanisms of overheating avoidance) and xylem water content. This latter attribute showed relatively low variation among species and was independent of those traits related to tissue density (wood density and leaf dry matter content) and bark water content. In addition, leaf retention time was related to both axes, i.e. it was positively associated with dense tissues and negatively related to xylem water content and leaf compoundness. Leaf mass per area was the only attribute not correlated with either of the two axes.

Zoom In Zoom Out Reset image size

The ANOVA applied to the scores of axis 1 of the PCA indicated that these values decreased from the lower to the upper parts of the slopes (figure 3(a); F_2,30 = 12.73, P ≤ 0.001) and were significantly higher on north-facing slopes than on those south-facing (F_1,30 = 6.27, P ≤ 0.01). The average values of axis 2 were more negative towards the upper parts of the slopes (figure 3(b); F_2,30 = 5.62, P ≤ 0.01), particularly on the southern slopes (F_2,30 = 4.65, P  ≤ 0.05, interaction term).

Zoom In Zoom Out Reset image size

In our study site, slope-aspect strongly affected physical variables (RAD, T and ET₀), but such effects varied in opposing directions between the wet and the dry season. During the rainy season, growth may be favored on the north-facing slope, as this receives higher radiation, while during the dry season, a higher radiation load, higher temperature and potential evapotranspiration may compromise plant growth and survival (see Méndez-Toribio et al 2016, for an analysis integrating physical variables across the year in the same study site). The finding of similar results for another tropical dry forest landscape in Mexico (Gallardo-Cruz et al 2009), indicates that this pattern of topographic variation in physical environment is not unique to our study site, raising the hypothesis that in tropical dry forest regions slope-aspect does not impose a clear filter on plant water use strategies.

It has been proposed that in undulating landscapes, superficial water movements impulsed by gravity from the top hill to the lower slope positions, typically with reduction in slope steepness, is commonly accompanied by an increase in both soil depth and soil water content, generating a gradient of soil drought risk for plants (Holland and Steyn 1975, Daws et al 2002, Tesfa et al 2009, Liu et al 2012, for empirical and modelling evidence, Tromp-van Meerveld and McDonnell 2006, for discussions on oversimplification issues). In our study landscape, the detected increase in topographic wetness index, a trend for increasing soil depth and reducing rock fragment content, together comply with the view, that the risk of soil drought increases from the lower slope positions to the top hills. Remarkably, observations from local farmers point to the same view⁵, however our conclusions need to be corroborated by a more extensive characterization of soil depth across the landscape, because other factors as micro-topography may play a role.

Intriguingly RAD observed slightly lower values in the mid slope position, but those changes did not reflect into parallel changes in temperature or evapotranspiration. Interestingly, the detected increase of air temperature and potential evapotranspiration, from the top hill to the lower slope position suggest that surface air gets warmer from the top hills to the lower slope positions, potentially increasing risk of drought above ground (particularly during the rainy season, when values became more affected by topographic positions). We are unable to size at what extent surface conditions counteract the soil drought gradient, however a stronger change of wetness index across topographic positions (40%), compared to a slight change of RAD, T or ET₀, (1%, 5%, 8%, respectively), suggests that lower positions are warmer, but subject to lower risk of drought, at least during the growing season while rain water is flowing across the terrain surface. Soil movement downhill may also result in changes of soil nutrient content towards the valleys, affecting plant traits (Silver et al 1994). In this study we miss this factor as no data on soil nutrients were recorded.

This study demonstrates that topographic position is an overriding factor structuring functional dominance in tropical dry forests. The lack of response of the functional traits to slope aspect suggests that the difference in total annual radiation between north- and south- facing slopes is a less important filtering factor, probably due to the fact that at inter-tropical latitudes, the difference in annual radiation load between slopes is relatively small compared to that of higher latitudes (Holland and Steyn 1975, Gallardo-Cruz et al 2009, Méndez-Toribio et al 2016). The fact that slope aspect in these landscapes does structure species composition of the community (Méndez-Toribio et al 2016), but not its functional composition, suggests a convergence towards similar functional traits among different plant lineages. An answer to this interesting question would require a phylogenetic analysis to be conducted in the studied forest.

This study proposed the hypothesis that, in tropical dry forest landscapes, water-scarce environments (e.g. the upper parts of the slopes, particularly those facing south) tend to be covered by vegetation communities with attributes such as hard wood and dense tissues, that allow plants a greater tolerance to desiccation, compared to those of less dry areas (e.g. lower parts of the slopes, particularly those facing north). However, our results indicated the opposite trend, supporting the notion that opportunistic species, which quickly evade drought by shedding their leaves or by using water stored in their tissues, have an advantage over other plants on the upper parts of slopes, especially those facing south, where the growth period is shorter, more variable and exposed to high levels of insolation (Pineda-García et al 2011).

Our study suggests that the marked dominance of avoidance strategies on the upper portions of the slopes may be the result of filtering processes related to a soil depth gradient, from shallow soils on the high parts, to deep soils on the lower sections of the slopes. This affects the water dynamics, both spatially and temporally. Shallow soils on the high portions of slopes have a low water retention capacity and dry quickly after rain events. This translates into a high risk of death for plants over prolonged periods of drought during both the growing season and the beginning of the dry season (Tromp-van Meerdel and McDonnell 2006, Pineda-García et al 2011, Méndez-Alonzo et al 2013). Under this scenario, it is likely that the rapid loss of foliage during drought may drastically reduce the risk of cavitation, especially in species with wide vessels, typically associated with low-density wood (see Hacke and Sperry 2001 for wood density-vessel anatomy relationships; Méndez-Alonzo et al 2012, for TDF species). Moreover, it is likely that trees with low-density wood and large water deposits in the parenchyma and/or bark are able to use this stored water to rehydrate the xylem and temporarily dissociate the plant water balance from the soil water potential, as previous studies suggest (Goldstein et al 1989, Daws et al 2002, Scholz et al 2007, Meinzer et al 2008, Pineda-García et al 2012, Rosell et al 2014). Other studies have advanced the idea that species with water stored in their stems (xylem or bark) tend to be highly efficient in terms of water conduction and photosynthetic activity, and may be able to respond to water pulses by rapidly developing leaves and roots (Sobrado 1993, Holbrook et al 1995, Holbrook and Zwieniecki 2005, Pineda-García et al 2015). Such attributes contribute to maximize total carbon assimilation in environments with short and unpredictable growth periods (Eamus 1999, Santiago et al 2004, Choat et al 2005, Franco et al 2005, Pineda-García et al 2011). Taken together, the results of our study combine to support the notion that water-stress delaying strategies and the capacity to exploit water pulses may both represent an adaptive advantage in temporarily heterogeneous environments with long episodes of drought and short growth periods.

The dominance of readily-deciduous and water-storing species on the upper portions of the slopes, as well as the dominance of species with dense tissue tolerant to water stress on the lower portions of slopes, may also be reinforced by selective pressures of soil depth on rooting depth, and the occurrence of a tight coordination of rooting habit with stem and leaf traits. It is therefore expected that shallow soils with high runoff and poor vertical percolation prevent the development of deep roots, favoring species with water-storing strategies, while deeper soils, with greater influx and percolation of water, favor species with deep roots that are capable of capturing water from deeper and longer-lasting deposits. This is a particularly interesting possibility, given the fact that a previous study done with young plants in a dry tropical forest (Paz et al 2015), reported evidence of a potential trade-off between the two water-stress delay mechanisms; storing water in the tissues or rooting more deeply. This hypothesis seems plausible and its further exploration calls for a detailed study of above and below ground traits along topographic gradients, coupled with extensive measurements of soil depth. The expected increase of distance to water table from the valleys to the top-hills may exaggerate the detected drought risk for plants and selection on rooting depth, however this important factor still awaits to be measured, as water table depth can vary with local specific stratigraphy (Moore and Thompson 1996).

In the dry forest, tree foliage is subject to strong thermal stress, especially during periods of low soil water availability. It is therefore common to find leaf structures, such as the pulvini and trichomes, that help decrease the heating load and the risk of overheating or water loss. Species with leaf pulvination optimize the functioning of photosynthetic structures in extremely hot environments by changing the position of leaves or leaflets, this in turn affects transpiration and thus the water relations of the plant (Comstock and Mahall 1985, Schulze et al 2002). In our study, the dominance of species with leaf pulvination on the lower portions of slopes suggests a filtering process through this mechanism, as was in these landscape units where higher air temperatures were recorded (Méndez-Toribio et al 2016). The fact that this attribute is strongly linked to certain lineages of the Fabaceae family is remarkable. This family dominates the lower parts of the slopes of the dry tropical forests under study (Méndez-Toribio et al 2016), which suggests the existence of a significant phylogenetic component in the functional structure of the forest (Pennington et al 2009). We are not aware that other attributes analyzed here have a phylogenetic component as extreme as leaf pulvination, although both the functional structure and phylogenetic relationships among species remain to be clarified.

It has been proposed that trichomes act as mechanical filters, reducing the incidence of ultraviolet radiation and potential damage to the leaf tissue (Levizou et al 2004) and that they attenuate the drying effect of wind due to the increased thickness of the boundary layer (Meinzer et al 1985, Brewer and Smith 1994), reducing water loss through transpiration (Brewer and Smith 1994) as well as increasing water retention via a decoupling from stressful environmental conditions (Konrad et al 2015). In this context, the dominance of trichomes towards the upper parts of slopes could result from the exposure of the leaves to environments with drier soils but relatively cooler temperatures, where higher exposure to wind and ultraviolet radiation are to be expected. Previous studies in semi-arid areas of North America also have identified dominance of plant species with trichomes, suggesting that this phenomenon is not unique to our study area (Sandquist and Ehleringer 1997).

It is interesting to note that leaf mass per area, which is important in the context of carbon foliar economy (Wright et al 2005), proved irrelevant to the characterization of functional strategies, since it showed no response to the topographical factors analyzed, nor was it included on the main axes of morpho-functional variation detected for this forest. This would confirm the proposition that, while this attribute is strongly related to the competitive ability of species in communities with marked vertical light gradients (Westoby et al 2002), it does not seem to be an important component in the functional strategies of trees in dry tropical forests, where heterogeneity in water availability strongly selects for a diversity of water use strategies among species (Pineda-García et al 2011).

Another attribute that was expected to vary significantly along the water topographic gradient is xylem water content, a key attribute related to the delay of water stress (Roger 2001). However, this attribute did not respond individually to any of the topographic features analyzed. This may be due to our failure to capture the most important water storage sections during branch sampling, which in certain species are found in the main stem (Gartner 1995, León and Espinoza 1999). In contrast, the strong co-variation between the topographic gradient and the bark water content and thickness suggests that the storing of water in thick barks may play an important role in maintaining water xylem water status during drought conditions (Gartner 1995). This could be equally or more important than the role of xylem water storage in community processes of functional differentiation along a topographic gradient, an aspect that deserves to be investigated in more detail (Rosell et al 2014).