Tree species identity and soil water uptake
Despite a large range of feasible source contributions for a given soil layer and considerable variation among replicates, three distinct water uptake strategies were identified for trees grown in monocultures: strong reliance on upper soil layer water sources (C. odorata), uptake from all soil layers in equal proportions (H. crepitans, A. excelsum and L. seemannii) and reliance mainly on deep soil water (T. rosea). The differences among species in their depth of water uptake were most pronounced during the dry season.
The depth of soil water uptake by plants has been linked to rooting patterns and active root area (Ogle et al., 2004). The species C. odorata and H. crepitans are reported to have most of their root biomass located in the upper horizon (Cintron, 1990; Coll et al., 2008; Jefferson Hall, personal communication 2010), which may explain their reliance on upper soil water sources. H. crepitans might have been able to cope with drier conditions by storing water in its particularly large trunk, which, at Sardinilla, is shaped like an elephant's foot, providing a buffer during periods of limited water availability (Delagrange et al., 2008). T. rosea has both lateral roots and deeply penetrating taproots as observed during root excavations at Sardinilla (Jefferson Hall, personal communication 2010), which enables T. rosea to extract water from deeper soil layers. Water uptake from deeper layers by L. seemannii during the dry season was also reported from a natural forest on Barro Colorado Island, Panama (Meinzer et al., 1999).
Plasticity in their patterns of water uptake was observed in T. rosea and to a lesser degree in A. excelsum and L. seemannii (Figure 3a–c). The ability to switch between soil water sources depending on water availability has been reported for a number of species in arid/semiarid regions (e.g. Dawson and Pate, 1996; Eggemeyer et al., 2009) and tropical rainforests subjected to seasonal droughts (e.g. Oliveira et al., 2005; Markewitz et al., 2010). A recent study conducted during the dry season in a tropical rainforest in French Guiana found changes in water uptake depth of up to 50 cm within a few days (Stahl et al., 2013). Plasticity in water uptake depth may put plants at an advantage if competition occurs within the ecosystem or water becomes extremely limited (Ehleringer and Dawson, 1992). Strong vertical gradients in nutrient levels are often found in tropical ecosystems (Jobbagy and Jackson, 2001). Accessing nutrient-rich water from the upper layer during the wet season suggests a dynamic water and nutrient use strategy and has implications on the stand nutrient and water budget as discussed later.
In summary, we found interspecific differences in soil water uptake patterns, although all species were exposed to the same environmental conditions. This contrasts with the view that belowground plant functional traits such as root distribution and water uptake depth will be similar among species, as these traits are closely related and governed by soil water availability (Dawson et al., 1998; Meinzer et al., 2001). Our findings also suggest that the observed differences in water acquisition strategies are species-specific traits, as we did not find a correlation between xylem water isotopic composition and tree size for a given species. No strong correlation between tree height and water uptake depth was found by Stahl et al. (2013) investigating trees of varying size in a tropical rainforest in French Guiana. The authors reported that the water uptake depth of shorter trees (less than 25 m) was highly variable ranging from 10 to 200 cm depth (Stahl et al., 2013). In contrast, a strong dependence between tree diameter and water uptake depth was reported by Meinzer et al. (1999) with larger trees extracting water mainly from the upper soil profile.
Phenology, tree transpiration and complementary water use
Our results demonstrated that water acquisition strategies were closely linked to leaf phenology especially during the dry season. Species able to tap into deeper soil water sources were able to produce and sustain transpiring leaves over the course of the dry season (Figure 5a, b). On Barro Colorado Island, Central Panama, the sap flux of A. excelsum increased, whereas xylem δ2H values became more depleted over the course of the dry season (Meinzer et al., 1999), indicating that A. excelsum was able to tap into deeper soil water sources during this period.
In monocultures, stand transpiration rates mediated by species-specific water uptake strategies had in turn an effect on seepage rates. Seepage rates calculated for the wet season 2008 were inversely correlated with stand transpiration rates with the highest seepage rate in H. crepitans (−6·13 mm day−1, Sprenger et al., 2013; stand transpiration rate = 0·69 mm day−1, Kunert et al., 2012) and the lowest seepage rate in L. seemannii (−5·18 mm day−1, Sprenger et al., 2013; stand transpiration rate = 2·62 mm day−1, Kunert et al., 2012). This suggests that species-specific water uptake patterns have implications on stand water budgets.
No significant differences were found in the soil water uptake pattern between trees grown in monoculture and mixtures. This suggests that for the species studied, the species-specific water acquisition pattern was invariant. One may argue that the investigated 7-year-old trees were too young or the tree density too low for trees to interact. Measuring the vertical and horizontal extension of the rooting system (roots > 2 mm) revealed that lateral expansion of roots exceeded the 3-m spacing of planting, suggesting belowground interaction among trees (Jefferson Hall, personal communication 2010).
We did find some evidence of complementary water use among co-occurring species, especially between C. odorata and T. rosea in 5-species plots in the dry season (Figure 4a). Further, any of the other three species is unlikely to strongly compete with C. odorata or T. rosea for water, as their water uptake depth differs at 10–30 and 30–50 cm depth (Figure 4a). In contrast, if H. crepitans, A. excelsum and L. seemannii occurred together in a mixture, there would be a considerable overlap in soil water uptake depths. In 2/3-species plots, these strategies remain the same for T. rosea, A. excelsum and L. seemannii (Figure 4b). Neighbourhood effects may explain the high variability and differences found for C. odorata and H. crepitans in 2/3-species plots compared with 5-species plots. Both species rely predominantly on water from the upper 30 cm, and when grown next to each other, as in plot T4, they may compete for water especially in the dry season (Figure 4b). In the 5-species plots, C. odorata and H. crepitans are not planted next to each other. H. crepitans did not grow well in mixtures with C. odorata and L. seemannii (T4) as indicated by the lower diameter (Table 1). Investigating the variation in tree diameter and height at Sardinilla, Potvin and Dutilleul (2009) found that the size of neighbours was the largest source of variation, and suggested that neighbourhood plays a central role in determining productivity. Thus, 2/3-species mixtures most likely vary in their degree of complementarity depending on species composition. Considerable differences in basal area, aboveground carbon pool, stand transpiration and nutrient stocks were found among the existing six 2/3-species plots (Oelmann et al., 2010; Zeugin et al., 2010; Povin et al., 2011; Kunert et al., 2012), highlighting the importance of species composition. However, not all possible 3-species combinations were established at Sardinilla and thus could not be tested.
In summary, we did find some evidence of complementary water use among co-occurring species, which may partly explain the twofold higher annual transpiration rates of trees growing in the 5-species mixtures (Kunert et al., 2012) and lower seepage rates (Sprenger et al., 2013).
It is important to note that our sampling approach only assessed the spatial partitioning of soil water sources along a vertical axis. Uptake along the vertical axis is mainly influenced by species-specific differences in root activity (Stratton et al., 2000). However, spatial partitioning can also occur along a horizontal axis (laterally) defined by the pattern of species distribution and spacing (Hinckley et al., 1991).