Study site

The BEF-China experiment is located near Xingangshan Township, Jiangxi Province (N29°08–11, E117°90–93), P.R. China. The mean annual temperature is 17.4°C and mean annual rainfall is 1635 mm. The climate of the study area is characterized by subtropical summer monsoon with a wet season from May to July and a dry winter. After the clear-cut of a Cunninghamia lanceolata plantation in 2008, an experimental forest was planted on a plot-level based approach with 400 tree individuals per plot (25.8 m x 25.8 m; planted in 20 rows of 20 tree individuals each), using a planting distance of 1.29 m and including a total of 24 tree species on 261 plots to investigate biodiversity effects on ecosystem functions (see [33] and [34] for detailed explanations). This study focuses only on monoculture plots of trees that ranged in mean height from 1.10 m to 5.76 m in 2013. At the time of study, the trees were five years old. No specific permissions were required for these locations and activities. The field studies did not involve endangered or protected species.

Experimental design and data sampling

TKE was measured during five rainfall events with an event-based approach for a total of 11 species in 17 monoculture plots in 2013. Within the central part of each plot (including 6 x 6 trees), eight randomly assigned positions with distinct distances to the tree stems were used to measure TKE (1) 15 cm away from tree stem, (2) in the middle of two tree individuals, (3) in the middle of four individuals, (4) 45 cm away from tree stem, (5) at the 45 cm x 120 cm intersection between two individuals, (6) below the first branch of a tree individual, (7) at the 75 cm x 75 cm intersection between two individuals, and (8) 30 cm away from tree stem). TKE was measured using splash cups [35] and representative values of J/m² were obtained by using a modified version of the function provided by [35]. Next to each splash cup, a rainfall collector was installed to quantify throughfall with a high spatial resolution. Rainfall events were registered by the BEF-China climate stations and classified by rainfall intensity, duration and total amount (S3 Table). A total of eight leaf traits and six architectural traits were analyzed. Leaf traits included leaf area (LA), specific leaf area (SLA), leaf pinnation (simple or pinnate), leaf margin (entire or dentate), trichome cover of upper leaf surface, leaf thickness, leaf toughness and leaf habit (deciduous or evergreen). These traits were measured on individuals planted in the experiment [36,37]. Architectural traits examined at each tree individual were total height, elliptic crown area, number of branches, stem diameter at ground level (GD) and crown base height (CBH) [38]. Leaf area index (LAI) was registered at each TKE measuring point under diffuse radiation conditions, using a Nikon D100 with a Nikon AF G DX 180° and HemiView V8 (Delta-T) [39]. Table 1 gives an overview of all tree species with leaf and architectural traits influencing TKE.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 1. Leaf and tree architectural traits of the tree species included in the present study according to a significant influence on throughfall kinetic energy.

https://doi.org/10.1371/journal.pone.0128084.t001

Data analyses

Species-specific variation of TKE was investigated using linear mixed-effect models fitted by restricted maximum likelihood. Rainfall event, species identity and the interaction of species identity with rainfall event were included as fixed factors. The interaction of species identity with rainfall event allowed detecting whether differences among species only played a role at certain rainfall events. Plot, measurement position within each plot, interaction of plot with rainfall event and interaction of plot with position entered the model as random effects. For testing effects within each rainfall event, rainfall event was not used as fixed factor. Contrasts were fitted before species identity to detect species which had significantly higher or lower TKE than mean of all others. Significant effects were detected using Wald Test statistics with Type I SS ANOVAs. In total, 625 data points entered the analyses (5 events x 17 plots x 8 positions– 55 failed measurements).

To specify possible effects of species identity, mediation analysis were constructed by fitting mediation trait variables before the species identity term. Mediation variables were detected as such, if significance of species identity was changed from significance to non-significance and if the mediation variable itself significantly influenced TKE. To identify the most important mediation variable, categorical levels were predicted and ranked by their magnitude of TKE differences (effect size). For continuous mediation variables, the difference in TKE was evaluated when increasing mediation variable by one standard deviation. Each model was only fitted with a single mediation variable to avoid multicollinearity among traits and overparameterization.

Additionally, a model was constructed consisting only of leaf traits (fitted first to avoid underrepresentation by larger effects), tree architectural traits and throughfall to test for their influences on TKE. In this model, plot and rainfall event were considered as random effects. Model simplification was done using step-wise backward selection with the maximum likelihood approach [40]. Hence, the final model only contained significant effects (P<0.05). Prior to the analyses, all covariates have been checked for collinearity (correlations were not allowed to exceed R = ± 0.7). Hence, leaf toughness, leaf thickness and crown area were omitted in the final model due to multicollinearity. Predictions were used to identify the effect size according to the method described above.

If a measuring position was influenced by more than one tree individuals, mean values of leaf and tree architectural traits of surrounding individuals have been calculated.

TKE data was log-transformed to ensure normal distribution. Model residuals did not show violation of model assumptions (normality and homogeneity of variances). Analyses were conducted using R 2.14.1 [41]. Linear mixed effects models were analyzed with R package “asreml” [42] and “lme4” [43].

Species-specific TKE variation (H1)

Our first hypothesis was confirmed by the significant influence of species identity on TKE. However, only three out of 11 species showed distinct differences in TKE compared to overall means. Among these, two species positively (Choerospondias axillaris and Sapindus saponaria) and one species negatively (Schima superba) affected TKE (Fig 1). Species-specific differences of throughfall amount or interception have been frequently reported [18,44]. Moreover, drop size distribution as an important driver of TKE has been found to be species-specific [27]. Nevertheless, preceding studies found no significant difference in TKE among certain species [5,16,21,44,45], which is in line with our findings. Furthermore, the highly significant interaction of species identity with rainfall event emphasizes the importance of abiotic characteristics in TKE distribution [18]. An influence of species identity on TKE was found at low-peak intensity rainfall events, whereas TKE at high rainfall intensities was not species-specific. An exception occurred at rainfall event 4 where species identity did not affect TKE although rainfall intensity was low. Higher intensity rainfall usually results in considerable canopy vibration, through which the drop sizes are reduced [20]. Therefore, the variation of TKE at high intensity rainfall could be much less than that at low intensity rainfall leading to no species-specific differences. However, this effect often is superimposed by an increase of total throughfall amount with higher rainfall intensities.

The species-specific effects of canopies of Choerospondias axillaris, Sapindus saponaria and Schima superba have strong implication for managing TKE. Planting Schima superba, which negatively affected TKE, has the potential to decrease soil erosion in early successional stages. Schima superba is also well-known for high values of canopy interception during rainfall [46]. This could be partially attributed to the high LAI and re-interception of rainfall by lower canopy layers [25,30]. Low TKE below Schima superba was also reported by [5]. These findings are as much more relevant as Schima superba represents one of the dominant tree species in the regional species pool [47,48]. Choerospondias axillaris increased TKE which is consistent with high runoff volumes found for this species in comparison to peanut crops [49]. However, despite a TKE increase, higher soil loss with Choerospondias axillaris can be counteracted by an intact litter cover [29].

Leaf and tree architectural traits mediate species-specific variation (H2)

First, the strong impact of rainfall event on TKE suggests that the TKE variation is pre-determined by the characteristics of rainfall events, such as duration, total rain amount, wind speed and rainfall intensity [18]. However, our study could not confirm a significant effect of rainfall duration and wind speed on TKE. Within a specific rainfall event, TKE differed among tree species as species responded differently to different rainfall intensities [50]. Independent of a specific rainfall event, the species identity effects on TKE were mediated by leaf and tree architectural traits. LA, leaf habit, leaf pinnation, GD, CBH, tree height, number of branches and LAI as biotic factors were found to be responsible in mediating species-specific TKE (Table 2). Moreover, the significant effect of throughfall measured at each splash cup position on TKE showed the influence of biotic and abiotic factors on TKE [18].

In our study, species-specific changes of TKE were induced most by leaf area. A higher leaf area can increase the gathering of rain water and thus may cause larger drops resulting in higher TKE [13,20,26]. In contrast, many studies have reported on the positive influence of leaf area on interception [44], which leads to decreasing throughfall amount and decreasing TKE. Therefore, in our study variation of interception might only play a minor role in explaining species-specific differences in TKE, since all rain events lasted long enough to compensate the effect of canopy storage at the beginning of each event. However, the high effect size of LA in our study might be an overestimation, since leaf areas of Choerospondias axillaris and Sapindus saponaria, both with largest TKE, were almost twice of the standard deviation above the mean. This is due to the fact that for measurements of leaf area, the leaflets of pinnate leaves are traditionally added up to a total value per pinnate leaf [51]. Furthermore, water might gather at the branch, where each leaflet splits, which in turn may result in increased drop size and thus TKE.

TKE varied second-most between deciduous and evergreen species where deciduous species showed higher TKE. Similarly, [5] found that Castanea henryi and Quercus serrata as examples of deciduous species yielded higher TKE than the evergreen species Schima superba. Leaf habit represents a dominant segregation for many leaf traits and has been found to influence core functional and physiological processes specifically in the study species [47] as well as globally [52–54]. Deciduous species tend to have leaves with higher SLA [55], which we found to positively affect TKE. In addition, evergreen species tended to have a larger crown length ratio (ratio of crown length to the total tree height). Two mechanisms might elucidate the great variation between deciduous and evergreen species: (i) A lower tree height decreases falling height of raindrops and thus, results in lower TKE; (ii) a larger crown area with lower tree height (higher crown length ration) may increase LAI which results in higher interception, leading to decreasing throughfall. Moreover, leaf pinnation (pinnate or simple) can alter drop sizes. On the basis of higher margin circumference in relation to total leaf area [56], pinnate leaves create more dripping points. In addition, pinnate leaves showed the highest leaf area (see Table 1) with the exception of Sapindus saponaria. Corresponding to the above, a higher leaf area increases TKE. However, different leaf margins contributed only marginally to species-specific changes in TKE. As demonstrated in former studies [13,28], tree height was the most important tree architectural parameter to describe species-specific differences in TKE. Increasing tree height can contribute to higher TKE by several processes: (i) higher drop velocity due to higher falling heights [57], (ii) larger crown width [38] that increases drop size through increased confluence, and (iii) larger crown width is associated with higher LAI, which creates more dripping points [26,28]. The species-specific differences in TKE were mediated by LAI which negatively affected TKE. It is known that high canopy thickness increases drop splitting by dripping on branches and leaves [20,25], which in turn may decrease raindrop sizes. Moreover, higher canopy thickness in young forest stands might decrease space between vegetation and surface resulting in lower rain drop velocities [5]. Additionally, with denser and thicker crown cover water storage in the canopy increases, but this effect can be neglected with regard to rainfall durations longer than a day.

CBH (with half of the effect size of tree height) contributed to species variances as indirect factor, as it is usually related to tree height. CBH may constitute “the last barrier” in releasing throughfall drops, determines the falling height and thus drop velocity. Yet, our data showed that CBH contributed to interspecific TKE much less as compared to tree height. One reason might be that the same CBH might occur at trees along a large range of tree height.

GD positively and branch number negatively mediated species identity of TKE, but only to a smaller extent (< 7% difference). This suggests that GD mediated species-specific changes as an indirect effect of tree growth characteristics (tree height and LAI). Furthermore, the number of branches might affect TKE through an indirect effect via LAI. Branches gather throughfall and release it at any random position or transfer it directly to the stem, which decreases throughfall and increases stemflow [26]. However, our results indicated that this was a weak effect and water might both, be distributed along the branches and transferred to the stem in equal proportions [26]. Moreover, the greater effect size of LAI demonstrated that leaves are much more important than the branches as regards the impacts of species-specific TKE.

Besides the significant effect of plant traits on TKE, throughfall amount was also highly correlated with TKE, but showed smaller effect sizes than the findings from other studies [19]. In most studies, throughfall amount was found to be the major driver of spatial variability of TKE [16,35]. However, our species comparison revealed that shifts in drop formation and drop velocity within a specific rainfall event might have a higher impact on TKE than the total amounts of rainfall. Thus higher throughfall amounts do not necessarily lead to higher TKE at rainfall event level.