Differential intra-specific stemflow funnelling efficiencies of Caragana korshinskii within arid desert ecosystems

Drylands cover about 41% of Earth's terrestrial surface (Reynolds et al. 2007; Huang et al. 2015), characterized by a two-phase mosaic composed of vegetated patches interspersed with bare patches (Aguiar & Sala 1999; Tongway et al. 2001; Rietkerk & Van de Koppel 2008), where precipitation is scarce and typically unpredictable and hence a key limiting factor of ecosystem functioning (e.g., Noy-Meir 1973). Shrubs are the dominant vegetation over the vast drylands, and play a crucial role in the hydrological and biogeochemical cycles in terms of redistributing incident precipitation (Schlesinger & Pilmanis 1998; Llorens & Domingo 2007; Zhang et al. 2016a). Precipitation falling on the shrub canopy either is intercepted and subsequently evaporated (interception loss) or reaches the ground diffusely by throughfall or concentrated by funnelling down the stems in the form of stemflow.

Stemflow is volumetrically small (normally 5–10% of incident precipitation) in comparison to the other components of the canopy water balance; it is, however, of high eco-hydrological and chemical importance due to its very local input to nature (Johnson & Lehmann 2006; Levia & Germer 2015). Nutrient-enriched stemflow funnels down the shrub stems and infiltrates into deep soil layers through preferential pathways such as roots, creating islands of soil moisture and nutrients (e.g., Mauchamp & Janeau 1993; Navar 2011; Schwärzel et al. 2012; Zhang et al. 2013, 2016b). Stemflow is thus considered to be an important biological transfer mechanism in contributing to the development of ‘fertile islands’ under shrub canopies in the arid and semi-arid environments where water and nutrients are typically limited (Garner & Steinberger 1989; Whitford et al. 1997).

Stemflow has been reported to be influenced by a suite of biotic and abiotic factors, including canopy structure and architecture (e.g., Crockford & Richardson 2000; Park & Cameron 2008; Wang et al. 2013), rainfall depth, intensity and duration (e.g., Dunkerley 2014; Zhang et al. 2015), antecedent dry period (Tobon et al. 2004), wind direction and speed (e.g., Xiao et al. 2000; Van Stan et al. 2011) and vapour pressure deficit (e.g., Staelens et al. 2008; Pugh & Small 2013). The complex interactions among these factors also make it difficult to determine how a single factor affects stemflow production (Levia & Germer 2015).

Stemflow shows a highly inter-specific variation (the variation among species) due to morphological differences among tree/shrub species in a wide range of ecosystems (Van Dijk & Bruijnzeel 2001; Levia & Frost 2003; Llorens & Domingo 2007; Swaffer et al. 2014; Janeau et al. 2015; Zhang et al. 2015). For example, Llorens & Domingo (2007) reported that in the European Mediterranean area, stemflow percentage averaged 3%, with a variation coefficient of 111%, ranging from about 1% for Picea abies, Pinus sylvestris and Quercus pyrenaica and 12% for Pinus nigra. Yet the intra-specific variation (the variation within the same species) of stemflow has been poorly explored. A handful of recent studies from forested ecosystems showing high stemflow generating efficiencies for small trees (Germer et al. 2010; Siegert & Levia 2014; Su et al. 2016) are stimulating further research in various ecosystems. In the present study, event-based measurements on stemflow of nine shrubs of C. korshinskii varying from small to large were made during three rainy seasons of 2011–2013 within a water-limited arid desert ecosystem. The objectives of our study were: (1) to quantify the amount of rainwater flowing down the shrub stem and (2) to evaluate the differential funnelling efficiencies among individual shrubs within the same species. Achievement of these research objectives is important to gain a better understanding of vertical and horizontal water distribution under shrub canopies, and of different water acquisition strategies of individual shrubs within the same species in water-limited arid desert ecosystems.

Field measurements were carried out during three growing seasons of 2011–2013 at the Water Balance Experimental Field (WBEF) (Figure 1(a)) of Shapotou Desert Research and Experiment Station of Chinese Academy of Sciences (37°32′ N, 105°02′ E, with an elevation of 1,300 m a.s.l.), on the southeastern fringe of the Tengger Desert in northwestern China. Mean annual precipitation is 191 mm with 80% of rain falling between July and September. Most storms are of a low amount and intensity, with around 80% of the rainfall intensities ≤5 mm h⁻¹ (Zhang et al. 2015). The groundwater is at a depth of 50–80 m, and therefore inaccessible to plant roots. Potential evapotranspiration is approximately 2,500 mm during the growing season (April–October), resulting in a large annual moisture deficit. Mean maximum and minimum air temperature is 24.7 °C in July and −6.1 °C in January, respectively. Annual mean wind velocity is approximately 2.8 m s⁻¹.

C. korshinskii is a multiple-stemmed deciduous perennial leguminous shrub with an inverted cone shape, and is one of the successful shrubs used in revegetation for protecting the Baotou-Lanzhou railway against encroaching sand dunes in the Shapotou area (Li et al. 2006; Li 2012). The stems are smooth, the leaves are pinnately compound and opposite or subopposite in arrangement and 6–10 cm long, and each pinna has five to eight pairs of ovate leaflets (7–8 mm in length and 2–5 mm in width). The average height and the average canopy diameter of C. korshinskii at WBEF was 145 and 130 cm, respectively.

Nine robust and healthy shrubs of C. korshinskii were selected for field observation (Figure 1(b)) according to the gradient of stem numbers of C. korshinskii in WBEF, and their morphological characteristics are quantified in Table 1. A shrub with more stems is also more likely to have a greater canopy area, thus we assume the nine shrubs represent the range of sizes of C. korshinskii at larger community scales. Stem diameter was measured with a vernier caliper at each stem base, and stem angle was determined by a protractor. The plant area index (PAI, m² m⁻²), one-sided plant area (stems, twigs, leaves) per ground area, was estimated using a LAI-2000 plant canopy analyser (Li-Cor, Inc., USA). Shrub height was measured at the centre of the canopy using a tape. The projected canopy area (approximated as an ellipse) was calculated by taking the east–west and north–south diameters through the centre of the fullest part of the canopy (Martinez-Meza & Whitford 1996). Shrub volume was estimated by considering canopy shape as an inverted elliptic cone. Basal area was defined as the sum of the cross-sectional area of individual stems of a given shrub.

Stemflow was collected using an aluminium foil collar that was fitted around the entire circumference of individual shrub stems (Figure 1(c)). Stemflow volume was measured by a graduated cylinder for each individual stem after the cessation of each rainfall event.

A standard tipping bucket rain gauge (Adolf Thies GMVH & Co. KG, Germany) with a resolution of 0.1 mm was installed in an open area approximately 50 m from the study plot. Individual events were separated by at least 4 h without rainfall.

Descriptive statistics were compiled for shrub morphological metrics, rainfall depth, stemflow volume and funnelling ratio. The correlations among shrub morphological metrics were analysed using Pearson correlation. The linear function was used to determine the relationships of stemflow volume with rainfall and shrub morphological metrics. The linear and the exponential decay functions were used to determine the relationships between funnelling ratio and shrub morphological metrics. All the descriptive statistics, correlations and fitted functions were performed using SPSS 16.0 statistical software (SPSS Inc., Chicago, IL, USA).

In total, 37 rainfall events (totalling 410.3 mm) that can produce stemflow were recorded during the study period, ranging in depth from 2.4 to 28.8 mm with a mean of 11.1 mm and a variation of coefficient (CV) of 61% (Table 2).

A great variability in stemflow volume can be found among rainfall events for individual shrubs. Stemflow volume increased significantly (P < 0.01) with rainfall depth for all studied shrubs (Figure 2). Total event stemflow volumes differed considerably among individual shrubs of C. korshinskii (Table 2). For example, the sum of stemflow volume produced by CK_7 (190,049 mL) was 43.3 times greater than CK_1 (4,387 mL). These findings highlight the important interactions between rain event magnitude and intra-specific stemflow yield.

Mean event funnelling ratio also differed considerably among individual sample shrubs of C. korshinskii, suggesting the differential funnelling efficiencies among shrubs (Table 3). For example, the funnelling ratio was 37 for CK_1 and 123 for CK_3. Generally, the event funnelling ratio for individual shrubs had an increased tendency with rainfall depth (Figure 3(b)) until a threshold value around 12 mm and then began to decline (Figure 3(a)).

The relationship between the total stemflow yield from all the individual stemflow events and shrub morphological metrics was analysed separately (Figure 4). Stemflow volume had a significant positive linear relationship with projected canopy area (R² = 0.92, P < 0.001), shrub volume (R² = 0.83, P < 0.001), basal area (R² = 0.94, P < 0.001), shrub height (R² = 0.49, P= 0.036), number of stems (R² = 0.45, P= 0.047), stem diameter (R² = 0.53, P < 0.001) and stem length (R² = 0.19, P < 0.001), respectively. It had an increased tendency with PAI (Figure 4(d)), although the relation was not statistically significant (P= 0.13). No clear relation was found between stemflow volume and stem angle (Figure 4(g)). Moreover, for morphological metrics (Table 1), significant positive relations (r > 0.7, P < 0.05) were found among canopy area, canopy volume, basal area and stem diameter.

Morphological metrics clearly affected stemflow funnelling efficiencies for shrubs within the same species (Figure 5). Funnelling ratio decreased at first and then became constant after a threshold value with increasing projected canopy area (Figure 5(a)), shrub volume (Figure 5(b)), basal area (Figure 5(c)) and stem diameter (Figure 5(h)), respectively, being fitted by an exponential decay function. Funnelling ratio was negatively correlated with PAI (Figure 5(d)), shrub height (Figure 5(e)) and number of stems (Figure 5(f)), respectively, although the relations were not statistically significant for the former two parameters; it increased with stem angle (Figure 5(g)), although the relation was rather weak (R² = 0.08). No clear relationship was found between funnelling ratio and stem length (Figure 5(i)).

Stemflow volume represents the quantity of stemflow flux. In our study, stemflow volume increased with some shrub size metrics such as canopy area, canopy volume, basal area and stem diameter (Figure 4), and significant positive Pearson correlations (r > 0.7, P < 0.05) were found among these metrics. This indicates that smaller shrubs generate lower quantities of stemflow than larger shrubs, which is supported by studies from forested ecosystems that larger trees within the same species are inclined to generate higher volumes of stemflow due to large canopies, i.e., rainwater collecting area (e.g., Levia et al. 2010; Honda et al. 2015). However, this does not mean that larger trees/shrubs profit more from stemflow than smaller ones, because basal area, not canopy area, is the true area over which stemflow is delivered to the soil (Herwitz 1986; Germer et al. 2010; Levia et al. 2011). Although stemflow volume is an important metric for calculating watershed inputs and overall budgets, it is unreasonable to use this metric to assess the hydrologic implications of stemflow by leaving out the stemflow-affected area (Germer et al. 2010). Also, this unstandardized metric is unsuitable for use in hydrologic models (Levia & Germer 2015). This metric thereby inevitably fails to compare the funnelling efficiencies among individual plants. Alternatively, the stemflow funnelling ratio is an effective quantitative tool to assess how efficient a tree/shrub is at funnelling water to its base, because it identifies the true delivering area (shrub basal area) of stemflow and allows researchers to readily compare results with other studies and achieve meaningful cross-site comparisons (Llorens & Domingo 2007; Levia & Germer 2015).

In our study, funnelling ratio differed considerably among individual shrubs within the same species (Table 3), with a mean funnelling ratio range of 37–123 (CV = 33%), suggesting that the shrub basal area of C. korshinskii can receive 37–123 times the amount of rainwater by stemflow as compared to an open area. This fell in the range of studies from arid and semi-arid environments (Navar 1993; Llorens & Domingo 2007; Li et al. 2008; Yang et al. 2008; Wang et al. 2011). For example, Li et al. (2009) reported an average funnelling ratio of 77.8 and 48.7 for shrubs H. scoparium and S. psammophila, respectively; Garcia-Estringana et al. (2010) observed that the average funnelling ratio for nine Mediterranean shrubs was 104 with a range of 30–250. Stemflow is thus considered as an essential water resource available for growth and survival of shrubs, which may further contribute to the stability of shrub communities in water-limited arid desert environments. Moreover, we found an exponential decay relationship between funnelling ratio and some shrub size metrics (Figure 5). This suggests that smaller shrubs gain access to rainfall via higher funnelling ratio than larger shrubs. Our findings thus highlight the importance of morphological metrics on stemflow yield, and in particular, small shrubs might profit from sporadic rainfall events. Small shrubs generally lack deep developed roots to draw water from deep soil layers, whereas small shrubs are more efficient at funnelling rainwater than their large counterparts. This is of eco-hydrological importance for the survival and growth of juvenile shrubs in drought conditions of arid desert ecosystems. Moreover, according to Kidron (2015), each millimetre of rain infiltrates to ∼10–11 mm in sand; it is therefore expected that long-term mean precipitation (191 mm) in our study area should result in an infiltration of ∼191–210 cm. Funnelling ratio is an important parameter for a better understanding of shrub-related infiltration (Levia & Germer 2015). In our study, shrubs had a high funnelling ratio with small shrubs possessing higher funnelling efficiencies than larger ones, it is therefore reasonable to estimate that the infiltration under shrub canopies should be greater than in bare sand soil; it is, however, difficult to precisely estimate infiltration depth due to complex root structure and the great difference between stemflow infiltration process and vertical rainfall infiltration process (Liang et al. 2009).

Not surprisingly, stemflow volume increased significantly (P < 0.01) with rainfall depth for all studied shrubs (Figure 2). For funnelling ratio, it first showed an increased tendency with rainfall depth until a threshold value around 12 mm and then began to decline (Figure 3). With increasing rainfall depth, a general increase in funnelling ratios (e.g., Figure 3(b)) can be found in the literature, whereas deviations exist as to this relationship after exceeding a certain point of rainfall depth. For co-occurring evergreen broadleaved trees (Cyclobalanopsis multinervis and Cyclobalanopsis oxyodon) and deciduous broadleaved trees (Fagus engleriana and Quercus serrata var. brevipetiolata) in a subtropical forest, Su et al. (2016) found that funnelling ratio displayed a tendency to increase with incident rainfall up to a rainfall depth of 50 mm, above which funnelling ratio remained relatively constant despite increasing rainfall. In a northern hardwood stand in southern Ontario under growing season conditions, Carlyle-Moses & Price (2006) found that stemflow funnelling ratios increased with increasing rainfall depth in a linear fashion until a peak (17.4 mm) was reached with funnelling ratios declining with greater rainfalls. Similar results as Carlyle-Moses & Price (2006) have been found for desert shrubs in China with a threshold value of 11 mm for C. korshinskii and 17 mm for Reaumuria soongorica and Tamarix ramosissima (Li et al. 2008), and for a juniper (Juniperus virginiana, redcedar) woodland and a tallgrass prairie with a threshold value of 35 mm in the south-central Great Plains of the US (Zou et al. 2015). According to Carlyle-Moses & Price (2006), a greater proportion of a tree becomes saturated with increasing rainfall input and thus the area contributing to stemflow increases until a threshold rainfall input that saturates all areas capable of producing stemflow is reached; once this threshold rainfall depth has been exceeded, funnelling ratios would be expected to decrease since the numerator in Equation (1) will be limited by the maximum contributing area of the tree, while the denominator will increase in proportion with the rainfall input.

Both stemflow volume and funnelling ratio differed considerably among individual shrubs within the same species. Greater incident rainfalls resulted in larger event stemflow volumes, while funnelling ratio seems to be less affected by rainfall depth. Smaller shrubs generate a lower quantity of stemflow, while gaining access to rainfall via higher funnelling ratio than larger shrubs, suggesting that small shrubs may have a favourable soil water condition around the shrub base for survival and growth when competing with larger ones in water-limited arid desert environments. Investigation into soil moisture variations related to the stemflow induced by shrubs of varying sizes is called for in the future to obtain direct and more convincing evidence of differential intra-specific stemflow funnelling efficiencies on the growth of shrubs.

This study was supported by the National Natural Science Foundation of China (grant nos. 41530750, 41501108, 41371101) and the CAS ‘Light of West China’ Program. The authors would like to express their gratitude to the associate editor and three anonymous reviewers for their constructive comments in improving the manuscript.