A study on crown interception with four dominant tree species: a direct measurement

Rainfall interception refers to the proportion of rainwater that is temporarily intercepted and stored on vegetation surfaces, and subsequently evaporates to the atmosphere or drains to the ground as throughfall (TF) or stemflow (SF) (Muzylo et al. 2009). Acting as a rainfall mitigating buffer, interception usually accounts for 10–60% gross precipitation (Llorens & Domingo 2007; Gerrits et al. 2010), and also is widely accepted to affect the subsequent hydrologic processes, such as evaporation, transpiration, infiltration, and surface runoff (Savenije 2004; Janeau et al. 2015).

Interception has usually been studied on three scales: the single branch level (with foliage) (Keim et al. 2006; Xiao & McPherson 2015), the individual tree level (Asadian & Weiler 2009; Xiao & McPherson 2011), and the forest stand level (Dietz et al. 2006; Safeeq & Fares 2014). In most of these studies, interception has been indirectly calculated by the difference between gross rainfall and the sum of TF and SF, indicating that only the intercepted water for evaporation has been estimated as a flux or stock. The mechanism of the interception process which includes wetting, saturation, and post-rainfall drainage has thus far been overlooked. However, a process-based interception is critical to clarify the rainfall-mitigating and time-lagging effect during rainfall, and thereby helps to control peak flow, flood, and erosion (Keim & Skaugset 2004). Limited results from previous studies indicated that almost 30% of intercepted rainwater tended to drip off after rainfall (Aston 1979; Pitman 1989; Keim et al. 2006). As a result, interception would be better characterized as a dynamic process.

According to previous investigations, interception is governed by both biotic and abiotic factors (Xiao et al. 2000a; Gerrits & Savenije 2011). Abiotic factors such as rainfall intensity shapes the way raindrops interact with leaf and branch surfaces: Price & Carlyle-Moses (2003) and Wang et al. (2007) reported that high rainfall intensity reduced interception amount because it caused splashing and shaking of the crown. Keim et al. (2006) and Xiao & McPherson (2015) found the opposite: interception increased with rainfall intensity given that the leaves were beyond saturated. In addition, rainfall frequency as well as air temperature, wind speed, and net radiation impacts evaporation and interception (Xiao et al. 2000b). On the other hand, biotic factors such as leaf morphology, roughness, inclination, and hydrophobicity tended to affect interception by modifying the intercepted-rainwater pathways inside the crown (Holder 2013; Nanko et al. 2013). Other factors that played a role included branch inclination, surface property, and bark texture (Herwitz 1987; Levia et al. 2015). Crown gap fraction controls interception as it represents the proportion of free TF that has no contact with the crown (Rutter et al. 1971). Among the biotic factors, leaf area index (LAI) and leaf area (LA) have been extensively used in previous analyses (Carlyle-Moses & Gash 2011). LAI, calculated as the total LA divided by crown projection area (CPA) (Watson 1947), has been measured indirectly in forest stands based on the transmission of light mechanism (Fleischbein et al. 2005). However, the overlapping canopy from different trees may have resulted in an underestimation of LAI (Breda 2003). On the other hand, few studies have quantified interception on an LA basis due to the labor-intensive and destructive measurement, although it has been recognized as an accurate indicator, and easy to scale up from single trees to forest stand level in several studies (van Dijk & Bruijnzeel 2001a, 2001b).

Therefore, the objective of this research is to quantify interception on a process basis. First, we aim to depict the wetting–saturation–drainage phases in interception. Second, we examine the influence of different biotic (LAI, LA, and leaf type) and abiotic factors (rainfall intensity) on interception.

Four common tree species in Northern China including two conifer species (Platycladus orientalis and Pinus tabulaeformis) and two broadleaf deciduous species (Quercus variabilis and Acer truncatum) were selected for this study (China Forest Editorial Committee 2001). Two four-year-old trees for each species were chosen from the field in Jiufeng National Forestry Park, Beijing, China (116°28′E, 39°34′N). The trees were then transplanted into a 2-m-diameter plastic cylinder and were transported from the field to a rainfall simulation laboratory, where several tree characteristics were measured before rainfall simulation began (Table 1). LAI was measured using an LAI-2200 Plant Canopy Analyzer (LICOR Inc., Lincoln, NE, USA). CPA was determined from photos taken from a height of 15 m above the tree, and was further analyzed using Photoshop software (Adobe System Inc., San Jose, CA, USA). No wilting phenomenon was observed before or during the experimental procedures.

The infrequency and unpredictability of natural rainfall makes it difficult for us to study interception on a process basis. To address this challenge, a rainfall simulator was used. Over the past few decades, a wide variety of rainfall simulators have been used to conveniently and rapidly obtain diverse rainfall intensities and interception rates (Moore et al. 1983). In general, at least three criteria should be taken into consideration when designing a simulator (Wilson et al. 2014): (1) the simulated rainfall should have physical characteristics similar to natural rainfall with respect to raindrop terminal velocity and kinetic energy; (2) the simulator should produce various rainfall intensities, and can be easily moved and operated to fit the study area; and (3) the cost must be low. Based on these principles, two types of rainfall simulators have been developed: drip tank and spray nozzles. The drip tank simulator with hypodermic needles usually operates at low pressure and thereby generates light rainfall with small raindrop diameter. The raindrop kinetic energy is mainly determined by the hanging height of the simulator (Battany & Grismer 2000). In contrast, simulators with spray nozzles, working under high pressure, can provide a wide range of raindrop diameters. Also, most of the simulated raindrops can reach 90% of the kinetic energy of natural rainfall. As a result, spray nozzle simulators have been extensively used in many studies (Fister et al. 2012).

The spray nozzle simulator used in this study was jointly manufactured in 2006 by Beijing Jiaotong University and Beijing Normal University. A more detailed description is presented by Zhang et al. (2007). The simulator is composed of a water supply system with a pump (YCGT90W, Chuanji Co. Ltd, Shanghai, China), a rainfall output system including a Veejet 80150 sprinkler with three rotation nozzles (Spraying Systems Co., Wheaton, IL, USA) and a computer-control system. The simulator is able to spray water over an area of 2.2 m × 1.5 m at a height of 4.5 m. A wide range of rainfall intensities (20 to 200 mm h⁻¹) can be simulated by a combination of different nozzle sizes and swing frequencies. Simulated raindrops have a 2.3 ± 0.3 diameter and can reach terminal velocity. The simulated rainfall uniformity is about 80%, which is in accordance with traits of natural rainfall. Since rainfall intensities over the past 50 years in Beijing have ranged from 2.6 to 80 mm h⁻¹ (Zhong et al. 2013), a range of low to moderate simulated rainfall intensities of 11.7, 25.2, 48.4, and 78.5 mm h⁻¹ were used in the experiment for 40 minutes.

To fully depict the interception process, two different interception storages were measured: the maximum interception storage (C_max) and the minimum interception storage (C_min). C_max is the amount of rainwater intercepted and detained by the tree crown immediately before rainfall cessation; C_min, taken as the amount of rainwater detained on the crown when drainage ceases after rainfall, can only be removed by evaporation (Li et al. 2013).

After the rainfall began, tree weight was recorded every 10 seconds in the first minute, then every 30 seconds in the next 9 minutes and every minute in the remaining 30 minutes. C_max was thereby calculated as the amount of intercepted water when simulated rainfall ceased. After the cessation, change in tree weight due to the leaf and branch drip was recorded every 10 seconds for the first minute, 30 seconds for the next minute, and each minute thereafter until a weight change of <0.1 g was obtained. Consequently, C_min was calculated as the difference in tree weight before the rainfall simulation and after the drainage process. After these primary interception experiments, tree crown was divided into three vertical layers from the bottom to the top. The leaves and branches of the bottom layer were initially removed and preserved for LA measurement. Similarly, tree traits such as crown depth, crown diameter, LAI, and CPA were remeasured. Afterwards, each tree was repositioned on the electronic weighing balance to be resubjected to the simulated rainfall. Likewise, rainfall simulation was repeated after the middle layer was removed. Total LA of broadleaves was determined by tracing the outline of every leaf, and then calculating the area using Adobe Photoshop software (Adobe System Incorporated). For needles, LA was calculated by measuring needle length and diameter to determine the two-dimensional area, and then scaled to column, three-dimensional, and four-dimensional cones based on needle shape.

It should be mentioned that evaporation (E) was not taken into consideration in the simulation process because the amount of wind and sunlight radiation in the laboratory was considered negligible. To minimize the impact of E on the drainage process, some water was poured on the wall of the laboratory to keep a relatively stable humidity. All the examined trees were moved outside the laboratory to naturally dry the crown for 24 hours. Each test run was repeated once daily, resulting in a total of 192 rainfall events.

Pearson's correlation method was used to test whether C_max and C_min were significantly correlated with rainfall intensity, LAI, and LA. Parameters were considered to be significantly correlated when they were at or above the 95% confidence level (p ≤ 0.05). Meanwhile, linear, polynomial, and non-linear regressions were used to analyze the variation trend between C_max and C_min and those factors. One-way analysis of variance (ANOVA) with the Fisher LSD (least significant difference) test at p ≤ 0.05 was used to determine if there were significant differences in C_max, C_min between the broadleaf and coniferous species. Related data were grouped according to leaf morphology (e.g., broadleaf species were represented by PO and PT, while coniferous species were represented by QV and AT). All statistical analyses were performed using IBM SPSS Statistics 20.0 software.

Contradicting relationships between RI and interception storages were reported in previous studies. Price & Carlyle-Moses (2003) found that interception storage capacity when RI < 7 mm h⁻¹ was larger than when RI > 7 mm h⁻¹. Calder (1996b) drew a similar conclusion when comparing the storage capacity at RIs of 36 and 45 mm h⁻¹. However, several researchers found that interception storage capacity was larger at higher RIs than lower RIs (e.g., Aston 1979; Humbert & Najjar 1992; Keim et al. 2006). These outcomes also indicated that additional light rainfall experiments should be conducted to determine the threshold RI in the interception process.

Beyond the role of leaves in intercepting rainfall, studies have recognized the importance of the role of branch and bark in interception and evaporation (Keim et al. 2006), particularly in leafless periods (Herbst et al. 2008). For instance, an index including both leaf and branch area, such as woody area index and plant area index, has been introduced in several studies (Livesley et al. 2014; Safeeq & Fares 2014). These expanded indices could provide stronger correlations with interception than we have observed with LA. Along those lines, it is likely that there is a positive relationship between interception rates and crown height (CH). This is because the number of overlapping leaves and branches will increase with increasing CH, which forms a multi-layer interception effect, and time delay. CH is an easily measurable metric that deserves further investigation.

Leaf type also played a relevant role in determining interception. The needle crown of P. tabulaeformis had the highest mean C_max and C_min on both a CPA and LA basis due to its dense clusters (Table 3). This resulted in raindrops that were intercepted multiple times by the overlapped branches and needles and thereby gradually accumulated in the crown. By comparison, Q. variabilis with its broadleaf crown showed the lowest mean C_max and C_min because its crown was relatively open and rainwater could pass through directly. Additionally, the smooth and hairless leaves likely failed to prevent rainwater from dripping. Generally, average C_max and C_min for the needle species were 0.71 (STD = 0.18) and 0.28 (STD = 0.10) mm on a CPA basis, which was 2.96 and 1.75 times larger than that for broadleaf species regardless of rainfall intensities. Moreover, ANOVA analysis suggested that there were significant differences (p = 0.000 and 0.000) in C_max and C_min between the two types. These results were in line with Barbier et al. (2009), who in reviewing 28 articles found that needle species intercepted more rainwater than broadleaf species. Therefore, the proportion of broadleaf and needle trees (i.e., LAI ratio) in mixed forests should be taken into consideration to precisely estimate interception.

The present study highlighted the interception process to address some critical challenges that have seldom been investigated. The first challenge was how to depict the rainfall interception process. Generally, the complete interception consisted of three phases, the rapid-dampening phase (0–10 minutes, 90% of C_max was reached), the stable-saturated phase (10–40 minutes, no evident I_C fluctuation), and the draining phase (40–70 minutes, 45% of C_max dripped). Almost 90% of the C_max amount was intercepted in the rapid-dampening phase. In the subsequent stable-saturated phase, the interception storage showed no evident fluctuation as the crown was gradually wetted and saturated. Approximately 45% of C_max drained off in the draining phase due to gravity. The second challenge was to evaluate the effect of rainfall characteristics and tree traits on interception. The major findings showed that LAI and LA were relevant indicators for predicting interception as they were found not only to be significantly correlated with C_max and C_min, but were adequately explained using polynomial and linear relationships. By comparison, rainfall intensity was not significantly correlated with C_max and C_min, but the co-variation trends implied that there should be a threshold rainfall intensity to influence interception. Last, needle crowns with dense branch and needle foliage distribution intercepted more rainfall water than the relatively open broadleaf crowns. Further studies are warranted to accurately quantify interception on a process basis, and establish related models with more easily measurable parameters such as CH, branch number, length, density, and inclination.

We thank the working staff from Beijing Normal University for their careful inspection on the rainfall simulator and the help of postgraduates Changjiang Xie, Yong Liu, Jingping Tan, Xiaoli Wu, and Yinghu Zhang. We also deeply thank the editor and two anonymous reviewers for their thoughtful recommendations on this manuscript. This study was supported by Fundamental Research Funds for the Central Universities (No. TD2011-03, No. BLYJ201406); Chinese Scholarship Council Fund; National Natural Science Fund of China (41171028); National Advanced Project of the 12th Plan, China (2011BAD38B05); and National Forestry Public Welfare Industry Research Project (201104005). All the funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript.