Evaluating the influential variables on rainfall interception at different rainfall amount levels in temperate forests

Highlights

• Rainfall interception (RI) was affected by rainfall amount levels (RMLs)

• RI amount was closely related to rainfall amount regardless of RMLs.

• RI proportion was most determined by humidity at heavy rain events.

• RI proportion was governed by canopy interception index at light and middle RMLs.

Abstract

Rainfall interception (RI) by forest canopy is an important part of the ecohydrology cycle, influenced by rainfall characteristics, weather conditions, forest structures and their interactions. Rainfall events with different amounts are likely to change the ways that rainfall interacts with forest canopies, and consequently vary the effects of influential variables on RI markedly. Thus, analyzing the influential variables on RI separately according to rainfall amounts need further attention. In this study, rainfall partitioning was measured in 60 plots of four forest types during June to September from 2017 to 2019. Weather conditions were obtained from four weather stations spatially representing all the study plots. Detailed canopy structural variables were retrieved from the terrestrial laser scanning. Then, boosted regression trees (BRT) analyses were performed to evaluate the contributing variables of RI at different rainfall amounts. As a result, a total of 77 rainfall events were measured, and 43, 19 and 15 events were classified as light rainfall events (LR, gross rainfall < 10 mm), middle rainfall events (MR, 10 mm ≤ gross rainfall < 25 mm) and heavy rainfall events (HR, gross rainfall ≥ 25 mm). The average values of RI amounts (the difference between gross rainfall and the sum of throughfall and stemflow) for LR, MR and LR were 2.0 mm, 3.5 mm and 6.6 mm, respectively. The average RI proportions (the proportion of RI amount to gross rainfall) for LR, MR and HR were 74.8 %, 20.2 % and 17.1 %, respectively. Both RI amount and RI proportion were significantly different among the rainfall amount levels (LR, MR and HR). BRT results showed that, for all rainfall amount levels together, RI amount was governed by rainfall amount, followed by canopy interception index (CII) and humidity. Whereas RI proportion was most influenced by CII, followed by rainfall amount, humidity and average canopy height (ACH). As expected, for a given rainfall amount level, rainfall amount was the most influential variable on RI amount. Naturally, the most influential variable on RI proportion varied with rainfall amount levels. For LR and MR, the most influential variables were CII; while it was humidity for HR. Furthermore, a variety of significant variables (with the relative influence value higher than the averaged value), such as rainfall intensity, wind speed and ACH, intricately and complexly affected RI amount and proportion at each rainfall amount level. In addition, the Rainfall amount exerted a stepwise positive correlation with RI amount. CII showed a stepwise positive and humidity a negative influence on RI proportion, respectively. A comprehensive understanding of the influence variable on RI at distinct rainfall amount levels provides valuable insights into how forest canopies intercept rainfall and greatly supports understanding of canopy hydrological processes in temperate forests.

Introduction

Rainfall interception (RI) is the rainfall retained in the canopy and then evaporates back into the atmosphere. Generally, RI varies greatly at the forest stand, with RI amount (the difference with gross rainfall and the sum of throughfall and stemflow) ranging from 1.8 mm to 6.5 mm, and RI proportion (the proportion of RI amount to gross rainfall) ranging from 10 % to 50 % in temperate forests (Barbier et al., 2009, He et al., 2014). RI has various ecological and hydrological functions, such as reducing the kinetic energy of rainfall drops (Goebes et al., 2015, Levia et al., 2017), mitigating soil erosion (Nanko et al., 2016a, Seitz et al., 2017), regulating the water storage in litter and soil layers (Nanko et al., 2010, Uhlenbrook, 2006), influencing chemical deposition (Kristensen et al., 2004, Zhu et al., 2018), total evaporation (Savenije, 2004) and global water supply (Murray, 2014). Because of its important role, it is essential to evaluate the influential variables on RI.

RI varies considerably for forest stands as a function of forest structures (e.g. tree height, basal area, surface area of leaf, branch and stem) and meteorological variables. Meteorological variables are further divided into two groups in studies (del Campo et al., 2018, Li et al., 2016a), including rainfall characteristics (e.g. rainfall amount, rainfall intensity, rainfall duration) and meteorological conditions (e.g. wind speed, temperature, humidity). However, up to now, there is no consensus on the effects of variables on RI. For example, some studies showed that RI increased with the increase of wind speed (Reid and Lewis, 2009, Staelens et al., 2008). This was because the higher wind speed raised the evaporation rate according to the Penman-Monteith equation. However, other studies suggested that higher wind speed resulted in lower RI (Iida et al., 2017; Van Stan et al., 2014). This could be explained by an assumption that wind shook the canopy, causing splash and channeling more rainfall into throughfall. Another example, the raindrop diameter was also regarded as either increasing or decreasing RI (Calder, 1996, Zabret et al., 2018). The conflicting information may be because that the three groups of influential variables are interdependent. To create a complete picture of effects of different variables on RI, it is essential to analyse these variables together.

Generally, the effect of variables on RI are different between the deciduous and coniferous tree species (Li et al., 2016a, Zabret et al., 2018), between the leafed and leafless periods (Nanko et al., 2016b, Tanaka et al., 2017, Zhang et al., 2019), between the young and old trees (Pypker et al., 2005), between general weather conditions (Zabret and Šraj, 2021), between canopy phenologies (Tanaka et al., 2015), and between the diverse tree species (del Campo et al., 2018, Li et al., 2016a, Van Stan et al., 2014). Therefore, in order to analyze the effects of variables on RI considering the species or morphologies of trees, the relationships between influential variables and RI are studied separately according to the tree characters mentioned above. Additionally, rainfall amount is reported as the most important variable that influences RI in many studies (Staelens et al., 2008, Zabret and Šraj, 2019). Rainfall events with different amounts vary the effects that rainfall interacts with forest canopy (Mali et al., 2020, Park and Cameron, 2008, Siles et al., 2010). For example, forest canopy redistributes rainfall, and rainfall simultaneously changes the inclination of leaves and branches, and consequently the changed forest canopy directly influences canopy storage capacity (Fathizadeh et al., 2017). The changed forest canopy, together with weather conditions, may also regulate the rate or amount of evaporation (Deguchi et al., 2006). Thus, rainfall events with different amounts are likely to vary the effect of influential variables on RI markedly. However, analyzing the influential variables on RI separately according to rainfall amounts has been less investigated in the context of all three-group variables.

In addition, previous RI studies involved forest stands and individual mature trees, where many structure characteristics could not be accurately measured due to the high three-dimensional spatial variability (Herwitz and Slye, 1995, Roth et al., 2007, Yu et al., 2020). To accurately investigate the relationship between RI and structural variables, the detailed structural variables are collected from shrubs (Yuan et al., 2016, Zhang et al., 2018) or small trees (Li et al., 2016a), in the field (Yuan et al., 2016) or laboratory (Li et al., 2016a). However, whether the relationships are applicable to forest stand in the field remains unclear. Terrestrial laser scanning (TLS) is a state of the art active remote-sensing technology, which can offer an opportunity to rapidly and accurately describe the three-dimensional forest structures on forest stand (Liang et al., 2016). Thus, the TLS-derived structure variables may allow a reliable relationship between RI and forest structure. For example, Yu et al. (2020) reported that the TLS-derived canopy interception index (CII) showed a stronger performance to RI than other structural variables. Meanwhile, the TLS-derived structure variables, such as LAI, PAI, CII, have not yet been fully explored in the analysis of the relationships between RI and forest canopies.

Hence, four forest types, including Korean pine (Pinus koraiensis) plantation forest (KPF), larch (Larix spp.) plantation forest (LPF), mixed broadleaved forest (MBF) and Mongolian oak (Quercus mongolica) forest (MOF) are common in Northeastern China, where water scarcity has become an increasing threat to sustainable development, and where large-scale re-forestation efforts are altering rainfall distribution and regional water budgets. With the aid of TLS technology to investigate forest structures, this study aimed to analyze rainfall characteristics, meteorological conditions and forest structures together, and to identify the most influential variables on RI among different levels of rainfall amounts. The main objectives of this study were as follows: (1) to compare RI amount (the difference of gross rainfall and the sum of throughfall and stemflow) and RI proportion (the proportion of RI amount to gross rainfall) among different rainfall amount levels; (2) to understand the influences of rainfall characteristics, meteorological conditions and forest structures on RI at all rainfall events; (3) to explore the influential variables on RI separately according to rainfall amount levels. The comprehensive combination of three groups of variables, especially the detailed structural variables extracted from the TLS, will draw more robust conclusions on influential variables on RI. The exploration of the influence variables on RI at distinct rainfall amount levels can provide insights that the effect ways of influence variables on RI may be different among rainfall amount levels.