Rainfall Partitioning in Chinese Pine ( Pinus tabuliformis Carr.) Stands at Three Di ﬀ erent Ages

: Chinese pine ( Pinus tabuliformis Carr.) is the main forest species in northern China, with the potential to dramatically a ﬀ ect biotic and abiotic aspects of ecosystems in this region. To discover the rainfall partitioning patterns of di ﬀ erent growth periods of Chinese pine forest, we studied the throughfall ( Tf ), stemﬂow ( Sf ) and canopy interception ( I ) in three stand ages (40-, 50-, 60-year-old) in Liaoheyuan Natural Reserve of Hebei Province during the growing seasons of 2013 and 2014, and analyzed e ﬀ ect of rainfall amount, rainfall intensity, and canopy structure on rainfall partitioning in Chinese pine forest. The results showed that throughfall decreased with the stand age, accounting for 78.8%, 74.1% and 66.7% of gross rainfall in 40-, 50- and 60-year-old Chinese pine forests, respectively. Canopy interception, on the other hand, increased with the stand age (20.4%, 24.8%, and 32.8%, respectively), while the pattern in stemﬂow was less clear (0.8%, 1.1%, and 0.6%, respectively). As rainfall intensity increased, the Tf and Sf increased and I declined. Additionally, our results showed that leaf area index (LAI) and the diameter at breast height (DBH) increased with age in Chinese pine stands, probably explaining the similar increase in canopy interception ( I ). On the other hand, the mean leaf angle, openness, gap fraction all decreased with the stand age. Stepwise regression analysis showed that the rainfall amount and LAI were the major determinants inﬂuencing the rainfall partition. Our study highlights the importance of stand age in shaping di ﬀ erent forest canopy structures, and shows how age-related factors inﬂuence canopy rainfall partitioning. This study also signiﬁcantly adds to our understanding the mechanisms of the hydrological cycle in coniferous forest ecosystems in northern China.


Introduction
Forests can influence hydrological processes and alter soil conditions due to their many impacts on water purification, runoff regulation and water and soil conservation [1]. In this capacity, forests can also serve to improve water availability at watershed, regional and worldwide levels [2]. The forest canopy plays an important role in this process, by partitioning rainfall and affecting additional hydrological processes [3], not only changing the spatial distribution of rainfall in the forest [4,5], but also affecting the chemical composition of the falling rain [6,7]. Previous studies of forest stands indicated that forest rainfall interception led to heterogeneous water distribution [8][9][10] and showed a large range of spatial variation relating to different forest structures [11].
Rainfall is redistributed through a forest canopy by throughfall, stemflow and canopy interception. Throughfall is the total volume of raindrops that penetrate through the forest canopy and reach to the ground. Stemflow is the part of rainfall flowing down from stems or trunks, and the amount intercept by the vegetation canopy and evaporate is the interception loss [12]. Throughfall accounts for the of a year is 7.7 • C and the mean precipitation of a year is 516. 9 mm (1980-2018), according to data from Pingquan County Meteorological Bureau, with most rainfalls occurring from May to August. The soil is a typical brown forest soil, 0-100 cm thick [30]. The dominant tree species is the evergreen conifer P. tabuliformis, or Chinese pine, which is widely distributed in north China. Spiraea pubescens and Lespedeza bicolor are the main shrub species. Carex rigescens, Saussurea nivea, and Dianthus chinensis are major herbaceous plant species.
Forests 2020, 11, x FOR PEER REVIEW 3 of 15 of a year is 7.7 °C and the mean precipitation of a year is 516. 9 mm (1980-2018), according to data from Pingquan County Meteorological Bureau, with most rainfalls occurring from May to August. The soil is a typical brown forest soil, 0-100 cm thick [30]. The dominant tree species is the evergreen conifer P. tabuliformis, or Chinese pine, which is widely distributed in north China. Spiraea pubescens and Lespedeza bicolor are the main shrub species. Carex rigescens, Saussurea nivea, and Dianthus chinensis are major herbaceous plant species. . We got the plot basic information on vegetation and site characteristic data by field investigation. Species name, height, crown breadth, diameter and density of trees with DBH ≥ 5 cm were recorded for each plot. Nine digital hemispherical photographs (WinScanopy 2010a) were taken for each plot to get LAI and other structure information on overcast days or near sunrise or sunset. The camera was adjusted to horizontal position at 1.3 m above ground [31]. Photographs were analyzed using WinScanopy software for calculating LAI, the gap fraction, the openness, and the mean leaf angle [26]. Our study was conducted based on the National Standards of the People's Republic of China or "Methodology for field long-term observation of forest ecosystem research" (GB/T 33027-2016). Data for each site are shown in Table 1. . We got the plot basic information on vegetation and site characteristic data by field investigation. Species name, height, crown breadth, diameter and density of trees with DBH ≥ 5 cm were recorded for each plot. Nine digital hemispherical photographs (WinScanopy 2010a) were taken for each plot to get LAI and other structure information on overcast days or near sunrise or sunset. The camera was adjusted to horizontal position at 1.3 m above ground [31]. Photographs were analyzed using WinScanopy software for calculating LAI, the gap fraction, the openness, and the mean leaf angle [26]. Our study was conducted based on the National Standards of the People's Republic of China or "Methodology for field long-term observation of forest ecosystem research" (GB/T 33027-2016). Data for each site are shown in Table 1.

Collection of P, Tf, and Sf
Precipitation (P), throughfall (Tf ) and stemflow (Sf ) were collected from May to September of 2013 and 2014. Precipitation was recorded by an automated tipping-bucket gauge (resolution 0.2 mm) at a meteorological station (CR1000, Compbell Scientific Inc., Logan, UT, USA), installed in an open hillside 1000 m away from the plots (Figure 2a). A single rainfall event was classified as rain exceeding 0.5 mm during a period with more than 6 h from the last rainfall event [32]. Rainfall events were divided into 4 intensity classes, including light rain (0.2 mm/d ≤ P < 10 mm), middle rain (10 mm/d ≤ P < 25 mm), heavy rain (25 mm/d ≤ P < 50 mm), rain storm (50 mm/d ≤ P < 100 mm), according to Criteria for Classification of Rainfall Intensity (Inland) produced by China's Meteorological Administration.

Collection of P, Tf, and Sf
Precipitation (P), throughfall (Tf) and stemflow (Sf) were collected from May to September of 2013 and 2014. Precipitation was recorded by an automated tipping-bucket gauge (resolution 0.2 mm) at a meteorological station (CR1000, Compbell Scientific Inc., Logan,UT USA), installed in an open hillside 1000 m away from the plots (Figure 2a). A single rainfall event was classified as rain exceeding 0.5 mm during a period with more than 6 h from the last rainfall event [32]. Rainfall events were divided into 4 intensity classes, including light rain (0.2 mm/d ≤ P < 10 mm), middle rain (10 mm/d ≤ P < 25 mm), heavy rain (25 mm/d ≤ P < 50 mm), rain storm (50 mm/d ≤ P < 100 mm), according to Criteria for Classification of Rainfall Intensity (Inland) produced by China's Meteorological Administration. Throughfall (Tf) was monitored manually by using 3 homemade plastic rain gauges (200 × 10 cm) randomly placed in each plot, which were set on a bracket 50 cm above the ground (Figure 2b). Each gauge kept an angle of 5° above ground to make drainage easily, the end of the gauge connected to a plastic tank (16 L), which was in a hole below the ground covered with plastic sheeting to avoid evaporation and splash. The volume of water in the tank was measured soon after each rainfall event by using a graduated cylinder with the resolution of 1 mL. The throughfall depth was calculated by the rain volume divided by the area of the gauge. The average of the three throughfall depths was adopted as the throughfall amount for each plot.  Throughfall (Tf ) was monitored manually by using 3 homemade plastic rain gauges (200 × 10 cm) randomly placed in each plot, which were set on a bracket 50 cm above the ground (Figure 2b). Each gauge kept an angle of 5 • above ground to make drainage easily, the end of the gauge connected to a plastic tank (16 L), which was in a hole below the ground covered with plastic sheeting to avoid evaporation and splash. The volume of water in the tank was measured soon after each rainfall event by using a graduated cylinder with the resolution of 1 mL. The throughfall depth was calculated by the rain volume divided by the area of the gauge. The average of the three throughfall depths was adopted as the throughfall amount for each plot.
The stemflow (Sf ) was collected by using a 5 mm thick white halved rubber hose, which was spirally wrapped around tree trunk and used Vaseline to create a water-tight environment between the stem and the hose. The lower side of the hose was 30 cm above the ground and connected to a plastic bucket (16 L), after each rainfall event the rain volume in the bucket was measured ( Figure 2c). Trees for stemflow experiments were selected according to the diameter at breast height (DBH) within the plot. In each plot all the trees were divided into five groups by DBH. DBH groups for three stands were divided as follows: the 40-year-old stand ( . Then one tree was chosen in each group as the standard tree to monitor stemflow for each plot in the studied forest. A total of five trees in each plot were selected as standard trees, information of which was shown in Table 2.
The depth of the stemflow is calculated by the following equation [14].
where Sf is the stemflow amount (mm), n is the number of stem groups in this study; Ci is the stemflow amount of monitored tree in ith group (mL); Mi is the number of trees in the ith group (n); S is the plot area (600 m 2 ). If total precipitation (P), throughfall (Tf ), and stemflow (Sf ) are known, canopy interception amount (I) can be measured according to the following equation [33]: where I is interception amount (mm), P is the precipitation (mm), Tf is the throughfall amount (mm), Sf is the stemflow amount (mm).
Canopy interception ratio (R) is then calculated as: where R is interception ratio (%), P is the precipitation (mm), I is the interception amount (mm).

Data Analysis
Throughfall depth under each rainfall event for each stand was average amount of the three gauges. We used cumulative data during the study period to reflect the overall rainfall partitioning pattern, and used mean values to distinguish the difference between stand ages and months.
All data preparations including the total precipitation, amount and ratio data were done by Microsoft Excel (2010 version, Microsoft Corporation, Redmond, Washington, USA). A one-way analysis of variance (ANOVA) by SPSS (IBM SPSS Statistics 19 version, and ratio data were done by Microsoft Excel (2010 version) was used to compare the statistical difference between different Chinese pine stand ages. Stepwise multiple linear regression was conducted with SPSS to find out significant factors for Chinese pine rainfall partitioning. According to the classification of the "precipitation intensity classification standard (inland part)" of China's Meteorological Administration, the precipitation events in the study area during the observation period were most often light rain, with 40 times total and rainfall amounting to 194.4 mm, accounting for 22.7% of total rainfall in growing seasons of 2013 and 2014 (Table 3). Increasingly intense rainfall events were less frequent. Rain storms occurred just twice, both in 2013. The gross rainfall amount of heavy rain was the largest, followed by middle rain, light rain, and rain storm.

Rainfall Partitioning Pattern Across Chinese Pine Stand Age
During the two years of the observation period, the accumulated rainfall division in (mm) and (%) showed different results among the three forest stands (Table 4). In each plot, Tf corresponded to the highest proportion of the total rainfall, while Sf was always the lowest proportion. Tf showed an expected decrease from 40 to 60 years. Stemflow (Sf ), on the other hand, did not show an obvious trend, since the 50-year-old stand resulted in the highest value among the three stands. Total interception (I), on the other hand, behaved similarly to Tf. An additional analysis, based on 64 sampled rainfall events (mm), showed I was the only variable that proved to have a significant statistical difference among the three forest stands, being the 60 years plot significantly greater than that of 40 and 50 years. Monthly rainfall distribution patterns for two years ( Figure 3) showed that throughfall, stemflow and interception amount of three stand ages in June and August were larger than the other three months. The Tf, Sf, and I of three stands presented the same trend across five months and the trend among three stands varied. Throughfall in each month decreased as the stand age increases, whereas interception increased with the stand age. The stemflow was largest in the 50-year-old stand in each month.

Dependence of Rainfall Partitioning on Rainfall Amount
The relationships between rainfall amount and throughfall, stemflow and canopy interception were examined (Figure 4), with the best fitting model and parameters of the model listed in Table 5. Throughfall and stemflow were positively related to the rainfall amount. According to the simulated model, throughfall would generate in the 40-year-old stand when rainfall amount reached more than 1.5 mm, and that for the 50-year-old and the 60-year-old stand was 1.9 and 1.86 mm, respectively. The stemflow occurred when rainfall was more than 4.9 mm for the 40-year-old stand, and the corresponding value for the 50-year-old, and the 60-year-old stand was 4.0 and 5.9 mm, respectively. That meant for 40-year-old stand, in general, the canopy was wetting under rainfall less than 1.5 mm. Forest canopy was saturated and could not hold more water when the rainfall reached 1.5 mm, after that throughfall occurred, and when the rainfall more than 4.9 mm, the water flowed down from the stem namely stemflow. This progress was basically consistent with monitored data. The interception and rainfall had binominal function relationship, with coefficient of determination > 0.67. Interception amount increased with the rainfall amount at first, after that reached to a maximum value then declined as shown in figures.

Dependence of Rainfall Partitioning on Rainfall Amount
The relationships between rainfall amount and throughfall, stemflow and canopy interception were examined (Figure 4), with the best fitting model and parameters of the model listed in Table 5. Throughfall and stemflow were positively related to the rainfall amount. According to the simulated model, throughfall would generate in the 40-year-old stand when rainfall amount reached more than 1.5 mm, and that for the 50-year-old and the 60-year-old stand was 1.9 and 1.86 mm, respectively. The stemflow occurred when rainfall was more than 4.9 mm for the 40-year-old stand, and the corresponding value for the 50-year-old, and the 60-year-old stand was 4.0 and 5.9 mm, respectively. That meant for 40-year-old stand, in general, the canopy was wetting under rainfall less than 1.5 mm. Forest canopy was saturated and could not hold more water when the rainfall reached 1.5 mm, after that throughfall occurred, and when the rainfall more than 4.9 mm, the water flowed down from the stem namely stemflow. This progress was basically consistent with monitored data. The interception and rainfall had binominal function relationship, with coefficient of determination > 0.67. Interception amount increased with the rainfall amount at first, after that reached to a maximum value then declined as shown in figures.

Effect of Rainfall Intensity on Rainfall Partitioning
Relative throughfall, stemflow and interception to gross rainfall varied with the rainfall intensity in three Chinese pine stand ages ( Figure 5). On the whole, as the rainfall intensity increased, the average Tf % and Sf % for all three stands increased, but I% decreased. The largest interception ratio was under the light rain level for all three stands. Except for rainfall storm class I% of the 60-year-old stand was dominantly larger than the 40-year-old stand, and Tf % was opposite.

Effect of Rainfall Intensity on Rainfall Partitioning
Relative throughfall, stemflow and interception to gross rainfall varied with the rainfall intensity in three Chinese pine stand ages ( Figure 5). On the whole, as the rainfall intensity increased, the average Tf% and Sf% for all three stands increased, but I% decreased. The largest interception ratio was under the light rain level for all three stands. Except for rainfall storm class I% of the 60-year-old stand was dominantly larger than the 40-year-old stand, and Tf% was opposite. Figure 5. Relative throughfall, stemflow, and interception categorized by rainfall classes for three Chinese pine stand ages. Different letters meant the significant difference (p < 0.05) among the three stand ages, and the alphabetical order showed the values from the largest to the smallest.

Effect of Canopy Features on Rainfall Partitioning
The rainfall redistribution of the three stands was also affected by the canopy structure ( Table  1). As the age of Chinese pine forests increased, the DBH, the tree dimension and the LAI increased, whereas the forest gap fraction, openness and the mean leaf angle decreased. The canopy interception ability of the 60-year-old stand is higher than that of 40-and 50-year-old stand, probably caused by the discrepancy of above mentioned structural characteristics. This result demonstrated that the stand structure in different growing stage of the Chinese pine forest was an important factor for rainfall partitioning.

Comprehensive Analysis of Factors
A stepwise regression analysis was conducted to understand key factors influencing the rainfall redistribution pattern. We used the measured factors including rainfall amount, rainfall intensity, LAI, density, average tree height, DBH, dimension, and mean leaf angle as independent variables, and used Tf, Sf and I as dependent variables, respectively. Results (Table 6) showed rainfall amount and structural properties (LAI) were dominant factors for Tf and I, while rainfall amount, density and mean leaf angle were dominant factors for Sf.

Effect of Canopy Features on Rainfall Partitioning
The rainfall redistribution of the three stands was also affected by the canopy structure ( Table 1). As the age of Chinese pine forests increased, the DBH, the tree dimension and the LAI increased, whereas the forest gap fraction, openness and the mean leaf angle decreased. The canopy interception ability of the 60-year-old stand is higher than that of 40-and 50-year-old stand, probably caused by the discrepancy of above mentioned structural characteristics. This result demonstrated that the stand structure in different growing stage of the Chinese pine forest was an important factor for rainfall partitioning.

Comprehensive Analysis of Factors
A stepwise regression analysis was conducted to understand key factors influencing the rainfall redistribution pattern. We used the measured factors including rainfall amount, rainfall intensity, LAI, density, average tree height, DBH, dimension, and mean leaf angle as independent variables, and used Tf, Sf and I as dependent variables, respectively. Results (Table 6) showed rainfall amount and structural properties (LAI) were dominant factors for Tf and I, while rainfall amount, density and mean leaf angle were dominant factors for Sf.

Age Dependence of Rainfall Partitioning
Our results showed a rainfall partitioning pattern in Chinese pine stand ages (40, 50 and 60-year-old). We found that throughfall decreased while the canopy interception increased with increasing age, this pattern was in accordance with conifer species rainfall partitioning trend from young to adult in temperate boreal forests [13] and with deciduous forest A. altissima [31]. The study on a series of succession stands of Liriodendron tulipifera forest in the southern Appalachian Mountains also found that canopy interception increased rapidly with the forest age until reaching a maximum of 21% [27]. Another study on different Chinese fir stand ages got the similar trend that interception ratio in mature stands significantly higher [29]. There seems to be a clear pattern, therefore, that in mature forests of many species the canopy interception ability is higher than that in young forests.
Throughfall, stemflow and interception ratio varied among the three stand ages of Chinese pine forests, all of which were within the range of results of Sun et al. reported [14]. The throughfall in our study was higher than that of A. altissima [31], although lower than redcedar in the USA with an average annual throughfall of 57.3% [8]. We suspect that the discrepancy among different plants is due to differences in climate, tree species characteristics and stand structure.
Compared with previous studies of Chinese pine forest on rainfall partitioning (Table 7), including reports at Miyun [34], at Daqingshan [35], at Hebei [36], at Taiyue Mountain [37] and Yeheshan [38]. I% in this study were higher than Taiyue Mountain reflected the young stand age had less canopy interception ability. Throughfall, stemflow and interception ratio of 33 years old Chinese pine forest in Miyun was similar to corresponding values for the 60-year-old stand in this paper, and Yeheshan result was similar to the 40-year-old forest in our study.
Compared with other coniferous forests worldwide (Table 7), the interception ability in our result was higher than that in mixed forests [39] and P. cembroides forest [9], lower than P. eldarica [40], Pinus nigra in Slovenia [41] and Larix gmelinii forest [42], and similar with other reports listed in the table [43,44]. Throughfall in this paper was higher than P. nigra and Chamaecyparis obtuse forest, but smaller than P. cembroides and mixed (upland) forest, within results range of other reports. The stemflow in the 50-year-old stand was larger than in the other two stand ages, though stemflow accounted for a small part of the rainfall (<2%) in all stands. This low percentage is consistent with the results of Chinese fir plantations that stemflow was the highest in the middle age category (12 years old forest). Stemflow of Chinese pine forest in our study was lower than most of the reports listed in the table, much lower than C. obtuse (22.6%), and higher than L. gmelinii and P. nigra forest. The low stemflow in this study may due to climate and also the accuracy of the experimental device of the stemflow in this study. Previous research confirmed that stemflow was the most variable part of the canopy partitioning (coefficient of variation = 107.8%) [17]. Furthermore, it is known that shrubs generate more stemflow than trees, which has been demonstrated in our results comparing with the results of stemflow in shrub species [45,46].

Effect of Rainfall Characteristic on Rainfall Partitioning
Rainfall directly influences the rainfall redistribution pattern, in redcedar forest rainfall amount accounts for 60-70% of the variation of the percentage of throughfall and stemflow [8]. In our study rainfall amount is also a dominant factor. According to Table 5, linear regression equations between the rainfall amount (x) and throughfall amount (y) showed positive coefficients of determination (R 2 ) > 0.97, whereas for stemflow coefficients of determination were R 2 > 0.79. This result was consistent with previous studies [47,48]. In a single rainfall event canopy gradually becomes moist, until rainfall amount reaches a threshold above which throughfall and stemflow are generated. Binominal function model fitted in this study accurately capture the relationship between the canopy interception and rainfall amount, which was different from previous study result that either the power [19] or the linear [49] relationship. The interception amount increased with rainfall until it reached plateaus. Some previous studies showed different best fit models probably due to the experiment designs and the vegetation features.
Rainfall intensity is also an important factor for the rainfall partitioning pattern. On the whole our results showed that throughfall and stemflow increased with rainfall intensity, while the canopy interception decreased, which meant that rainfall intensity could make more net rainfall under the canopy. The larger interception under light rainfall events in this study demonstrated that forest canopy had a strong interception effect under light rainfall events. Some reports reveal that rainfall partitioning was most influenced by rainfall intensity. For example, Nytch et al. [16] reported that rainfall intensity had a significant effect on delay in Tf reaching the ground and Li et al. found that the maximum and minimum interception storage were only correlated with rainfall intensity in Chinese pine forest [50]. However, Liang et al. found that stemflow generation was mainly affected by canopy structure not by rainfall intensity [51], and Li et al. [46] showed the rainfall intensity increased stemflow with which less than 2 mmh −1 and decreased with which greater than 2 mmh −1 . In light of this variation among studies, more research should be done to understand the impact of rainfall intensity on rainfall redistribution.

Forest Structure Dependence of Rainfall Partitioning
Our results showed that the rainfall partitioning pattern in stands with three different stand ages was correlated with the canopy structure. Under the same rainfall events, the variation of throughfall, stemflow and interception were caused by the difference in the crown structure at different growth stage. The LAI of the 60-year-old stand were larger than the 40-year-old stand, meaning that older trees had larger crowns than younger trees, and are therefore able to intercept more water. Previous studies also analyzed forest structural parameters including basal area, bark sickness, leaf area index across the forest, which can contribute to the rainfall redistribution [52][53][54]. For example, larger canopy surface increases interception ratio, and a denser mid-canopy and taller trees can increase cloud interception [1]. Some studies have found that forest structure and meteorological events importantly influence interception ratios, especially in drier sites [23]. Structural properties including LAI and aggregation index were correlated with canopy interception [14]. Stand density is an informative structural factor for all the rainfall partitioning components in coniferous plantations [22]. Smooth barks in some species perform higher stemflow than with rough bark [13]. Significant changes in rainfall partitioning across an invasive chronosequence of A. altissima were underpinned by canopy structure and trunk parameters [31]. Fang et al. deemed canopy structure as a key element for spatial variation of throughfall in small rainfall events [55]. Our results also revealed that forest structure was an important factor influencing rainfall redistribution as shown in previous studies. Through this study, we show that stand age is a determining factor, causing the divergence of canopy structures that lead to differences in rainfall distribution patterns. Further work should be done to discover the rainfall partitioning laws in a large range of forest ages. Above all, both the rainfall amount and stand structure significantly influenced rainfall partitioning. The last point, we simulated the relationship between the rainfall amount with Tf, Sf and I, and calculated the tipping points, this is very useful and convenient. But advanced methods should be used like rainfall interception modeling, which includes more intercept progresses and can give deep insight to interception mechanisms [56]. Liang et al. concluded that the revised gash model was able to accurately simulate the weekly canopy interception of Chinese pine forest [36]. The results of canopy interception of Chinese pine from model [36,38] were compared with our results, a similar pattern was found, but more useful information from the models that should be our future work.

Conclusions
Our study monitored rainfall partitioning pattern of Chinese pine forest at three stand ages and provided deep insights to understand the rainfall interception dynamics. Specifically, we analyzed the dynamics of three rainfall partitioning aspects (throughfall, stemflow, and interception) reflecting the relationship with rainfall amount and intensity, which helps to better understand the progress and relevant facets. Across three stand ages, the majority of rainfall became throughfall, though this decreased with increasing stand age. We attribute this decrease in throughfall to an increase in interception as canopy structure changes with forest age. Our results also highlighted that the rainfall amount is also an important factor affecting the rainfall partitioning across different stand age of Chinese pine. This study provides insight into how rainfall amount, stand age and elements of forest structure affect rainfall partitioning, which has important implications for water conservation and forest management.
Author Contributions: H.H. and F.K. designed the experiments; J.Z., X.S. investigated the field data; L.D. analyzed and wrote the paper; X.C. revised the paper and modified the language. All authors have read and agreed to the published version of the manuscript.