Abstract

Recently, the role of urban trees in stormwater management has received increasing interest. The interception of rainfall by urban trees has been proposed to provide substantial benefits by reducing runoff rates and quantities. However, few data are available for rainfall interception of trees in typical urban settings, in contrast to research from natural forests that have dense standings of trees. Additional needed information includes:

• how interception changes for different seasonal changes in urban tree canopies for different types of trees,
• how these interception values vary for different rains; and
• how interception affects urban stormwater for typical urban settings.

This paper describes a series of direct interception (throughfall) measurements under urban trees and calibrated modeling using WinSLAMM to provide some data to address these questions. This study used a standard rain gauge located in an open area and rain gauges under deciduous water oaks (Quercus nigra) and evergreen loblollly pines (Pinus taeda) trees. A total of 85 rain events were monitored from early December 2018 through January 2020 and were statistically evaluated. It was found that tree type had the most important effect on tree canopy interception, followed by rain amount, while seasonal effects were not as important. The interception under the pine was only important for the smallest rain events, while interception under the oaks varied from about 30% to 50%, depending on the rain amount.

1 Introduction

1.1 Background

Urban trees have been linked to enhanced livability and community wellbeing (Millward and Sabir 2011; Beatley and Newman 2013; Beatley 2017), mitigation of urban heat island effects (Ng 2015) and increased biodiversity (Corlett et al.,1988). They are also associated with some problems, such as inlet and drainage clogging by leaves and other organic debris that results in localized flooding (Palla et al. 2018) and damage from failure and windthrow (Li et al. 2018). Roy et al. (2012) summarize both the positive and negative aspects of urban trees. However, there are also detrimental water quality effects associated with urban trees associated with leaf debris (such as nutrient releases) and increased habitat for urban wildlife (and associated increased bacteria). Measurements of roof and parking area runoff having extensive tree coverage have indicated increased bacteria levels due to increased numbers of birds and squirrels in the area (Shergill 2004). Nutrient contributions associated with leaf fall on paved areas have also been quantified (Janke et al. 2017). These other effects associated with urban trees over paved areas also need to be considered. Obviously, trees add substantially to the quality of life in urban areas but can increase nuisance conditions and the need for public works leaf removal.

This paper describes a series of direct interception measurements under urban trees to quantify some of these hydrologic benefits for inclusion in urban stormwater models, such as WinSLAMM. The study presented in this paper included a standard rain gauge located in an open area and rain gauges under deciduous oak and evergreen pine trees to measure throughfall. A total of 85 events monitored from early December 2018 through February 2020 have been statistically evaluated and summarized. These results have been used to add urban tree interception benefits to WinSLAMM for appropriate conditions (tree overstory above directly connected paved areas). Only direct interception is considered, as trunk flow is assumed to infiltrate near the base of an urban tree in the surrounding landscape or tree planter box. These tests are part of research being conducted at the University of Alabama which will incorporate similar interception measurements from other locations, along with other associated tests.

1.2 Literature Review on Benefits of Urban Trees

Data needed for calculating the benefits of urban trees in stormwater management have been frequently recognized at many different locations. The following briefly reviews some of the recent studies that have described these data needs.

Livesley et al. (2016) examined 14 studies on the effect of trees on water, heat, and pollution cycles at different local scales. They outlined a framework of multidisciplinary studies to obtain additional understanding of the biogeochemical aspects of urban trees. Berland et al. (2017) examined the literature on how interacting mechanisms associated with urban trees affect urban hydrology. They concluded that many of these interactions are poorly understood, especially at the spatial and temporal scales most important for stormwater management. Based on a literature review, Kuehler et al. (2017) concluded that “inadequate research quantifying the urban tree contribution to rainfall–runoff processes limits their promotion by stormwater managers.” Researchers in Belgium (Smets et al. 2019) reported that “an important knowledge gap in current urban hydrological models [is caused by lack of] reliable, generic data about interception storage capacities of small urban plant species.” They conducted several interception modeling studies to examine the sensitivity of tree characteristics. Based on the available tree characteristics database, biomass (tree height and diameter) was determined to be most important for calculating interception.

Studies have relied on modeling to help identify the important factors affecting rain interception associated with urban trees. A tree rainfall interception model was developed by California researchers (Xiao et al. 2000) and compared to direct measurements for throughfall and stemflow. Interception losses were about 15% for a pear tree and 27% for an oak tree. Interception was greatest at the start of the rain events. They also found that rainfall frequency was more important than rain intensity in affecting interception losses. Increased rain intensity and wind speed increased stemflow while reducing throughfall. In a Canadian study, Kirnbauer et al. (2013) modeled the effects of planting trees on vacant urban land to benefit stormwater. They found that planted trees could reduce (intercept and evaporate) from 7% to 27% of the rainfall from a planted lot. Local tree growth information was needed to improve the reliability of the modeled values. Gonzalez-Sosa et al. (2017) modeled the benefits of literature-based tree interception losses on different aspects of urban hydrology and their benefits in combined sewer areas. They concluded that 10% to 20% runoff volume reductions and peak delays of 10 min to 15 min could be expected by using street trees. Indoor simulated rainfall interception experiments were conducted by Baptista et al. (2018) to study the factors affecting the tree’s rainfall storage capacity. They found that canopy rainfall storage capacity was well correlated to plant surface area and area density, reflected in types and abundance of leaves on the trees.

Field experiments have also been conducted at many locations to directly measure interception and other losses associated with urban trees, although most of these were limited in the number and types of trees and extent of the investigations. Some of these studies are mentioned below. Guevara-Escobar et al. (2007) examined the distribution patterns of precipitation around the canopy of a single evergreen tree during 19 storms in Mexico. During late summer to early fall months, they measured 38% throughfall, 2.4% stemflow, and 60% interception by the tree canopy. They also measured an average time of 20 min before canopy saturation. Rainfall screening was also important in the area surrounding the tree (about 18% losses). Kermavnar and Vilhar (2007) measured urban tree interception from a mixed upland forest in the city center, a riparian pine forest and a floodplain hardwood forest in Slovenia. Tree species, canopy cover and tree dimensions were the main determining factors affecting interception. The mixed forest had an average interception amount of 18%, while the pine forest had only 4% interception of the rainfall and the mixed hardwood forest had 7% interception. Changing tree characteristics (leaf vs leafless periods) and rainfall intensity were found to be important factors that affected the portioning of the losses during the seasons. Asadian and Weiler (2009) studied rainfall interception under six different urban trees during seven rain events in British Columbia. Average canopy interception during these events for Douglas fir and western red cedar were about 49% and 60%. The trees also caused a delay in the rain throughfall reaching the ground. These losses were found to be dependent on canopy structure, climatic conditions and rain characteristics. As part of a research program in Australia, Livesley et al. (2014) studied interception of two eucalyptus tree species used as street trees. The sites having a greater tree density intercepted 44% of the annual rainfall compared to 29% for trees that were less densely planted. Also, stemflow (only 5% of the intercepted amount) was less for rough bark specimens compared to smooth bark specimens. Overall, they expected up to 20% runoff reductions associated with the presence of street trees. Van Stan et al. (2015) studied two common northeastern United States urban tree species (beech and poplar). Poplars, having rougher bark, lower branch inclination and thinner canopy, had greater losses compared to the beech trees, which allowed greater amounts of the rainfall to reach the underlying soil. Rainfall throughfall measurements were made during 10 rain events in Brazil by Alves et al. (2018). They found that rainfall interception was highly dependent on tree species. Researchers in Slovenia (Zabret et al. 2018) studied pine and birch rainfall losses over 180 rain events. The amount of rainfall was found to be the most important variable, with rain intensity and number of raindrops also being important. San Juan researchers (Nytch et al. 2019) studied 6 trees during 13 storms to measure factors affecting interception losses. Deciduous trees had 23% interception losses while evergreen trees had 17% interception losses. The tree species affected the interception for low and moderately intense rains but not for high intensity rains.

Besides interception of rainfall by tree canopies, trees can also reduce runoff quantities and flow rates by enhancing underlying soils, as noted by the following researchers. Schooling and Carlyle-Moses (2015) stated that “generalizations that deciduous canopies reduce stormwater are based largely on closed canopy forests, highlighting the need for more detailed study of isolated urban trees.” They examined isolated deciduous trees in Germany. Stemflow was a maximum of 23% of the canopy interception and funneling was about 20% of the maximum stemflow. They concluded that infiltration at the base of isolated urban trees needs to be considered as part of stormwater management schemes. Kuehler et al. (2017) found that soil benefits of urban trees should also be considered and integrated into the design of urban tree stormwater practices. Improved soil conditions beneath urban trees in Germany were studied by Rahmana et al. (2019). Higher infiltration rates were associated with trees having fast growth rates and increased root mass.

Table 1 summarizes the interception percentages of the rainfall reported by the above cited literature. These range from a low of 4% to a high of 60%, depending on the tree species and climatic conditions. Although these locations are widespread, there are overall few locations and tree species represented in the urban tree interception literature. Obviously, additional data are needed before comprehensive conclusions can be made for all situations. Luckily, these data are relatively inexpensive to collect and should be considered as part of comprehensive stormwater monitoring programs.

Table 1 Examples of reported urban tree interception values.

Location	Reference	Interception	Description
California	Xiao et al. 2000	15%	pear
		27%	oak
Mexico	Guevara-Escobar et al. 2007	60%
Slovenia	Kermavnar and Vilhar 2007	18%	mixed urban forest
		4%	mixed hardwood urban forest
British Columbia	Asadian and Weiler 2009	49%	Douglas fir
		60%	western red cedar
Australia	Livesley et al. 2014	44%	high density eucalyptus
		29%	less dense eucalyptus
San Juan	Nytch et al. 2019	23%	deciduous trees
		17%	evergreen trees

The experiments described in this paper were conducted to examine canopy interception by direct measurements of throughfall under isolated or low density stands of mature urban deciduous and evergreen trees in the southeastern United States. These measurements were made for 85 rain events at the open grass site, 75 events at the pine site, and 72 events at the oak site (fewer interception events occurred under trees as some small events did not produce any throughfall and some events were not recorded due to clogging or data logging issues). All seasons were also monitored, although the fall season had few events due to drought conditions and data collection problems. These data were analyzed to determine statistically significant relationships for use in urban stormwater models. Additional interception measurements are currently being made under smaller urban trees (Japanese maples, Acer palmatum) and will be reported in a follow up paper.

When modeling, there is a possibility for over- or under-counting some of the benefits of urban trees. As in all modeling, it is important that calibration areas be similar to test areas. For example, calibrated stormwater models rely on monitored outfall flow measurements of existing areas. These areas have varying amounts of trees through their landscapes. Adding additional interception to models that were calibrated in areas having extensive tree coverage can result in improper estimates of runoff. Similarly, errors would also occur if the calibrated area had few trees, but the model analyses assumed extensive trees. Figures 1 through 3 are examples of monitored medium density residential areas used in previous WinSLAMM calibrations showing the contrast of mature trees in older areas with the few young trees in new developments. These residential areas were separated based on age of development to account for the differences in vegetation. However, the outfall monitored runoff characteristics did not indicate any differences between the old and new developments, beyond that which was explained by differences in directly connected impervious area types and other land surface areas, implying little hydrologic effects associated with the trees.

If a tree is located in a pervious area of the watershed (i.e. the canopy is over lawns or other nonpaved areas), interception may not greatly affect outfall runoff quantities; any unintercepted rainfall (throughfall) is likely to be infiltrated with or without the trees. However, trees likely maintain good soil characteristics and minimize compaction, which increases rainfall infiltration. The largest hydrologic benefit of urban trees occurs when directly connected impervious areas (such as roofs, walkways, parking areas and streets) are heavily covered by an overstory of trees. If tree covered impervious areas are directly connected to the drainage system, these benefits are greatest, but if tree covered impervious areas drain to pervious areas (such as disconnected roofs or walks surrounded by lawns), the benefits are lower.

This paper describes a series of direct measurements of throughfall under urban trees (canopy interception) for inclusion in WinSLAMM to quantify some of these hydrologic benefits. This study included a standard rain gauge located in an open area and rain gauges under deciduous oak and evergreen pine trees. The results of the 85 events monitored from early December 2018 through February 2020 were statistically evaluated and summarized. These results were used to add urban tree interception benefits to WinSLAMM for appropriate conditions (tree overstory above directly connected paved areas for different types of trees, seasons, and rain quantities). Only direct canopy interception is considered as stemflow is assumed to infiltrate near the base of an urban tree in the surrounding landscaping or tree planter box. These tests are part of research being conducted at the University of Alabama that includes similar interception measurements from other locations, along with other associated tests, building upon early research on urban landscaping evapotranspiration (ET) as summarized in Bean and Pitt (2012).

2 Methods

2.1 Site Description

The test site was located at the home of one of the co-authors in Hoover, Alabama, a suburb of Birmingham. The site has an elevation of 312 m (1030 ft) and is shown on Figure 4.

Figure 4 Topographic map of Alabama showing location of test site (Department of Geology, University of Alabama).

Figure 5 is an aerial photograph of the home with the three rain gauge locations indicated. The neighborhood was developed >50 years ago and has many mature trees.

Figure 5 Location of test trees (Bing photo).

This lot was developed in 1957 with most of the pines and oaks likely present before home construction. The site is well wooded with mature trees in the front and sides of the lot, with a large open grass area in the rear of the lot. Appendix A includes photographs of the test sites. These photographs are generally of the rain gauge, the overhead branches and leaves, and any debris on the rain gauge screen before calibration, cleaning, and recalibration; photos were taken each time the rain gauge data were downloaded. The test periods between data downloads varied depending on travel schedules and weather, ranging from about 3 weeks to almost 4 months.

The evergreen test tree was a loblolly pine. According to Wikipedia (2020-11-30), “Pinus taeda, commonly known as loblolly pine, is one of several pines native to the southeastern United States, from central Texas east to Florida and north to Delaware and southern New Jersey. The wood industry classifies the species as a southern yellow pine. U.S. Forest Service surveys found that loblolly pine is the second most common species of tree in the United States, after red maple. For its timber, the pine species is regarded as the most commercially important tree in the southeastern U.S.“ The Conifer Society also states that it is an important timber species outside of its native range, with substantial plantations in South Africa, Zimbabwe, southern Brazil, Argentina, China and Australia (https://conifersociety.org/conifers/pinus-taeda/). The test tree is located in the front yard and is one of a group of three pines surrounded by oaks. These trees are mature (about 0.4 m in trunk diameter and 20 m tall with the first branch about 5 m from the ground). The upward aiming photographs in Appendix A above the pine show that the needle and branch density is relatively sparse, especially compared to the oak.

The deciduous test tree is a water oak. According to Wikipedia (2020-11-30), Quercus nigra, the water oak, is an oak in the red oak group (Quercus sect. Lobatae), native to the eastern and south-central United States, found in all the coastal states from New Jersey to Texas, and inland as far as Oklahoma, Kentucky, and southern Missouri. Quercus nigra is a medium-sized deciduous tree, growing to 30 m (100 ft) tall with a trunk up to 1 m (3 ft) in diameter.” The test oak is one of several widely spaced oaks in the front yard. It is mature (about 1 m in trunk diameter and 30 m tall, with the first branch ~3 m from the ground). The upward aiming photographs in Appendix A show the highly seasonal leaf densities and massive limbs and branches.

2.2 Weather Conditions during Monitoring Period

The National Weather Service (https://www.weather.gov/bmx/climo_2019review) reported that the total annual rainfall for Birmingham, AL in 2019 (the year when most of the data for these analyses were collected) was 132.1 cm (52.00 in.), 4.37 cm (1.72 in.) below normal. Abnormally dry conditions occurred during March, April, May, and July. The service summarized the statewide 2019 rainfall as: “2019 will go down in the record books as a fairly normal year hydrology-wise in many respects, with varied rainfall, scattered episodes of river flooding, and periodic flash flood events from localized heavy rainfall. Mixed in the middle of this was also an episode of drought from late summer into the fall season. Overall, rainfall in 2019 averaged between 110 cm and 150 cm (45 in.–60 in.) with localized higher totals. These amounts were generally below normal over the southeast half of central Alabama and above normal over the northwest sections. This produced a variety of hydrologic events throughout the year in central Alabama.”

2.3 Rain Gauges and Their Calibration

The rain gauges (HOBO Data Logging Rain Gauge RG3) were installed on 2018-12-10. The first rain event occurred on the following day. Figures 6 through 8 show these rain gauge installations, while Appendix A includes photographs from throughout the monitoring period.

The data loggers record temperature (every 5 min) and time for every tip of the rain gauge (0.01 in., 0.25 mm). A Davis Vantage Pro2 weather station was also located in the open area for most measurements (wind, temp, rain, UV, humidity, pressure, ET), but these data were only available in 30 min time steps.

Calibration tests were conducted before the rain and throughfall data were downloaded from the data loggers. If there was any debris on the rain gauge screens, another calibration test was conducted after the debris was removed. If the calibration ratios were close to 1.00 (rain gauge recorded depth ratio to standard depth), then the prior data were deemed satisfactory. The first calibration for the oak rain gauge was only 0.85 and the last calibration at this location was only 0.63, both possibly due to observed long leaf stems interfering with the tipping mechanism. When the event data were compared to the grass rain gauge data for the periods prior to these problem calibrations, it was determined that some of the data were questionable and were therefore not included in the summary or analyses.

Figure 9 shows the calibration unit on top of the grass area rain gauge. The calibration orifice is off center to ensure that water did not directly enter the rain gauge funnel opening. The calibration unit was an FC-500 from Texas Electronics (model TR-525101) with 0.01 in./tip (0.25 mm) and 100 tips for 1.00 in. (25.4 mm) full capacity.

Figure 9 Calibration setup.

The data were downloaded and the calibrations conducted on the following dates: 2019-01-07, 2019-01-26, 2019-02-25, 2019-03-23, 2019-04-16, 2019-06-03, 2019-07-25, 2019-11-20, 2020-01-24 and 2020-03-27.

Figure 10 is a time series of the calibration ratios (actual total depth during calibration to standard depth of 1 in., 25.4 mm).

Figure 10 Calibration ratios time series.

Table 2 summarizes the calibration ratios (Figure 10) for the study period. Nine calibration replications were also conducted at the grass location on 2019-03-24. The average calibration ratios were 1.04 for the pine and oak locations and 1.05 for the grass location. The replicated average calibration ratio at the grass location was 1.06. The coefficients of variation for each data set were 0.01 and 0.02.

Table 2 Summary of calibration statistics.

	Grass calibrations	Pine calibrations	Oak calibrations	Grass calibration replicates
Number	12	15	14	9
Minimum	1.02	1.02	1.02	1.04
Maximum	1.09	1.07	1.07	1.08
Average	1.05	1.04	1.04	1.06
COV	0.02	0.01	0.01	0.02

As noted above, if any debris was found on the screen, the screen was cleaned and a second calibration test was conducted. Paired two sample t-tests for means were used to determine if the before–after cleaning calibrations were significantly different. Table 3 summarizes these calculations. The before and after cleaning calibration ratios were very close and were not found to be significantly different. Therefore the previous plots and data summaries included both the paired calibration ratio values.

Table 3 Paired calibration values.

	Grass	Pine	Oak	All combined
Number of pairs	2	5	4	11
Before cleaning mean	1.060	1.042	1.035	1.043
Aafter cleaning mean	1.060	1.040	1.048	1.046
Two tail P value	n/a	0.82	0.39	0.54

There were several periods of missing data for various reasons, including:

• grass area rain gauge: 85 events recorded with no missing data;
• pine rain gauge: 75 events recorded, missing data for 10 events between 2020-01-26 and 2020-02-19 due to debris interfering with tipping mechanism; and
• oak rain gauge: 72 events recorded, missing data for 3 events between 2018-12-31 and 2019-01-03 due to debris interfering with tipping mechanism, missing data for 7 events between 2019-12-06 and 2020-01-26 events due to debris interfering with tipping mechanism, and missing data for 3 events from 2019-11-23 to 2019-12-01 due to data logger failure.

Another reason for unequal total numbers of rain events recorded at each location is that no rain was observed under the trees for some of the small rain events.

3 Results

Appendix B lists the observed rainfall conditions at the three rain gauges for the period 2018-7-10–2020-02-20. A total of 85 rains events were monitored during this period, 42 during winter (December–February), 26 during spring (March–May), 14 during summer (June–August), and 3 during fall (September–November). Few fall rain events were monitored due to an unusual dry period and equipment problems (data logger resetting due to cold weather and interference of the tipping mechanism by leaf stem). The rain depths ranged from 0.3 mm to 139 mm, with a median rain depth of 18.2 mm during winter, 6.2 mm during spring, 6.5 mm during summer, and 11.7 mm during fall. The maximum rain depths were 139 mm during winter, 63.2 mm during spring, 36.6 mm during summer, and 41.4 mm during fall. The interevent periods ranged from ~7 h to 24 d. Rain duration ranged from ~0.1 h to 70 h, with a median duration of ~14 h during winter, 4 h during spring, 0.6 h during summer, and 11 h during fall.

Figure 11 is a scatterplot of the tree throughfall and rain depths for all of the monitored events. The oak throughfall values were mostly smaller than the pine throughfall values for all rain depths. The equivalence line indicates equal rain and throughfall values. Most of the throughfall values were all lower than this equivalence line for the rains <10 mm. The pine throughfall values approach and sometimes exceed this line for larger rain events, while the oak throughfall values remain below the line for almost all large events. The tree interceptions (as a fraction of the rain amounts) are therefore greater for the small rain events compared to the large rain events, especially for the pine values. The scatterplots of the data for any rain depth were similar for both tree types, reflecting similar measurement errors and rainfall variations, but the pine throughfall values have more events larger than the rain amount because of their greater values compared to the oak throughfall values.

Figure 11 Scatterplot of tree throughfall and rain depths.

A 2³ full factorial analysis (Box et al. 1978) was conducted to identify any significant effects on the throughfall to rain ratio values. This was conducted by organizing the throughfall ratios to the grass rains into eight categories corresponding to all combinations of season, tree type and rain depth category. The ratios of throughfall to rain depths are the throughfall amounts recorded under the trees divided by the total rain depth recorded at the grass location. Smaller ratios indicate greater amounts of interception by the trees. In some cases this was found to be >1.0, as noted above , due to typical measurement errors.

Table 4 lists the 8 factorial categories and the associated average ratios of throughfall to rain depths. Fall rains were not included in these analyses due to the few rain events available. The number of events in each category and the expected errors (as a fraction of the average) for 95% confidence and 80% power are also shown (Burton and Pitt 2002) and are in the range 0.2–0.5.

Table 5 shows the seven main factors and interactions, their calculated effects, and rankings. Only the main factors were found to be significant (using three times the grouped standard error as the critical value which was calculated to be 0.21). These significant factors were also confirmed by probability and dot plots of the calculated effects.

Table 4 2³ full factorial categories. 

	Average ratio of throughfall to rain depth	COV	Number of events in category	Fraction error with alpha = 0.05 and beta = 0.20
Ratio of pine throughfall/grass rain
Winter large rains (>25 mm)	1.04	0.33	11	0.4
Winter small rains (<25 mm)	0.98	0.46	14	0.4
Spring plus summer large rains (>25 mm)	0.94	0.11	6	0.4
Spring plus summer small rains (<25 mm)	0.54	0.52	34	0.2
Ratio of oak throughfall/grass rain
Winter large rains (>25 mm)	0.70	0.29	8	0.4
Winter small rains (<25 mm)	0.45	0.59	12	0.5
Spring plus summer large rains (>25 mm)	0.54	0.19	6	0.5
Spring plus summer small rains (<25 mm)	0.43	0.56	34	0.2

Table 5 Main factors, interactions and calculated effects.

Main Factors and Interactions	Calculated Effects	Ranked Effects
S (season)	−0.20 (marginally significant)	3
R (rain depth)	−0.23 (significant)	2
T (tree species)	−0.37 (significant)	1
SR (season and rain depth interaction)	−0.02	7
ST (season and tree species interaction)	0.12	4
RT (rain depth and tree species interaction)	0.05	6
SRT (season, rain depth and tree species interaction)	0.09	5

All three main factors were significant when determining the throughfall ratio (rain depth under the trees vs rain depth at the grass location). The tree species (oak vs pine) had the greatest effect on the throughfall, followed by the rain depth, while the seasonal differences had only marginally significant effects. These were found to be independent of each other as no significant interactions were identified.

Paired t-tests were also conducted to compare the individual rain depths (in log10 space) measured under the two trees with the rain depths measured at the gauge surrounded by grass. The grass vs. pine gauge and the grass vs. oak gauge data differences were significant (p ≤0.001), as expected. The data were further subdivided by season, tree species, and rain depth category (<12 mm; 12 to 50 mm; and >50 mm) for comparisons. Kruskal-Wallis One Way Analysis of Variance on Ranks tests indicated a number of significant differences between these subgroups, as shown in the Figure 12 box and whisker plots.

The box and whisker plots in Figure 12 indicate that the throughfall ratios were greater (largest abstractions) for the small events than the larger events. Many of the throughflow values recorded under the pine in the winter (cases 1, 8 and 15 on the chart) were larger than the grass rainfall amounts. The pine needle density was lower during the winter and the limbs tended to have a greater effect on the rain falling through the tree, possibly causing funneling of rain into the rain gauge from overhanging branches. Also, the pine throughfall to rain ratio observations were consistently greater than the oak ratios. The fall distance from the lowest branches to the rain gauge under the pine was ~5 m, making it unlikely that a stream of water would preferentially and consistently enter the gauge. If possible, multiple gauges could be placed under each tree to better indicate the variability (this is planned for future monitoring).

Figure 12 Ratios of tree canopy throughfall to rain for tree species, season and rain depth categories.

The most notable difference is the spring, summer, and fall pine conditions for the small rain events; they are similar to the oak values and much less than observed for the winter pine conditions (as indicated in the box and whisker plot categories).

The factorial analyses indicated that tree species, rain depth, and season were all significant in affecting the throughfall amounts. Regression analyses were conducted considering these three factors. The scatterplots in Figure 13 are for the total throughfall amounts under the pine tree or the oak tree plotted against the rain depths at the grass gauge for the winter, spring, summer, and fall (only for pine) seasons. The significant regression coefficients are shown in Table 6. The plots and ANOVA statistical tests were conducted on log10 transformed rain depth data to obtain reasonably even distributions of observations spread across the rainfall depths to provide acceptable residual analysis results (Draper and Smith 1981). Table 6 shows that the resulting regressions and coefficients are highly significant, with the exception of the fall–oak regression that has a marginal overall regression significance due to the few data available for that condition. The pine data for the winter and fall periods did not result in significant constant (intercept) terms (regression through the plot origin), so those regressions only have a slope coefficient term, while all of the oak data and spring and summer pine data had both significant intercept and slope coefficients. The negative intercept values indicate that some rain needed to occur before throughfall occurred (similar to initial detention storage terms in hydrology).

The residual analyses (example shown in Figure 14 for winter pine conditions) indicated satisfactory patterns.

The time delays until the onset of recorded rain under the pine and oak trees were also examined. There were substantial variations in the rain delays, with medians of ~12 min for the pine tree to 7 min for the oak tree. No significant relationships were found affecting the delays for different rain depths, intensities, or wind speeds, although there were apparent larger delays associated with smaller rain events, smaller rain intensities and lighter winds in comparison with larger rain events, intensities and winds. These trends were more apparent for the oak data than for the pine data.

Wind data were also available from the weather station located near the grass rain gauge. Full factorial analyses were conducted to identify significant factors that may affect the observed interception ratios. However, even with 85 data sets, there were some missing and underrepresented conditions for the eight possible combinations (2³) when average rain intensity, total rain and peak wind speed were evaluated. When only total rain and peak wind speed were examined (2² = 4 combinations), large rains with low winds were still underrepresented (i.e. stormy conditions associated with large rains likely have large winds). Apparent relationships of increased throughfall with total rain and peak rain intensity were observed. As previously noted, the 2³ full-factorial analyses that considered tree species, season and rain depth and their interactions showed that all three main factors were significant, with the tree species and the rain depth being the most significant factors, and seasons having somewhat smaller effects.

Figure 13 Scatterplots of throughfall for tree species and season.

Table 6 Regression coefficients of throughfall vs grass rainfall (log10 mm transformations).

	Intercept term	P-value for intercept	Slope term	P-value for slope term	R2	Regression significance F
Winter
Pine	0.00	n/a	1.01	<0.001	0.98	<0.001
Oak	−0.39	0.00	1.09	<0.001	0.92	<0.001
Spring
Pine	−0.36	<0.001	1.21	<0.001	0.97	<0.001
Oak	−0.34	<0.001	1.06	<0.001	0.95	<0.001
Summer
Pine*	−0.75	<0.001	1.59	<0.001	0.95	<0.001
Oak	−0.49	<0.001	1.21	<0.001	0.95	<0.001
Fall
Pine	0.00	n/a	0.90	0.0099	0.98	0.064**
Oak	n/a***		n/a

*Example equation form:
 summer pine log10 total rain (mm) = −0.75 + 1.59 × Summer grass log10 total rain (mm).
**Marginal significance due to few observations.
***No oak fall season observations available.

Figure 14 Example residual plots for equation describing canopy interception for pine species during the winter.

4 Discussion

Highly significant regression equations relating rain depth and throughfall were developed for conifer and deciduous trees for the different seasons for implementation in WinSLAMM , the Source Loading and Management Model. These relationships are expected to be useful for other urban stormwater models for use with similar climatic and tree conditions. WinSLAMM is an urban stormwater model that considers land development characteristics, specifically the types of land cover in different land uses. It can be used for single urban lots to complex urban catchments. WinSLAMM is based on field monitoring observations covering a wide range of scales with unique hydrologic and pollutant related components that focus on urban systems. Several presentations and papers associated with the annual urban water modeling conferences and the Journal of Water Management Modeling have described WinSLAMM and its attributes; amongst others are Pitt 1997, 1999; Pitt and Lantrip 2000; Pitt and Voorhees 2011. The addition of tree interception allows for the calculated benefits of urban trees on urban hydrology, as described below in an example.

As noted previously, tree interception effects on throughfall in stormwater management are only relevant for trees that shade directly connected impervious areas. Counting the benefits of existing trees in a calibrated model would likely result in double counting the benefits. Also, the benefits of new trees shading uncompacted soils during small and intermediate rain events would be small as the throughfall would likely be almost completely infiltrated, as would the total rainfall for these areas. During large rain events, the canopy interception fraction of the rain is greatly reduced, which results in minimal differences in runoff compared to uncompacted soil areas having no trees. WinSLAMM was therefore modified to directly calculate the benefits of trees over directly connected impervious areas, as shown in Figures 15 and 16.

Figure 15 Tree canopy interception shading over directly connected impervious areas (shown in red outline).

The screen shot of a paved parking area in WinSLAMM (Figure 16) shows how the tree canopy shading values are entered for directly connected areas.

Figure 16 Paved parking area information input screen.

Tables 7 and 8 shows the seasonal and annual runoff reductions for the continuous WinSLAMM analyses for one year (1977, previously identified as being close to an average rain year in total amount, monthly distribution and numbers of events) of rains for Birmingham, Alabama from a 0.40 ha paved parking area having varying amounts of shading by tree canopies. WinSLAMM is a continuous model that calculated the tree interception amounts using the previously presented regression equations for different rain depths, seasons and types of tree.

The deciduous trees show the greatest potential benefit. With 100% shading, the deciduous trees may provide about 39% reductions in runoff from paved areas. The benefits are linear, with, for example, 50% canopy shading giving half this maximum benefit. The conifer calculations resulted in much smaller benefits, especially for the winter and spring seasons which have the larger rains with much reduced interception of rainfall by the pines. The annual benefit of shading of impervious areas by conifers is only about 4%. This is expected to be greater in areas having more moderate rainfalls that this example area. Both the pine and oak had almost complete interception of the smallest rains monitored, but the pine’s interception benefits decreased much more rapidly as the rain depths increased.

Table 7 Runoff reductions from paved areas shaded by conifers.

	No trees	10% conifer	25% conifer	50% conifer	100% conifer
Winter	0.0	0.2	0.6	1.2	2.3
Spring	0.0	0.1	0.4	0.7	1.5
Summer	0.0	1.0	2.4	4.9	9.7
Fall	0.0	0.7	1.8	3.6	7.1
Annual	0.0	0.4	1.0	2.0	4.0

Table 8 Runoff reductions from paved areas shaded by deciduous trees.

	No trees	10% deciduous	25% deciduous	50% deciduous	100% deciduous
Winter	0.0	3.7	9.3	18.5	37.1
Spring	0.0	3.8	9.6	19.1	38.3
Summer	0.0	4.1	10.2	20.4	40.9
Fall	0.0	3.8	9.4	18.9	37.7
Annual	0.0	3.9	9.6	19.3	38.6

The relatively low runoff reduction values for all conifer examples and for the low amounts of deciduous tree coverage are in contrast with rural forested areas, where the runoff amounts of heavily wooded areas (both conifers and deciduous trees) are very small. These major forest benefits are mostly associated with the forest duff (thick layers of partially and completely decomposed organic material) beneath the trees and large infiltration rates through undisturbed soils. In urban areas (especially for thinly planted or isolated trees, if relatively young and with common leaf removal by homeowners) the benefits of trees on underlying soils are important, but much reduced compared to thick stands of mature trees having deep layers of organic material covering the soil. Duff has no effect on paved areas, although it may build up near the trunk in tree planter boxes or other small areas. Therefore, the main direct effect of urban trees on urban hydrology is the limited canopy interception amounts.

5 Conclusions

Literature reviews indicate that the interacting mechanisms of urban trees that affect urban hydrology are poorly understood. Past canopy throughfall measurements of urban trees have identified important differences between tree species and study area. Projected runoff volume reductions due to extensive use of urban trees have been found to be about 10%–20%. Field studies have also concluded that stemflow is usually a small portion of the total tree runoff yield to runoff (usually <10% of the canopy throughfall). Soil characteristics under urban trees are also expected to affect understory runoff yields, with trees expected to improve soil structure (decreased compaction and increased organic matter).

Very limited data of throughfall are available for urban trees in the southeastern United States. Measurements of throughfall were made during 85 rains from December 2018 through February 2020, resulting in throughfall data for 72 events for the deciduous oak tree and for 75 events for the evergreen conifer. The rain depths for the monitored rains ranged from 0.3 to 139 mm, with a median rain depth of 9.7 mm.

A full 2³ factorial analysis was conducted using the throughfall and corresponding meteorological data. The tree species (oak vs pine) had the greatest effects on the throughfall, followed by the rain depth, while the seasonal differences had only marginally significant effects. Kruskal-Wallis One Way Analysis of Variance on Ranks tests indicated a number of significant differences between tree types, rain category and seasonal subgroups. The ratios of throughfall amounts to the rain depths were smallest (largest percentage abstractions) for the small events compared to the larger events. Also, the pine throughfall to rain ratio observations were much greater than the oak throughfall to rain ratios (less abstractions). There were no large differences in the seasonal rain ratios for each category, except for the winter vs spring pine values for the smallest rains.

The literature indicated several areas lacking information for a comprehensive hydrologic understanding of urban trees. These needs mostly relate to additional interception data for additional urban tree species in more climates. Other issues needing information are the role of urban trees on soil structure under and surrounding trees, specifically mitigation of soil compaction.

Tree interception measurements are recommended to be a part of urban stormwater monitoring projects to directly measure the benefits of urban trees. Continued throughfall measurements are planned for smaller urban trees as part of this research. Other planned monitoring includes throughfall measurement variations under trees and rainfall distributions surrounding trees.

Acknowledgment

The authors would like to thank the anonymous reviewers for their helpful suggestions which resulted in a much-improved paper.

References

1. Alves, P.L., K.T.M. Formiga, and M.A.B. Traldi. 2018. “Rainfall interception capacity of tree species used in urban afforestation.” Urban Ecosystems 21: 697–706. https://doi.org/10.1007/s11252-018-0753-y
2. Asadian, Y., and M. Weiler. 2009. “A new approach in measuring rainfall interception by urban trees in coastal British Columbia.” Water Quality Research Journal Canada 44 (1): 16–25. https://iwaponline.com/wqrj/article-pdf/44/1/16/229544/wqrjc0440016.pdf
3. Baptista, M.D., S.J. Livesley, E.G. Parmehr, M. Neave, and M. Amati. 2018. “Variation in leaf area density drives the rainfall storage capacity of individual urban tree species.” Hydrological Processes 32: 3729–40. https://doi.org/10.1002/hyp.13255
4. Bean, R., and R. Pitt “Advancing stormwater beneficial uses: ET mapping in urban areas.” Water Environmental and Water Resources Congress 2012. ASCE-EWRI. Congress Proceedings. Albuquerque, NM, May 20-25 2020. http://unix.eng.ua.edu/~rpitt/Publications/6_Urban_Hydrology/ET%20in%20urban%20areas%20Bean_Pitt_EWRI2012_1.pdf
5. Beatley, T. 2017. “Biophilic cities and healthy societies.” Urban Planning 2 (4): 1–4.
6. Beatley, T., and P. Newman. 2013. “Biophilic cities are sustainable, resilient cities.” Sustainability 5 (8): 3328–45.
7. Berland, A., S.A. Shiflett, W.D. Shuster, A.S. Garmestani, H.C. Goddard, D.L. Herrmann, and M.E. Hopton. 2017. “The role of trees in urban stormwater management.” Landscape and Urban Planning 162: 167–77.
8. Box, G.E.P., W.G. Hunter, and J.S. Hunter. 1978. Statistics for Experimenters. New York: John Wiley and Sons.
9. Burton, G.A. Jr., and R. Pitt. 2002. Stormwater Effects Handbook: A Tool Box for Watershed Managers, Scientists, and Engineers. Boca Raton, Fl: CRC Press. ISBN 0-87371-924-7.
10. Corlett, R., and T. Bukit Timah. 1988. “The history and significance of a small rain-forest reserve.” Environmental Conservation 15 (1): 37–44.
11. Draper, N., and H. Smith. 1981. Applied Regression Analysis. New York: John Wiley and Sons.
12. Gonzalez-Sosa, E., I. Braud, R.B. Piña, C.A.M. Loza, N.M.R. Salinas, and C.V. Chavez. 2017. “A methodology to quantify ecohydrological services of street trees.” Ecohydrology and Hydrobiology 17: 190–206. https://doi.org/10.1016/j.ecohyd.2017.06.004
13. Guevara-Escobar, A., E. González-Sosa, C. Véliz-Chávez, E. Ventura-Ramos, and M. Ramos-Salinas. 2007. “Rainfall interception and distribution patterns of gross precipitation around an isolated Ficus benjamina tree in an urban area.” Journal of Hydrology 333: 532–41. https://doi.org/10.1016/j.jhydrol.2006.09.017
14. Janke, B.D., J.C. Finlay, and S.E. Hobbie. 2017. “Trees and streets as drivers of urban stormwater nutrient pollution.” Environmental Science and Technology 51 (17): 9569–79. https://doi.org/10.1021/acs.est.7b02225
15. Kermavnar, J., and U. Vilhar. 2007. “Canopy precipitation interception in urban forests in relation to stand structure.” Urban Ecosyststems 20: 1373–87. https://doi.org/10.1007/s11252-017-0689-7
16. Kirnbauer, M.C., B.W. Baetz, and W.A. Kenney. 2013. “Estimating the stormwater attenuation benefits derived from planting four monoculture species of deciduous trees on vacant and underutilized urban land parcels.” Urban Forestry and Urban Greening 12: 401–7. https://doi.org/10.1016/j.ufug.2013.03.003
17. Kuehler, E., J. Hathaway, and A. Tirpak. 2017. “Quantifying the benefits of urban forest systems as a component of the green infrastructure stormwater treatment network.” Ecohydrology 10: e1813. https://doi.org/10.1002/eco.1813
18. Li, Y., H. Rahardjo, K.N. Irvine, and A.W.K. Law. 2018. “CFD analyses of the wind drags on Khaya Senegalensis and Eugenia Grandis.” Urban Forestry & Urban Greening 34: 29–43.
19. Livesley, S.J., B. Baudinette, and D. Glover. 2014. “Rainfall interception and stem flow by eucalypt street trees: The impacts of canopy density and bark type.” Urban Forestry and Urban Greening 13: 192–7. https://doi.org/10.1016/j.ufug.2013.09.001
20. Livesley, S.J., E.G. McPherson, and C. Calfapietra. 2016. “The urban forest and ecosystem services: Impacts on urban water, heat, and pollution cycles at the tree, street, and city scale.” Journal of Environmental Quality. January 11 2016. https://doi.org/10.2134/jeq2015.11.0567
21. Millward, A.A., and S. Sabir. 2011. “Benefits of a forested urban park: What is the value of Allan Gardens to the city of Toronto, Canada?” Landscape and Urban Planning 100 (3): 177–88.
22. Ng, Y.X. 2015. “A study of urban heat island using local climate zones: The case of Singapore.” British Journal of Environment and Climate Change 5 (2): 116–33.
23. Nytch, C.J., E.J. Meléndez-Ackerman, M. Pérez, and J.R. Ortiz-Zayas. 2019. “Rainfall interception by six urban trees in San Juan, Puerto Rico.” Urban Ecosystems 22: 103–15.  https://doi.org/10.1007/s11252-018-0768-4
24. Palla, A., M. Colli, A. Candela, G.T. Aronica, and L.G. Lanza. 2018. “Pluvial flooding in urban areas: the role of surface drainage efficiency.” Journal of Flood Risk Management 11: S663–76.
25. Pitt, R. 1999. “Small storm hydrology and why it is important for the design of stormwater control practices.” In Advances in Modeling the Management of Stormwater Impacts, vol. 7, edited by W. James, 61–91.Guelph: Computational Hydraulics Int.
26. Pitt, R. 1997. “Unique Features of the Source Loading and Management Model (SLAMM).” In Advances in Modeling the Management of Stormwater Impacts, vol 6, edited by W. James, 13–37. Guelph: Computational Hydraulics Int.
27. Pitt, R., and J. Lantrip. 2000. “Infiltration through disturbed urban soils.” In Advances in Modeling the Management of Stormwater Impacts, vol 8, edited by W. James, 1–22. Guelph: Computational Hydraulics Int.
28. Pitt, R., and J. Voorhees. 2011. “Modeling green infrastructure components in a combined sewer area.” Monograph 19. ISBN 978-0-9808853-4-7. In Modeling Urban Water Systems. Cognitive Modeling of Urban Water Systems, edited by W. James, K.N. Irvine, James Y. Li, E.A. McBean, R.E. Pitt, and S.J. Wright, 139–56. Guelph: Computational Hydraulics Int.
29. Rahmana, M.A., A. Moserb, M. Andersona, C. Zhangb, T. Rötzerb, and S. Pauleita. 2019. “Comparing the infiltration potentials of soils beneath the canopies of two contrasting urban tree species.” Urban Forestry and Urban Greening 38: 22–32. https://doi.org/10.1016/j.ufug.2018.11.002
30. Roy, S., J. Byrne, and C.A. Pickering. 2012. “A systematic quantitative review of urban tree benefits, costs, and assessment methods across cities in different climatic zones.” Urban Forestry & Urban Greening 11 (4): 351–63.
31. Schooling, J.T., and D.E. Carlyle-Moses. 2015. “The influence of rainfall depth class and deciduous tree traits on stemflow production in an urban park.” Urban Ecosystems 18: 1261–84. https://doi.org/10.1007/s11252-015-0441-0
32. Shergill, S.S. 2004. Quantification of Escherichia Coli and Enterococci Levels in Wet Weather and Dry Weather Flows. Tuscaloosa, Al: University of Alabama Department of Civil and Environmental Engineering. MSCE Thesis. http://unix.eng.ua.edu/~rpitt/Publications/11_Theses_and_Dissertations/Shergill_Bacteria_sources_thesis.pdf
33. Smets, V., W. Akkermans, B. Verbeiren, M. Hermy, and B. Somers. 2019. “Ex-situ estimation of interception storage capacity of small urban plant species.” Journal of Hydrology 572: 869–83. https://doi.org/10.1016/j.jhydrol.2019.03.047
34. Van Stan II, J.T., D.F. Levia Jr., and R.B. Jenkins. 2015. “Forest canopy interception loss across temporal scales: Implications for urban greening initiatives.” The Professional Geographer 67 (1): 41–51. https://doi.org/10.1080/00330124.2014.888628
35. Xiao, Q., E.G. McPherson, S.L. Ustin, M.E. Grismer, and J.R. Simpson. 2000. “Winter rainfall interception by two mature open-grown trees in Davis, California.” Hydrological Processes 14: 763–84.
36. Zabret, K., J. Rakovec, and M. Šraj. 2018. “Influence of meteorological variables on rainfall partitioning for deciduous and coniferous tree species in urban area.” Journal of Hydrology 558: 29–41. https://doi.org/10.1016/j.jhydrol.2018.01.025

Appendices

Appendices are in the PDF version of the paper, available here: https://www.chijournal.org/Journals/PDF/C475