See discussions, stats, and author profiles for this publication at: https://www.researchgate.net/publication/260683953

Meteorological influences on stemflow generation across diameter size classes of two morphologically distinct species

Article in International Journal of Biometeorology · March 2014 DOI: 10.1007/s00484-014-0807-7 · Source: PubMed

CITATIONS READS 13 52

3 authors, including:

John Toland Van Stan Jarrad Van Stan Georgia Southern University Harvard Medical School, Massachusetts Gen…

64 PUBLICATIONS 398 CITATIONS 25 PUBLICATIONS 105 CITATIONS

SEE PROFILE SEE PROFILE

Some of the authors of this publication are also working on these related projects:

Canopy influences over microbial processes in critical zones View project

Effects of Urbanization on Forest Ecohydrology View project

All content following this page was uploaded by John Toland Van Stan on 16 July 2015.

The user has requested enhancement of the downloaded file. Int J Biometeorol (2014) 58:2059–2069 DOI 10.1007/s00484-014-0807-7

ORIGINAL PAPER

Meteorological influences on stemflow generation across diameter size classes of two morphologically distinct deciduous species

John T. Van Stan II & Jarrad H. Van Stan & Delphis F. Levia Jr.

Received: 15 November 2013 /Revised: 11 February 2014 /Accepted: 19 February 2014 /Published online: 11 March 2014 # ISB 2014

Abstract Many tree species have been shown to funnel sub- meteorological conditions). Multiple regressions performed on stantial rainfall to their stem base as stemflow flux, given a leafless canopy stemflow resulted in an inverse relationship favorable stand structure and storm conditions. As stemflow is with wind speeds, likely decoupling stemflow sheltered solely a spatially concentrated flux, prior studies have shown its on surfaces from VPD influences. presence generally impact on ecohydrological and biogeochemical processes increased direct stemflow associations with rainfall intensity, can be significant. Less work has been performed examining yet diminished stemflow-rainfall relationships. F. grandifolia stemflow variability from meteorological conditions com- canopies (exemplifying structures of smoother bark and greater pared to canopy structural traits. As such, this study performs branch angle) strengthened differences in stemflow associations multiple regressions: (1) to examine stemflow variability due with rainfall/mean wind speed between leaf states. These find- to event-based rainfall amount, intensity, mean wind speeds, ings are placed in a conceptual interception loss path analysis, and vapor pressure deficit; (2) across three diameter size which shows the potential to alter common interception loss classes (10–20, 21–40, and >41 cm DBH); and (3) for two estimates in high stemflow stands. common tree species in the northeastern USA of contrasting canopy morphology—Liriodendron tulipifera L. (yellow pop- Keywords Stemflow . Fagus grandifolia . Liriodendron lar) versus Fagus grandifolia Ehrh. (American beech). On the tulipifera . Canopy structure . Meteorological conditions . whole, multiple regression results yielded significant positive Tree size correlations with stemflow for rainfall amount, intensity, and mean wind speed and a significant negative correlation for vapor pressure deficit (VPD). Tree size altered stemflow- Introduction meteorological condition relationships, where larger trees strengthened indirect stemflow-VPD and direct stemflow- Forest canopy cover drastically alters hydrologic cycling at rainfall and stemflow-intensity associations. Canopies of rough- the surface through the physical interception and rerouting of er bark and lower branch angle (represented by L. tulipifera) inputs. Much of the intercepted rainfall, for in- enhanced correlations for nearly all meteorological conditions stance, is stored on canopy surfaces until evaporation back to via greater stemflow residence time (and longer exposure to the atmosphere (Carlyle-Moses and Gash 2011). The remain- ing droplets reach the forest floor via leaf and branch J. T. Van Stan II (*) pathways as a drip flux (throughfall) or a funneled flow down Department of Geology and Geography, Georgia Southern the main trunk (stemflow) (Levia et al. 2011). Of these two University, Statesboro, GA 30640, USA rainfall redistribution pathways, throughfall covers a larger e-mail: [email protected] infiltration area and represents the dominant proportion of J. H. Van Stan rainfall: approximately 75 % on average (Llorens and Center for Interprofessional Studies and Innovation, MGH Institute Domingo 2007; Levia et al. 2011). Throughfall has also been of Health Professions, Boston, MA 02114, USA linked to spatial patterns in moisture (Raat et al. 2002; Guswa 2012), fine (Ford and Deans 1978), groundwater D. F. Levia Jr. Departments of Geography and & Soil Sciences, University of recharge (Guswa and Spence 2012), and streamwater chem- Delaware, Newark, DE 19716, USA istry (Inamdar and Mitchell 2007). As a result, throughfall has 2060 Int J Biometeorol (2014) 58:2059–2069 historically been more researched than stemflow (Levia and Kuraji et al. 2001; Levia et al. 2011) and prevailing wind Frost 2003; Levia et al. 2011). Yet, for particular species or directions show the potential to further augment stemflow storm conditions, the stemflow process is capable of applying yield by preferentially saturating dominant canopies substantial loads over small infiltration areas around the (Herwitz and Slye 1995) or canopies of a particular structure stem (e.g., Herwitz 1986;Germeretal.2010; Van Stan and (Van Stan et al. 2011). Atmospheric moisture conditions— Levia 2010). Where this occurs, researchers have linked i.e., vapor pressure deficit (VPD)—are also expected to stemflow to the distribution of epiphytes and forest floor influence stemflow generation as Llorens et al. (1997) vegetation (Crozier and Boerner 1984; Hauck et al. 2002), noted enhanced evaporation under drier atmospheric runoff generation (Neave and Abrahams 2002), preferential conditions (which could diminish water entrained on groundwater recharge (Durocher 1990;Liangetal.2011), canopy surfaces contributing to stemflow). Muzylo soil solution chemistry (Chang and Matzner 2000), and et al. (2012) also suggested the need for analysis on hydropedological processes (Li et al. 2009)—increasing stemflow generation including VPD, indicating that it interest from the scientific community regarding the investi- may account for some of the high variability observed gation of factors controlling stemflow variability (e.g., Germer in past research (e.g., Staelens et al. 2008). et al. 2010; Levia et al. 2011). This study seeks to examine stemflow variability due to Stemflow variability has been linked to canopy structure storm meteorological conditions across three diameter size and meteorological conditions (Levia et al. 2011; Pypker et al. classes (10–20, 21–40, and >41 cm DBH) for two species of 2011). Structural traits within tree canopies—bark microrelief, contrasting canopy morphology: the rough-bark, low branch leaf and gap patterns, branch angle, etc.—are naturally het- angle, thin canopy of Liriodendron tulipifera L. (tulip poplar) erogeneous across space and time, which can hamper or versus the smooth-bark, high branch angle, thick canopy of facilitate rain droplet capture and entrainment as stemflow Fagus grandifolia Ehrh. (American beech). Specific questions (Crockford and Richardson 2000; Van Stan and Levia 2010; to address with multivariate regressions include the following: Frost and Levia 2014). Intuitively, trees of larger diameter at (1) How do storm meteorological variables influence breast height (DBH), projected area, and canopy den- stemflow from these species? Moreover, which meteorologi- sity can generate greater stemflow volumes (Ford and Deans cal factors exert greater control over stemflow after account- 1978; Aboal et al. 1999; Park and Hattori 2002; Pypker et al. ing for rainfall amount? (2) Do individual meteorological 2011). However, this simple tree size-to-stemflow yield rela- variables controlling stemflow, or the order of influence tionship varies considerably with the geometric orientation of among these variables, differ between the two morphological- branching and leaf surfaces, where more steeply inclined ly contrasting species? (3) Are meteorological controls differ- branches and (erectophile) enhance stemflow genera- ent across trees of different sizes regardless of species? An- tion compared to more moderately (plagiophile), zero- swers to these questions would improve our understanding of (planophile), or negatively sloped branching patterns trunk storage and drainage processes in interception loss (Hutchinson and Roberts 1981; André et al. 2008; Van Stan models—e.g., the most commonly used Rutter- and Gash- et al. 2011; Frost and Levia 2014). Stemflow generation can type models (Rutter et al. 1971;Gashetal.1995; Valente be further affected by bark structure of the branches and trunk. et al. 1997)—and stemflow’s interactions with reliant Smoother bark and bark with tighter, high frequency ridge ecohydrological processes. patterns have lower water storage capacities, allowing earlier (and greater) drainage of rain droplets entrained on the branches as branchflow and ultimately stemflow (Brown and Site description Barker 1970; Levia et al. 2010; Van Stan and Levia 2010). A direct relationship between stemflow volume and rainfall Stemflow and meteorological measurements were made at the amount has been well established (e.g., Clements 1972; André Fair Hill Natural Resources Management Area (NRMA) from et al. 2008; Van Stan and Levia 2010). However, stemflow July 2007 through April 2008. The experimental forest plot yield responds to many more meteorological factors. Past is at 72 m asl in northeastern Maryland, USA at the studies found stemflow volumes negatively corresponded to Pennsylvania-Delaware-Maryland border (39° 42′ N, 75° 51′ storm intensity, indicating that heavy rainfall may overload W). Fair Hill NRMA site’s close proximity to large water stemflow pathways through the canopy (Carlyle-Moses and bodies (Delaware and Chesapeake Bays) results in a humid Price 2006; Staelens et al. 2008). Past research also identified maritime climate, where 30-year mean annual precipitation is wind influences, linking greater stemflow yields to higher about 1,200 mm and most rainfall is delivered via frontal wind speeds (Xiao et al. 2000;Kurajietal.2001)andto storms between the months of September and May particular wind directions (Van Stan et al. 2011). High wind (Maryland State Climatologist Office 2012). June–August speeds during rainfall may wet larger portions of the tree rainfall events are primarily convective rainfall (Maryland canopy, boosting stemflow production (Xiao et al. 2000; State Climatologist Office 2012). Regardless, rainfall is Int J Biometeorol (2014) 58:2059–2069 2061 relatively evenly distributed throughout the year, with winter Stemflow monitoring being the driest season and generating slight, short-lived snowpack (Maryland State Climatologist Office 2012). Be- Stemflow monitoring began July 2007 and ended in April tween 1981–2010, highest mean monthly air temperature was 2008, resulting in the collection of stemflow from 33 discrete during July (25.7 °C), whereas the coldest mean monthly air storm events. Thirty trees—fifteen trees of each co-dominant temperature occurred during January (−0.1 °C) (Maryland species divided equally into three size classes (10–20 cm State Climatologist Office 2012). The 33 storms monitored DBH, 20.1–40 cm DBH, and >40 cm DBH)—were instru- in this study spanned a wide range of meteorological condi- mented with stemflow collars connected to plastic collection tions experienced at the site, ranging between 0.8–36 mm bins. Stemflow collars were fashioned out of 3.2 cm outer rainfall, 2–34 h duration, 0.1–4.1 mm h−1 intensity, 0.2– diameter vinyl tubing that had been longitudinally cut and 7.5 m s−1 mean event wind speed, and 0.04–0.49 kPa vapor connected to the tree bole at a moderate slope to ensure pressure deficit. complete drainage of captured stemflow water. Silicon sealant Upper canopy cover of the study plot is co-dominated was used to seal the edges of these vinyl stemflow collars by F. grandifolia and L. tulipifera. Mean diameter at against the bark surface. Each stemflow collar end was breast height (DBH) for trees >10 cm DBH ranges from inserted into two plastic collection bins connected in series 14.4–48.6 cm for F. grandifolia and 16.5–67.5 cm for so that large stemflow events could overflow from the initial L. tulipifera across diameter classes, respectively. Plant bin into the secondary bin. Calibration curves were generated area index ranges from 1.2–5.3 m2 m–2 for leafless and for the stemflow collection bins so that millimeter depths leafed periods. These co-dominant tree species drastically could be converted to liters of stemflow. After each storm differ in canopy morphology: with L. tulipifera canopies event, stemflow volumes within the collection bins were having thinner canopies of rougher and thicker bark, measured with rulers and emptied. lower branch inclination, and lower leaf area index com- pared to F. grandifolia. L. tulipifera’ s bark form and texture Multiple regression analyses also become rougher and more furrowed with increasing DBH, resulting in a higher overall bark microrelief for Thirty-six multiple regression analyses evaluated the simulta- L. tulipifera than for F. grandifolia trees of similar size neous effects of four meteorological variables (rainfall (Van Stan and Levia 2010). Greater detail on these canopy amount, intensity, mean wind speed, and VPD) on stemflow structural differences may be found in Van Stan and Levia from (1) all thirty trees for all storms; (2) ten small, ten (2010). medium, and ten large trees for all storms; (3) all fifteen beech and (4) yellow poplar trees for all storms; (5) five small, five medium, and five large beech and (6) poplar trees for all Materials and methods storms; and (7) then repeated these analyses for seventeen leafed season and sixteen leafless season storms. Since mete- Meteorological monitoring orological variables act upon stemflow in a simultaneous, not sequential, fashion, it was decided that a standard multiple Meteorological conditions were continuously monitored by a regression would be more appropriate to statistically analyze meteorological station from the Delaware Environmental Ob- their effects instead of a hierarchical approach. Step-wise serving System (DEOS) situated in Fair Hill NRMA about multiple regression analysis was also not chosen due to this 0.5 km from the experimental plot. Gross precipitation was method’s well-established pitfalls, where the “design” of a collected by a Geonor T-200B vibrating wire gauge (Oslo, best set of predictors through parameter interference can, Norway) equipped with an alter-type wind shield. A discrete oddly enough, interfere with the model itself by producing storm event was considered any continuous rainfall preceded parameter biases, over-fitting, and altered significance by a minimum eight hour antecedent dry period that resulted (Devore and Peck 1993). Before attempting the multiple re- in measurable stemflow generation from the subject trees. An gressions, bivariate Pearson correlations were completed eight hour minimum dry period was selected as this is the using all dependent and independent variables to ensure min- minimum time necessary to allow bark to dry based on obser- imal collinearity (Devore and Peck 1993). Assumptions re- vational evidence (Van Stan and Levia 2010). Meteorological garding normality of distribution and homoscedasticity were conditions also monitored include wind speed and maximum also confirmed (Devore and Peck 1993). After the four mete- gust (m s−1), wind direction (°) air temperature (°C), and orological variables were regressed with stemflow, any non- relative (%). Vapor pressure deficit (VPD) was cal- significant contributing independent variables were removed culated using temperature and relative humidity data to deter- and the multiple regression was computed again. If the overall mine the difference between actual vapor pressure and satu- relationship weakened significantly (meaning one of the pre- ration vapor pressure. vious significant contributing variables became nonsignificant 2062 Int J Biometeorol (2014) 58:2059–2069 or the overall r2 decreased 10 % or more), the original multiple Table 1 Multiple regression r values and standard errors (SE) for differ- regression using all meteorological variables was recorded as ing size trees of both species and each species across all seasonal condi- tions (annual, leafed, and leafless) the final result. If the overall relation did not change (meaning all previous significant contributing variables remained sig- Both spp. L. tulipifera F. grandifolia nificant and the overall r2 decreased less than 10 %), then the r SE r SE r SE multiple regression using the least amount of meteorological variables was recorded as the final result. Evaluation of the Annual multivariate regression models was not performed as our All sizes 0.64 0.061 0.77 0.077 0.61 0.067 objective was not to generate a quantitative predictive model Small 0.68 0.088 0.71 0.131 0.67 0.112 for stemflow yield from these trees and species. Rather, this Medium 0.64 0.098 0.76 0.137 0.75 0.069 study seeks to use multivariate regression analysis to investi- Large 0.66 0.114 0.86 0.118 0.79 0.072 gate the previously unexamined holistic influence of dominant Leafed storm-based meteorological conditions (rainfall amount, All sizes 0.67 0.087 0.84 0.093 0.66 0.089 VPD, mean wind speed, and rainfall intensity) and, thereby, Small 0.75 0.117 0.80 0.168 0.75 0.142 generate a conceptual model of how the stemflow hydrologic Medium 0.69 0.146 0.87 0.162 0.80 0.086 process is influenced by these conditions within canopies of Large 0.63 0.171 0.88 0.134 0.85 0.094 contrasting architecture. Such conceptual models are funda- Leafless mental to the development of process parameterizations and All sizes 0.64 0.081 0.74 0.106 0.59 0.101 the incorporation of these formulas into existing interception Small 0.67 0.121 0.67 0.178 0.69 0.166 loss models for improved interception loss estimation in high Medium 0.65 0.120 0.72 0.175 0.73 0.106 stemflow-yield stands. Large 0.69 0.147 0.86 0.169 0.79 0.095

Results amount on stemflow generation from all trees, per variable- specific β weight, was greatest for the large size class, which Overall multiple regression results was also nearly equivalent to the small size class (Table 2). After accounting for rainfall amount, the dominant trend in the All multiple regression r values for all conditions were greater magnitude of each variable’s effect on stemflow volume be- than 0.6, except for the leafless F. grandifolia condition neath all trees was as follows: VPD > mean wind speed > including tree subjects from all size classes (Table 1). rainfall intensity (Table 2). VPD and wind speed effects were Multiple regressions for leafed middle-sized L. tulipifera opposite in their relationship to stemflow yield (Table 2). stemflow generation against meteorological conditions Rainfall intensity, however, accounted for very little stemflow produced the largest r values and standard errors (Table 1). variability, was insignificant for all trees in the small size class Multiple regressions resulted in greater r and standard and was not included in the “all trees” large size class analysis errors for L. tulipifera stemflow generation for all trees due to a combination of insignificant β weight and false regardless of size class compared to F. grandifolia (Table 1). enhancement of mean wind speed’s effect and significance ’ The large size class F. grandifolia trees multiple regression (Table 2). Mean wind speed and VPD most affected the large results produced the greatest correlation with meteorolog- size class of all trees; whereas, when rainfall intensity was ical conditions for that species (Table 1). The small size statistically relevant, it exerted the strongest influence on the F. grandifolia class had the smallest r, but largest standard medium size class for all trees (Table 2). For the small size error, for that species (Table 1). Regardless of species (see class of all monitored trees, mean wind speed trumped the β both spp. column of Table 1), standard error of the multiple weight and correlation of VPD (Table 2). regression increased with size class (Table 1).

Individual meteorological condition influence on stemflow Interspecific differences in meteorological influences over stemflow Multiple regression results for all trees in all size classes identified all four storm-based meteorological variables as All F. grandifolia canopies’ stemflow generation correlated, significant and accounting for some level of variability in almost singularly, to rainfall amount (Table 3). Only a small the stemflow dataset (Table 2). Rainfall amount, as was ex- effect (0>β>−0.05), but statistically significant (0.1>p≥ pected, accounted for the majority of stemflow variability 0.05), could be assigned to any other storm-based meteoro- from all monitored trees and produced the strongest correla- logical condition, specifically VPD (Table 3). This phenome- tions across size classes (Table 2). Relative “effect” of rainfall non is best represented by the regression results from the Int J Biometeorol (2014) 58:2059–2069 2063

Table 2 Overall multiple regression results for all trees regardless of species

r β

Variables All Small Medium Large All Small Medium Large

Rainfall (mm) 0.41*** 0.43*** 0.40*** 0.57*** 0.55 0.59 0.54 0.60 Rain intensity (mm hr−1)0.05**– 0.08* NI 0.07 0.09 0.11 NI Mean wind speed (m s−1) 0.11*** 0.11** 0.08* 0.14** 0.13 0.13 0.09 0.16 Vapor pressure deficit (kPa) −0.11*** −0.08** −0.10** −0.17*** −0.14 −0.10 −0.13 −0.19

Level of statistical significance for each variable’scomponentr is indicated by superscripts. Dashes indicate insignificant variable r,and“NI” shows when statistically insignificant variables were not included in analysis due to their inclusion falsely enhancing statistical significance of other significant variables *0.1>p≥0.5; **0.05>p≥0.01; ***p<0.01

F. grandifolia small size class where all other variables were observed for all trees and F. grandifolia,wereoppositeintheir insignificant (p>0.1), as to not merit inclusion (Table 3). relationship to stemflow yield (Table 3). In the overall Influence of atmospheric moisture conditions increases with L. tulipifera analysis (without regard to size), variable rainfall the medium and large F. grandifolia size classes as indicated intensity represented little effect on stemflow and was insig- by growing β and r values (Table 3). VPD even goes from low nificant or not included in the small and large size class (0.1>p≥0.05) to high (p<0.01) statistical significance with analyses, respectively (Table 3). Rainfall intensity most affect- regard to stemflow generation between the medium and large ed stemflow from the medium L. tulipifera size class (Table 3). size F. grandifolia classes, respectively (Table 3). The general Stemflow produced from large L. tulipifera trees appear to effect of rainfall intensity on F. grandifolia stemflow genera- have the strongest, and most significant (p<0.01),relationship tion, although statistically irrelevant for the small and medium with atmospheric moisture conditions and mean wind speeds size classes, becomes significant for the large size class of all the size classes (Table 3). (Table 3). For all storm conditions except rainfall amount, L tulipifera L. tulipifera stemflow volumes for all trees significantly stemflow production more strongly correlated than for correlated with all four conditions represented in the multiple F. grandifolia stemflow (Table 3). This is particularly evident regression (Table 3). The overall trend regarding the magni- in the “all” and small size class results where rainfall amount tude of each meteorological variable’s relative effect and is the primary—and nearly sole—meteorological driver of statistical significance, after the well-established connection F. grandifolia stemflow production, yet L. tulipifera stemflow to rain amount, was as follows: VPD ≥ mean wind speed >> variability is significantly correlated to at least two more storm rain intensity (Table 3). The effects of VPD and wind speed, as conditions beyond rainfall amount (Table 3). The relative

Table 3 Multiple regression results per tree species across the size classes

r β

Variables All Small Medium Large All Small Medium Large

Fagus grandifolia Rainfall (mm) 0.40*** 0.67*** 0.54*** 0.50*** 0.61 0.67 0.69 0.68 Rain intensity (mm hr−1) – NI – 0.09* 0.08 NI 0.08 0.12 Mean wind speed (m s−1) – NI NI – 0.06 NI NI 0.09 Vapor pressure deficit (kPa) −0.07* NI −0.09* −0.15*** −0.03 NI −0.10 −0.19 Liriodendron tulipifera Rainfall (mm) 0.48*** 0.44*** 0.45*** 0.71*** 0.65 0.60 0.61 0.74 Rain intensity (mm hr−1)0.05* – 0.11** NI 0.07 0.09 0.15 NI Mean wind speed (m s−1) 0.16*** 0.13** 0.13** 0.23*** 0.20 0.16 0.16 0.27 Vapor pressure deficit (kPa) −0.17*** −0.14** −0.13** −0.25*** −0.21 −0.17 −0.17 −0.29

Level of statistical significance for each variable’scomponentr is indicated by superscripts. Dashes indicate insignificant variable r,and“NI” shows when statistically insignificant variables were not included in analysis due to their inclusion falsely enhancing statistical significance of other significant variables *0.1>p≥0.5; **0.05>p≥0.01; ***p<0.01 2064 Int J Biometeorol (2014) 58:2059–2069 stemflow response to rainfall amount was about equal be- relationship, and its significance, between mean wind speed tween species regardless of canopy structure despite stemflow and stemflow production for all trees also diminished after leaf magnitudes significantly differing (Table 3). Mean wind speed abscission, particularly for the medium and large size classes effect was only statistically relevant, particularly in the large (Fig. 1c). This wind-stemflow volume interaction actually size class, for L. tulipifera (Table 3). β weight for the VPD reversed between leafed and leafless states, with mean wind influence on stemflow generation from all L. tulipifera trees speed being inversely related to stemflow generation within was seven times, and two significance levels, greater than for bare canopies and positively correlated within leafed canopies all F. grandifolia trees (Table 3). The VPD effect on stemflow (Fig. 1c). A reversal of relationship was also the result of production remained lower for medium F. grandifolia trees in multivariate regressions across leaf states for VPD, where comparison to L. tulipifera of the same size class, but this leafless stemflow production became directly associated with comparison was less drastic (Table 3). Stemflow generated drier atmospheric conditions (Fig. 1d). The curious leafless from large trees of both species showed the highest β weights, season stemflow-VPD results were, however, of lower effect, correlation, and significance of all size classes (Table 3). correlation, and significance for all trees and size classes Interestingly, rainfall intensity played an insignificant role in besides the small trees (Fig. 1d). stemflow variability from large L. tulipifera trees but ex- As seen for all trees, stronger effects of similar (high) plained some portion of stemflow variability for the smooth- significance were observed for rainfall amount with regard barked, steep-angled branching structure of large F. grandifolia to stemflow generation from leafless, compared to leafed, trees (Table 3). F. grandifolia canopies across size classes (Fig. 2a). Stemflow from small L. tulipifera trees during the leafed season, how- Leaf state alterations to -stemflow interactions ever, exhibited greater correlation and slightly greater effect from rainfall amounts than when leafless (Fig. 3a). All other The presence or absence of leaves modified variable-specific L. tulipifera size classes conformed to the trend of enhanced β weights, correlation, and the significance of these effects for stemflow generation under leafless conditions (Fig. 3a). all trees across species (Fig. 1). Effect of rainfall amount on Greater effects and correlations with rainfall amount during stemflow production, for instance, on all trees across all size the leafed season were generally observed for L. tulipifera classes strengthened with leaf absence despite lower correla- than F. grandifolia, particularly in the results from all trees, tions (Fig. 1a). Conversely, Fig. 1b shows that multivariate and the large and small size classes (Figs. 2a and 3a). Greater regression results were insignificant for rainfall intensity’s rainfall-stemflow relationships in the leafless season were influence over leafless stemflow generation for all analyses relatively evenly split between the species, where L. tulipifera barring the small size class, where β, r,andp values were trumped F. grandifolia in the all and large tree models, yet this lower than observed during the leafed season. The interspecific comparison reverses for the medium and small

Fig. 1 Comparison of leafed and leafless condition β weights relating a rainfall amount (mm), b rain intensity (mm hr−1), c mean event wind speed (m s−1), and d vapor pressure deficit (kPa) to stemflow generation from all trees and size classes regardless of species. Variable-specific r values are provided (when significant) with statistical significance level indicated by superscripts,where 0.1>p≥0.5, 0.05>p≥0.01, and p<0.01 are denoted by a, b,andc, respectively. Two dashes indicate insignificant variable-specific r values, and an asterisk shows when statistically insignificant variables were not included in analysis due to their inclusion falsely enhancing statistical significance of other significant variables Int J Biometeorol (2014) 58:2059–2069 2065

Fig. 2 Comparison of leafed and leafless condition β weights relating a rainfall amount (mm), b rain intensity (mm hr−1), c mean event wind speed (m s−1), and d vapor pressure deficit (kPa) to stemflow generation from all F. grandifolia trees and size classes. Variable-specific r values are provided (when significant) with statistical significance level indicated by superscripts,where 0.1>p≥0.5, 0.05>p≥0.01, and p<0.01 are denoted by a, b,andc, respectively. Two dashes indicate insignificant variable-specific r values, and an asterisk shows when statistically insignificant variables were not included in analysis due to their inclusion falsely enhancing statistical significance of other significant variables

tree models (Figs. 2a and 3a). No significant relationship was strengthened in effect and correlation with increasing found for rainfall intensity during the leafless season for F. grandifolia size class. A clear leafed canopy stemflow- stemflow volume from either species despite statistically rel- VPD relationship with regard to tree size was not apparent evant interactions during the leafed season (Figs. 2b and 3b). for L. tulipifera (Fig. 3d). As seen in the analysis of all trees Leafed canopy stemflow generation showed the strongest regardless of species, multivariate results for F. grandifolia positive relationship with rainfall intensity for the large and L. tulipifera leafless season stemflow generation indicate F. grandifolia (Fig. 2a) and medium L. tulipifera trees a positive statistically significant correlation with VPD (Fig. 3b). (Figs. 2d and 3d). Figure 2c shows leafless F. grandifolia canopies’ stemflow is generally reduced with increasing mean wind speed during rainfall for all trees and all size classes, while wind tends to Discussion increase stemflow generation for leafed canopies. The nega- tive leafless stemflow-wind correlation grows stronger with Meteorological influences over stemflow generation decreasing F. grandifolia size class (Fig. 2c). This trend is relatively apparent for L. tulipifera as well, where mean wind Multivariate regression results from F. grandifolia and speed went from insignificant and falsely enhancing other L. tulipifera stemflow indicate clear influences of meteorolog- variables (therefore, not included in analysis) for the large size ical conditions beyond the commonly reported correlation class to just insignificant in the medium size class and to low with rainfall amount (Tables 1 and 2; Figs. 1, 2, and 3). statistical significance in the small size class (Fig. 3c). The Further, there appear to be shared meteorological influences mean wind speed effect on L. tulipifera stemflow was quite over stemflow generation across tree sizes and the contrasting similar in β, r,andp value across size classes during the leafed canopy architectures represented by the rough-barked, low season (Fig. 3c). F. grandifolia leafless canopy stemflow cor- branch angle, thinner L. tulipifera and oppositely structured related more strongly with mean wind speed than was ob- F. grandifolia canopies (Table 2; Fig. 1). These shared results served for L. tulipifera regardless of size class (Figs. 2c and are compiled into a conceptual path analysis adapted from the 3c). Oppositely, L. tulipifera stemflow-wind interactions were Valente et al. (1997) reformulated Rutter interception model greater (and more consistent across tree sizes) during the framework to illustrate their potential impact on interception leafed season than for F. grandifolia (Figs. 2c and 3c). Trends loss modeling in stands of high stemflow generation (Fig. 4). in stemflow reduction due to VPD during the leafed season The most significant shared influence of storm conditions on were weaker for F. grandifolia across all size classes (Fig. 2d) stemflow production was the negative correlation with atmo- than for L. tulipifera (Fig. 3d), although this relationship spheric dryness, VPD (Table 2). Staelens et al. (2008)also 2066 Int J Biometeorol (2014) 58:2059–2069

Fig. 3 Comparison of leafed and leafless condition β weights relating a rainfall amount (mm), b rain intensity (mm hr−1), c mean event wind speed (m s−1), and d vapor pressure deficit (kPa) to stemflow generation from all L. tulipifera trees and size classes. Variable-specific r values are provided (when significant) with statistical significance level indicated by superscripts,where 0.1>p≥0.5, 0.05>p≥0.01, and p<0.01 are denoted by a, b,andc, respectively. Two dashes indicate insignificant variable-specific r values, and an asterisk shows when statistically insignificant variables were not included in analysis due to their inclusion falsely enhancing statistical significance of other significant variables

found this significant negative stemflow-VPD interaction, between wind conditions and stemflow during leafless periods and, in our study, it was not solely restricted to the overall while partially accounting for the positive correlations ob- (across species) analyses, but is also apparent in the individual served between stemflow generation and VPD under leafless species’ multivariate regression results (Tables 2 and 3). Al- conditions (Fig. 4b). though VPD is low during rainfall (0.04–0.5 kPa), this atmo- Mean wind speed generally enhanced stemflow generation spheric water demand could be enough to divert canopy and from both species and across all size classes, agreeing with trunk storage water from their respective drainage components observations and results from other studies (Herwitz and Slye during rainfall, or shortly after the event while stemflow 1995; André et al. 2008; Van Stan et al. 2011;Muzyloetal. continues to drain (Fig. 4a). Increased interception losses from 2012). Greater wind speeds may boost stemflow volumes by enhanced VPD have been observed by others (Llorens et al. saturating larger portions of the tree canopy through the scat- 1997; Staelens et al. 2008). As VPD’s inverse control over tering of rain into finer droplets and applying them vertically stemflow was most influential during the leafed season—and to leaf and bark surfaces (Xiao et al. 2000;Kurajietal.2001). the positive VPD-stemflow relationship during the leafless Greater rainfall diversion to the canopy and trunk storage season (Figs. 1d, 2d, 3d)—it is likely that stemflow generation components likely heightened stemflow production at the is primarily reduced by evaporation of water between the expense of the throughfall drainage pathway (Fig. 4a). Wind canopy storage-to-drainage pathway, rather than the trunk speeds overall, therefore, reinforced the canopy storage-to- storage-to-drainage pathway (Fig. 4a). This VPD-related trunk drainage pathway (Fig. 4a). The enhanced effect of wind stemflow reduction pathway is also supported by past research speed on leafless season stemflow, however, was not observed indicating that trunk evaporative losses may be lower due to in this study (Figs. 1c, 2c, 3c) as it was observed in others bark structure’s more effective use of surface tension to entrain (André et al. 2008). Rather, mean wind speed negatively rain droplets in comparison to leaves lower microrelief sur- correlated with leafless season stemflow generation—and less faces (Liu 1998; Llorens and Gallart 2000;LeviaandHerwitz significantly than for the positive leafed season correlation 2000, 2005). (Figs. 1c, 2c, 3c)—indicating mean wind speed for these study Bark microrelief likely provides a better sheltered micro- species may mechanically divert water ensnared along the climate, whereas the leaves are less likely to resist atmospheric canopy storage-to-trunk drainage pathway to the throughfall water demands (Fig. 4a). It is plausible that a combination of generation pathway via wind sway (Fig. 4b). Rainfall intensity bark “sheltering” of stemflow pathways (particularly in deeper produced the lowest effect on, and correlation with, stemflow furrows where surface tension and capillary forces are stron- generation of all the storm-based meteorological conditions, ger) and lower winds could decouple stemflow generation oftentimes being insignificant or excluded due to false alter- from VPD. This may explain the observed inverse correlation ation of other variables (Tables 1 and 2). Int J Biometeorol (2014) 58:2059–2069 2067

Fig. 4 Conceptual path analysis overlaying stemflow relationships ob- season conditions, as well as b the leafless condition. Line color indicates served in this study between VPD (event vapor pressure deficit), I (event direction of the relationship (red=inverse, blue=direct) and line direction rainfall intensity), and W (mean event wind speed), on the Rutter and indicates the canopy hydrological component to which potential Gash interception models’ framework of canopy-based hydrological stemflow water was diverted or enhanced by the interaction of these processes (adapted from Valente et al. (1997)) for a annual and leafed meteorological conditions

Many have reported a negative correlation between along the stemflow pathway (Fig. 4a). With decreasing size stemflow and storm intensity with multivariate regression, class, trees are more sheltered from meteorological conditions suggesting that heavy rainfall may overload stemflow path- measured outside the canopy, diminishing relationships with ways through the canopy (Crockford and Richardson 2000; VPD (Tables 1 and 2; Figs. 1d, 2d, 3d in leafed season). For Levia and Frost 2003; Carlyle-Moses and Price 2006). Our that same reason, dominant canopies of large trees have been results, at the storm scale, found the opposite: stemflow pro- shown to more efficiently divert rain to stemflow under high duction increased with rainfall intensity regardless of canopy winds (Herwitz and Slye 1995; Levia et al. 2011). Interesting- structural characteristics (Tables 1 and 2;Figs.1, 2,and3). ly, the relationship between tree size, stemflow generation and However, this positive correlation does not mean that mean wind speed appears to be the opposite to that of rainfall stemflow pathways may not become overloaded and diminish amount and VPD, most notably in the leafless season where the efficiency of stemflow drainage pathways, resulting in wind is negatively correlated to stemflow (Figs. 1c, 2c, 3c). diminished stemflow partition compared to throughfall and The diminishing effect of wind speed on stemflow as tree size interception loss partitions. In fact, Levia et al. (2010) increases is not very apparent in annual analyses (Tables 1 and witnessed a sharp decrease in stemflow funneling ratios as 2), but much clearer in leaf state analyses (Figs. 1c, 2c, 3c). the amount of rain increased in a 5-minute period. But, more During the leafless season, where wind appears to have the intense rain can enhance stemflow production along the can- largest toll on stemflow volumes, larger trees may not be opy storage-to-trunk drainage pathway (Fig. 4a–b), despite shaken as easily as smaller trees (which are more flexible). lowering its proportion of precipitation (Levia et al. 2010). Thus, the larger the tree, the less likely it is to give up canopy Rainfall intensity correlations with stemflow are also, in part, and trunk drainage water to the throughfall pathway from obscured for medium and smaller canopy trees through con- wind-related mechanical manipulations (Fig. 4b). tact with larger trees altering original intensities (Tables 1 and 2; Fig. 2b). Interaction of species-specific canopy morphology It is likely that tree size influenced correlations between with meteorological controls over stemflow generation stemflow and rainfall intensity and amount, VPD, and mean wind speed. Larger trees, for example, are most exposed to The canopy structure of L. tulipifera appears to enhance storm conditions, likely enhancing the effect of atmospheric meteorological controls over stemflow generation in compar- moisture conditions (Tables 1 and 2;Figs.1d, 2d, 3d in leafed ison to F. grandifolia. This is probably on account of season). Enhanced exposure to atmospheric moisture condi- L. tulipifera producing a thick-limbed branching architecture tions increases opportunity for evaporative loss from trunk of low branch angle and high bark microrelief which impedes and canopy storage which would otherwise have drained trunk drainage and could result in long hydraulic retention 2068 Int J Biometeorol (2014) 58:2059–2069 times for trunk drainage water on the bark surfaces (Levia relationships with rain amount and intensity, yet weaken the et al. 2010; Van Stan and Levia 2010). These long in-canopy inverse relationship with wind (possibly due to greater stiffness retention times permit greater exposure of canopy and trunk with size). Leafed canopy results agreed with the annual mul- drainage water to meteorological conditions, enhancing their tiple regression, producing an inverse association between effect on, and correlation with, stemflow variability (Table 3; stemflow and VPD, and a direct association with mean wind Figs. 2 and 3). For example, stemflow from L. tulipifera speed. Alternatively, multiple regressions performed on leafless responded more strongly to VPD than F. grandifolia (Table 3; canopy stemflow resulted in an inverse relationship with wind Figs. 2d, 3d) due to this longer opportunity for evaporation to speeds, likely decoupling stemflow sheltered within bark fur- claim the water between the canopy storage-drainage pathway rows from VPD. Leaf presence generally increased direct or, even to some extent, the trunk storage-drainage pathway stemflow relationships with rain intensity while diminishing (Fig. 4a). F. grandifolia canopies are oppositely structured— its association with rain amount. The F. grandifolia canopy steep, highly ramified branches covered in smooth bark— structure showed stronger differences in the stemflow-rainfall allowing little meteorological influence apart from rainfall and stemflow-mean wind speed interactions between leaf amount and intensity (Table 3). As F. grandifolia canopy states. These findings, couched within a conceptual path anal- structure encourages canopy and trunk drainage towards ysis based on the reformulated Rutter interception model frame- stemflow (Levia et al. 2010; Van Stan and Levia 2010), it work, show clear potential to alter interception loss modeling in may be unable to resist intensity changes, explaining the stands of high stemflow generation. Thus, we suggest future strong positive correlations with rainfall intensity (Table 3). work seek to incorporate these relationships into the trunk In the leafed season analyses, dominant F. grandifolia cano- storage and drainage processes of commonly used models pies correlate better to intensity (higher β and r)thansmall (e.g., Rutter- and Gash-type models) to improve our under- trees (Fig. 2b). Again, this diminished relationship between standing and prediction of stemflow’s myriad interactions with stemflow and rainfall intensity with F. grandifolia tree size forest ecohydrological processes in high stemflow stands. An- could be a result of intensity being altered by canopy biomass other avenue for future work should also include evaluating the above the smaller trees’ less dominant canopies. Stemflow impact of stand structure (e.g., spatial distribution of trees) on from L. tulipifera trees (particularly during the leafless season) the influence of meteorological conditions over stemflow. did not respond significantly to intensity because the drainage- impeding canopy structure may resist rainfall intensity chang- Acknowledgments The authors gratefully acknowledge the staff and es (Table 3; Fig. 3b). The leafed season L. tulipifera stemflow- administration of the Fair Hill Natural Resource Management Area for wind relationships were stronger than F. grandifolia,probably having granted site access, providing surveillance, and incorporating our since L. tulipifera structure has a higher trunk storage capacity research results and infrastructure in their educational outreach activities. We also thank those who helped with site maintenance: Patrick Kelley- where enhanced wetting from wind more substantially aids Hauske, Charles Apple, and John Fry. stemflow drainage (Fig. 3c). Alternatively, F. grandifolia can- opy architecture has low water storage, and wind-induced wetting may only help in reaching stemflow production for small events (Table 3). References

Aboal JR, Morales D, Hernández M, Jiménez MS (1999) The measure- Conclusions ment and modelling of the variation of stemflow in a laurel forest in Tenerife, Canary Islands. J Hydrol 221:161–175 André F, Jonard M, Ponette Q (2008) Influence of species and rain event Multiple regression analyses have shown a significant meteo- characteristics on stemflow volume in a temperate mixed -beech rological influence over stemflow production from three size stand. Hydrol Process 22:4455–4466 classes of two morphologically contrasting tree species com- Brown JH, Barker AC (1970) An analysis of throughfall and stemflow in – mon to the northeastern USA (L. tulipifera and F. grandifolia) mixed oak stands. Water Resour Res 6:316 323 Carlyle-Moses DE, Gash JHC (2011) Rainfall interception loss by forest across leaf states. Regardless of differing canopy structure, tree canopies. In: Levia DF, Carlyle-Moses DE, Tanaka T (eds) Forest size, or leaf state, positive correlations existed between event and biogeochemistry: synthesis of past research and rainfall amount, rainfall intensity, and mean wind speed while a future directions (Ecological Studies Series 216). Springer, – negative correlation was observed for VPD. However, the Heidelberg, pp 407 423 Carlyle-Moses DE, Price AG (2006) Growing-season stemflow produc- L. tulipifera canopy structure generally strengthened correla- tion within a deciduous forest of southern Ontario. Hydrol Process tions and effect weights for meteorological conditions by en- 20:3651–3663 hancing stemflow residence time and, therefore, exposure to Chang SC, Matzner E (2000) The effect of beech stemflow on spatial meteorological conditions. Tree size also altered stemflow- patterns of soil solution chemistry and seepage fluxes in a mixed beech/oak stand. Hydrol Process 14:135–144 meteorological associations, where taller trees tended to Clements JR (1972) Stemflow in a multi-storied aspen community. Can J strengthen indirect relationships with VPD and direct For Res 2:160–165 Int J Biometeorol (2014) 58:2059–2069 2069

Crockford RH, Richardson DP (2000) Partitioning of rainfall into Li X-Y, Yang Z-P, Li Y-T, Lin H (2009) Connecting ecohydrology and throughfall, stemflow and interception: effect of forest type, ground hydropedology in desert shrubs: stemflow as a source of preferential cover and climate. Hydrol Process 14:2903–2920 flow in . Hydrol Earth Syst Sci 13:1133–1144 Crozier CR, Boerner REJ (1984) Correlations of understory herb distri- Liang W, Kosugi K, Mizuyama T (2011) Soil water dynamics around a bution patterns with microhabitats under different tree species in a tree on a hillslope with or without rainwater supplied by stemflow. mixed mesophytic forest. Oecologia 62:337–343 Water Resour Res 47, W02541 Devore J, Peck R (1993) Statistics, the exploration and analysis of data. Liu S (1998) Estimation of rainfall storage capacity in the canopies of Duxbury Press, Belmont, 881 p cypress wetlands and slash pine uplands in north-central Florida. J Durocher MG (1990) Monitoring spatial variability of forest interception. Hydrol 207:32–41 Hydrol Process 4:215–229 Llorens P, Domingo F (2007) Rainfall partitioning by vegetation under Ford ED, Deans JD (1978) The effects of canopy structure on stemflow, Mediterranean conditions. A review of studies in Europe. J Hydrol throughfall and interception loss in a young Sitka spruce plantation. 335:37–54 J Appl Ecol 15:905–917 Llorens P, Gallart F (2000) A simplified method for forest water storage Frost EE, Levia DF (2014) Hydrologic variation of stemflow yield across capacity measurement. J Hydrol 240:131–144 co-occurring dominant canopy trees of varying mortality. Ecohydrol Llorens P, Poch R, Latron J, Gallart F (1997) Rainfall interception by a In press Pinus sylvestris forest patch overgrown in a Mediterranean moun- Gash JHC, Lloyd CR, Lachaud G (1995) Estimating sparse forest rainfall tainous abandoned area I. monitoring design and results down to the interception with an analytical model. J Hydrol 170:79–86 event scale. J Hydrol 199:331–345 Germer S, Werther L, Elsenbeer H (2010) Have we underestimated Maryland State Climatologist Office (2012) www.atmos.umd.edu/ stemflow? Lessons from an open . J Hydrol 395: ~climate [accessed June 2012] 169–179 Muzylo A, Llorens O, Domingo F (2012) Rainfall partitioning in a Guswa AJ (2012) Canopy vs. roots: production and destruction of vari- deciduous forest plot in leafed and leafless periods. Ecohydrology ability in soil moisture and hydrologic fluxes. Vadose Zone J. doi: 5:759–767 10.2136/vzj2011.0159 Neave M, Abrahams AD (2002) Vegetation influences on water yields Guswa AJ, Spence CM (2012) Effect of throughfall variability on re- from grassland and shrubland ecosystems in the Chihuahuan Desert. charge: application to hemlock and deciduous in western Earth Surf Process Landf 27:1011–1020 Massachusetts. Ecohydrology 5:563–574 Park H-T, Hattori S (2002) Applicability of stand structural characteristics Hauck M, Hesse V, Runge M (2002) The significance of stemflow chem- on stemflow modelling. J For Res 7:91–98 istry for epiphytic diversity in a dieback-affected spruce forest Pypker TG, Levia DF, Staelens J, Van Stan JT (2011) Canopy structure in on Mt Brocken, northern Germany. Lichenologist 34:415–427 relation to hydrological and biogeochemical fluxes. In: Levia DF, Herwitz SR (1986) Infiltration-excess caused by stemflow in a cyclone- Carlyle-Moses DE, Tanaka T (eds) Forest hydrology and biogeo- probe tropical rainforest. Earth Surf Process Landf 11:401–412 chemistry: synthesis of past research and future directions. Springer, Herwitz SR, Slye RE (1995) Three-dimensional modeling of canopy tree Heidelberg, pp 371–388 interception of wind-driven rainfall. J Hydrol 168:205–226 Raat KJ, Draaijers GPJ, Schaap MG, Tietema A, Verstraten JM (2002) Hutchinson I, Roberts MC (1981) Vertical variation in stemflow genera- Spatial variability of throughfall water and chemistry and forest tion. J Appl Ecol 18:521–527 floor water content in a Douglas fir forest stand. Hydrol Earth Syst Inamdar SP, Mitchell MJ (2007) Storm event exports of dissolved organic Sci 6:363–374 nitrogen (DON) across multiple catchments in a glaciated forest Rutter AJ, Kershaw KA, Robins PC, Morton AJ (1971) A predictive watershed. J Geophys Res 112, G02014 model of rainfall interception in forests, 1. Derivation of the model Kuraji K, Yuri T, Nobuaki T, Isamu K (2001) Generation of stemflow from observations in a plantation of Corsican pine. Agric Meteorol volume and chemistry in a mature Japanese cypress forest. Hydrol 9:367–384 Process 15:1967–1978 Staelens J, de Schrijver A, Verheyen K, Verhoest NEC (2008) Rainfall Levia DF, Frost EE (2003) A review and evaluation of stemflow literature partitioning into throughfall, stemflow, and interception within a in the hydrologic and biogeochemical cycles of forested and agri- single beech (Fagus sylvatica L.) canopy: influence of foliation, rain cultural ecosystems. J Hydrol 274:1–29 event characteristics, and meteorology. Hydrol Process 22:33–45 Levia DF, Herwitz SR (2000) Physical properties of stemflow water in Valente F, David JS, Gash JHC (1997) Modelling interception loss for two relation to leachate dynamics: implications for cycling. Can sparse eucalypt and pine forests in central Portugal using reformulated JForRes30:662–666 Rutter and Gash analytical models. J Hydrol 190:141–162 Levia DF, Herwitz SR (2005) Interspecific variation of bark water storage Van Stan JT, Levia DF (2010) Inter- and intraspecific variation of capacity of three deciduous tree species in relation to stemflow yield stemflow production from Fagus grandifolia Ehrh. (American and solute flux to forest soils. Catena 64:117–137 beech) and Liriodendron tulipifera L. (yellow poplar) in relation to Levia DF, Van Stan JT, Mage SM, Kelley-Hauske PW (2010) Temporal bark microrelief in the eastern United States. Ecohydrology 3:11–19 variability of stemflow volume in a beech-yellow poplar forest in Van Stan JT, Siegert CM, Levia DF, Scheick CE (2011) Effects of wind- relation to tree species and size. J Hydrol 380:112–120 driven rainfall on stemflow generation between codominant tree Levia DF, Keim RF, Carlyle-Moses DE, Frost EE (2011) Throughfall and species with differing crown characteristics. Agric For Meteorol stemflow in wooded ecosystems. In: Levia DF, Carlyle-Moses DE, 151:1277–1286 Tanaka T (eds) Forest hydrology and biogeochemistry: synthesis of Xiao Q, McPherson EG, Ustin SL, Grismer ME, Simpson JR (2000) past research and future directions (Ecological Studies Series 216). Winter rainfall interception by two mature open grown trees in Springer, Heidelberg, pp 425–443 Davis, California. Hydrol Process 14:763–784

View publication stats