Small-scale topographic variability influences tree species distribution and canopy throughfall partitioning in a temperate deciduous forest

Forest Ecology and Management 359 (2016) 109–117

Contents lists available at ScienceDirect

Forest Ecology and Management journal homepage: www.elsevier.com/locate/foreco

Small-scale topographic variability influences tree species distribution and canopy throughfall partitioning in a temperate deciduous forest C.M. Siegert a,⇑, D.F. Levia b,c, S.A. Hudson b, A.L. Dowtin b, F. Zhang d, M.J. Mitchell e a

Department of Forestry, Mississippi State University, Mississippi State, MS 39762, USA Department of Geography, University of Delaware, Newark, DE 19716, USA Department of Plant and Soil Sciences, University of Delaware, Newark, DE 19716, USA d Key Laboratory of Western China’s Environmental Systems (MOE), Lanzhou University, Lanzhou, China e College of Environmental Science and Forestry, State University of New York, Syracuse, NY 13210, USA b c

a r t i c l e

i n f o

Article history: Received 15 June 2015 Received in revised form 15 September 2015 Accepted 19 September 2015

Keywords: Throughfall Catchment topography Forest hydrology Deciduous forest

a b s t r a c t At very large spatial scales, the impacts of topography, elevation, and aspect on throughfall variability are apparent. However, within relatively small catchments (<50 ha), differences in species composition induced by slight changes in elevation coupled with slope orientation, could result in sufficient canopy variability whereby throughfall hydrology would be definitively different across small gradients. This study aims to (1) quantify the differences in throughfall hydrology across small topographic gradients, (2) determine the variability of throughfall across such gradients, and (3) determine the seasonal effects on throughfall hydrology resulting from differences in species composition and growing niches. Throughfall partitioning was measured during 15 sampling periods at 4 landscape positions including 3 hillslopes with aspects facing north (NF), west (WF), and south (SF) in addition to a flat area (F) situated in the center of a 12 ha deciduous catchment. Throughfall partitioning was significantly lower on the steepest SF plot (TF = 75.0%) than on the moderately sloping NF (TF = 83.9%, p = 0.001) and F (TF = 81.7%, p = 0.037) plots. SF also had the largest degree of throughfall variability (CV = 20.1), resulting from overlapping canopies, which led to higher rates of canopy interception. NF and WF plots exhibited the largest inter-seasonal differences with decreases in throughfall partitioning of 13.2% (p = 0.013) and 12.1% (p = 0.052), respectively, and corresponded to the largest differences in plant canopy indices (PAI) between seasons. Although slope and aspect were found to be distinguishing variables in our study, it was the influence of these variables on species composition that led to differences in throughfall quantity. Our study illustrates the systematic distribution of water resources across topographic positions within a relatively small forested catchment and highlights the need for additional consideration of topographyinduced controls on microclimate and growing space, which ultimately influence water quality and quantity for effective management strategies. Ó 2015 Elsevier B.V. All rights reserved.

1. Introduction Forested ecosystems impact terrestrial water and nutrient budgets through the partitioning of precipitation into interception, throughfall, and stemflow. The partitioning of water into these three pathways is largely controlled by physiological and morphological traits related to forest composition; seasonality and the presence/absence of foliage; precipitation characteristics; and meteorological conditions. Subsequently, these pathways may become enriched in nutrients and other solutes via washoff of ⇑ Corresponding author. E-mail address: [email protected] (C.M. Siegert). http://dx.doi.org/10.1016/j.foreco.2015.09.028 0378-1127/Ó 2015 Elsevier B.V. All rights reserved.

dry deposition that accumulated during antecedent dry periods (Kazda, 1990) and/or canopy leaching (Lovett and Lindberg, 1984). The spatial variability of throughfall hydrology is the result of heterogeneous canopy cover. Variability in canopy density may arise from several inherent physiological traits such as stand density, crown cover percentage, and leaf area index (LAI), which contribute to throughfall variability. In managed pine plantations with evenly spaced canopies, throughfall increased systematically with decreasing stand density (Stogsdill et al., 1989). In tropical plantations, crown traits (e.g., LAI, crown openness, crown depth) and rainfall characteristics interacted to produce interspecific differences in throughfall partitioning among four commercially grown trees species (Park and Cameron, 2008). In a natural

110

C.M. Siegert et al. / Forest Ecology and Management 359 (2016) 109–117

mixed-deciduous stand, intra- and interspecific differences in canopy density such as the contribution of branch drainage to throughfall generation at higher stand densities were observed to influence variability (André et al., 2011). At even smaller scales within a single canopy, spatial variability is induced by physical characteristics such as distance from stem, although a consensus does not exist whether this variability is systematic (Ford and Deans, 1978; Nanko et al., 2011; Staelens et al., 2006). However, throughfall variability has been shown to be stable across storm events at specific points beneath a leafed canopy (Keim et al., 2005). Consequently, throughfall monitoring requires careful attention to stand characteristics and tree dynamics to accurately quantify the hydrologic flux to the forest floor (Durocher, 1990). Topography is also a significant factor determining the distribution of water resources across the landscape. Stands located at higher elevations, especially those above the average cloud base, are capable of generating greater total volumetric throughfall via cloud water condensation (Holder, 2004), which may also increase total solute flux (Köhler et al., 2015). The presence of tree canopies creates an abrupt increase in surface roughness, which subsequently introduces frictional drag to the horizontal movement of air masses (Weathers et al., 2001), resulting in increased deposition of airborne gases and sediments on foliar and woody surfaces in forested ecosystems. This property, in addition to the increased potential for atmospheric deposition resulting from the complex geometries of leaf surfaces, provides forest canopies with enhanced scavenging abilities and preferential access to passing air masses (Griffith et al., 2015; Hofhansl et al., 2010; Lovett et al., 1996). At lower elevations, differences in throughfall biogeochemistry between uplands and floodplains are less apparent, especially when studies are designed to monitor forest communities of comparable species composition between both landscape positions (Peterson and Rolfe, 1982). However, catchment aspect has been shown to affect throughfall partitioning, whereby slopes oriented towards oncoming storms receive more precipitation, resulting in overall greater throughfall (McJannet et al., 2007). At very large spatial scales, the impacts of topography, elevation, and aspect on throughfall variability are obvious. However, within relatively small catchments, differences in these factors can create microclimate conditions that provide competitive advantages for specific species. Differences in species composition across small spatial gradients can be the result of slight changes in elevation and slope orientation that produce understory light and moisture conditions more favorable for regeneration and establishment of certain species (Siegert and Levia, 2011). As a result, these microclimate conditions result in forest canopies composed of predictable species assemblages, which influence throughfall hydrology and nutrient flux in unique ways, even across small topographic gradients. There are always instances of individual outliers resulting from random seed dispersal mechanisms (Nathan and Muller-Landau, 2000), but on the whole, microclimate conditions most greatly influence competition dynamics. Therefore, it is the objective of our study to (1) quantify the differences in throughfall hydrology across small topographic gradients, (2) determine the variability of throughfall across such gradients, and (3) determine the seasonal effects on throughfall hydrology resulting from differences in species composition and growing niches.

2. Site description Data were collected at Fair Hill Natural Resources Management Area (FH-NRMA) in Fair Hill, Maryland in a 12 ha forested catchment (Fig. 1). Fair Hill NRMA is located in northeastern Maryland

(39°420 N, 75°500 W) at an average elevation of 70 m above sea level within the Chesapeake Bay watershed. FH-NRMA is situated in the Piedmont physiographic region and is characterized as a humid subtropical climate with well-defined seasons. Mean 30-year (1981–2010) summertime (JJA) maximum temperature in northeastern Maryland is 30.9 °C (87.6°F) and average summertime minimum temperature is 18.3 °C. Mean 30-year wintertime (DJF) maximum temperature is 0.7 °C and average minimum wintertime temperature is 4.3 °C (MD State Climatologist Office, 2013). Mean 30-year total annual precipitation is approximately 1200 mm, with an average of 523 mm winter seasonal snowfall and the rest falling as rainfall with little annual variation (MD State Climatologist Office, 2013). The wettest season is autumn (320 mm), followed by summer (314 mm), spring (308 mm), and winter (262 mm). Frontal precipitation patterns are typical for fall, winter, and spring; convective precipitation events dominate the summer. Orographic precipitation does not occur at this field site as it is situated relatively close to sea level and changes in elevation throughout the catchment are minor (79.2–97.1 m). Sampling periods that experienced precipitation falling as snow or during periods of below-freezing temperatures were omitted from this study as solid state precipitation requires different collection methods and were not under the purview of this study. Four study plots were selected within the 12-ha catchment based on landscape position in the watershed and include (1) a north-facing slope (NF), (2) a flat central area (F), (3) a westfacing slope (WF), and (4) a south-facing slope (SF) (Table 1). The site has a tree density of 225 trees ha1 and a basal area of 36.8 m2 ha1. Mean diameter at breast height (dbh) is 40.8 cm and mean canopy tree height is 27.8 m. The forest canopy is comprised of Acer rubrum L. (red maple), Betula lenta L. (sweet birch), Fagus grandifolia Ehrh. (American beech), Liriodendron tulipifera L. (yellow poplar), and Quercus species (white and red oak) as dominant canopy species while the midstory is principally F. grandifolia saplings (Fig. 2). Species distribution across the four landscape positions is variable and represents interspecific preferences for microclimate conditions induced by soil moisture and light availability. The dominant canopy trees are approximately 80–100 years old with a total Plant Area Index (PAI) of 5.3 m2 m2. Leaf emergence begins in May with the growing season lasting through October when senescence begins. The dormant season is categorized as November through April.

3. Materials and methods 3.1. Throughfall collection Throughfall gauges were constructed using 3.78 L high density polyethylene (HDPE) collectors fitted with 20.32 cm diameter funnels, which drained into the collecting apparatus and prohibited evaporation between sampling periods. At each of the four landscape positions, a 50 m by 50 m centrally located plot was designated for throughfall monitoring. Ten collectors were randomly placed on the hillslope of each of the four plots (NF, F, WF, and SF). During each collection period, throughfall volume from each gauge was measured, the gauges emptied, and randomly relocated on the respective hillslopes. Roving gauges, in contrast to stationary gauges, were selected for this study as roving gauges are shown to be more accurate in determining throughfall volume (Durocher, 1990; Lloyd and Marques-Filhode, 1988) and reduce the total number of gauges needed to correctly sample for the throughfall mean (Helvey and Patric, 1965; Rodrigo and Àvila, 2001). Throughfall volume was converted to a depth equivalent by dividing by the orifice area of the funnel (324.3 cm2). Throughfall hydrologic fluxes were measured for one year beginning in September 2011 during

111

C.M. Siegert et al. / Forest Ecology and Management 359 (2016) 109–117

(A) Areal Extent

(B) Aspect

100

SF 95

SF 80

WF

85

F NF

F

100

NF

85

95 90

Aspect

75

E

SW

N

SE

W

NE

S

NW

100 Text 0

0.05

0.1

WF

0

0.05

0.1

0.2

Kilometers

(C) Slope

95

0.2

Kilometers

(D) Orthographic View

SF

F

WF

NF

E. Regional Location

Slope (% Rise)

10-15%

25-30%

0-5%

15-20%

>30%

5-10%

20-25%

Fig. 1. Map of 12-ha research catchment at Fair Hill Natural Resources Management Area (MD) including (A) aerial extent, (B) aspect, (C) slope, (D) orthographic view of the four throughfall plots and (E) the regional location of FH-NRMA in Maryland, USA. Plots are labeled as NF (north-facing), F (flat), WF (west-facing), and SF (south-facing).

Table 1 Study plot site descriptions. Species composition is measured as a percentage of the number of stems (>10 cm DBH) of an individual species relative to the total number of stems per subplot. Plant Area Index (PAI) accounts for both woody and foliar tree components.

Species composition (%) A. rubrum B. lenta F. grandifolia L. tulipifera Quercus spp. Misc. spp. Stems (ha1) Basal area (m2 ha1) PAI (m2 m2, leafless) PAI (m2 m2, leafed) DPAI Slope (°) Aspect (°) Mean elevation (m) Elevation range (m) Gauges (n)

NF

F

WF

SF

4.8 9.7 38.7 11.3 30.6 4.8 248 32.8 1.09 5.58 4.49 7.6 312.1 83.9 8.7 10

14.0 21.1 28.1 14.0 19.3 3.5 228 26.4 1.76 5.51 3.75 3.9 NA 84.6 4.0 10

9.7 31.9 20.8 6.9 22.2 8.3 288 42.0 0.99 5.37 4.41 9.0 257.9 90.9 11.5 10

8.8 20.6 22.5 8.8 29.4 9.8 408 51.6 1.21 5.14 3.93 15.0 141.1 88.8 14.2 10

14 sampling periods (Table 2). Throughfall fluxes were measured approximately bi-weekly, but were variable depending on rainfall conditions and sampling logistics (i.e., whether sufficient rainfall occurred between sampling periods). Samples were not collected

from December 2011 through February 2012 because of sub-zero temperatures and frozen precipitation. 3.2. Meteorological observations Meteorological observations were obtained from a Delaware Environmental Observing System (DEOS) station located 0.5 km south of the forested site in a clearing, a distance which is within the range of those found in similar studies (André et al., 2011; Houbao et al., 1999; Oyarzún et al., 2011; Zimmermann et al., 2007). The DEOS station monitors air temperature, solar radiation, wind speed and direction, soil moisture, and relative humidity. Rainfall was also monitored at the DEOS site with a TE525MM tipping-bucket rain gauge (Texas Electronics, Dallas, Texas, USA) attached to a Campbell Scientific CR1000 datalogger (Logan, Utah, USA). All data were aggregated to coincide with the throughfall sampling periods (Table 2). 3.3. Plant Area Index Plant Area Index (PAI-m2 m2) is a measure of the horizontal aboveground vegetation (e.g., leaves, stems, twigs, and branches) per unit of land area. For each hillslope, PAIs were measured during the dormant season when the canopy was in the leafless phase (March 2012) and during the growing season when the canopies were leafless (June 2012). PAI measurements were obtained using a Li-Cor LAI-2000 Plant Canopy Analyzer (Lincoln, Nebraska, USA).

112

C.M. Siegert et al. / Forest Ecology and Management 359 (2016) 109–117

Flat

100

North

75

DBH Distribution (%)

50

25 DBH Class 10−20 cm

0 South

100

West

20−30 cm 30 cm <

75

50

25

0

Ar

Bl

Fg

Lt

Misc

Q

Ar

Bl

Fg

Lt

Misc

Q

Species Fig. 2. Distribution of species as a function of percent of stems relative to the total number of stems for each species across each of the study plots. Species abbreviations are as follows Ar: A. rubrum; Bl: B. lenta; Fg: F. grandifolia; Lt: L. tulipifera; Misc: miscellaneous species; Q: Q. rubra, Q. alba, and Q. velutina. Table 2 Summary of throughfall collection periods from September 2011 to October 2012. Throughfall values from discrete sampling periods are given in mm with partitioning percentage in parentheses. Sample period

1 2 3 4 5a 6a 7a 8a 9 10 11 12b 13 14 Mean throughfall (%) CV MAD a b

Deployment date

30-Sep-2011 10-Oct-2011 18-Oct-2011 20-Oct-2011 31-Oct-2011 18-Nov-2011 15-Mar-2012 29-Mar-2012 04-May-2012 10-Jul-2012 02-Aug-2012 16-Aug-2012 28-Aug-2012 10-Sep-2012

Collection date

10-Oct-2011 18-Oct-2011 20-Oct-2011 31-Oct-2011 18-Nov-2011 30-Nov-2011 29-Mar-2012 04-May-2012 11-May-2012 30-July-2012 16-Aug-2012 28-Aug-2012 10-Sep-2012 24-Sep-2012

Precipitation

19.9 18.6 13.7 29.3 20.2 89.4 6.4 84.9 19.7 63.5 41.6 37.6 33.6 53.7

Throughfall NF

F

WF

SF

14.8 (74.2) 13.4 (71.9) 12.0 (87.6) 23.7 (80.7) 17.8 (89.0) 90.0 (100.7) 6.1 (95.0) 75.4 (88.8) 19.1 (96.7) 44.8 (70.5) 31.4 (75.5) 40.5 (107.7) 22.8 (67.8) 51.4 (95.7) 83.9 17.0 1.85

14.7 (74.0) 13.6 (73.0) 10.5 (76.8) 23.6 (80.5) 17.5 (87.4) 88.6 (99.1) 4.4 (68.3) 71.5 (84.3) 18.3 (93.1) 45.9 (72.3) 36.3 (87.3) 41.6 (110.7) 26.5 (78.9) 42.0 (78.3) 81.9 18.2 2.72

14.2 (71.3) 12.3 (65.9) 11.0 (80.2) 23.3 (79.4) 16.0 (80.1) 86.7 (97.0) 5.9 (91.8) 70.0 (82.4) 17.2 (87.5) 44.7 (70.4) 28.0 (67.3) 35.4 (94.2) 22.4 (66.6) 45.5 (84.8) 79.5 15.9 2.42

11.5 (57.6) 11.8 (63.5) 12.1 (88.0) 24.9 (85.5) 17.5 (87.7) 79.4 (88.9) 4.9 (76.6) 60.5 (71.3) 15.3 (77.8) 36.0 (56.7) 26.5 (63.7) 39.4 (104.7) 20.4 (60.8) 45.0 (83.8) 75.0 20.1 2.47

Storm events observed during complete defoliated canopy conditions. Scattered thunderstorms during this sampling period were too isolated to be included in statistical analysis but are still included in graphical figures.

For each landscape position, PAI measurements were obtained along two transects, one running north–south and one running east–west. An initial PAI reading was taken in a nearby clearing and then a series of seven measurements were taken along the respective transects at five meter intervals. PAI was averaged over each transect and then over each landscape position to determine a representative estimate of PAI over the entire landscape position of interest. PAI measurements on both dates were collected during overcast conditions to reduce instrument error and were collected in approximately 1 h during stable overcast conditions. 3.4. Statistics Coefficient of variation (CV) is a common metric used to measure throughfall spatial variability (Keim et al., 2005; Staelens et al., 2006; Zimmermann et al., 2007) and is calculated as:

CV ¼ r=l

ð1Þ

where r is the standard deviation of throughfall gauge volume, and l is the mean throughfall gauge volume. CVs were calculated for each of the four landscape positions. Median absolute deviation (MAD) is another common metric and is calculated as:

MAD ¼ Mi ðjxi  Mj ðxj ÞjÞ

ð2Þ

where xi is the observed throughfall volume for a given gauge i and Mj is the median throughfall volume for all gauges on the landscape position j. MAD is less susceptible to influence by outliers and is a more robust statistical measure than standard deviation. Kruskal–Wallis tests for non-parametric population distributions were performed to determine whether throughfall was statistically different across landscape positions for each collection period. Precipitation, and subsequently throughfall, follows a non-Gaussian distribution, most frequently classified using a

C.M. Siegert et al. / Forest Ecology and Management 359 (2016) 109–117

gamma distribution, although skewed normal distributions have also been proposed (Dunn, 2004). Wilcoxon Rank Sums pairwise tests were then used to determine which of the specific landscape positions had statistically different median throughfall partitioning rates. To test for seasonal differences, Welch’s t-tests or Wilcoxon Rank Sums were employed, depending on normality assumptions, to determine whether throughfall partitioning and variability was statistically different between the leafed and leafless canopy phases for a given landscape position. For purposes of determination of water resources, seasonal to annual hydrologic fluxes are most important to forest managers, therefore these tests were performed for the sum total of all storms in a given canopy leaf phase. 3.5. GIS analysis Transects across each of the four plots were established with a handheld GPS unit (Garmin GPSmap 62st, Olathe, Kansas, USA) with a location accuracy of ±5 m. Plots were laid out during leafless canopy conditions to reduce the error associated with the GPS instrument. The endpoints of the survey transects were then connected into polygons in ArcGIS 10.2.2 (ESRI 2013). The centroid of these polygons were used to create four 50  50 m plots, defining the four study plots and the boundaries for the basal area survey. Slope and aspect were derived from 1/9 arc second DEM from the National Elevation Dataset (USGS 2013), with slope calculated as percent rise. Minimum, maximum, mean, and standard deviation values for slope values were calculated using the Zonal Statistics toolset in Spatial Analyst (ArcGIS). Descriptive statistics for aspect values were calculated using the Raster Calculator tool, where the sine and cosine of each aspect cell were combined using the ATAN2 function resulting in the real mean aspect. The 1 m contour lines were generated by a built in function, and six inch areal imagery was incorporated (Maryland 2015). 4. Results 4.1. Landscape topography and biophysical properties Throughfall partitioning was measured during 14 sampling periods at four landscape positions including three hillslopes with aspects facing north (NF), west (WF), and south (SF) in addition to a flat area (F) situated in the center of the catchment. Sample period 12 was omitted from statistical analysis because the storm delivered significantly more precipitation to the throughfall collection site than was observed at the meteorological station just 0.5 km to the south. This sampling period occurred during the summer, when isolated convective thunderstorms are most common. Of the remaining 13 sampling periods, 4 occurred from November through April when the canopy was in the leafless phase and 10 occurred from May through October when the canopy was in the leafed phase (Table 2). The steepest hillslope was located on SF with an average slope of 15.0% followed in steepness by WF with a 9.0% slope, NF landscape with a 7.6% slope, and F with a 3.9% slope (Table 1). During the leafed canopy phase, PAIs ranged from a maximum 5.58 m2 m2 on NF to a minimum of 5.14 m2 m2 on SF, with intermediate values of 5.37 m2 m2 and 5.51 m2 m2 on WF and F, respectively (Table 1). During the leafless canopy phase, PAI ranged from 0.99 m2 m2 on WF to 1.76 m2 m2 on F with intermediate values of 1.09 m2 m2 and 1.21 m2 m2 on NF and SF, respectively (Table 1). The largest change in PAI between seasons was observed on NF, followed by WF and SF with the smallest PAI change observed on F (Table 1). However, PAI was greatest on F during the leafless season (Table 1).

113

4.2. Spatial variability of throughfall across landscape positions The spatial distribution of throughfall exhibited a clear relationship to landscape position and hillslope orientation. Nonparametric statistical tests indicate that throughfall partitioning was statistically different underneath tree canopies on different hillslope positions (Kruskal–Wallis chi-squared = 15.4249, df = 3, p = 0.001487) (Figs. 3 and 4). Overall, throughfall gauges on NF recorded the largest partitioning of incident precipitation into throughfall with an annual average of 83.9% (SD = 18.7%), followed by F (81.9%, SD = 17.0%), WF (79.5%, SD = 15.5%), and SF (75.0%, SD = 18.1%) (Table 2). Throughfall partitioning by tree canopies on SF was statistically different from NF (p = 0.0012) and F (p = 0.0372) was not significantly different than WF (Table 3). SF also had significantly different rates of throughfall partitioning than F at the 90% confidence interval (Table 3). Average throughfall partitioning values ranged from 67.8% to 100.7% on NF, from 68.3% to 99.1% on F, from 65.9% to 97.0% on WF, and from 56.7% to 88.9% on SF (Table 2 – excluding Sampling Period 12). Throughfall partitioning values greater than 100% observed on NF during sampling period 6 indicate the presence of drip points from the canopy whereby throughfall is preferentially deposited to the forest floor in quantities greater than those of observed precipitation. Throughfall partitioning values were greatest on NF during 7 monitoring periods, greatest on F during 4, and greatest on SF during 2 (Table 2). Throughfall partitioning values were least on SF during 8 monitoring periods, least on WF during 2, least on F during 3 (Table 2). Throughfall partitioning values were intermediate during all observed sampling periods (Fig. 3). At the annual scale, SF had the largest coefficient of variation (CV) at 20.1 followed by F (18.2), NF (17.0) and WF (15.9) (Table 2). Throughfall CV was most variable between plots during rainfall amounts of intermediate magnitude. Smaller rainfall amounts (<25 mm) and the largest rainfall amounts (>75 mm) resulted in throughfall CVs that were most similar while intermediate rainfall amounts resulted in the most variable throughfall CV (Fig. 3). Spatial heterogeneity of throughfall distribution was also assessed using the median absolute deviation (MAD). At the annual scale, F had the largest MAD (2.72) followed by similar values on SF (2.47) and WF (2.42), and NF (1.85) (Table 2). Similar to CV, MAD values indicate greater spatial variability on SF and F, with less spatial variability on NF and WF. Throughfall MAD also exhibited a larger degree of variability across plots as rainfall amounts increased (between 25 and 50 mm) and then became more similar across plots at even higher rainfall amounts (>50 mm) (Fig. 3). However, MAD increased across all plots as rainfall increased, while no clear trend was identifiable with CV across the range of observed rainfall. 4.3. Temporal variability of throughfall across landscape positions The presence of foliage in the forest canopy during the leafed canopy phase reduced the amount of incident precipitation partitioned into throughfall on all four plots in comparison to the leafless canopy phase (Table 4). When foliage was present, throughfall partitioning was reduced by 13.2% on NF (p = 0.013), by 12.1% on WF (p = 0.052), by 9.4% on SF (p = 0.142), and by 4.3% on F (p = 0.556). NF and WF exhibited the largest inter-seasonal differences in throughfall partitioning (Fig. 4) and also exhibited the largest change in LAI between phenoseasons (Table 1), which may explain the decrease in throughfall partitioning observed on these two plots (Table 4). There were no significant differences in seasonal throughfall partitioning on F or SF plots (Table 4). On all four landscape positions, the presence of foliage in the leafed canopy phase resulted in an increase in CV (Table 4). The lar-

114

C.M. Siegert et al. / Forest Ecology and Management 359 (2016) 109–117

Plot

120 NF

Mean TF Partitioning (%)

Mean Throughfall (mm)

75 F WF SF 50

25

80

40

0 25

50

25

75

Precipitation (mm)

75

15

Median Absolute Deviation

50

40

Coefficient of Variation

50

Precipitation (mm)

30

20

10

5

10

0

0 25

50

75

Precipitation (mm)

25

50

75

Precipitation (mm)

Fig. 3. Observed values in each of the four study plots for mean throughfall partitioning in mm, mean throughfall partitioning as a percentage of incident precipitation, mean throughfall coefficient of variation (CV), and throughfall median absolute deviation (MAD). Legend symbols denote plots as NF (north-facing), F (flat), WF (west-facing), and SF (south-facing).

gest inter-seasonal differences in CV were observed on F with an increase of 10.0 when foliage was present (p = 0.030) and on WF with an increase of 7.4 (p = 0.034), both of which were significant seasonal differences. CV also increased on NF by 5.7 (p = 0.203) and by 7.7 (p = 0.604) on SF when foliage was present, although these trends were not statistically significant (Table 4). MAD increased on NF, F, and SF and was slightly reduced on WF when foliage was present in the forest canopy however these interseasonal differences in MAD were not significant (Table 4). 5. Discussion The controls on the spatial distribution of throughfall are numerous and complex, stemming from individual canopy characteristics (Nanko et al., 2011; Staelens et al., 2006), ecoregion (Bryant et al., 2005; Zimmermann et al., 2007), storm characteristics (Van Stan et al., 2012), and seasonality (Deguchi et al., 2006). Such influencing factors were observed in this study (Tables 1 and 4, Fig. 3) in addition to identifying a new factor— landscape position. However, landscape position itself is merely a proxy variable representing species composition as it is determined by microclimate growing conditions as a result of topographic characteristics. In our study, SF had consistently lower volumetric throughfall fluxes while NF and F had the highest fluxes, suggesting that the latter two plots behaved the most similarly in terms of canopy

partitioning. In flat landscapes, tree canopies compete for growing space and in the process develop gaps around individual crowns through the process of crown shyness (Putz et al., 1984), which allows for greater amounts of throughfall partitioning. We therefore hypothesize that throughfall was most reduced on SF because it was the steepest, leading to overlapping canopies and additional opportunities for canopy interception. In fact, SF had many more stems and a much higher basal area per plot area than the other plots (Table 1). However, SF had the smallest PAI (5.14 m2 m2) during the leafed canopy phase compared to the largest PAI on NF (5.58 m2 m2) (Table 1), which traditionally is an important factor in limiting canopy throughfall (Barbier et al., 2009; Morris et al., 2003). The incongruity between plots with low PAI and high canopy interception suggests that individual species characteristics such as leaf and branch orientation (Park and Cameron, 2008), leaf texture (Holder, 2007), and canopy position (Herwitz and Slye, 1995) are also strong determinants of rainfall partitioning. Therefore, it is likely that the importance of microsite characteristics and preferential growth niches for tree species enhances the impacts of these characteristics on throughfall. In terms of abiotic controls, volumetric throughfall flux was greatest under intermediate rainfall amounts (Fig. 3). At low rainfall amounts the forest canopy intercepts the majority of incident rainfall when rainfall amount is equivalent to canopy storage capacity, resulting in minimal throughfall across plots. At large rainfall amounts, the forest canopy storage capacity is saturated

115

C.M. Siegert et al. / Forest Ecology and Management 359 (2016) 109–117

75

dormant

Throughfall (mm)

50

25

0

75

growing

50 Plot NF

25

F WF SF

0 25

50

75

Precipitation (mm) Fig. 4. Throughfall variability measured at each of the study plots over 14 collection periods. Average throughfall for each study plot is plotted as a circle with 95% confidence intervals plotted as vertical lines. Collection periods during the dormant season are located in the top plot and during the growing season in the bottom plot. The diagonal dotted line represents a 1:1 line. Legend symbols denote plots as NF (north-facing), F (flat), WF (west-facing), and SF (south-facing).

Table 3 P-value results from pairwise comparisons of throughfall partitioning using Wilcoxon rank sum test.

F WF SF

NF

F

WF

0.575 0.190 0.001**

– 0.575 0.037*

– – 0.190

Significance codes: ***p < 0.001; **p < 0.010; *p < 0.050

and most of the rainfall occurring after this saturation point is converted into throughfall. We hypothesize that at intermediate rainfall amounts, the interspecific differences between canopy structure, morphology, and canopy storage capacity would be most apparent (e.g., Hörmann et al., 1996 – Table 1). Species with lower canopy storage values would preferentially generate more throughfall while species with higher canopy storage values would generate less at intermediate rainfall values.

In our study, we used non-stationary gauges to capture such variability. Throughfall gauges were centrally located within the study plots under canopy conditions most representative of each landscape position in terms of species composition and canopy structure. Random relocation of the gauges within the small plot areas enabled us to sample the variability induced by speciesspecific structural traits that we hypothesize to occur in a given location due to topographic–microclimate constraints. As such, our sampling scheme was designed to capture seasonal to annual trends in throughfall partitioning—an appropriate temporal resolution for determining water resource availability. Considerable efforts have been made to determine the impact of individual storm characteristics such as duration, intensity, wind speed, wind direction, and intra-storm evaporation on throughfall partitioning (e.g., Crockford and Richardson, 2000; Link et al., 2004; Marin et al., 2000; Nanko et al., 2011; Staelens et al., 2008) but are not in the purview of this paper. Instead, we show that in addition to these factors, the control on the distribution of tree species across small landscapes from microclimate conditions can also impact

Table 4 Plot-scale throughfall descriptive statistics. p-values determined using Welch’s two sample t-test when sample populations were normally distributed and using Wilcoxon’s Rank Sums when sample populations were not normally distributed. Data are derived from throughfall collected during 10 rainfall events in the leafed canopy phase and 4 rainfall events in the leafless canopy phase.

Mean TF (%) CV MAD

North-facing (NF) Test

Leafless

Leafed

p-value

Flat (F) Test

Leafless

Leafed

p-value

Welch’s Welch’s Wilcoxon

93.2 13.0 3.15

80.0 18.7 3.39

0.013* 0.203 0.711

Welch’s Welch’s Welch’s

84.6 11.3 3.55

80.3 21.3 4.20

0.556 0.030* 0.755

West-facing (WF)

Mean TF (%) CV MAD

South-facing (SF)

Test

Leafless

Leafed

p-value

Test

Leafless

Leafed

p-value

Welch’s Wilcoxon Welch’s

87.7 10.7 2.45

75.6 18.1 2.40

0.052° 0.034* 0.968

Welch’s Wilcoxon Wilcoxon

80.9 14.8 3.24

71.5 22.5 3.54

0.142 0.604 0.940

Significance codes: ***p < 0.001; **p < 0.010; *p < 0.050; °p < 0.100

116

C.M. Siegert et al. / Forest Ecology and Management 359 (2016) 109–117

throughfall partitioning. For example, spatial variability of throughfall volumetric flux was highest on SF and F (Table 2). These landscape positions had the steepest and flattest slopes, respectively (Table 1). In this instance, we hypothesize that the effect of canopy spatial heterogeneity was maximized on these two landscape positions in contrast to intermediately steep hillslopes. In very flat landscapes, stand and gap dynamics led to large degrees of canopy heterogeneity, while the overlapping canopies on the steepest landscapes (see previous discussion) were responsible for such heterogeneity. Seasonally, NF and WF exhibited the largest inter-seasonal differences in throughfall partitioning while throughfall partitioning on F was most similar (Table 4). NF and WF also exhibited the largest change in LAI between phenoseasons and may be responsible for the decrease in throughfall partitioning experienced on these plots (Table 1). Furthermore, F and WF displayed significant increase in CV when foliage was present (Table 4), although it is unclear why this would be the case. Further insights into individual landscape controls on species and throughfall partitioning are necessary. At present, the overall size of the study catchment and the viable representative area of canopy cover within each hillslope plot prevents replicate subplots. However, replication of findings from our study in forests across geographically similar regions with similar species distributions is an avenue for future research activities. Additionally, we hypothesize that the controls on throughfall hydrology by species distribution are likewise important on throughfall biogeochemistry. For example, NF has the largest percentage of F. grandifolia stems by landscape position (Table 1). This species has characteristically marcescent leaves that are retained in the forest canopy throughout the dormant season until new buds emerge in the spring. The persistence of canopy leaves in addition to the dominance of F. grandifolia on this slope could increase concentrations of solutes derived from canopy leaching such as DON and DOC (Hofhansl et al., 2012). As this landscape position was also associated with the greatest volumetric throughfall fluxes, it is possible that the largest mass flux of these solutes could also occur on this landscape position.

may not be as susceptible to differences induced by topographicspecies composition, but for studies where throughfall values are aggregated to longer time scales (seasonal, annual) these differences become apparent. For watershed management purposes, it is precisely these longer time scales which are more important for determination of surface water resources. Throughfall spatial variability is controlled at many scales: from the individual tree canopy to the catchment. Our study suggests that hillslope position within a catchment must also be considered for its effects on microclimate growing conditions and species distribution. For relatively flat catchments, such as those analyzed by Keim et al. (2005), the influence of hillslope on species composition may be insignificant, but for studies in more diverse topographic regions, gauge distribution in a small subplot of a catchment may be unrepresentative of the catchment as a whole. The preference of certain tree species to specific landscape positions is well known (e.g., bottomlands versus uplands, north versus south slopes), but the degree to which these preferences influence canopy hydrology, even at small spatial scales, has not received the same attention. We demonstrate that these local features also matter and propose that the dynamics of growth regimes on canopy partitioning and surface water availability are also important for successful outcomes in forest and watershed management. Acknowledgements The authors would like to acknowledge the financial support received from the US National Science Foundation (Ref. Nos. EAR0724971, BCS-1233592) and the University of Delaware Mather Research Award (2012). This material is also based upon work that is supported by the National Institute of Food and Agriculture, U.S. Department of Agriculture, McIntire Stennis project #1002493. Many thanks are given to the Delaware Environmental Observing System (DEOS) for meteorological data; to Ranger Rachel Temby and the Maryland Department of Natural Resources for access to the research site at Fair Hill Natural Resource Management Area. The authors extend sincere thanks to the Associate Editor and anonymous reviewers that provided comments to improve this manuscript.

6. Conclusion References By design, the configuration of a watershed is typically horseshoe shaped with drainage hillslopes facing in three directions that drain into a single outlet point. Slope orientation and aspect are large controls of species composition due to surface energy and soil moisture regimes. The catchment monitored in our study consists of three distinctly oriented hillslopes and a flat central area, which resulted in four different microclimates and growing conditions within the relatively small 12 ha catchment. Steeper slopes, as those located on NF and SF, support species that prefer welldrained soils such as L. tulipifera and F. grandifolia (Beck, 1990; Tubbs and Houston, 1990). Flatter slopes experience poorer drainage, which leads to more mesic species compositions that tolerate soils with poorer drainage conditions such as A. rubrum and B. lenta (Lamson, 1990; Walters and Yawney, 1990). Even within a relatively small area, this study shows that throughfall variability is also influenced by slope and aspect, which directly impact surface microclimate, growing conditions, and competition. As throughfall gauges were relocated between sampling periods, it is likely that the random resampling led to alternating rankings between plots with the highest throughfall fluxes, namely the north-facing and flat plots. However, it is important to note that throughfall partitioning must be considered across longer sampling periods to notice true topographic influences at the spatial scale of this study. Variations during individual collection periods

André, F., Jonard, M., Jonard, F., Ponette, Q., 2011. Spatial and temporal patterns of throughfall volume in a deciduous mixed-species stand. J. Hydrol. 400, 244– 254. Barbier, S., Balandier, P., Gosselin, F., 2009. Influence of several tree traits on rainfall partitioning in temperate and boreal forests: a review. Ann. For. Sci. 66, 602. Beck, D.E., 1990. Yellow poplar. In: Burns, R.S., Honkala, B.H. (Eds.), Silvics of North America, vol. 2. U.S. Department of Agriculture, Agriculture Handbook 654, Hardwoods, pp. 406–416. Bryant, M.L., Bhat, S., Jacobs, J.M., 2005. Measurements and modeling of throughfall variability for five forest communities in the southeastern US. J. Hydrol. 312, 95–108. Crockford, R.H., Richardson, D.P., 2000. Partitioning of rainfall into throughfall, stemflow and interception: effect of forest type, ground cover and climate. Hydrol. Process. 14, 2903–2920. Deguchi, A., Hattori, S., Park, H.T., 2006. The influence of seasonal changes in canopy structure on interception loss: application of the revised Gash model. J. Hydrol. 318, 80–102. Dunn, P.K., 2004. Occurrence and quantity of precipitation can be modelled simultaneously. Int. J. Climatol. 24, 1231–1239. Durocher, M.G., 1990. Monitoring spatial variability of forest interception. Hydrol. Process. 4, 215–229. Ford, E.D., Deans, J.D., 1978. The effects of canopy structure on stemflow, throughfall and interception loss in a young Sitka spruce plantation. J. Appl. Ecol. 15, 905–917. Griffith, K.T., Ponette-González, A.G., Curran, L.M., Weathers, K.C., 2015. Assessing the influence of topography and canopy structure on Douglas fir throughfall with LiDAR and empirical data in the Santa Cruz mountains, USA. Environ. Monit. Assess. 187, 4486. Helvey, J.D., Patric, J.H., 1965. Canopy and litter interception of rainfall by hardwoods of Eastern United States. Water Resour. Res. 1, 193–206.

C.M. Siegert et al. / Forest Ecology and Management 359 (2016) 109–117 Herwitz, S.R., Slye, R.E., 1995. Three-dimensional modeling of canopy tree interception of wind-driven rainfall. J. Hydrol. 168, 205–226. Hofhansl, F., Wanek, W., Drage, S., Huber, W., Weissenhofer, A., Richter, A., 2012. Controls of hydrochemical fluxes via stemflow in tropical lowland rainforests: effects of meteorology and vegetation characteristics. J. Hydrol. 452–453, 247– 258. Hofhansl, F., Wanek, W., Drage, S., Huber, W., Weissenhofer, A., Richter, A., 2010. Topography strongly affects atmospheric deposition and canopy exchange processes in different types of wet lowland rainforest, Southwest Costa Rica. Biogeochemistry 106, 371–396. Holder, C.D., 2004. Rainfall interception and fog precipitation in a tropical montane cloud forest of Guatemala. For. Ecol. Manage. 190, 373–384. Holder, C.D., 2007. Leaf water repellency of species in Guatemala and Colorado (USA) and its significance to forest hydrology studies. J. Hydrol. 336, 147–154. Hörmann, G., Branding, A., Clemen, T., Herbst, M., Hinrichs, A., Thamm, F., 1996. Calculation and simulation of wind controlled canopy interception of a beech forest in Northern Germany. Agric. For. Meteorol. 79, 131–148. Houbao, F., Wei, H., Zhuang, M., Kosuke, W., 1999. Acidity and chemistry of bulk precipitation, throughfall and stemflow in a Chinese fir plantation in Fujian. China. For. Ecol. Manage. 122, 243–248. Kazda, M., 1990. Sequential stemflow sampling for estimation of dry deposition and crown leaching in beech stands. In: Harrison, A.F., Ineson, P., Hehl, O.H. (Eds.), Nutrient Cycling in Terrestrial Ecosystems. Elsevier, London, pp. 46–55. Keim, R.F., Skaugset, A.E., Weiler, M., 2005. Temporal persistence of spatial patterns in throughfall. J. Hydrol. 314, 263–274. Köhler, L., Leuschner, C., Hauck, M., Hertel, D., 2015. Cloud water interception and element deposition differ largely between Norway spruce stands along an elevation transect in Harz Mountains, Germany. Ecohydrol. 8, 1048–1064. http://dx.doi.org/10.1002/eco.1563. Lamson, N.I., 1990. Sweet birch. In: Burns, R.S., Honkala, B.H. (Eds.), Silvics of North America, vol. 2. U.S. Department of Agriculture, Agriculture Handbook 654, Hardwoods, pp. 148–152. Link, T.E., Unsworth, M., Marks, D., 2004. The dynamics of rainfall interception by a seasonal temperate rainforest. Agric. For. Meteorol. 124, 171–191. Lloyd, C.R., Marques-Filhode, A.de O., 1988. Spatial variability of throughfall and stemflow measurements in Amazonia rainforest. Agric. For. Meteorol. 42, 63– 73. Lovett, G.M., Lindberg, S.E., 1984. Dry deposition and canopy exchange in a mixed oak forest as determined by analysis of throughfall. J. Appl. Ecol. 21, 1013–1027. Lovett, G.M., Nolan, S.S., Driscoll, C.T., Fahey, T.J., 1996. Factors regulating throughfall flux in a New Hampshire forested landscape. Can. J. For. Res. 26, 2134–2144. Marin, C.T., Bouten, W., Sevink, J., 2000. Gross rainfall and its partitioning into throughfall, stemflow and evaporation of intercepted water in four forest ecosystems in western Amazonia. J. Hydrol. 237, 40–57. McJannet, D., Wallace, J., Reddell, P., 2007. Precipitation interception in Australian tropical rainforests: I. Measurement of stemflow, throughfall and cloud interception. Hydrol. Process. 21, 1692–1702.

117

Morris, D.M., Gordon, A.G., Gordon, A.M., 2003. Patterns of canopy interception and throughfall along a topographic sequence for black spruce dominated forest ecosystems in northwestern Ontario. Can. J. For. Res. 33, 1046–1060. Nanko, K., Onda, Y., Ito, A., Moriwaki, H., 2011. Spatial variability of throughfall under a single tree: experimental study of rainfall amount, raindrops, and kinetic energy. Agric. For. Meteorol. 151, 1173–1182. Nathan, R., Muller-Landau, H.C., 2000. Spatial patterns of seed dispersal, their determinants and consequences for recruitment. Trends Ecol. Evol. 15, 278– 285. Oyarzún, C.E., Godoy, R., Staelens, J., Donoso, P.J., Verhoest, N.E.C., 2011. Seasonal and annual throughfall and stemflow in Andean temperate rainforests. Hydrol. Process. 25, 623–633. Park, A., Cameron, J.L., 2008. The influence of canopy traits on throughfall and stemflow in five tropical trees growing in a Panamanian plantation. For. Ecol. Manage. 255, 1915–1925. Peterson, D.L., Rolfe, G.L., 1982. Precipitation components as nutrient pathways in floodplain and upland forests of central Illinois. For. Sci. 28, 321–332. Putz, F.E., Parker, G.G., Archibald, R.M., 1984. Mechanical abrasion and intercrown spacing. Am. Midl. Nat. 112, 24–28. Rodrigo, A., Àvila, A., 2001. Influence of sampling size in the estimation of mean throughfall in two Mediterranean holm oak forests. J. Hydrol. 243, 216–227. Siegert, C.M., Levia, D.F., 2011. Stomatal conductance and transpiration of cooccurring seedlings with varying shade tolerance. Trees 25, 1091–1102. Staelens, J., De Schrijver, A., Verheyen, K., Verhoest, N.E.C., 2006. Spatial variability and temporal stability of throughfall water under a dominant beech (Fagus sylvatica L.) tree in relationship to canopy cover. J. Hydrol. 330, 651–662. Staelens, J., Schrijver, A..De., Verheyen, K., Verhoest, N.E.C., 2008. Rainfall partitioning into throughfall, stemflow, and interception within a single beech (Fagus sylvatica L.) canopy: influence of foliation, rain event characteristics, and meteorology. Hydrol. Process. 22, 33–45. Stogsdill, W.R., Wittwer, R.F., Hennessey, T.C., Dougherty, P.M., 1989. Relationship between throughfall and stand density in a Pinus taeda plantation. For. Ecol. Manage. 29, 105–113. Tubbs, C.H., Houston, D.R., 1990. American beech. In: Burns, R.S., Honkala, B.H. (Eds.), Silvics of North America, vol. 2. U.S. Department of Agriculture, Agriculture Handbook 654, Hardwoods, pp. 325–331. Van Stan, J.T., Levia, D.F., Inamdar, S.P., Lepori-Bui, M., Mitchell, M.J., 2012. The effects of phenoseason and storm characteristics on throughfall solute washoff and leaching dynamics from a temperate deciduous forest canopy. Sci. Total Environ. 430, 48–58. Walters, R.S., Yawney, H.W., 1990. Red maple. In: Burns, R.S., Honkala, B.H. (Eds.), Silvics of North America, vol. 2. U.S. Department of Agriculture, Agriculture Handbook 654, Hardwoods, pp. 60–69. Weathers, K.C., Cadenasso, M.L., Pickett, S.T.A., 2001. Forest edges as nutrient and pollutant concentrators: potential synergisms between fragmentation, forest canopies, and the atmosphere. Conserv. Biol. 15, 1506–1514. Zimmermann, A., Wilcke, W., Elsenbeer, H., 2007. Spatial and temporal patterns of throughfall quantity and quality in a tropical montane forest in Ecuador. J. Hydrol. 343, 80–96.