Forest Floor Interception

Hydrological Processes
Spatial and temporal variability of canopy and forest floor

interception
Journal: Hydrological Processes
Manuscript ID: HYP-09-0460

Fo
Wiley - Manuscript type: Research Article
Date Submitted by the

15-Sep-2009
Author:
r
Complete List of Authors: Gerrits, Miriam; Delft University of Technology, Water Management
Pfister, Laurent; CRP-Gabriel Lippmann, EVA
Pe
Savenije, Hubert; Delft University of Technology, civil Eng and

Geosciences
canopy interception, forest floor interception, spatial variability,

Keywords:
er
temporal variability, semi-variograms, time stability plots

Re
vi
ew
http://mc.manuscriptcentral.com/hyp
Page 1 of 30 Hydrological Processes
1
2
3
4
5
6
7
8
9
10
11
12
Spatial and temporal variability of canopy and
13
14
forest floor interception
15
16 A.M.J. Gerrits, L. Pfister, H.H.G. Savenije
17
18 September 15, 2009
19
20
Fo
21
22
Contents
23 1 Introduction 2
24
rP
25 2 Methodology 3
26
27 3 Site description and materials 4
28
29
ee
4 Temporal variation in interception measurements 6

30 4.1 Canopy interception . . . . . . . . . . . . . . . . . . . . . . . . . 6
31 4.2 Forest floor interception . . . . . . . . . . . . . . . . . . . . . . . 8
32
rR
33 5 Temporal variation in storage capacity 9

34 5.1 Canopy interception . . . . . . . . . . . . . . . . . . . . . . . . . 9
35 5.2 Forest floor interception . . . . . . . . . . . . . . . . . . . . . . . 10
36
37 6 Effect of variability in storage capacity on Rutter model pre-
ev
38 dictions 12
39 6.1 Rutter Model description . . . . . . . . . . . . . . . . . . . . . . 12
40 6.2 Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14
41
iew
42 7 Spatial variation in throughfall and infiltration 19

43
44 8 Conclusions 24
45
46 9 Acknowledgment 27
47
48
49
50
51
52
53
54
55
56
57 1
58
59
60
Hydrological Processes Page 2 of 30
1
2
3
4
5
6
7
8 Abstract
9 Depending on season, rainfall characteristics, and forest species, in-
10 terception amounts to 15-50% of total precipitation in a forest under
11 temperate climates. Many studies have investigated the importance of
12 interception of different tree species in all kinds of different climates. Of-
13 ten authors merely determine interception storage capacity of that specific
14 species and the considered event, and only sometimes a distinction is made
15 between foliated and non-foliated trees. However, interception is highly
16 variable in time and space. Firstly, because potential evaporation is higher
17 in summer, but secondly because the storage capacity has a seasonal pat-
18 tern. Besides weather characteristics like wind and rain intensity, snow
causes large variations in the maximum storage capacity. In an experi-
19
mental beech plot in Luxembourg we found storage capacity of canopy
20
interception to show a clear seasonal variability varying from 0.1 mm to
Fo
21 1.2 mm. The capacity of the forest floor appears to be rather constant at
22 1.8 mm. However, both show variations as high as ±100%, which result in
23 only about 11% difference in evaporation estimates when a Rutter model
24 is applied. Hence the number of raindays and the potential evaporation
rP
25 are stronger driving factors on interception. Furthermore, the spatial

26 correlation of the throughfall and infiltration has been investigated with
27 semi-variograms and time stability plots. Within 6-7 m distance through-
28 fall and infiltration are correlated and the general persistence is rather
29 weak.
ee
30
31
32 1 Introduction
rR
33
In a forest, part of the precipitation is intercepted by vegetation. Rain drops
34
35 first hit the leaves and branches before they fall on the forest floor, where again
36 a part is intercepted by litter on the forest floor. Progressively, the leaves and
37 the litter dry-out by evaporation. Both processes are here considered to be part
ev
38 of the same interception process, I, which equals the sum of the change in in-
39 terception storage (Si ) and the evaporation from this stock (Ei ):
40 dSi
41 I = Ei + (1)
iew
dt
42
43 Interception is considered to be about 15 to 50% of the total incoming precip-
44 itation on forests in temperate humid latitudes (e.g., Rutter et al. [1975], Viville
45 et al. [1993], Hörmann et al. [1996], Savenije [1997]). Since the intercepted and
46 evaporated water is not recharging the forest soils, the interception process is
47 important to the soil moisture balance of forests. The intercepted water does
48 not contribute to the soil reservoir and hence is not available to vegetation. In
49 addition, the interception process plays an important role as a re-distributor
50 of rainfall in space. In general, throughfall is less than gross precipitation, but
51 often drip points are observed where the canopy and branches funnel the rain-
52 fall causing locally higher intensity throughfall (e.g., Germer et al. [2006] and
53 Gerrits et al. [2009]). This results in concentrated infiltration, which again may
54 be the trigger for sub-surface flow. These throughfall patterns appear to have
55
56
57 2
58
59
60
1
2
3
4
5
6
7
8 an important influence on the the soil moisture patterns (Bouten et al. [1992]).
9
10 In literature, many studies on interception can be found (see Kittredge
11 [1948], Zinke [1967], and Breuer et al. [2003] and references herein). Most of
12 these studies consider interception of events or short duration periods (in the
13 order of weeks). However, interception is highly seasonal. First of all the poten-
14 tial evaporation changes throughout the year and so does the storage capacity
15 for deciduous trees. At best a distinction is made between leaf-on and leaf-off
16 periods (e.g., Rutter et al. [1975], Rowe [1983], Hörmann et al. [1996], Zhang
17 et al. [2006], Fenicia et al. [2008], and Herbst et al. [2008]), but the transition
18 between these states is rarely described. The spatial variability also causes large
19 variations in the storage capacity and differs per season (Staelens et al. [2006]).
20 Furthermore, the storage capacity depends on rainfall conditions that vary over
Fo
21 time (e.g., snow/no snow, heavy rain/drizzle, wind/windless). Hence, it is not

22 possible to define one single storage capacity for a certain tree species in con-
23 trast to what many authors claim. This is especially important to realize when
24 these storage capacities are used for interception modelling.
rP
25
26 The aim of this paper is to investigate how the interception process changes
27 over the seasons and how this affects evaporation predictions of an intercep-
28 tion model. This model is used to investigate the spatial distribution and the
29
ee
persistence of spatial throughfall and infiltration patterns.

30
31
32 2 Methodology
rR
33
34 To investigate the effect of forest interception both canopy and forest floor in-
35 terception have been measured. Canopy interception is defined as precipitation
36 (Pg ) minus the sum of throughfall (Tf ) and stemflow (Ts ) (Equation 2):
37
ev
38 dSc
Ei,c + = Pg − T f − T s (2)
39 dt
40
where Ei,c is evaporation from the canopy and Sc the storage on the canopy.
41
iew
42
Forest floor interception is defined as:
43
44 dSf
45 Ei,f + = Tf − F (3)
dt
46
47 where Ei,f is evaporation from the forest floor and Sf the storage on the forest
48 floor and F the infiltration.
49
50 For both interception types we have investigated the temporal variation of
51 evaporation from interception.
52
53 The maximum storage capacity, Sc , is determined by the ’mean-method’ of
54 Klaassen et al. [1998], where the storage capacity is the negative intercept with
55
56
57 3
58
59
60
1
2
3
4
5
6
7
8 the y-axis of the linear regression line of accumulated gross rainfall versus ac-
9 cumulated throughfall of an event. In this way the effect of evaporation during
10 the event is taken into account (when the slope of the regression line is less than
11 1:1). For the storage capacity of the forest floor, Sf , the same procedure is used
12 on accumulated throughfall versus accumulated infiltration.
13
14 These storage capacities are determined for all events under different weather
15 conditions to study the seasonality (average variation over time) and variability
16 (variation within a certain season or period).
17
18 Successively, we have analyzed the consequences of this variability on pre-
19 diction with a Rutter model (Rutter et al. [1971]) to asses the prediction error
20 due to uncertainty in the storage capacity.
Fo
21
22 Finally, the spatial variability has been investigated through semi-variograms
23 and time stability plots.
24
rP
25
26 3 Site description and materials
27
28 The study area is located in the Huewelerbach basin near the village of Hov-
29 elange in Luxembourg (Figure 1a). The experimental beech stand (Fagus Syl-
ee
30 vatica) of 120 years of age is located in the North-West of the basin and has a
31 total area of 0.0596 ha. The density of the stand is of 168 trees/ha, while the
32 tree height varies from 30 to 40 m (see Figure 1b). The tree in the Northeast
rR
33 corner of the stand has been cut in December 2004 causing a gap in the canopy
34 coverage. The climate is modified oceanic with mild winters and temperate
35 summers. The average rainfall is 845 mm/a and the average temperature is 8◦ C
36 [Pfister et al., 2005].
37
ev
38 Gross precipitation, temperature, humidity, radiation, wind speed and wind

39 direction are measured every 5 minutes at the meteorological station (Camp-
40 bell Scientific) in the alluvial valley. An additional rain collector collects gross
41
iew
precipitation. Cumulative rainfall in the collector is read at a weekly time-step

42 and helps verifying the correct functioning of the tipping bucket.
43
44 Throughfall is measured both continuously and at a biweekly to monthly
45 time-step. In the experimental plot, a network of 3 gutters has been built (total
46 collecting area of 1.07 m2 ). The gutters are connected to a tipping bucket [RM
47
Young, 52203-L, 0.1 mm], so as to provide the temporal structure of throughfall
48
below the canopy. Since the representativity of these gutters is extremely diffi-
49
cult to assess, because of the high spatial variability of the canopy density, an
50
additional network of rain collectors has been installed inside the experimental
51
stand. Eighty-one collectors (41 before May 2004) have been put in a network
52
at a spacing of 3 meters. This network of rain collectors gives both an indica-
53
54 tion of the spatial variability of throughfall, as well as a precise measurement
55
56
57 4
58
59
60
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19 (a) (b)
20
Fo
21
22 Figure 1: (a) Location of the interception plot in the Huewelerbach catchment
23 (Luxembourg) and (b) the location of the equipment inside the plot
24
rP
25
26 of the amount of throughfall at a biweekly time-step. The biweekly through-
27 fall amounts are disaggregated via the measurements obtained by the tipping
28 bucket gauges that are connected to the gutters below the canopy.
29
ee
30 Stemflow is measured by an open flexible tube (width 3 cm) wrapped around

31 the trunk of the trees. The water collected from the trees is directed to a tip-
32 ping bucket (RM Young, 52203-L, 0.1 mm & Isco 674, 0.1 mm). Stemflow is
rR
33 upscaled by calculating the average stemflow per tree and dividing by the stand
34 area (Gerrits et al. [2009]).
35
36 Forest floor interception is measured with a specially developed device (Fig-
37
ev
ure 2). It consists of two aluminium basins mounted above each other. The
38
upper basin is filled with beech leaves and has a permeable bottom, so water
39
can drain into the second basin. The lower basin is watertight and is emptied
40
every day by an electronic valve. Both basins are weighed and logged every
41
iew
5 minutes. The evaporation from the forest floor is calculated from the water
42
43 balance (Equation 3), where F = dSl . A detailed description of the device can
dt
44 be found in Gerrits et al. [2007]. Additionally, the flushed water from the valve
45 is collected in a big barrel for verification.
46
47 Since the weighing sensors are made of a piece of metal with strain gauges,
48 temperature changes have an effect on the sensor output due to the expansion
49 and compression of the metal. In Gerrits et al. [2007] it is described how to
50 compensate for this effect. From 2006 onwards an additional sensor (so-called
51 ’dummy sensor’) is installed to compensate for the temperature influence. The
52 dummy sensor is a free hanging sensor with no weight on it and thus only reacts
53 on temperature changes. If the relation between the dummy sensor and the
54
55
56
57 5
58
59
60
1
2
3
4
5
6
7
8 other sensors is determined in the laboratory, we can compensate the observed
9 data.
10
11
12
13
14
15
16
17
18
19
20
Fo
21
22
23
24
rP
25
26
27 Figure 2: Schematic drawing of forest floor interception device with Ei,f evapo-
28 ration from the forest floor, El evaporation from the lower basin, and Su (= Sf )
29 and Sl the storage in the upper and lower basin respectively.
ee
30
31
32
4 Temporal variation in interception measure-
rR
33
34 ments
35
36 4.1 Canopy interception
37
ev
38 Since October 2003 throughfall is intensively measured in the Huewelerbach

39 with 41 collectors and since May 2004 with 81 collectors. In Figure 3a the gross
40 precipitation, the average throughfall, and stemflow are depicted. In Figure 3b,
41 evaporation from canopy interception as a percentage of gross rainfall is shown.
iew
42 As can be seen, there is a clear seasonal trend. When there are leaves on the
43 trees from April until September (indicated by the green bar) evaporation from
44 interception is on average about 25% of the precipitation compared to 5% in
45 winter.
46
47 Furthermore, some positive and negative outliers can be seen in the percent-
48 age of canopy interception (Figure 3b). These events often coincide with small
49 rainfall amounts (triangles), causing relatively large errors in the readings, or
50 coincide with snow or ice events (asterixs). Snow and ice can prevent precipi-
51 tation to be caught by the tipping bucket or the collector, or the recording to
52 be delayed.
53
54
55
56
57 6
58
59
60
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19 a) 140
Huewelerbach
Precipitation
20 120
Throughfall
Stemflow
Fo
21 100
22
[mm/period]
80
23 60
24 40
rP
25 20
26 0
23/02/04 07/07/05 19/11/06 02/04/08 15/08/09
27 Date [dd/mm/yy]
28
29
ee
b) 80
Canopy interception
30 60
Snow/ice
Mean Pg < 0.02 mm/h
31 40
Leaves on canopy
32
[% of P ]
g
rR
20
33
34 0
35 −20
36 −40
37 23/02/04 07/07/05 19/11/06 02/04/08 15/08/09

ev
Date [dd/mm/yy]
38
39
40
41 Figure 3: Precipitation, collector readings of throughfall (average over 81 collec-
iew
42 tors), and stemflow (a) and b) canopy interception percentage (of precipitation)
43 over time in the Huewelerbach.
44
45
46
47
48
49
50
51
52
53
54
55
56
57 7
58
59
60
1
2
3
4
5
6
7
8 4.2 Forest floor interception
9
In Figure 4 the results obtained from the forest floor interception device are
10
shown. Unlike with canopy interception, there is no clear seasonal trend. Al-
11
though, the amount of data is limited, we can see that evaporation from the
12
forest floor is rather constant over the year, with a slight increase in summer.
13
14 On average, evaporation is about 20% of the throughfall when extreme values,
15 related to snow events, are not included. Snow is causing high water equivalents
16 to be first stored in the upper basin, which will drain towards the lower basin
17 only when the temperature is again above 0◦ C. Similar to canopy interception,
18 it appears that small rainfall events (and thus throughfall events) cause rela-
19 tively high forest floor interception values.
20
Fo
21 a) Huewelerbach
150 //
22 Throughfall
Infiltration
23
24 100
rP
[mm/period]
25
26 50
27
28
29
ee
0
10/09/04 29/03/05 15/10/05
// 19/11/06 07/06/07 24/12/07 11/07/08 27/01/09 15/08/09
Date [dd/mm/yy]
30
31
32 b) 100
Forest floor interception
rR
90 Snow/ice
33 80
Mean Tf < 0.05 mm/h
//
Leaves on canopy
34 70
35 60
[% of T ]
f
50
36 40
37 30
ev
20
38 10
39 0
10/09/04 29/03/05 15/10/05
// 19/11/06 07/06/07 24/12/07 11/07/08 27/01/09 15/08/09
40 Date [dd/mm/yy]
41
iew
42
43
44 Figure 4: Throughfall and infiltration (a) and temporal variation of forest floor
45 interception as percentage of throughfall (b) (four throughfall collectors nearest
46 to forest floor interception device).
47
48 Combining results from canopy evaporation and forest floor evaporation,
49 we can conclude that in winter 24% of the precipitation is intercepted and in
50 summer as much as 40%. Hence the seasonal effect of interception, if forest
51 floor interception is included, is less strong when only based on canopy inter-
52 ception. On annual basis this amounts to 40-45% of the actual evaporation in
53 the Huewelerbach catchment, which is a considerable part of the total evapo-
54 ration, taking into account the modest dimensions of the interception reservoir
55
56
57 8
58
59
60
1
2
3
4
5
6
7
8 compared to the soil moisture storage capacity. The reason for the relatively
9 large interception flux lies in the high frequency of rainfall events as compared
10 to infiltration events.
11
12
13 5 Temporal variation in storage capacity
14
15 To determine the storage capacity of the canopy and forest floor, only those
16 events have been selected that meet the following criteria:
17 • The event is large enough to saturate the storage capacity, causing through-
18 fall to occur from the canopy or infiltration from the forest floor.
19
20 • The time between the events is long enough for the intercepted water to
Fo
21 evaporate and that no water is stored before the start of the next event.
22
23
24
5.1 Canopy interception
rP
25 For the analysis of storage capacity 52 recorded events meet the criteria defined
26 above. In Figure 5 the results are shown for the different months of the year. As
27 can be seen, there is a clear seasonal variability in the canopy storage capacity.
28 In winter the capacity is on average low: 0.4 mm and in summer when the leaves
29
ee
are on the trees the capacity is on average 0.9 mm.

30
31
32 Canopy storage capacity (Huewelerbach)
rR
33 2
34
1.5
35
36 1
37
ev
number of data points

storage capacity [mm]
38 0.5
39
0 0
40
41
iew
2
42
43 4
44
45 6
46
8
47 2 4 6 8 10 12
month of the year
48
49
50 Figure 5: Variability of the canopy storage. The black line is the mean of all
51 events in a certain month, and the grey area defines a bound ± one standard
52 deviation. The number of events is indicated by the black bars.
53
54
55
56
57 9
58
59
60
1
2
3
4
5
6
7
8 The grey area indicates a band of ± one standard deviation in the storage
9 capacity of the different events. In winter the deviation is ±0.2 mm (cv = 46%)
10 compared to summer ±0.5 mm (cv = 54%). Hence it is difficult to determine
11 one single capacity value, especially when the canopy is developed. One of the
12 reasons for the variability may be the applied regression of Klaassen et al. [1998]
13 method to determine the storage capacity. As Rowe [1983] concluded this ap-
14 proach is strongly influenced by wind speed and intensity. For example, Horton
15 [1919], Klaassen et al. [1996] and Hörmann et al. [1996] found that with increas-
16 ing wind speed the measured storage capacity is less, due to the fact that the
17 wind shakes the rain water off the leaves. This effect is also visible in Figure
18 6a, although the relationship is weak. We also tested the relationship between
19 maximum windspeed during the event and storage capacity, but the relationship
20 is even worse.
Fo
21
22 There is no consensus in the literature on the effect of rainfall intensity on the
23 storage capacity. Horton [1919] and Wang et al. [2007] concluded that the higher
24 the rainfall intensity the lower the capacity, because high rainfall intensities
rP
25 cause splashing and shaking of leaves. On the other hand, Aston [1979] and
26 Keim et al. [2006] found the opposite: high rainfall intensities coincide with
27 high storage capacities, due to dynamic storage. Based on our data we do not
28 see a clear relationship, only a weak decline in storage capacity with increasing
29
ee
average intensity (see Figure 6b). Also no relationship has been found with
30
maximum rainfall or storm size.
31
32 a) Relation wind on storage capacity b) Relation intensity on storage capacity
rR
33 2 summer (JJAS)
fall (ON)
2 summer (JJAS)
fall (ON)
34 1.8 winter (DJFM)
spring (AM)
1.8 winter (DJFM)
spring (AM)
2 2
35 1.6 Linear R = 0.20 1.6 Linear R = 0.41
1.4 1.4
36 1.2 1.2
37
ev
1 1
38 0.8 0.8
39 0.6 0.6
40 0.4 0.4
0.2 0.2
41
iew
0 0
42 0 0.5 1 1.5 2 2.5
average wind speed [m/s]
3 3.5 0 0.5 1 1.5
rainfall intensity [mm/h]
2
43
(a) (b)
44
45
46 Figure 6: Effect of average wind speed (a) and average rainfall intensity (b) on
47 canopy storage capacity during different seasons.
48
49
50 5.2 Forest floor interception
51
52 For the forest floor 74 recorded events were available. In contrast to the canopy
53 there is no apparent seasonal trend (see Figure 7). The storage capacity is
54 rather constant over the year (1.8 mm), with a temporal increase during fall
55
56
57 10
58
59
60
1
2
3
4
5
6
7
8 (2.8 mm). This can be explained by the fresh litter. The fresh leaves dry out
9 quickly after they fall off the tree and obtain a curled shape, which has a high
10 storage capacity.
11
12
13 4
Forest floor storage capacity (Huewelerbach)
14
15 3
16
2
17
number of data points

18 1
19
0 0
20
Fo
21 5
22
23 10
24
rP
15
25
26 2 4 6 8 10 12
20
27 month of the year
28
29
ee
Figure 7: Variability of the storage capacity of the forest floor. The black line is
30 the mean of all events in a given month, and the grey area defines the standard
31 deviation. The number of events is indicated by the black bars
32
rR
33
34 Furthermore, it can be seen that the variability is changing from winter to
35 summer. In contrast to the canopy storage capacity, the variability in summer
36 is lower ±0.4 mm (cv = 19%) than in winter ±0.8 mm (cv = 50%). This can be
37 explained by the effect of snow, which has been repeatedly documented through
ev
38 field observations. If a snow event occurs, the leaves are completely flattened
39 due to the snow weight, causing a small storage capacity. If no snow occurs,
40 the leaves retain their original shape, with a large storage capacity. Hence, the
41 capacity really depends on whether snow events occurred or not. This causes a
iew
42 high variability in winter values.

43
44 Additional variability may be caused by throughfall intensity. Putuhena
45 and Cordery [1996], Sato et al. [2004] and Guevara-Escobar et al. [2007] found
46 that with increasing intensity the dynamic, and also the static storage capacity
47 increased. If we test this hypothesis on our data, we find a weak positive rela-
48 tionship (Figure 8). The effect of wind is not investigated, since it is not likely
49 that wind can shake off water at forest floor level.
50
51
52
53
54
55
56
57 11
58
59
60
1
2
3
4
5
6
7
Relation intensity on storage capacity
8
summer (JJAS)
9 3.5 fall (ON)
winter (DJFM)
10 3
spring (AM)
2
Linear R = 0.10
11

2.5
12 2
13
1.5
14
15 1
16 0.5
17 0
0 0.5 1 1.5 2 2.5 3 3.5 4 4.5
18 throughfall intensity [mm/h]
19
20 Figure 8: Effect of average throughfall intensity on forest floor storage capacity
Fo
21 throughout the seasons.

22
23
24 6 Effect of variability in storage capacity on Rut-
rP
25 ter model predictions

26
27 6.1 Rutter Model description
28
29 To investigate the effect of the observed uncertainties on interception predictions
ee
30 we applied a modified Rutter model. The modified Rutter model (Rutter et al.
31 [1971]) has a forest floor interception reservoir added. In Figure 9 an overview
32 of the model structure is given. Drainage is modelled as an exponential function
rR
33 with D0 taken from Rutter et al. [1971]. Drainage is also possible when Scl <
34 Scl ,max , taking care of the shake off of rain drops by wind.
35
36 The partitioning of the gross rainfall into canopy input, free throughfall and
37 trunk input is based on our measurements. The factor p is determined from
ev
38 the leaf area index (LAI ). To determine the leaf area index the Equation of
39 Lambert-Beer [Montheith and Unsworth, 1990] is used.
40
41 p = exp(−k × LAI ) (4)
iew
42
43 with k the extinction coefficient equal to 0.43 for beech trees (Bréda [2003]).
44
45 The fraction pt is determined as in Gash and Morton [1978] and Kittredge
46 [1954]. With pt equal to 0.07. Since pt is not changing over time, we consider pt
47 to be constant. The negative intercept with the y-axis determines the storage
48 capacity of the trunk. We assume the trunk storage capacity to be a constant
49 equal to 0.11 mm.
50
51 The maximum storage capacity of the canopy (Scl ,max ) and the forest floor
52 (Sf ,max ) are determined as described in Section 2 and are dependent on the
53 season. For calibration of the adjusted Rutter model we used the mean values
54
55
56
57 12
58
59
60
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15 Canopy evaporation Trunk evaporation
Gross rainfall 
16   S l
  t Sct 
min  E p l c
, Scl  , Scl < Scl ,max

Pg min  ε E p S t , Sct  , Sct < Sct ,max

17 Eil,c =  Eit,c =   
 Sc ,max  c ,max

 min (ε E p , Sc )

( ) , Sct ≥ Sct ,max
t t
18  min E p , S l
c , Scl ≥ Scl ,max
19
20
Fo
21
22 Canopy input Free throughfall Trunk input
23 ( 1- p- pt ) Pg p Pg p t Pg
24
rP
25
26
27 S cl S ct
Scl,max
28 Sct,max
29
ee
30 Drainage
31 
D=
0 (
min D exp  d (S l − S l ) , S l
 c c ,max  c ) , Scl < Scl ,max
32 (
min D0 exp  d (Scl − Scl ,max ) , Scl ) , Scl ≥ Scl ,max

rR
33 Stemflow
34  0 , Sct < Sct ,max
Ts = 
35  min (S t
c − S t
c , S t
c ) , Sct ≥ Sct ,max
36 Throughfall, Tf
Forest floor evaporation
37
ev
  Sf 

 min  E p S ,Sf  , S f < S f ,max
38 Ei. f =  f ,max 
39 

 min (E p , S f ) , S f ≥ S f ,max
40
41
iew
42 Sf
Sf,max
Infiltration
43
44  0 , S f < S f ,max
F =
45 min (F0 exp( f ), S f ) , S f ≥ S f ,max
46
47 Figure 9: Overview of the adapted Rutter model.
48
49
50
51
52
53
54
55
56
57 13
58
59
60
1
2
3
4
5
6
7
8 from Figure 5 and 7, respectively.
9
10 The forest floor reservoir is modelled in the same way as the canopy reser-
11 voir: a threshold (Sf ) and an exponential infiltration rate, F ; however infiltra-
12 tion equals zero if Sf < Sf ,max , because wind will not shake off rain drops at
13 forest floor level.
14
15 Potential evaporation (Ep ) is calculated with the Penman equation [Allen
16 et al., 1998]. The potential evaporation for the forest floor is slightly lowered
17 by increasing the relative humidity with 2%. In Table 1 the model parame-
18 ters are given. The model is evaluated on throughfall, infiltration, forest floor
19 evaporation, and forest floor wetness.
20
Fo
21 Parameter Value Parameter Value

22 p [-] Equation 4 ǫt [-] 0.5
23 pt [-] 0.07 D0 [L T−1 ] 0.03 mm/15 min
24 Scl ,max [L] Figure 5 d [L−1 ] 0.9 mm−1
rP
25 Sct ,max [L] 0.11 mm F0 [L T−1 ] 0.04 mm/15 min

26 Sf ,max [L] Figure 7 f [L−1 ] 2.0 mm−1
27
28 Table 1: Model parameters of adjusted Rutter model.
29
ee
30
31
32 6.2 Results
rR
33 In Table 2 the model results of the adjusted Rutter model are presented for
34 6 simulation periods. Given are the Root Mean Square Error (RMSE), the
35 relative bias and the Nash-Sutcliffe efficiency. Overall the performance of the
36 model is reasonable, although the goodness-of-fit measures are sometimes poor.
37 RMSE is not very good due to fluctuations that may be caused by the temper-
ev
38 ature sensitivity of the forest floor interception device and phase lags in loggers.
39 However, the bias is generally small. Throughfall is best modelled, especially in
40 winter time. In summer time, due to the variability in the storage capacity, the
41
iew
performance is lower. However, it should also be mentioned that the results are
42 not independent. If throughfall is wrongly modelled, this causes a wrong input
43 to the forest floor reservoir.
44
45
Furthermore, the effect of the variability in the storage capacity is visible.
46
The Rutter model is calibrated on the mean storage capacity per month. How-
47
ever, it can really vary between years when for example the leaf growth or fall
48
starts. This timing has a large impact on the actual storage capacity and can
49
thus deviate from the mean storage capacity for that specific month. This effect
50
can be the reason for the lower performance of the simulations in fall and spring.
51
52
53 To investigate the effect of the variability of the storage capacity on the
54 modelled evaporation we applied the model with a low, a mean, and a high
55
56
57 14
58
59
60
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16 Run RMSE rel. bias NS
17 [%] [-]
18 SIM1 Tf 3.55 mm/d 4 0.59
19 02-11-04 F 2.61 mm/d -5 0.62
20 to Ei,f 0.74 mm/d 5 0.24
Fo
21 01-12-04 Sf 0.37 mm 11 0.53

22 SIM2 Tf 9.34 mm/d 9 0.30
23 21-06-05 F 11.07mm/d 30 0.02
24 to Ei,f 2.06 mm/d 0 -1.02
rP
25 17-08-05 Sf 0.64 mm -40 0.22

26 SIM3 Tf 5.80 mm/d 8 0.71
27 14-12-06 F 4.26 mm/d 1 0.66
28 to Ei,f 1.14 mm/d 40 0.16
29
ee
11-01-07 Sf 0.60 mm -21 0.66

30 SIM4 Tf 4.81 mm/d 3 0.83
31
07-02-07 F 4.21 mm/d 0 0.72
32
to Ei,f 1.41 mm/d -9 0.10
rR
33
27-02-07 Sf 1.10 mm -4 0.37
34
SIM5 Tf 3.52 mm/d 7 0.81
35
04-12-07 F 7.14 mm/d 4 -0.18
36
37 to Ei,f 0.98 mm/d 98 0.14
ev
38 07-01-08 Sf 0.93 mm -34 0.37

39 SIM6 Tf 5.16 mm/d 11 0.70
40 07-01-08 F 4.43 mm/d -3 0.65
41 to Ei,f 0.81 mm/d 45 -0.44
iew
42 23-01-08 Sf 0.37 mm 14 0.42

43
44 Table 2: Model results of the adjusted Rutter model for throughfall (Tf ), infil-
45 tration (F ), forest floor evaporation (Ei,f ), and forest floor wetness (Sf ). Given
46 are the root mean square error, relative bias, and Nash-Sutcliffe efficiency.
47
48
49
50
51
52
53
54
55
56
57 15
58
59
60
1
2
3
4
5
6
7
8 storage capacity. For Scl ,max and Sf ,max we used the upper and lower limits
9 of Figure 5 and 7. In Figure 10 the results for SIM1 are shown and in Table 3
10 the mean evaporation rates for all six simulations. In Figure 11 the results are
11 graphically shown.
12
13 As can be seen in Figure 10 the observed data fits between the upper and
14 lower estimates. However, sometimes the model prediction deviates from the
15 observed data. This is mainly the case during the wetting, thus on the rising
16 limb, and not on the falling limb during drainage and evaporation. During the
17 falling limb the model performs well, showing that the Rutter model is capable
18 of predicting evaporation. The rising limb gives more problems, which can be
19 explained by variability of the storage capacity in the instrument. This may,
20 for example, be due to unequal litter distribution in the instrument caused by
Fo
21 wind or snow. Figure 3 and 4 illustrate the temporal variability of the storage
22 capacity.
23
24 As can be seen from Table 3 the effect of the storage capacity on canopy
rP
25 interception evaporation is limited. On average the increase or decrease is about

26 5% with an average coefficient of variation in the storage capacity of 56%. Hence
27 a large variation in the storage capacity has a low impact on the evaporation
28 predictions, and thus canopy interception is more driven by the number of rain-
29
ee
day and the potential evaporation than by the storage capacity.

30
31
This is the opposite for the forest floor. Here the average coefficient of vari-
32
ation of the forest floor storage capacity is 48%, and the average increase or
rR
33
decrease is 12%. Hence the influence of the storage capacity on evaporation
34
predictions is higher, indicating that forest floor interception is more driven by
35
36 the storage capacity than by the number of rainday and the potential evapora-
37 tion. This supports the findings of Baird and Wilby [1999] and Gerrits et al.
ev
38 [2007].
39
40 Furthermore, the effect of the lower limit of the canopy storage capacity is,
41 as expected, higher than of the upper limit of the storage capacity: 6% and 4%
iew
42 respectively. This is also the case with the forest floor: 16% for the lower limit
43 and 9% for the upper limit.
44
45 The impact of uncertainties in the storage capacity (which can be as high as
46 ±100%) on the total interception evaporation is about 11% and the difference in
47 the lower or upper storage capacity is 15% and 8% respectively. This indicates
48 that interception is more influenced by the rainfall pattern than by the storage
49 capacity. Hence, in interception modelling, the value of the storage capacity is
50 of minor concern.
51
52 Furthermore, Table 3 shows the relative importance of canopy, trunk, and
53 forest floor interception evaporation. The model results show that in fall (SIM1)
54 4% of precipitation evaporates from the canopy (6% observed) and 33% from
55
56
57 16
58
59
60
1
2
3
4
5
6
7
8
9
10
11
12
13
14 02−Nov−2004 to 01−Dec−2004
15 Throughfall
16 60
Modelled−avg
17 Modelled−lower
Modelled−upper
40
18 Observed
[mm]
19 20
20
Fo
21 0
0 500 1000 1500 2000 2500 3000
22 Infiltration
23 40
Modelled−avg
24 30 Modelled−lower
rP
Modelled−upper
25 Observed
[mm]
20
26
10
27
28 0
0 500 1000 1500 2000 2500 3000
29
ee
Evaporation forest floor

30 20
Modelled−avg
Modelled−upper
32 Observed
[mm]
10
rR
33
5
34
35 0
0 500 1000 1500 2000 2500 3000
36 Wetness forest floor
37 8
ev
Modelled−avg
Modelled−upper
39 Observed
[mm]
4
40
2
41
iew
42 0
0 500 1000 1500 2000 2500 3000
43 time (x 15 min)
44
45
46
47
48 Figure 10: Model results for SIM1. The green line represents the results with
49 the lower limit storage capacity, the blue the mean, and the red line the upper
50 limit storage capacity.
51
52
53
54
55
56
57 17
58
59
60
1
2
3
4
5
6
7
8
9
10
11
12
13
Run P̄
[mm/day]
Fo S max E¯i,c
l
[mm/day]
E¯i,c
t
[mm/day]
E¯i,c = E¯i,c
[mm/day]
l + E¯t
i,c
obs.
E¯i,f
[mm/day] obs.
Ēi
[mm/day] obs.
rP
lower 0.03 (2%) 0.04 (3%) 0.08 (5%) 0.42 (25%) 0.50 (30%)
14 SIM1 1.69 (100%)
mean 0.02 (1%) 0.04 (3%) 0.07 (4%) [6%] 0.56 (33%) [31%] 0.63 (37%) [37%]
15
(02-11-04 to 01-12-04) upper 0.02 (1%) 0.04 (3%) 0.07 (4%) 0.65 (39%) 0.72 (43%)
16
ee
17 lower 0.19 (9%) 0.07 (3%) 0.26 (12%) 0.70 (32%) 0.90 (43%)
SIM2 2.22 (100%)
18 mean 0.23 (10%) 0.07 (3%) 0.29 (13%) [20%] 0.81 (36%) [42%] 1.10 (50%) [63%]
19 (21-06-05 to 17-08-05) upper 0.30 (13%) 0.07 (3%) 0.36 (16%) 0.85 (38%) 1.2 (54%)
20
21
22
SIM3 3.11 (100%)
(14-12-06 to 11-01-07)
lower
mean
upper
0.06 (2%)
0.06 (2%)
0.07 (2%) rR
0.05 (2%)
0.05 (2%)
0.05 (2%)
0.11 (3%)
0.11 (4%)
0.11 (4%)
[7%]
0.45 (14%)
0.51 (16%)
0.55 (18%)
[13%]
0.55 (18%)
0.62 (20%)
0.66 (21%)
[19%]
ev
18
23 lower 0.19 (4%) 0.08 (2%) 0.27 (6%) 0.53 (11%) 0.80 (17%)
SIM4 4.78 (100%)
24 mean 0.19 (4%) 0.08 (2%) 0.27 (6%) [6%] 0.67 (14%) [14%] 0.94 (20%) [27%]
iew
25 (07-02-07 to 27-02-07) upper 0.19 (4%) 0.08 (2%) 0.27 (6%) 0.74 (15%) 1.00 (21%)
26 lower 0.04 (2%) 0.04 (2%) 0.07 (3%) 0.42 (18%) 0.49 (21%)
SIM5 2.29 (100%)
27 mean 0.04 (2%) 0.04 (2%) 0.08 (3%) [4%] 0.46 (20%) [13%] 0.54 (23%) [17%]
28 (04-12-07 to 07-01-08) upper 0.05 (2%) 0.04 (2%) 0.08 (4%) 0.48 (21%) 0.56 (25%)
29 lower 0.07 (2%) 0.06 (2%) 0.13 (4%) 0.58 (18%) 0.72 (22%)
SIM6 3.28 (100%)
30 mean 0.08 (2%) 0.06 (2%) 0.13 (4%) [10%] 0.71 (22%) [13%] 0.85 (26%) [24%]
31 (07-01-08 to 23-01-08) upper 0.08 (2%) 0.06 (2%) 0.13 (4%) 0.79 (24%) 0.93 (28%)
32
33 Table 3: Model results of average interception evaporation compared to average rainfall over the simulation periods. In bold
34 the observed percentages.
35
36
37
38
39
40
41
42
43
44
45 http://mc.manuscriptcentral.com/hyp
46
47
1
2
3
4
5
6
7
8 Results Rutter−model
100
9 Canopy
10 90 Forest floor
Total
11 80
12
70
13 Modelled [% of P]
14 60
15 50
16
40
17
18 30
19 20
20
Fo
10
21
22 0
0 20 40 60 80 100
23 Observed [% of P]
24
rP
25
26 Figure 11: Graphical representation of Table 3, showing the performance of
27
the Rutter model for canopy interception, forest floor interception, and total
28
interception.
29
ee
30
31 the forest floor (31% observed). In winter (SIM 3-6) this is 4% (observed 7%)
32 from the canopy and 18% (observed 13%) from the forest floor. There are no
rR
33 simulations in spring. And in summer (SIM2) 13% (observed 20%) from the
34 canopy and 36% (observed 42%) from the forest floor.
35
36
37
ev
38 7 Spatial variation in throughfall and infiltra-

39
40 tion
41
iew
Besides the temporal variation, interception varies also in space, mainly due to
42
the heterogeneity of the vegetation density. In Figure 12a an example of the
43
spatial distribution of observed throughfall is given. It encompasses a period in
44
45 spring 2006 with 70 mm of gross rainfall. It can be clearly seen that the forest
46 canopy redistributes the rainfall. In general throughfall is lower around trees,
47 due to interception. However, the trees can also funnel the rainfall, as can be
48 seen near the tree at coordinates (15m, 15m). The structure of this tree really
49 acts like a funnel, causing even higher throughfall than gross rainfall close to the
50 tree, and lower throughfall values around the tree. This effect is also mentioned
51 by (Germer et al. [2006] and Gerrits et al. [2009])
52
53 In Figure 12b the resulting canopy interception pattern is shown. Evapora-
54 tion from canopy interception is here calculated as gross rainfall minus through-
55
56
57 19
58
59
60
1
2
3
4
5
6
7
8 fall and hence stemflow is neglected. Because throughfall can exceed gross pre-
9 cipitation, this results in negative interception evaporation values. These values
10 have been removed from the analysis.
11
12
13 A.
24−Mar−2006 to 06−Apr−2006
P=69.6mm C.
Throughfall [mm] Infiltration [mm]
14 30
(observed)
30
(modelled)
15 trees
collectors
160
trees
collectors 140
16 25
N 140
25
N
120
17 20 20
Distance [m]
Distance [m]
120
18 15 15
100
19 100 80
20 10 10
Fo
80 60
21 5 5
60
22 0 0
40
23 0 5 10 15
Distance [m]
20 25 0 5 10 15
Distance [m]
20 25
24
rP
25 B.
Canopy interception evaporation [mm]
D.
Forest floor interception evaporation [mm]
(observed) (modelled)
26 30
trees
30
trees
18.6
27 25
collectors
20 25
collectors
N N 18.4
28 18.2
20 20
29
ee
15
Distance [m]
Distance [m]
18
30 15
10
15 17.8
31 10 10
17.6
32 5
17.4
rR
17.2
33 5 5
17
34 0
0 5 10 15 20 25
0
0
0 5 10 15 20 25
35 Distance [m] Distance [m]
36
37
ev
38
Figure 12: Spatial variability of canopy interception (observed) and forest floor
39
interception (modelled with adjusted Rutter model) in mm over the period 24
40
March to 6 April 2006. Triangles indicate the position of the beech trees and
41
iew
the circles the positions of the collectors.

42
43
44 Since we do not have spatial observations of forest floor interception, we used
45 the adjusted Rutter model. To reduce the modelling error as much as possible
46 we used observed throughfall data as input for the forest floor module instead of
47 modelled throughfall. In Figure 12c and 12d the model results of the infiltration
48 and forest floor evaporation are shown. In order to see if the spatial pattern of
49 canopy interception is different to that of the forest floor, and to investigate if
50 the pattern changes throughout the seasons we calculated the spatial correlation
51 with semi-variograms.
52
53 We used the method as described by Keim et al. [2005] to calculate the
54
55
56
57 20
58
59
60
1
2
3
4
5
6
7
8 semi-variogram:
P 2
9 ex,y − N
N ex,y+h
n(h)
10 γ(h) = (5)
11 2n(h)
12 Where h is the lag, n(h) is the number of measurements pairs in the data set
13 that are distance h apart. And N ex,y the normalized throughfall or infiltration
14 at measuring point (x, y):
15
16 Nex,y = Nx,y − N̄ (6)
σ(N )
17
18 with σ the standard deviation. We used normalized data to be able to compare
19 throughfall and infiltration with each other.
20
Fo
21 In Figure 13 the semi-variograms of the throughfall and infiltration are shown

22 per season as well as per year. The range r is defined as the lag h, whereby the
23 variance (γ) is 95% of the sill c, and is a measure for the correlation between
24 the points. High spatial correlation between the collectors causes the range to
rP
25 be high and vice versa. We fitted an exponential model (Chilès and Delfiner
26 [1999]):
27 −3h
γ(h) = c(1 − exp( )) (7)
28 r
29
ee
As expected the range in winter is higher (higher spatial correlation) after leaf
30 senescence; however, this not a clear relation. Furthermore, it seems that in
31 summer and winter the range is relatively higher than during the transition
32 seasons. This is both the case for throughfall and for infiltration.
rR
33
34
To investigate if these patterns are persistent in time, time stability plots
35
(TSP) can show if there exist persistent dry or wet collectors by ranking the
36 ex,y . From a time stability plot two types of
37 normalized throughfall/infiltration N
ev
38 persistence can be derived: extreme persistence and general persistence (Keim

39 et al. [2005] and Zimmermann et al. [2008]). Extreme persistence occurs if steep
40 tails exists, and general persistence refers to the overall slope of the middle
41 range. In Figure 14 the time stability plots of throughfall (a) and (modelled)
iew
42 infiltration (b) are presented. As can be seen, the difference between the TSP of
43 throughfall and infiltration is small and extreme persistence only occurs in the
44 ’wet’ tail. The general persistence is rather weak. Only 25%, 7%, 14%, 7% and
45 7% of the collectors are significantly (65%) drier than the mean for summer,
46 fall, winter, spring, and annual respectively. This suggests that the variation in
47 winter and summer is smaller than in the transition seasons.
48
49 In Figure 15 and 16 the mean time stability plots (red line in Figure 14) are
50 plotted in space to investigate where the drier and wetter spots are located and
51 how they vary throughout the seasons. As can be seen, the collector near the tree
52 at coordinates (15m, 15m) is consistently wetter than the surrounding collectors.
53 Furthermore, it can be seen that a second drip point develops at coordinates
54 (20m, 0m) in fall which vanishes again in spring. Hence this drip point is really
55
56
57 21
58
59
60
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18 Summer (JJAS) Summer (JJAS)
2 2
19 Range= 6.73
Sill= 1.00 fitted curve
Range= 6.55
2 2
20
γ (h)
γ (h)
1 R = 0.55 1 R = 0.53
Fo
21 0 0
0 5 10 15 20 0 5 10 15 20
22 h [m]
Fall (ON)
h [m]
Fall (ON)
23 2
Range= 5.87
2
Range= 5.54
24 2 2
γ (h)
γ (h)
1 R = 0.52 1 R = 0.46
rP
25 0 0
0 5 10 15 20 0 5 10 15 20
26 h [m] h [m]
Winter (DJFM) Winter (DJFM)
27 2
Range= 7.39
fitted curve
2
Range= 6.82
fitted curve
Sill= 1.00 Sill= 1.00
28 2 2
γ (h)
γ (h)
1 R = 0.54 1 R = 0.53
29
ee
0 0
30 0 5 10
h [m]
15 20 0 5 10
h [m]
15 20
31 3
Range= 5.85
Spring (AM)
3
Range= 5.73
Spring (AM)
fitted curve fitted curve

32 2
Sill= 1.00
2 2
Sill= 1.00
2
γ (h)
γ (h)
R = 0.48 R = 0.44
rR
33 1 1
34 0
0 5 10
h [m]
15 20
0
0 5 10
h [m]
15 20
35 3
Annual
3
Annual
Range= 6.35 Range= 6.19
36 2
Sill= 1.00
2
data
fitted curve 2
Sill= 1.00
2
data
fitted curve
γ (h)
γ (h)
R = 0.53 R = 0.50
37 1 1
ev
38 0
0 5 10 15 20
0
0 2 4 6 8 10 12 14 16 18
h [m] h [m]
39
40
41
iew
(a) (b)
42
43
44
45 Figure 13: Semi-variogram of throughfall (a) and (modelled) infiltration (b) per
46 season and per year.
47
48
49
50
51
52
53
54
55
56
57 22
58
59
60
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18 Summer (JJAS) Summer (JJAS)
19 6
4
6
4
20 2 2
Fo
0 0
21 −2
0 10 20 30 40 50 60 70 80
−2
0 10 20 30 40 50 60 70 80
22 Fall (ON) Fall (ON)
23 6
4
6
4
24 2 2
rP
0 0
25 −2 −2
0 10 20 30 40 50 60 70 80 0 10 20 30 40 50 60 70 80
26 Winter (DJFM) Winter (DJFM)
27 6
4
6
4
28
ex,y
ex,y
2 2
N
0 0
29
ee
−2 −2
0 10 20 30 40 50 60 70 80 0 10 20 30 40 50 60 70 80
30 Spring (AM) Spring (AM)
31 6 6
4 4
32 2 2
rR
33 0
−2
0
−2
0 10 20 30 40 50 60 70 80 0 10 20 30 40 50 60 70 80
34 Annual Annual
35 6 6
4 4
36 2 2
37 0 0
ev
−2 −2
38 0 10 20 30 40
collector
50 60 70 80 0 10 20 30 40
collector
50 60 70 80
39
40
41 (a) (b)
iew
42
43
44 Figure 14: Time stability plot of normalized throughfall (a) and (modelled)
45 infiltration (b) per season and per year. The grey areas indicates one time the
46 standard deviation.
47
48
49
50
51
52
53
54
55
56
57 23
58
59
60
1
2
3
4
5
6
7
8 determined by the branch structure and less by the canopy structure. It is also
9 possible to see by the dark blue rings around the drip points that the funneled
10 rain is collected from the surrounding areas. Furthermore, the effect of the
11 canopy development over time is visible. In summer the spatial pattern is much
12 more heterogeneous. Also the forest gap in the upper left corner receives in
13 summer relatively more throughfall than the covered parts.
14
15
16
Summer (JJAS) Fall (ON)
17 30 30
trees trees
18 collectors 5 collectors 5
25 25
19 N
4
N
4
20 20 20
Fo Distance [m]
Distance [m]
3 3
21 15 15
2 2
22
23 10 1 10 1
24 5 0 5 0
rP
25 −1 −1
0 0
26 0 5 10 15
Distance [m]
20 25 0 5 10 15
Distance [m]
20 25
27
28 Winter (DJFM) Spring (AM)
29
ee
30 30
trees trees
collectors collectors
30 25
5
25
5
N N
31 4 4
20 20
32
Distance [m]
Distance [m]
3 3
rR
33 15
2
15
2
34 10 10
1 1
35
36 5 0 5 0
37 −1 −1
ev
0 0
0 5 10 15 20 25 0 5 10 15 20 25
39
40
41
iew
Figure 15: Spatial persistence of normalized throughfall. Triangles indicate the

42
43 position of the beech trees and the circles the positions of the collectors.
44
45
46
47 8 Conclusions
48
Evaporation from the canopy shows a clear seasonal trend. In summer evapora-
49
tion from the canopy is about 25% of the precipitation and in winter about 5%.
50
There is less seasonal variation in the forest floor evaporation. Here the evapo-
51
ration is relatively constant about 20% of the throughfall. Both in the canopy
52
and on the forest floor outliers exitst, which often coincide with low rainfall
53
54 amounts (causing high ratios) or with snow events (causing low or even nega-
55
56
57 24
58
59
60
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18 30
Summer (JJAS)
30
Fall (ON)
19 trees
collectors 5
trees
collectors 5
20 25
N
25
N
Fo
4 4
21 20 20
Distance [m]
Distance [m]
3 3
22
15 15
23 2 2
24 10 1 10 1
rP
25 5 0 5 0
26 −1 −1
0 0
27 0 5 10 15
Distance [m]
20 25 0 5 10 15
Distance [m]
20 25
28
29
ee
Winter (DJFM) Spring (AM)

30 30
trees
30
trees
31 25
collectors 5
25
collectors 5
32 N
4
N
4
rR
33 20 20
Distance [m]
Distance [m]
3 3
34 15 15
2 2
35
10 10
36 1 1
37 5 0 5 0
ev
38 0
−1
0
−1
0 5 10 15 20 25 0 5 10 15 20 25
40
41
iew
42
43 Figure 16: Spatial persistence of normalized (modelled) infiltration. Triangles
44 indicate the position of the beech trees and the circles the positions of the
45 collectors.
46
47
48
49
50
51
52
53
54
55
56
57 25
58
59
60
1
2
3
4
5
6
7
8 tive ratios). Combining canopy and forest floor interception, we can conclude
9 that in winter 24% and in summer 40% of the rainfall evaporates. On annual
10 basis this amounts to 40-45% of the actual evaporation in the Huewelerbach
11 catchment, which is a considerable part of the total evaporation.
12
13 A similar seasonal effect can be found for the storage capacities: for the
14 canopy the average storage capacity ranges from 0.4 mm with a deviation of
15 ±0.2 mm (cv = 46%) in winter to 0.9 mm in summer with a deviation of ±0.5
16 mm (cv = 54%). Hence the variation in the storage capacity is slightly higher in
17 summer. The storage capacity of the forest floor is rather constant around 1.8
18 mm, with a temporal increase to 2.8 mm in fall. The highest variability of the
19 storage capacity of the forest floor is in winter (±0.8 mm, cv = 50%), mainly
20 due to the occurrence, or no occurrence, of snow which decreases the storage by
Fo
21 flattening the leaves. The variation in summer is ±0.4 mm (cv = 19%).

22
23 Hence, we can conclude that besides the seasonal trend there is variability.
24 This variability can be caused by wind or by rain intensity effects; however,
rP
25 none of these factors has revealed itself as a strong forcing of the documented
26 variabilities. To investigate these uncertainties we applied the Rutter intercep-
27 tion model with an extra forest floor reservoir. The model was calibrated on
28 the mean storage capacities of both the canopy and the forest floor for six sim-
29
ee
ulation periods. The overall performance of the model was good, although the
30
error measure sometimes indicates poor performance, likely due to temperature
31
sensitivity of the forest floor device and difference in synchronization of the dif-
32
ferent loggers.
rR
33
34
The model shows that a variation in the canopy storage capacity of 56%
35
36 results in only 5% difference in evaporation predictions from the canopy, while
37 a variation of 48% in the storage capacity of the forest floor results in 12%
ev
38 difference in evaporation from the forest floor. Hence, we can conclude that
39 canopy interception is less determined by the storage capacity, but more by the
40 number of raindays and the potential evaporation, and that for the forest floor
41 the storage capacity is more important.
iew
42
43 The Rutter model was also applied to investigate the spatial distribution
44 of throughfall and infiltration. The spatial patterns show that also the canopy
45 coverage can cause variation in the calculated storage capacity. For example,
46 it makes a difference if the storage capacity is determined in a relatively open
47 area or in a dense location in the forest. Furthermore, trees can create hot
48 spots, where throughfall is higher than gross precipitation. In this case it even
49 becomes impossible to determine the storage capacity.
50
51 For the experimental beech plot in Luxembourg we found that the spatial
52 correlation is about 6-7 meters and that in summer and winter the spatial corre-
53 lation is slightly higher than in fall and spring. Similar results are found for the
54 infiltration pattern. Also the stability of the patterns was analyzed. Although
55
56
57 26
58
59
60
1
2
3
4
5
6
7
8 the extreme persistence at the wet tail was high, the general persistence was
9 rather weak. Plotting the time stability plots in space we found that one drip
10 point remains throughout the season, but that another drip point only occurs
11 in winter. Similar results are found for the (modelled) infiltration patterns.
12
13 Overall, we can conclude that interception is a highly variable process. Both
14 in time as in space the differences are considerable. Although the variation in
15 storage capacity can be as high as ±100%, the effect on evaporation estimates is
16 relatively low (ca. 11%). This indicates that interception is more influenced by
17 the rainfall pattern than by the storage capacity. As a consequence, in intercep-
18 tion modelling, the value of the storage capacity appears to be of minor concern.
19
20
Fo
21
22 9 Acknowledgment
23
24 We are very grateful to Jean-François Iffly and Cyrille Tailliez for regularly
rP
25 reading out the equipment and for maintaining the plot. Furthermore, the
26 authors would like to thank the Ministry of Culture, Higher Education and
27 Research of Luxembourg (FNR) and Delft Cluster, the Netherlands, for their
28 support of this research.
29
ee
30
31 References
32
Allen, R. G., Pereira, L. S., Raes, D. and Smith, M.: 1998, Crop evapotran-
rR
33
34 spiration - Guidelines for computing crop water requirements, Vol. 56, FAO -
35 Food and Agriculture Organization of the United Nations.
36 Aston, A.: 1979, Rainfall interception by eight small trees, Journal of Hydrology
37
ev
42(3-4), 383–396.
38
URL: http://www.sciencedirect.com/science/article/B6V6C-487HR7J-
39
4M/2/819359a615abca2086d7e2c93967506d
40
41
iew
Baird, A. J. and Wilby, R. L.: 1999, Eco-hydrology - Plants and water in ter-
42 restrial and aquatic environments, Routledge.
43
44 Bouten, W., Heimovaara, T. J. and Tiktak, A.: 1992, Spatial patterns of
45 throughfall and soil water dynamics in a Douglas fir stand, Water Resources
46 Research 28, –.
47 URL: http://dx.doi.org/10.1029/92WR01764
48
49 Bréda, N. J. J.: 2003, Ground-based measurements of leaf area index: a review
50 of methods, instruments and current controversies, Journal of Experimental
51 Botany 54(392), 2403–2417.
52
Breuer, L., Eckhardt, K. and Frede, H.-G.: 2003, Plant parameter values for
53
models in temperate climates, Ecological Modelling 169(2-3), 237 – 293.
54
55
56
57 27
58
59
60
1
2
3
4
5
6
7
8 URL: http://www.sciencedirect.com/science/article/B6VBS-49MF193-
9 1/2/f99ae4fa953b53b8e557fac84c9b3a57
10
Chilès, J. P. and Delfiner, P.: 1999, Geostatistics. Modeling spatial uncertainty,
11
12 John Wiley & Sons.
13 Fenicia, F., Savenije, H. H. G. and Avdeeva, Y.: 2008, Anomaly in the rainfall-
14 runoff behaviour of the Meuse catchment. Climat, land use, or land use man-
15 agement?, Hydrology and Earth System Sciences Discussion 5, 1787–1819.
16
17 Gash, J. H. C. and Morton, A. J.: 1978, An application of the rutter model
18 to the estimation of the interception loss from thetford forest, Journal of
19 Hydrology 38(1-2), 49 – 58.
20 URL: http://www.sciencedirect.com/science/article/B6V6C-487DCSY-
Fo
21 MS/2/d5715bd3aa80497ccff840853f51bd01
22
23 Germer, S., Elsenbeer, H. and Moraes, J. M.: 2006, Throughfall and temporal
24 trends of rainfall redistribution in an open tropical rainforest, south-western
rP
25 Amazonia (Rondonia, Brazil), Hydrology and Earth System Sciences 10, 383–
26 393.
27 Gerrits, A. M. J., Savenije, H. H. G., Hoffmann, L. and Pfister, L.: 2007, New
28 technique to measure forest floor interception - an application in a beech forest
29
ee
in Luxembourg, Hydrology and Earth System Sciences 11, 695–701.

30
31 Gerrits, A. M. J., Savenije, H. H. G. and Pfister, L.: 2009, Canopy and forest
32 floor interception and transpiration measurements in a mountainous beech
rR
33 forest in Luxembourg, IAHS Redbook 326, 18–24.

34
35 Guevara-Escobar, A., Gonzalez-Sosa, E., Ramos-Salinas, M. and Hernandez-
36 Delgado, G. D.: 2007, Experimental analysis of drainage and water storage
37 of litter layers, Hydrology and Earth System Sciences 11(5), 1703–1716.
ev
38 URL: http://www.hydrol-earth-syst-sci.net/11/1703/2007/
39
Herbst, M., Rosier, P. T., McNeil, D. D., Harding, R. J. and Gowing, D. J.:
40
2008, Seasonal variability of interception evaporation from the canopy of a
41
iew
mixed deciduous forest, Agricultural and Forest Meteorology 148(11), 1655–

42
1667.
43
URL: http://www.sciencedirect.com/science/article/B6V8W-4SVM0V3-
44
45 1/2/c7084bf875e30e4de8cf982afbc4d6f8
46 Hörmann, G., Branding, A., Clemen, T., Herbst, M., Hinrichs, A. and Thamm,
47 F.: 1996, Calculation and simulation of wind controlled canopy interception
48 of a beech forest in Northern Germany, Agricultural and Forest Meteorology
49 79(3), 131–148.
50 URL: http://www.sciencedirect.com/science/article/B6V8W-3Y3YRYM-
51 1/2/bae721e87b88cd210a892c32e19bc0a8
52
53 Horton, R. E.: 1919, Rainfall interception, Monthly Weather Review 47(9), 603–
54 623.
55
56
57 28
58
59
60
1
2
3
4
5
6
7
8 Keim, R. F., Skaugset, A. E. and Weiler, M.: 2005, Temporal persistence of
9 spatial patterns in throughfall, Journal of Hydrology 314, 263–274.
10
11 Keim, R. F., Skaugset, A. E. and Weiler, M.: 2006, Storage of water on verge-
12 tation under simulated rainfall of varying intensity, Advances in Water Re-
13 sources 29, 974–986.
14 Kittredge, J.: 1948, Forest influences, McGraw-Hill Book Co, New York.
15
16 Kittredge, J.: 1954, The influence of shade and litter of pine on evaporation
17 from a clay loam soil at Berkeley, California, Ecology 35(3), 397–405.
18
19 Klaassen, W., Bosveld, F. and deWater, E.: 1998, Water storage and evap-
20 oration as constituents of rainfall interception, Journal of Hydrology 212-
Fo
21 213, 36–50.
22
23 Klaassen, W., Lankreijer, H. J. M. and Veen, A. W. L.: 1996, Rainfall
24 interception near a forest edge, Journal of Hydrology 185(1-4), 349–361.
rP
25 URL: http://www.sciencedirect.com/science/article/B6V6C-3VVT4KR-
26 M/2/63dcb387fd4f3e63a33300ede2fc2a64
27 Montheith, J. L. and Unsworth, M. H.: 1990, Principles of Environmental
28 Physics, 2 edn, Edward Arnold.
29
ee
30 Pfister, L., Wagner, C., Vansuypeene, E., Drogue, G. and Hoffmann, L.: 2005,
31 Atlas climatique du Grand-Duché de Luxembourg, Johnen-Druck GmbH &
32 Co, Bernkastel-Kues.
rR
33
34 Putuhena, W. and Cordery, I.: 1996, Estimation of interception capacity of the
35 forest floor, Journal of Hydrology 180, 283–299.
36
37 Rowe, L.: 1983, Rainfall interception by an evergreen beech forest, Nelson,
ev
38 New Zealand, Journal of Hydrology 66(1-4), 143–158.

39 URL: http://www.sciencedirect.com/science/article/B6V6C-487D1PJ-
40 2D/2/581da412e6d1878de1d36213da66622d
41
iew
Rutter, A. J., Kershaw, K. A., Robins, P. C. and Morton, A. J.: 1971, A predic-
42 tive model of rainfall interception in forests. I. Derivation of the model and
43
comparison with observations in a plantation of Corsican pine., Agricultural
44
Meteorology 9, 367–384.
45
46 Rutter, A. J., Morton, A. J. and Robins, P. C.: 1975, A predictive model of
47 rainfall interception in forests. II Generalization of the model and comparison
48 with observations in some coniferous and hardwood stands, Journal of Applied
49 Ecology 12, 367–380.
50
51 Sato, Y., Kumagai, T., Kume, A., Otsuki, K. and Ogawa, S.: 2004, Experi-
52 mental analysis of moisture dynamics of litter layers - the effect of rainfall
53 conditions and leaf shapes, Hydrological Processes 18, 3007–3018.
54
55
56
57 29
58
59
60
1
2
3
4
5
6
7
8 Savenije, H. H. G.: 1997, Determination of evaporation from a catchment water
9 balance at a monthly time scale, Hydrology and Earth System Sciences 1, 93–
10 100.
11
12 Staelens, J., De Schrijver, A., Verheyen, K. and Verhoest, N. E. C.: 2006, Spatial
13 variability and temporal stability of throughfall water under a dominant beech
14 (Fagus sylvatica L.) tree in relationship to canopy cover, Journal of Hydrology
15 330, 651–662.
16
Viville, D., Biron, P., Granier, A., Dambrine, E. and Probst, A.: 1993, Inter-
17
ception in a mountainous declining spruce stand in the Strengbach catchment
18
(Vosges, France), Journal of Hydrology 144, 273–282.
19
20 Wang, A., Diao, Y., Pei, T., Jin, C. and Zhu, J.: 2007, A semi-theoretical model
Fo
21 of canopy rainfall interception for a broad-leaved tree, Hydrological Processes

22 21(18), 2458–2463.
23 URL: http://dx.doi.org/10.1002/hyp.6413
24
rP
25 Zhang, G., Zeng, G. M., Jiang, Y. M., Huang, G. H., Li, J. B., Yao, J. M., Tan,
26 W., Xiang, R. J. and Zhang, X. L.: 2006, Modelling and measurement of two-
27 layer-canopy interception losses in a subtropical mixed forest of central-south
28 China, Hydrology and Earth System Sciences 10, 65–77.
29
ee
30 Zimmermann, A., Germer, S., Neill, C., Krusche, A. V. and Elsenbeer, H.:
31 2008, Spatio-temporal patterns of throughfall and solute deposition in an
32 open tropical rain forest, Journal of Hydrology 360(1-4), 87–102.
rR
33 URL: http://www.sciencedirect.com/science/article/B6V6C-4T2S8XJ-
34 1/2/8bdb8ca4081de629249dccead13d6811
35
36 Zinke, P. J.: 1967, Forest Hydrology, Pergamon Press, Oxford, chapter Forest
37 interception studies in the United States, pp. 137–161.
ev
38
39
40
41
iew
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57 30
58
59
60

Forest Floor Interception

Uploaded by

Document Information

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Forest Floor Interception

Uploaded by

Copyright:

Available Formats

Hydrological Processes

Spatial and temporal variability of canopy and forest floor

Journal: Hydrological Processes

Manuscript ID: HYP-09-0460

Date Submitted by the

Savenije, Hubert; Delft University of Technology, civil Eng and

canopy interception, forest floor interception, spatial variability,

temporal variability, semi-variograms, time stability plots

4 Temporal variation in interception measurements 6

33 5 Temporal variation in storage capacity 9

42 7 Spatial variation in throughfall and infiltration 19

25 are stronger driving factors on interception. Furthermore, the spatial

21 time (e.g., snow/no snow, heavy rain/drizzle, wind/windless). Hence, it is not

persistence of spatial throughfall and infiltration patterns.

38 Gross precipitation, temperature, humidity, radiation, wind speed and wind

precipitation. Cumulative rainfall in the collector is read at a weekly time-step

30 Stemflow is measured by an open flexible tube (width 3 cm) wrapped around

38 Since October 2003 throughfall is intensively measured in the Huewelerbach

37 23/02/04 07/07/05 19/11/06 02/04/08 15/08/09

are on the trees the capacity is on average 0.9 mm.

number of data points

storage capacity [mm]

number of data points

27 month of the year

42 high variability in winter values.

storage capacity [mm]

21 throughout the seasons.

25 ter model predictions

21 Parameter Value Parameter Value

25 Sct ,max [L] 0.11 mm F0 [L T−1 ] 0.04 mm/15 min

21 01-12-04 Sf 0.37 mm 11 0.53

25 17-08-05 Sf 0.64 mm -40 0.22

11-01-07 Sf 0.60 mm -21 0.66

38 07-01-08 Sf 0.93 mm -34 0.37

42 23-01-08 Sf 0.37 mm 14 0.42

25 interception evaporation is limited. On average the increase or decrease is about

day and the potential evaporation than by the storage capacity.

Evaporation forest floor

38 7 Spatial variation in throughfall and infiltra-

the circles the positions of the collectors.

21 In Figure 13 the semi-variograms of the throughfall and infiltration are shown

38 persistence can be derived: extreme persistence and general persistence (Keim

fitted curve fitted curve

Figure 15: Spatial persistence of normalized throughfall. Triangles indicate the

Winter (DJFM) Spring (AM)

21 flattening the leaves. The variation in summer is ±0.4 mm (cv = 19%).

in Luxembourg, Hydrology and Earth System Sciences 11, 695–701.

33 forest in Luxembourg, IAHS Redbook 326, 18–24.

mixed deciduous forest, Agricultural and Forest Meteorology 148(11), 1655–

38 New Zealand, Journal of Hydrology 66(1-4), 143–158.

21 of canopy rainfall interception for a broad-leaved tree, Hydrological Processes

You might also like