HYDROLOGICAL PROCESSES

Hydrol. Process. 24, 3011– 3025 (2010)

Published online 8 June 2010 in Wiley Online Library
(wileyonlinelibrary.com) DOI: 10.1002/hyp.7712

Spatial and temporal variability of canopy and forest floor

interception in a beech forest
A. M. J. Gerrits,1 * L. Pfister2 and H. H. G. Savenije1
1 Water Management, Delft University of Technology, Delft, The Netherlands
2 EVA, CRP-Gabriel Lippmann, Belvaux, Luxembourg

Abstract:
Depending on season, rainfall characteristics and tree species, interception amounts to 15–50% of total precipitation in a forest
under temperate climates. Many studies have investigated the importance of interception of different tree species in all kinds
of different climates. Often authors merely determine interception storage capacity of that specific species and the considered
event, and only sometimes a distinction is made between foliated and non-foliated trees. However, interception is highly
variable in time and space. First, since potential evaporation is higher in summer, but secondly because the storage capacity
has a seasonal pattern. Besides weather characteristics, such as wind and rain intensity, snow causes large variations in the
maximum storage capacity. In an experimental beech plot in Luxembourg, we found storage capacity of canopy interception
to show a clear seasonal pattern varying from 0Ð1 mm in winter to 1Ð2 mm in summer. The capacity of the forest floor
appears to be rather constant over time at 1Ð8 mm. Both have a standard deviation as high as š100%. However, the process
is not sensitive to this variability resulting only in 11% variation of evaporation estimates. Hence, the number of raindays and
the potential evaporation are stronger driving factors on interception. Furthermore, the spatial correlation of the throughfall
and infiltration has been investigated with semi-variograms and time stability plots. Within 6–7 m distance, throughfall and
infiltration are correlated and the general persistence is rather weak. Copyright  2010 John Wiley & Sons, Ltd.

KEY WORDS canopy interception; forest floor interception; spatial and temporal variability; beech; Rutter model storage capacity

Received 15 September 2009; Accepted 18 March 2010

INTRODUCTION are observed where the canopy and branches funnel the
In a forest, part of the precipitation is intercepted by rainfall causing locally higher intensity throughfall (e.g.
vegetation. Rain drops first hit the leaves and branches Germer et al., 2006; Gerrits et al., 2009). This results in
before they fall on the forest floor, where again a part concentrated infiltration, which again may be the trigger
is intercepted by litter on the forest floor. Progressively, for subsurface flow. These throughfall patterns appear to
the canopy and the litter dry out by evaporation. Both have an important influence on the soil moisture patterns
processes are here considered to be part of the same (Bouten et al., 1992).
interception process, I, which equals the sum of the In literature, many studies on interception can be found
change in interception storage (Si ) and the evaporation (see Kittredge, 1948; Zinke, 1967; Breuer et al., 2003 and
from this stock (Ei ): references herein). Most of these studies consider inter-
ception of events or short duration periods (in the order
dSi of months). However, interception is highly seasonal.
I D Ei C
1
dt First of all, the potential evaporation changes throughout
Interception is considered to be about 15–50% of the year and so does the storage capacity for decidu-
the total incoming precipitation on forests in temper- ous trees. At best a distinction is made between leaf-
ate humid latitudes (e.g. Rutter et al., 1975; Viville on and leaf-off periods (e.g. Rutter et al., 1975; Rowe,
et al., 1993; Hörmann et al., 1996; Savenije et al., 2004). 1983; Hörmann et al., 1996; Zhang et al., 2006; Feni-
Because the intercepted and evaporated water is not cia et al., 2008b; Herbst et al., 2008), but the transition
recharging the forest soils, the interception process is between these states is rarely described. The spatial vari-
important to the forest soil moisture balance of forests. ability of the canopy coverage also causes large vari-
The intercepted water does not contribute to the soil ations in the storage capacity and differs per season
reservoir and hence is not available to vegetation. In addi- (Staelens et al., 2006). Furthermore, the storage capacity
tion, the interception process plays an important role as depends on precipitation conditions that vary over time
a re-distributor of rainfall in space. In general, through- (e.g. snow/no snow, heavy rain/drizzle, low wind/strong
fall is less than gross precipitation, but often drip points wind). Hence, it is not possible to define one single
storage capacity for a certain tree species in contrast to
what many authors claim. This is especially important
* Correspondence to: A. M. J. Gerrits, Water Management, Delft Univer-
sity of Technology, Delft, The Netherlands. when these storage capacities are used for interception
E-mail: [email protected] modelling.

Copyright  2010 John Wiley & Sons, Ltd.

3012 A. M. J. GERRITS, L. PFISTER AND H. H. G. SAVENIJE

Figure 1. Overview of the adapted Rutter model

The aim of this paper is to investigate how the as:

interception process changes over the seasons and how dSf
Ei,f C D Tf
F
3
this affects evaporation predictions of an interception dt
model. This model is used to investigate the spatial
distribution and the persistence of spatial throughfall and
where Ei,f is evaporation from the forest floor and Sf
infiltration patterns.
the storage on the forest floor and F the infiltration.
For both interception types, we have investigated the
temporal variation of evaporation from interception.
METHODOLOGY The maximum storage capacity, Sc , is determined by
the ‘mean-method’ of Klaassen et al. (1998), where the
To investigate the effect of forest interception, both
canopy and forest floor interceptions have been measured. storage capacity is the negative intercept with the y-axis
Canopy interception is defined as precipitation (Pg ) of the linear regression line of accumulated gross rainfall
minus the sum of throughfall (Tf ) and stemflow (Ts ) versus accumulated throughfall of an event. In this way,
(Equation 2): the effect of evaporation during the event is taken into
account (when the slope of the regression line is less than
dSc 1 : 1). For the storage capacity of the forest floor, Sf ,
Ei,c C D Pg
Tf
Ts
2 the same procedure is used on accumulated throughfall
dt
versus accumulated infiltration.
where Ei,c is evaporation from the canopy and Sc the Only those events have been selected that meet the
storage on the canopy. Forest floor interception is defined following criteria:

Copyright  2010 John Wiley & Sons, Ltd. Hydrol. Process. 24, 3011– 3025 (2010)
 SPATIAL AND TEMPORAL VARIABILITY OF CANOPY AND FOREST FLOOR INTERCEPTION 3013

(a) (b)

Figure 2. (a) Location of the interception plot in the Huewelerbach catchment (Luxembourg) and (b) the location of the equipment inside the plot

exponential function with D0 taken from Rutter et al.

(1971). Drainage is also possible when Scl < Sc,max l
,
taking care of the shake off of rain drops by wind.
The partitioning of the gross rainfall into canopy
input, free throughfall and trunk input is based on
our measurements. The factor p is determined by light
interception measurements. A fish-eye lens is connected
to a light sensible resistance, and the ratio of light
intensity under (I) and above (I0 ) the canopy equals p.
The fraction pt is determined as in Gash and Morton
(1978) and Kittredge (1948). We found a sympt fraction
equal to 0Ð07. Because pt is not changing over time, we
consider pt to be constant. The negative intercept with
the y-axis determines the storage capacity of the trunk.
We assume the trunk storage capacity to be a constant
equal to 0Ð11 mm.
l
The maximum storage capacity of the canopy (Sc,max )
and the forest floor (Sf,max ) are derived from the mea-
surements as described above and are dependent on the
Figure 3. Schematic drawing of forest floor interception device with Ei,f season. For calibration of the adjusted Rutter model, we
evaporation from the forest floor, El evaporation from the lower basin and
Su
D Sf  and Sl the storage in the upper and lower basin respectively used the mean storage capacity per month.
The forest floor reservoir is modelled in the same
way as the canopy reservoir: a threshold (Sf ) and
ž The event is large enough to saturate the storage
an exponential infiltration rate, F; however infiltration
capacity, causing throughfall to occur from the canopy
equals zero if Sf < Sf,max , because wind will not shake
or infiltration from the forest floor.
off rain drops at forest floor level.
ž The time between the events is long enough for the
Potential evaporation (Ep ) is calculated with the Pen-
intercepted water to evaporate and that no water is
man equation (Allen et al., 1998). The potential evapora-
stored before the start of the next event.
tion for the forest floor is slightly lowered by increasing
The storage capacities are determined for all events the relative humidity with 2%.
under different weather conditions to study the season- Finally, the spatial variability has been investigated
ality (average variation over time) and variability (varia- through semi-variograms and time stability plots. We
tion within a certain season or period). Also the relation used the method as described by Keim et al. (2005) to
between storage capacity and rain intensity and wind calculate the semi-variogram:


speed is investigated.  x,y
N

N  x,yCh 2
To investigate the effect of the observed uncertainties n
h
on interception predictions, we applied a modified Rutter 
h D
4
2n
h
model (Rutter et al., 1971) which includes a forest floor
interception reservoir. In Figure 1, an overview of the where h is the lag, n
h the number of measurement pairs
model structure is given. Drainage is modelled as an  x,y the
in the data set that are distance h apart, and N

Copyright  2010 John Wiley & Sons, Ltd. Hydrol. Process. 24, 3011– 3025 (2010)
3014 A. M. J. GERRITS, L. PFISTER AND H. H. G. SAVENIJE

Huewelerbach
(a) 140
Precipitation
Throughfall
120 Stemflow

100
[mm/period]

80

60

40

20

0
01/07/03 01/01/04 01/07/04 01/01/05 01/07/05 01/01/06 01/07/06 01/01/07 01/07/07 01/01/08 01/07/08 01/01/09 01/07/09
Date [dd/mm/yy]

(b) 80
Canopy interception
Snow/ice
60 Mean Pg < 0.02 mm/h
Leafed canopy
40
[% of Pg]

20

−20

−40
01/07/03 01/01/04 01/07/04 01/01/05 01/07/05 01/01/06 01/07/06 01/01/07 01/07/07 01/01/08 01/07/08 01/01/09 01/07/09
Date [dd/mm/yy]

Figure 4. (a) Precipitation, collector readings of throughfall (average over 81 collectors) and stemflow and (b) canopy interception percentage (of
precipitation) over time in the Huewelerbach

(a) Huewelerbach
150
Throughfall
Infiltration

100
[mm/period]

50

0
01/01/05 01/07/05 01/01/06 01/07/06 01/01/07 01/07/07 01/01/08 01/07/08 01/01/09
Date [dd/mm/yy]

(b) 100
Forest floor interception
90 Snow/ice
Mean Tf < 0.16 mm/h
80
Leafed canopy
70
60
[% of Tf]

50
40
30
20
10
0
01/01/05 01/07/05 01/01/06 01/07/06 01/01/07 01/07/07 01/01/08 01/07/08 01/01/09
Date [dd/mm/yy]

Figure 5. (a) Throughfall and infiltration and (b) temporal variation of forest floor interception as percentage of throughfall (four throughfall collectors
nearest to forest floor interception device)
Copyright  2010 John Wiley & Sons, Ltd. Hydrol. Process. 24, 3011– 3025 (2010)
 SPATIAL AND TEMPORAL VARIABILITY OF CANOPY AND FOREST FLOOR INTERCEPTION 3015

normalized throughfall or infiltration at measuring point area of 1Ð07 m2 ). The gutters are connected to one tipping
(x,y): bucket (RM Young, 52203-L, 0Ð1 mm), so as to provide
N x,y D Nx,y
N
5 the temporal structure of throughfall below the canopy.

N Because the representativity of these gutters is difficult
to assess, because of the high spatial variability of the
with  the standard deviation. We used normalized data canopy density, an additional network of rain collectors
to be able to compare throughfall and infiltration with (SPIEA 1650-01; A D 400 cm2 ) has been installed inside
each other. the experimental stand. Eighty-one collectors (41 before
May 2004) have been put in a network at a spacing
of 3 m. This network of rain collectors gives both an
SITE DESCRIPTION AND MATERIALS indication of the spatial variability of throughfall, as well
The study area is located in the Huewelerbach basin near as a precise measurement of the amount of throughfall
the village of Hovelange in Luxembourg (Figure 2a). at a biweekly to monthly time-step. The biweekly to
The experimental beech stand (Fagus sylvatica L.) of monthly throughfall amounts are disaggregated via the
120 years of age is located in the northwest of the measurements obtained by the tipping bucket gauges that
basin (49° 420 5700 N, 5° 530 1300 E) and has a total area of are connected to the gutters below the canopy.
0Ð0596 ha with an average slope of 5%. The density of Stemflow is measured by an open flexible tube (width
the stand is of 168 trees/ha, whereas the tree height varies 3 cm) wrapped around the trunk of two trees from
from 30 to 40 m (Figure 2b). The average tree diameter January 2004 until July 2009. The water collected from
at breast height is 70 cm. The tree in the northeast corner the trees is directed to a tipping bucket (RM Young,
of the stand has been cut in December 2004 causing a 52203-L, 0Ð1 mm & Isco 674, 0Ð1 mm). Stemflow is
gap in the canopy coverage. The leaf area index varies upscaled by calculating the average stemflow per tree
between 2 and 7, with leave burst in April and leave fall and multiplying by the mean number of trees per square
in October. The climate is modified oceanic with mild metre in the stand (Gerrits et al., 2009).
winters and temperate summers. The average rainfall is Forest floor interception is measured with a specially
845 mm/year and the average temperature is 8 ° C (Pfister developed device (Figure 3) from November 2004 until
et al., 2005). March 2009. It consists of two aluminium basins of
Gross precipitation, temperature, humidity, radiation, 1 m2 mounted above each other. The upper basin is
wind speed and wind direction are measured every 5 min filled with beech leaves and has a permeable bottom, so
at the meteorological station (Campbell Scientific) in the water can drain into the second basin. The lower basin
alluvial valley. To verify the correct functioning of the is watertight and is emptied every day by an electronic
tipping bucket, an additional rain collector collects gross valve. Both basins are weighed and logged every 5 min.
precipitation and is read at a weekly time-step. Rainfall The evaporation from the forest floor is calculated from
was observed from January 2004 until July 2009. the water balance (Equation 3), where F D dSl /dt. A
Throughfall is measured both continuously and at a detailed description of the device can be found in Gerrits
biweekly to monthly time-step from October 2003 until et al. (2007). Additionally, the water flushed from the
July 2009. In the experimental plot, a network of three valve is collected in a big barrel for verification.
gutters (16Ð5 ð 215Ð5 cm) has been built (total collecting Because the weighing sensors are made of a piece
of metal with strain gauges, temperature changes have
an effect on the sensor output due to the expansion
2 and compression of the metal. From 2004 to 2006,
the device was compensated with the relation between
1.5
temperature and weight change observed in a dry period
1 by linear regression (see Gerrits et al., 2007, for further
details). From 2006 onwards, an additional sensor (so-
storage capacity [mm]

number of data points

0.5 called dummy sensor) is installed to compensate for

the temperature influence. The dummy sensor is a free
0 0
hanging sensor with no weight on it and thus only
2
reacts on temperature changes. If the relation between
the dummy sensor and the other sensors is determined in
4 the laboratory, we can compensate the observed data.

6
TEMPORAL VARIATION IN INTERCEPTION
8
2 4 6 8 10 12 MEASUREMENTS
month of the year
Canopy interception
Figure 6. Variability of the canopy storage. The black line is the mean of
all events in a certain month, and the grey area defines a bound of one The observed gross precipitation (starting from January
standard deviation. The number of events is indicated by the black bars 2004), average throughfall and stemflow are depicted

Copyright  2010 John Wiley & Sons, Ltd. Hydrol. Process. 24, 3011– 3025 (2010)
3016 A. M. J. GERRITS, L. PFISTER AND H. H. G. SAVENIJE

(a) (b)
2 summer (JJAS) 2 summer (JJAS)
fall (ON) fall (ON)
1.8 winter (DJFM) 1.8 winter (DJFM)
spring (AM) spring (AM)
1.6 2
Linear R = 0.20 1.6 Linear R2= 0.41
storage capacity [mm]

storage capacity [mm]

1.4 1.4

1.2 1.2

1 1

0.8 0.8

0.6 0.6

0.4 0.4

0.2 0.2

0 0
0 0.5 1 1.5 2 2.5 3 3.5 0 0.5 1 1.5 2
average wind speed [m/s] rainfall intensity [mm/h]

Figure 7. Effect of average 15-min wind speed per event (a) and average 15-min rainfall intensity per event (b) on canopy storage capacity during
different seasons

4
summer (JJAS)
3.5 fall (ON)
3 winter (DJFM)
spring (AM)
3
2 Linear R2= 0.10
storage capacity [mm]

number of data points

storage capacity [mm]

1 2.5

0 0 2

5 1.5

10
1
15
0.5
20
2 4 6 8 10 12
0
month of the year 0 0.5 1 1.5 2 2.5 3 3.5 4 4.5
throughfall intensity [mm/h]
Figure 8. Variability of the storage capacity of the forest floor. The black
line is the mean of all events in a given month, and the grey area defines Figure 9. Effect of average throughfall intensity on forest floor storage
the standard deviation. The number of events is indicated by the black capacity throughout the seasons
bars

in Figure 4. We observe a clear seasonal pattern in the data logger, damage by branches that fell on the device,
evaporation from canopy interception as a percentage of battery failures, and so on. Even a completely new device
gross rainfall (Figure 4b). When there are leaves on the was built that causes in total a data gap of more than one
trees from April until September (indicated by the green year in 2006.
bar), evaporation from interception is on average 18% of Unlike with canopy interception, there is no clear
the precipitation compared with 5% in winter. seasonal trend. Although the amount of data is limited,
Furthermore, some positive and negative outliers we can see that evaporation from the forest floor is rather
can be seen in the percentage of canopy interception constant over the year, with a slight increase in summer.
(Figure 4b). These events often coincide with small rain- On average, evaporation is 22% of the throughfall when
fall amounts (triangles), causing relatively large errors in extreme values, related to snow events, are not included.
the readings, or coincide with snow or ice events (aster- Snow is causing high water equivalents to be first
isks). Snow and ice can prevent precipitation to be caught stored in the upper basin, which will drain towards
by the tipping bucket or the collector, or the recording to the lower basin only when the temperature is again
be delayed. above 0 ° C. Similar to canopy interception, it appears that
small rainfall events (and thus throughfall events) cause
Forest floor interception relatively high forest floor interception values.
In Figure 5, the results obtained from the forest floor Combining results from canopy evaporation and forest
interception device are shown. The large data gaps are floor evaporation, we can conclude that in winter 26% of
caused by several child deceases of the newly developed the precipitation is intercepted and in summer as much as
device. Lots of data was lost due to problems with the 36%. Hence, the seasonal effect of interception, if forest

Copyright  2010 John Wiley & Sons, Ltd. Hydrol. Process. 24, 3011– 3025 (2010)
 SPATIAL AND TEMPORAL VARIABILITY OF CANOPY AND FOREST FLOOR INTERCEPTION 3017

Table I. Model parameters of adjusted Rutter model Canopy interception

Prameter Value Parameter Value For the analysis of storage capacity, 52 recorded events
meet the criteria. To ensure that the canopy is dry before
p (
) Leaf area εt (
) 0Ð5 an event, at least 2 days of no rain had to separate two
index (LAI) events. As can be seen in Figure 6, there is a clear
pt (
) 0Ð07 D0 (L/T) 0Ð03 mm/15 min
l seasonal variability in the canopy storage capacity. In
Sc,max (L) Figure 6 d (L
1 ) 0Ð9 mm
1
t
Sc,max (L) 0Ð11 mm F0 (L/T) 0Ð04 mm/15 min winter, the capacity is on average low 0Ð4 mm and in
Sf,max (L) Figure 8 f (L
1 ) 2Ð0 mm
1 summer when the leaves are on the trees the capacity is
on average 0Ð9 mm.
The grey area indicates a band of one standard devi-
Table II. Model results of the adjusted Rutter model for through- ation in the storage capacity of the different events.
fall (Tf ), infiltration (F), forest floor evaporation (Ei,f ) and forest In winter, the deviation is š0Ð2 mm (cv D 46%) com-
floor wetness (Sf ). Given are the root mean square error (RMSE),
pared with š0Ð5 mm (cv D 54%) in summer. Hence,
relative bias, and Nash–Sutcliffe efficiency
it is difficult to determine one single capacity value,
Run RMSE Relative bias NS especially when the canopy is developed. One of the
(%) (
) reasons for the variability may be the applied regres-
sion of Klaassen et al. (1998) method to determine
SIM1 Tf 3Ð55 mm/day 4 0Ð59
the storage capacity. As Rowe (1983) concluded this
02-11-04 F 2Ð61 mm/day
5 0Ð62
to Ei,f 0Ð74 mm/day 5 0Ð24 approach is strongly influenced by wind speed and
01-12-04 Sf 0Ð37 mm 11 0Ð53 intensity. For example, Horton (1919), Klaassen et al.
SIM2 Tf 9Ð34 mm/day 9 0Ð30 (1996) and Hormann et al. (1996) found that with
21-06-05 F 11Ð07mm/day 30 0Ð02 increasing wind speed the measured storage capacity
to Ei,f 2Ð06 mm/day 0
1Ð02 is less, because the wind shakes the rain water off
17-08-05 Sf 0Ð64 mm
40 0Ð22 the leaves. This effect is also visible in Figure 7a,
SIM3 Tf 5Ð80 mm/day 8 0Ð71 although the relationship is weak. We also tested the
14-12-06 F 4Ð26 mm/day 1 0Ð66
relationship between maximum wind speed during the
to Ei,f 1Ð14 mm/day 40 0Ð16
11-01-07 Sf 0Ð60 mm
21 0Ð66 event and storage capacity, but the relationship is even
SIM4 Tf 4Ð81 mm/day 3 0Ð83 worse.
07-02-07 F 4Ð21 mm/day 0 0Ð72 There is no consensus in the literature on the effect of
to Ei,f 1Ð41 mm/day
9 0Ð10 rainfall intensity on the storage capacity. On the one hand,
27-02-07 Sf 1Ð10 mm
4 0Ð37 Horton (1919) and Wang et al. (2007) concluded that
SIM5 Tf 3Ð52 mm/day 7 0Ð81 the higher the rainfall intensity the lower the capacity,
04-12-07 F 7Ð14 mm/day 4
0Ð18 because high rainfall intensities cause splashing and
to Ei,f 0Ð98 mm/day 98 0Ð14
shaking of leaves. On the other hand, Aston (1979)
07-01-08 Sf 0Ð93 mm
34 0Ð37
and Keim et al. (2006) found the opposite: high rainfall
SIM6 Tf 5Ð16 mm/day 11 0Ð70
07-01-08 F 4Ð43 mm/day
3 0Ð65 intensities coincide with high storage capacities, because
to Ei,f 0Ð81 mm/day 45
0Ð44 of dynamic storage. On the basis of our data, we do not
23-01-08 Sf 0Ð37 mm 14 0Ð42 see a clear relationship, only a weak decline in storage
capacity with increasing average intensity (Figure 7b).
Also no relationship has been found with maximum
floor interception is included, is less strong when only rainfall or storm size.
based on canopy interception. The annual interception
amounts to 26–36% of gross rainfall and amounts to Forest floor interception
38–42% of the actual evaporation in the Huewelerbach
catchment, on the basis of water balance data from 2004 For the forest floor, 74 recorded events were available
until 2006 (Atlas, 2004, 2005, 2006). This is a consid- (we also used data from the gaps presented in Figure 5
erable part of the total evaporation, taking into account in case the lower basin was working). The weight of
the modest dimensions of the interception reservoir com- upper basins was used to determine if the forest floor
pared with the soil moisture storage capacity. The reason was dry before an event. In contrast to the canopy, there
for this relatively large interception flux lies in the high is no apparent seasonal trend (Figure 8). The storage
number of rainfall events as compared with infiltration capacity is rather constant over the year (1Ð8 mm), with
events. a temporal increase during early fall (2Ð8 mm). This can
be explained by the fresh litter. The fresh leaves dry out
quickly after they fall off the tree and obtain a curled
shape, which has a high storage capacity. However, the
TEMPORAL VARIATION IN STORAGE CAPACITY
curled shape can quickly disappear after rain or snow
To determine the storage capacity of the canopy and storms.
forest floor, only those events have been selected that Furthermore, it can be seen that the variability is
meet the criteria in the Methodology Section. changing from winter to summer. In contrast to the

Copyright  2010 John Wiley & Sons, Ltd. Hydrol. Process. 24, 3011– 3025 (2010)
3018 A. M. J. GERRITS, L. PFISTER AND H. H. G. SAVENIJE

Throughfall
60
Modelled–avg
Modelled–lower
40 Modelled–upper
Observed

[mm]
20

0
30/10/04 09/11/04 19/11/04 29/11/04
Infiltration
40
Modelled–avg
30 Modelled–lower
Modelled–upper
Observed
[mm]

20

10

0
30/10/04 09/11/04 19/11/04 29/11/04
Evaporation forest floor
20
Modelled–avg
15 Modelled–lower
Modelled–upper
Observed
[mm]

10

0
30/10/04 09/11/04 19/11/04 29/11/04

Wetness forest floor

8
Modelled–avg
6 Modelled–lower
Modelled–upper
Observed
[mm]

0
30/10/04 09/11/04 19/11/04 29/11/04
time (x 15 min)
Figure 10. Model results for SIM1. The green line represents the results with the lower limit storage capacity, the blue the mean and the red line the
upper limit storage capacity

canopy storage capacity, the variability in summer is increasing intensity the dynamic and also the static stor-
lower š0Ð4 mm (cv D 19%) than in winter š0Ð8 mm age capacity increased. If we test this hypothesis on our
(cv D 50%). This can be explained by the effect of data, we find a weak positive relationship (Figure 9).
snow, which has been repeatedly observed during field The effect of wind is not investigated because it is
observations. If a snow event occurs, the leaves are not likely that wind can shake off water at forest floor
completely flattened due to the snow weight, causing level.
a small storage capacity. If no snow occurs, the leaves
retain their original shape, with a large storage capacity.
EFFECT OF VARIABILITY IN STORAGE
Hence, the capacity really depends on whether snow
CAPACITY ON RUTTER MODEL PREDICTIONS
events occurred or not. This causes a high variability in
winter values. The consequences of the variability in storage capacity
Additional variability may be caused by through- on prediction with a Rutter model (Rutter et al., 1971)
fall intensity. Putuhena and Cordery (1996), Sato et al. have been analysed to asses the prediction error because
(2004) and Guevara-Escobar et al. (2007) found that with of uncertainty in the storage capacity.

Copyright  2010 John Wiley & Sons, Ltd. Hydrol. Process. 24, 3011– 3025 (2010)
 SPATIAL AND TEMPORAL VARIABILITY OF CANOPY AND FOREST FLOOR INTERCEPTION 3019

100 winter time. In summer time, because of the variabil-

Canopy
90 Forest floor
ity in the storage capacity, the performance is lower.
Total However, it should also be mentioned that the results
80
are not independent. If throughfall is wrongly modelled,
70 this causes a wrong input to the forest floor reser-
Modelled [% of P]

60 voir.
Furthermore, the effect of the variability in the storage
50
capacity is visible. The Rutter model is calibrated on
40 the mean storage capacity per month. However, it can
30 really vary between years when, for example, the leaf
growth or fall starts. This timing has a large impact on
20
the actual storage capacity and can thus deviate from
10 the mean storage capacity for that specific month. This
0 effect can be the reason for the lower performance of the
0 20 40 60 80 100
Observed [% of P]
simulations in fall and spring.
To investigate the effect of the variability of the stor-
Figure 11. Graphical representation of Table III, showing the perfor- age capacity on the modelled evaporation, we applied the
mance of the Rutter model for canopy interception, forest floor inter-
ception and total interception model with a low, a mean and a high storage capacity.
l
For Sc,max and Sf,max , we used the upper and lower limits
of Figures 8 and 8. In Figure 10, the results for SIM1 are
shown and in Table III the mean evaporation rates for all
In Table I, the model parameters are given. The six simulations. In Figure 11, the results are graphically
model is evaluated on throughfall, infiltration, forest floor shown.
evaporation and forest floor wetness. As can be seen in Figure 10, the observed data fit
In Table II, the model results of the adjusted Rutter between the upper and lower estimates. However, some-
model are presented for six simulation periods. Given times the model prediction deviates from the observed
are the root mean square error (RMSE), the relative data. This is mainly the case during the wetting, thus
bias and the Nash–Sutcliffe efficiency (Nash and Sut- on the rising limb, and not on the falling limb dur-
cliffe, 1970). Overall the performance of the model is ing drainage and evaporation. During the falling limb
reasonable, although the goodness-of-fit measures are the model performs well, showing that the Rutter model
sometimes poor. RMSE is not very good because of is capable of predicting evaporation. The rising limb
fluctuations that may be caused by the temperature gives more problems, which can be explained by vari-
sensitivity of the forest floor interception device and ability of the storage capacity in the instrument. This
phase lags in loggers. However, the bias is gener- may, for example, be due to unequal litter distribution
ally small. Throughfall is best modelled, especially in in the instrument caused by wind or snow. Figures 4

Table III. Model results of average interception evaporation compared with average rainfall over the simulation periods. In bold the
observed percentages
l t l t
Run P Smax Ei,c Ei,c Ei,c D Ei,c C Ei,c obs. Ei,f obs. Ei obs.
(mm/day) (mm/day) (mm/day) (mm/day) (mm/day) (mm/day)

SIM1 1Ð69 (100%) lower 0Ð03 (2%) 0Ð04 (3%) 0Ð08 (5%) 0Ð42 (25%) 0Ð50 (30%)
mean 0Ð02 (1%) 0Ð04 (3%) 0Ð07 (4%) [6%] 0Ð56 (33%) [31%] 0Ð63 (37%) [37%]
(02-11-04 to 01-12-04) upper 0Ð02 (1%) 0Ð04 (3%) 0Ð07 (4%) 0Ð65 (39%) 0Ð72 (43%)
SIM2 2Ð22 (100%) lower 0Ð19 (9%) 0Ð07 (3%) 0Ð26 (12%) 0Ð70 (32%) 0Ð90 (43%)
mean 0Ð23 (10%) 0Ð07 (3%) 0Ð29 (13%) [20%] 0Ð81 (36%) [42%] 1Ð10 (50%) [63%]
(21-06-05 to 17-08-05) upper 0Ð30 (13%) 0Ð07 (3%) 0Ð36 (16%) 0Ð85 (38%) 1Ð2 (54%)
SIM3 3Ð11 (100%) lower 0Ð06 (2%) 0Ð05 (2%) 0Ð11 (3%) 0Ð45 (14%) 0Ð55 (18%)
mean 0Ð06 (2%) 0Ð05 (2%) 0Ð11 (4%) [7%] 0Ð51 (16%) [13%] 0Ð62 (20%) [19%]
(14-12-06 to 11-01-07) upper 0Ð07 (2%) 0Ð05 (2%) 0Ð11 (4%) 0Ð55 (18%) 0Ð66 (21%)
SIM4 4Ð78 (100%) lower 0Ð19 (4%) 0Ð08 (2%) 0Ð27 (6%) 0Ð53 (11%) 0Ð80 (17%)
mean 0Ð19 (4%) 0Ð08 (2%) 0Ð27 (6%) [6%] 0Ð67 (14%) [14%] 0Ð94 (20%) [20%]
(07-02-07 to 27-02-07) upper 0Ð19 (4%) 0Ð08 (2%) 0Ð27 (6%) 0Ð74 (15%) 1Ð00 (21%)
SIM5 2Ð29 (100%) lower 0Ð04 (2%) 0Ð04 (2%) 0Ð07 (3%) 0Ð42 (18%) 0Ð49 (21%)
mean 0Ð04 (2%) 0Ð04 (2%) 0Ð08 (3%) [4%] 0Ð46 (20%) [13%] 0Ð54 (23%) [17%]
(04-12-07 to 07-01-08) upper 0Ð05 (2%) 0Ð04 (2%) 0Ð08 (4%) 0Ð48 (21%) 0Ð56 (25%)
SIM6 3Ð28 (100%) lower 0Ð07 (2%) 0Ð06 (2%) 0Ð13 (4%) 0Ð58 (18%) 0Ð72 (22%)
mean 0Ð08 (2%) 0Ð06 (2%) 0Ð13 (4%) [10%] 0Ð71 (22%) [13%] 0Ð85 (26%) [24%]
(07-01-08 to 23-01-08) upper 0Ð08 (2%) 0Ð06 (2%) 0Ð13 (4%) 0Ð79 (24%) 0Ð93 (28%)

Copyright  2010 John Wiley & Sons, Ltd. Hydrol. Process. 24, 3011– 3025 (2010)
3020 A. M. J. GERRITS, L. PFISTER AND H. H. G. SAVENIJE

24-Mar-2006 to 06-Apr-2006
P=69.6mm
(a) Throughfall [mm] (b) In ltration [mm]
(observed) (modelled)
30 30
trees trees
160 140
collectors collectors
25 25
N 140 N
120
20 20
Distance [m]

Distance [m]
120 100
15 15
100 80
10 10
80 60
5 5
60 40
0 0
0 5 10 15 20 25 0 5 10 15 20 25
Distance [m] Distance [m]

(c) Canopy interception evaporation [mm] (d) Forest oor interception evaporation [mm]
(observed) (modelled)
30 30
trees trees
18.6
collectors collectors
25 20 25
N N 18.4

18.2
20 15 20
Distance [m]

18

15 15 17.8
Distance [m]

10
17.6
10 10
17.4
5
5 5 17.2

17
0
0 0
0 5 10 15 20 25 0 5 10 15 20 25
Distance [m] Distance [m]
Figure 12. Spatial variability of canopy interception (observed) and forest floor interception (modelled with adjusted Rutter model) in millimetre over
the period 24 March to 6 April 2006. Triangles indicate the position of the beech trees and the circles the positions of the collectors

and 5 illustrate the temporal variability of the storage forest floor interception is more driven by the storage
capacity. capacity than by the number of rainday and the potential
As can be seen from Table III, the effect of the storage evaporation. This supports the findings of Baird and
capacity on canopy interception evaporation is limited. Wilby (1999) and Gerrits et al. (2007).
On average, the increase or decrease is about 5% with an The impact of uncertainties in the storage capacity
average coefficient of variation in the storage capacity of (which can be as high as š100%) on the total interception
56%. Hence, a large variation in the storage capacity has evaporation is about 11% and the difference in the lower
a low impact on the evaporation predictions, and thus or upper storage capacity is 15% and 8%, respectively.
canopy interception is more driven by the number of This indicates that interception is more influenced by the
rainday and the potential evaporation than by the storage rainfall pattern than by the storage capacity. Hence, in
capacity. interception modelling, the value of the storage capacity
This is the opposite for the forest floor. Here, the is of minor concern.
average coefficient of variation of the forest floor storage Furthermore, Table III shows the relative importance
capacity is 48%, and the average increase or decrease of canopy, trunk and forest floor interception evaporation.
is 12%. Hence, the influence of the storage capacity The model results show that in fall (SIM1) 4% of
on evaporation predictions is higher, indicating that precipitation evaporates from the canopy (6% observed)

Copyright  2010 John Wiley & Sons, Ltd. Hydrol. Process. 24, 3011– 3025 (2010)
 SPATIAL AND TEMPORAL VARIABILITY OF CANOPY AND FOREST FLOOR INTERCEPTION 3021

(a) Summer (JJAS) (b) Summer (JJAS)

2 2
Range = 6.73 Range = 6.55
fitted curve fitted curve
Sill = 1.00 Sill = 1.00
2 2

γ (h)
R = 0.55
γ (h)

R = 0.53
1 1

0 0
0 5 10 15 20 0 5 10 15 20
h [m] h [m]
Fall (ON) Fall (ON)
2 2
Range = 5.87 Range = 5.54
fitted curve fitted curve
Sill = 1.00 Sill = 1.00
2
R2 = 0.52
γ (h)

γ (h)
R = 0.46
1 1

0 0
0 5 10 15 20 0 5 10 15 20
h [m] h [m]
Winter (DJFM) Winter (DJFM)
2 2
Range = 7.39 Range = 6.82
fitted curve fitted curve
Sill = 1.00 Sill = 1.00
2 2
γ (h)

R = 0.54

γ (h)
R = 0.53
1 1

0 0
0 5 10 15 20 0 5 10 15 20
h [m] h [m]
Spring (AM) Spring (AM)
3 3
Range = 5.85 Range = 5.73
fitted curve fitted curve
Sill = 1.00 Sill = 1.00
2 2 2 2
γ (h)

γ (h)

R = 0.48 R = 0.44
1 1

0 0
0 5 10 15 20 0 5 10 15 20
h [m] h [m]
Annual Annual
3 3
Range = 6.35 Range = 6.19
data data
Sill = 1.00 Sill = 1.00
2 2 fitted curve 2 fitted curve
γ (h)

R = 0.53 2
γ (h)

R = 0.50
1 1

0 0
5 10 15 20 0 2 4 6 8 10 12 14 16 18
0
h [m] h [m]

Figure 13. Semi-variogram of throughfall (a) and (modelled) infiltration (b) per season and per year

and 33% from the forest floor (31% observed). In winter is also mentioned by Germer et al. (2006) and Gerrits
(SIM 3–6), this is 4% (observed 7%) from the canopy et al. (2009)
and 18% (observed 13%) from the forest floor. There are In Figure 12b, the resulting canopy interception pattern
no simulations in spring, and in summer (SIM2) 13% is shown. Evaporation from canopy interception is here
(observed 20%) from the canopy and 36% (observed calculated as gross rainfall minus (observed) throughfall
42%) from the forest floor. and hence stemflow is neglected. Because throughfall
can exceed gross precipitation, this results in negative
interception evaporation values. These values have been
SPATIAL VARIATION IN THROUGHFALL removed from the analysis.
AND INFILTRATION Because we do not have spatial observations of forest
Besides the temporal variation, interception varies also in floor interception, we used the adjusted Rutter model. To
space, mainly due to the heterogeneity of the vegetation reduce the modelling error as much as possible, we used
density. In Figure 12a, an example of the spatial dis- observed throughfall data as input for the forest floor
tribution of interpolated (triangle-based cubic) observed module instead of modelled throughfall. In Figure 12c
throughfall is given. It encompasses a period in spring and 12d, the model results of the infiltration and forest
2006 with 70 mm of gross rainfall. It can be clearly seen floor evaporation are shown. In order to see if the
that the forest canopy redistributes the rainfall. In gen- spatial pattern of canopy interception is different from
eral, throughfall is lower around trees, due to interception. that of the forest floor and to investigate if the pattern
However, the trees can also funnel the rainfall, as can be changes throughout the seasons, we calculated the spatial
seen near the tree at coordinates (15 m, 15 m). The struc- correlation with semi-variograms.
ture of this tree really acts like a funnel, causing even In Figure 13, the semi-variograms of the throughfall
higher throughfall than gross rainfall close to the tree, and infiltration are shown per season as well as per year.
and lower throughfall values around the tree. This effect The range r is defined as the lag h, whereby the variance

Copyright  2010 John Wiley & Sons, Ltd. Hydrol. Process. 24, 3011– 3025 (2010)
3022 A. M. J. GERRITS, L. PFISTER AND H. H. G. SAVENIJE

(a) Summer (JJAS) (b) Summer (JJAS)

6 6
4 4
2 2
0 0
−2 −2
0 10 20 30 40 50 60 70 80 0 10 20 30 40 50 60 70 80
Fall (ON) Fall (ON)
6 6
4 4
2 2
0 0
−2 −2
0 10 20 30 40 50 60 70 80 0 10 20 30 40 50 60 70 80
Winter (DJFM) Winter (DJFM)
6 6
4 4
Nx,y

Nx,y
2
~

~
2
0 0
−2 −2
0 10 20 30 40 50 60 70 80 0 10 20 30 40 50 60 70 80
Spring (AM) Spring (AM)
6 6
4 4
2 2
0 0
−2 −2
0 10 20 30 40 50 60 70 80 0 10 20 30 40 50 60 70 80
Annual Annual
6 6
4 4
2 2
0 0
−2 −2
0 10 20 30 40 50 60 70 80 0 10 20 30 40 50 60 70 80
collector collector

Figure 14. Time stability plot of normalized throughfall (a) and (modelled) infiltration (b) per season and per year. The grey areas indicates one time
the standard deviation

() is 95% of the sill c and is a measure for the correlation and extreme persistence only occurs in the ‘wet’ tail.
between the points. High spatial correlation between the The general persistence is rather weak. Only 25%, 7%,
collectors causes the range to be high and vice versa. We 14%, 7% and 7% of the collectors are significantly (65%)
fitted an exponential model (Chilés and Delfiner, 1999): drier than the mean for summer, fall, winter, spring and
annual, respectively. This suggests that the variation in

h D c[1
exp

3h/r]
6 winter and summer is smaller than in the transition sea-
sons.
As expected (Western et al., 1998) the range in winter
In Figures 15 and 16, the mean time stability plots (red
is higher (higher spatial correlation) after leaf senescence;
line in Figure 14) are plotted in space to investigate where
however, this not a clear relation. Furthermore, it seems
the drier and wetter spots are located and how they vary
that in summer and winter the range is relatively higher
than during the transition seasons. This is the case both throughout the seasons. As can be seen, the collector near
for throughfall and for infiltration. the tree at coordinates (15 m, 15 m) is consistently wetter
To investigate if these patterns are persistent in time, than the surrounding collectors. Furthermore, it can be
time stability plots (TSP) can show if there exist per- seen that a second drip point develops at coordinates
sistent dry or wet collectors by ranking the normalized (20 m, 0 m) in fall which vanishes again in spring.
 x,y . From a time stability plot,
throughfall/infiltration N Hence, this drip point is really determined by the branch
two types of persistence can be derived: extreme persis- structure and less by the canopy structure. It is also
tence and general persistence (Keim et al., 2005; Zim- possible to see by the dark blue rings around the drip
mermann et al., 2008). Extreme persistence occurs if points that the funnelled rain is collected from the
steep tails exists, and general persistence refers to the surrounding areas. Furthermore, the effect of the canopy
overall slope of the middle range. In Figure 14, the time development over time is visible. In summer, the spatial
stability plots of throughfall (a) and (modelled) infiltra- pattern is much more heterogeneous. Also the forest gap
tion (b) are presented. As can be seen, the difference in the upper left corner receives in summer relatively
between the TSP of throughfall and infiltration is small more throughfall than the covered parts.

Copyright  2010 John Wiley & Sons, Ltd. Hydrol. Process. 24, 3011– 3025 (2010)
 SPATIAL AND TEMPORAL VARIABILITY OF CANOPY AND FOREST FLOOR INTERCEPTION 3023

Summer (JJAS) Fall (ON)

30 30
trees trees
collectors 5 collectors 5
25 25
N N
4 4
20 20
Distance [m]

Distance [m]
3 3

15 15
2 2

10 1 10 1

5 0 5 0

−1 −1
0 0
0 5 10 15 20 25 0 5 10 15 20 25
Distance [m] Distance [m]

Winter (DJFM) Spring (AM)

30 30
trees trees
collectors 5 collectors 5
25 25
N N
4 4
20 20
Distance [m]

Distance [m]

3 3

15 15
2 2

10 1 10 1

5 0 5 0

−1 −1
0 0
0 5 10 15 20 25 0 5 10 15 20 25
Distance [m] Distance [m]

Figure 15. Spatial persistence of normalized throughfall. Triangles indicate the position of the beech trees and the circles the positions of the collectors

CONCLUSIONS capacity of the forest floor is in winter (š0Ð8 mm, cv D

50%) mainly due to the occurrence, or no occurrence, of
Evaporation from the canopy has a clear seasonal trend. snow which decreases the storage by flattening the leaves.
In the leaf on period evaporation from the canopy is on The variation in summer is š0Ð4 mm (cv D 19%). The
average 18% of the precipitation and in the leaf off period variability can also be related to wind or rainfall intensity.
5%. There is less seasonal variation in the forest floor However, none of these factors revealed itself as a strong
evaporation: 22% of the throughfall. Both in the canopy effect.
and on the forest floor, there are outliers which coincide To investigate the sensitivity to different conditions,
with low rainfall amounts (causing high ratios) or snow we applied a Rutter interception model with an additional
events (causing low or even negative ratios). Combined, forest floor reservoir. The model shows that a variation
canopy and forest floor interception amount to 26% of in the canopy storage capacity of 56% results in only 5%
the rainfall in winter and 36% in summer. Annually, difference in evaporation predictions from the canopy,
this amounts to 38–42% of the total evaporation in the whereas a variation of 48% in the storage capacity of
Huewelerbach catchment, which is a lot for a flux which the forest floor results in 12% difference in evaporation
is often disregarded or lumped with other evaporation from the forest floor. Hence, we can conclude that canopy
fluxes. interception is not so much determined by the storage
A similar seasonal effect can be found in the storage capacity, but rather by the number of raindays and the
capacities: for the canopy the average storage capacity potential evaporation, and that for the forest floor the
ranges from 0Ð4 mm in winter to 0Ð9 mm in summer. The storage capacity is more important.
storage capacity of the forest floor is less variable over The Rutter model was also applied to investigate the
the year: around 1Ð8 mm on average with a maximum of spatial distribution of throughfall and infiltration. The
2Ð8 mm during the fall. spatial patterns show that also the canopy coverage can
Besides this seasonal effect, there is variability at a cause variation in the calculated storage capacity. For
given point in time. The highest variability of the storage example, it makes a difference if the storage capacity

Copyright  2010 John Wiley & Sons, Ltd. Hydrol. Process. 24, 3011– 3025 (2010)
3024 A. M. J. GERRITS, L. PFISTER AND H. H. G. SAVENIJE

Summer (JJAS) Fall (ON)

30 30
trees trees
collectors 5 collectors 5
25 25
N N
4 4
20 20
Distance [m]

Distance [m]
3 3

15 15
2 2

10 1 10 1

5 0 5 0

−1 −1
0 0
0 5 10 15 20 25 0 5 10 15 20 25
Distance [m] Distance [m]

Winter (DJFM) Spring (AM)

30 30
trees trees
collectors 5 collectors 5
25 25
N N
4 4
20 20
Distance [m]

Distance [m]

3 3

15 15
2 2

10 1 10 1

5 0 5 0

−1 −1
0 0
0 5 10 15 20 25 0 5 10 15 20 25
Distance [m] Distance [m]

Figure 16. Spatial persistence of normalized (modelled) infiltration. Triangles indicate the position of the beech trees and the circles the positions of
the collectors

is determined in a relatively open area or in a dense ACKNOWLEDGEMENTS

location in the forest. Furthermore, trees can create hot We are very grateful to Jean-François Iffly and Cyrille
spots, where throughfall is higher than gross precipita- Tailliez for regularly reading out the equipment and for
tion. maintaining the plot. Furthermore, the authors would like
For the experimental beech plot in Luxembourg, we to thank the Ministry of Culture, Higher Education and
found that the spatial correlation of throughfall is about Research of Luxembourg (FNR) and Delft Cluster, the
6–7 m and that in summer and winter the spatial Netherlands, for their support of this research.
correlation is slightly higher than in fall and spring.
Similar results are found for the infiltration pattern. Also
the stability of the patterns was analysed. REFERENCES
Overall, we can conclude that interception is a highly Allen RG, Pereira LS, Raes D, Smith M. 1998. Crop Evapotranspiration
variable process, both in time and in space. Although - Guidelines for Computing Crop Water Requirements, vol. 56,
the variation in storage capacity can be as high as FAO(Food and Agriculture Organization of the United Nations): Rome,
Italy.
š100%, the effect on evaporation estimates is relatively Aston A. 1979. Rainfall interception by eight small trees. Journal of
low (ca 11%). This indicates that interception is more Hydrology 42(3-4): 383– 396.
Atlas2004. 2004. Atlas Hydro-climatologique du Grand-Duché de
influenced by the rainfall pattern than by the storage Luxembourg 2004, CRP-Gabriel Lippmann.
capacity. As a consequence, in interception modelling, Atlas2005. 2005. Atlas Hydro-climatologique du Grand-Duché de
the value of the storage capacity appears to be of minor Luxembourg 2005, Luxembourg.
Atlas2006. 2006. Atlas Hydro-climatologique du Grand-Duché de
concern. The research shows that the combined effect Luxembourg 2006, Technical report.
of canopy and forest floor interception is a considerable Baird AJ, Wilby RL. 1999. Eco-hydrology - Plants and Water in
flux in the rainfall runoff process (about 40% of the total Terrestrial and Aquatic Environments. Routledge: London, UK.
Bouten W, Heimovaara TJ, Tiktak A. 1992. Spatial patterns of
evaporation), which cannot be denied in rainfall–runoff throughfall and soil water dynamics in a Douglas fir stand. Water
modelling. Resources Research 28: 3227– 3233.

Copyright  2010 John Wiley & Sons, Ltd. Hydrol. Process. 24, 3011– 3025 (2010)
 SPATIAL AND TEMPORAL VARIABILITY OF CANOPY AND FOREST FLOOR INTERCEPTION 3025

Breuer L, Eckhardt K, Frede H-G. 2003. Plant parameter values for Nash JE, Sutcliffe JV. 1970. River flow forecasting through conceptual
models in temperate climates. Ecological Modelling 169(2-3): models. Part I: a discussion of principles. Journal of Hydrology 10:
237– 293. 282– 290.
Chilès JP, Delfiner P. 1999. Geostatistics. Modeling Spatial Uncertainty. Pfister L, Wagner C, Vansuypeene E, Drogue G, Hoffmann L. 2005.
John Wiley & Sons: USA. Atlas Climatique du Grand-Duché de Luxembourg. Johnen-Druck
Fenicia F, Savenije HHG, Avdeeva Y. 2008. Anomaly in the rainfall- GmbH & Co: Bernkastel-Kues.
runoff behaviour of the Meuse catchment. Climate, land use, or land Putuhena W, Cordery I. 1996. Estimation of interception capacity of the
use management? Hydrology and Earth System Sciences Discussion 5: forest floor. Journal of Hydrology 180: 283–299.
1787– 1819. Rowe L. 1983. Rainfall interception by an evergreen beech forest, Nelson,
Gash JHC, Morton AJ. 1978. An application of the rutter model to the New Zealand. Journal of Hydrology 66(1-4): 143– 158.
estimation of the interception loss from thetford forest. Journal of Rutter AJ, Kershaw KA, Robins PC, Morton AJ. 1971. A predictive
Hydrology 38(1-2): 49–58. model of rainfall interception in forests. Part I. Derivation of the model
Germer S, Elsenbeer H, Moraes JM. 2006. Throughfall and temporal and comparison with observations in a plantation of Corsican pine.
trends of rainfall redistribution in an open tropical rainforest, south- Agricultural Meteorology 9: 367– 384.
western Amazonia (Rondonia, Brazil). Hydrology and Earth System Rutter AJ, Morton AJ, Robins PC. 1975. A predictive model of rainfall
Sciences 10: 383–393. interception in forests. Part II. Generalization of the model and
Gerrits AMJ, Savenije HHG, Hoffmann L, Pfister L. 2007. New comparison with observations in some coniferous and hardwood stands.
technique to measure forest floor interception—an application in a Journal of Applied Ecology 12: 367–380.
beech forest in Luxembourg. Hydrology and Earth System Sciences Sato Y, Kumagai T, Kume A, Otsuki K, Ogawa S. 2004. Experimental
11: 695– 701. analysis of moisture dynamics of litter layers—the effect of rainfall
Gerrits AMJ, Savenije HHG, Pfister L. 2009. Canopy and forest floor conditions and leaf shapes. Hydrological Processes 18: 3007– 3018.
interception and transpiration measurements in a mountainous beech Savenije HHG. 2004. The importance of interception and why we should
forest in Luxembourg. IAHS Redbook , IAHS: IAHS Publication No. delete the term evapotranspiration from our vocabulary. Hydrological
326: Eynsham, UK; 18–24. Processes 18: 1507– 1511.
Guevara-Escobar A, Gonzalez-Sosa E, Ramos-Salinas M, Hernandez- Staelens J, De Schrijver A, Verheyen K, Verhoest NEC. 2006. Spatial
Delgado GD. 2007. Experimental analysis of drainage and water variability and temporal stability of throughfall water under a dominant
storage of litter layers. Hydrology and Earth System Sciences 11(5): beech (Fagus sylvatica L.) tree in relationship to canopy cover. Journal
1703– 1716. of Hydrology 330: 651–662.
Herbst M, Rosier PT, McNeil DD, Harding RJ, Gowing DJ. 2008. Viville D, Biron P, Granier A, Dambrine E, Probst A. 1993. Interception
Seasonal variability of interception evaporation from the canopy of a in a mountainous declining spruce stand in the Strengbach catchment
mixed deciduous forest. Agricultural and Forest Meteorology 148(11): (Vosges, France). Journal of Hydrology 144: 273– 282.
1655– 1667. Wang A, Diao Y, Pei T, Jin C, Zhu J. 2007. A semi-theoretical model
Hörmann G, Branding A, Clemen T, Herbst M, Hinrichs A, Thamm F. of canopy rainfall interception for a broad-leaved tree. Hydrological
1996. Calculation and simulation of wind controlled canopy Processes 21(18): 2458– 2463.
interception of a beech forest in Northern Germany. Agricultural and Western AW, Blöschl G, Grayson RB. 1998. How well do indicator
Forest Meteorology 79(3): 131– 148. variograms capture the spatial connectivity of soil moisture?
Horton RE. 1919. Rainfall interception. Monthly Weather Review 47(9): Hydrological Processes 12(12): 1851– 1868.
603– 623. Zhang G, Zeng GM, Jiang YM, Huang GH, Li JB, Yao JM, Tan W,
Keim RF, Skaugset AE, Weiler M. 2005. Temporal persistence of spatial Xiang RJ, Zhang XL. 2006. Modelling and measurement of two-layer-
patterns in throughfall. Journal of Hydrology 314: 263–274. canopy interception losses in a subtropical mixed forest of central-south
Keim RF, Skaugset AE, Weiler M. 2006. Storage of water on vergetation China. Hydrology and Earth System Sciences 10: 65–77.
under simulated rainfall of varying intensity. Advances in Water Zimmermann A, Germer S, Neill C, Krusche AV, Elsenbeer H. 2008.
Resources 29: 974–986. Spatio-temporal patterns of throughfall and solute deposition in an
Kittredge J. 1948. Forest Influences. McGraw-Hill: New York. open tropical rain forest. Journal of Hydrology 360(1-4): 87–102.
Klaassen W, Bosveld F, deWater E. 1998. Water storage and evaporation Zinke PJ. 1967. Forest interception studies in the United States. Forest
as constituents of rainfall interception. Journal of Hydrology 212-213: Hydrology. Pergamon Press: Oxford; 137–161.
36– 50.
Klaassen W, Lankreijer HJM, Veen AWL. 1996. Rainfall interception
near a forest edge. Journal of Hydrology 185(1-4): 349– 361.

Copyright  2010 John Wiley & Sons, Ltd. Hydrol. Process. 24, 3011– 3025 (2010)