WATER RESOURCES RESEARCH, VOL. 48, W01544, doi:10.

1029/2011WR010705, 2012

Forest harvesting effects on the magnitude and frequency of peak

flows can increase with return period
Piotr K. Kuras,1,2 Younes Alila,1 and Markus Weiler3
Received 22 April 2011; revised 28 November 2011; accepted 3 December 2011; published 28 January 2012.
[1] Paired watershed studies have limited researchers wishing to disentangle road and
harvesting effects on peak ows or to study management schemes other than the existing
scenario. The outcomes of many paired watershed studies examining peak ows have also
recently been challenged since only an approach that pairs peak ows by frequency can
adequately evaluate the effects of harvesting on peak ows. This study takes advantage of a
model that has been developed and extensively tested at a site containing a rich set of
internal catchment process observations to examine the isolated and combined effects of
roads and harvesting on the peak ow regime of a snow-dominated catchment for return
periods of up to 100 years. Contrary to the prevailing perception in forest hydrology, the
effects of harvesting are found to increase with return period, which is attributable to the
uniqueness of peak ow runoff generation processes in snow-dominated catchments.
Planned harvesting (50% harvest area) is found to have a signicant effect (9%25% over
control) on peak ows with recurrence intervals ranging 10100 years. Peak ow frequency
increases after harvesting increase with return period, with the largest events (100 year)
becoming 56.7 times more frequent, and medium-sized events (10 year) becoming 1.72
times more frequent. Such changes may have substantial ecological, hydrological, and
geomorphological consequences within the watershed and farther downstream. Study
ndings suggest that peak ow regimes are fairly tolerant to the current level of harvesting
in this particular watershed but that further harvesting may affect this element signicantly.
Citation: Kuras, P. K., Y. Alila, and M. Weiler (2012), Forest harvesting effects on the magnitude and frequency of peak flows can
increase with return period, Water Resour. Res., 48, W01544, doi:10.1029/2011WR010705.

1. Introduction improved our understanding of how forest harvesting affects

[2] Following road construction and forest harvesting,
the hydrograph in a snow-dominated watershed can be
expected to change through four mechanisms altering run-
off generation processes: an increase in snow accumulation
and melt, a decrease in evapotranspiration (including can-
opy interception), a decrease in channel roughness from
large woody debris removal, and extension of the channel
network by roads [Jones and Grant, 1996]. An increase in
snow accumulation and melt combined with a decrease in
evapotranspiration resulting from canopy removal can affect
the hillslope water balance by increasing peak ow and
stormow [Jones and Grant, 1996; Troendle and Reuss,
1997]; while a decrease in channel roughness with an exten-
sion of the channel network by roads affects ow routing and
can speed stormow, potentially advancing and/or increasing
the peak ow without a change in volume [Jones, 2000].
Over the years, many process-based stand-level studies have
1
Department of Forest Resources Management, University of British
Columbia, Vancouver, British Columbia, Canada.
2
Northwest Hydraulic Consultants Ltd., North Vancouver, British
Columbia, Canada.
3
Institute of Hydrology, University of Freiburg, Freiburg, Germany.
Copyright 2012 by the American Geophysical Union
0043-1397/12/2011WR010705
hydrological and snow processes [Schmidt and Troendle,
1989; Troendle and Reuss, 1997; Buttle et al., 2001; Spittle-
house and Winkler, 2002; Murray and Buttle, 2003; Thyer
et al., 2004; Winkler et al., 2005]. This knowledge and how
it translates to the catchment-scale mechanisms of stream-
ow generation, and associated changes in peak ow
regimes, has traditionally been investigated by the paired
watershed approach [Rothacher, 1970; Ziemer, 1981; Harr
et al., 1982; Duncan, 1986; Wright et al., 1990; Jones and
Grant, 1996; Burton, 1997; Jones, 2000], and our prevalent
scientic perception of the relation between forest harvesting
and the peak ow regime is shaped by decades of paired
watershed studies. Alila et al. [2009] recently challenged the
outcomes of such studies since these are based on research
hypotheses that focus on a change in magnitude between pre-
harvesting and postharvesting peak ows that are paired by
storm input (for pluvial regimes) or freshet season (for nival
regimes). The authors illustrate that this type of chronologi-
cal event pairing leads to a misleading change in the magni-
tude of peak ows since it does not account for changes in
frequency, and consequently strips a fundamental part of the
physics from the science evaluating these effects. Alila et al.
[2010] assert that only an approach that pairs peak ows
by frequency, well established in other disciplines, can evalu-
ate the effects of forest harvesting on the inextricably linked
magnitude and frequency of peak ows, and call for a

W01544 1 of 19
W01544 KURAS ET AL.: FOREST HARVESTING IMPACTS ON PEAK FLOWS
reevaluation of past studies and the century-old, precon-
ceived scientic perception of the relation between forest
harvesting and peak ows.
[3] Paired watershed studies have also limited research-
ers that wish to disentangle the combined effects of roads
and harvesting or study management schemes other than
the existing scenario [Bowling et al., 2000]. The character-
istically short period of record in paired watershed studies
have also been inadequate in statistically testing changes in
peak ows that have return periods greater than 10 years
[Jones and Grant, 1996], and the typically shorter pretreat-
ment versus posttreatment period makes such analyses
unreliable [Thomas and Megahan, 1998]. Though their
debate on forest road and harvesting effects on peak ows
was unresolved, both Thomas and Megahan and Jones and
Grant concurred that process-based studies should be
coupled with the development and validation of physically
based, distributed hydrologic models in order to forecast
the effects of forest cutting and roading activities on a given
watershed [Thomas and Megahan, 1998, p. 3403] and
will improve our understanding of forest harvest effects
on these rare big oods and also should address the geomor-
phic and ecological consequences of changes in all sizes of
peak discharge events [Jones and Grant, 2001, p. 177].
[4] A brief overview of prior modeling studies that
examined the relation between forest management and
peak ows follows, with a focus on studies that evaluated
peak ow events paired by frequency and not chronology.
Schnorbus and Alila [2004a] combined hydrologic numeri-
cal modeling (Distributed Hydrology Soil Vegetation
Model (DHSVM) [Wigmosta et al., 1994]) with syntheti-
cally generated long-term meteorological inputs to assess
the immediate impacts of forest harvesting (vegetation con-
ditions were held stationary) on the peak ow regime of
Redsh Creek. The authors did not consider operationally
feasible harvesting scenarios, however, as harvest areas
were restricted along elevation bands. Schnorbus and Alila
[2004a] also emphasized the necessity for future research
to investigate the impact of roads (not included in simula-
tions), especially since prior modeling exercises for the
watershed attributed poor performance to the exclusion
of road networks in simulations [Whitaker et al., 2003].
Hydrological processes in Redsh Creek differ substantially
from 241 Creek, as the former is a steep-gradient watershed
with rain-on-snow events frequently dominating the annual
peak ow hydrograph, while the latter is snowmelt domi-
nated with a atter, subdued topography that typically pro-
duces a nearly uniformly melting snowpack. La Marche
and Lettenmaier [2001] isolated the effects of forest roads
on peak ows at the Deschutes River with DHSVM, and
found that model predicted increases in peak ows as a
result of roads increased with return period, while vegeta-
tion effects decreased; however, because of data limitations,
the authors were only able to predict the effects of roads
and harvesting up to the 10 year event. Bowling and Letten-
maier [2001] also conducted a similar analysis with retro-
spective simulations at Hard and Ware Creeks and found
that mean annual peak ows were predicted to have
increased because of roads, the effects of which were addi-
tive and roughly equal in magnitude to harvesting.
[5] To date, no study has collectively examined simu-
lated peak ow regime changes in snow-dominated water-
2 of 19
W01544
sheds for a wide range of return intervals (1.003100
years), with the application of various harvesting levels
with and without roads, including a focus on the effects of
forest roads alone. Such an assessment is only possible
with a model that has been developed and calibrated at a
site containing a rich set of internal catchment process
observations for evaluation of model performance and in-
ternal structure, such as the DHSVM application developed
by Thyer et al. [2004] for 241 Creek. This watershed is the
treatment at the Upper Penticton Creek (UPC) Watershed
Experiment, which is situated in the Okanagan region
of south central British Columbia (BC) [Winkler et al.,
2003]. The subdued topography of this small scale, snow-
dominated watershed is unique, and snow melts nearly uni-
formly during the melt season with no clear topography or
elevation related snowline. The model has also proven to
be successful in realistically simulating spatiotemporally
variable runoff generation processes in further performance
evaluations [Kuras et al., 2011], and is believed to be a reli-
able tool for contributing to the ongoing debate on the
effects of forest roads and different levels of harvesting on
peak ow regimes in snow-dominated watersheds.
[6] The purpose of this research is to apply a physically
based, spatially distributed hydrologic model (DHSVM) in
combination with long-term synthetically generated mete-
orological data to examine the simulated effects of various
rates of harvesting on streamows in 241 Creek, with a par-
ticular focus on road effects. Simulations are based on sta-
tionary vegetation cover, which reects the immediate
impact of treatment on peak ow regimes prior to any sub-
stantial forest regrowth (and hence hydrologic recovery)
[Ward, 1971; Schnorbus and Alila, 2004a]. This paper is
organized as follows: a description of the study area is pro-
vided in section 2; methods are presented in section 3,
which includes a general overview of DHSVM with a sum-
mary of the UPC application, as well as a summary of the
treatment scenarios, synthetic meteorological data genera-
tion, and data analysis techniques. Research results are pre-
sented in section 4, and section 5 discusses these ndings,
with concluding remarks provided in section 6. Table 1
provides a summary and denitions of acronyms and sym-
bols used in text.
2. Study Area
[7] This study focuses on 241 Creek, part of the ongoing
long-term interdisciplinary UPC Watershed Experiment
(49 400 N, 119 240 W) [Winkler et al., 2003], located
approximately 26 km northeast of Penticton, BC (Figure 1),
which was designed for the purpose of understanding the
relation between forest management and hydrology. The
experiment includes a network of all-season weather sta-
tions, snow survey sites, continuous streamow (Water
Survey of Canada) and groundwater (perched water table)
monitoring sites [Kuras et al., 2007] (Figure 1). Field data
collected to date include observations of forest and open
radiation balance, snow accumulation and melt [Winkler
et al., 2005], rainfall interception [Spittlehouse, 1998], soil
water content [Spittlehouse, 2000], tree transpiration [Spit-
tlehouse, 2002], and spatially intensive measurements of
road and stream network ows [Kuras et al., 2007], and
with both control and treatment watersheds and prelogging
W01544 KURAS ET AL.: FOREST HARVESTING IMPACTS ON PEAK FLOWS W01544

Table 1. Denitions of Acronyms and Symbols Used in Texta 750 mm, approximately half of which falls as snow, and
snow cover generally lasts from late October until early
Acronym or Symbol Definition
June [Winkler et al., 2004]. The late winter snowpack is
DHSVM Distributed Hydrology Soil Vegetation Model normally 1 to 1.5 m deep, with 1 April snow water equiva-
UPC Upper Penticton Creek lents (SWE) averaging 265 mm. The annual hydrograph is
BC British Columbia dominated by snowmelt, and the freshet peak ow typically
SWE snow water equivalent
AMS annual maximum series occurs in late spring to early summer, with annual water
QP annual maximum peak ow yields ranging approximately 170630 mm (as a depth over
T return period the entire basin). Soil textures in the study area are predom-
QT,h return period dependent quantiles for hourly inantly coarse sandy loams and loamy sands ranging in
QP
QT,d return period dependent quantiles for daily QP
depth from 0.14 m, and are derived from glacial tills and
QT,7 return period dependent quantiles for 7 day QP coarse-grained granitic rock [Hope, 2001]. Soils are low in
DEM digital elevation model clay and high in coarse fragment content, with forest oors
LAI leaf area index generally less than 4 cm thick. Soils are generally well
GIS geographic information system drained and have a low water holding capacity, with late
Op20 preexisting (20% clear-cut) operational
harvesting scenario in 241 Creek summer eld observations verifying hydrophobicity in the
Op30 current (30% clear-cut) operational harvesting upper soil layers.
scenario in 241 Creek
Op50 future (50% clear-cut) operational harvesting
scenario in 241 Creek 3. Methods
Op20R, Op30R, addition of roads to each scenario, represented 3.1. Overview
and Op50R by the sufx R
Control scenario representing the pristine condition of [8] The sensitivity of streamow metrics to preexisting,
the watershed current and future harvest scenarios, with and without
CR Control with roads scenario roads, was assessed using long-term numerical simulation
GEV generalized extreme value distribution
QT peak ow quantiles with DHSVM. Simulations were driven with long-term
DQT percent change in treatment QT relative to synthetically generated meteorological data, producing
Control QT (frequency pairing) streamow time series of 100 years duration at an hourly
QHT treatment QT resolution for a total of eight operational harvesting scenar-
QCT Control QT
Q100/Q2 ratio of 100 to 2 year T peak ows
ios. Only forest cover parameters were adjusted among sce-
SY simulation year narios to reect the various areas of harvesting, while all
DQP percent change in treatment QP relative to other basin parameters were held constant. The sample an-
Control QP (chronological pairing) nual maximum series (AMS) extracted from each time se-
ANOVA analysis of variance ries was used to estimate the frequency distribution of the
ANCOVA analysis of covariance
annual maximum peak ow (QP) for each particular sce-
a
In order of appearance. nario. The procedure employed for numerical simulations
and the assessment of outputs is presented as a owchart in
Figure 2. Each scenario was driven with identical meteoro-
and postlogging measurements, it is the only experiment of logical inputs with forest parameters held constant (i.e., no
its kind in the interior of BC [Winkler et al., 2004]. 241 vegetative recovery), thus generating a stationary ood fre-
Creek is a headwater tributary to Penticton Creek and quency response reective of the rst few critical years
drains an area of 4.74 km2, which ranges in elevation from posttreatment (i.e., prior to hydrologic recovery). The
16002025 m. The lower 1.5 km2 of the watershed is rela- effects of treatment on the peak ow regime can only truly
tively at (<7% slope), and 75% of the watershed area has be interpreted by separating ow regimes into class meas-
slopes less than 30%; however, the remaining upper area is urements (streamow metrics) that describe peak ows in a
signicantly steeper and accounts for 45% of catchment range from the near instantaneous (hourly) to sustained
relief. 241 Creek is the treatment catchment in the paired (daily and 7 day). Only reporting a single temporal scale of
watershed study, and the most recent harvesting entry streamow may not represent the full range of hydrologic,
(FebruaryMarch 2003) has brought the total clear-cut area geomorphic and ecological consequences of annual peak
to 30% (Figure 1). The watershed is situated within the dry ows [Schnorbus and Alila, 2004a].
Engelmann Spruce Subalpine Fir biogeoclimatic subzone. [9] The hourly peak ow metric was used to describe the
The predominant forest cover type in the watershed is effects of treatment on near instantaneous ows, which for
mature lodgepole pine (Pinus contorta Dougl.), with small example, are important in engineering design. The daily
amounts of Engelmann spruce (Picea engelmannii Parry) and 7 day (extracted from a 7 day rolling average) peak
and subalpine r (Abies lasiocarpa (Hook.) Nutt) (Figure 1; ow metrics were used to describe the effects of treatment
details are in Table 2). The trees in the study site are over on more sustained peak ows, and the combination of all
100 years old, reaching a maximum height of 2026 m, three metrics provides a means of characterizing the fre-
with canopy densities ranging from 35% to 50%. The quency, magnitude, and duration of geomorphologically
understory is composed of lichens, mosses, and small and ecologically signicant peak ows. Similar to the work
shrubs (<0.5 m in height). Mean summer (June to August) by Schnorbus and Alila [2004a], streamow metrics repre-
and winter (November to March) air temperatures are 11 C senting the hourly, daily, and 7 day QP were used to derive
and 5 C, respectively, with an average temperature of return period (T)-dependent quantiles for hourly (QT,h),
2 C over the entire year. The mean annual precipitation is daily (QT,d), and 7 day (QT,7) QP.

3 of 19
W01544 KURAS ET AL.: FOREST HARVESTING IMPACTS ON PEAK FLOWS W01544

Figure 1. Location of the Upper Penticton Creek study site in south central British Columbia (49 400 N,
119 240 W), with detailed current maps of 241 Creek; vegetation class details are provided in Table 2.

3.2. DHSVM UPC Application
[10] A brief description of DHSVM is provided below,
with a summary of the steps taken in the development of
the UPC application by Thyer et al. [2004], and further
model performance evaluation by Kuras et al. [2011]. For a
more detailed description of model structure and equations,
the reader is referred to Wigmosta et al. [2002].
[11] DHSVM is a physically based, spatially distributed
hydrologic model, which explicitly solves the water and
energy balance for each cell of a watersheds digital eleva-
tion model (DEM). Grid cells are able to exchange water
with potentially eight of their adjacent neighbors, allowing
for a three-dimensional representation of surface and sub-
surface water interactions across the landscape. The spatial
distribution of canopy interception, evapotranspiration, soil
moisture, water table depth, snow distribution and melt,
and runoff can be simulated at hourly or longer time
steps. The model uses a two layer canopy representation
for interception and evapotranspiration, a two layer energy
balance model for snow accumulation and melt, a multi-
layer unsaturated soil model, and a saturated subsurface
ow model. Meteorological data inputs include: air tem-
perature, precipitation, relative humidity, wind speed, and
incoming shortwave and longwave radiation. Digital eleva-
tion data are used to model topographic controls on
solar radiation, precipitation, air temperature and down-
slope water movement. Precipitation and air temperature
for each model pixel were determined from point (climate
station) measurements with an elevation gradient approach.
The land surface may be composed of overstory and/or

Table 2. Structural Properties of Vegetation Classes for DHSVM Vegetation Input Filesa
Class Overstory Description Dominant Height (m) Canopy Closure (0.01.0) LAI (m2 m2) Understory Present

1 rock N/A N/A N/A no

4 lodgepole pine 19 0.5 4.0 no
5 lodgepole pine 25 0.5 4.0 yes
8 Engelmann spruce 23 0.2 2.4 no
9 Engelmann spruce 28 0.4 3.8 yes
10 clear-cut N/A N/A N/A no
a
Source: Thyer et al. [2004].

4 of 19
W01544 KURAS ET AL.: FOREST HARVESTING IMPACTS ON PEAK FLOWS
Figure 2. Flowchart illustrating the procedure employed for numerical simulations and assessment of
outputs.
understory vegetation, and soil. The overstory may cover
all or a prescribed fraction of the land surface, while the
understory, if present, covers the entire land surface. The
complexity of DHSVM provides the opportunity to analyze
more than just changes in streamow as a result of treat-
ment, and offers a closer look into internal catchment
processes.
[12] Thyer et al. [2004] evaluated the performance and
internal structure of DHSVM using 19982001 data from
the UPC Watershed Experiment. Parameter equinality is a
common issue in numerical modeling if the parameters of a
physically based hydrologic model are adjusted with
streamow as the only constraint. This was not the case for
the model developed by Thyer et al. [2004], however, and
because of the richness of observations at UPC the authors
were able to diagnose the internal performance of the
model and calibrate it to the watershed accordingly. Hourly
measurements of air temperature, precipitation, relative hu-
midity, shortwave radiation, wind speed, and snow and soil
temperatures since August 1997 and September 1999 were
used from the P1 and PB climate stations, respectively
(Figure 1). Hourly streamows were measured for both the
treatment and control (not shown) watershed outlets since
November 1983. Supplementary data included measure-
ments of: SWE in 2 clear-cuts and adjacent forest stands,
snow albedo, snowmelt in a clear-cut, rainfall interception,
tree transpiration (over a 6 day summer period), and below
canopy shortwave and longwave radiation.
5 of 19
W01544
[13] The model was calibrated to the streamows of 240
Creek (the adjacent control watershed) rather than to 241
Creek because of the completeness of the former catch-
ments streamow record. The stepwise calibration of the
model with a sensitivity analysis of parameters followed the
approach taken by Whitaker et al. [2003]. The calibrated
vegetation and soil parameters for 240 Creek were derived
from measurements for water year 1999, and SWE and
streamow data for the remaining 3 years was used to eval-
uate model performance. The authors found that the model
successfully simulated streamows, forest and clear-cut
SWE, tree transpiration, clear-cut snowmelt rates, and was
able to capture the differences between canopy rainfall
interception for small and large storms, and the differences
in soil moisture between wet and dry summers. The stream
ows of 241 Creek were used to assess the transferability of
the model parameters between the two basins, where 241
Creek has a higher stream network density and slightly
more southerly aspect. Results were found to be good and
differences between the two watersheds such as the earlier
and more pronounced peak ows of 241 were successfully
simulated for the 4 years, which included both the preharv-
est and postharvest periods.
[14] Three forest radiation schemes were developed by
Thyer et al. [2004], and this study utilized Scheme 2, which
was determined to be most suitable for 241 Creek through
model performance evaluations. The shortwave radiation
balance for this scheme was calculated using an alternate
W01544 KURAS ET AL.: FOREST HARVESTING IMPACTS ON PEAK FLOWS
approach to DHSVMs original formulation by Wigmosta
et al. [1994], and was based on fractional forest cover
rather than leaf area index (LAI). Fractional forest cover
was used instead of LAI because higher-resolution forest
cover maps were available for the two watersheds, and LAI
was only an estimator. Longwave radiation exchange at the
snow surface was also an alternate to original formulation,
and utilizes the canopy view factor, which is the proportion
of the canopy as would be observed from the snow pack
surface. The authors found that the alternative forest radia-
tion balance was consistent with measurements of short-
wave transmittance and the hemispherical canopy view
factor components of the forest cover, and that this
improved model efciency in simulating streamows from
0.84 to 0.91.
[15] In addition to further validating the model with an
additional 4 years of observed hydrographs since the time
of Thyer et al. [2004], Kuras et al. [2011] also evaluated
the internal performance of the model using spatially inten-
sive measurements of the premelt season snowpack, road
and stream network ows, and groundwater (perched water
table) uctuations [Kuras et al., 2007]. Overall, the model
was quite successful in reproducing the outlet hydrograph,
and in simulating the spatial distribution of the premelt sea-
son snowpack. Even though the UPC DHSVM application
had not previously been evaluated or calibrated for internal
catchment surface and subsurface ow generation, Kuras
et al. [2011] also found that the model realistically simu-
lated the spatiotemporal variability of road and stream net-
work ows, and subsurface responses in the watershed. The
extensive and overall successful testing of the UPC
DHSVM application in simulating both surface and subsur-
face internal catchment processes has constrained the
potential for model parameter equinality, and this applica-
tion is considered to be a reliable tool for evaluating the
effects of various forest management scenarios on the peak
ow regime of 241 Creek.
3.3. Road and Harvest Scenarios
[16] The road network in 241 Creek is divided into two
sections: the east and west roads (Figure 1). The total length
of roads in the watershed is 4.7 km, with 0.9 km of road
draining directly into streams (estimated in a geographic in-
formation system (GIS)). The proportion of road segments
directly draining into streams increases the drainage density
of the watershed by at least 7%; however, if it is assumed
that the entire road network contributes to drainage density,
this gure potentially increases to 37% (Table 3). The road
network surface and right of way account for 4% of the area
in the watershed. The stream network for the watershed
reects eld observations (Figure 1).
[17] Three scenarios were developed representing preex-
isting (20% clear-cut), current (30% clear-cut) and future
(50% clear-cut) operational harvesting scenarios in 241
Creek, referred to as Op20, Op30 and Op50, respectively,
from this point forward (Figure 3). Simulations were run
with and without roads, and the addition of roads to each
scenario is represented by the sufx R (i.e., Op20R,
Op30R, and Op50R). The Op30 scenario reects the cur-
rent vegetation coverage in 241 Creek (Figures 1 and 3d),
while the Control scenario represents the pristine condition
of the watershed (Figure 3a) prior to any anthropogenic
6 of 19
W01544
Table 3. Summary of Road Implications in 241 Creek Derived
From GIS Analysis
Percent
Descriptor 241 Creek Change
Original stream channel length LS (km) 12.6
Drainage density LS /AD (km km2) 2.7
Length of roads LR (km) 4.7
Length of roads connected to streams LRS (km) 0.9
Length of roads draining to the hillslope (km) 3.8
Extended drainage density due to stream connected 2.8 7%
roads (LRS LS)/AD (km km2)
Potential extended drainage density due to entire 3.6 37%
road network (LR LS)/AD (km km2)
Area cut for road construction (ha) 19.0 4%a
a
As a percent of total watershed area.
activity. The Control with roads scenario (CR) was created
to isolate the hydrological effects of roads in the watershed
and reects the harvest area that was necessary for the road
right of way (Figure 3b).
3.4. Data Analysis
[18] The generated peak ow regime for each scenario
and streamow metric was quantied with ood frequency
analysis of the simulated AMS. The sample AMS extracted
from each scenario (100 simulated QP events from 100
simulation years) was used to estimate the frequency distri-
bution of the annual maximum peak ow for each stream-
ow metric. The generalized extreme value (GEV)
distribution t was used to estimate peak ow quantiles,
and parameters for the distribution were estimated using
the method of L moments as described in detail by
Stedinger et al. [1993]. L moments are a way to describe
the statistical properties of hydrologic data, and are linear
combinations of ranked observations. The GEV distribution
is a mathematical form that incorporates Gumbels type I,
II, and III extreme value distributions for maxima, and uses
three parameters in terms of L moments (location, scale,
and shape) that are based on the sample distribution.
Assuming that the GEV parameters estimated from each
sample AMS are true parameters, peak ow quantiles
(QT) for each streamow metric were estimated using
Monte Carlo simulation with 10,000 runs, for a range of T:
1.003, 2, 10, 20, 50, and 100 years (corresponding to non-
exceedance probabilities of 0.003, 0.5, 0.9, 0.95, 0.98, and
0.99, respectively). The quantiles of the GEV distribution
are based on the cumulative probability of each T, which
were found using the empirical probability calculation
(probability 1  [1/T]). The 10,000 estimates of the Con-
trol quantile estimator were used to approximate variance :
the estimates were assumed to be asymptotically normally
distributed [Stedinger et al., 1993], and the resulting var-
iance tends to increase with a decrease in sample size and
with increasing T [Schnorbus and Alila, 2004a]. The GEV
t for each scenario was used to compare treatment and
Control QT for a range of T (1.003100 years), with Control
condence limits constructed using a level of signicance
 0.05. DQT describes the percent change in treatment
C
QT (QH T ) relative to the Control QT (QT ), by comparing
events of equal frequency (i.e., equivalent T).
W01544 KURAS ET AL.: FOREST HARVESTING IMPACTS ON PEAK FLOWS W01544

Figure 3. Spatial distribution of vegetation classes for the following scenarios: (a) Control, (b) Control
with roads (CR), (c) Op20, (d) Op30, and (e) Op50. Vegetation class details are provided in Table 2.

3.5. Synthetic Meteorological Data
[19] Long-term synthetic meteorological data used to
drive model simulations was generated using methods
described by Schnorbus and Alila [2004b]. In brief, short-
duration meteorological records were extended by using a
combined stochastic-empirical technique to generate syn-
thetic meteorological data in three steps (for more details
the reader is referred to Schnorbus and Alila [2004b]) :
hourly precipitation was generated using a rectangular
pulses point process; daily meteorological data were gener-
ated using a multivariate, rst-order, autoregressive pro-
cess; and nal hourly nonprecipitation meteorological data
were derived by disaggregating daily meteorology. Long-
term meteorological data were generated using measure-
ments of air temperature, precipitation, relative humidity,
wind speed, and incoming global radiation from the P1 cli-
mate station (Figure 1) for the period 19972005.
[20] A comparison was made between the short-term
(19972005) and long-term (19712005) climate of the
region to verify that the generated long-term meteorological
data were an accurate reproduction of the current climate
period. This was done to prevent bias in the synthetic long-
term meteorological data by verifying that the observed
variability in the short-term data set from which these were
generated was representative of longer-term regional climate
trends. Two long-term stations were used for the comparison,
with daily average air temperature and total precipitation
obtained from Penticton Airport (Environment Canada sta-
tion 1126150) and rst of the month SWE observations
from the Greyback reservoir (located 10 km from the study
at an elevation of 1550 m; BC Snow Survey Network sta-
tion 2F08). It was found that the short-term climate period
provides a range of conditions with respect to temperature
and precipitation, and mean daily, absolute minimum and
maximum daily temperatures, and mean daily precipitation
during this period were consistent with long-term observa-
tions. Likewise, mean SWE observations for the 1997
2005 period were found to be in general agreement with
long-term measurements (19712005).
4. Results
4.1. Coupling of Synthetic Meteorological Data With
DHSVM
[21] In addition to prior model performance evaluations
conducted by Kuras et al. [2011], the performance of cou-
pling the long-term synthetically generated meteorological
data with DHSVM was evaluated by assessing the com-
bined ability of these models to reproduce the general char-
acteristics of ood frequency curves based on observed
streamow data. Hourly discharge ood frequency curves
derived from streamows observed during the pretreatment

7 of 19
W01544 KURAS ET AL.: FOREST HARVESTING IMPACTS ON PEAK FLOWS W01544

period in the watershed (15 observations, 19841998) and
simulated streamow using the Control scenario (100 simu-
lated QP events) are shown in Figure 4. The two ood fre-
quency curves were in general agreement for the entire
range of T from 1.003 to 100 years, and the generated curve
was not statistically different from the observed ( 0.05).
Slopes between the two ood frequency curves were also
consistent, with the ratio of the 100 to 2 year T peak ows,
Q100/Q2, being 2.4 and 2.0 for the observed and simulated
AMS, respectively. Simulated quantiles, however, were
visibly lower for T > 20 years, but some discrepancy was
expected as observed peak ows were derived from the pe-
riod 19841998, while simulated peak ows were generated
using meteorological data representative of the period
19972005.
[22] It is critical to note that the objective of generating
100 years of synthetic meteorological data was not to
reproduce past or to predict future climatic conditions, but
to produce a long-term meteorological time series that
allowed for hydrological simulations to be driven by events
with a wide range of recurrence intervals that are a plausi-
ble realization of the current climate period.
4.2. Scenario Simulations
[23] The Op50 and Op50R scenario GEV ts are pre-
sented visually in comparison to the Control (Figure 5),
while values of DQT,h, DQT,d, and DQT,7 for all scenarios
with and without roads are presented in Tables 4, 5, and 6,
respectively, for T 1.003 to 100 years.
[24] The CR, Op20, and Op20R scenarios did not have a
statistically signicant impact on the peak ow regime of
the watershed, and were therefore considered to be the
most benign treatments in this regard. QT,h, QT,d, and QT,7
for the Op50 scenario signicantly ( 0.05) increased
over the Control for T 2 to 100 years, where DQT,h,
DQT,d and DQT,7 ranged from 1433%, 9%33%, and
11%29%, respectively. The relative increase in Op50
peak ow quantile magnitudes was greatest for the smallest
(T 1.003 years) and largest events (T  50 years), with
moderate changes in medium-sized events. Signicant DQT
generally increased with T for each streamow metric, indi-
cating that following treatment, there is a more substantial
response in streamows for larger recurrence intervals.
Increases in QP frequency following harvesting were also
found to increase with T (for T  2 years), where the
expected frequency of the largest events (T 100 years) in
Control increased roughly 5 to 6.7 times following harvest-
ing, while the expected frequency of medium-sized events
(T 10 years) in Control increased roughly 1.7 to 2 times
following harvesting (Figure 5). This trend was slightly
greater for sustained events (7 day and daily) than for the
near instantaneous (hourly).
[25] The GEV distribution provided a better t for me-
dium-sized events (T  1.25 to 5 years; Figure 5) for all
streamow metrics in each scenario. The variance of the
Control quantile estimator was larger with decreasing sam-
ple size and with increasing T [Schnorbus and Alila,
2004a] and the resulting condence bands used to test the
statistical signicance of the GEV quantile estimators fol-
lowing treatment were narrower for medium-sized T, and
broadened out for larger and smaller T (Figure 5). It was
consequently more difcult to reject the null hypothesis
that quantile estimators for the Control and treatment were
equal for the largest and smallest return period quantiles.
For example, the smallest return period peak ow quantiles
(T 1.003 years) for the Op50 scenario had the greatest
relative increase in magnitude, but were not found to have
signicantly increased over the Control (Tables 46). The

Figure 4. Comparison of generalized extreme value (GEV) t to observed (n 15) and simulated
(Control, n 100) hourly annual peak ow frequency for 241 Creek [after Schnorbus and Alila, 2004a];
also shown is the 5th and 95th percentile range for observations (dashed lines) and Control (bars).

8 of 19
W01544 KURAS ET AL.: FOREST HARVESTING IMPACTS ON PEAK FLOWS W01544

Figure 5. Comparison of GEV ts for Control and (a) Op50 hourly, (b) Op50R hourly, (c) Op50 daily,
(d) Op50R daily, (e) Op50 7 day, and (f) Op50R 7 day scenarios ; included is the sample annual maxi-
mum series (AMS) for each treatment scenario, with dashed lines representing 95% condence intervals
for the Control GEV t. Arrows and notation in Figures 5a, 5c, and 5e indicate event frequency increases
following harvesting.

Table 4. The 241 Creek Hourly QCT and Hourly DQT by Scenario and T
DQT (Percent Change in Hourly) by Treatment Scenario
3 1
T (Years) Hourly Control QCT (m s ) CR Op20 Op20R Op30 Op30R Op50 Op50R

1.003 0.35 1 0 1 10 8 43 37
2 1.14 5 1 4 3 2 14a 7
10 1.70 5 3 2 6 1 19a 12a
20 1.88 5 4 1 8 3 23a 16a
50 2.11 5 5 0 11 5 28a 21a
100 2.26 4 6 1 12 6 33a 25a
a
Two-tailed Z test P (QH C H C
T < QT or QT > QT ) is signicant at  0.05.

9 of 19
W01544 KURAS ET AL.: FOREST HARVESTING IMPACTS ON PEAK FLOWS W01544

Table 5. The 241 Creek Daily QCT and Daily DQT by Scenario and T
DQT (Percent Change in Daily) by Treatment Scenario
3 1
T (Years) Daily Control QCT (m s ) CR Op20 Op20R Op30 Op30R Op50 Op50R

1.003 0.27 1 4 4 13 10 39 33
2 0.93 5 1 6 0 5 9a 2
10 1.35 5 4 1 7 1 16a 9a
20 1.48 5 6 1 10a 3 21a 13a
50 1.63 4 8 3 14a 7 28a 19a
100 1.73 4 11 5 18a 10 33a 25a
a
Two-tailed Z test P (QH C H C
T < QT or QT > QT ) is signicant at  0.05.

GEV t also tended to slightly underestimate DQT for T
20 to 50 years, and this was most apparent in the hourly
AMS for the Op50 and Op50R scenarios (Figure 5).
[26] Adding roads to the Op50 scenario mitigated peak
ows in comparison to the no road scenario, and reduced
the GEV quantile estimators of each streamow metric for
the entire range of T (Tables 46). The mitigating effect of
roads was substantial enough to narrow the T range of
quantile estimators that were found to be signicantly dif-
ferent following treatment, with signicant increases in
DQT reduced to T 10 to 100 years. Statistically signi-
cant DQT was reduced in comparison to Op50 and ranged
from 12% to 25%, 9% to 25%, and 10% to 20% for QT,h,
QT,d, and QT,7, respectively. As was observed for the Op50
scenario, streamows had a more substantial response to
treatment for larger T. The relative decrease in Op50 peak
ow quantiles with the addition of roads was comparable
for all T in each streamow metric (6%9%), with roads
decreasing peak ows across the board. The mitigating
effect of roads on peak ows was quite apparent for all
three streamow metrics of the CR scenario (Tables 46),
where DQT was negative for all T (with the exception
DQT,h, DQT,d 1% for T 1.003 years, which was simply
due to an overestimation of the GEV t).
[27] QT,d and QT,7 for the Op30 scenario were found to
have signicantly increased over the Control for T 20
100 and 10100 years, respectively, where signicant DQT
ranged from 10% to 18% and 8% to 14% for QT,d and QT,7,
respectively. The relatively small variation in DQT for this
scenario indicates that streamows responded similarly to
small and large T events following forest harvesting, which
was contrary to what was found for the Op50 scenario. The
Op30 hourly peak ow quantile estimators had no signi-
cant change over the Control, indicating that the current
state of forest management in 241 Creek has a greater
impact on sustained (daily and 7 day) ows than on the
near instantaneous (hourly). Similar to what was observed
in the Op50 scenario, the addition of roads decreased peak
ow quantiles across the board, and the Op30R peak ow
quantiles were mitigated enough to no longer be signi-
cantly different from the Control in either of the streamow
metrics. Plausible rationale for this simulated response to
roads in 241 Creek are presented in section 4.3, with exam-
ples of the year to year variability in harvesting and road
effects provided in section 4.4.
4.3. Roads
[28] The simulation average value of water intercepted
by roads (both surface and subsurface) is 22% of the CR
average annual water yield (443 mm, Table 7). Roads
potentially extend the stream channel network density in
241 Creek by 7% (Table 3), with 8% of simulated water
intercepted by roads being directly discharged into streams
on an average annual basis. The total surface runoff
(excluding ow in streams) for an average simulation year
is increased by 55% over the Control, indicating the magni-
tude of surface water transported by roads (Table 8).
[29] The CR scenario most clearly illustrates the simu-
lated effect of roads since only 4% of forest cover is
removed for the road network right of way (Figure 3b). It
was initially expected that the small area harvested for the
road right of way would slightly decrease the amount of
evapotranspiration relative to Control; however, the aver-
age annual evapotranspiration component increases by
0.8%, with total sublimation from ground and canopy inter-
cepted snow [Wigmosta et al., 2002] decreasing by 4.3%
(Table 8). On an average annual basis, the loss of forest
cover for the road right of way reduces the amount of subli-
mation resulting from canopy intercepted snow. In addi-
tion, snow covering roads melts more quickly relative to
sublimation under the forest canopy, leading to a further
reduction in the sublimation component. While harvesting

Table 6. The 241 Creek 7 day QCT and 7 day DQT by Scenario and T
DQT (Percent Change in 7 Day) by Treatment Scenario
3 1
T (Years) 7 day Control QCT (m s ) CR Op20 Op20R Op30 Op30R Op50 Op50R

1.003 0.24 1 1 3 1 3 28 22
2 0.70 4 1 3 3 2 11a 5
10 0.98 4 5 0 8a 2 18a 10a
20 1.06 4 6 1 10a 4 21a 13a
50 1.15 4 8 3 12a 6 25a 17a
100 1.22 4 9 4 14a 7 29a 20a
a
Two-tailed Z test P (QH C H C
T < QT or QT > QT ) is signicant at  0.05.

10 of 19
W01544 KURAS ET AL.: FOREST HARVESTING IMPACTS ON PEAK FLOWS
Table 7. Average Annual Road Runoff Processes in the CR Scenario of 241 Creek Based on Simulation Averages
Process Annual Average (mm)
Water intercepted by roads WR 96.6
Culvert discharge to streams CS 7.6
Culvert discharge to hillslope CH 77.9
Excess, WR  CS  CH 11.1
a
Simulation average water yield is 443 mm.
of vegetation in the small area designated for roads reduces
transpiration, the increase in melt water from the greater
snowpack on roads has a relatively higher contribution to
evaporation, which on an average annual basis results in a
slight increase in the collective evapotranspiration compo-
nent of DHSVM.
[30] The mitigating effect of roads on streamow is
found to be most prominent around the time of QP for each
simulation year (SY). Some road segments may only
respond to larger events, potentially acting to divert ows
out of the watersheds natural drainage [Bowling and Let-
tenmaier, 2001]. SY 29 most clearly demonstrates this phe-
nomenon (Figure 6a), with a 2.4% loss of the annual water
yield from the system in comparison to the Control.
Streamow initially increases (relative to Control) on the
rising limb of the hydrograph, however, as road corridors
act as a permanent clear-cut with greater snow accumula-
tion than forested areas and a greater resultant contribution
of melt during the early melt season period [Bowling and
Lettenmaier, 1997]. At rst glance, it appears that roads
may actually be routing ows more rapidly to the outlet on
the rising limb of the hydrograph, which reduces stream-
ows during the subsequent peak ow and falling limb
(Figure 6a). It is important to note, however, that 4% of the
watershed has been harvested for roads in the CR scenario,
with no harvesting in the Control (Figure 3a), and the initial
streamow augmentation in CR is primarily due to a reduc-
tion in the evapotranspiration component. This is consistent
with eld observations, where road surfaces have a consid-
erably greater snowpack than adjacent forest zones, which
generally melts prior to the peak ow period (except in
shaded areas where it may last longer). In addition, SY 41
hydrographs for the Op50 and Op50R scenarios (Figure
6b), where the areas of harvest are equivalent, illustrate the
typical effect of roads in any of the harvesting scenarios,
where streamows are substantially decreased during peak
ow periods, with no recuperation of ows later in the
hydrograph. Examples of this simulated effect are also pro-
vided in section 4.4. The reduction in annual water yield
Table 8. Average Annual Changes in Simulated Hydrological
Processes, Resulting From the Application of Road Networks in
241 Creek, Based on Differences Between CR and Control
Average Annual Difference, Change Relative
Process CR  Control (mm) to Control (%)
Evapotranspiration 1.9 0.8
Vapor mass ux (ground 2.8 4.3
and canopy sublimation)
Surface runoff 75.9 55
Water intercepted by roads 96.6
11 of 19
W01544
Relative to Average Relative to Total Water
Annual Water Yielda (%) Intercepted by Roads (%)
22
2 8
18 81
2 11
resulting from the application of roads to the Control is
consistent for all simulation years (2.3% average), with
approximately 11% of water intercepted by roads being
lost from the system (Table 7).
[31] Because of the extent of saturation during peak ow
periods, lower sections of the watershed have been
observed in the eld to experience near- and above-surface
ows [Kuras et al., 2007], with the watershed outlet often
inundated. Thus critical road segments in the watershed
were evaluated in simulations to determine whether the
addition of roads to the Control scenario furthered this pro-
cess. It was found that the simulated road segment hydro-
graph for the southerly most segment on the west road
(Figure 7), plateaued for a time interval that coincided with
the CR peak ow period (Figure 6a). As is the case in real-
ity, this particular road segment drains out of the watershed
and as a result, corresponds to the total volume of water
(the east road did not convey ows directly out of the ba-
sin) transported directly out of the watershed via road sur-
face runoff and ditch ow. The volume of water conveyed
by the road segment however, is miniscule in comparison
to the total loss of water for SY 29, but the plateauing of
the road segment hydrograph nevertheless illustrates the
simulated accumulation of near-surface water around the
outlet of the watershed. Water table maps for SY 29 of
the Control and CR scenarios are presented in Figure 8
for the hour of QP (8 May, 16:00 LT), where pixels repre-
sent the simulated water table depth below the surface.
Above surface water levels may only be queried at the
DEM pixel scale, and areas where the water table depth is
equal to zero (white pixels in Figure 8) indicate that the
level of water is either at the surface or above. The spatial
distribution of subsurface ow in the Control was altered
with the addition of roads, and areas C and D (Figure 8a)
experienced an overall reduction in ows following the
application of roads, while ows in areas around locations
A and B (Figure 8b) increased. Roads appear to be divert-
ing ows to hillslope areas that do not contribute to stream-
ow at the weir and further transport water out of the basin
by both surface and subsurface ow paths.
[32] SY 29 groundwater level uctuations at DEM pixels
A and B (Figure 8b) are presented in Figure 9 for CR,
which demonstrate at and slightly above (not visible) sur-
face ows during the QP period. This is consistent with
prior eld observations around the outlet of the watershed
during peak ow periods. The level of simulated inundation
in this area, however, is not nearly as substantial as would
be expected on the basis of eld observations. This is most
likely due to model ow routing being dictated by a DEM
with the assumption that there is no vertical saturated
groundwater ow in a DEM pixel. While return ow and
W01544 KURAS ET AL.: FOREST HARVESTING IMPACTS ON PEAK FLOWS W01544

Figure 6. Comparison of hourly streamow hydrographs for (a) simulation year (SY) 29 CR (Treat-
ment, T) and Control (C) and (b) SY 41 Op50R (T) and Op50 (C) scenarios, together with residual time
series (T-C) and meteorological model hourly precipitation and temperature.

saturation overland ow are simulated in locations where
grid cell water tables intersect the ground surface [Wig-
mosta et al., 2002], DHSVM will tend to overestimate rein-
ltration of surface ows and underestimate rising
groundwater levels.
[33] Simulated surface ows potentially owing from
the watershed through pathways other than the outlet do
not solely account for the change in water yield following
the application of roads (2.3% on average), and a much
Figure 7. Road segment hydrograph for the lower west
road (Figure 1).
greater portion of these ows may be routed out of the
basin through subsurface means. The area surrounding
location B (Figure 8b), with its close proximity to the
watershed boundary and large increase in ows following
the addition of roads, is expected to be a potential pathway
for this process.
4.4. Variable Responses to Harvesting and Roads
[34] SY 10 provides an example of a large decrease in
treatment QP relative to Control (DQP) (Op50 DQP
40%) with no change in timing (Figure 10a). The exposed
snowpack in the treatment scenario (augmented in clear-cut
areas compared to Control) melts early in the season during
two temperature increases (1214 and 2125 April), which
raises early season peak ows. Treatment causes the main
snowmelt period to advance in time, but melt rates are not
high enough to increase early season peak ows to the Con-
trol QP level or above. The duration and rate of early sea-
son snowmelt determines the timing of treatment QP,
which can either occur earlier (relative to Control) or with
no change in timing. Treatment QP is still synchronized
with a rainfall event (15 May) that increases it above early
season peak ows, but the availability of basin-wide SWE
and the contribution of melt to QP is less during the original
timing (Control) of the event, and QP does not increase
over pretreatment levels. SY 57 is an example of a rela-
tively smaller Control QP that is also reduced with

12 of 19
W01544 KURAS ET AL.: FOREST HARVESTING IMPACTS ON PEAK FLOWS W01544

Figure 8. Water table maps (depth below surface) for the (a) Control and (b) CR at the time of QP for
SY 29 (5 May at 16:00 LT). Pixels at locations A and B were queried for CR groundwater level uctua-
tions (Figure 1). Areas C and D are referred to in section 5.1.

treatment (Figure 10b), but the main freshet period for this
particular SY is primarily a function of radiation induced
melt with virtually no rainfall occurring this time. Treat-
ment reduces QP (Op50 DQP 26%) for SY 57 and
advances the peak ow by 15 days.
[35] SY 10 also provides an example of treatment effect
on reducing late melt season ows (relative to Control: 31
May to 30 June) due to an increase in early season melt.
The mitigating effect of roads when combined with har-
vesting is most apparent for SY 10 during the early freshet

Figure 9. CR groundwater levels at digital elevation model pixels A and B (Figure 8b) for SY 29; the
CR hydrograph has been included for reference.

13 of 19
W01544 KURAS ET AL.: FOREST HARVESTING IMPACTS ON PEAK FLOWS W01544

Figure 10. Comparison of hourly streamow hydrographs for (a) SY 10 Op50R (T), Op50 (C), and
Control and (b) SY 57 Op50R (T), Op50 (C), and Control, together with residual time series (T-C) and
meteorological model hourly precipitation and temperature.

period, which is dominated by snowmelt (Figure 10a).
Road mitigation does not necessarily occur during high
ows, but appears to be more dependent on the level of sat-
uration in the lower elevations of the watershed [Kuras
et al., 2007, 2011]. The 25 April treatment peak ow, for
example, results purely from snowmelt and the mitigating
effect of roads is very apparent at this time; on the other
hand, the effect of roads is negligible later in the season (15
May QP) during a larger event that is more strongly inu-
enced by rainfall. This is due to steady snowmelt inputs
that saturate the lower elevations of the watershed during
the peak ow period, and create favorable conditions for
roads to indirectly transport ows out of the basin (i.e.,
through subsurface pathways).
[36] SY 86 provides an example of an increase in
treatment QP relative to Control (Op50 DQP 34%,
Figure 11a). Two main peak ows for this freshet corre-
spond to two large increases in temperature (1012 April
and 29 May) and with harvesting, a high rate of early sea-
son snowmelt from the augmented and exposed snowpack
in clear-cut areas increases the rst peak ow over the Con-
trol, making it the largest for the year. A lack of SWE later
in the treatment freshet lowers the second peak ow and
QP occurs 25 days earlier relative to Control. The magni-
tude of a peak ow increase following harvesting may be
more dependent on temperature during the snowmelt period
than the size of the snowpack [Whitaker et al., 2002]. Some
of the largest peak ows can occur later in the melt season,
at a time when the snowpack is so ripe for melt that a sud-
den increase in temperature results in an abrupt and rapid
rate of snowmelt input, and the nearly uniformly melting
snowpack in 241 Creek only enhances this outcome. For
this process to occur, early season temperature increases
must evolve in such a way as to allow for a majority of the
snowpack to remain until the time of the peak ow. A
prime example is SY 49, which experiences a large
increase in QP following treatment, without a change in
timing (Figure 11b). Both precipitation and radiation
induced melt contribute to QP for this SY, whereby the
snowpack melts rapidly during a sharp temperature
increase (27 April to 2 May) combined with a rainfall event
(1 May), which results in increasing QP by 71% relative to
Control.
5. Discussion
5.1. Road Implications
[37] Combining road networks in 241 Creek with either
of the scenarios (Control, Op20, Op30 or Op50) results in a
reduction of simulated streamows reported at the water-
shed outlet. The mitigating effect of roads on peak ows
was unexpected, as it is the general consensus in forest

14 of 19
W01544 KURAS ET AL.: FOREST HARVESTING IMPACTS ON PEAK FLOWS W01544

Figure 11. Comparison of hourly streamow hydrographs for (a) SY 86 Op50R (T), Op50 (C), and
Control and (b) SY 49 Op50R (T), Op50 (C), and Control, together with residual time series (T-C) and
meteorological model hourly precipitation and temperature.

road hydrology that roads tend to increase peak ow dis-
charge by routing water more quickly to streams [e.g.,
Wemple et al., 1996; Alila et al., 2009]. The reduction of
peak ows by roads is not a novel concept [e.g., King and
Tennyson, 1984; Wigmosta and Perkins, 2001], however,
and simulations indicate that roads transport water out of
the watershed either directly through roadside ditches (in
agreement with what has been observed in the eld), or
indirectly through culvert outows discharging onto hill-
slope areas that further transport water out of the basin
through surface and subsurface pathways. Simulations
imply that the latter process has the most dominant contri-
bution. In essence, ows leaving the watershed through
pathways other than the outlet are lost, creating a water
decit in the input-output balance of the model. If the scope
of this study were able to encompass a larger watershed
area (i.e., examining changes further downstream at a point
that encompasses the entire road network), then it would be
expected that roads would either increase or have no effect
on QP.
5.2. Harvesting and Roads
[38] Short duration streamow metrics (hourly in this
study) help to identify watershed responses that would be
masked by longer-duration metrics, including: quick catch-
ment response (typically 46 h for 241 Creek), diurnal vari-
ability during the snowmelt season [Kuras et al., 2007],
effects of rainfall during snowmelt, and the inuence of
evapotranspiration on streamow during low-ow periods.
Longer-duration metrics (daily and 7 day in this study), on
the other hand, best assess sustained ows in snowmelt-
dominated watersheds, which provide a better depiction of
longer-term snowmelt contributions to streamow than
does the hourly metric. In addition, channel forming events
are driven by a combination of both high ows and fre-
quently occurring ows [Wolman and Miller, 1960]. Along
these lines, the daily and 7 day streamow metrics provide
a means of characterizing sustained high ows that may
maintain channel morphology. Changes in metrics that
quantify ows over longer sampling windows, may indicate
changes in geomorphologically effective oods. 241 Creek
is an upper tributary to Penticton Creek, which supplies
municipal water to the City of Penticton. Such changes can
precipitate changes in channel form, which could propagate
downstream and affect community water supplies. Forest
harvesting activities in this watershed can lead to changes
in channel form that could result in channel destabilization
and stream bank erosion, which would have a direct impact
on downstream water quality.
[39] Peak ow regime responses following harvesting in
241 Creek exhibit similar trends for the Op20, Op30 and
Op50 scenarios, but the smallest events (T 1.003 years)

15 of 19
W01544 KURAS ET AL.: FOREST HARVESTING IMPACTS ON PEAK FLOWS
are unparalleled in relation, and have the relatively largest
but statistically insignicantresponse to the future harvest-
ing scheme (Op50) (Tables 46). Op50 DQT,h,, DQT,d and
DQT,7 are found to increase with T for T  2 years, which
is similar to the ndings of Schnorbus and Alila [2004a] for
QT,d and QT,7 with a similar harvest area in Redsh Creek;
however, the most extreme events at that particular study
site are typically derived from rain on melting snow (not
the case for 241 Creek) and QT,h did not continue to
increase for larger T, as the effects of interception for such
events are marginal [Schnorbus and Alila, 2004a]. The CR,
Op20, and Op20R scenarios are not found to have signi-
cantly (in practical or statistical terms) affected QT relative
to Control while all streamow metrics for Op50 have a
signicant impact on QT for T  2 years. Op30 also has no
signicant effect on QT,h, but QT,d and QT,7 are signicantly
affected for T  20 years and T  10 years, respectively.
The current harvesting scheme, as such, is found to have an
increasingly signicant effect on more sustained ows in
the watershed, which may have a direct impact on channel
morphology. Because of the mitigating effect of roads (sec-
tion 5.1), however, DQT decreases to statistically insigni-
cant levels in the Op30R scenario, indicating that the
existing condition of the watershed (i.e., with roads) does
not signicantly affect QT for any of the Op30R streamow
metrics. It is important to note, however, that changes in
streamow due to treatment vary spatially in the watershed,
and while streamow metrics reported at the watershed
outlet could be insignicant, changes elsewhere could be
substantial enough to affect channel morphology.
[40] Planned harvesting for the watershed (Op50) has a
signicant effect (with and without roads) on QT for all
streamow metrics, and the ranges of signicant DQT,h,
DQT,d and DQT,7 are approximately twice the magnitude
than what was found in the Redsh Creek long-term mod-
eling study [Schnorbus and Alila, 2004a]. The temporal
progression of basin melt differs for the two watersheds
because of differences in basin physiography, as the steep
gradient of Redsh Creek produces a clear snowline retreat,
while the snowpack in 241 Creek typically melts more uni-
formly in the subdued topography with greater short-term
snowmelt inputs. In addition, 40% of the Redsh Creek
upper basin area is inoperable, sparsely vegetated subalpine
parkland, and can reduce the apparent impact of harvesting
in the watershed [Schnorbus and Alila, 2004a].
[41] It is difcult to compare the ndings of numerical
modeling and paired watershed studies as, because of an
inappropriate type of event pairing and relatively short
postharvest record periods, the ndings of the latter are of-
ten only relevant to average peak ow conditions (T  2
years). Paired watershed studies in similar snow-dominated
environments that have found increases in mean annual
daily maximum peak ows ranging 11%66% [Troendle
and King, 1985; Cheng, 1989; Burton, 1997; Troendle
et al., 2001], as such, can only be compared to DQT,d (T
2 years), and indicate that the predicted value for Op50
(9%, Table 5) is at the lower end of the published range.
[42] The relative decrease in Op50 peak ow quantiles
with the addition of roads is similar by T for each stream-
ow metric, and roads decrease peak ow discharges
across the board (Tables 46). This is contrary to the nd-
ings of La Marche and Lettenmaier [2001], who modeled
16 of 19
W01544
road effects and harvesting with DHSVM and found that
road effects increased with increasing T, while the effects
of forest harvesting decreased. Similar to the ndings of
King and Tennyson [1984], the spatial arrangement of
roads in 241 Creek alters the natural drainage pattern and
acts to indirectly route ows out of the basin through path-
ways other than the outlet.
[43] In this study, forest harvesting has the greatest effect
both on the magnitude and the frequency of the largest
events, while medium-sized events are only moderately
affected (most clearly illustrated in Figures 5a and 5b). In
essence, the mean of the peak ow distribution is not as
affected following harvesting as the variability around the
mean, with greater deformation at the extremes or tails of
the distribution. Prior to logging (Control), the distribution
of the AMS is best t by a near straight line, whereas the
rearrangement of the peak ow distribution following har-
vesting results in a curvature in the t that is most pro-
nounced for the smallest (T 1.003 years) and largest
(T > 20 years) events (Figure 5). This observation is best
explained by the variable simulation year responses to har-
vesting with examples presented in section 4.2, which illus-
trate the ability of the model to realistically simulate the
complex processes involved.
[44] Model simulations reveal three intriguing patterns in
this snow-dominated catchment: (1) forest harvesting shifts
peak ows of all sizes upward, (2) forest harvesting affects
larger peak ows more so than small peak ows (for T > 2
years), and (3) increases in peak ow frequency following
harvesting increase with T (for T > 2 years). Current forest
hydrology literature (summarized in several reviews such
as those by Eaton and Church [2001], Troendle et al.
[2001, 2010], MacDonald and Stednick [2003], Scherer
and Pike [2003], Moore and Wondzell [2005], and Grant
et al. [2008]) suggests that forest harvesting has a progres-
sively smaller effect on larger peak ow events. The effects
of forest harvesting on peak ows have been analyzed in a
majority of this literature by pairing peak ow events by
chronology (chronological pairing), where the response of
a managed watershed (e.g., some level of harvesting and
perhaps road construction) has been compared to that of an
unmanaged watershed using the analysis of variance
(ANOVA) or analysis of covariance (ANCOVA) proce-
dures. The results of these types of analyses, however, have
recently been challenged by Alila et al. [2009] since a peak
ow regime inherently includes both frequency of occur-
rence and magnitude components that cannot be analyzed
simultaneously with these techniques.
[45] As a result of Alila et al. [2009], some high-prole
regulatory agencies have recognized the awed nature of
decades of literature based on chronological pairing [U.S.
Department of Agriculture, 2010, pp. 37; Natural Resour-
ces Canada/Canadian Forest Service, 2011, p. 19]. In con-
trast, a coarse review of literature published in top rated
science journals in the last two years reveals that chronolog-
ical pairing is still the modus operandi of research methods
in experimental [e.g., Bathurst et al., 2011a], observational
[e.g., Jones and Perkins, 2010; Tran et al., 2010], and nu-
merical modeling [e.g., Bathurst et al., 2011b; Birkinshaw
et al., 2010; Zegre et al., 2010; Seibert and McDonnell,
2010; Seibert et al., 2010] studies evaluating the effects of
forests, deforestation, and forest harvesting on oods. It is
W01544 KURAS ET AL.: FOREST HARVESTING IMPACTS ON PEAK FLOWS
time for the forest hydrology community to put an end to
the confusion, misconception, and misinformation spread
by the use of such methods. Our study ndings are inconsis-
tent with the prevailing perception shaped by decades of
chronological pairingbased, peer-reviewed studies that
suggest the effects of forest harvesting decrease with
increasing peak ow T, and simulations in this study indi-
cate that these effects actually increase with T. This trend
can only be revealed with an analysis that pairs peak ow
events by frequency [Alila et al., 2009, 2010], the results of
which appear to be physically linked to peak ow runoff
generation processes in snowmelt-dominated catchments.
The largest peak ow events generally occur in years when
an average or greater snowpack remains late into the melt
season, and rapid melt inputs (potentially combined with
precipitation) follow an abrupt increase in temperature. For
these types of conditions, the greater the premelt season
snowpack, the greater the potential for the occurrence of a
larger return period event; harvesting can only increase this
potential because of an increase in snow accumulation and
changes in snowmelt energetics and runoff synchronization
[Schnorbus and Alila, 2004a]. Furthermore, peak ow fre-
quency curves in snow-dominated watersheds have a rela-
tively mild slope, which translates into larger return period
events being more easily affected by modest changes in
these parameters [Alila et al., 2009].
6. Summary and Conclusions
[46] The 241 Creek DHSVM application [Thyer et al.,
2004] was coupled with long-term synthetically generated
meteorological data to assess the impacts of roads and har-
vesting (independently and in combination) in this snow-
dominated catchment. Streamows were simulated for
prior (20% harvest area), existing (30% harvest area) and
future (50% harvest area) harvesting scenarios with and
without roads, and model output was used to assess the sen-
sitivity of peak ow quantiles for the hourly, daily, and
7 day discharge metrics. Meteorological generation is not
intended to predict future or to replicate past long-term cli-
mate data, but instead to produce a plausible realization of
the current climate period that can be used to drive hydro-
logic simulations for a sufcient number of water years in
order to assess operational impacts for a wide range of re-
currence intervals. This approach assesses the immediate
operational impacts on the peak ow regime for the critical
years following treatment (vegetation is held stationary
during simulations), which overcomes the confounding
factors of short sample size and forest regrowth induced
hydrologic recovery associated with traditional paired
watershed studies. Prior evaluations for this particular
model application [Kuras et al., 2011] have shown general
success in realistically simulating spatially variable internal
catchment ow routings (roads, streams and perched water
table dynamics), and imply that the model can be used as a
reliable tool for disentangling road and harvesting effects
in the current application.
[47] Model simulations indicate that prior (20% harvest
area) and current (30% harvest area) operational conditions
in the watershed do not have a signicant effect on the
peak ow regime. The current condition has a signicant
effect on sustained ows if the effect of roads is removed,
17 of 19
W01544
but the inclusion of the road network reduces this impact to
insignicant levels. The application of roads to any of the
harvesting scenarios reduces the average annual water
yield, and roads are responsible for transporting water both
indirectly and directly out of the basin. Roads divert ows
from the natural drainage to lower-elevation areas close to
the watershed boundary, and spatially extensive saturation
typically observed in this area during snowmelt events aids
in transporting ows out of the basin throughow paths
(surface and subsurface) other than the outlet. Roads also
transport water directly out of the watershed through road-
side ditches and surface runoff, but the volume transported
through these pathways is minute. Field observations have
indicated that the watershed boundary does not entirely
enclose both surface and subsurface ows, and the model
appears to indicate the routes and processes involved in
producing slight water budget decits. It is also critical to
note, however, that if the scope of this study were able to
encompass a larger watershed area (i.e., examining changes
further downstream at a point that encompasses the entire
road network), then it would be expected that roads would
either have no additional effect or increase the impacts of
harvesting on peak ows.
[48] While roads reduced peak ow quantiles for the
forthcoming harvest plan (50% harvest area) by 6%9% for
all T in each streamow metric, this only results in reducing
the mean annual peak ow (T 2 years) to insignicant
levels. Planned harvesting (50% harvest area with roads) is
found to have both a statistically and physically signicant
effect (9%25% over Control) on peak ows with recur-
rence intervals ranging 10100 years for all three stream-
ow metrics. This indicates that the effects of this particular
scenario may have substantial ecological, hydrological and
geomorphological consequences. This is especially the case
since increases in peak ow frequency following harvesting
increase with T, with the largest events (T 100 years)
becoming 56.7 times more frequent (compared to Control).
This can lead to changes in channel form, for example, that
could result in channel destabilization and stream bank ero-
sion with a direct impact on downstream water quality.
Contrary to the prevailing perception in forest hydrology,
the effects of harvesting are found to increase with T, which
is attributable to the uniqueness of peak ow runoff genera-
tion processes in snow-dominated catchments. The ndings
of this research suggest that peak ow regimes are fairly
tolerant to the current level of harvesting in this particular
watershed, but that further harvesting operations may affect
this element signicantly. The reported changes in the fre-
quency and the magnitude of peak ows for our study
watershed should only be extended to other basins with
care, as it is anticipated that these changes are strongly de-
pendent upon basin physiography, climate, and forestry
practices. It is therefore necessary to determine the extent
to which our reported changes in peak ow magnitude and
frequency can be generalized to other basins, by conducting
similar studies in different hydroclimate regimes. Sugges-
tions for future studies also include (1) sensitivity analyses
of hypothetical harvest levels and spatial distributions to
determine basin-specic maximum thresholds below which
peak ow regimes are not signicantly affected and (2)
power spectrum analyses to assess changes in the periodic-
ity of ows following harvesting.
W01544 KURAS ET AL.: FOREST HARVESTING IMPACTS ON PEAK FLOWS
[49] Acknowledgments. In memory of Richard Finlayson. This work
was partially funded by the National Science and Engineering Research
Council of Canada Discovery grant RGPIN 194388-11 to Y. Alila and
Forest Investment Account Forest Science Program project funding
(Y073115) to R. Winkler. The authors wish to thank the entire research
team at the Upper Penticton Creek Watershed Experiment for making this
research possible and for providing much needed long-term data, eld sup-
port, and advice. Further acknowledgments are also owed to A. and H.
Kuras for much needed support, the British Columbia Ministry of Forests
and Range (Kamloops Forest Region) for providing study site facilities,
and the Weyerhaeuser Company for allowing access to their eld opera-
tions. Further grateful acknowledgment is owed to the unerring assistance
provided by friends and colleagues at the University of British Columbia,
particularly J. Maedel for his expert advice in Geographical Information
Systems analysis and M. Schnorbus and A. Anderson for their aid in com-
puter programming. We would also like to thank A. Canon and A. Zim-
mermann for discussions.
References
Alila, Y., P. K. Kuras, M. Schnorbus, and R. Hudson (2009), Forests and
oods: A new paradigm sheds light on age-old controversies, Water
Resour. Res., 45, W08416, doi:10.1029/2008WR007207.
Alila, Y., R. Hudson, P. K. Kuras, M. Schnorbus, and K. Rasouli (2010),
Reply to comment by Jack Lewis et al. on Forests and oods: A new
paradigm sheds light on age-old controversies, Water Resour. Res., 46,
W05802, doi:10.1029/2009WR009028.
Bathurst, J. C., et al. (2011a), Forest impact on oods due to extreme rain-
fall and snowmelt in four Latin American environments 1: Field data
analysis, J. Hydrol., 400, 281291, doi:10.1016/j.jhydrol.2010.11.044.
Bathurst, J. C., et al. (2011b), Forest impact on oods due to extreme rain-
fall and snowmelt in four Latin American environments 2: Model analy-
sis, J. Hydrol., 400, 292304, doi:10.1016/j.jhydrol.2010.09.001.
Birkinshaw, S. J., J. C. Bathurst, A. Iroume, and H. Palacios (2010), The
effect of forest cover on peak ow and sediment discharge: An integrated
eld and modelling study in central-southern Chile, Hydrol. Processes,
25(8), 12841297, doi:10.1002/hyp.7900.
Bowling, L. C., and D. P. Lettenmaier (1997), Evaluation of the effects of
forest roads on streamow in Hard and Ware Creeks, Washington, Water
Resour. Ser. Tech. Rep. 155, 189 pp., Univ. of Wash., Seattle.
Bowling, L. C., and D. P. Lettenmaier (2001), The effects of forest roads
and harvest on catchment hydrology in a mountainous maritime environ-
ment, in Land Use and Watersheds: Human Inuence on Hydrology and
Geomorphology in Urban and Forest Areas, Water Sci. Appl. Ser.,
vol. 2, edited by M. S. Wigmosta and S. J. Burges, pp. 145164, AGU,
Washington, D. C.
Bowling, L. C., P. Storck, and D. P. Lettenmaier (2000), Hydrologic effects
of logging in western Washington, United States, Water Resour. Res.,
36, 32233240.
Burton, T. A. (1997), Effects of basin-scale timber harvest on water yield
and peak streamow, J. Am. Water Resour. Assoc., 33(6), 11871196.
Buttle, J. M., P. W. Hazlett, C. D. Murray, I. F. Creed, D. S. Jeffries, and R.
Semkin (2001), Prediction of groundwater characteristics in forested and
harvested basins during spring snowmelt using a topographic index,
Hydrol. Processes, 15, 33893407.
Cheng, J. D. (1989), Streamow changes after clear-cut logging of a pine
beetle-infested watershed in southern British Columbia, Canada, Water
Resour. Res., 25, 449456.
Duncan, S. H. (1986), Peak stream discharge during thirty years of sus-
tained yield timber management in two fth order watersheds in Wash-
ington state, Northwest Sci., 60(4), 258264.
Eaton, B., and M. Church (2001), Hydrological effects of forest harvest in
the Pacic Northwest, Riparian Decis. Tool Tech. Rep., 3, 56 pp., Dept.
Geogr., Univ. of British Columbia, Vancouver, B. C., Canada.
Grant, G. E., S. L. Lewis, F. J. Swanson, J. H. Cissel, and J. J. McDonnell
(2008), Effects of forest practices on peak ows and consequent channel
response: A state-of-science report for western Oregon and Washington,
Gen. Tech. Rep., PNW-GTR-760, 76 pp., Pac. Northwest Res. Stn., U.S.
Dept. Agric., Portland, Ore.
Harr, R. D., A. Levno, and R. Mersereau (1982), Streamow changes after
logging 130-year-old Douglas r in two small watersheds, Water Resour.
Res., 18, 637644.
Hope, G. (2001), Soil descriptions for Penticton Creek experimental water-
sheds, report, 19 pp., Kamloops For. Reg., B. C. Minist. of For. and
Range, Kamloops, B. C., Canada.
18 of 19
W01544
Jones, J. A. (2000), Hydrologic processes and peak discharge response to
forest removal, regrowth, and roads in 10 small experimental basins,
western Cascades, Oregon, Water Resour. Res., 36, 26212642.
Jones, J. A., and G. E. Grant (1996), Peakow responses to clear-cutting
and roads in small and large basins, western Cascades, Oregon, Water
Resour. Res., 32, 959974.
Jones, J. A., and G. E. Grant (2001), Comment on Peakow responses to
clear-cutting and roads in small and large basins, western Cascades, Ore-
gon: A second opinion by R. B. Thomas and W. F. Megahan, Water
Resour. Res., 37, 175178.
Jones, J. A., and R. M. Perkins (2010), Extreme ood sensitivity to snow
and forest harvest, western Cascades, Oregon, United States, Water
Resour. Res., 46, W12512, doi:10.1029/2009WR008632.
King, J. G., and L. C. Tennyson (1984), Alteration of streamow character-
istics following road construction in north central Idaho, Water Resour.
Res., 20, 11591163.
Kuras, P. K., M. Weiler, and Y. Alila (2007), The spatiotemporal variability
of runoff generation and groundwater dynamics in a snow-dominated
catchment, J. Hydrol., 352, 5066, doi:10.1016/j.jhydrol.2007.12.021.
Kuras, P. K., Y. Alila, M. Weiler, D. Spittlehouse, and R. Winkler (2011),
Internal catchment process simulation in a snow-dominated basin: Per-
formance evaluation with spatiotemporally variable runoff generation and
groundwater dynamics, Hydrol. Processes, 25, 31873203, doi:10.1002/
hyp.8037.
La Marche, J. L., and D. P. Lettenmaier (2001), Effects of forest roads on
ood ows in the Deschutes River, Washington, Earth Surf. Processes
Landforms, 26, 115134.
MacDonald, L. H., and J. Stednick (2003), Forests and water: A state of
the-art review for Colorado, Completion Rep., 196, 65 pp., Colo. Water
Resour. Res. Inst., Fort Collins.
Moore, R. D., and S. M. Wondzell (2005), Physical hydrology and the
effects of forest harvesting in the Pacic Northwest: A review, J. Am.
Water Resour. Assoc., 41, 763784.
Murray, C. D., and J. M. Buttle (2003), Impacts of clearcut harvesting on
snow accumulation and melt in a northern hardwood forest, J. Hydrol.,
271, 197212.
Natural Resources Canada/Canadian Forest Service (2011), Mountain pine
beetle: A synthesis of the ecological consequences of large-scale distur-
bances on sustainable forest management, with emphasis on biodiversity,
Inf. Rep. BC-X-426, NRC/CFS, Victoria, B. C. [Available at http://
publications.gc.ca/collections/collection_2011/rncan-nrcan/Fo143-2-426-
eng.pdf.]
Rothacher, J. (1970), Increases in water yield following clear-cut logging in
the Pacic Northwest, Water Resour. Res., 6, 653658.
Scherer, R., and G. P. Pike (2003), Effects of forest management activities
on streamow in the Okanagan Basin, British Columbia: Outcomes of a
literature review and workshop, FORREX Ser. 9, 52 pp., For. Res. Ext.
Partnership (FORREX), Kamloops, B. C., Canada.
Schmidt, R. A., and C. A. Troendle (1989), Snowfall into a forest and clear-
ing, J. Hydrol., 110, 335348.
Schnorbus, M., and Y. Alila (2004a), Forest harvesting impacts on the
peakow regime in the Columbia Mountains of southeastern British Co-
lumbia: An investigation using long-term numerical modeling, Water
Resour. Res., 40, W05205, doi:10.1029/2003WR002918.
Schnorbus, M., and Y. Alila (2004b), Generation of an hourly meteorologi-
cal time series for an alpine basin in British Columbia for use in numeri-
cal hydrologic modeling, J. Hydrometeorol., 5(5), 862882.
Seibert, J., and J. McDonnell (2010), Land-cover impacts on streamow: A
change-detection modelling approach that incorporates parameter uncer-
tainty, Hydrol. Sci. J., 55(3), 316332, doi:10.1080/02626661003683264.
Seibert, J., J. J. McDonnell, and R. D. Woodsmith (2010), Effects of wild-
re on catchment runoff response: A modelling approach to detect
changes in snow-dominated forested catchments, Hydrol. Res., 41(5),
378390, doi:10.2166/nh.2010.036.
Spittlehouse, D. L. (1998), Rainfall interception in young and mature conifer
forests in British Columbia, paper presented at 23rd Conference on Agri-
cultural and Forest Meteorology, Am. Meteorol. Soc., Albuquerque, N. M.
Spittlehouse, D. L. (2000), Using time domain reectometry in stony forest
soil, Can. J. Soil Sci., 80, 311.
Spittlehouse, D. L. (2002), Sap ow in old lodgepole pine trees, paper pre-
sented at 25th Conference on Agricultural and Forest Meteorology, Am.
Meteorol. Soc., Norfolk, Va., 2024 May.
Spittlehouse, D. L., and R. D. Winkler (2002), Modelling snowmelt in a for-
est and clearcut, paper presented at 25th Conference on Agricultural and
Forest Meteorology, Am. Meteorol. Soc., Norfolk, Va., 2024 May.
W01544 KURAS ET AL.: FOREST HARVESTING IMPACTS ON PEAK FLOWS
Stedinger, J. R., R. M. Vogel, and E. Foufoula-Georgiou (1993), Frequency
analysis of extreme events, in Handbook of Hydrology, edited by D. R.
Maidment, chap. 18, McGraw-Hill, New York.
Thomas, R. B., and W. F. Megahan (1998), Peakow responses to clear-
cutting and roads in small and large basins, western Cascades, Oregon:
A second opinion, Water Resour. Res., 34, 33933403.
Thyer, M., J. Beckers, D. Spittlehouse, Y. Alila, and R. Winkler (2004),
Diagnosing a distributed hydrologic model for two high-elevation for-
ested catchments based on detailed stand- and basin-scale data, Water
Resour. Res., 40, W01103, doi:10.1029/2003WR002414.
Tran, P., F. Marincioni, and R. Shaw (2010), Catastrophic ood and forest
cover change in the Huong river basin, central Viet Nam: A gap between
common perceptions and facts, J. Environ. Manage., 91(11), 21862200,
doi:10.1016/j.jenvman.2010.05.020.
Troendle, C. A., and R. M. King (1985), The effect of timber harvest on
the Fool Creek watershed, 30 years later, Water Resour. Res., 21, 1915
1922.
Troendle, C. A., and J. O. Reuss (1997), Effect of clear cutting on snow
accumulation and water outow at Fraser, Colorado, Hydrol. Earth Syst.
Sci., 1(2), 325332.
Troendle, C. A., M. S. Wilcox, G. S. Bevenger, and L. S. Porth (2001), The
Coon Creek water yield augmentation project: Implementation of timber
harvesting technology to increase streamow, For. Ecol. Manage., 143,
179187.
Troendle, C. A., L. MacDonald, C. H. Luce, and I. J. Larsen (2010), Fuel
management and water yield, Cumulative Watershed Effects of Fuel Man-
agement in the Western United States, edited by W. J. Elliot, I. S. Miller
and L. Audin, Gen. Tech. Rep. RMRS-GTR-231, pp. 126148, U.S. Dept.
Agr., Forest Serv., Rocky Mountain Res. Stat., Fort Collins, Col.
U.S. Department of Agriculture (2010), Final environmental impact state-
ment: Fernow Experimental Forest, report, North. Res. Stn., For. Serv.,
Newton Square, Pa. [Available at http://nrs.fs.fed.us/ef/locations/wv/
fernow/EIS/resources/Final_EIS_Fernow_Exp_Forest_04Oct.pdf.]
Ward, R. C. (1971), Small Watershed Experiments: An Appraisal of Con-
cepts and Research Developments, 254 pp., Univ. of Hull, Hull, U. K.
Wemple, B. C., J. A. Jones, and G. E. Grant (1996), Channel network exten-
sion by logging roads in two basins, Western Cascades, Oregon, Water
Resour. Bull., 32(6), 11951207.
Whitaker, A., Y. Alila, and J. Beckers (2002), Evaluating peak ow sensi-
tivity to clear-cutting in different elevation bands of a snowmelt-
dominated mountainous catchment, Water Resour. Res., 38(9), 1172,
doi:10.1029/2001WR000514.
Whitaker, A., Y. Alila, J. Beckers, and D. Toews (2003), Application of the
distributed hydrology soil vegetation model to Redsh Creek, British
19 of 19
W01544
Columbia: Model evaluation using internal catchment data, Hydrol.
Processes, 17, 199224.
Wigmosta, M. S., and W. A. Perkins (2001), Simulating the effects of forest
roads on watershed hydrology, in Land Use and Watersheds: Human
Inuence on Hydrology and Geomorphology in Urban and Forest Areas,
Water Sci. Appl. Ser., vol. 2, edited by M. S. Wigmosta and S. J. Burges,
pp. 127143, AGU, Washington, D. C.
Wigmosta, M. S., L. W. Vail, and D. P. Lettenmaier (1994), A distributed
hydrology-vegetation model for complex terrain, Water Resour. Res., 30,
16651679.
Wigmosta, M. S., B. Nijssen, P. Storck, and D. P. Lettenmaier (2002), The
distributed hydrology soil vegetation model, in Mathematical Models of
Small Watershed Hydrology and Applications, edited by V. P. Singh and
D. K. Frevert, pp. 742, Water Resour. Publ., Littleton, Colo.
Winkler, R. D., D. L. Spittlehouse, B. A. Heise, T. R. Giles, and Y. Alila
(2003), The Upper Penticton Creek Watershed Experiment: A review at
year 20, in Water Stewardship: How Are We Managing? Proceedings
of the 56th Annual Meeting, pp. 5158, Can. Water Resour. Assoc.,
Cambridge, Ont. [Available at http://www.for.gov.bc.ca/hre/pubs/pubs/
1304.htm.]
Winkler, R., D. Spittlehouse, T. Giles, B. Heise, G. Hope, and M. Schnorbus
(2004), Upper Penticton Creek: How forest harvesting affects water
quantity and quality, Streamline Watershed Manage. Bull., 8(1), 1820.
Winkler, R. D., D. L. Spittlehouse, and D. L. Golding (2005), Measured
differences in snow accumulation and melt among clearcut, juvenile,
and mature forests in southern British Columbia, Hydrol. Processes, 19,
5162.
Wolman, M. G., and J. P. Miller (1960), Magnitude and frequency of forces
in geomorphic processes, J. Geol., 68, 6474.
Wright, K., K. H. Sendek, R. H. Rice, and R. B. Thomas (1990), Logging
effects on streamow: Storm runoff at Casper Creek in northwestern
California, Water Resour. Res., 26, 16571667.
Zegre, N., A. E. Skaugset, N. A. Som, J. J. McDonnell, and L. M. Ganio
(2010), In lieu of the paired catchment approach: Hydrologic model
change detection at the catchment scale, Water Resour. Res., 46,
W11544, doi:10.1029/2009WR008601.
Ziemer, R. R. (1981), Storm ow response to road building and partial cutting
in small streams of northern California, Water Resour. Res., 17, 907917.
Y. Alila and P. K. Kuras, Department of Forest Resources Management,
University of British Columbia, 2045-2424 Main Mall, Vancouver, BC
V6T 1Z4, Canada. (peterkuras@gmail.com)
M. Weiler, Institute of Hydrology, University of Freiburg, Fahnenberg-
platz, D-79098 Freiburg, Germany.