Forest cover modifies snow accumulation and ablation rates via canopy interception and changes in sub-canopy energy balance
processes. However, the ways in which snowpacks are affected by forest canopy processes vary depending on climatic, topographic and
forest characteristics. Here we present results from a 4-year study of
snow–forest interactions in the Oregon Cascades. We
continuously monitored snow and meteorological variables at paired forested and open sites at three elevations representing the Low,
Mid, and High seasonal snow zones in the study region. On a monthly to bi-weekly basis, we surveyed snow depth and snow water
equivalent across 900
Snowpacks the world over are changing. Increasing global temperatures and accompanied climatic changes are altering snowpack characteristics and shifting melt timing earlier (McCabe and Clark, 2005; Mote, 2006; Mussleman et al., 2017). The timing, intensity, and duration of snowmelt depend on climatic and physiographic variables. In the topographically diverse western US the distribution of snow cover is governed by regional climate, elevation, vegetation presence/absence, and forest structure (Elder et al., 1998; Harpold et al., 2013). Forests overlap with mountains across this region and modify snow accumulation and ablation rates through canopy interception and a recasting of the sub-canopy energy balance (Hedstrom and Pomeroy, 1998; López-Moreno and Stähli, 2008; Varhola et al., 2010). Recently, a considerable amount of effort has been expended in research into the snow–forest processes that control the distribution of snow in mountainous regions (Stähli and Gustafsson, 2006; Jost et al., 2007; López-Moreno and Latron, 2008; Musselman et al., 2008; Ellis et al., 2013; Moeser et al., 2015). While these studies have focused on cold, predominately continental snowpacks, few have investigated snow–forest process interaction in warm maritime environments where snow is especially sensitive to changes in energy balance (Storck et al., 2002; Lundquist et al., 2013). Maritime snowpacks accumulate and reside at temperatures near the melting point. Such snowpacks do not fit the simple accumulation–ablation model of a monotonic increase until peak snow water equivalent (SWE) followed by a monotonic decrease to snow disappearance. Such temperature sensitive snowpacks may experience disproportionate effects of climate warming and changing forest cover (Nolin and Daly, 2006; Dickerson-Lange et al., 2015). Ramifications of these impacts have far reaching eco-hydrological impacts across the snowmelt dependent western US, highlighting the continued need for research into snow–forest process interactions in maritime montane settings (Mote, 2006; Harpold et al., 2015; Vose et al., 2016).
In the Pacific Northwest, United States (PNW), mountain environments are a disparate composite of forest cover driven by forest harvest, regrowth, and natural disturbance. Forest disturbance can have significant impacts on snow processes, whose effects can range from immediate (Boon, 2009) to decadal (Lyon et al., 2008; Gleason and Nolin, 2016). At the stand scale, forests attenuate wind speeds, thereby suppressing turbulent mixing of the near-surface atmosphere (Liston and Sturm, 1998); modify the radiation received at the snow surface through shifts in shortwave and longwave contributions and reduced surface albedo (Sicart et al., 2004; O'Halloran et al., 2012; Gleason et al., 2013); and temporally shift seasonal- and event-scale accumulation and ablation patterns through canopy snowfall interception (Varhola et al., 2010). Natural and anthropogenic alterations in forest cover such as mountain pine beetle infestation, forest management practices, and forest fire affect snow processes by modifying forest structure, i.e., canopy cover and gap size (Boon, 2009; Bewley et al., 2010; Ellis et al., 2013) and snow albedo (Gleason et al., 2013; Gleason and Nolin, 2016). The frequency and intensity of forest fires have been increasing (Westerling et al., 2006; Miller et al., 2009; Spracklen et al., 2009), impacting accumulation and ablation rates (Gleason et al., 2013), and are anticipated to continue increasing (Moritz et al., 2012; Westerling et al., 2011), while prolonged droughts, and a future of increasing drought prevalence, have increased water stress, creating changes in forest characteristics across the western US (Allen, 2010; Choat, 2012; Dai, 2013). Disturbances of this type alter the snow–forest dynamic through a modification of the magnitudes of central process relationships, often resulting in unanticipated outcomes (Lundquist et al., 2013). The present reality and specter of continued future change to climate and forest cover underscores the increasing importance of characterizing vegetation impacts on snow accumulation and ablation within warm, topographically varied terrains.
Elevation (as a proxy for temperature) and forest canopy cover are important
controls on peak snow accumulation (Geddes et al., 2005; Jost et al., 2007).
Elevation drives snow accumulation and is the principle predictor of peak
snow water equivalent (Gray, 1979; Elder et al., 1991; Sproles et al., 2013).
The partitioning of precipitation between rainfall and snowfall is determined
by atmospheric temperature and the elevation of the rain–snow transition can
be described as a function of the temperature lapse rate. Forest canopies
intercept snow, reducing sub-canopy accumulation (Schmidt and Gluns, 1991;
Hedstom and Pomeroy, 1998; Musselman et al., 2008). The magnitude and rate of
canopy interception are also affected by air temperature. Air temperature has
been shown to have an inverse relationship with canopy interception
(Andreadis et al., 2009) and a nonlinear correlation with event size
(Hedstrom and Pomeroy, 1998); these relationships are often based on a few
measurements and at a single point. Forests also reduce solar radiation
reaching the snowpack surface (Link and Marks, 1999; Hardy et al., 2004) and
increase longwave radiation at the snowpack surface (Lundquist et al., 2013),
thus modifying net radiation (Sicart et al., 2004). Forest cover reduces wind
speed, thereby reducing latent and sensible heat flux at the snowpack surface
(Link and Marks, 1999; Boon, 2009). The direct effect of wind speed on canopy
snow interception has not been explicitly studied, with most research
focusing on wind redistribution of snow (Gary, 1974; Pomeroy et al., 1997;
Liston and Sturm, 1998; Woods et al., 2006). Research demonstrates that
forests reduce wind speed and can lead to increased snow accumulation in
canopy gaps or forest clearcuts where wind speeds decline and snow is
released from upwind canopy flow (Gary, 1974). These combined forest effects
on sub-canopy energy and mass balance can accelerate or delay the onset and
rate of snowmelt (Varhola et al., 2010). These studies highlight the key
differences between forested and open areas, and the effects of elevation on
snowpack evolution. With strong agreement that the western US will be facing
warmer winters in the future and new understanding that snow in forested
regions is more sensitive to increased temperatures than snow in non-forested
regions (Lundquist et al., 2013), it is critical that we measure,
characterize, and understand maritime snow–forest interactions. This study
examines and evaluates the combined effects of forest cover, climate
variability, and elevation on snow accumulation and ablation in a maritime
montane environment. Specifically, we focus on the following research
questions.
To what extent do forests modify snow accumulation and ablation in a maritime temperate forest? How does canopy interception affect sub-canopy snowpack evolution across an
elevation gradient? How does forest cover affect the sub-canopy snow surface energy balance
relative to adjacent open areas and what are the principal drivers of melt?
In subsequent sections, we describe the study area; present research methods
for field measurements, energy balance calculations, and snow modeling;
present our key findings; and conclude with a description of potential
applications and future steps.
The Oregon ForEST network sites of the McKenzie River basin.
The McKenzie River basin (MRB) is part of the greater Willamette
River basin in western Oregon, USA (Fig. 1). It covers an area of
3041
The Oregon Forest Elevation Snow Transect (ForEST) network extends
from the rain–snow transition zone through the seasonal snow zone in the
Oregon Cascades with paired forested and open sites at three elevations, Low
(1150
At each of the six sites within the ForEST network tower-based instruments
continuously measured snow depth, incoming and reflected shortwave radiation,
air temperature, relative humidity, wind speed, wind direction, and soil
temperature and soil moisture (Table S1 in the Supplement). Sensor
measurement frequency was 15 s with output values averaged over
a 10
Site forest characteristics with the associated SD for each measurement.
Forest structure characteristics at each site were quantified using
ground-based conventional forest inventory methods. At transect locations
coinciding with SWE measurements, individual tree characteristics were
measured within each quadrat and averaged for that particular site, i.e.,
diameter at breast height (DBH), crown radius, tree height, and
tree species (Table 1). Forest density was performed using a plotless density
estimator approach described in Elzinga et al. (1998). The forest canopy at
each site was further characterized using skyward looking hemispherical
photographs acquired using a Nikon Coolpix 990 digital camera equipped with
a FC-E8 fisheye converter, which has a 180
During the snow accumulation period forest canopy plays a large role in
reducing snowpack by intercepting incoming snowfall, prohibiting
a significant portion from accumulating on the forest floor. A forest canopy
is the integrated sum of the forest overlaying the ground surface; this
includes needles, leaves, branches, and trunks. The canopy structure is the
primary control on canopy interception, followed by event-specific variables,
i.e., event size, air temperature, and wind speed (Varhola et al., 2010).
Canopy snow interception is inherently difficult to accurately quantify due
to the temporally sensitive impacts of local climate on the canopy itself and
the limited measurement capabilities to directly measure canopy interception
(Martin et al., 2013; Friesen et al., 2014). From measured snowfall at each
climate station within the ForEST network we calculated percent canopy
interception efficiency (
A snow surface energy balance was calculated at a daily time step using
aggregated 10
A critical component within the snow surface energy balance calculations is
the determination of the snow surface temperature,
Incoming and reflected solar radiation were each measured using an upward
facing and downward facing LI-200s
Longwave radiation is rarely directly measured in the seasonal snow zone due
to the high cost in both absolute, e.g., instrument cost, and relative terms,
e.g., energy requirements. Longwave radiation balance was calculated as
A variety of empirically derived formulas exist for calculating incoming
longwave radiation under clear (
Following Flerchinger et al. (2009) we performed a comparative analysis of
various longwave radiation algorithms and measured net longwave radiation.
Table S2 shows two clear sky algorithms and three cloud correction algorithms
used in the comparison, totalling six combinations in all, with the
“best-fit” algorithm determined by root mean squared error (RMSE). We
measured longwave radiation using a Huskeflux NR1 net radiometer during
spring 2013 for a 2-week period in a forested site within the MRB (Gleason
et al., 2013) and for a 10-day period in an adjacent open area, excluding
a 4-day period of rain. The NR1 measures four separate components of the
surface radiation balance, separately measuring incoming and reflected solar
radiation and both incoming and outgoing far infra-red radiation. The
pyrogeometers have a built-in Pt100 temperature sensor for calculation of
both the sky and surface temperature. Additionally, they are heated, with
temperature compensation, to avoid moisture build-up on the thermopile
sensors. The predicted incoming longwave radiation results of each method
were then compared to the NR1 measured incoming longwave radiation using RMSE
(Table S3). We found that the best approximation for incoming longwave energy
was the clear sky algorithm of Dilley and O'Brien (1998) combined with the
cloud adjustment of Crawford and Duchon (1999). The combined Crawford–Dilley
method was therefore used in all longwave calculations going forward and is
calculated as
The turbulent fluxes of latent and sensible heat are calculated using
indirect methods. Latent heat exchange was calculated using the method found
by Kustas et al. (1994):
The bulk aerodynamic approach is guided by stability conditions in the air
above the snow surface. The stability of the air column is determined by
application of the dimensionless bulk Richardson number
Average snow water equivalent (SWE) for the Open and Forest sites within the ForEST network, WY 2012–2014.
Values for 1 April SWE, as calculated from the NRCS SNOTEL stations, range from 9 % (WY 2015) to 139 % (WY 2012) of the 30-year median reference period (1981–2010). Snow surveys conducted at the Low and Mid elevation sites for WY 2012–14 show SWE at the Open site to be consistently greater and snow cover lasting longer into the spring than at the adjacent Forest site (Fig. 2). During the average snow year of WY 2013 (93 % of the 30-year median) the Low and Mid sites showed substantial differences between Open and Forest SWE throughout the accumulation and ablation seasons, whereas at the High sites SWE amounts were similar in Open and Forest. Conversely, snow lasted longer into the spring at the High Forest site relative to the High Open site. Because 1 April SWE may not accurately represent annual peak SWE at low and mid elevations within the PNW, we use the date of peak SWE in the following analysis. Therefore, peak SWE at the Low Open site was 209, 215, 225, and 242 % of the Forest site peak SWE, respectively, for WY 2012–WY 2015. Peak SWE at the Mid Open site was 200, 280, 328, and 302 % of the Forest site peak SWE, respectively, for WY 2012–WY 2015. However, SWE at the High Forest site is consistently higher than at the High Open site, 111, 103, 125, and 110 % for WY 2012–WY 2015, respectively.
Summary snow statistics for WY 2012–WY 2014 – Oregon ForEST network.
Excluding the historically low snowpack of WY 2015 (Sproles et al., 2017),
the 3-year average snow depth ablation rates in the Forest sites at Low and
Mid elevation were 1.3 and 1.2
Canopy interception depth vs. event snowfall within the ForEST network.
Results show that
Calculated daily mean energy balance in
To better understand the energy balance effect of forest canopies on snow accumulation and ablation, we calculated the mean daily energy balance components for the low and mid elevation sites for WY 2012–WY 2015 and for WY 2014 and WY 2015 for both high elevation sites (Fig. 4). Net radiation is the major component at all sites, while the turbulent fluxes and sensible and latent heat are only significant at the High Open site. Turbulent fluxes at all other sites are only episodically important and do not account for any significant amount of energy at the monthly or annual timescales. On an annual basis, shortwave radiation is the primary component of the energy balance at all Open sites, whereas longwave radiation dominates at all Forest sites. There is a strong dominance of shortwave (longwave) energy at the Low and Mid Open (Forest) sites, where it accounts for 89 and 71 % (93 and 92 %) of the average annual net energy balance, respectively. At the High sites this trend persists, although the magnitudes change. Within the High Forest site, shortwave radiation accounts for the majority of energy received at the snow surface, but the annual total is reduced by 53 %, with net longwave radiation accounting for 47 %. Conversely, at the High Open site solar radiation accounts for 71 % of the annual total, while longwave radiation is reduced to 7 %. The turbulent fluxes account for the remaining 22 %.
Calculated daily mean energy balance component magnitudes (bars) and
the daily measured snow depth (dashed line) for Mid Open
The stable atmospheric conditions at all sites, except the High Open site,
reduce the turbulent fluxes to consistently insignificant values at the daily
timescale, with only a few days over the course of the study period where
these fluxes persist (Fig. 4). Not surprising then is the importance of the
radiative fluxes for the net energy balance at all sites outside of the High
Open site. Longwave radiation dominates at the Low and Mid Forest sites
regardless of elevation or year (Figs. S1–S4 in the Supplement). Snowpack
melt response to the increased longwave radiation in the forest from lasting
events can be substantial. For example, at the Mid Forest site during an
8-day mid-January period, longwave radiation at the snow surface increased
71
Boxplot of average monthly air temperature for each site within the ForEST network, WY 2012–WY 2015.
Air temperature is a first-order control in longwave radiation calculation
and therefore it is expected that the lower and thus warmer sites will
experience a larger percentage of net radiation in the form of longwave
radiation. Average monthly air temperatures show that the High Forest site is
1.9 and 1.8
Daily average wind speed (heavy solid line) and the range of wind speeds (shaded area) for each site within the ForEST network, WY 2012–WY 2015.
Wind speeds at all sites except at the High Open site are relatively weak and
inconsistent, resulting in little turbulent mixing. Sustained (annual
average) wind speeds at the High Open site are over 5 times greater than at
any other site, with peak daily maxima more than 9 times greater (Fig. 7). At
the High Open site high wind speeds occur frequently, while all other sites
experience low wind speeds and little variability. Mean winter wind speed for
the High Open site is 3.6
Forest structure at the Low and Mid Forest sites is typified by average crown
diameters of 9.4 and 6.7
In maritime snow zones where winter precipitation is often a mix of rain and
snow, multiple mechanisms align to contradict the conventional wisdom that
snow is retained longer in forests than in open areas (Link and Marks, 1999;
Jost et al., 2007; Musselman et al., 2008). Multi-layered forest cover and
a relatively warm forest increase canopy interception efficiency, resulting
in significant reductions in sub-canopy snow accumulation (Storck et al.,
2002). While no significant relationship existed between daily air
temperature and
While few studies in maritime forested environments on the energy balance
exist, there is evidence of longwave radiation as the dominating term during
rain on snow (ROS) events within forests (Berris and Harr, 1987; Mazurkiewicz
et al., 2008; Garvelmann et al., 2014). Berris and Harr (1987) showed that
longwave radiation accounted for 38–88 % of all ROS event snowmelt.
Garvelmann et al. (2014) found that in two ROS events longwave
radiation accounted for 55.1 and
38.8 % of the net energy balance, although this may be biased low due to
the inability to accurately capture tree trunk temperature. Although
Mazurkiewicz et al. (2008) did not differentiate between radiation terms,
they found that net radiation was the largest contributor to melt. The highly
nonlinear relationship between air temperature and incoming longwave
radiation formulation is apparent in the net radiation budget analysis.
Infrequent cloud-free days and the warm, dense forests of the study area
combine to emit a significant amount of longwave radiation to the snow
surface (Berris and Harr, 1987; Sicart et al., 2004; Garvelmann et al.,
2014). This leads to a positive net snow surface energy balance and
mid-winter melt events, most pronounced at the warmer lower elevation sites.
With prolonged exposure to longwave radiation emitted by the canopy and the
high efficiency of warm forest canopy interception capabilities, low
elevation maritime sub-canopy snowpacks are relatively thin and do not
persist long enough into the spring season to benefit from forest shading.
This creates a radiative paradox where the longwave radiation emitted by
dense and relatively warm forest cover exceeds the resulting reduction in
shortwave radiation due to forest shading (Sicart et al., 2004; Lawler and
Link, 2011; Lundquist et al., 2013). The higher elevation sites experience
colder air temperatures, higher wind speeds, and lower forest density, which
combine to decrease
Here, we considered that wind may have an impact on canopy snow unloading and
subsequent increases in sub-canopy snow accumulation. While a seasonal mean
presents a general view of the wind environment at each Open site, it masks
the variability of wind gusts that can drive snow redistribution. Using the
10
The effects of elevation position within a watershed and forest structure on
snow persistence can have serious implications within a warming climate.
Sproles et al. (2013) documented a 150
This paper highlights the complex snow–forest process relationships and suggests that forest cover is a principal control on snow persistence due to reduced accumulation from canopy interception and earlier/faster melt due to increased longwave radiation. High density, relatively warm forests have high canopy interception efficiency that controls sub-canopy snowpack evolution and mediates the amount of springtime solar shading of the snowpack. The cooler and less dense High Forest site has a reduced interception efficiency and acts as a snow deposition reservoir for the nearby windy High Open site. Net radiation drives the snow surface energy balance, with the partitioning between longwave and shortwave a function of forest complexity. Our study demonstrates the sensitivity of Pacific Northwest snowpack development to temperature and forest cover. Nolin and Daly (2006) demonstrated that much of the Oregon Cascade snowpack is at risk, the ForEST network included, by looking at temperature only. Similarly, Sproles et al., 2013 showed that the lower boundary of the snow zone has little resilience to a warming world. Our paper demonstrates that understanding the snowpack energy budget is key to understanding how forests influence snow accumulation and melt. By quantifying the mechanisms of how vegetation affects sub-canopy snowpack energy balance, the results of this study provide the basis for understanding the sensitivity of maritime snowpacks to a changing climate. As climate continues to warm, we anticipate reduced snow accumulation at elevations where snowfall shifts to a rain–snow mix, and amplified sub-canopy melt rates due to longwave radiative heating in warmer forests, thereby reducing overall forest snow retention. However, higher elevation colder sites with a less dense forest can mitigate that to some extent by retaining the snowpack longer through lower relative forest longwave emission and lower canopy interception. A key finding within this study is that throughout the study duration, one that saw high inter-annual snowfall variability, a definitive pattern emerged within the energy budget and snowpack dynamics across the network. The energy budget format that we present here goes beyond the temperature only approach while getting at the causal effects and mechanisms of the challenge of vegetation–snowpack interactions for a warming climate.
While these results are focused on the Oregon Cascades, they have broader implications for other relatively warm forested snow environments with elevation gradients, such as parts of the California Sierra Nevada, the Japanese and European Alps, and the Pyrenees (Lundquist et al., 2013). These results will aid in improving parameterizations of snow–forest interactions in physically based snow hydrology models and land surface models. Additionally, as climate change alters regional snow deposition patterns across the western US, our findings are applicable to land and water managers, seeking to improve forest snowpack retention, enhance forest health, and improve streamflow forecasting. This study demonstrates the value of plot-scale snow–forest process studies for improving our understanding of the forest effects on snowpack evolution. Future work will focus on a multi-scale approach that incorporates remote sensing and snow hydrology modeling to identify forest structure metrics that are well suited to accurately modeling snow–forest interactions. Such an approach will allow the snow community to quantify the improvement of snow–forest interactions across spatial scales and enhance model prediction for landscape and regional applications.
The data used in this study are freely available online
from the Oregon State University ScholarsArchive:
The authors declare that they have no conflict of interest.
This research was made possible by funding provided by the National Science Foundation (EAR 1039192) and from a NASA Earth Science Student Fellowship (16-EARTH16F-0426). We thank Willamette National Forest for providing access permits for the ForEST network. Additional material support was provided by the Western Ecology Division office of the Environmental Protection Agency, with special thanks to Ron Waschmann. We thank the many student interns who assisted in snow surveys and site maintenance. Edited by: Jan Seibert Reviewed by: three anonymous referees