Snowpack influences spatial and temporal soil nitrogen dynamics
in a western U.S. montane forested watershed
Y 1,URIKO YANO , 
 C Q ,1 Z H ,2 2 3LAIRE UBAIN ACH OLYMAN KELSEY JENCSO, AND JIA HU
1Department of Ecology, Montana State University, 310 Lewis Hall, Bozeman, Montana 59717 USA
2Department of Forest Management, University of Montana, 32 Campus Drive, Missoula, Montana 59812 USA
3School of Natural Resources and the Environment, University of Arizona, 1064 East Lowell Street, Tucson, Arizona 85712 USA
Citation: Yano, Y., C. Qubain, Z. Holyman, K. Jencso, and J. Hu. 2019. Snowpack influences spatial and temporal soil
nitrogen dynamics in a western U.S. montane forested watershed. Ecosphere 10(7):e02794. 10.1002/ecs2.2794
Abstract. Declines in winter snowpack have increased the severity of summer droughts in western U.S.
forests, with the potential to also impact soil available nitrogen (N). To understand how snowpack controls
spatiotemporal N availability, we examined seasonal N dynamics across elevation, aspect, and topographic
position (hollow vs. slope) in a forested watershed in the northern Rocky Mountains. As expected, peak
snow-water equivalent (SWE) was generally greater at higher elevations and on north-facing aspects. How-
ever, the effects of topographic position and snowdrift led to variability in snow accumulation at smaller
spatial scales. Spatial patterns of the snowpack, in turn, influenced soil moisture and temperature, with
greater SWE leading to generally higher soil moisture levels during the summer and smaller temperature
fluctuations throughout the year. Wetter conditions in spring or fall generally supported greater inorganic
N pools, but at the driest locations (low-elevation slope), pulses of N mineralization in summer may have
played important roles in overall N dynamics. More importantly, soil moisture during the summer
appeared to be more influenced by antecedent snowpack from the previous year than by current-year sum-
mer rain. Subsequently, N mineralization under snowpack may be strongly influenced by soil moisture
and temperature conditions from the previous fall, before snowpack accumulation. Together, our results
indicate that snowpack strongly influences N dynamics beyond the current growing season in western
coniferous forests through mediation of soil moisture and temperature, and suggest that further decline in
winter snowpack may affect these forests through constraints in both water and N availability.
Key words: antecedent snow effect; conifer forest; nitrogen availability; nitrogen cycling; snowpack decline; snow-water
equivalent.
Received 11 January 2019; revised 25 May 2019; accepted 31 May 2019. Corresponding Editor: Debra P. C. Peters.
Copyright: © 2019 The Authors. This is an open access article under the terms of the Creative Commons Attribution
License, which permits use, distribution and reproduction in any medium, provided the original work is properly cited.

 E-mail: yuriko.yano@montana.edu
INTRODUCTION coupled through biological N transformation
and leaching (Schimel et al. 1997, Wang et al.
In temperate forests, nitrogen (N) is often con- 2015). Furthermore, because water availability in
sidered to be the most common limiting nutrient these ecosystems largely depends on snowmelt
(Vitousek and Howarth 1991), and many water, spatial distribution and inter-annual varia-
experimental studies demonstrate increases in tions in snowpack may indirectly control N
aboveground productivity with increased N availability during the growing season. For
availability (summarized in LeBauer and Trese- example, the amount of snowpack and the tim-
der 2008). In semi-arid western U.S. conifer ing of spring snowmelt collectively affect proxi-
forests, water and N likely co-limit tree growth mate soil moisture during the snow-free period
because the water and N cycles are tightly (Blankinship et al. 2014, Maurer and Bowling
 ❖ www.esajournals.org 1 July 2019 ❖ Volume 10(7) ❖ Article e02794
YANO ET AL.
2014, Harpold and Molotch 2015). While this throughout the summer and fall, and then the
suggests that snowpack declines and earlier following year snowpack forms over these wet
snowmelt under warmer climate can alter sea- soils, the insulative properties of a deep snow-
sonal N availability, we currently lack a basic pack may enhance microbial activity. Conversely,
understanding of N dynamics at the landscape if winter snowpack forms over relatively dry
scale to evaluate these potential outcomes. soils, the insulative properties of a deep snow-
Snowpack controls N dynamics through multi- pack may not be as important for microbial activ-
ple abiotic and biotic processes, which have been ity, because soil moisture levels may be too low.
identified by many plot-scale snow-manipula- However, the antecedent soil moisture condi-
tion studies thus far. Deep and continuous snow- tions prior to snowpack formation and its influ-
pack can insulate soils from cold air ence on overwinter microbial N dynamics have
temperatures and increase N availability through not received much attention. Furthermore, in
sustained microbial N transformation through- many landscapes across the western United
out winter. However, deeper snowpack can also States, complex terrain can lead to heterogeneity
lead to higher gaseous N losses through denitrifi- in snowpack depth, leading to variability in soil
cation (Williams et al. 1998, Brooks et al. 2011). moisture and N dynamics over small spatial
On the other hand, shallow and discontinuous scales.
snowpack often results in freezing soils and Topography controls the distribution of
reduced microbial activities (Schimel et al. 2004, organic matter, soil particle size, and water in
Brooks et al. 2011, Duran et al. 2014). Reduced many ways, influencing spatial and temporal
microbial activities subsequently lower N avail- variations of biogeochemical processes (McClain
ability due to decreased microbial N mineraliza- et al. 2003, Bernhardt et al. 2017). For example,
tion (Schimel et al. 2004, Groffman et al. 2009) or soil depth, soil texture, organic matter, and nutri-
increase nitrate leaching due to decreased N ent pools all vary across topographic positions
immobilization (Brooks et al. 1998, Fitzhugh (Moore et al. 1993, Amundson and Jenny 1997,
et al. 2001). Most of our current knowledge is Hook and Burke 2000), which in turn affect water
derived from plot-scale studies that have focused and N availability. Topography also affects snow
on the insulative effect of snowpack on N accumulation through controls on energy inputs
dynamics mediated by soil microbes because low (i.e., solar radiation, temperature) and snowdrift
soil temperature is thought to be the main factor patterns (Elder et al. 1991, Luce et al. 1998,
limiting microbial activity in many mesic snow- Anderton et al. 2004). Furthermore, topography
dominated ecosystems (Brooks et al. 1998, 2011, influences the movement of snowmelt water
Schimel et al. 2004, Groffman et al. 2009, Duran within both the vadose and bedrock layers
et al. 2014). (McGlynn et al. 1999). While studies on the eco-
In the semi-arid western United States, interac- hydrology of montane ecosystems have greatly
tions between snowpack and N dynamics are advanced in recent years (McGlynn and McDon-
more complicated, because snowpack controls nell 2003, Jencso et al. 2009), fewer studies have
both soil temperature and moisture (Molotch extended the linkages between topography and
et al. 2009, Brooks et al. 2011, Maurer and Bowl- water availability to include nutrient dynamics
ing 2014, Barnhart et al. 2016). Soil moisture con- (Powers 1990, Griffiths et al. 2009). It is still
tent starts high during snowmelt and then unclear how variations in snowpack across space
decreases during the summer and into early fall and time would affect N availability during the
(Blankinship et al. 2014, Maurer and Bowling growing season in semi-arid ecosystems.
2014). Low soil moisture during this period can In this study, we took advantage of a topo-
lead to limitations on soil microbial activity graphic gradient within a watershed, with vari-
(Johnson et al. 2009). During the fall, however, ability in snow cover, soil moisture, and soil
precipitation may increase soil water content temperature, to examine the hydroclimatic
prior to snowpack formation, creating a favor- impacts on spatial and temporal N availability in
able environment for overwinter N dynamics a western montane forest. Our specific questions
(Maurer and Bowling 2014). Alternatively, if a were as follows: (1) How does topography (ele-
large snowpack leads to relatively wet soils vation, aspect, and hillslope position) influence
 ❖ www.esajournals.org 2 July 2019 ❖ Volume 10(7) ❖ Article e02794
YANO ET AL.
soil water and temperature both spatially and while high-elevation sites receive 664 mm
temporally? (2) How do shifts in soil moisture across a 30-yr normal (NRCS SNOTEL, 1981–
and temperature affect N dynamics? And more 2010). The average temperature is 4.2°C and
broadly, (3) how do inter-annual variations in 3.0°C (1981–2010) for low and high elevation,
snowpack influence N dynamics in water-limited respectively. Atmospheric N-deposition rate in
ecosystems? We hypothesized that topography the area is close to the global background level
and variations in snowpack together influence and is estimated to be 1.3 kg-Nha
1yr
1 (Saros
temporal and spatial soil temperature and water et al. 2005).
availability, which in turn affect N dynamics. We Trees across the even-aged forest are approxi-
also hypothesized that antecedent soil moisture mately 80 yr old and naturally regenerated from
would play an important role in N dynamics logging in the early 1900s. The dominant tree
under snowpack. Collectively, these three ques- species were Douglas fir (Pseudotsuga menziesii),
tions are critical to advance our understanding of Engelmann spruce (Picea engelmannii), subalpine
the spatiotemporal dynamics of soil N and water fir (Abies lasiocarpa), ponderosa pine (Pinus pon-
availability in semi-arid ecosystems. derosa), and western larch (Larix occidentalis).
Nitrogen-fixing alder (Alnus vividis or incana)
METHODS was present only at the hollow positions of high-
elevation sites.
Study sites The underlying lithology of the watershed is
The study site was located at the Lubrecht predominantly composed of quartz monzonite,
Experimental Forest, Montana, USA (47° N, 113° with the periphery of the watershed composed of
W), within the North Fork Elk Creek watershed. Mesoproterozoic metasedimentary mudstones
The typical climate in this region is characterized and fine-grained sandstones (Belt Supergroup).
by cold and snowy winters followed by warm Soil textures within the study area ranged from
and relatively dry summers. Since 1970, mean gravely loamy sands to well-drained silty loams,
annual precipitation is 650 mm and snowfall although soils were quite heterogeneous even
constitutes 24% and 41% of annual precipitation within sites (NRCS 2017). The soil was neutral
at 1426 m a.s.l. and 1905 m a.s.l., respectively with a mean soil pH of 6.3  0.12 (mean  1
(NRCS SNOTEL, stations 604 and 657). standard error [SE]) for the top 15 cm. The high-
We established four sites, each consisting of a elevation sites had higher organic matter concen-
small zero-order catchment. The sites were laid trations (21–24% carbon; C) than the lower eleva-
out across a gradient that contained contrasting tion sites (5–8% C).
elevations, aspects, and slope–hollow positions At each site, we recorded air temperature and
(slope represented areas of high divergence and relative humidity at 30-min intervals, using
hollow represented areas of high convergence). EM50 dataloggers (METER, Pullman, Washing-
We established two north-facing and two south- ton, USA) and a VP3 sensor (METER). Precipita-
facing sites at high (~1800 m) and low (~1280 m) tion was recorded at the same frequency using a
elevation, resulting in four total sites: North-low tipping bucket rain gauge (ECH2O ECRN-100;
(46.876317,
113.3448023), North-high (46.8863188, METER). Within each site, we also recorded soil

113.2975332), South-low (46.8821088,
113.3444993), temperature and volumetric water content at 30-
and South-high (46.8909935, 
113.2971493). The min intervals at 10, 30, and 50 cm depths, using
main goal for this site design was to not treat 5TE sensors (METER). At each site, both the hol-
each site as a categorical variable; instead, each low and slope landscape positions were instru-
site represented a set of conditions across a range mented as shown in the meteorological stations
of moisture and temperature conditions (e.g., hot in Fig. 1.
and dry sites to cool and wet sites). Within each
site, we established 25 plots along five evenly Sampling and analyses
spaced transects (five plots per transect), which We measured the variability in snowpack dis-
ran across the catchment, intersecting the hollow tribution across the watershed by measuring
(Fig. 1). On average, the low-elevation sites snow-water equivalent (SWE) within each site
receive 514 mm of accumulated precipitation, during near peak SWE (hereafter “peak SWE”) in
 ❖ www.esajournals.org 3 July 2019 ❖ Volume 10(7) ❖ Article e02794
YANO ET AL.
Fig. 1. Layout of sampling plots within each site in the North Fork of the Elk Creek Watershed, Montana,
USA. The four sites include north-facing low-elevation (N-low), south-facing low-elevation (S-low), north-facing
high elevation (N-high), and south-facing high elevation (S-low). Within each site, soil samples were collected
within five plots that ran along a transect that extended from slope to hollow back to slope. Five total transects
were established at each site for a total of 25 plots at each site.
late February 2017. The peak SWE of the Center for Stable Isotope Biogeochemistry (CSIB)
2016–2017 water year was generally similar to on the University of California, Berkeley.
the previous two winters, but the snowpack To examine seasonal N dynamics in the soil,
formed later and was preceded by a relatively we first determined extractable inorganic N
wet fall (Table 1). At each site, we recorded pools across our study sites. In 2015 and 2016,
the depth and weight of 15 snow cores, five cores we collected soil cores (15 cm depth) from each
at hollow positions and 10 at slope positions, of the 25 plots at each of the four sites, starting
totaling 60 cores (15 cores 9 4 sites). from near the end of spring snowmelt to the end
After the snowpack melted in June, we mea- of the snow-free season. Soil cores were collected
sured total C and N in bulk soil by collecting 25 approximately monthly and transported on ice
soil samples per site from the top 15 cm within and then homogenized after live plant material
each site (Fig. 1). Soils were dried at 60°C for and rocks (>2 mm) were removed. Soils were
≥48 h and ground to flour-like consistency, and homogenized and extracted with 1 M KCl within
soil %C and soil %N were determined at the 48 h of soil collection (soil: KCl = 1:4 by weight),
 ❖ www.esajournals.org 4 July 2019 ❖ Volume 10(7) ❖ Article e02794
YANO ET AL.
Table 1. Precipitation during snow-covered and snow-free seasons in during study period.
Precipitation (mm)†
Percentage growing Start date of
Elevation Water year Peak SWE (mm) Fall Growing season Total season‡ snow-free season
Low 2014–2015 155 69 157 437 36 March 20
2015–2016 97 69 246 523 47 March 26
2016–2017 163 127 . . . . . . . . . April 1
High 2014–2015 257 48 84 411 20 May 4
2015–2016 259 56 277 630 44 May 3
2016–2017 249 114 . . . . . . . . . May 21
Notes: SWE, snow-water equivalent. Data were recorded at the low (1426 m)- and high (1905 m)-elevation SNOTEL
stations. An ellipsis indicates no data available.
†Fall precipitation was estimated as a total precipitation between October 1 and the first day of winter snowpack formation.
Growing season precipitation was estimated as a total precipitation between the first day of snow-free day in spring and
September 30.
‡Percent growing season precipitation of total annual precipitation.
and ammonium (NH4-N) and nitrate (NO3-N) we verified these values by measuring net
were determined using a Lachat Quickchem mineralization and nitrification rates during the
Autoanalyzer (Lachat Instruments, Loveland, 2016–2017 winter (Nadelhoffer et al. 1985). In
Colorado, USA). Subsamples of the homoge- November 2016, before the snowpack developed,
nized soils were dried at 60°C for ≥48 h, and we cored soils (0–15 cm) from the same resin-
gravimetric soil moisture was determined. probe plots. Half the cores were immediately
While extractable N provides information on extracted for NH4-N and NO3-N, and the other
the size of available N pools at the time of sam- halves were placed into a plastic bag and buried
pling (i.e., potential N availability in soils), it back in the soil. These buried bags were retrieved
does not provide N transformation rates (i.e., in spring, and NH4-N and NO3-N were extracted.
realized N availability in soils). To estimate the Net mineralization and nitrification rates were
supply rates of available N in soils, we measured calculated by subtracting the final NH4-N and
NH +4 and NO


3 ions captured by ion-exchange NO3-N values from the values of the initial soils.
resin probes. At each of the same four sites, we In June 2015, to examine the influence of SWE
installed 24 pairs of cation- and anion-exchange on NO3-N leaching within the watershed, we
resin membrane probes (Western Ag, Saskatoon, installed six tension lysimeters (Prenart Equip-
Saskatchewan, Canada): 12 pairs across three ment, Frederiksberg, Denmark) per site in the
plots on the slope and 12 pairs across three plots same plots corresponding to the resin probes (a
in the hollow (4 pairs per plot, 6 plots per site, total of 24 lysimeters across the watershed;
Fig. 1). The resin probes were inserted into the Fig. 1). The lysimeters were installed at 30 cm
soils, at a ~45° angle such that the resin surfaces deep and ~30° angle. Soil water was collected by
were in contact with the soil at approximately applying 50 psi vacuum pressure; the lysimeters
10–15 cm deep. The probes were deployed and were purged by discarding first collections. A
retrieved approximately monthly during the sufficient amount of soil water for analysis could
snow-free season in 2015 and then left in the field be collected only during snowmelt and a short
over the 2015–2016 winter. By calculating the period following the snowmelt. Lysimeter water
amount of resin-exchangeable NH4-N and NO3- samples were retrieved within 2 weeks of vac-
N over a given period, we were able to compare uum application, when the sites were not accessi-
available N supply rates in the soil across the ble by motorized vehicles and the soil
landscape. The probes were pooled by plot (three temperature remained <5°C, and within 48 h
plots on hollow and three plots on slope per site) otherwise. Stream samples were collected from
and analyzed by Western Ag for resin-exchange- the two high-elevation sites at the time of lysime-
able NH4-N and NO3-N. ter sample retrieval, where a first-order stream
Because NH4-N and NO3-N supply rates dur- runs in the hollow year-round. The water sam-
ing 2015–2016 winter were lower than expected, ples were kept cold, while they were transported
 ❖ www.esajournals.org 5 July 2019 ❖ Volume 10(7) ❖ Article e02794
YANO ET AL.
to the laboratory, stored frozen after filtered elevation were not included in the LME regres-
through acid-washed grass fiber filters (0.7 lm sion analysis because of their associations with
pore size), and then filtered through membrane soil temperature and moisture. To meet the nor-
filters (0.2 lm pore size) prior to chemical analy- mal distribution assumption, the response vari-
sis NO3-N using a continuous segmented flow ables were natural log-transformed before the
analyzer (SEAL QuAAtro, Seal Analytical, regression analysis. Due to the high spatial
Mequon, Wisconsin, USA). heterogeneity of soil moisture within a watershed
and even within one site, we did not use the con-
Statistical analysis tinuous soil moisture values collected from the
To compare SWE, total soil C and N, and dataloggers at each of the four sites for the LME
resin-exchangeable N between elevation and regression analysis. Instead, for a more accurate
landscape position, we used ANOVA for a given estimate, we used the gravimetrically measured
sampling date, followed by Tukey post-hoc test. soil moisture values from each corresponding
When necessary, log transformation was used to soil core. All statistical tests were conducted
accomplish the normal distribution and equal using R (R Core Team 2013), and statistical differ-
variance before ANOVA. Net N transformation ences were determined at P < 0.05.
rates were compared using the non-parametric To assess the representative soil conditions of
Kruskal–Wallis test followed by Dunn test. each of the four sites around the time of snow-
To compare extractable N across time and to pack formation and disappearance, we used con-
have the data be placed in an ecologically mean- tinuous data collected from the dataloggers at
ingful context, we plotted our time series data to each site. We calculated mean soil temperature
account for differences in the timing of the snow- and moisture over two-week periods before and
melt. Because snowpack disappeared an entire after the first day of snowpack formation in the
month earlier from the low-elevation sites than fall and before and after the first snow-free day
the high-elevation sites, and from the slope than in the spring for each year sampled.
the hollow locations, we used number of days
since the first snow-free day instead of day of RESULTS
year. This allowed us to compare seasonal pat-
terns of soil N availability across a range of Soil characteristics
hydroclimatic conditions from the four sites. We Soil %C varied across the watershed, with the
approximated the first snow-free day by the time highest total %C occurring at the high-elevation
when we first observed a quick rise in soil tem- hollow sites (32.0%  5.6%, mean  1 SE), fol-
perature with diurnal fluctuations (Johnson et al. lowed by the high-elevation slope sites
2009) at 10 cm depth following an isothermal (16.3%  1.5%). Both low-elevation sites (hollow
period over winter from each datalogger location and slope) had the lowest %C values (7.9  0.9
(hollow and slope, Fig. 1). For the hollow loca- for hollow and 6.6%  1.0% for slope; Fig. 2).
tions in the N-high site, where data were not Spatial patterns of soil %N were similar to %C,
available, the first snow-free day was approxi- with the highest total N in the hollows at high-
mated by that of S-high hollow. elevation sites (1.06%  0.18%), intermediate N on
To determine the environmental factors that the slopes at high-elevation sites (0.52%  0.05%),
influence extractable N, we used linear mixed- and the lowest N at low-elevation sites (0.39% 
effects (LME) regression with step-wise model 0.03% for hollow and 0.27%  0.04% for slope;
selection, using Akaike information criterion Fig. 2). Soil bulk density was higher at the
(AIC) and nlme package in R (R Core Team low-elevation sites than the high-elevation sites.
2013). When there were multiple models with
identical AIC scores, a model with fewer insignif- Topographic influence on snowpack, soil moisture,
icant parameters were chosen. The fixed effects and soil temperature
used in the LME regression included: soil tem- Elevation, aspects, and topographic position
perature, soil moisture, soil %C, and soil %N. (slope vs. hollow) interacted to create a complex
Our sampling plots (25 plots per site, Fig. 1) were distribution of snowpack across our watershed,
used as a random effect in the model. Aspect and and this complexity influenced the onset of the
 ❖ www.esajournals.org 6 July 2019 ❖ Volume 10(7) ❖ Article e02794
YANO ET AL.
Fig. 3. Near peak snow-water equivalent (SWE) col-
lected in February 2017. The labels are as follows:
north-facing low elevation (N-low), south-facing low
elevation (S-low), north-facing high elevation (N-high);
south-facing high elevation (S-low). Snow cores were
collected from the hollow (open boxes, n = 5 per site)
and slope (filled boxes, n = 10 per site). See Fig. 2
legend for the boxplot parameters. The lowercase let-
ters indicate significant differences at P < 0.05 found
by ANOVA followed by Tukey test.
additional factor influencing snow accumulation,
where large snowdrifts regularly accumulated
Fig. 2. Soil characteristics across the topographic on both hollow and slope at the South-high site
gradient, including soil %C (A), soil %N (B), and soil (Fig. 3). Generally, locations with high peak SWE
bulk density (C). Labels are as follows: low-hollow (low (e.g., high-elevation hollows) were associated
elevation, hollow landscape position), low-slope (low with later snow disappearance and higher spring
elevation, slope), high-hollow (high elevation, hollow), soil moisture, while locations with low peak
and high-slope (high elevation, slope). The lower and SWE (e.g., low-elevation slope) were associated
upper sides of the boxes represent the first and third with earlier snow disappearance and lower
quantiles; the thick line within each box represents the spring soil moisture (Figs. 3, 4).
median, and the whiskers indicate the minimum and We evaluated how differences in topography
maximum values. The lowercase letters indicate signifi- and snowpack distribution across the watershed
cant differences at P < 0.05 by ANOVA following influenced soil temperature and moisture, and
Tukey’s test. The number of samples is (from left to found greater differences between low- and
right on the x-axis) 9, 37, 9, 36 for %C (A); 7, 27, 9, 36 for high-elevation sites in the hollow landscape posi-
%N (B); and 116, 445, 115, 445 for bulk density (C). tions compared to slope landscape positions. For
example, the higher-elevation hollow sites (high
snow-free season. Although elevation and peak SWE) entered the snow-free season at ~2°C
aspects were generally good predictors for peak higher soil temperature than low-elevation hol-
SWE, as expected, this was true only for the hol- low sites (low peak SWE), despite the one-month
low positions (Fig. 3). By contrast, peak SWE delay in the onset of the snow-free season at high
was generally low at slope positions, regardless elevation (Fig. 4A, I). The soil temperature at the
of elevation or aspect. Snowdrift was an high-elevation hollow also remained elevated for
 ❖ www.esajournals.org 7 July 2019 ❖ Volume 10(7) ❖ Article e02794
YANO ET AL.
Fig. 4. Seasonal patterns in soil temperature, moisture, extractable NH4-N, and NO3-N in 2015 (A–H) and
2016 (I–P). Hollow positions are shown in the left column, and slope positions are shown in the right column.
The x-axes are adjusted for the first snow-free day; the dotted lines indicate the date of first snow-free date. Some
sampling dates are provided in the panels as references. Soil temperature (A, B, I, J) and moisture (C, D, K, L)
were monitored at 10 cm depth. The NH4-N (E, F, M, N) and NO3-N concentrations (G, H, O, P) (mean  1 stan-
dard error) were for the top 15 cm, and the number of samples for each NH4-N and NO3-N data point is as fol-
lows: n = 5 for hollow, and n = 20 for slope.
 ❖ www.esajournals.org 8 July 2019 ❖ Volume 10(7) ❖ Article e02794
YANO ET AL.
two to four months of the snow-free season (up concentrations in 2015 (1 SE) for NH4-N and
to ~5°C). However, by mid-summer (July– NO3-N were 15.5  1.8 and 3.4  0.5 mg-N/kg,
August), soil temperatures in the low-elevation respectively, whereas mean concentrations in
hollows surpassed those at the high-elevation 2016 were 7.1  1.0 and 1.4  0.3 mg-N/kg,
sites and remained higher for the rest of the sea- respectively. In contrast, the inter-annual differ-
son, with the South-low site recording the high- ences in concentrations were much smaller at the
est temperature. Unlike the hollow locations, the driest locations (low-elevation slope); mean con-
effects of elevation and aspect on soil tempera- centrations in 2015 (1 SE) for NH4-N and NO3-
ture were minimal on the slope locations, reach- N were 2.5  0.1 and 1.0  0.1 mg-N/kg,
ing similarly high temperatures at both the low- respectively, whereas mean concentrations in
and high-elevation sites (Fig. 4B, J). 2016 were 1.8  0.1 and 0.3  0.02 mg-N/kg,
Soil moisture was higher for the high-elevation respectively.
sites and the hollows for almost the entire snow- We evaluated the interactions among environ-
free season, and similar to the case for the tem- mental factors (soil temperature and moisture)
perature, the differences across sites were smaller and soil C and N pools that drive these differ-
on the slopes than in the hollows (Fig. 4C, D, K, ences in N availability across the watershed,
L). The overall differences in the early-season soil using LME regression with step-wise model
moisture levels generally reflected SWE leading selection. We found that the soil moisture and
into spring (Fig. 3); sites with the highest SWE N pools, either alone or together, were impor-
values had higher soil moisture values at the tant factors influencing extractable NH4-N and
start of the snow-free season (>40% w/w), while NO3-N concentrations. However, the impor-
sites with the lowest SWE had low soil moisture tance and interactions of soil moisture and N
(~20% w/w), even at the start of the snow-free pools shifted greatly with soil temperature,
season (Figs. 3, 4C, D, K, L). Soils also became resulting in relatively low explanatory power
progressively drier mid-summer at all four sites when all the data points were fitted to one
(Fig. 4C, D, K, L), despite some summer rain model (Table 2).
events (Table 1). Soil moisture rose only after To better understand the relationships among
considerable early fall rains in 2016 (56 mm total) soil temperature, moisture, and C and N pools,
at the high-elevation sites (Fig. 4K, L); average we fitted LME regression models separately by
October soil moisture across all sites remained four different soil temperature brackets (≤5°C,
low in 2015 (6.8%  1.2%, v/v) but increased 5–10°C, 10–15°C, and >15°C; Table 2, Fig. 5). We
considerably in 2016 (14.2%  2.2%, data not found relatively strong and simple relationships
shown). among extractable N and soil environmental fac-
tors when the soils were cold (<5°C, Table 2);
Extractable N pools extractable NH4-N was positively influenced by
Across all four sites, extractable inorganic N both soil moisture and %N (R2 = 0.507,
concentrations changed across both space and P < 0.0001), whereas extractable NO3-N was
time, with a strong influence from topographic strongly influenced solely by soil moisture
position (hollow vs. slope). Generally, soil NH4- (R
2 = 0.682, P < 0.0001). However, these strong
N and NO3-N were highest at the high-elevation relationships disappeared when soil transitioned
hollow locations throughout the two-year study between cold (<5°C) and warm (5–10°C, Table 2).
period. By contrast, the concentrations were At higher temperature ranges (>10°C), soil mois-
consistently low throughout the year across all ture and %N continued to explain more than
the slope locations (high and low elevation; half of the variability for NH4-N (R
2 > 0.586,
Fig. 4E–H, M-P). In addition to the seasonal vari- P < 0.0001), but these factors alone could not
ations within a year, we also found inter-annual sufficiently explain much of the variability for
differences in extractable N concentrations. Both NO3-N (Table 2). Generally, soil C content was
NH4-N and NO3-N concentrations were higher only marginally important for NH4-N and was
in 2015 compared to 2016 (P < 0.001), with the little important for NO3-N.
largest differences occurring in the wettest loca- While soil temperature and C can both influ-
tions; at high-elevation hollow locations, mean ence N pools, our results suggest that soil
 ❖ www.esajournals.org 9 July 2019 ❖ Volume 10(7) ❖ Article e02794
YANO ET AL.
Table 2. Linear mixed-effects regression for the natural log of NH4-N and NO3-N.
Temperature brackets Model n R2 †
lnNH4-N (mg/kg)
All 
0.03*T + 0.86*W + 0.01C 
 0.75*N 
 0.11*T:N 
 0.12*W:C + 4.76*W:N + 0.62 760 0.363
<5°C 1.90*W 
 0.03C + 1.51*N + 0.39 131 0.507
5–10°C 
0.13*T + 1.03*N + 1.43 252 0.204
10–15°C 
0.03T 
 4.00*W + 1.60*N + 0.44T:W + 0.22 189 0.597
>15°C 7.29*W + 0.15C 
 2.39*N 
 0.90*W:C + 22.00*W:N 
 0.38 187 0.586
lnNO3-N (mg/kg)
All 
0.07*T + 0.90*N 
 0.47 760 0.099
<5°C 2.72*W 
 0.22 130 0.682
5–10°C 0.30T + 6.42*W + 0.01*N 
 1.37*T:W + 3.96*N:W 
 2.93 252 0.185
10–15°C 
0.08*T 
 33.45*W + 18.06*N 
 1.04*T:N + 2.10*T:W + 0.05 187 0.273
>15°C 
0.08*T 
 33.45W + 18.06*N 
 1.04T:N + 2.10*T:W + 0.05 187 0.159
Notes: Models are fitted for all data or by the soil temperature brackets, which correspond to those in Fig. 5. The abbrevia-
tions of the factors are T, soil temperature at 10 cm; W, soil moisture content (w/w); C, bulk soil C concentration; N, bulk soil N
concentration. The colons in the models indicate interactions of the factors. All models were statistically significant
(P-values < 0.001). The bold letters indicate significant parameters (P < 0.05).
†Conditional R2 (Lefcheck 2016).
moisture was one of the strongest drivers of soil moisture on extractable N concentrations across
extractable NH4-N and NO3-N. For example, our watershed.
when soil moisture was <15% w/w, both NH4-N
and NO3-N concentrations were consistently N supply rates and N leaching
lower across all temperature brackets than when Nitrogen supply rates, determined as resin-
soil moisture was >15% w/w (Fig. 5); the overall exchangeable NH4-N and NO3-N, differed across
means for NH4-N and NO3-N when soil mois- the landscape and between years. Contrary to
ture was <15% were two-thirds (4.6 mg/kg) and our expectations, NH4-N supply rates were low-
about one half (0.7 mg/kg) less, respectively, than est during the winter, even under a deep snow-
when soil moisture was >15%. In other words, pack at high-elevation hollows (Fig. 6A, B). The
when soil moisture was low (<15% w/w), soil highest resin NH4-N values were actually
temperature and N content had minimal effect observed at the hottest and driest soil conditions
on NH4-N and NO3-N (Fig. 5). (low-elevation slope) in 2015 (lower summer pre-
The soils that fell within the highest soil tem- cipitation year); however, this pattern was not as
perature bracket (>15°C) were also drier, with distinctive in 2016 (higher summer precipitation
soil moisture content always less than ~40% year; Fig. 6B, Table 1). Resin NO3-N values, on
(w/w; Fig. 5). These dry and warm conditions the other hand, were typically low for all the sites
generally occurred in mid-summer at the low- but high at one of the wettest locations (N-high
elevation slope locations, where soil %N was hollow; Fig. 6C, D); at this location, NO3-N
relatively low (Fig. 2B). Since soil %N is corre- accounted for on average 63% of total exchange-
lated with soil extractable N, but mediated by able inorganic N during our study years, while
soil moisture (Table 2), we also analyzed the rela- NO3-N accounted for 34% in all other hollow
tionship between soil %N and NH4-N and locations and 18% in all slope locations (Fig. 6).
between soil %N and NO3-N, separating the Net nitrification rates under snowpack were also
relatively wet year, 2015, from the dry year, 2016. highest at the wettest locations (high-elevation
We found that, although soil %N positively influ- hollows), compared to other drier locations
enced the extractable N concentrations in both (Fig. 7). Despite these relatively high overwinter
years, its influence was more pronounced in the NO3-N supply rates, NO3-N leaching appeared
wetter year than in the dry year. In the drier year, minimal; NO3-N concentrations in the streams
extractable N was consistently low across the and soil water at these wettest locations were
entire range of the soil N (Appendix S1: Fig. S1). consistently low during the 2016 snowmelt
This further supports the importance of soil (Appendix S1: Fig. S2).
 ❖ www.esajournals.org 10 July 2019 ❖ Volume 10(7) ❖ Article e02794
YANO ET AL.
Fig. 5. The relationships between extractable NH4-N and NO3-N with soil moisture, separated by four temper-
ature brackets (≤5°C, 5–10°C, 10–15°C, and >15°C; the temperature ranged from 0.4°C to 20°C), are shown for
NH4-N (left) and NO3-N (right). All the data across the topographic positions and years were combined for the
analysis. The lines represent best-fit models selected by Akaike information criterion. All the models were statis-
tically significant (P < 0.0001). The outliers that were not included in the analysis are shown in open circles.
 ❖ www.esajournals.org 11 July 2019 ❖ Volume 10(7) ❖ Article e02794
YANO ET AL.
Fig. 6. Resin-exchangeable N during 2015 and 2016. NH4-N (A, B) and NO3-N (C, D) are shown for hollow
(left) and slope (right). The error bars are 1 standard error. The asterisks indicate overwinter resin probes. The
lowercase letters indicate significant differences of means for each site (n = 3) found by ANOVA at P < 0.05, fol-
lowed by Tukey test.
Antecedent soil moisture effects of snowpack on precipitation in 2015, soil moisture and extracta-
extractable N ble N were higher in 2015 than in 2016. Further
We examined the precipitation patterns during investigation revealed that the seasonal snow-
2015 and 2016 water years to understand how pack at high elevation during the 2014–2015 win-
winter snow and summer precipitation impacted ter formed over wetter soils during fall 2014
soil moisture and subsequent seasonal N avail- (Fig. 8C, ~20% v/v, S-high and N-high, left-most
ability. We found that both water years had simi- arrow on left; Fig. 8E, Start 
2 weeks and Start
lar peak SWE and snowmelt timing, but growing +2 weeks), compared to the 2015–2016 winter
season precipitation during 2015 was 70% (at (Fig. 8C, ~15% v/v, S-high and N-high, second
high elevation) and 36% (at low elevation) lower arrow on right; Fig. 8E, Start 
2 weeks and Start
than the 2016 growing season (Table 1). How- +2 weeks). The difference in soil moisture
ever, despite lower snow-free season between years was magnified even more at the
 ❖ www.esajournals.org 12 July 2019 ❖ Volume 10(7) ❖ Article e02794
YANO ET AL.
of snowmelt, soil moisture and temperature were
similar between the two years.
DISCUSSION
Within mountain ecosystems, complex topog-
raphy can lead to heterogeneity in climate,
hydrology, and soil properties. While mountain
topography often results in predictable gradients
in precipitation and temperature, we found large
differences in snowpack accumulation across rel-
atively small spatial scales, such as across hollow
and slope topographic positions. This variability
led to large spatial and temporal heterogeneity in
soil moisture and temperature, which, in turn,
influenced soil N dynamics. Our findings also
suggest that the influence of snowpack on N
availability can extend beyond the current grow-
ing season; large snowpack years can keep soil
moisture elevated throughout summer and into
fall, creating wetter and warmer soil conditions
before snowpack formation, leading to optimal
conditions for N transformation over winter.
Lastly, this study suggests that the spatial and
Fig. 7. Net N mineralization and nitrification over temporal patterns of N dynamics in western con-
winter. The in situ soil core (0–15 cm) incubation ifer forests are quite different from other pub-
started on 9 November 2016 at all locations and ended lished studies from temperate mesic forests
at the end of snowmelt: on 7 April 2017 at the low ele- (Nadelhoffer et al. 1984, Hentschel et al. 2009,
vation and on 4 May 2017 at the high elevation. Signifi- Groffman et al. 2011, Campbell et al. 2014).
cant differences were found by non-parametric Unlike the mesic temperate forests from the east-
Kruskal–Wallis test followed by Dunn test. The differ- ern United States or Europe, which experience
ent lowercase letters without parentheses indicate sta- consistent precipitation throughout the year, our
tistical difference (P < 0.05); the letter in the
findings suggest that the mid-summer drought
parenthesis indicates marginal difference (P < 0.1). creates sufficiently dry soils such that summer
precipitation does not induce the same soil N
low-elevation sites; fall soil moisture was twice response as soil moisture derived from snow-
as high prior to the 2014–2015 winter (Fig. 8C, melt.
~15% v/v, S-low and N-low, left-most arrow;
Fig. 8E) compared to the 2015–2016 winter Topographic influence on snowpack and soil
(Fig. 8C, ~8% v/v, S-low and N-low, second environmental factors
arrow on right; Fig. 8E). Thus, the higher soil Spatially variable snowpack distribution
moisture during the 2015 growing season was across the watershed had a strong influence on
most likely due to a carryover effect from the soil moisture and temperature. As expected, the
above-average SWE during the 2013–2014 winter snowpack at the higher-elevation sites (North-
(135% of the 30-yr average, Fig. 8A). This also high and South-high) had generally higher peak
suggests that despite more rain falling in 2016 SWE than the lower elevation sites (North-low
during the snow-free period (Table 1), rain con- and South-low), but these differences in SWE
tributed little or minimally to soil moisture. The were often confounded by topographic positions
mean soil temperature at the time of snowpack (hollow vs. slope) and aspect. Across mountain
formation was slightly higher (approximately 1– ecosystems, increasing elevation often leads to
2°C) in 2015 than 2014 (Fig. 8D), but by the end decreasing air temperature and increasing
 ❖ www.esajournals.org 13 July 2019 ❖ Volume 10(7) ❖ Article e02794
YANO ET AL.
Fig. 8. Snow-water equivalent (SWE) (A), soil temperature (B), soil moisture (C), mean soil temperature (D),
and soil moisture (E) during two-week periods before (
2 weeks) and after (+2 weeks) the start and end of
snowpack-covered period are shown for 2014–2015 and 2015–2016 snowpack. All the temperature and moisture
data shown here are from the hollows at 10 cm deep. The mean temperature and moisture shown in (D) and (E)
are from the S-high site. The error bars are 1 standard error. SNOTEL data were used to determine the start
date of snow cover, whereas the first snow-free day (see Methods for details) was used as the end date of snow
cover. The solid and broken arrows in (A) indicate 1981–2010 average SWE at the high- and low-elevation SNO-
TEL sites, respectively. The arrows in (B) and (C) panels correspond to the soil temperature and moisture com-
parison between the years shown in (D) and (E).
precipitation, resulting in generally deeper snow- the main form of water input in many western
pack at higher elevations (Elder et al. 1991). U.S. forests, soil moisture peaks during snow-
However, the complex shape of the landscape, as melt and continues to decline as the summer pro-
well as differences in insolation (as a result of gresses (Blankinship et al. 2014, Maurer and
aspect) and snowdrift patterns, reduced our abil- Bowling 2014). Higher soil moisture also tends to
ity to predict SWE across small spatial scales. occur at higher elevations and in more conver-
This heterogeneity in SWE across montane gent landscape positions (hollows) compared to
watersheds has been well demonstrated (Elder lower elevation and divergent landscape posi-
et al. 1991, Luce et al. 1998, Anderton et al. 2004) tions (slope). This general pattern across the
and highlights the variable microclimates in com- landscape occurs partly because of differences in
plex terrain. atmospheric demand for water; the demand is
While the heterogeneity in SWE can result in generally low at a high elevation and high at a
different levels of soil moisture following snow- low elevation/southern aspect (Martin et al.
melt, some general trends in soil moisture do 2017). Furthermore, hollows typically have
occur across our watershed. Because snowmelt is higher soil moisture than slopes due to a
 ❖ www.esajournals.org 14 July 2019 ❖ Volume 10(7) ❖ Article e02794
YANO ET AL.
combination of factors, including gravitational >50% of variability in available N; however,
flow of soil water from upslope areas (Jencso when the soils were warm in summer (>15°C),
et al. 2009), lower insolation (more vegetation the models became more complex (e.g., for NH4-
and shading; Nemani and Running 1989), and N) or explanatory power of the models
cold air drainage during the evening/early morn- decreased (e.g., for NO3-N, Table 2).
ing (Novick et al. 2016). The heterogeneity of soil One likely explanation for the higher complex-
moisture within watersheds can also contribute ity of models or the lack of strong explanatory
to spatial patterns of soil C and N across topog- power under warmer soil temperature is the
raphy. Organic C and N tend to accumulate in changes in the balance among N mineralization,
convergent landscape positions, because lateral nitrification, N uptake, and N losses. When soils
flows of water transport particulates downhill are cold, abiotic (e.g., leaching, sorption) and
(Amundson and Jenny 1997, Weintraub et al. microbial processes likely dominate N cycling,
2015), but higher soil moisture also promotes but once soil and root temperatures reach >6°C,
leaching and gaseous losses (Perakis and Sink- plants would actively take up available N
horn 2011, Malone et al. 2018). However, despite (Alvares-Uria and Ko€rner 2007) and begin to
these general trends, soil moisture tends to be influence extractable N pools. Soil microbial
highly heterogeneous and remains one of the communities also shift dramatically from winter
most difficult parameters to model within mon- to summer (Schadt et al. 2003), with the different
tane watersheds. communities responding differently to soil tem-
Our data also indicate that deeper snowpack perature, moisture, and soil %C and %N. In
not only maintained mild winter soil tempera- snowmelt-dominated ecosystems, the end of
ture but also helped the soil maintain smaller snowmelt marks the transition from fungal-
temperature and moisture swings throughout dominated winter microbial communities to bac-
the year (Figs. 4A, I and 8B). While the direct teria-dominated summer microbial communities
insulating effect of snowpack on winter soil tem- (Schadt et al. 2003), which in turn control N
perature has been well demonstrated (Brooks transformation and immobilization (Edwards
et al. 1998, Groffman et al. 2009), our results sup- et al. 2006). In an alpine dry meadow in Color-
port the idea that snowpack may indirectly influ- ado, United States, this community transition led
ence soil temperature across the season via latent to a large pulse of soluble proteins to be released
heat exchange (Brooks et al. 2011) by sustaining from the winter microbes shortly after snowmelt,
high soil moisture availability. resulting in a pulse of extractable inorganic N
(Lipson et al. 1999). This mechanism could
Spatial and temporal effects on N dynamics explain why we found larger extractable N peaks
Landscape heterogeneity can create high spa- during the wetter year in 2015, compared to the
tial and temporal variability in patterns of soil drier year in 2016.
available N by mediating soil temperature and In addition to soil temperature and moisture,
moisture. Our data confirmed the importance of soil N content is another key driver that influ-
both moisture and temperature on soil available ences N transformation rates. In mesic ecosys-
N, but we also found that the strength of the rela- tems with mean annual precipitation over
tionships shifted considerably across a range of 1500 mm, studies have found soil %N to
soil temperatures. For example, the explanatory strongly control extractable soil N (Schuur and
power (R2-values) of our mixed-effect regression Matson 2001, Griffiths et al. 2009, Perakis and
models for all data points was weak, but the Sinkhorn 2011, Weintraub et al. 2015). While our
power improved considerably by fitting models findings generally confirm the positive relation-
separately by four different temperature brackets ship between soil %N and extractable N
(Table 2, Fig. 5). This suggests that the relation- (Table 2), we also found that the effect of soil %N
ships among soil temperature, moisture, and on extractable N was strongly mediated by soil
both %C and %N of soils were not constant moisture (Appendix S1: Fig. S1). This suggests
throughout the year. For example, in early spring that N dynamics is strongly influenced by water
when the soils were cold (<5°C), soil moisture availability in western semi-arid forest
alone or moisture and %N together could explain
 ❖ www.esajournals.org 15 July 2019 ❖ Volume 10(7) ❖ Article e02794
YANO ET AL.
ecosystems, and suggests that N availability fluc- soil microbes to pulses of moisture availability in
tuates with inter-annual variations in soil mois- water-limited systems highlight the complex nat-
ture. ure of soil N dynamics and further support the
Traditionally, in arid ecosystems, such as those importance of water in driving many of the spa-
experiencing a Mediterranean climate, microbial tial and temporal patterns of available N.
N transformation in soil was thought to be Leaching of N below the rooting zone can also
absent in the middle of dry summer. Although occur during snowmelt (Brooks et al. 1998, Wil-
water availability limits biological activities and liams et al. 2009) or upon thunderstorms during
the transport of biochemical substrates (Stark summer droughts (Reichmann et al. 2013) when
and Firestone 1995, Wang et al. 2015), an increas- net nitrification exceeds N uptake. Although the
ing number of studies have shown that N cycling low NO3-N concentrations in soil lysimeter and
can be active in the midst of summer drought in stream water in 2016 spring suggest NO3-N
these arid ecosystems (Parker and Schimel 2011, leaching may be less important in this forest
Homyak et al. 2016). Our finding of high resin- (Appendix S1: Fig. S2), NO3-N leaching could be
exchangeable NH4-N at the driest sites during important in other years, when overwinter nitri-
the driest time of the year is consistent with these fication is greater under snowpack because of
studies. At the driest sites (Low-slope), the large higher soil moisture in the fall (discussed further
discrepancy between extractable NH4-N (lowest) in Antecedent fall soil moisture vs. summer precipita-
and resin NH4-N (highest) in the summer can be tion below). Although we did not measure gas-
explained by pulses of N mineralization cap- eous N loss rates in this study, denitrification
tured by the resin probes following summer rain under saturated soil conditions (Williams et al.
events in the relatively coarse-textured soils 1998) is another possible pathway that con-
(Austin et al. 2004, Schwinning and Sala 2004). It tributed to the declines in extractable N in soils
is likely that these NH4-N pulses were quickly toward the end of snowmelt (Fig. 4E, G).
followed by microbial N immobilization or by
nitrification, and denitrification, resulting in the Antecedent fall soil moisture vs. summer
low levels of resin-exchangeable NO3-N (Fig. 6B, precipitation
D). Recent studies in arid ecosystem have also The insulating properties of snowpack for win-
shown that these processes can lead to more N ter N mineralization have been well documented
losses during a drought than compared to a wet- in a variety of snowmelt-dominated ecosystems
ter season (Parker and Schimel 2011, Homyak (Schimel et al. 2004, Groffman et al. 2009, Brooks
et al. 2016). For example, in a semi-arid grass- et al. 2011). Given this widely occurring phe-
land, N mineralization, nitrification, and both nomenon, we had expected differences in snow-
nitrification and denitrification potential all pack SWE between our two study years to
increased during a summer drought, when plant explain the differences in seasonal soil N dynam-
N uptake was absent (Parker and Schimel 2011). ics. However, we were surprised to find that both
Furthermore, the largest nitric oxide (NO) emis- 2015 and 2016 had almost identical SWE, as well
sions occurred quickly upon rewetting of these as the timing of snowmelt (Table 1), while extrac-
dry soils, because the substrates for NO-produ- table NH4-N and NO3-N were more than twice
cing processes that were isolated in soil pores as high in 2015. This suggested that additional
became suddenly accessible to NO-producing factors related to soil moisture and temperature
microbes through diffusion (Homyak et al. were responsible for these differences. We
2016). In our study, similar mechanisms may hypothesized that different antecedent soil mois-
have been responsible for the high resin NH4-N ture conditions during the fall just before snow-
levels observed at the driest locations (low-eleva- pack formation may create different conditions
tion, slope) during the drier summer of 2015. By for winter soil microbes and N mineralization
contrast, substrates for N mineralization were under snowpack. The importance of antecedent
likely more accessible to soil microbes through- soil conditions has been observed in the past,
out the wetter summer of 2016. This may have where a drier fall led to lower winter soil micro-
resulted in the lack of large NH4-N pulses bial biomass and inter-annual differences in
(Fig. 6B, D). These relatively fast responses by nutrient concentrations (Edwards and Jefferies
 ❖ www.esajournals.org 16 July 2019 ❖ Volume 10(7) ❖ Article e02794
YANO ET AL.
2013). In our study, soils were wetter and cooler extractable N or resin-exchangeable N during the
in the fall 2014 than in fall 2015 (Fig. 8C–E). The 2016 growing season (Figs. 4, 6), despite 70%
reason for the higher soil moisture during fall more rainfall in 2016 (Table 1). This also provides
2014 was largely due to the higher-than-average evidence that N availability during the growing
snowpack during the winter of 2013–2014 season may be more strongly influenced by
(Fig. 8A) combined with long-term memory of snowmelt water than summer rain, and that a
anomalies by soil moisture (Palmer 1965, Dai shift from snow to rain under the warmer climate
et al. 2004). The large snowpack and subsequent may alter the availability of N during the grow-
late melt kept soil moisture elevated (Maurer and ing season.
Bowling 2014, Harpold and Molotch 2015)
throughout 2014 and ultimately led to wetter CONCLUSION
soils during snowpack formation in the fall of
2014. Additionally, our net N mineralization and Topographic controls on soil N availability
nitrification rates measured over winter follow- manifest primarily through its influence on local
ing the relatively wet 2016 fall (Figs. 7, 8C, hydrometeorology and soil properties. In our
Table 1) were also high, and comparable to those study, we found significant differences in snow-
rates observed in a wetter environment such pack accumulation across a catchment, with
eastern U.S. forests and arctic tundra (Groffman these differences in snowpack having lasting
et al. 2001a, b, Schimel et al. 2004). Thus, in effects on soil moisture and temperature
snow-dominated semi-arid temperate forests, throughout the growing season. In turn, these
soil conditions in the previous fall may play a differences shaped both spatial and temporal
critical role in N availability in the following patterns of soil extractable NH4-N and NO3-N.
growing season. Our data confirmed the generally observed pat-
While greater summer precipitation may also tern of higher extractable N associated with
contribute to higher soil moisture in the fall, our higher soil %N across the landscape. However,
data suggest that this may be true only for the soil extractable N was more than twice as high in
cooler and wetter locations (the hollows at high the wet year (2015) than in the dry year (2016).
elevation) and not for the other drier locations We suggest that antecedent soil moisture condi-
(all slopes or at low elevation). For these drier tions before snowpack formation, along with
locations, fall precipitation appeared to have a high soil moisture derived from snowmelt, are
stronger influence on fall soil moisture than sum- collectively responsible for high N mineralization
mer precipitation. Despite more precipitation in rates over winter and N availability in the fol-
the 2016 growing season than 2015, average soil lowing season. This antecedent moisture effects
moisture at the end of August was similarly low may be particularly important in ecosystems that
for both years (Table 1). The differences in fall experience summer drought. Future studies that
soil moisture between the two years developed can take advantage of long-term N dynamic
only after the summer; average October soil datasets combined with field manipulation study
moisture across all sites was still low in 2015, but to test antecedent soil conditions would help
was twice as high in 2016. Fall precipitation may address potential carryover effects and lag
enhance soil moisture in the drier locations more responses.
than summer precipitation because most of the
summer rain tends to be either intercepted by the ACKNOWLEDGMENTS
canopy and evaporated or absorbed by the forest
litter before infiltrating the soils (Reynolds and We thank F. Maus and L. Nitz for facilitating access
Knight 1973, Hu et al. 2010). High atmospheric to Lubrecht Experimental Forest and extend our grati-
demand for water in the summer at the lower tude to E. Anderson and T. Field for field assistance.
elevation also intensifies water losses through This project was supported in part by the Montana
evaporation and transpiration, whereas lower EPSCoR Program via the Montana Institute on Ecosys-
atmospheric demand for water in fall would help tems, and the USDA, National Institute of Food and
retain more rainwater in the soils. This could Agriculture Grant, Award number 2015-67020-23454.
partly explain why we did not observe higher
 ❖ www.esajournals.org 17 July 2019 ❖ Volume 10(7) ❖ Article e02794
YANO ET AL.
LITERATURE CITED and nutrients in wet and dry low-Arctic sedge
meadows. Soil Biology & Biochemistry 57:83–90.
Alvares-Uria, P., and C. Ko€rner. 2007. Low tempera- Edwards, K. A., J. McCulloch, G. P. Kershaw, and R. L.
ture limits of root growth in deciduous and ever- Jefferies. 2006. Soil microbial and nutrient dynam-
green temperate tree species. Functional Ecology ics in a wet Arctic sedge meadow in late winter
21:211–218. and early spring. Soil Biology & Biochemistry
Amundson, R., and H. Jenny. 1997. Thinking of biol- 38:2843–2851.
ogy: on a state factor model of ecosystems. BioS- Elder, K., J. Dozier, and J. Michaelsen. 1991. Snow
cience 47:536–543. accumulation and distribution in an alpine water-
Anderton, S. P., S. M. White, and B. Alvera. 2004. Eval- shed. Water Resources Research 27:1541–1552.
uation of spatial variability in snow water equiva- Fitzhugh, R. D., C. T. Driscoll, P. M. Groffman, G. L.
lent for a high mountain catchment. Hydrological Tierney, T. J. Fahey, and J. P. Hardy. 2001. Effects of
Processes 18:435–453. soil freezing disturbance on soil solution nitrogen,
Austin, A. T., L. Yahdjian, J. M. Stark, J. Belnap, A. Por- phosphorus, and carbon chemistry in a northern
porato, U. Norton, D. A. Ravetta, and S. M. Schaef- hardwood ecosystem. Biogeochemistry 56:215–238.
fer. 2004. Water pulses and biogeochemical cycles Griffiths, R. P., M. D. Madritch, and A. K. Swanson.
in arid and semiarid ecosystems. Oecologia 2009. The effects of topography on forest soil char-
141:221–235. acteristics in the Oregon Cascade Mountains
Barnhart, T. B., N. P. Molotch, B. Livneh, A. A. Har- (USA): implications for the effects of climate
pold, J. F. Knowles, and D. Schneider. 2016. Snow- change on soil properties. Forest Ecology and Man-
melt rate dictates streamflow. Geophysical agement 257:1–7.
Research Letters 43:8006–8016. Groffman, P. M., C. T. Driscoll, T. J. Fahey, J. P. Hardy,
Bernhardt, E. S., J. R. Blaszczak, C. D. Ficken, M. L. R. D. Fitzhugh, and G. L. Tierney. 2001a. Colder
Fork, K. E. Kaiser, and E. C. Seybold. 2017. Control soils in a warmer world: a snow manipulation
points in ecosystems: moving beyond the hot spot study in a northern hardwood forest ecosystem.
hot moment concept. Ecosystems 20:665–682. Biogeochemistry 56:135–150.
Blankinship, J. C., M. W. Meadows, R. G. Lucas, and S. Groffman, P. M., C. T. Driscoll, T. J. Fahey, J. P. Hardy,
C. Hart. 2014. Snowmelt timing alters shallow but R. D. Fitzhugh, and G. L. Tierney. 2001b. Effects of
not deep soil moisture in the Sierra Nevada. Water mild winter freezing on soil nitrogen and carbon
Resources Research 50:1448–1456. dynamics in a northern hardwood forest. Biogeo-
Brooks, P. D., P. Grogan, P. H. Templer, P. Groffman, chemistry 56:191–213.
M. G. O€quist, and J. Schimel. 2011. Carbon and Groffman, P. M., J. P. Hardy, S. Fashu-Kanu, C. T. Dris-
nitrogen cycling in snow-covered environments. coll, N. L. Cleavitt, T. J. Fahey, and M. C. Fisk.
Geography Compass 5:682–699. 2011. Snow depth, soil freezing and nitrogen
Brooks, P. D., M. W. Williams, and S. K. Schmidt. 1998. cycling in a northern hardwood forest landscape.
Inorganic nitrogen and microbial biomass dynam- Biogeochemistry 102:223–238.
ics before and during spring snowmelt. Biogeo- Groffman, P. M., J. P. Hardy, M. C. Fisk, T. J. Fahey,
chemistry 43:1–15. and C. T. Driscoll. 2009. Climate variation and soil
Campbell, J. L., A. M. Socci, and P. H. Templer. 2014. carbon and nitrogen cycling processes in a north-
Increased nitrogen leaching following soil freezing ern hardwood forest. Ecosystems 12:927–943.
is due to decreased root uptake in a northern hard- Harpold, A. A., and N. P. Molotch. 2015. Sensitivity of
wood forest. Global Change Biology 2663–2673. soil water availability to changing snowmelt tim-
Dai, A., K. E. Trenberth, and T. Qian. 2004. A global ing in the western US. Geophysical Research Let-
dataset of Palmer drought severity index for 1870– ters 42:8011–8020.
2002: relationship with soil moisture and effects of Hentschel, K., W. Borken, T. Zuber, C. Bogner, B.
surface warming. Journal of Hydrometeorology Huwe, and E. Matzner. 2009. Effects of soil frost on
5:1117–1130. nitrogen net mineralization, soil solution chemistry
Duran, J., J. L. Morse, P. M. Groffman, J. L. Campbell, and seepage losses in a temperate forest soil. Glo-
L. M. Christenson, C. T. Driscoll, T. J. Fahey, M. C. bal Change Biology 15:825–836.
Fisk, M. J. Mitchell, and P. H. Templer. 2014. Winter Homyak, P. M., J. C. Blankinship, K. Marchus, D. M.
climate change affects growing-season soil micro- Lucero, J. O. Sickman, and J. P. Schimel. 2016. Arid-
bial biomass and activity in northern hardwood ity and plant uptake interact to make dryland soils
forests. Global Change Biology 20:3568–3577. hotspots for nitric oxide (NO) emissions. Proceed-
Edwards, K. A. and R. L. Jefferies. 2013. Inter-annual ings of the National Academy of Sciences USA 113:
and seasonal dynamics of soil microbial biomass E2608–E2616.
 ❖ www.esajournals.org 18 July 2019 ❖ Volume 10(7) ❖ Article e02794
YANO ET AL.
Hook, P. B., and I. C. Burke. 2000. Biogeochemistry zones to catchment runoff. Water Resources
in a shortgrass landscape: control by topography, Research 39:1310.
soil texture, and microclimate. Ecology 81:2686– McGlynn, B. L., J. J. McDonnell, J. B. Shanley, and C.
2703. Kendall. 1999. Riparian zone flowpath dynamics
Hu, J., D. J. P. Moore, S. P. Burns, and R. K. Monson. during snowmelt in a small headwater catchment.
2010. Longer growing seasons lead to less carbon Journal of Hydrology 222:75–92.
sequestration by a subalpine forest. Global Change Molotch, N. P., P. D. Brooks, S. P. Burns, M. Litvak, R.
Biology 16:771–783. K. Monson, J. R. McConnell, and K. Musselman.
Jencso, K. G., B. L. McGlynn, M. N. Gooseff, S. M. 2009. Ecohydrological controls on snowmelt parti-
Wondzell, K. E. Bencala, and L. A. Marshall. 2009. tioning in mixed-conifer sub-alpine forests. Ecohy-
Hydrologic connectivity between landscapes and drology 2:129–142.
streams: transferring reach- and plot-scale under- Moore, I. D., P. E. Gessler, G. A. Nielsen, and G. A.
standing to the catchment scale. Water Resources Peterson. 1993. Soil attribute prediction using ter-
Research 45:W04428. rain analysis. Soil Science Society of America Jour-
Johnson, D. W., W. W. Miller, R. B. Susfalk, J. D. Mur- nal 57:443–452.
phy, R. A. Dahlgren, and D. W. Glass. 2009. Biogeo- Nadelhoffer, K. J., J. D. Aber, and J. M. Melillo. 1984.
chemical cycling in forest soils of the eastern Sierra Seasonal patterns of ammonium and nitrate uptake
Nevada Mountains, USA. Forest Ecology and Man- in nine temperate forest ecosystems. Plant and Soil
agement 258:2249–2260. 80:321–335.
LeBauer, D. S., and K. K. Treseder. 2008. Nitrogen limi- Nadelhoffer, K. J., J. D. Aber, and J. M. Melillo. 1985.
tation of net primary productivity in terrestrial Fine roots, net primary production, and soil nitro-
ecosystems is globally distributed. Ecology 89:371– gen availability: a new hypothesis. Ecology
379. 66:1377–1390.
Lefcheck, J. S. 2016. piecewiseSEM: Piecewise struc- Nemani, R. R., and S. W. Running. 1989. Estimation of
tural equation modelling in r for ecology, evolu- regional surface-resistance to evapotranspiration
tion, and systematics. Methods in Ecology and from ndvi and thermal-ir avhrr data. Journal of
Evolution 7:573–579. Applied Meteorology 28:276–284.
Lipson, D. A., S. K. Schmidt, and R. K. Monson. 1999. Novick, K. A., A. C. Oishi, and C. F. Miniat. 2016. Cold
Links between microbial population dynamics and air drainage flows subsidize montane valley
nitrogen availability in an alpine ecosystem. Ecol- ecosystem productivity. Global Change Biology
ogy 80:1623–1631. 22:4014–4027.
Luce, C. H., D. G. Tarboton, and R. R. Cooley. 1998. NRCS (Natural Resources Conservation Service). 2017.
The influence of the spatial distribution of snow on Web soil survey. United States Department of Agri-
basin-averaged snowmelt. Hydrological Processes culture, Washington, D.C., USA. http://websoil
12:1671–1683. survey.nrcs.usda.gov/app/HomePage.htm
Malone, E. T., B. W. Abbott, M. J. Klaar, C. Kidd, M. Palmer, W. C. 1965. Meteorological drought. Research
Sebilo, A. M. Milner, and G. Pinay. 2018. Decline in Paper No. 45, U.S. Department of Commerce
ecosystem delta C-13 and mid-successional nitro- Weather Bureau, Washington, D.C., USA.
gen loss in a two-century postglacial chronose- Parker, S. S., and J. P. Schimel. 2011. Soil nitrogen avail-
quence. Ecosystems 21:1659–1675. ability and transformations differ between the
Martin, J., N. Looker, Z. Hoylman, K. Jencso, and J. summer and the growing season in a California
Hu. 2017. Hydrometeorology organizes intra- grassland. Applied Soil Ecology 48:185–192.
annual patterns of tree growth across time, space Perakis, S. S., and E. R. Sinkhorn. 2011. Biogeochem-
and species in a montane watershed. New Phytolo- istry of a temperate forest nitrogen gradient. Ecol-
gist 215:1387–1398. ogy 92:1481–1491.
Maurer, G. E., and D. R. Bowling. 2014. Seasonal Powers, R. F. 1990. Nitrogen mineralization along an
snowpack characteristics influence soil tempera- altitudinal gradient – interactions of soil-tempera-
ture and water content at multiple scales in interior ture, moisture, and substrate quality. Forest Ecol-
western US mountain ecosystems. Water Resources ogy and Management 30:19–29.
Research 50:5216–5234. R Core Team. 2013. R: A language and environment
McClain, M. E., et al. 2003. Biogeochemical hot spots for statistical computing. R Foundation for Statisti-
and hot moments at the interface of terrestrial and cal Computing, Vienna, Austria.
aquatic ecosystems. Ecosystems 6:301–312. Reichmann, L. G., O. E. Sala, and D. P. C. Peters. 2013.
McGlynn, B. L., and J. J. McDonnell. 2003. Quantifying Water controls on nitrogen transformations and
the relative contributions of riparian and hillslope stocks in an arid ecosystem. Ecosphere 4:1–17.
 ❖ www.esajournals.org 19 July 2019 ❖ Volume 10(7) ❖ Article e02794
YANO ET AL.
Reynolds, J. F., and D. H. Knight. 1973. The magnitude Schwinning, S., and O. E. Sala. 2004. Hierarchy of
of snowmelt and rainfall interception by litter in responses to resource pulses in arid and semi-arid
lodgepole pine and spruce-fir forests in Wyoming. ecosystems. Oecologia 141:211–220.
Northwest Science 47:50–60. Stark, J. M., and M. K. Firestone. 1995. Mechanisms for
Saros, J. E., T. J. Michel, S. J. Interlandi, and A. P. Wolfe. soil moisture effects on activity of nitrifying bacte-
2005. Resource requirements of Asterionella ria. Applied and Environmental Microbiology
formosa and Fragilaria crotonensis in oligotrophic 61:218–221.
alpine lakes: implications for recent phyto- Vitousek, P. M., and R. W. Howarth. 1991. Nitrogen
plankton community reorganizations. Canadian limitation on land and in the sea – how can it occur.
Journal of Fisheries and Aquatic Sciences 62:1681– Biogeochemistry 13:87–115.
1689. Wang, L., S. Manzoni, S. Ravi, D. Riveros-Iregui, and
Schadt, C. W., A. P. Martin, D. A. Lipson, and S. K. K. Caylor. 2015. Dynamic interactions of ecohydro-
Schmidt. 2003. Seasonal dynamics of previously logical and biogeochemical processes in water-lim-
unknown fungal lineages in tundra soils. Science ited systems. Ecosphere 6:1–27.
301:1359–1361. Weintraub, S. R., P. G. Taylor, S. Porder, C. C.
Schimel, J. P., C. Bilbrough, and J. A. Welker. 2004. Cleveland, G. P. Asner, and A. R. Townsend.
Increased snow depth affects microbial activity 2015. Topographic controls on soil nitrogen avail-
and nitrogen mineralization in two Arctic tundra ability in a lowland tropical forest. Ecology
communities. Soil Biology & Biochemistry 36:217– 96:1561–1574.
227. Williams, M. W., P. D. Brooks, and T. Seastedt. 1998.
Schimel, D. S., B. H. Braswell, and W. J. Parton. 1997. Nitrogen and carbon soil dynamics in response to
Equilibration of the terrestrial water, nitrogen, and climate change in a high-elevation ecosystem in the
carbon cycles. Proceedings of the National Acad- Rocky Mountains, U.S.A. Arctic and Alpine
emy of Sciences USA 94:8280–8283. Research 30:26–30.
Schuur, E. A. G., and P. A. Matson. 2001. Net primary Williams, M., C. Seibold, and K. Chowanski. 2009.
productivity and nutrient cycling across a mesic to Storage and release of solutes from a subalpine sea-
wet precipitation gradient in Hawaiian montane sonal snowpack: soil and stream water response,
forest. Oecologia 128:431–442. Niwot Ridge, Colorado. Biogeochemistry 95:77–94.
SUPPORTING INFORMATION
Additional Supporting Information may be found online at: http://onlinelibrary.wiley.com/doi/10.1002/ecs2.
2794/full
 ❖ www.esajournals.org 20 July 2019 ❖ Volume 10(7) ❖ Article e02794