During spring, daily stream flow and groundwater dynamics in forested subalpine catchments are to a large extent controlled by hydrological processes that respond to the day–night energy cycle. Diurnal snowmelt and transpiration events combine to induce pressure variations in the soil water storage that are propagated to the stream. In headwater catchments these pressure variations can account for a significant amount of the total pressure in the system and control the magnitude, duration, and timing of stream inflow pulses at daily scales, especially in low-flow systems. Changes in the radiative balance at the top of the snowpack can alter the diurnal hydrologic dynamics of the hillslope–stream system, with potential ecological and management consequences.
We present a detailed hourly dataset of atmospheric, hillslope, and streamflow measurements collected during one melt season from a semi-alpine headwater catchment in western Montana, US. We use this dataset to investigate the timing, pattern, and linkages among snowmelt-dominated hydrologic processes and assess the role of the snowpack, transpiration, and hillslopes in mediating daily movements of water from the top of the snowpack to local stream systems. We found that the amount of snowpack cold content accumulated during the night, which must be overcome every morning before snowmelt resumes, delayed water recharge inputs by up to 3 h early in the melt season. These delays were further exacerbated by multi-day storms (cold fronts), which resulted in significant depletions in the soil and stream storages. We also found that both diurnal snowmelt and transpiration signals are present in the diurnal soil and stream storage fluctuations, although the individual contributions of these processes are difficult to discern. Our analysis showed that the hydrologic response of the snow–hillslope–stream system is highly sensitive to atmospheric drivers at hourly scales and that variations in atmospheric energy inputs or other stresses are quickly transmitted and alter the intensity, duration, and timing of snowmelt pulses and soil water extractions by vegetation, which ultimately drive variations in soil and stream water pressures.
In snow-dominated headwater catchments, hillslope hydrologic
processes and streamflow during spring are largely determined by fluxes of
snowmelt
Diurnal streamflow fluctuations have been observed over a range of climates
and scales
The study of diurnal cycles of snowmelt, water table, and stream recharge
fluctuations in snow-fed river systems complements the study of diurnal
streamflow signals in areas dominated by evapotranspiration
Solar radiation and air temperature are considered the two major cyclic
forcings that induce thaw and evapotranspiration, exerting a strong control
on soil moisture, recharge to the shallow groundwater system, and soil
hydraulic gradients
In this study, we present a comprehensive set of hourly atmospheric,
hillslope, and streamflow measurements from a semi-alpine headwater catchment
in western Montana, US, to investigate the timing, pattern, and linkages among
snowmelt-dominated subdaily hydrologic processes. The initial assumption is
that in our study site, during the melt season, diurnal variations in the
snowpack energy state, as well as alterations in flow timings and pathways
induced by changes in snowpack depth and density, are larger controls on the
amount and timing of diurnal and seasonal water inputs into streams than
transpiration or contributions from the mountain block groundwater system.
However, how these processes evolve and combine to generate the observed
diurnal hillslope and stream response is unclear
Study area in the upper Lost Horse Creek watershed, including location
in western Montana
Our study site is a small (2.9 km
We instrumented a 40 m long, north-facing hillslope at the bottom of the
study drainage with a meteorological station, five shallow monitoring wells
(Fig.
Five wells were placed in the anticipated path of groundwater flow along the
study hillslope. These wells were placed with a drive-point rod and sleeve,
pounded to the point of refusal, and installed with 12.7 mm diameter PVC
encasing. Wells were backfilled with sand and capped with bentonite. These
five wells were instrumented with HOBO pressure transducers (error
Sap flow velocity was monitored at 10 min intervals through the stem of
two adjacent Engelmann spruce of different age using the Granier-type Dynamax
thermal dissipation probe (TDP-50) sap flow monitoring system
The analysis focuses on one melt and growing season spanning from 1 April to
1 September 2012. All datasets except stream stage cover this period without
gaps. The stream stage record starts on 26 April, once ice had melted
sufficiently to permit pressure-transducer installation. All datasets were
aggregated to hourly time steps by simple averaging to reduce noise and
homogenized with a common hourly timestamp index. The sap flow velocity
signal of the four probes was highly correlated (Pearson's
The methodological hypothesis followed in our analysis is that pressure dynamics in the soil saturated layer are the sum of separable horizontal (soil throughflow) and vertical (snowmelt and transpiration) water fluxes. Furthermore, we posit that local daily pressure variations are induced by superimposed cyclic snowmelt and evapotranspiration pulses and later modified by processes that delay or dampen the signal, such as nighttime snowpack refreezing, snowpack depth, or changes in soil absorptivity.
Our analysis of the hillslope fluxes follows the work by
In our site the soil saturated layer is very local and inflows of water to recover pressure losses at a given location come from nearby, so unlike in Loheide's concept we expect that pressure variations at a point and at the recovery source covary. The effect of this is that the overall rate of change of the water table is highly variable and complex with periods of increase and decrease as the extent of the soil saturated layer grows and shrinks. Furthermore, in addition to diurnal fluctuations by plant water uptake during the growing season, we also have to consider diurnal increases in the water table caused by snowmelt. In our case, then, the regional aquifer needs to act not only as a recovery source, but also as a sink of excess pressure.
Omitting deep percolation water losses to bedrock, the mass balance for the
soil saturated layer at any point in the hillslope is
We argue that the local time trend of the water table time series is an
approximation of
Furthermore, in the context of this paper, infiltration is primarily
generated by snowmelt, which, like evapotranspiration, is responsive to the
diurnal solar cycle. In that case,
The hour of daily maxima was calculated from the detrended time series as the time the maximum detrended flow occurred each day. Because of the circular nature of hourly times series, a late diurnal peak of a given day can occur early in the morning of the following day. To solve this problem, and maintain ordinal consistency we unwrapped these peak times into the following day by adding 24 h when peaks in the morning are preceded by peaks in the evening.
We also calculated rolling (time-evolving) correlation coefficients between daily total snowmelt and daily maximum groundwater level, between maximum daily groundwater levels and maximum diurnal stream levels, between maximum daily transpiration and minimum daily groundwater level, and between maximum daily transpiration and minimum streamflow levels. We used groundwater levels measured at well 4 because it records the clearest series of diurnal fluctuations. The rolling coefficients for each day were calculated with the points within a 16-day moving window centered at the specific day.
Finally, we plotted the empirical phase space of the hillslope–stream system using the detrended time series of well 4 and streamflow. The components of the velocity vectors showing the trajectories in the phase-space plane were calculated by differentiating the respective detrended pressure head time series.
Atmospheric forcing (Fig.
Figure
Transpiration – observed as relative sap – increased from mid-May,
when minimum daily temperatures were consistently above freezing and daytime
periods grew sufficiently long (Fig.
The daily response of wells to melt inputs varied according to their relative
position along the hillslope, but the seasonal response of all wells exhibits
a similar pattern (Fig.
From 10 May to 25 June, large sections of the hillslope were fully saturated. Because the soil water storage capacity was exhausted in some parts of the hillslope, diurnal snowmelt pulses did not contribute to the soil saturated layer. During this period melt pulses were less clearly recorded at some wells. Saturation conditions decreased during cold front 2 and cold front 3, which produced a refreezing of the snowpack, a cessation of snowmelt inputs, and a recession in the soil saturated layer. During these cold fronts water level diurnal fluctuations disappeared or were heavily dampened in all wells. From 25 June to 10 July, well levels receded when the snowpack melted out and water inputs into the hillslope ceased. In general, and as a result of upslope water subsidies, downslope wells decreased at a slower rate and had more extended recessions than upslope wells. Note, however, that well 4 starts draining earlier than well 3 situated upstream. In general this well showed higher sensitivity to periods of no snowmelt indicating higher draining capacity, perhaps because bedrock permeability at this location is higher and the soil loses water to the bedrock aquifer at a faster rate.
Time series of atmospheric inputs and hydrologic states;
During the melt season stream stage responded to diurnal energy inputs and to
the level of soil saturation (Fig.
When the water supply to the stream from the soil saturated layer ceased,
soon after the snowpack melted, streamflow quickly receded to its base flow
levels (Fig.
Detrended versions of the time series presented in Fig.
Diurnal cycles caused by snowmelt have a sharp rise and a gradual decline
Another clear indication that transpiration is extracting water from the soil
was in its relationship with minimum daily flows. The magnitude of the daily
snowmelt pulses was directly correlated with the magnitude of the water
level maxima in the soil. Also, water level maxima were correlated with
streamflow maxima. Similarly, maximum daily transpiration was inversely
correlated with diurnal minima in the soil and the stream (Fig.
Detrended water levels for wells 1 to 5
Declining diurnal snowmelt pulses were associated with a decline in the
amplitude of the diurnal groundwater pulses. In general, groundwater
oscillations had larger amplitude during the first half of May, but a trend
can only be evaluated in well 4 because the soil at the other wells was
close to saturation during significant parts of June and did not fully
register diurnal snowmelt events during much of the study period. Because
amplitudes are truncated in these wells, we did not analyze their
oscillations. At well 4 (Fig.
Black circles indicate the hour of the day when daily pressure peaks
occur in wells
Additive interference of two sinusoidal pressure signals, one caused
by transpiration
Moving correlations between the following variables. Red dashed line: total diurnal snowmelt and maximum daily water level at well 4; red solid line: maximum daily water level at well 4 and maximum water level in stream; blue dashed line: maximum daily transpiration and minimum daily water levels at well 4; blue solid line: maximum daily transpiration and minimum water level at stream. Well 4 was chosen because it contains the largest number of diurnals. Correlations for each day were performed using the points within a centered 16 day window.
Assuming that fluxes in the soil and stream are proportional to water levels
in a roughly linear way, the ratio of the amplitude of the diurnal signal to
the amplitude of the local trend can be used to approximate the contribution
of vertical fluxes to the total mass balance in the soil or stream
(Fig.
Blue dashed line, left axis: estimation of the contribution of
diurnal (vertical) fluxes to lateral (horizontal) fluxes at
well 4
Daily replenishing time during the melt season. Each marker indicates the time taken by daily radiative inputs to compensate for nightly radiative losses during the previous night. This calculation provides an estimate of the amount of time required each day to return the snowpack to the water output phase. The summer solstice is highlighted with a vertical line.
Radiative exchanges are a major driver of diurnal cycles in high-elevation
environments, controlling snowpack energetics and transpiration. Stream
recharge from diurnal melt events had a prompt, same-day response to
radiative forcing, indicating a strong atmosphere–snowpack–soil–stream
hydraulic connection very sensitive to the day–night energy exchanges at the
top of the snowpack. Turbulent energy exchanges (i.e., latent and sensible
heat) between the snowpack and the lower atmosphere in semi-alpine
environments are arguably of secondary importance to radiative exchanges
An earlier onset of snowmelt output caused by a declining nighttime energy
deficit does not explain why water table and streamflow daily maxima peaked
earlier as the melt season advanced. Rather, changes in snowpack thickness
and residual saturation, changes in the thickness and saturation of the soil
unsaturated zone, and the continuity of the soil saturated layer determine
how quickly water moves from the top of the snowpack to the stream. In the
subalpine hillslope instrumented for this study, the unsaturated zone was
minimal or nonexistent during the spring melt when hillslopes were mostly or
entirely saturated. Therefore, the reduction in the delay from the moment the
snowpack was isothermal and saturated to the moment when groundwater peaks
occur from May to early June (Fig.
Reconstruction of the phase portrait of the hillslope–stream system from the record of observations. The states are the detrended hydraulic heads at well 4 and at the stream. The lines and the vectors indicate the trajectory and velocity of the states. Colors indicate the time of day at which the system was at that corresponding state.
Despite the variability in the timing among the five wells, they show a timing
pattern very similar to that in
Once the daily water pulse is in the saturated layer of the soil, it moves
downhill toward the channel. Although the five piezometers were installed
following the downhill gradient and at relatively short distances from each
other, we did not find clear patterns or consistent lags in the timing of
peaks between the five wells in the transect. This may reflect water losses
by seepage into the bedrock and the multiplicity and evolving dynamics of
groundwater flow paths in the soil–bedrock interface
Regardless of the source, diurnal fluctuations only appeared in the stream when a saturated layer was present in the soil. Correlations between transpiration and snowmelt fluxes and water level extrema in the soil and the stream, as well as recognizable interference patterns in the diurnal signals, provided evidence of the superposition of both signals, but we did not find more direct indications of the balance of individual contributions such as recognizable changes in the symmetry or the emergence of multimodality in water level diurnal cycles. A reason for this is that the strength of snowmelt and transpiration signals is different. Water inputs into the soil from snowmelt in a typical day are on the order of tens of millimeters (10–30 mm per day), while transpiration losses are much smaller, on the order of a few millimeters (2–4 mm per day). With an extensive snowpack on the ground, snowmelt fluxes dominate diurnal hillslope storage fluctuations. This and the varying interaction between the signals due to the shift in their timings make it difficult to directly observe or determine the individual contributions of diurnal water inputs and uptakes on the observed hillslope response.
Attribution was further complicated by the fact that no diurnal stream
fluctuations were observed in summer flows, when the saprolite supplied water
to the stream, suggesting that either the main source of water for
transpiration was the soil or uptake from the saprolite was minimal.
However, it is increasingly recognized as the role that bedrock moisture has in
sustaining plant transpiration
Diurnal fluxes accounted for between 2 % and 12 % of the pressure variations
in the stream, with the remaining being contributions from older water stored in
the soil and saprolite. These diurnal contributions are on the lower end of
the 5 % to 25 % reported for high-elevation rivers in the western US during
the melt season
The amplitude of the oscillations in the hillslope was dampened as the season
progressed, because the recovery of cold content in the snowpack during
nighttime decreases and reduces the period when melt is shutoff, but also by
changes in the absorptivity and specific yield of the soil unsaturated layer
as evapotranspiration dries the soil. The hydraulic capacitance (specific
moisture capacity) of wet soil is relatively low compared to dry soils, which
magnifies the oscillations of soil water levels in response to diurnal
inputs and outputs. As the unsaturated zone dries and its specific yields and
soil absorptivity increase, oscillations in the soil saturated layer are
damped and the trajectories in Fig.
The periodic water level trajectories presented in Fig.
Although the pressure variations observed in the hillslope–stream system are
small in absolute terms, they are a significant portion of the total
hydraulic head in soils and streams of low-flow systems and control, among
other things, the arrival of water pulses, which maintain the perennial
nature of some first-order streams and extend the upslope extent of riverine
systems
Low-flow, snow-dominated headwater catchments are sensitive and therefore vulnerable to alterations in energy and precipitation regimes, highlighting the importance of understanding linkages between local atmospheric, hillslope, and fluvial processes. In this paper we presented and analyzed a unique, high-temporal-resolution observational dataset that contributes to advancing our understanding of how diurnal snowmelt and transpiration cycles drive the hydrology of a snow-dominated, semi-alpine headwater catchment during the melt and growing seasons.
We found that in this type of environment the snowpack mediates the hydrologic response of hillslopes to atmospheric drivers by (i) accumulating cold content at night and during cold multi-day storms and (ii) becoming an energy sink during the day until such deficit is replenished. Overcoming this energy hurdle delays the production of snowmelt into the soil by several hours, especially early in the melt season. The freeze and thaw cycles create pulse-like infiltration events that induce substantial pressure oscillations in the soil storage system, which are subsequently transmitted to the stream. As the thickness of the snowpack declines with the melt season, the travel times of melt water to the base of the snowpack and to the soil saturated layer are reduced, shifting diurnal pressure peaks progressively to early in the day. Changes in the energy balance at the surface of the snowpack can potentially reduce melt response times and accelerate the hydrologic response of headwater catchment, with potentially ecological and hydrologic implications not only for first-order catchments, but also for the timing of spring freshets in higher-order downstream hydrologic systems.
Even in a small, well-connected system, the interactions between diurnal solar cycles, snowmelt, and the response of water levels in the hillslope and the stream showed high variability and complexity. With the onset of the growing season, the timing and dynamics of soil and stream water oscillations are further altered by diurnal plant-water-uptake cycles. We interpreted the beat-like patterns in the observed water pressure oscillations as evidence of the interaction between these two periodic signals (snowmelt and transpiration). The existence of this interaction is further corroborated by a correlation analysis between diurnal snowmelt and sap velocity cycles and diurnal maxima and minima in soil and stream pressures. Although this interaction was indirectly detected, we could not disentangle the direct contributions of diurnal snowmelt inputs and transpiration uptakes on the production of diurnal pressure fluctuations in the soil and stream system, among other things because snowmelt and transpiration cycles are correlated and driven by the same solar cycle. This difficulty in disentangling and attributing the contribution of each of the signals is further complicated by the dominance of daily snowmelt inputs, which are on average on the order of tens of millimeters per day, while diurnal transpiration uptake rates are expected to be on the order of 2 to 4 mm per day in this type of environment. This dominance of snowmelt inputs throughout the melt season may be a reason we did not find differences in the shape of diurnal soil and stream water oscillations at the beginning and at the end of the snowmelt season. Changes in the symmetry of the oscillation have been reported in the literature for studies conducted in riparian zones and lower-elevation catchments when oscillations transition from being snowmelt driven to transpiration driven.
Snowmelt and transpiration pulses are tied to the rotation of the earth, and while the period of these cycles is not expected to change significantly in the future, their intensity can change with alteration in the amount of energy inputs from the atmosphere. Our analysis further showed that the hydrologic response of the snow–hillslope–stream system is highly sensitive to atmospheric drivers at hourly scales and that variations in atmospheric energy inputs or other stresses are quickly transmitted and alter the intensity, duration, and timing of snowmelt pulses and soil water extractions by vegetation, which ultimately drive variations in soil and stream water pressures. From our analysis it was also clear that at small scales (well to hillslope) our high-elevation semi-alpine hydrologic system did not damp high-frequency pressure variation as quickly as the local diffusive nature of flow in porous media may have lead us to believe. Furthermore, we did observe the soil damping mechanisms varied in intensity during the season, and we discussed that, when the unsaturated one is wetter, pressure oscillations are propagated with less attenuation. Clearly, the speed and damping of pressure variations in the system will depend on the soil physical characteristics and the geometry of the hillslope, but we show here that high-frequency pressure oscillation in the hillslope and stream systems can be easily induced and that these may play an important environmental role that warrants further research.
All data used in this study are available upon request. Address inquiries to the corresponding author.
BWB installed field equipment, collected data, and drafted the paper. MPM designed the study and helped with field data collection. All authors contributed to data interpretation and writing the manuscript.
The authors declare that they have no conflict of interest.
This article is part of the special issue “Assessing impacts and adaptation to global change in water resource systems depending on natural storage from groundwater and/or snowpacks”. It is not associated with a conference.
This material is based upon work supported in part by the National Science Foundation EPSCoR Cooperative Agreement IIA-1443108 and EPS-1101342. Edited by: David Pulido-Velazquez Reviewed by: two anonymous referees