Abstract: More than 85% of NO3− losses from watersheds in the northeastern United States are exported during winter months (October 1 to May 30). Interannual variability in NO3− loads to individual streams is closely related to interannual climatic variations, particularly during the winter. The objective of our study was to understand how climatic and hydrogeological factors influence NO3− dynamics in small watersheds during the winter. Physical parameters including snow depth, soil temperature, stream discharge, and water table elevation were monitored during the 2007-2008 winter in two small catchments in the Adirondack Mountains, New York State. Snowpack persisted from mid-December to mid-April, insulating soils such that only two isolated instances of soil frost were observed during the study period. NO3− export during a mid-winter rain-on-snowmelt event comprised between 8 and 16% of the total stream NO3− load for the four-month winter study period. This can be compared with the NO3− exported during the final spring melt, which comprised between 38 and 45% of the total four-month winter NO3− load. Our findings indicate that minor melt events were detectable with changes in soil temperature, streamflow, groundwater level, and snow depth. But, based on loading, these events were relatively minor contributors to winter NO3− loss. A warmer climate and fluctuating snowpack may result in more major mid-winter melt events and greater NO3− export to surface waters.
Nitrogen (N) availability in forested watersheds controls many aspects of ecosystem function, including forest productivity, competitive relationships between tree species, and increased N pollutant exports to receiving waters (Vitousek et al., 1997). Excess N losses from forests can reduce water quality downstream, negatively impact plant function and growth, contribute to forest decline, and lead to changes in forest species composition (Aber et al., 1989, 2003; Mitchell, 2011). Despite the detrimental impacts of excess N, the controls on N export from forested watersheds are not well understood, particularly during the winter, in catchments that have long periods of snow cover. Although regional variation in stream NO3− concentrations and associated export are related to atmospheric N deposition and soil nitrification rates, annual NO3− loads in individual streams may be more closely correlated to interannual climatic variations (Murdoch et al., 1998; Park et al., 2003; Ross et al., 2012). Hence, other factors besides deposition, including changes in climate, need to be evaluated to understand intraannual and interannual variations in stream NO3− loads.
Although the winter season is sometimes considered to be “dormant,” due to the cold temperatures, vegetation dormancy, and persistent snow cover, winter microbial processes can have a critical impact on N cycling in the Northeast United States (U.S.) (Campbell et al., 2005). Watersheds throughout the Northeast U.S. export more than 85% of the annual NO3− loss during the dormant season (Mitchell et al., 1996) with most of this export occurring during spring snowmelt (Campbell et al., 2005) and mid-winter rain-on-snow events (Casson et al., 2010). Changing winter climate in the Northeast, which may include warmer temperatures and changes in precipitation (Rosenzweig et al., 2011a,b), will alter snowpack dynamics (Rosenzweig et al., 2011a,b) and thus likely influence N cycling and hydrologic transport to stream channels (Campbell et al., 2005). Higher temperatures in winter may reduce the amount and duration of snowpack, which insulates soils from very cold above-ground temperatures (Hardy et al., 2001). Snow cover’s insulating properties can maintain soil temperatures sufficiently high to allow root growth, microbial respiration, and other biotic activities to continue (Campbell et al., 2005; Groffman et al., 2009). If snowpack is absent, soils are more likely to freeze resulting in physical and chemical changes, including the killing of fine roots and the alteration of soil microbial processes (Tierney et al., 2001). Monson et al. (2006) have observed that, in winters with high amounts of late winter snow, soil temperatures are higher and ecosystem respiration rates are higher, compared to years with less late winter snow. Also, as temperatures fluctuate around freezing, mid-winter melt events and rain-on-snow events alter hydrological pathways and the transport of NO3− from shallow soils, increasing winter NO3− loads in streams. Soil freezing can modify winter hydrology, resulting in reduced infiltration and recharge during snowmelt events, although this is variable depending on the spatial patchiness of soil freezing, the timing of soil thaw relative to melt events, and other factors (Shanley and Chalmers, 1999). Individual rain-on-snow events can contribute as much as 40% of annual NO3− export from forested watersheds in the Northeast U.S. and the contribution of rain-on-snow events to annual and winter NO3− loads has generally increased in recent decades (Eimers et al., 2007). Warmer winters have more rain-on-snow events than cooler winters (Casson et al., 2010) and winters with more rain-on-snow events have earlier export of NO3− (Casson et al., 2012), so rain-on-snow events may play a critical role in the observed correlation between winter temperatures and annual NO3− loads (Park et al., 2003).
Even a small winter temperature increase of a few degrees may have a considerable impact on Adirondack ecosystems of New York State, where winter air temperatures hover around freezing and the form of precipitation (i.e., snow vs. rain) can be markedly affected by small shifts in air temperature (Campbell et al., 2005). A regional climate study in the Adirondack Mountains and surrounding area found that, over the past 50 years, there was a general increase in precipitation of 105.7 mm along with an increase in annual minimum temperatures of 0.76°C (Dello, 2008). In addition, winter temperatures increased by 0.42°C over this period. As a result of warmer temperatures in the Northeast, less winter precipitation has fallen as snow and more as rain (Frumhoff et al., 2007) and some winter months including February have been drier (Dello, 2008). It has been predicted that, during the current century, the Adirondacks region will experience a decrease in snowpack duration and depth (Hodny and Mather, 1995) and a 25 to 50% decrease in “snow-covered” days (Frumhoff et al., 2007). Past research suggests that snowpack depth and duration control soil temperature, which in turn controls the amount of inorganic N sequestration in soil (Brooks and Williams, 1999). Precipitation as rain infiltrates and flushes soils, but with precipitation as snow there is no infiltration until snowmelt, allowing ions to accumulate in the soil (Campbell et al., 2006).
Although NO3− dynamics in forested watersheds in winter are of great interest, detailed data sets of winter biogeochemistry and hydrology are generally lacking due to the logistical challenges of working in forested watersheds in winter. The Huntington Wildlife Forest, located in the Adirondack Mountains of New York State, has been the site of numerous studies linking hydrology and biogeochemistry (Park et al., 2003; Campbell et al., 2006; Christopher et al., 2006, 2008; Piatek et al., 2009). Studies on watershed biogeochemistry have focused on the Arbutus Watershed, including Archer Creek, the major inlet to Arbutus Lake. In the upper reaches of the Archer Creek Watershed, the stream waters of two small catchments (WS14 and WS15) have markedly different NO3− concentrations (Christopher et al., 2006).
The first objective of this study was to observe relationships between temporal patterns of air and soil temperatures, snowpack depth, soil freezing, and NO3− concentrations in stream, soil, and groundwater during the winter months in two paired catchments with markedly different NO3− concentrations in soil and associated waters. We hypothesized that NO3− is hydrologically flushed from soils during melt events, contributing to higher winter and annual NO3− export rates overall. To address this hypothesis, our second objective was to compare stream chemistry with soil and groundwater chemistry in two small catchments (WS14 and WS15) during melt events. We also compared event type and timing with the observed NO3− concentrations during the winter months. This research has yielded one of the most spatially and temporally detailed characterizations of winter watershed hydrology and nitrogen biogeochemistry currently available and therefore provides unique insights on winter processes.
WS14 and WS15 are two adjacent headwater catchments within the Arbutus Lake Watershed at the Huntington Wildlife Forest (Figure 1; Table 1). The Huntington Wildlife Forest is located in the Adirondack Park near the town of Newcomb, New York. The regional climate is humid and cool, receiving a mean of 104.6 cm of precipitation annually between 1940 and 2007. Mean monthly precipitation ranges from 7.6 cm in January to 9.6 cm in November (Campbell et al., 2006). During winter, snow cover persists on average 131 days with the average maximum snow depth of 77.9 cm occurring in February and average yearly snowfall totaling 303 cm (Campbell et al., 2006). The monthly mean maximum daily air temperature, as recorded on site from 1941 to 2004, was lowest in January, at −4°C, and highest in July, at 25°C (Campbell et al., 2006). The long-term, regional climate values reported here are based on data collected at the Newcomb NOAA Meteorological Station, which has operated since 1941. Since 2007, meteorological data, including air temperature, snow depth, and precipitation, has been recorded continuously at the Ackerman clearing tower, which is adjacent to Catchments WS14 and WS15 (Figure 1). Meteorological data reported in the results were collected at the Ackerman clearing tower.
Mean stream NO3− concentration during winter3 (mg/l)
Tree species composition1,2
More base-rich indicator species (sugar maple, eastern hophornbeam, American basswood)
More base-poor indicator species (American beech, eastern white pine)
WS14 and WS15 are similar with respect to their physical dimensions (Table 1) and respond similarly to hydrologic events. The two catchments have a similar topographic index frequency distribution, which suggests that they should have similar source areas during hydrologic events (Christopher et al., 2006). Indeed, the observed relative contributions of soil water and groundwater sources to streamflow are similar in WS14 and WS15 (Christopher et al., 2008), with winter streamflow primarily derived from groundwater during base flow with increasing contributions of soil water sources during melt events. Christopher et al.’s (2008) conceptual model of streamflow generation during melt events in WS14 and WS15 follows a typical soil flushing model; the water table is low in elevation before snowmelt, and streamflow is primarily sourced from relatively NO3−-poor groundwater. During snowmelt, the water table rises and stream water is increasingly derived from relatively NO3−-enriched shallow soil water. Some subtle differences in hydrology, specifically the relative differences in contribution of different solute sources, between the two catchments have been noted (Christopher et al., 2008); WS14 is located in an upland ridge top area, which has been attributed to increasing the movement of seep and groundwater flow to the stream (McHale et al., 2002; Mitchell et al., 2006; Christopher et al., 2008).
Catchments WS14 and WS15 have been of particular interest in recent work (Campbell et al., 2006; Christopher et al., 2006, 2008; Page and Mitchell, 2008) because although the adjacent catchments have very similar physical characteristics (e.g., slope, aspect, elevation, and size), they have markedly different soil and water chemistry and tree species composition (Christopher et al., 2006). The differences in water chemistry between the two catchments are due to differences in the soil composition and tree species distribution, rather than differences in physical hydrology (Christopher et al., 2006). WS14 has more sugar maple (Acer saccharum Marsh), eastern hophornbeam (Ostrya virginiana (Mill.) Koch), and American basswood (Tilia Americana L.) (base-rich indicator trees) and fewer American beech (Fagus grandifolia Ehrh.) and eastern white pine (Pinus strobus L.) (base-poor indicator trees), which is due to the much higher soil Ca2+ concentrations in WS14. Sugar maple has been linked to higher net nitrification in soils resulting in the surrounding soils having higher NO3− concentrations (Lovett and Mitchell, 2004; Page and Mitchell, 2008). The NO3− in soil water is transported to the stream, contributing to WS14’s higher stream water NO3− concentration (Christopher et al., 2006, 2008). The local bedrock is granite gneiss and gabbro-amphibolite from the Precambrian period overlain with Wisconsin glacial till and shallow soils (∼1 m) consisting of Becket-Mundall series sandy loams (Christopher et al., 2006).
Field Measurements of Snow Density and Depth
Snow depth and density were measured approximately every two weeks from December 19, 2007 through March 31, 2008 and more frequently during the major end-of-season snowmelt (April 1 to April 19, 2008). Snow course transects were established perpendicular to the streams to capture a range of elevations along the hillslopes in each of the two catchments (Figure 1). The snow courses were 10 m in length and were sampled at 1-m intervals using an Adirondack Snow Sampling Tube (http://www.rickly.com). At each interval, the snow depth was recorded and a snow core was collected and weighed to calculate snow water equivalence (SWE) as
where SWE is the snow water equivalence (m), d is the snow depth (m), ρs is the snow density (kg/m3), and ρw is water density (1,000 kg/m3). Snow density was calculated as
where Ms is the mass of the snow core (kg) and Vs is the volume of the snow core (m3). The average SWE for each catchment on each sample day was computed as the average of the 10 measurements taken along the respective snow course. In addition, every other snow core in each snow course was collected, combined, and returned to the laboratory for chemical analysis.
Groundwater and Stream Water Sampling
The stream outlets in WS14 and WS15 and five water table wells in the two catchments were equipped with Campbell Scientific CR1000 data loggers and pressure transducers (Campbell Scientific, Logan, UT) to record water height every 15 min. The streams in WS14 and WS15 are equipped with 90° V-notch weirs at their outlets (Figure 1), and rating curves were established for each weir for calculating discharge from stage height. The two rating curves were established by calibrating the standard weir Equation (3) using known discharge rates and the corresponding water-level heights that were recorded during the summer and fall of 2008. Discharge rates were calculated as
where Cw is a weir coefficient (unitless), g is the acceleration due to gravity (9.8 m/s2), θ is the angle of the weir (1.57 radians), and h is the water height above the crest of the weir (m). Values of 0.450 and 0.725 were derived for Cw for WS14 and WS15, respectively. Discharge rates were normalized to watershed area for cross-watershed comparisons.
Stream water samples were collected from WS14 and WS15 approximately every two weeks between December 19, 2007 and April 19, 2008 and at a much higher frequency during snowmelt. During routine winter sampling, grab samples were taken from the overflow of the weirs. During snowmelt events, time-activated ISCO automated samplers (Teledyne Technologies Incorporated, Lincoln, NE) were programmed to sample from the streams several times a day in order to obtain a more detailed characterization of stream chemistry during snowmelt.
Stream NO3− loading () was calculated for each stream from December to April as
where [NO3−]s is the stream NO3− concentration (kg/l) and Qs is the streamflow (L/ha/day). During periods of biweekly sampling, observed NO3− concentrations were assumed constant for the week prior to and following sampling. During periods of subdaily sampling, 15-min streamflow data were used with subdaily NO3− concentrations to compute loads for each stream water sampling interval. These daily and subdaily loads were summed to get total loads for specified time intervals during the study.
Six water table wells (W33, W34, W35, W36 in WS14, and W12 and W13 in WS15), installed between 2000 and 2002, were used for this study because they were installed several years prior to the field investigation, leaving the surrounding soil undisturbed since installation (Figure 1). Water table wells have screen lengths between 0.31 and 0.59 m screened at depths from 0.44 to 2.32 m below the land surface in glacial till (Table 2). All wells except W13 were equipped with automated data loggers, which recorded water height every hour during the study period. Well casings have not been surveyed to a common datum, so well water heights are reported relative to the lowest level observed during the study period. A groundwater sample was collected for chemical analysis at each well approximately biweekly during winter and more frequently during snowmelt.
Table 2. Groundwater Well Characteristics.
Screen Length (m)
Screened Depth Interval (m)
Mean NO3− Concentration (mg/l)
Soil temperature was measured by temperature data loggers installed in profile at depths of 5, 15, 30, and 50 cm in the soil at six “microsites” within each watershed (Figure 1). We used iButton® temperature data loggers (Maxim Integrated Products, Dallas Semiconductor, Inc., Sunnyvale, CA) programmed to record temperature at 3-h intervals from November 3, 2007 through April 19, 2008. Two iButton models were used, DS1921G (0.5°C resolution and ±1.0°C accuracy) and DS1921Z (0.125°C resolution and ±1.0°C accuracy). In addition to the temperature data loggers, each of the 12 microsites consisted of a deep (50 cm) and shallow (15 cm) soil lysimeter and a “frost tube.”
Depth of soil freezing was measured approximately every two weeks from December 19, 2007 through March 31, 2008 and more frequently during the major end-of-season snowmelt (April 1 to April 19, 2008). One “frost tube” was installed at each microsite to provide a more vertically precise frost depth than provided by the temperature measurements (Ricard et al., 1976). An advantage of frost tubes is that they are able to record any depth of frost in a vertical column, rather than simply recording the temperature at fixed depth intervals. Frost tubes were assembled according to Ricard et al. (1976) and consisted of a piece of clear plastic tubing filled with methylene blue dye, capped at both ends, and installed vertically in the soil to 50 cm. Methylene blue indicator dye is blue at temperatures above freezing and turns clear when frozen. Frost depth was manually measured from frost tubes approximately every two weeks during the winter and more frequently during snowmelt.
Soil Water Chemistry
Soil water samples were collected with 5-cm diameter porous-cup tension lysimeters that were previously installed at the 12 microsites during 2002 (Campbell et al., 2006). Preexisting lysimeters were used to ensure that the soil had adequate time to recover from any disturbance that may have occurred during installation. The lysimeters were installed in pairs, one to collect shallow water at 15 cm depth and one to collect deeper water at 50 cm depth. Six pairs of lysimeters in both WS14 and WS15 were used, for a total of 12 lysimeter pairs. All lysimeters were sampled approximately every two weeks between January 7, 2008 and April 19, 2008. During snowmelt (April 1 to April 12, 2008), lysimeters were sampled every 24 h. On occasion, lysimeter samples were not collected at specific sites because the lysimeters did not produce water due to dry soil or because the lysimeter tubing became frozen, preventing sample collection. Each lysimeter was pumped dry prior to evacuation of air to ensure that any residual water in the lysimeter would not be included in the sample, after which 40 kPa of tension was applied to the lysimeters at least 6 h prior to sampling.
Air temperature, precipitation, and snow depth were recorded every 15 min at the Ackerman Clearing Meteorological Station (Figure 1). Air temperature was measured by a Campbell Scientific 107-l temperature probe. The probe has a Steinhart-Hart error of ±0.01°C and an interchangeability error of ±0.2°C. Snow depth was measured by a Campbell Scientific SR50A-L CSC Sonic Ranging Sensor, which has a resolution of 0.25 mm and an accuracy of ±1.0 cm or 0.4% of distance to target (whichever is greatest). Precipitation was measured by a TE525WS tipping-bucket rain gage (Texas Electronics, Dallas, TX), which has an accuracy of ±1% up to precipitation rates of 2.5 cm/h.
Chemical Data Analysis and Sample Preparation
Within 24 h of snowpack, soil water, groundwater and stream water sample collection, a WTW® 340i multiparameter field probe (WTW Wissenschaftlich-Technische Werkstätten GmbH, Weilheim, Germany) was used to measure pH and specific conductance. All samples were then filtered through 0.45-μm precombusted glass fiber filters. Water samples were then transported on ice to the laboratory, where they were stored in the dark at 3°C until they were analyzed. A Dionex ICS-2000® ion chromatograph (Thermo Scientific, Sunnyvale, CA) was used to measure nitrate (NO3−) and sulfate (SO42−). A Tekmar-Dohrmann Phoenix 8000® total carbon analyzer (Teledyne Tekmar, Mason, OH) was used to measure dissolved organic carbon (DOC). A Perkin-Elmer® OPTIMA 3300DV Inductively Coupled Plasma Emission Spectrometer (Perkin Elmer, Waltham, MA) was used to measure total calcium (Ca2+) and magnesium (Mg2+).
Air Temperature and Snow Cover
Average daily air temperatures ranged from −21.0 to 16.1°C, with the warmest temperature recorded during the final day of the study period (Figure 2). The average air temperature per month during the study period was always below freezing, except during April, the month of the major spring snowmelt (Table 3). During the 123-day study period, from December 19 through April 19, the average daily temperature was above freezing for 29 days, 5 of which occurred in early January and 16 of which occurred in mid-April.
Table 3. Summary of Meteorological Data for the 2007 to 2008 Winter Season in the Arbutus Lake Watershed from the Ackerman Meteorological Station.
Average Daily Air Temperature (°C)
Average Snow Depth (mm)
Total Precipitation (mm)
*Includes only April 1 through 19.
In both catchments, increases in SWE were positively correlated with increases in snow depth. The snow density on average was >250 kg/m3 when snow depth was decreasing during melt events. Outside of the melt events, the snow density increased linearly from about 150 kg/m3 in late January to about 280 kg/m3 in early April. Detailed snow course measurements taken in WS14 and WS15 indicate that the snow depth data recorded by the Ackerman Clearing Meteorological Station were representative of the average snow depths in both catchments (Figure 2). The average percent differences between the snow depth measurements at the snow courses and the snow depths recorded at the meteorological station were 5.6% for WS14 and −12.4% for WS15. The higher temporal resolution data from the meteorological station were used when actual measured snow course results were not available.
Snow cover was continuous from November 15 until April 15, with a peak snow depth of 81.3 cm observed at the Ackerman Clearing Meteorological Station. Some intermittent decreases in snow (i.e., >30 cm decline in snow depth) occurred mid-winter, but never resulted in total loss of snow in either catchment (Figure 2). The length of continuous snow cover at the Ackerman Clearing Meteorological Station (141 days) during our study was longer than the long-term (1941-2004) mean length of continuous snow cover (131 days) and the maximum snow depth on March 3 (81.3 cm) was greater than the long-term mean maximum snow depth (77.9 cm) (Campbell et al., 2006). From the snow course measurements, WS14 averaged a snow depth 25% deeper than in WS15.
One notable mid-winter melt event occurred between January 7 and 12, 2008, in addition to the final major spring snowmelt in April (Figure 2). Melt events corresponded with the elevated air temperatures, specifically when the mean air temperature was above 2°C for two days or longer (Figure 2). During the January 7-12 melt event, average daily temperatures were as high as 11.8°C and remained above freezing for five of the six days. A total of 41.2 mm of precipitation was also recorded during this melt event. As a result of this “rain-on-snow event,” snow depth decreased by about 0.3 m, snow density doubled from 150 to 280 kg/m3, and SWE decreased by about 30 mm.
Soil Temperature and Freezing
Soil temperatures remained at or above 0°C at all observed depths (5, 15, 30, and 50 cm) throughout the winter at 11 of the 12 microsites (Figures 3 and 4). The one microsite that experienced subzero soil temperatures, Site 1D in WS15, did so only for short periods of time (i.e., <9 h) at a shallow depth of 5 cm during periods when there was presumably infiltrating melt water. Site 1D, in WS15, recorded temperatures at or below zero at a depth of 5 cm on December 23, February 5 and 18, March 18, and April 1, 8, and 9. On those days, average air temperatures hovered around 0°C, presumably generating some melting of snow and short-term infiltration of near 0°C melt water. During those events, several other sites recorded short-term decreases in soil temperature from the previous three-day average, but all of these temperature changes were <1°C and no temperatures fell below 0°C (Figures 3 and 4). Interestingly, no sites experienced subzero soil temperatures during or immediately following the major rain-on-snowmelt event that occurred between January 7 and 12. The largest mid-winter melt event occurring during the 2007-2008 winter at this site was not associated with any observed soil freezing.
Data loggers at all 12 sites, at all four depths, recorded rapidly declining temperatures until the end of December when temperatures remained stable due to the development of a snowpack >50 cm (Figures 3 and 4). These soil temperatures remained steady until mid-winter snowmelts occurred, which resulted in soil temperatures changing from the previous three-day average by ≥ 0.12°C. Although the soil temperature was very responsive to snowmelt, only two incidents of freezing in the frost tubes were recorded. At Site 8A, in WS15, a frost depth of 0.5 cm was observed on December 19 and at Site 6C, in WS15, a frost depth of 2.5 cm was observed on January 25. The average air temperatures for the four days leading up to and including December 19 and January 25 were −12.7 and −11.8°C, respectively, which were more than 5°C colder than the four-day average leading up to any other soil frost sampling date. The December 19 and January 25 sampling dates corresponded to two of the five coldest periods during the winter (Figure 2). No soil freezing was observed in any of the frost tubes in WS14. The observed freezing in the frost tubes did not directly correspond to those measured soil temperatures that were <0°C, but soil freezing depths observed in the frost tubes were <5 cm, the depth where temperature was recorded.
Soil NO3− Concentrations
Soil NO3− concentrations were highly variable in space and time. In both catchments the soil water NO3− concentrations at both depths (15 and 50 cm) rapidly approached a relatively uniform, minimum concentration of <5 mg/l during the final spring snowmelt (Figures 5 and 6). In general, concentrations of NO3− in shallow soil water (15 cm) were greater than at depth (50 cm), although the difference varied greatly by location and time. NO3− concentrations in soil water during the mid-winter snowmelt (January 10) did not show a consistently higher or lower concentration than other winter sampling dates (Figures 5 and 6), indicating no generalizable response of soil water NO3− concentrations to mid-winter melt. It may have been possible that some of the short-term dynamics of changing chemistry may have been missed for these smaller episodes due to our two-week sampling interval, although soil water chemistry was measured on January 10, in the middle of the largest mid-winter, rain-on-snowmelt event.
NO3− concentrations in soil varied greatly within each catchment by location and depth. Soil water NO3− concentrations in WS14 ranged from 0.1 to 16.5 mg/l, with the lowest concentration at a 50-cm lysimeter (Site 2B) and highest concentration at a 15-cm lysimeter (Site 11B) (Figure 5). Within WS14, five of the six microsites show the same pattern of a higher NO3− concentration at the 15-cm depth compared with the 50-cm depth. The one site (8D) that did not fit this pattern had similar NO3− concentrations at both depths (4.9 to 10.9 mg/l at 15 cm and 4.8 to 11.7 mg/l at 50 cm). Soil water NO3− concentrations in WS15 ranged from 0.1 to 24.3 mg/l, with the lowest concentration at a 50-cm lysimeter (Site 1A) and the highest concentration at a 15-cm lysimeter (Site 6C). Within WS15, four of the six microsites showed the same pattern as WS14, with higher NO3− at the 15-cm depth. The two locations that did not show this pattern (Sites 8A and 1D) had similar NO3− concentrations at both depths (Figure 6).
Groundwater Hydrographs and NO3− Concentrations
Water levels in all five gaged wells responded to melting snow with rapid increases in the water table height (Figure 7). Groundwater NO3− concentrations showed considerable spatial variability (Figure 7). The wells in WS14 had NO3− concentration ranges as low as 0.1 to 0.8 mg/l (W34) and as high as 7.6 to 10.5 mg/l (W36) (Figure 7). The two wells in WS15 also showed spatial variability, with NO3− concentration ranges of <0.1 to 0.2 mg/l at W12 and 1.2 to 1.7 mg/l at W13 (Figure 7). The wells in both catchments showed similar temporal response with little change in NO3− concentration during the mid-winter snowmelt events, but a marked decline in NO3− concentrations during the large final snowmelt event and a gradual monotonic decrease of NO3− concentrations from the beginning of December through the remainder of the winter. It is important to note that observations of groundwater chemistry were initiated on December 19, 2007, and therefore do not capture potential increases in NO3− concentrations in the fall or early winter.
Stream Hydrograph and Solute Concentrations
Stream hydrographs responded most markedly to the January 7-12 melt event (Figures 8 and 9) and the final snowmelt (Figures 10 and 11). During the January melt event, streamflow increased from 3.9 to 23.5 mm/day in WS14 and from 3.5 to 52.3 mm/day in WS15. During the end-of-season snowmelt in April, streamflow peaked at 59.5 mm/day in WS14 and at 88.0 mm/day in WS15. Previous work in these catchments has shown that WS15 is a flashier stream, which is not surprising given its smaller physical size (Christopher et al., 2006). During the final spring snowmelt, streamflow fluctuated diurnally due to increased melting in late afternoon, during peak solar radiation and air temperature, and reduced melting in early morning, following a night of minimal radiation and cold air temperatures (Figures 10 and 11).
The streams in both WS14 and WS15 showed changes in the concentrations of Ca2+, NO3−, and DOC during snowmelt events. Changes in stream water ion concentrations were more evident than those in soil water and groundwater for the mid-winter, rain-on-snowmelt event between January 7 and 12 (Figures 8 and 9). In both streams during the final snowmelt, the NO3− concentration increased only slightly at the onset of snowmelt but subsequently decreased to below presnowmelt NO3− concentrations whereas the hydrograph was still rising (Figures 10 and 11). The calculated NO3− load of the stream showed the same pattern, but in addition showed an increase in the NO3− load over time during the January event. The DOC concentrations respond in the same manner as the NO3−, with an increase in concentration during the January event and then again during the final snowmelt. Concentrations of Ca2+ responded by decreasing during both melt events followed by increases to premelt concentrations (Figures 8 and 9).
Snow Cover, Soil Freezing, and Temperature
Results indicate that reduced snow cover following mid-winter, rain-on-snowmelt events does not necessarily reduce soil insulation sufficiently to lower soil temperatures such that subsequent below freezing air temperatures following the melt cause soil freezing. Results also indicate that a decrease in soil temperature is an important indicator that soil is being flushed by infiltrating water from melting snow. Past studies have focused on soil freezing and how it affects soil biogeochemical processes (Brooks and Williams, 1999; Groffman et al., 2001). During our study, there was not extensive soil freezing and this was likely a function of the early snow that persisted until the middle of April (Figure 2). Frost depth data were taken every two weeks, so it is possible that some deeper freezing events were not observed using the frost tubes. Hardy et al. (2001) found that soil frost depth was strongly regulated by snow depth as a result of snow cover’s insulating properties. Although the exact snow depth required to provide insulation was not given in that study, it was suggested that winters of low snowfall and mild temperatures were enough to influence soil freezing (Hardy et al., 2001). In WS14 and WS15, the snowpack began to form on November 15, when the soil temperatures were still >3°C, and subsequently reached a depth of about 50 cm at the end of December (Figures 3 and 4). Between December 1, 2007 and April 8, 2008, snow depth was consistently >0.28 m. The winter of 2007-2008 in WS14 and WS15 provided an opportunity to study the effects of fluctuating soil temperature and soil flushing with meltwater on water biogeochemical composition in the absence of freezing events.
Soil temperatures did not respond uniformly to the major, mid-winter melt event that occurred between January 7 and 12. In some cases, soil temperatures decreased at all depths, as expected, presumably due to infiltrating melt water that was close to 0°C (e.g., Site 1D in WS15, Figure 4). In other cases, the opposite relationships were observed and all soil temperatures increased in response to the melt event (e.g., Site 6C, WS14, Figure 3). This variable temperature response suggests that soil temperature may have been affected differently within the catchments depending on the relative importance of the downward flux of melt water vs. the upward movement of relatively warmer groundwater due to a rising water table. Site 6C, WS14, is located only 5 m from the stream, a likely variable source area where the water table is expected to be very close to the land surface and may rise above the surface during melt events. The groundwater temperatures during the winter varied spatially from 2 to 7°C, but remained relatively constant over time at individual wells. At two sites (e.g., 11B, WS14) there was a rise in soil temperature response at a depth of 5 cm and a decrease in temperature at all other depths (15 to 50 cm). This response was likely due to a loss of snow cover at these individual sites that allowed for solar heating at shallow depths of 5 cm and cooling due to infiltrating melt water at deeper depths. The spatial variability of the soil temperature response to melt events suggests that drawing conclusions from plot scale studies to watershed-scale response may be problematic (e.g., Groffman et al., 2001).
Although only one large mid-winter melt event occurred between January 7 and 12, other smaller events are indicated by temporally coincident water table fluctuations (Figure 7) and soil temperature changes (Figures 3 and 4). During these minor melt events, average air temperatures were <1°C and/or above freezing for only one day. Also, snow depth decreased by <10 cm and in a few cases even increased due to concomitant snowfall during these events. The mid-winter and end-of-season snowmelt events resulted in cold water infiltrating the soil, causing soil temperature to temporarily (i.e., <7 h) decrease at many sites by as much as 2.1°C (Figures 3 and 4) reflecting the movement of meltwater through the soil. The influx of this melt water contributed to the dilution of soil NO3− as well as other solutes as the winter progressed (Figures 5 and 6).
Spatial Variability in NO3− Chemistry
NO3− concentrations in deep soil (i.e., 50 cm depth), stream, and groundwater in WS14 were generally higher than those in the WS15 catchment, despite similar NO3− snowpack concentrations (Figure 12). However, in the current study, the soil NO3− concentrations at 10 cm were not higher in WS14 when compared with WS15 as observed in previous studies. Higher NO3− concentrations in catchment WS14 were expected based on previous work by Campbell et al. (2006), Christopher et al. (2006, 2008), and Page and Mitchell (2008) which show how soil chemistry and vegetation influence NO3− concentrations in these catchments. Higher soil Ca2+ concentrations in WS14 support base-rich tree species including sugar maple, American basswood, and eastern hophornbeam (Christopher et al., 2006; Page and Mitchell, 2008). Differences in the composition of the overstory tree composition (Table 1) results in substantial differences in N mineralization and nitrification rates between the two catchments (Campbell et al., 2006; Christopher et al., 2006; Page and Mitchell, 2008; Ross et al., 2012). Higher net nitrification occurs in soils under sugar maples (Lovett and Mitchell, 2004), generating NO3−, which enters streams through various flow paths depending on the hydrological conditions (Christopher et al., 2006, 2008). In addition, the presence of basswood is correlated with higher soil NO3− concentrations in WS14 (Page and Mitchell, 2008). Basswood, a calciphilic species, benefits from the local Ca2+-rich geology of WS14 and produces leaf litter that is Ca2+ rich and labile. As a result, basswood creates patchy NO3− hotspots in locations of favorable soil pH and the presence of easily mineralized substrate for nitrification (Page and Mitchell, 2008). Spatial variability in soil NO3− was found within each of the catchments in this study, indicating the continued presence of “hot spots” of NO3− accumulation that persist over the winter. These hotspots were likely major contributors to the elevated NO3− concentrations in stream water during spring snowmelt. Previous isotopic assessments of NO3− during winter at this site indicate that there is limited transformation of NO3− in the snowpack, nitrification produces the majority of NO3− in soil water, and terrestrial sources account for the majority of NO3− in stream water (Campbell et al., 2006).
Soil water NO3− concentrations are generally higher in catchment WS14 than WS15, but both catchments have high spatial variability with “hotspots” for N mineralization, nitrification, and/or denitrification found in both catchments. Soil water NO3− concentrations are as high as 16.5 and 24.3 mg/l in WS14 and WS15, respectively, and as low as 0.1 mg/l in both catchments (Figure 12). The “hotspots” with high NO3− concentration ranges fit Campbell et al.’s (2006) observations that higher rates of nitrification are occurring at specific locations in these catchments, whereas locations with low NO3− concentration ranges suggest the variable importance of denitrification and/or microbial uptake of nitrogen. The importance of spatial variability of nitrification as well as denitrification, which would result in the removal of NO3−, has been demonstrated by studies at other sites (McClain et al., 2003; Groffman et al., 2009). Campbell et al. (2006) observed in 2002 through 2004 that the soils for the WS14 and WS15 catchments remained unfrozen below 10 cm and had elevated NO3− concentrations due to presumed microbial activity occurring at temperatures near and even below 0°C (Dorland and Beauchamp, 1991; Clein and Schimel, 1995).
Across both catchments, the NO3− concentrations at each microsite varied by depth (Figures 5 and 6). In WS14, five of the six microsites had higher NO3− concentrations at the 15-cm depth and within the forest floor compared with the 50-cm depth deep within the mineral soil and four of the six microsites in WS15 had the same pattern with depth. In WS14, groundwater NO3−concentrations were relatively enriched compared with soil water in both catchments and groundwater in WS15 (Figure 12). Similar patterns with depth were also observed in previous studies in the WS14 and WS15 catchments (Campbell et al., 2006; Christopher et al., 2008). Previous isotopic work in WS14 and WS15 indicates that NO3− is produced in the forest floor by nitrification and NO3− found in deeper soils and groundwater is also derived from nitrification (Campbell et al., 2006). In WS14, the availability of inorganic N may exceed biological demand, allowing N in soil water to migrate to depth, resulting in higher NO3− concentrations in groundwater in WS14 relative to WS15 (Christopher et al., 2008).
NO3− concentrations in groundwater varied between catchments as well as within each catchment (Figures 7 and 12). The groundwater sampled in the two wells in WS15 (W12 and W13) had a lower NO3− concentration range (0.02 to 1.7 mg/l) with the highest observed concentration 1.7 mg/l. A study in northeastern Vermont also observed NO3− concentrations decreasing with depth; surficial soil waters and shallow groundwater consistently had higher concentrations than deep groundwater (Sebestyen et al., 2008). In contrast to well water NO3− concentrations in WS15, for WS14 our results differed from those of Campbell et al. (2006) and Christopher et al. (2008), who observed that all groundwater NO3− concentrations were low when compared with other water sources within the catchment, such as soil and stream water (Figure 12). McHale et al. (2002) and Piatek et al. (2009) concluded that deep groundwater had a major influence on stream water chemistry in WS14. In the current study, NO3− concentrations in groundwater were sampled from several wells in WS14. W36 concentrations ranged from 7.6 to 10.5 mg/l, which is greater than the highest stream NO3− concentration of 7.3 mg/l. Groundwater sampled from W33 and W35 had NO3− concentrations similar to those observed in the soil water, whereas W34 concentrations were lower in relation to the other water sources. These NO3− levels suggest a greater concentration of groundwater NO3− during our study, showing there can be substantial variation among years in these concentrations (Campbell et al., 2006; Christopher et al., 2008).
Temporal Variability in NO3− Chemistry
During our study, there were no mid-winter melt events that resulted in total loss of snowpack. There was, however, some evidence of elevated NO3− export during mid-winter melt events associated with partial loss of snow cover. NO3− export during the five days of the January rain-on-snowmelt event comprised 8 and 16% of the total stream NO3− load for the four-month winter study period (December 19 through April 19, 2008) for WS14 and WS15, respectively. This can be compared with the NO3− export during the final spring melt, which comprised 39 and 47% of the total four-month winter NO3− load in WS14 and WS15, respectively. Previous work has shown that mid-winter rain-on-snow events account for about 60% of the total NO3− export in winter in forested watersheds, on average, despite large differences in total NO3− load across sites (Casson et al., 2010).
Both catchment streams showed similar temporal responses of NO3−, DOC, and Ca2+ (Figures 8 and 9). NO3− and DOC concentrations increased with the stream hydrographs during the January 7-12 melt event and during the final spring snowmelt. Ca2+ concentrations had a similar, but inverse response to the same melt events, decreasing as discharge increased. The concomitant increases in NO3− and DOC concentrations suggest that these solutes were being mobilized from the forest floor. The NO3− concentrations decreased slightly before the peak discharge in the final snowmelt, with a more gradual decrease in DOC concentrations. Past studies in other locations have also observed highest NO3− concentrations during snowmelt events, despite differing NO3− sources (Murdoch et al., 1998; Ohte et al., 2004; Petrone et al., 2007; Sebestyen et al., 2008). The decrease of Ca2+ concentrations was more visually evident in WS14, which had higher Ca2+ concentrations overall compared with WS15, but the percent change in Ca2+ concentrations was similar in both catchments (a decrease of between 30 and 35% from April 1 to mid-melt). Hence, the melting snow with its low solute concentrations had a more marked effect on the solute concentrations in WS14 stream water. Christopher et al. (2008) developed a conceptual model to explain the stream water geochemical response during melt events in WS14 and WS15. They proposed that during snowmelt events, the water table rises, reaching its maximum just prior or simultaneously as the stream hydrograph peaks resulting in greater shallow soil water contribution to streamflow. The influx of soil water increases stream NO3− and DOC concentrations and reduces base cation concentrations. Therefore, both WS14 and WS15 showed a flushing response (e.g., Hornberger et al., 1994) during snowmelt; the water table rises, flushing shallow soil water that is enriched in NO3− due to nitrification that occurred in the shallow forest floor to the stream. NO3− dynamics are different during late-summer/fall rain events than winter NO3− dynamics. Christopher et al. (2008) observed different solute concentration responses to rain events in WS14 and WS15. Stream water concentrations of NO3− and Ca2+ were highest prior to rain events, decreased as the hydrograph increased, and returned to base-flow levels when the hydrograph receded. DOC concentrations followed the hydrograph, increasing and decreasing in proportion to streamflow. In comparison, WS15 NO3− concentrations increased and decreased with streamflow, the opposite of WS14 observations. They attributed these differing responses to different NO3− concentrations of the stream water sources. WS14 had much higher NO3− concentrations in deep groundwater, which contributed to base flow and as the catchment got wetter during rain events, more streamflow was from the shallower soil water with lower NO3− concentrations. WS15 deep groundwater had lower NO3− concentrations; therefore, the inclusion of shallow soil waters to streamflow increased stream NO3− concentrations.
Soil water NO3− concentrations were spatially variable and exhibited a general decline during the winter, particularly for those lysimeters with initially high NO3− concentrations and declined markedly toward the end of the melt (Figures 5 and 6). Compared with other water compartments, soil water NO3− concentrations had relatively high concentrations, particularly in the forest floor (Figure 12). During the terminal melt in April, which included daily sampling, there was a clear decline in soil NO3− concentrations.
Groundwater NO3− concentrations varied temporally with an overall general decline over the winter (Figure 7). The overall decline is most apparent in wells with initially higher NO3− concentrations and suggests a gradual decline of NO3− in groundwater as a result of the dilution of this pool by meltwater with low solute concentrations. The lower relative declines in groundwater NO3− concentrations compared with soil water likely reflect the greater volume of the groundwater pool. NO3− concentrations in the soil and groundwater did approach a more uniform, minimum value of ≤5 mg/l by the end of the spring snowmelt.
Watershed NO3− dynamics have long been of interest (e.g., Chen et al., 1984; Schaefer et al., 1990), but most recently there has been more focus on the role that the winter season and snowmelt have on NO3− export due to the predominance of this period in contributing to the amount of NO3− discharged (e.g., Park et al., 2003; Campbell et al., 2006, 2009; Petrone et al., 2007; Piatek et al., 2009). Although during our study there was only one large mid-winter melt event and no catchment scale soil freezing, the effects of snow cover and smaller mid-winter events in affecting soil water movement and solute dynamics, including NO3− chemistry, were still evident. Detectable changes in soil temperature changes over time and the flushing of NO3− from the forest floor were observed.
We used physical parameters (snow depth, density and SWE, groundwater level, stream level, air temperature, and soil temperature) to observe the role of snowmelt events in NO3− dynamics in a forested watershed. Our results clearly indicated small melt events occurred mid-winter, as evidenced by soil temperature decreases as snowmelt water infiltrated the soil. These events coincided with increased stream level and groundwater level, and/or decreased snow depth. Such changes in soil temperature in response to minor mid-winter melt events have not been documented in previous studies. The use of soil temperature data for evaluating the spatial and temporal pattern of water movement during snowmelt has great potential for future research. These temperature data when combined with solute chemistry provide new insight on watershed hydrological and biogeochemical dynamics.
Our findings on the dynamics of NO3− support the results of other studies that have found “hotspots” of high NO3− concentrations that vary both in space and time (McClain et al., 2003; Groffman et al., 2009). NO3− concentrations in the snowpack were very low, and there was a general decline of NO3− with depth from the forest floor (15 cm) to deeper mineral soil (50 cm) to groundwater throughout the winter. Spatially, soil water NO3− concentrations ranged from <0.1 to 24.3 mg/l, suggesting the spatial variability of nitrogen uptake and transformations within these catchments. During the final melt in April, soil water and groundwater showed a marked decline in NO3− concentrations, concomitant with increases in stream water NO3− concentrations.
These findings help link the role of snowmelt to NO3− export in forest watersheds and these results have implications related to climate change. This study shows that continuous snowpack prevented soil freezing. Under climate-change scenarios that result in less snow, soil freezing may become more prominent. An increase in soil freezing has the potential to alter the site hydrology, specifically flow paths including infiltration rates and surface runoff. Both the timing and intensity of snowmelt and mid-winter rain events would impact soil flushing events and the export of solutes. In addition, our findings indicated that minor melt events were detectable as changes in soil temperature, streamflow, groundwater level, and snow depth but based on loading were relatively minor contributors to NO3− loss. A warmer climate and fluctuating snowpack may result in more major, mid-winter melt events and greater NO3− export to surface waters. This potential for NO3− concentrations in soil, ground, and surface waters to increase as the climate warms, also indicates the potential of downstream water quality being impacted by the increased depletion of cations and the resulting acidification, and demonstrates the need for a better understanding of climate change’s potential impact on NO3− export during the winter season.
This research was made possible by a grant from the USDA McIntire-Stennis program, administered through the SUNY College of Environmental Science and Forestry. Funding was also provided by the New York State Energy Research Development Authority (NYSERDA). We would like to thank Tom Sesto, Mark Fabian, Ken Hubbard, Laura Schifman, David Lyons, and Rachel Ribaudo for help sampling during challenging field conditions in winter. A special thank you to Pat McHale for help with all aspects of data collection and laboratory analysis for this project. The authors thank the three anonymous reviewers and the associate editor for thoughtful and detailed comments that greatly improved the quality of the paper.