Transcription

PUBLICATIONSWater Resources ResearchRESEARCH ARTICLESnowmelt Timing as a Determinant of Lake Inflow Mixing10.1002/2017WR021977D. C. Roberts1,2Special Section:Responses to EnvironmentalChange in Aquatic MountainEcosystemsKey Points: Snowpack magnitude affects therelative timing of snowmelt andspring lake warming in a longresidence time, snowmelt-fed lake Reduced snowpack increases theproportion of annual inflow enteringthe lake prior to stratification andwith near-neutral buoyancy A projected shift toward decreasedsnowpack may increase nearshoreand near-surface mixing of inflowloads in snowmelt-fed lakesSupporting Information:Supporting Information S1 Correspondence to:D. C. Roberts,[email protected]:Roberts, D. C., Forrest, A. L., Sahoo,G. B., Hook, S. J., & Schladow, S. G.(2018). Snowmelt timing as adeterminant of lake inflow mixing.Water Resources Research, 54. https://doi.org/10.1002/2017WR021977Received 3 OCT 2017Accepted 26 JAN 2018, A. L. Forrest1,2, G. B. Sahoo1,2, S. J. Hook3, and S. G. Schladow1,21Department of Civil & Environmental Engineering, University of California, Davis, CA, USA, 2UC Davis TahoeEnvironmental Research Center, Incline Village, NV, USA, 3Jet Propulsion Laboratory, California Institute of Technology,Pasadena, CA, USAAbstract Snowmelt is a significant source of carbon, nutrient, and sediment loads to many mountainlakes. The mixing conditions of snowmelt inflows, which are heavily dependent on the interplay betweensnowmelt and lake thermal regime, dictate the fate of these loads within lakes and their ultimate impact onlake ecosystems. We use five decades of data from Lake Tahoe, a 600 year residence-time lake where snowmelt has little influence on lake temperature, to characterize the snowmelt mixing response to a range ofclimate conditions. Using stream discharge and lake profile data (1968–2017), we find that the proportionof annual snowmelt entering the lake prior to the onset of stratification increases as annual snowpackdecreases, ranging from about 50% in heavy-snow years to close to 90% in warm, dry years. Accordingly, in8 recent years (2010–2017) where hourly inflow buoyancy and discharge could be quantified, we find thatdecreased snowpack similarly increases the proportion of annual snowmelt entering the lake at weak topositive buoyancy. These responses are due to the stronger effect of winter precipitation conditions onstreamflow timing and temperature than on lake stratification, and point toward increased nearshore andnear-surface mixing of inflows in low-snowpack years. The response of inflow mixing conditions to snowpack is apparent when isolating temperature effects on snowpack. Snowpack levels are decreasing due towarming temperatures during winter precipitation. Thus, our findings suggest that climate change maylead to increased deposition of inflow loads in the ecologically dynamic littoral zone of high-residence time,snowmelt-fed lakes.Plain Language Summary Winter climate conditions can affect both the timing of snowmelt andthe timing of spring lake warming. Relative stream-lake temperature is an important control on how inflowplumes mix through lakes, distributing nutrients, carbon, and sediment relevant to lake ecosystems andwater quality. This study examines how snow conditions have affected relative stream-lake temperature,and thus inflow mixing conditions, over the past 50 years at Lake Tahoe. Years of lower snowpack are foundto favor nearshore and near-surface mixing of snowmelt inflows. Given trends toward reduced snowpackdue to a warming climate, this result may indicative of future conditions in large, snowmelt-fed lakes.1. IntroductionAll Rights Reserved.Spring snowmelt is a major component of the inflows that transport terrestrial carbon, nutrients, and sediments into many lakes. The thermal stratification of receiving lakes and their temperature relative to thetemperature of inflowing snowmelt controls the mixing and fate of transported particulates and solutes(Alavian et al., 1992; Rueda et al., 2007; Spigel et al., 2005). Annual snowpack is decreasing across westernNorth America (Fyfe et al., 2017), causing a shift toward earlier snowmelt (Barnett et al., 2005; Stewart, 2013;Stewart et al., 2004). The same climate patterns affecting snowpack are also expected to affect lake thermalregime (Sahoo et al., 2013, 2016; Sahoo & Schladow, 2008). In shorter residence-time lakes, snowmeltinflows can be a major driver of the lake thermal regime (e.g., Cort es et al., 2017). However, in longresidence-time lakes, where inflows annually replace only a small fraction of the lake volume, lake thermalregime is largely unaffected by inflows; the onset of stratification and snowmelt timing are independentlydriven through regional meteorological patterns. In such lakes, understanding how changing climate patterns will separately affect lake temperature and stream characteristics is important to understanding theeffect of climate change on the delivery of stream constituents that affect nutrient, carbon, and suspendedsediment concentrations in the lake.ROBERTS ET AL.1C 2018. American Geophysical Union.V

Water Resources Research10.1002/2017WR021977Both inflow timing and lake thermal dynamics have been affected by climate change at Lake Tahoe, a verylong-residence-time lake in the Sierra Nevada Mountains, USA. The length of the stratified season hasincreased by 24 days since 1968, with the onset of stratification advancing by 5 days between 1968 and2015, and expected to advance an additional 16 days (about 2 day/decade) by the end of the century (Sahooet al., 2016). Coats (2010) observed that peak snowmelt timing is shifting earlier at a faster rate, approximately 4 days/decade since the 1960s, than the shift in the timing of onset of stratification. Coats (2010) alsofound that the percentage of precipitation falling as snow is decreasing in the Tahoe basin. The changetoward a more rain-dominated precipitation regime, causing a shift from spring to winter stream discharge,is expected to be most pronounced in hydrologic ‘‘transition zones,’’ like the Tahoe basin, that are significantly affected by snowpack but where winter air temperatures average about 0 C (Njissen et al., 2001). Thisshift toward reduced snowpack and earlier peak streamflow is shown to be driven by warming air temperatures, rather than a reduction in precipitation, and is therefore likely to continue under projected climate scenarios in the Sierra Nevada (Hamlet et al., 2005; Kang et al., 2016; Pederson et al., 2011; Stewart et al., 2005).As peak snowmelt shifts earlier relative to spring lake warming, a greater proportion of inflows would beexpected to enter ambient lake waters of similarly cold temperatures. Low density differences betweeninflowing and ambient water would support high rates of near-field mixing as quantified by initial dilution(Alavian et al., 1992; Johnson et al., 1989). Reduced density differences would also lead to plume insertioncloser to the surface. The latter effect could be compounded by positive buoyancy in late-spring and earlysummer inflows due to warmer stream temperatures associated with reduced snowpack (Ficklin et al.,2013). The insertion depth and post-insertion propagation of inflow plumes, herein referred to as far-fielddynamics, is affected by lake stratification (Wells & Nadarajah, 2009). Under the expected timing shifts insnowmelt and lake stratification, plumes will be less vertically confined and will intrude at lower velocities(Imberger & Hamblin, 1982). The reduced inertial length scales associated with lower velocity plumes makethe plumes susceptible to rotation closer to shore (Horner-Devine, 2009; Horner-Devine et al., 2006). In summary, as snowmelt shifts earlier relative to the onset of thermal stratification, we would expect thicker,slower-moving, and more rotationally influenced inflow plumes, as illustrated in Figures 1c/1d.These conditions are expected to become more prevalent under a warmer climate and will affect the fateof inflow loads. Higher rates of near-field mixing would disperse inflow constituents through the ecologically dynamic littoral zone rather than allowing them to plunge into the pelagic zone with minimal dilution(Rueda et al., 2007). More near-surface insertion would potentially deposit a greater proportion of the nutrient load in the photic zone (Cort es et al., 2014). Slower-moving plumes would increase the proportion ofsuspended sediments deposited in the littoral zone (Scheu, 2016). Increased rotational influence coulddirect inflows and their constituents alongshore, rather than offshore, influencing littoral productivity asdemonstrated in the coastal ocean by Kudela and Peterson (2009) and Kudela et al. (2010).Here we explore the effect of a range of winter climate conditions on two factors known to influence inflowmixing. Long-term meteorological and stream discharge data (1968–2017) are used to quantify the effect ofclimate conditions on the timing of snowmelt relative to the onset of lake stratification. Shorter-term,higher-frequency stream and lake-littoral data are used to analyze the buoyancy conditions of inflows undera range of cold-to-warm and dry-to-wet conditions from 2010 to 2017. Serendipitously, this latter periodencompassed a broad range of winter/spring climate scenarios, including years of average conditions, yearsof extreme drought, and near-record snowpack years. Our results confirm the hypothesis that inflow mixingconditions at Lake Tahoe are strongly affected by snowpack. The projected reduction in snowpack, due to awarming climate, favors higher rates of nearshore and near-surface mixing of inflows.2. MethodsSnowmelt and lake stratification timings are calculated from long-term daily discharge data and lake temperature profile data from 1968 to 2017, and are compared to annual snowpack data. The response ofinflow buoyancy conditions to snowpack is quantified by calculating a time series of inflow buoyancy usinga combination of measured and modeled stream and littoral temperature data. This hourly buoyancy timeseries, along with hourly discharge data, allows for quantification of both the volume and proportion ofannual discharge entering the lake above a buoyancy threshold. These annual values, calculated for theROBERTS ET AL.2

Water Resources Research10.1002/2017WR021977Figure 1. (a) Map of Lake Tahoe and its watershed. The lake is shown in white with 100 m depth contours. The watershed is shown in gray with 250 m elevationcontours referenced to lake surface elevation of 1,897 m. The watershed for the representative stream, Blackwood Creek (underlined), is shown as a speckledregion. Other streams indicated were used as part of this study. Data collection station names are: RB—Rubicon; HW—Homewood; SS—Sunnyside; TC—TahoeCity NCEI; USCG—United States Coast Guard Station; TV—Tahoe Vista; IC—Incline Creek; TBx—NASA JPL Tahoe Buoys. (b) Example of the annual cycle of thermalstratification over the upper 100 m of the water column; dashed line at 25 May, 2015 shows the day of onset of stratification as defined by Sahoo et al. (2016).(c) Schematic of inflow mixing conditions favored by pre-stratification snowmelt—high rates of initial dilution; thick, less-confined surface plumes that are influenced by rotation nearshore. (d) Schematic of inflow mixing conditions favored by postonset of stratification snowmelt—less initial dilution; thinner, more confined, plunging plumes that are influenced by rotation farther from shore.period for which measurement occurred or modeling was possible (2010–2017), are then compared toannual winter climate conditions.2.1. Study Site and PeriodLake Tahoe is perched in the Sierra Nevada Mountains (surface elevation about 1,897 m) on the border ofCalifornia and Nevada, USA (Figure 1a). The lake is deep (maximum depth of 501 m; average depth ofROBERTS ET AL.3

Water Resources Research10.1002/2017WR021977305 m) and voluminous (approximately 156 km3 ). Owing to a comparatively small watershed (800 km2 ),the mean residence time of Lake Tahoe is extremely long, estimated at 600–700 years. This partiallyaccounts for the lake’s famed clarity (average annual Secchi depth greater than 20 m). The extremely lowinflow-to-lake volume ratio implies that inflows have a minimal effect on the energy balance of the lake.Instead, lake temperature and stratification are controlled predominantly by surface heat exchange. LakeTahoe is monomictic, typically stratifying from late-May to late-December and mixing to several hundredmeters in late-winter or early spring (Figure 1b). Due to its great depth and associated thermal inertia, LakeTahoe does not freeze.About 85% of Tahoe basin precipitation falls between November and April; summers (June–September) aredry and mild. At lake-level, the Tahoe basin averages 55 days per year with mean-daily temperature belowfreezing. However, the watershed topography ascends to over 3,000 m, with colder temperatures and moresnowfall at elevation. High-elevation snowpack drives a snowmelt-dominated inflow that typically startsincreasing toward the end of February, peaks in May or June, and asymptotes toward base flow in July. Winter rain events are not uncommon; these rain-on-snow events often drive large spikes in stream discharge.We use the annual period of 1 November to 1 August to bracket the time period in which streamflows aremost affected by snow conditions. Annual climate conditions, lake stratification, and snowmelt timing values were calculated for 1968–2017. Inflow buoyancy conditions were quantified for 2010–2017.2.2. Field Data Collection2.2.1. Historical Stream Discharge, Lake Temperature, and Climate DataBlackwood Creek, on the west shore of Lake Tahoe (Figure 1a), was chosen as a representative inflow forthis system. It has the fifth largest watershed in the Tahoe basin and exhibits a strongly snowmelt-drivendischarge signal. Discharge is above the annual mean between February and June, typically peaking at over300% of the annual mean in May or June, and decreasing to 6% of annual mean in September. Eighty-eightpercent of the 30.1 km2 watershed is undeveloped (Tahoe Regional Planning Agency Open Data). Thewatershed elevation ranges from 1,897 to 2,687 m, with 51% of the land area at over 2,200 m elevation. TheUnited States Geological Survey (USGS; www.waterdata.usgs.gov) has collected Blackwood Creek dailymean discharge data since 1960 (USGS 1033660). Gaps in the recent 2017 daily record were filled usingmean-daily values from the hourly discharge record described below.Water temperature profiles have been regularly collected at Lake Tahoe since 1968. Profiles are conductedto a depth of 150 m at the LTP index station (near TB3) every 7–10 days. From 1968 to 2005, data were collected at variable intervals using reversing thermometers. Since 2005, water temperature profiles have beenmeasured using a Sea-Bird Instruments CTD, offering submeter resolution data.The National Centers for Environmental Information (NCEI; National Oceanographic and AtmosphericAdministration; https://www.ncdc.noaa.gov) have compiled daily precipitation and maximum and minimumtemperature records near lake-level in Tahoe City, CA since 1903. Yearly maximum snow depth has beenrecorded since 1960 at Donner Summit, CA (elevation 2,100 m) by the UC Berkeley Central Sierra Snow Lab(CSSL). While Donner Summit does not lie within the Lake Tahoe basin, it is close to the basin rim and ispositioned windward relative to the prevailing storm pattern. Annual maximum snow water equivalent(SWE), recorded at CSSL since 1980, correlates strongly with yearly maximum snow depth (supporting information Figure S4); we believe that the yearly maximum snow depth data provide a reliable representationof longer-term historical snow patterns for the Lake Tahoe basin.2.2.2. Data for Inflow Buoyancy CalculationsNearshore temperature data have been collected at the Homewood nearshore station (Figure 1) since September 2014. The nearshore station includes an RBR Maestro conductivity-temperature-depth (CTD) instrument and a Turner Designs C3 fluorometer mounted to a weighted frame at approximately 2 m depth.Thirty-second measurements of CTD and colored-dissolved organic matter fluorescence (fCDOM) arerecorded by a dock-mounted Campbell Scientific CR1000 data-logger cabled to the instruments. Instrumentposition on the lakebed is periodically adjusted by UC Davis Tahoe Environmental Research Center (TERC)divers to maintain consistent depth with varying lake-level. Small data gaps (up to 3 h) were filled using linear interpolation. Data records are generally continuous except one long-term gap from 27 January to 3May 2016.ROBERTS ET AL.4

Water Resources Research10.1002/2017WR021977Near-surface water temperature data were collected from four NASA/JPL midlake buoys beginning in October2009 (TBx; Figure 1). Short thermistor chains measure temperature at 0.5, 1, 1.5, 2, 3, 4, 5, and 5.5 m beloweach buoy at two-minute intervals. The midlake water temperature record was constructed at a 1 h time stepfrom the TB3 buoy data. Gaps were filled using the average between TB1, TB2, and TB4 temperature data ateach depth. Remaining gaps up to 5 h were filled using linear interpolation. Eight remaining single-day gapswere filled using data from the following day, matching fill-data to the time-of-day of the gaps.Fifteen-minute Blackwood Creek water temperature data were collected at USGS gauge 1033660 beginningon 20 January 2015 up through the end of the study period on 1 August 2017. Data time step was increasedto 1 h, and gaps up to 5 h were filled using linear interpolation. Gaps longer than 5 h but shorter than 3days were filled with data from days immediately post-gap, matching fill-data to the time-of-day of gaps.This process yielded a complete hourly stream temperature record for the period 20 January 2015 to 1August 2017.The USGS has recorded 15 min discharge data at Blackwood Creek since 1989, but discharge data fromadditional streams were needed to calibrate stream-to-stream relationships for filling gaps in the BlackwoodCreek record. The data record for Blackwood Creek was reduced to a 1 h time step, and gaps up to 5 h werefilled using linear interpolation. Larger gaps remaining in the discharge record for the study streams werefilled using calibrated interstream discharge relationships (see supporting information Text and Figure S1).In addition to stream discharge data, shortwave radiation (SW) and air temperature data were needed tocalibrate, validate, and run a stream temperature model used for extending the hourly stream temperaturerecord. Ten-minute SW data from TERC’s United States Coast Guard meteorological station (USCG; Figure1a) were used to populate the hourly record. Gaps were filled using data from the Incline Creek meteorological station (Figure 1a) maintained by the Desert Research Institute (DRI). Remaining gaps were filled usingPAR data, collected on the TERC rooftop (near Incline Creek; Figure 1a) and converted to SW. The hourly airtemperature data were from the USCG station. Gaps were filled with the average of the available data fromthree other west-shore meteorological stations maintained by TERC: Tahoe Vista, Sunnyside, and Rubicon(Figure 1a). Remaining gaps were filled with air temperature data from the DRI Incline Creek station.2.3. Analysis of Historical Discharge, Lake Temperature, and Climate Data (1968–2017)Data back to 1961 were used to quantify annual climate conditions and snowmelt timing, and data back to1968 were used to calculate annual timing of onset of lake stratification.To place each year in the context of climate conditions, annual metrics along cold-warm and dry-wet axeswere calculated. We quantified the cold-warm axis using the temperatures at which precipitation fell ratherthan mean seasonal air temperatures. Using the long-term NCEI meteorological data, we calculated the proportion of precipitation that fell with mean-daily temperatures at or below 48C and considered this to besnow. We use this above-freezing threshold because the majority of the basin is at higher elevation thanthe NCEI instruments and thus sees colder temperatures. Accordingly, this precipitation-as-snow (PAS)value, calculated for the 1 November to 1 August study period for each year from 1961 to 2017, correlatesmore strongly with snowpack data than PAS values calculated with a typical 0 C threshold (not shown).Years are classified along a dry-wet axis using the mean of the daily 1 November to 1 August dischargefrom the representative stream.The timing of the center of mass of snowmelt inflows (center timing—CT) was calculated from BlackwoodCreek daily discharge data following the method described in Stewart et al. (2004):XnXnðt q Þ CT5q(1)1 i i1 iti is time and qi is discharge at a given step i in the data. Yearly day-of-year of CT was calculated using dailydischarge data between 1 November and 1 August. The data record yielded complete yearly CT timing values for Blackwood Creek from 1961 to 2017.We use the simplified stability index (SI) to determine the annual timing of onset of stratification (Sahooet al., 2016)XzHSI5ðz2 z Þqz(2)zoROBERTS ET AL.5

Water Resources Research10.1002/2017WR021977Here zo is surface depth (zero), zH is depth limit of the chosen section of water column, z is depth, z is thegeometric centroid of the section of water column, and qz is the density at depth z. Following Sahoo et al.(2016), we calculated the day-of-year of onset of stratification as the day on which SI becomes greater than600 kg m2 over the upper 100 m of the water column. An example thermal profile at SI 600 kg m2 isshown at the dashed line in Figure 1b.2.4. Calculation of Annual Inflow Buoyancy Metrics (2010–2017)To better examine inflow mixing conditions under a range of climate conditions, annual buoyancy metricshad to be calculated for more years than were available in the nearshore and stream temperature datarecords. This section outlines the details of compiling and synthesizing nearshore temperature, stream temperature, and stream discharge data to complete the hourly 1 November 2009 to 1 August 2017 recordneeded for quantifying annual inflow buoyancy metrics over eight study seasons (2010–2017) representinga range of climate conditions.Relationships between midlake temperatures and nearshore temperatures were used to reconstruct nearshore temperatures from 1 November 2009 up to the installation of the nearshore station in September2014, and to fill larger gaps in the nearshore station record through 2017. The development of this empiricalmodel is described in more detail in supporting material (supporting information Text and Figure S2). Thenearshore-midlake (NS-ML) relationship was used to generate a complete hourly time series of nearshoretemperature over the 2010–2017 study period. Where measured data were available (most of 2015–2017),they were used to replace the modeled data in the long-term record.Meteorological and stream discharge and temperature data were used to train, test, and validate the empirical artificial neural-network stream temperature model developed for Tahoe basin streams by Sahoo et al.(2009). Discharge, air temperature, and SW data were used to force the model over the entire 1 November2009 to 1 August 2017 period to populate a complete hourly time series of stream temperature for Blackwood Creek. Where measured data were available, most of 2015–2017, modeled data were replaced withmeasured data in the record. Additional details and validation of the model are shown in the supportinginformation Text and Figure S3.Water density time series for streams and nearshore lake water were calculated as a function of temperatureand conductivity using the TEOS-10 MATLAB package (IOC et al., 2010) which follows the Feistel (2008)approach. Long-term conductivity data normalized to 25 C show consistent specific conductivity (SpC) inthe range of 90–100 lS cm at the nearshore station. We therefore calculate nearshore water density as afunction of variable temperature but constant SpC (90 lS cmÞ. Since stream SpC is expected to vary withdischarge, a discharge-SpC relationship was fit using using USGS Blackwood Creek SpC data recorded1980–1983 (supporting information Figure S5), and the relationship was used to generate the time series ofstream SpC used in inflow density ðqo Þ calculations. Ambient receiving water density (qa Þ is assumed equalto density as calculated from the two-meter depth nearshore temperature measurements and estimates.Buoyancy is defined as the difference between nearshore and stream densities: Dq5qa 2qo .With continuous hourly time series of discharge and relative stream-nearshore density, we calculate boththe volume-as-buoyant (VAB) and proportion-as-buoyant (PAB) of each year’s 1 November to 1 Augustinflows entering the lake with weak to positive buoyancy:XVAB5Qi ti where Dqi 20:1 kg m3(3)PAB5 VAB XQi ti(4)The inflow buoyancy value of 20:1 kg m3 , slightly negatively buoyant inflows, was chosen as a generalthreshold for dividing inflows that are likely to mix at or near the surface from those that are more likely toplunge or to insert at depth; we herein use the word ‘‘plunge’’ to describe any inflow that is at least 0.1 kg m3denser than the littoral waters. Cabbeling, the process whereby water on opposite sides of the temperature ofmaximum density mixes to form a dense, plunging flow (Carmack et al., 1979), introduces some uncertaintyinto this simplistic buoyancy analysis. When stream temperatures are below 3.9 C (approximate temperatureof maximum density when SpC 5 70 lS cm) and nearshore temperatures are warm enough for a maximumdensity plume to classify as plunging (Dq 20:1 kg m3 Þ, our thresholding analysis could classify inflows asROBERTS ET AL.6

Water Resources Research10.1002/2017WR021977‘‘buoyant’’ when cabbeling might actually generate a plunging flow. The potential effect of cabbeling on theresults of the buoyancy analysis is considered in the results and discussion.3. Results3.1. Climate, Stream, and Lake Stratification ConditionsFrom 1961 to 2017, the annual maximum snowpack at Donner Summit averaged 305 cm, with a minimum of 79 cm in 2015 and a maximum of 615 cm in 1969. Over this same period, on average 80.5% of 1November to 1 August precipitation fell with mean-daily temperatures below 48C at the NCEI rain gauge,near lake-level. This metric of precipitation-falling-as-snow was as low as 52.7% in 2015 and as high as97.8% in 1962. Total precipitation (as rain or snow) at the NCEI gauge ranged from 15.4 cm in 1976 to144.7 cm in 2017, with a mean of 75.3 cm. The highlighted years in Figure 2 (1997 and 2010–2017) offer arange of climate conditions to discuss winter climate effects on inflow mixing conditions in greater detail.Total November–August discharge is not significantly affected by PAS (p 0:6; r 2 0.001). Discharge volume is driven mostly by quantity of precipitation (p 10215 ; r 2 5 0.87) and base flow rather than the formin which precipitation falls. However discharge timing is affected by both snowpack and PAS. The CT shiftslater with increased snowpack (7.8 days later per 100 cm of snow depth) and with increased PAS (12 dayslater per 10% increase in PAS); see supporting information Figure S6.Based on temperature profile data from 1968 to 2017, annual onset of stratification at Lake Tahoe occurs,on average, on 29 May. In 1997, a very warm year, stratification occurred as early as 11 May. The latestrecorded onset of stratification at Lake Tahoe occurred on 24 June 1971. Onset of lake stratification is notsignificantly affected by stream discharge, the temperature at which precipitation falls, or the amount ofsnowpack (see supporting information Figure S7); lake stratification responds to climate variability independently of the way that spring streamflow timing and temperature are affected.3.2. Effect of Winter Climate Conditions on Snowmelt Relative to Lake StratificationSince the snowpack affects snowmelt timing but does not affect the timing of stratification, winter conditions influence the relative timing of inflows and spring lake warming, as shown in Figure 3. The proportionFigure 2. Annual 1 November to 1 August climate conditions relative to the long-term mean values (1961–2017). Difference between long-term mean PAS (80.5%) and annual PAS defines the horizontal warm-to-cold axis. Difference betweenlong-term mean study-period discharge at Blackwood Creek (1:32 m3 sÞ and annual mean study-period dischargedefines a vertical dry-to-wet axis. Point size is scaled to snowpack, ranging from 79 cm (2015) to 615 cm (1969).ROBERTS ET AL.7

Water Resources Research10.1002/2017WR021977Figure 3. (a) Proportion and (b) volume of annual 1 November to 1 August Blackwood Creek discharge entering the lakeprior to the onset of stratification versus annual snowpack (1968–2017). 1971 is shown as an outlier year where the lakestratified 25 days later than the long-term average due to cold late-spring air temperatures. Excluding 1971: (a) r 2 50:64and p 1027 ; (b) r 2 50:32 and p 1024 . Including 1971: (a) r 2 50:48 and p 1027 ; (b) r 2 50:33 and p 1024 .of inflows entering the lake prior to the onset of stratification (discharge pre-stratification; DPS) is significantly affected by snowpack. In years of lower snowpack, a greater proportion of annual inflows enter LakeTahoe prior to the onset of stratification (Figure 3a). However, increased annual snowpack still leads to agreater volume of snowmelt inflow entering the lake prior to stratification (Figure 3b).3.3. Effect of Winter Climate Conditions on Inflow BuoyancyAnnual patterns in inflow buoyancy are dependent on winter climate conditions (Figure 4). In Novemberand early December, inflows tend to be negatively buoyant. Low streamflows are cool due to cold fall airFigure 4. Hourly 1 November to 1 August Blackwood Creek inflow conditions (2010–2017). First row: Relative nearshorestream density with dashed line at Dq5

snowmelt and lake thermal regime, dictate the fate of these loads within lakes and their ultimate impact on lake ecosystems. We use five decades of data from Lake Tahoe, a 600 year residence-time lake where snow-melt has little influence on lake temperature, to characterize the s