The Panamint Range, located partially within Death Valley National Park of California, consists of the Cottonwood Mountains, Panamint Mountains (an informal name, but common usage; Belcher & Sweetkind, 2010), and Owlshead Mountains. Our study was conducted entirely in the central Panamint Mountains which we will refer to as the Panamint Range hereafter. The Panamint Range hosts numerous springs despite regional hyperaridity due to its location within the rain shadow of the southern Sierra Nevada (Figure 1). For context, the ratio of evaporation to precipitation is 90:1 (Hunt, Robinson, Bowles, & Washburn, (1966)). While seasonal snowpacks in the southern Sierra Nevada can be greater than 4 m in thickness annually (Barbour, Berg, Kittel, & Kunz, 1991), snowpacks are typically much thinner in the Panamint Range due to its location in the rain shadow. It is surprising that a mountain range that receives so little snow in comparison to the southern Sierra Nevada can support over 180 springs (Figure 1).
Unfortunately, very little hydrogeological research has been conducted in the Panamint Range. The work of King and Bredehoeft (1999) provides data on nine springs in the Panamint Range. As a consequence, the sources of recharge and magnitude of recharge supporting these springs remain poorly quantified. This creates a multifaceted problem given the importance of desert springs in the region (and worldwide). The desert springs emerging in the Panamint Range were critical to the survival of early Native American Timbisha Shoshone communities who would move from winter homes in the valleys to mountain homes during the spring, summer and fall (Crum, 2002;Wallace, 1980;White, 2006). These same springs were important sources of water for mining camps and mining activities which moved into the area in the late 1800s (Lingenfelter, 1986;Miller, 2008;White, 2006). Today, multiple studies have shown that the springs in the southern Great Basin are critical to aquatic ecosystem functioning since they often provide the only surface water for bird, mammal, and insect species (Chambers, Miller, & MacMahon, 2004;Hannah et al., 2007;Sada, Fleishman, & Murphy, 2005). For example, the springs emerging in the Panamint Range support nesting habitat for over 70 species of birds (Wauer, 1964) and some of the springs support unique crenobiontic (obligate spring-dwelling) animals, namely springsnails such as P. Turbatrix (Hershler, Liu, & Howard, 2014;Hershler & Sada, 2002;Sada & Pohlmann, 2007). Thus, a spatially comprehensive quantification of the sources of groundwater recharge that support these springs and the magnitude of recharge occurring in the Panamint Range will provide valuable information on the vulnerability of these springs to the effects of climate change and human alteration (Frisbee, Phillips, White, Campbell, & Liu, 2013;Manning et al., 2012;Rademacher, Clark, Clow, & Hudson, 2005).
Groundwater recharge has an explicit definition: it is water that has infiltrated the soil and percolated into the fully saturated porous media below the water table (Freeze & Cherry, 1979). In the American F I G U R E 1 Regional map showing the location of the Panamint Range and surrounding mountain ranges and spring locations (the red box in the inset shows the location of the study area relative to a basemap of the USA). Basins are labelled using abbreviations: BW, Badwater Basin and PV, Panamint Valley. Yellow Triangles represent precipitation collection sites. Information on the location of springs is sourced from the USGS National Hydrography Database (https://nhd. usgs.gov) Southwest, groundwater recharge occurring in the mountains contributes significant amounts of water to surrounding basin aquifers because precipitation is usually higher than evapotranspiration (ET) in the mountains, especially during the snowbound and snowmelt seasons (Bresciani et al., 2018;Wilson & Guan, 2004). In the basins, evaporation commonly exceeds precipitation year-round, thus, the likelihood of substantial recharge occurring in basins is negligible. This is especially true for basins with climatic conditions similar to Death Valley (Figure 1) where the ratio of ET to precipitation exceeds 90:1. Substantial recharge can, however, occur when surface water flowing from the mountain block infiltrates alluvial sediments at and beyond the mountain front. Therefore, it is important to evaluate recharge mechanisms as well as sources and magnitudes of recharge since springs emerging in different parts of the landscape may depend on different recharge mechanisms.
Groundwater recharge in mountainous terrain commonly occurs via three different mechanisms (or a combination of these mechanisms): (i) mountain-block recharge (MBR), (ii) mountain-front recharge (MFR; Wilson & Guan, 2004;Bresciani et al., 2018) or (iii) the combination of MBR and MFR collectively known as mountain-system recharge (MSR; Wahi, Hogan, Ekwurzel, Baillie, & Eastoe, 2008;Bresciani et al., 2018). MBR occurs at the highest elevations of the mountain block and supports groundwater circulating primarily within the mountain block. This recharge, for example, supports springs or seeps that emerge within the mountain block (Bresciani et al., 2018;Manning & Solomon, 2003;Wilson & Guan, 2004). MFR occurs when surface water flows from the mountain block and recharges alluvial sediments located beyond the mountain front (Bresciani et al., 2018;Manning & Solomon, 2003). Springs emerging beyond the mountain front, on alluvial fans or bajadas, and/or at the terminus of alluvial fan and bajadas are commonly supported by MFR. However, springs which emerge at mountain-front faults or in the adjacent basin may also be supported by MSR.
The perennial nature of many of the springs in the Panamint Range suggests that the springs must be dependent on substantial annual recharge or that they are dependent on paleorecharge (e.g. Pleistocene recharge). The latter requires substantial subsurface storage and/or a hydraulic connection to a regional-flow system. The paucity in precipitation across the rain shadow in the southern Great Basin indicates that if substantial modern recharge is occurring, then it must be occurring primarily in the mountains (Ajami, Troch, Maddock, Meixner, & Eastoe, 2011). Here we address the following questions in this study. (i) What is/are the source(s) of groundwater recharge in the Panamint Range? Are springs recharged primarily by snowmelt or rainfall? (ii) Where is the recharge occurring (mountain-block, mountainfront, or mountain-system)? (iii) How much recharge occurs in the Panamint Range? We answer questions (i) and (ii) using stable isotopes of spring waters and precipitation. We answer question (iii) using a chloride mass-balance approach (Aishlin & McNamara, 2011;Frisbee et al., 2013;Russell & Minor, 2002) which is compared to a derivation of the Maxey-Eakin equation (Maxey & Eakin, 1949;Stephens & Umstot, 2019;Wilson & Guan, 2004). The Panamint Range study is a small component of a larger research project, which seeks to quantify the vulnerability of desert springs across the southern Great Basin from the Spring Mountains located west of Las Vegas, NV to Owens Valley, CA. This research provides much-needed hydrogeological data that can be used to assess the vulnerability of these springs to climate change and other disturbances (Frisbee et al., 2013;Manning et al., 2012;Rademacher et al., 2005). The results presented here have critical implications for other desert springs in the southern Great Basin and are broadly applicable to desert springs globally.
The The geology of the Panamint Range reflects its unique tectonic history. The Panamint Range is located within a zone of extreme Cenozoic extension (Miller, 1987). The mountain block comprises primarily Proterozoic sedimentary rock with several phases of granitic intrusive bodies that were emplaced between the Triassic period and Miocene epoch (Workman, et al., 2016). Younger Cambrian and Ordovician sedimentary rocks can be found on the flank of the eastern side of the Panamint Range but are largely absent on the western flank (Norton, 2011;Stewart, 1983). The range is cut by steep ephemeral channels draining to both Panamint Valley and Badwater Basin (Figure 2). Flow from these streams recharges alluvium at the mountain front.
Mountain recharge is likely confined to the Noonday Formation, Johnnie Formation, and Kingston Peak Formation which outcrop broadly across the crest and highest elevations of the Panamint Range (Labotka & Albee, 1990;Norton, 2011;Workman, et al., 2016). Talus slopes with little vegetation and rocky slopes supporting sparse vegetation are widespread along the crest. Aquifers in the Panamint Range are most likely constrained to the Precambrian sedimentary strata because the carbonate rocks prevalent in these units are thought to have higher hydraulic conductivity than the intrusive bodies and heavily metamorphosed rocks also present in the mountain block. We cannot provide ranges of hydraulic conductivities since very little hydrogeological research has been completed in the study area. However, our inference is supported by field observations noting that many springs emerge at contacts between dolomitic units, such as the Noonday and Beck Spring Dolomites, and formations associated with marine regression, such as the Johnnie Formation and the Kingston Peak formation.
Chemical weathering in the carbonate units and faulting associated with extensional forces have the potential to enhance the porosity of these units and create highly connected flowpaths for groundwater within the mountain block. Chemical weathering of carbonate rocks may also enhance infiltration and subsequent recharge in the mountain-block. In comparison, the upper greenschist to lower amphibolite-grade metamorphism of the Precambrian rock likely limits hydraulic conductivity and deep circulation in these units (Miller, 1987). Faulting is widespread throughout the Panamint Range due to the rifting of the southern Great Basin (Miller, 1987;Petterson, Prave, Wernicke, & Fallick, 2011;Workman, et al., 2016). These faults have variable depths and orienta- , Bour, & Scibek, 2013;Miller, 1987). Fault planes on the eastern flank of the mountain block dip 20-30 eastward toward Death Valley (Maxson, 1950;Miller, 1987).
Due to the remote and rugged nature of the terrain, there are no permanent weather or SNOTEL stations in the Panamint Range. As a consequence, there are no continuous datasets of precipitation and temperature for the Panamint Range. However, as elevation increases, the amount of total annual precipitation increases, the amount of snow compared to rain increases, and the mean annual and mean daily temperatures decrease due to the orographic effect created by the extreme relief (Roe, 2005). The California Department of Water Resources lists the average rainfall as 8-10 cm year -1 along the western Panamint Valley floor and less than 5 cm year -1 on the eastern Badwater Basin side of the range (Wauer, 1964). Average precipitation at the higher elevations of the Panamint Range can be estimated using precipitation models such as PRISM (PRISM, 2004). PRISM data indicate that the highest elevations of the Panamint Range receive an average annual precipitation rate of 486 mm year -1 (48.6 cm year -1 ).
We use the PRISM estimated precipitation in the recharge analyses described in this article (see Section 3.4).
Air temperatures and evaporation rates increase sharply with decreasing elevation in the Panamint Range. S1 in Supporting Information for weather station locations and identification). These plots indicate that: (i) little, if any, snow falls at elevations less than 1350 mrsl, (ii) the majority of snow falls during December, January and February and (iii) snow can fall during November and March but only at elevations higher than 2000 mrsl. Although these ELR were created across the southern Great Basin, it is likely that the evaporation and sublimation losses are higher in the Panamint Range because of its higher mean annual temperatures compared to the southern Sierra Nevada, the Argus Range, and possibly the Spring Mountains (DWR, 1964). Average summer temperatures in the greater Death Valley basin range from 30 to 46 C while average winter temperatures range between 4 and 19 C (Stachelski, 2013). Panamint Valley, however, is located at a higher elevation than Death Valley (Figure 2) and has correspondingly milder temperatures. According to the US National Oceanic and Atmospheric Administration (NOAA) Climate Data Online, the average annual summer temperatures from nearby Trona, CA (ID # GHCND:USC00049035; 11S, 464673 mE, 3957799 mN; https://www.ncdc.noaa.gov/cdo-web/datasets/GHCND/sta-tions/GHCND:USC00049035/detail) range from 22 to 40 C while the average annual winter temperatures range between 2 and 16 C. The only data available within the Panamint Range itself is from a USAF WBAN Station (ID # 746190 99999; Superior Valley Gunnery Range; 11S, 491025 mE, 4020556 mN) located in Wood Canyon near Emigrant Pass at an elevation of 1225 mrsl, far from the expected recharge elevation. These data show average annual summer temperatures between 17 and 34 C and winter temperatures between 0 and 12 C. Data on evaporation and ET rates are lacking for the Panamint Range. However, Hunt, Robinson, Bowles, & Washburn (1966) state that pan evaporation rates in the basin equal 304.8 cm year -1 and can exceed 393.7 cm year -1 . During 1958-1961, a pan evaporation station was monitored near the Death Valley National Park Service Headquarters. During the winter months of November, December, January,
February and March, an average evaporation of 71.5 cm was recorded. This is in stark comparison to the 307.5 cm that was recorded over the remainder of the year (Hunt, Robinson, Bowles, & Washburn, 1966). In fact, the high evaporation rates result in an evaporation to precipitation ratio of 90:1. An evaporation "tub" was placed at 610 mrsl in Hanaupah Canyon and the data from this pan conform to the 90:1 ratio (Hunt, Robinson, Bowles, & Washburn, 1966). Thus, precipitation which does reach the land surface at elevations less than 1000 mrsl is likely lost to evaporation and does not recharge. Altogether, the sharp elevation gradient combined with the strong elevation dependencies between temperature and precipitation create conditions conducive to the sky island effect. As elevation increases, there are corresponding increases in precipitation and decreases in temperature that create stark ecotones (Coe, Finch, & Friggens, 2012;Gottfried, Gebow, Eskew, & Edminster, 2005;McCormack, Huang, & Knowles, 2009). In fact, the Panamint Range has every life zone in the western U.S. except for the Arctic life zone (Wauer, 1964). Wauer (1964) divided the Panamint Range into 8 ecological zones based on predominant vegetation and landscape placement (e.g. alluvial fan vs.
hillslope vs. high mountain) and found that the number of bird species and diversity of species changed as a function of ecological zone and water availability. Thus, the relationship between ecological metrics such as species diversity and elevation are themselves expressions of the subtle feedbacks between hydrology and climatology that are characteristic of other sky islands in the American Southwest.
Springs were selected for this study to achieve a broad spatial distribution with respect to elevation, geologic setting, and climatic setting across the Panamint Range (Figure 2). Samples of spring water were collected for stable isotopic analyses during a single two-week field campaign in the Panamint Range from 23 May 2017 to 02 June 2017.
Water samples were collected at each spring using a portable peristaltic pump (Miller & Frisbee, 2018) and Masterflex silicon tubing. The tubing was placed in the spring orifice (where possible) and a 0.22 μm polyethersulfone membrane Sterivex-GP pressure filter was attached to the other end of the tubing for filtering. Water was pumped directly into a collection vial after purging the tubing for 10 min. Most samples were collected directly from the spring orifice or source. Some spring emergences, however, were diffuse or their true emergences were inaccessible due to the extremely rugged terrain or presence of dense vegetation surrounding the spring emergence. In these cases, spring runs were sampled downstream of the spring emergence (these samples are indicated in Table 1). In total, 18 springs were sampled during the field campaign; 7 of which were sampled downstream of their emergence point. Samples were stored unrefrigerated in the field in
and δ 2 H values of Panamint range spring samples Spring name Elevation (mrsl) UTM coordinates zone 11 S (m E, mN) Date sampled (mm/dd/yy) Temperature ( C) δ 2 H (‰) δ 18 O (‰) Jail Spring 2434 491216, 4005046 5/24/2017 9.0 -101 -14.1 Thorndike Spring 2337 493294, 4009949 5/25/2017 9.8 -103 -14.2 Uppermost Spring 1633 496410, 4012068 5/31/2017 16.5 -95 -12.9 Apron Spring 1606 493581, 4017809 5/27/2017 17.7 -98 -13.2 High Noon Spring 1419 493756, 4018743 5/27/2017 17.3 -98 -13.3 Main Hanaupah Spring #2 a 1265 496928, 4004586 5/28/2017 15.1 -100 -13.7 Main Hanaupah Spring #1 a 1258 497013, 4004384 5/28/2017 20.5 -93 -12.5 Limekiln Spring 1223 486446, 3996617 6/1/2017 19.4 -99 -13.3 Unnamed Panamint spring C a 1206 486503, 3996454 6/1/2017 16.5 -97 -13.4 Wilson Spring 1195 499326, 3993877 5/29/2017 20.4 -53 -7.7 South Hanaupah Spring #3 a 1154 498063, 4004323 5/28/2017 16.1 -92 -12.4 Unnamed Panamint Spring E 963 484864, 3987455 5/26/2017 18.8 -94 -12.7 Unnamed Panamint Spring F a 803 483892, 3987614 5/26/2017 19.4 -92 -12.5 Lower Warm Spring B 760 506108, 3980234 5/30/2017 34.4 -93 -12.7 Lower Warm Spring A 755 506301, 3980186 5/30/2017 34.3 -93 -12.4 Wheel Spring b 748 499209, 4019868 5/26/2017 22.6 -61 -8.5 Post Office Spring c 321 479772, 3988537 6/2/2017 18.7 -78 -8.8 Warm Sulphur Spring 318 480753, 3997248 6/1/2017 32.0 -95 -13.0 Tule Spring -77 510652, 4010962 3/14/2017 27.4 -103 -13.4 Poplar Spring A 1225 465797, 4097533 12/19/2016 17.7 -101 -14.0 Upper Emigrant Spring 1231 482674, 4031167 05/19/2016 19.8 -100 -13.5 a
Springs that were sampled downstream of their true emergence (they were sampled in the springrun).
To collect precipitation in the Panamint Range, we deployed two precipitation collectors in 14 October 2017 using the design described in Earman, Campbell, Phillips, and Newman (2006) and Frisbee, Phillips, Campbell, Hendrickx, and Engle (2010). One collector was placed near the Mahogany Flats Campground (2478 mrsl, 11S, 493858 mE, 4009470 mN) and one was placed near Thorndike Campground (2262 mrsl, 11S, 493513 mE, 4010232 mN). The locations of the precipitation collectors were chosen because they are dirt-road accessible.
Sample collection was further restricted by seasonal road closures and poor driving conditions during snowmelt. Mineral oil was poured in the reservoir of the collector to minimize the impact of evaporation on isotopic fractionation (Earman et al., 2006;Friedman, Smith, Gleason, Wardern, & Harris, 1992;Frisbee et al., 2010;Scholl, 2006).
This collector type is suitable for rugged terrain in which the collector must remain unchecked for months at a time (Earman et al., 2006;Friedman et al., 1992;Frisbee et al., 2010;Scholl, 2006).
Samples of precipitation were taken from the collectors and the
collectors were reset on 12 March 2018, 18 May 2018 and 18 September 2018. Precipitation samples were decanted and separated from the mineral oil before they were stored in refrigerated 2 ml glass vials with screw-caps. One bulk snow sample was collected from the remnant snowpack near Thorndike Campground on 12 March 2018. This snow sample was melted within a sealed, clean 1 L sampling bottle and then decanted into a smaller vial. The δ 18 O and δ 2 H of all spring and precipitation samples were measured by the University of California Davis Stable Isotope Facility using a Los Gatos Research Laser Water Isotope Analyzer V2. The reported precision is 0.3 ‰ for δ 18 O and 2.0 ‰ for δ 2 H (SIF, 2018). The accuracy of the analyses is 0.11 ‰ for δ 18 O and 0.54 ‰ for δ 2 H (based on multiple [n = 11] analyses of known standards). All stable isotopic values are reported relative to Vienna Standard Mean Ocean Water (VSMOW). A local meteoric water line (LMWL) was created for the Panamint Range using the δ 18 O and δ 2 H of precipitation that was collected in this study. The springs of the Panamint Range were then compared to the LMWL and the global meteoric water line (GMWL; Craig, 1961). Since our precipitation collection was short and we only sampled the springs in the Panamint Range once, we assessed the δ 18 O and δ 2 H values of the Panamint Range springs and the LMWL by comparing them to published stable isotopic compositions of precipitation and springs measured across the southern Great Basin. Three additional LMWLs were created from published data capturing the variability of precipitation and springs across the rain shadow of the southern Sierra Nevada: (i) a LMWL for Death Valley, CA was created using data compiled from Friedman, Smith, Johnson, and Moscati (2002), (ii) a LMWL
for Owens Valley, CA was created using data that was compiled from Friedman et al. (1992) and (iii) a LMWL was created for the Spring Mountains using data that was compiled from Ingraham and Taylor (1991), Ingraham, Lyles, Jacobson, and Hess (1991) and Winograd, Riggs, and Coplen (1998).
Two-component mixing models are commonly used to identify and partition (separate) the sources of waters contributing to a stream or spring based on known endmembers (Frisbee et al., 2010;Genereux, 1998;Liu, Williams, & Caine, 2004;Sklash & Farvolden, 1979;Winograd et al., 1998). The same model can be applied to quantify the sources of recharge assuming that all sources of recharge are known and measured. In this case, rain and snowmelt are the two possible sources of recharge. The set of equations that defines this model as applied to spring recharge is shown below:
where, δ 18 O spring is the stable composition of the spring, f rain is the fraction of recharge from rainfall, δ 18 O rain is the stable isotopic composition of rainfall, f snow is the fraction of recharge from snow, and δ 18 O snow is the stable isotopic composition of snow. Although the δ 18 O of snow likely varies with elevation in the Panamint Range, we choose to separate springflow using seasonal endmembers (e.g. summer rain and winter snow) because we do not have sufficient data to assess stable isotope lapse rates during the winter and summer nor their impact on the partitioning. An unintended consequence of this decision is that the fraction of rain may be lower than estimated since low-elevation snow may be heavier than high-elevation snow. We do, however, assess this uncertainty by varying the δ 18 O snow from -17.2 ‰ (representing the lowest value measured in our study) to -14.5 ‰ (representing the highest value measured in our study). We do not use volume-weighted average stable isotopic values for precipitation in the separation.
Uncertainties in the fractions of recharge from snow and rain were calculated following the work of Genereux (1998) and Liu et al. (2004). We modify the uncertainty equation from Liu et al. (2004) to:
where, W fsnow is the uncertainty in the fraction of snow and rain.
Please note that W fsnow = W frain . C snow is the average δ 18 O of the snow endmember (-16.0 ‰) and was calculated as the average of all snow samples taken from the collectors (n = 5; shown in blue in Table 2).
C rain is the average δ 18 O of the rain endmember (-7.7 ‰) and was calculated as the average of all rain samples taken from the collectors (n = 2, Table 2). W csnow is the standard deviation of the snow samples collected in the Panamint Range (1.2 ‰) and was calculated as the average standard deviation of the five snow samples shown in Table 2. W crain is the standard deviation of the rain samples collected in the Panamint Range (0.1 ‰) and was calculated as the average standard deviation of the two rain samples shown in Table 2. W cspr is the standard deviation of the spring samples. This is a difficult value to provide since we do not have repeat samples of individual springs.
Furthermore, the standard deviation of all the Panamint Range springs is 1.8 ‰; this describes the spatial variability of the δ 18 O of springs and may not be meaningful in the context of uncertainty for individual springs. However, as suggested in Genereux (1998), there is flexibility in what value is used for this parameter. Therefore, we assume that W cspr is equal to the analytical uncertainty (0.3 ‰). If W cspr is increased to 1.0 ‰, then it results in an overall increase in uncertainty of 4-6%.
To estimate the amount of recharge occurring in the Panamint Range, we used a chloride mass-balance approach (Aishlin & McNamara, 2011;Frisbee et al., 2013;Russell & Minor, 2002) and compared those calculations to a derivation of the Maxey-Eakin equation (Wilson & Guan, 2004). The advantage of the chloride mass-balance approach is that it is specific to each spring but is applicable only if chloride is not added to the groundwater flowpaths by bedrock weathering sources or in the spring emergence (Frisbee et al., 2013). Chloride/ bromide ratios were calculated for all springs and only those springs with a Cl -/Br -less than 200 were used to estimate recharge by the chloride mass-balance method. We chose a ratio of 200 since Davis, Whittemore, and Fabryka-Martin (1998) reports a Cl -/Br -range of 80-160 for clean, shallow groundwater that only reflects chloride from precipitation. For the springs with a Cl -/Br -<200 (n = 10), we use the chloride massbalance approach developed in (Frisbee et al., 2013) where MBR is in units of mm year -1 :
where, Cl Precip is the maximum chloride concentration in precipitation.
A Cl -concentration for precipitation of 0.6 mg L -1 was reported by the National Atmospheric Deposition Program (NADP) for nearby Furnace Creek. However, our measured Cl -concentrations in precipitation more closely match the Cl -concentration of 1.3 mg L -1 which was reported for the Nevada Test Site by Tyler et al. (1996). We use the higher value of 1.3 mg L -1 since it likely represents an upper limit on the expected Cl -in regional precipitation. Cl spring is equal to the measured chloride concentration of the spring water, and P PR is the average annual precipitation of the Panamint Range (mm year -1 ). The runoff term for Cl -was omitted from Equation ( 5) based on two assumptions: (i) runoff is usually short lived prior to infiltration in the mountain-block and therefore does not acquire a significant Cl -concentration and (ii) the volume of snowmelt runoff during the snowmelt season is large such that Cl -from dry deposition is diluted in runoff.
There is very little data available to assess the variability of the
P PR term because the Panamint Range is ungauged and there have been few studies in the Range. Hunt, Robinson, Bowles, & Washburn (1966) measured precipitation at Aguereberry Point located in the northern Panamint Range (11 S, 495701 mE, 4023709 mN) at an elevation of 1961 mrsl. They report that the annual precipitation at this elevation is 381 mm year -1 . Webb, Steiger, and Turner (1987) incorrectly state that this precipitation was measured at the "top of the Panamint Range" (p. 479). For comparison, PRISM calculates an annual precipitation total of 188 mm year -1 at Aguereberry Point. We chose to use the annual precipitation of 486 mm year -1 reported using T A B L E 2 δ 18 O and δ 2 H values from precipitation collectors at Mahogany Flats Campground (2478 mrsl) and Thorndike Campground (2262 mrsl), and a fresh snow sample from Thorndike Campground Collection period Season Sample name δ 2 H (‰) δ 18 O (‰) 10/14/2017-03/12/2018 Winter 2017-18 Thorndike Collector -123 -16.6 10/14/2017-03/12/2018 Winter 2017-18 Mahogany Flats Collector -130 -17.2 10/14/2017-03/12/2018 Winter 2017-18 Thorndike Fresh Snow -124 -16.9 03/12/2018-05/18/2018 Spring 2018 Thorndike Collector -107 -14.9 03/12/2018-05/18/2018 Spring 2018 Mahogany Flats Collector -104 -14.5 05/18/2018-09/18/2018 Summer 2018 Thorndike Collector -47.9 -7.76 05/18/2018-09/18/2018 Summer 2018 Mahogany Flats Collector -49.2 -7.71 PRISM for Telescope Peak (see Section 2.3) since it is likely more representative of the amount of precipitation occurring along the crest of the Panamint Range at elevations greater than 2500 mrsl. Equation (5) was solved for all springs with Cl -/Br -<200 and then an average MBR was calculated from that group of springs. Jeton, Watkins, Lopes, and Huntington (2005) report that PRISM estimates of precipitation were within ±15% of the precipitation measured by the NWS and other agencies. Thus, we assume that annual precipitation at the highest elevations in the Panamint Range likely varies between 413 and 559 mm year -1 . Maxey and Eakin (1949) provide an alternative calculation for MBR that is empirical in nature since it is based on an experimental dataset from White River Basin, NV. The Maxey-Eakin method requires three steps: (i) divide a study area into distinct zones of mean annual precipitation (e.g. 30.5-38.1 cm, 38.1-50.8 cm, and so on), (ii) apply a scaling factor to each precipitation band that accounts for losses of water due to ET and surface runoff and (iii) sum the estimated recharge in each zone. It should be noted that the general form of the Maxey-Eakin equation is used to calculate MFR, not MBR.
However, Wilson and Guan (2004) state that since the general form of the Maxey-Eakin equation accounts for losses due to direct runoff and ET, we can assume that MFR equals MBR. Wilson and Guan (2004) show that a power-law trend describes the relationship between Maxey-Eakin estimated recharge for each precipitation zone and the midpoint of each precipitation zone. The equation, shown below, is valid when P m is less than 600 mm year -1 :
where, P m is the mean annual precipitation (mm year -1 ). This assumption is valid for the Panamint Range. Berger, Halford, Belcher, and Lico (2008) suggest that the Wilson and Guan approximation (Equation ( 6))
has one advantage over the original Maxey-Eakin method; it provides an estimate of recharge when P m is less than 20.3 cm, albeit a small amount of recharge, whereas the scaling factor for the Maxey-Eakin method is assumed to be equal to zero (Maxey & Eakin, 1949). In addi-
Results of the stable isotope analyses of the Panamint Range spring waters are reported in Table 1. The δ 18 O values in these spring waters range from -14.2 ‰ to -7.7 ‰ with an average value of -12.6 ‰, while δ 2 H values range from -104 ‰ to -53 ‰ with an average value of -93 ‰. The stable isotopic compositions of precipitation samples collected in the Panamint Range are reported in Table 2. Snow samples ('Winter 2017-18' samples shown in Table 2) are lighter (more depleted) than spring waters and rainfall and are nearly identical to the snowpack sample collected in March 2018 (see 'Snow near Thorndike' in Table 2). In comparison, the precipitation samples collected in May 2018 are comprised of late winter snow and early spring rainfall and are isotopically heavier than the snow samples. Finally, the summer precipitation samples collected in September 2018 represent only summer rain and are enriched in heavy isotopes. Since we only have two sampling sites, there is insufficient data to quantify spatial variability or create an isotopic lapse rate. T A B L E 3 Data used to create LMWLs across the study area Owens valley (CA) LMWL Sample location δ 2 H (‰) δ 18 O (‰) Time of sample collection Citation Lone Pine, CA -89 -12 October 1985 Friedman et al. (1992) Lone Pine, CA -60 -7 April 1986 Friedman et al. (1992) Lone Pine, CA -119 -16 October 1986 Friedman et al. (1992) Lone Pine, CA -30 -3 April 1987 Friedman et al. (1992) Inyokern, CA -106 -14 October 1986 Friedman et al. (1992) Inyokern, CA -37 -3 April 1987 Friedman et al. (1992) Bishop, CA -27 -4 Summer 1991 Friedman et al. (2002) Bishop, CA -107 -14 Winter 1992 Friedman et al. (2002) Bishop, CA -108 -13 Winter 1993 Friedman et al. (2002) Bishop, CA -109 -15 Winter 1994 Friedman et al. (2002) Death valley (CA) LMWL Dante's view -40 -5.7 Summer 1991 Friedman et al. (2002) Dante's view -103 -14.5 Winter 1991 Friedman et al. (2002) Dante's view -68 -8.2 Summer 1992 Friedman et al. (2002) Dante's view -97 -13.7 Winter 1992 Friedman et al. (2002) Dante's view -108 -14.1 Winter 1994 Friedman et al. (2002) Furnace creek -18 0 Summer 1991 Friedman et al. (2002) Furnace creek -75 -10 Winter 1991 Friedman et al. (2002) Furnace creek -69 -8 Summer 1992 Friedman et al. (2002) Furnace creek -71 -9 Winter 1992 Friedman et al. (2002) Furnace creek -51 -4 Winter 1994 Friedman et al. (2002) Spring mountains (NV) LMWL Not given -116 -16 Not given Ingraham et al. (1991) Not given -103 -14 Not given Ingraham et al. (1991) Not given -89 -12 Not given Ingraham et al. (1991) Not given -75 -10 Not given Ingraham et al. (1991) Not given -61 -8 Not given Ingraham et al. (1991) Not given -48 -6 Not given Ingraham et al. (1991) Not given -34 -4 Not given Ingraham et al. (1991 Not given -20 -2 Not given Ingraham et al. (1991) low-discharge spring. Therefore, Wheel Spring is not included in 3).
The calculated fractions of recharge from snow and rainfall (and associated uncertainties) are listed in Table 4. The uncertainties shown in
Table 4 represent the uncertainty calculated based on the assumed Thorndike Spring 79 21 12 Jail Spring 77 23 12 Main Hanaupah Spring #2 72 28 11 Limekiln Spring 68 32 10 Apron Spring 67 33 10 High Noon Spring 68 32 10 Unnamed Panamint Spring C 68 32 10 Warm Sulphur Spring 64 36 10 Uppermost Spring 62 38 10 Unnamed Panamint Spring E 61 39 9 Lower Warm Spring B 60 40 9 Lower Warm Spring A 57 43 9 Main Hanaupah Spring #1 58 42 9 South Hanaupah Spring #3 57 43 9 Unnamed Panamint Spring F 58 42 9 Wilson Spring 1 99 4 Poplar Spring A 76 24 12 Tule Spring 70 30 11 Upper Emigrant Spring 69 31 11
Note. Wilson Spring (grey shading) is not included in the calculation of the average, minimum and maximum recharge. Uncertainty is estimated in percent.
between elevation and the fraction of recharge from snow is weak (Figure 6; f s = 0.0085 x z(m) + 55.7, R 2 = .38, p-value < .001, where z (m) equals the elevation of the spring emergence), but in general, the fraction of recharge from snow increases with increasing elevation.
Estimates of MBR based on the chloride mass-balance method range from 12 mm year -1 (at an elevation of 800 mrsl) to 388 mm year -1
(at an elevation of 2434 mrsl; Figure 7). The average MBR is 91 mm year -1 using the chloride mass-balance method (Table 5). The elevation dependency found by creating a simple linear regression to the data presented in and the isotopic composition of snowmelt changes prior to recharge due to evaporation and mixing in the soil zone (Beria et al., 2018;Earman et al., 2006 ;Frisbee et al., 2010). These processes tend to leave snowmelt recharge isotopically heavier and sometimes slightly evaporated compared to fresh snow (Earman et al., 2006;Frisbee et al., 2010). If the isotopic composition of snow becomes heavier, then snowmelt recharge should plot closer to the springs in Figure 4, and the estimated fraction of recharge sourced from snow should increase. For example, suppose that the snow endmember becomes 2 ‰ heavier (this value was chosen based on the snow evolution line observed in Frisbee et al., 2010), then the "evolved snow" endmember accounts for 75 (±15) to 104 (±20) percent of spring discharge.
The stable isotopic composition of snow likely decreases with increasing elevation, thus the proportion of recharge from rain may again be too large since low-elevation snow may have a higher δ 18 O. We do not have sufficient field data to assess the stable isotopic lapse rate, however if we assume a reasonable lapse rate of -2 ‰ km -1 (Poage & Chamberlain, 2001), then we can vary the stable isotopic composition of fresh snow accordingly from -16.0 ‰ (used in our model and collected between 2200 and 2500 mrsl). Our snow samples ranged from -17.2 to -14.5 ‰ and this largely conforms to the assumed lapse rate. Table 6 shows the uncertainty associated with
Relationship between the estimated fraction of recharge sourced from snow and mean elevation of spring emergence relative to sea level. The blue dots represent the fraction of recharge sourced from snow assuming that the δ 18 O of snow is equal to -16.0 ‰ (the average snow composition of this study). The red dashes represent the fraction of recharge sourced from snow assuming that the δ 18 O of snow is equal to-17.2 ‰ (assumed to be either fresh snow or snow at elevations >3000 mrsl if the stable isotopic lapse rate is approximately -2.0 ‰ km -1 ). The green dashes represent the fraction of recharge sourced from snow assuming that the δ 18 O of snow is equal to -14.5 ‰ (assumed to be late season snowpack or snow at elevations <2000 mrsl if the stable isotopic lapse rate is approximately -2.0‰ km -1 )
Relationship between the estimated MBR (mm year -1 ) and elevation of spring emergence
T A B L E 5 Data used to estimate groundwater recharge. Data in yellow shaded cells have Cl -/Br -greater than 200 Spring Name Elevation (m) Cl -(mg L -1 ) B r -(mg L -1 ) C l -/Br - Recharge (mm year -1 ) Jail Spring 2434 3.28 0.024 137 193 High Noon Spring 1419 21.5 0.171 126 29 Main Hanaupah Spring #2* 1265 1.63 0.011 148 388 Poplar Spring A 1225 23.9 0.17 141 26 Limekiln Spring 1223 8.35 0.075 111 76 Unnamed Panamint Spring C* 1206 7.57 0.042 180 83 Wilson Spring 1195 11.7 0.078 150 54 Unnamed Panamint Spring E 963 30.9 0.2 155 20 Unnamed Panamint Spring F* 803 32.5 0.19 171 19 Lower Warm Spring A 755 25.9 0.13 199 24 Apron Spring 1606 18.3 0.07 261 35 Lower Warm Spring B 760 25.8 0.11 235 24 Thorndike Spring 2337 7.44 0.005 1488 -Uppermost Spring 1633 12.9 0.005 2580 -Main Hanaupah Spring #1* 1258 6.95 0.005 1390 -Upper Emigrant Spring 1231 38.8 0.005 7760 -South Hanaupah Spring #3* 1154 6.7 0.005 1340 -Post Office Spring 321 1520 2.21 688 -Warm Sulphur Spring 318 873 1.3 687 -Tule Spring -77 998 0.14 6931 -Note. Thorndike Spring, Uppermost Spring, Main Hanaupah Spring # 1, Upper Emigrant Spring, and South Hanaupah Spring # 3 springs which have high Cl -/Br -due to Br -concentrations which were non-detects, so Br -was set equal to ½ the detection limit. Post Office Spring and Warm Sulfur Spring are interacting with evaporites in the emergence. Tule Spring is mixing with basin brines. The springs with high Cl -/Br -ratios were not used to calculate the average groundwater recharge. T A B L E 6 Uncertainty in Partitioning of recharge; where f s is the fraction of snow, f r is the fraction of rain, and Unc is uncertainty in percent Spring Name fs (%) f r (%) Unc. (%) fs (%) f r (%) Unc. (%) Jail Spring 0.67 0.33 9 0.94 0.06 17 Thorndike Spring 0.69 0.31 9 0.96 0.04 18 Unnamed Panamint Spring E 0.53 0.47 7 0.74 0.26 14 Unnamed Panamint Spring F 0.51 0.49 7 0.71 0.29 13 Wheel Spring 0.08 0.92 3 0.11 0.89 5 Apron Spring 0.58 0.42 8 0.81 0.19 15 Main Hanaupah Spring #2 0.63 0.37 9 0.88 0.12 16 Main Hanaupah Spring #1 0.51 0.49 7 0.71 0.29 13 South Hanaupah Spring #3 0.49 0.51 7 0.69 0.31 13 Wilson Spring 0.00 1.00 3 0.01 0.99 5 Lower Warm Spring A 0.50 0.50 7 0.69 0.31 13 Lower Warm Spring B 0.52 0.48 7 0.73 0.27 14 Uppermost Spring 0.54 0.46 8 0.76 0.24 14 Limekiln Spring 0.59 0.41 8 0.83 0.17 15 Unnamed Panamint Spring C 0.60 0.40 8 0.83 0.17 15 Warm Sulphur Spring 0.56 0.44 8 0.78 0.22 15 Note. The partitioning shown in the white column was created assuming that the δ 18 O of snow equals -17.2 ‰ and the δ 18 O of snow equals -7.7 ‰. The partitioning shown in the grey column was created assuming that the δ 18 O of snow equals -14.5 ‰ and the δ 18 O of snow equals -7.7 ‰. These values represent the lowest and highest the δ 18 O of snow, respectively, measured in the study. at high elevations and discharge at lower elevations (Anderson, 2005;Saar, 2011). The effects of heat conduction from the surface are often assumed to be negligible in mountainous terrain where the aquifer is deeper than 10 m (Anderson, 2005;Manga & Kirchner, 2004).
Environmental lapse rates bounding the months of the western snow season of November, December, January, February and March strongly indicate that snowfall is unlikely at elevations less than 1400 mrsl and that the majority of snow occurs during December and F I G U R E 8 Map of National Land Cover Database (NLCD) land cover classes January (Figure 3a). PRISM calculates 299 mm year -1 near Telescope Peak during the months of November, December, January, February and March as compared to 169 mm year -1 for the remainder of the year when the effects of ET are also the highest. The temperatures of all but two spring waters plot between the ELR for December (the coldest month) and July (the hottest month) indicating that springs are receiving recharge consistent with cooler conditions (Figure 3b). After recharge, the groundwater warms slightly as it circulates within the mountain block (Figure 3b). The temperatures of springs in the Panamint Range, however, plot to the left of the line describing the geothermal gradient (Figure 9); they are cooler than the geothermal gradient. The surface temperature in Figure 9 is assumed to be equal to the temperature of Jail Spring, one of the highest elevation springs.
The average geothermal gradient of the Panamint Range is 35.2 ± 1.7 C km -1 (Coolbaugh et al., 2005). Thus, we infer that the Table 5). MBR ranges from 19 to 24 mm year -1 at elevations less than 1000 mrsl, while at elevations greater than 1000 mrsl but less than 2000 mrsl, MBR ranges from 26 to 388 mm year -1 . MBR ranges from 85 to 193 mm year -1 at elevations greater than 2000 mrsl. We infer that MFR and MSR also support springs in the Panamint Range. The Panamint Range is fault-bounded such that alluvial fans are short on the west side of the range and large alluvial fans coalesce into bajadas on the east. Warm Sulfur Spring and Post Office Spring both emerge on the western side of the Range. Warm Sulfur Spring emerges at a mountain-front fault and is supported by MSR. Post Office Spring emerges at a fault at the terminus of the Pleasant Canyon alluvial fan, thus this spring is supported by MFR on the alluvial fan and MSR at the fault (Figure 2). Tule Spring emerges at the transition from the bajada draining Hanaupah Canyon and Badwater Basin on the east (Figure 2) and is supported by MFR. Tule Spring is surprisingly isotopically light for a basin spring (Table 1) and has a stable isotopic composition comparable to the headwater springs of Hanaupah Canyon (Main Hanaupah Spring # 2). Warm Sulphur Spring receives 58 percent of its recharge from snow, yet this spring emerges over 2134 m below the snowline. Therefore, there must be some groundwater connectivity through the mountain block to these springs consistent with MSR.
We do not see evidence for enhanced evaporation at these springs.
Adjacent basins (Badwater and Panamint Valley) typically receive less than 6 cm year -1 of precipitation per year compared to the high Reductions in recharge are also expected in the scenario where more rain and less snow falls on the mountainous systems of the western U.S. (Easterling et al., 2017;Knowles, Dettinger, & Cayan, 2006). It is extremely unlikely that groundwater recharge will remain stable or actually increase in this scenario (Niraula et al., 2017). The water balance of mountainous systems is hypsometric meaning that components of the water balance are elevation dependent. Goulden et al. (2012) discuss the hypsometric nature of the water balance in the Upper Kings River basin located in the southern Sierra Nevada. In their work, the difference (P-ET) between precipitation (P) and modelled ET increases with increasing elevation above ~2500 m (see Figure 8 of Goulden et al., 2012).
This implies that there is more effective precipitation (P-ET) available at higher elevations and that this water must be partitioned between surface runoff and groundwater recharge. The amount of effective precipitation at high elevations is critically important because, in a warming climate, vegetation is expected to expand upslope to higher elevations, which have been largely vegetationfree since the Last Glacial Maximum. When this occurs, P-ET will decrease implying that less water is available for groundwater recharge, and that recharge will occur over a potentially smaller area. These scenarios do not bode well for the springs of the Panamint Range. King and Bredehoeft (1999).
Changes in recharge will ultimately impact groundwater flowpath distributions, mean residence times of springs, and geochemical processes. These changes, in extension, will affect the permanence and ecosystem integrity of other springs in the Panamint Range in the future. Quantifying these fundamental relationships will improve our understanding of the processes and metrics, which define the permanence or vulnerability of desert springs in the U.S. and worldwide, especially those springs reliant on snowmelt recharge.