Quantifying water sources to a semiarid riparian ecosystem, San Pedro River, Arizona

Transcription

1 JOURNAL OF GEOPHYSICAL RESEARCH, VOL. 112,, doi: /2006jg000263, 2007 Quantifying water sources to a semiarid riparian ecosystem, San Pedro River, Arizona Matthew N. Baillie, 1,2,3 James F. Hogan, 1,2 Brenda Ekwurzel, 1,2 Arun K. Wahi, 1,2 and Christopher J. Eastoe 2,4 Received 18 July 2006; revised 5 February 2007; accepted 8 May 2007; published 19 July [1] The Upper San Pedro River Basin (Southeastern Arizona, United States) contains one of the few desert riparian areas in the Southwest, a system that is dependent on both shallow groundwater to support phreatic vegetation and baseflow for aquatic plants and animals. Proper management decisions for sustaining this biodiversity hotspot require understanding the hydrology of the riparian system and its interaction with the basin aquifer. To meet this need and to assess whether the techniques used would be efficient for evaluating other semiarid riparian ecosystems, we addressed the following questions. What are the contributions of different water sources (e.g., local recharge during monsoon flood events versus inflow of basin groundwater) to riparian groundwater and river baseflow? How does the spatial variability in water sources relate to gaining and losing reaches along of the river? We first characterize the possible water sources to the riparian system using a suite of geochemical tracers. Results indicate that, of the possible sources, basin groundwater recharged along the Huachuca Mountains to the west and local recharge of monsoon floodwaters are the dominant riparian water sources. Then, using their geochemical composition, we quantify these sources using a two end-member mixing model. We find that riparian groundwater composition varies between gaining and losing reaches. Locally recharged monsoon floodwater comprises 60 to 85% of riparian groundwater in losing reaches whereas that of gaining reaches contains only 10% to 40%. Baseflow, sampled year round, also contains a significant component of monsoon floodwater ranging from 80% on the upstream end and decreasing to 55% after passing though several gaining reaches. These results highlight the significance of local recharge during monsoon flood events as a water source for desert riparian systems, a fact that should be addressed when constructing and calibrating hydrologic models used to evaluate these future water management decisions. Citation: Baillie, M. N., J. F. Hogan, B. Ekwurzel, A. K. Wahi, and C. J. Eastoe (2007), Quantifying water sources to a semiarid riparian ecosystem, San Pedro River, Arizona, J. Geophys. Res., 112,, doi: /2006jg Introduction [2] Riparian corridors in the semiarid Southwest are rare areas of flowing surface water that provide habitat for the majority of local species and serve as an important stopping point for migratory species [Krueper et al., 2003; Glennon, 2002]. Riparian ecosystem biodiversity depends on both shallow groundwater to support phreatic vegetation and baseflow (streamflow derived from groundwater) for animals and aquatic plants. Loss of riparian ecosystems could mean the elimination of several sedentary endangered 1 Department of Hydrology and Water Resources, University of Arizona, Tucson, Arizona, USA. 2 SAHRA, University of Arizona, Tucson, Arizona, USA. 3 Now at New Mexico Institute of Mining and Technology, Socorro, New Mexico, USA. 4 Department of Geosciences, University of Arizona, Tucson, Arizona, USA. Copyright 2007 by the American Geophysical Union /07/2006JG species as well as abandonment by the numerous bird species that use these areas as stopovers on their annual migrations [Glennon, 2002]. This potential loss of biodiversity can have impacts that range beyond the ecological; the local economy can be reduced by the loss of tourism often associated with desert riparian areas [Orr and Colby, 2002]. [3] One of the greatest threats to riparian ecosystems is the increased stress on already scarce water resources caused by the recent rapid population growth in the semiarid southwestern United States. In many basins, water demand has increased such that pumping has outpaced natural or artificial recharge of groundwater supplies. The results have included lowered water tables, land subsidence and diminished perennial surface water flow resulting in the loss of riparian habitat. In order to best balance the preservation of riparian areas with increased water demand due to growth within their basins, water and land managers first need to understand the hydrology of the riparian area and how it is connected to the near-river riparian aquifer and the 1of13

2 management decisions to preserve surface water flow and riparian vegetation. Figure 1. Map of the study area highlighting sampling locations. The river is divided into perennial and intermittent reaches [Leenhouts et al., 2006] and gaining and losing reaches [Stromberg et al., 2006]. The outline of the gaining and losing reaches corresponds to the extent of the San Pedro River National Conservation Area. Potential groundwater sources regions (G1 G3) are also highlighted; see text and Table 2 for detailed description of these waters. larger, regional, basin aquifer. Essential to this is knowing not only how much water enters the riparian aquifer and the river but the sources of this water, the spatial and temporal variability of these sources. [4] The goal of this study was to address the need for improved understanding of riparian water sources. We focused our sampling of groundwater and surface water sampling on the San Pedro Basin in southern Arizona and analyzed these samples for a suite of geochemical tracers including anions (Cl,F,Br,NO 3 2,NO 2,SO 4 3, and PO 3 4 ), stable isotopes (d 2 H, d 18 O, d 13 C, and d 34 S), and radioactive isotopes ( 3 H and 14 C). We used the results of these analyses to (1) identify the possible water sources; (2) chemically characterize each of these sources; (3) determine which sources contribute significantly to riparian groundwater and baseflow; and (4) quantify the contribution of each source and identify differences along the length of the river. Our hope is that this information will help identify the most important or sensitive source water regions and processes to aid land and water managers as they make 2. Study Area [5] The San Pedro River in southeastern Arizona (Figure 1) is one river in the Southwest that remains perennial in places and supports a locally thriving riparian area. Additionally, aquifers contained within the larger basin provide water supplies to the rapidly growing city of Sierra Vista, a burgeoning rural population, and the Fort Huachuca Military Installation; the population of Sierra Vista alone grew 13.5% over the period 1990 to The basin is bounded on the east by the Mule Mountains and on the west by the taller Huachuca Mountains. Between these two ranges is an extensional basin filled with alluvium to a depth of up to 1,675 meters below land surface (bls) [Gettings and Houser, 2000]. [6] Precipitation within the San Pedro Basin has a generally bimodal occurrence throughout the year. Precipitation totals approximately 353 mm, with 244 mm falling during the summer wet season (mid-june through mid- October), 81 mm during the winter wet season (early December through late March), and the remainder, 28 mm, falling throughout the rest of the year [Hereford, 1993; Pool and Coes, 1999]. Summer season moisture typically originates south of Baja California and moves north through Mexico before entering the basin [Wright et al., 2001]. Summer precipitation transitions from monsoon storms through the early part of the season to moisture from tropical depressions at the end of the season. Monsoon storms are short, intense rainstorms that generate significant amounts of flooding and ephemeral flow (i.e., flow in ephemeral channels on the basin floor) and represent, on average, the bulk of summer moisture. Throughout this article, we will use the words summer and monsoon interchangeably. Winter precipitation primarily results from frontal systems that move east across California, originating in the Pacific Ocean [Sellers and Hill, 1974]. Winter storms are longer, gentler rainstorms that generally result in greater infiltration. The El Niño-Southern Oscillation (ENSO) affects winter precipitation, with a positive ENSO index increasing precipitation [Webb et al., 2004] and recharge rates [Pool, 2005]. [7] Baseflow and storm runoff volumes have both decreased throughout the period of record (since 1905), changes that have, in part, been attributed to shifts in precipitation patterns [Thomas and Pool, 2006]. Historically the Upper San Pedro River had perennial flow from south of Palominas to Charleston (Figure 1), however two of the seven river reaches are now classified as intermittent [Leenhouts et al., 2006]. Evidence would indicate that it may be increasingly intermittent as on June , no flow was recorded at the USGS gage at Charleston for the first time in recorded history [Davis, 2005]. In addition to differences in flow frequency along the river, the character of the hydraulic connection between the river and the riparian aquifer varies, with some sections losing water to the riparian aquifer and others gaining from it. Gaining and losing reaches have been classified based on vegetation type and growth, groundwater hydrology, and frequency of streamflow [Stromberg et al., 2006]. The reach from 2of13

3 Cottonwood to Charleston (Figure 1) is generally gaining because of an impermeable bedrock high present just south of Charleston reduces the vertical thickness of the alluvial aquifer [Pool and Coes, 1999]. [8] Prior water budget estimates for the Sierra Vista Subwatershed circa 1990 (compiled by Pool and Coes [1999]) estimate natural recharge to be between 16,500 and 19,000 acre-feet per year (afy) (205 to km 3 /yr). Well withdrawals amounted to about 11,000 afy ( km 3 / yr), significantly less than the estimated natural recharge. However, the addition of riparian evapotransipration 6,200 afy ( km 3 /yr) and baseflow 5,900 afy ( km 3 /yr) gives a total basin-wide output of 23,100 afy ( km 3 /yr), an amount significantly greater than estimates of natural recharge. 3. Methods [9] We performed sampling throughout the western half of the Upper San Pedro Basin, between the river and the Huachuca Mountains (Figure 1). We chose this study area because the Huachuca Mountains constitute the highest range in the area, and the basin west of the river has the best well coverage in the area. [10] Precipitation samples (54 total) were collected at a site within the City of Sierra Vista (Figure 1) from mid- November 2003 to mid-november Located at an elevation of approximately 1,400 m above sea level (asl), this site is representative of the basin floor. The precipitation collector used typically contained a coating of mineral oil to prevent post-precipitation evaporation. Precipitation samples were analyzed for stable isotopes of hydrogen and oxygen as well as anions in some samples. [11] We collected a total of 36 surface water samples at four sites along the San Pedro River from mid-december 2003 to early November 2004: Palominas, Hereford, Highway 90, and Charleston; see Figure 1. Sample sites were chosen based on accessibility, presence of long-term United States Geological Survey (USGS) stage and chemical monitoring, or proximity to sampled wells. We also incorporated USGS chemical data from these four sample sites from [Pool and Coes, 1999; USGS Water Quality Database, [12] In addition to our precipitation samples, the USGS collected precipitation for a three year period at Greenbush Draw (elevation 1,350 m asl) and the Upper Babocomari River (1,300 m asl) [Coes and Pool, 2005]. Precipitation was collected twice yearly as composite samples from mid- April through mid-october and mid-october through mid- April. To determine average winter and summer values over the entire period we volume-weighted our data and the USGS 6-month composite samples and combined the two datasets. [13] We collected 43 groundwater samples throughout the study area, of which 8 samples from the basin groundwater aquifer and 16 from the riparian area are discussed here. The analysis and interpretation of the 27 spring samples, 6 mountain block well samples, and 13 mountain front well samples are detailed elsewhere [Wahi, 2005]. Prior to collecting a sample, we purged at least three well volumes if a well was not continuously used prior to the site visit. In addition, we monitored ph, electrical conductivity (EC), temperature, and dissolved oxygen (DO) in the purge water; samples were collected only after these parameters were constant, to ensure that the collected sample represented aquifer water. [14] We performed a variety of chemical analyses on the samples, with all sample preparation, extraction, and measurement conducted at the University of Arizona. Anions (F, Cl, Br, NO 2,NO 3,SO 4, and PO 4 ) were analyzed using a Dionex Ion Chromatograph located at the Department of Hydrology and Water Resources following the methods described in the work of Dionex Corporation [2004]. The detection limit for anions was approximately ppm except PO 4, which had a detection limit of 0.10 ppm. Precision was 5% or better for concentrations greater than 1 ppm and 10% or better for concentrations less than 1 ppm. Every tenth sample was analyzed in duplicate. [15] The parameters d 18 O, d 2 H, and d 13 C were measured on a Finnigan Delta S gas source isotope ratio mass spectrometer located in the Stable Isotope Laboratory of the Department of Geosciences according to the methods outlined in the work of Craig [1957] and Gehre et al. [1996]. Precisions are ±0.9% for d 2 H, ±0.08% for d 18 O, and ±0.15% for d 13 C. d 34 S was analyzed on a Finnigan Deltra PlusXL continuous-flow gas-ratio mass spectrometer, with a precision of ±0.15% or better in the Stable Isotope Laboratory of the Department of Geosciences. Samples were prepared by precipitation of BaSO 4 followed by mixing with V 2 O 5, then ignition at 1030 C in a Costech Elemental Analyzer. [16] Tritium was analyzed on a Quantulus 1220 Spectrophotometer located underground at the Department of Geosciences according to the methods described in the work of Theodórsson [1996]. The detection limit for tritium was 0.5 TU, and the precision 0.18 to 0.37 TU. 14 C was analyzed according to the methods in the work of Polach et al. [1973] on one of two General Ionex Accelerator Mass Spectrometers, either 2.5 or 3 MV located in the Accelerator Mass Spectrometry Lab, shared between the Departments of Geosciences and Physics. This analysis has a detection limit of 0.2 pmc and a precision of 0.1 to 0.9 pmc. [17] To quantify water sources to baseflow and riparian groundwater, we performed a simple two-end member mixing analysis. This method is comparable to the commonly used isotopic method for hydrograph separation of storm events into new water and old water [Genereux and Hooper, 1998]. However, it differs in that, rather than looking at one location during a dynamic storm event, we are investigating the spatial variability of sources to integrated riparian groundwater and temporally averaged baseflow. As with the more commonly applied hydrograph separation technique, there is a small amount of error introduced from the uncertainty in the analytical analyses, while the bulk of the error results from uncertainty in the end-member characterization, especially if the end member values are variable in space or time [e.g., Rice and Hornberger, 1998]. To address this we first examined all potential water sources to the riparian area using both isotopic and geochemical data. With our end members identified, we then characterized our end members in d 2 H versus d 18 O space, including the uncertainty in their value, and calculated the percentage of each source contributing to a sample using a linear mixing model. Finally, we evaluated the potential uncertainty in this calculation 3of13

4 with an r 2 value of This equation has a lower slope than the global meteoric water line (GMWL, slope = 8.20 ± 0.07 [Rozanski et al., 1993]). This may indicate that evaporation occurs during precipitation in this semiarid region, or that there is a source area effect decreasing the slope of the equation. This is most pronounced for summer precipitation, which falls along a weighted trendline with the equation (r 2 = 0.810): d 2 H ¼ ð6:8 1:5Þd 18 O þ ð0:4 9:4Þ In contrast winter precipitation has the following relationship (r 2 = 0.975): d 2 H ¼ ð8:6 0:6Þd 18 O þ ð20:4 7:0Þ Figure 2. Stable isotopes for precipitation, groundwaters and surface waters collected in the study area. (a) Average (±1s) for different precipitation, groundwater and surface water groups. Note that all water groups are bracketed between the summer and winter precipitation end members. (b) Detail of all riparian groundwater and baseflow samples. Note that riparian groundwaters span the full range between average basin groundwater (shaded circle) and average summer precipitation (open square) end members whereas baseflow values are more tightly constrained toward the summer precipitation end member. Local Meteoric Water Line (LWML) also depicted. using a method adapted from hydrograph separation studies [Genereaux, 1998]. 4. Results 4.1. Precipitation [18] We analyzed event based precipitation samples, collected at elevation 1,400 m asl, for d 2 H and d 18 O values in order to determine the extent of isotopic differences between summer (April 16 to October 15) and winter precipitation (October 16 to April 15). We chose these periods to reflect the general bimodal nature of precipitation in the study region as mentioned above and to approximately coincide with previous sampling done by the United States Geological Survey (USGS) at two gauges in the area [Coes and Pool, 2005]. Stable isotope values for these two periods of precipitation are distinctly different, with summer rain more enriched in the heavier stable isotopes (Figure 2). A weighted linear regression of all precipitation samples (amount-weighted averages of our data and the USGS data) yields the following relationship between d 2 Handd 18 O values: d 2 H ¼ ð7:1 0:4Þd 18 O þ ð3:1 3:1Þ with a higher slope, indicating that evaporation is a less important control on the isotopic composition of winter precipitation. It should be noted that, while there is some overlap in slope within the 1s uncertainty, the patterns we observe are consistent with many other studies from the Southwest United States [e.g., Wright, 2001; Kalin, 1994] Groundwater [19] This research focuses on basin and riparian wells to resolve surface water-groundwater dynamics in the riparian system. We consider a riparian well to be one located in close proximity to the river, generally within the boundary of the San Pedro Riparian National Conservation Area. These wells are also all below 1,300 m asl in elevation. The results of other groundwater samples collected between the Huachuca Mountains and the San Pedro River are considered part of the basin groundwater system and exhibit no trend in value along this flow path. The range of d 2 H and d 18 O values for basin wells (average 9.6 ± 0.2% d 18 O and 67 ± 1% d 2 H) is narrow compared to the isotopic variability of precipitation input (Figure 2a). Riparian well samples (average 8.1 ± 1.0% d 18 O and 56 ± 5% d 2 H) are more variable than the basin wells, ranging from similar in composition to basin wells to significantly more enriched values. [20] The activity of radioactive isotopes can be used to estimate the mean residence time of groundwater throughout the study area. Basin groundwater varies from a high of 2.1 TU near the mountain front to below detection in the middle of the basin. Radiocarbon activity in basin groundwater ranges from 33 to 90 pmc, again with the highest activities near the mountain front. These results are consistent with groundwater flow from the mountains to the river, and also indicate that basin floor ephemeral channel recharge, which occurs during and immediately after intense precipitation in channels that are dry most of the year, does not appear to significantly alter the chemical composition of basin groundwater wells. Recent results based on two years of sampling [Coes and Pool, 2005] have estimated basin floor ephemeral channel recharge at 2,400 afy ( km 3 /yr). At 13 to 15% of the basin groundwater budget this may be consistent with the results of this study because: (1) only the shallowest wells could detect this portion of recharge compared with deeper wells; (2) the isotopic data represent long-term recharge averages which may be lower than the two year data set of Coes and Pool; 4of13

5 Figure 3. Riparian groundwater (circles) and average baseflow (triangles) chemistry at four sampling sites plotted versus distance downstream from the US-Mexico border. Baseflow values represent average values with ranges representing ±1s of samples collected at four sites from 1994 to 2005 by the USGS [Coes and Pool, 2005], and 2003 to 2004 in the course of this study. (a) d 18 O values at each site; values on the right hand side represent the percentage of monsoon floodwater for samples based on two end member mixing analysis described in the text. (b) SO 4 /Cl ratios, also noted are the range of SO 4 /Cl in monsoon floodwaters and basin groundwater (Table 2). (3) given higher horizontal flow rates compared to vertical flow the ephemeral channel recharge may not mix with the larger basin aquifer, and (4) spring and summer ephemeral channel recharge would be efficiently captured by vegetation during the growing season. [21] As with the stable isotopes, there is greater variation in radioisotope activity in the riparian wells than in the basin wells. Tritium activity in riparian groundwater ranges from no detectable tritium to values as great as 5.2 TU, with different vertical gradients in different parts of the riparian area. In some areas, like Palominas, the highest level of tritium is found in a medium-depth well, about 20 meters bls, whereas in other areas such as Cottonwood, the highest tritium activity is found in the shallowest well, about 5 meters bls. Radiocarbon was sparsely sampled along the riparian area, but varies from as low as 11 pmc up to 83 pmc. The large range in residence time tracers in the riparian wells reflects the dynamic mixing between older basin water and younger floodwater recharge. The lowest activity of 11 pmc occurred in a deep well (200 m), and is indicative of basin groundwater, whereas higher values at shallower depths are more indicative of mixed recharge sources. [22] In addition to stable isotopes of H and O and radioisotopes, other isotopes and solute ratios provide potentially useful flow path tracers. We evaluated Cl/Br but found that it did not vary significantly; basin wells average 115 ± 43 (n = 7) and riparian wells average 99 ± 15 (n = 13). These values are typical of a precipitation source for chloride [Davis et al., 1998] indicating that this is likely the only source of chloride within the basin, as other sources typically have much higher ratios. In contrast, SO 4 /Cl values exhibit significant variation. In general, chloride and sulfate are greatest in the riparian wells (7.4 ± 4.1 mg/l Cl and 28.7 ± 27.0 mg/l SO 4 ) and mountain wells [Wahi, 2005], and lowest in the basin wells (4.8 ± 2.2 mg/l Cl and 5.9 ± 3.7 mg/l SO 4 ). As a result, SO 4 /Cl ratios are high in the mountain wells [Wahi, 2005], decreasing to the lowest values in the basin wells (1.2 ± 0.4), and increasing again in the riparian wells (3.5 ± 2.1) to levels similar to the mountain wells. Averages represent the mean of all samples ±1s. The sulfur isotopic composition of sulfate however showed little systematic variation with basin wells having an average d 34 S value of 6.1 ± 1.9% (n = 6), whereas riparian wells average 6.3 ± 2.7% (n = 12). In contrast, d 13 C values show some variation (P = 0.1), with basin wells averaging 7.6 ± 1.3% (n = 8) and riparian wells averaging 9.4 ± 1.6% (n = 4). The more enriched d 13 C values in the middle of the basin may indicate carbonate dissolution along the groundwater flow path as carbonates have a d 13 C value around 3% [Wahi, 2005]. Other analytes did not show significant or patterned differences between the basin and riparian groundwater Surface Water and Riparian Groundwater [23] The chemistry of surface water in the San Pedro River varies both temporally and spatially. Stable isotope values and SO 4 /Cl ratios have similar patterns of spatial variability (Figure 3), with stable isotopes most enriched and SO 4 /Cl ratios highest at Palominas, after which SO 4 /Cl ratios quickly decrease and stable isotopes become more depleted along the rest of the river. The riparian groundwater broadly mimics this spatial pattern, although in almost all cases groundwater is more depleted in stable isotopes and has lower SO 4 /Cl ratios than the nearby surface water. [24] The isotopic compositions of surface water samples at our four sampling sites are intermediate between basin groundwater and summer precipitation and similar to riparian groundwater (Table 1 and Figure 2). The average isotopic and chemical progression mirrors the geographical progression of the sites; Palominas is furthest upstream as well as the only losing reach among the four locations, and is the most enriched isotopically; Charleston is furthest downstream, and most depleted (Figures 2b and 3; Table 1). [25] The stable isotopic composition of floodwater is slightly more depleted and more variable than that of baseflow (Table 1). Limited samples (n = 6) indicate that winter floodwater, as expected, is more similar to winter precipitation than is summer floodwater; however, there is not a great degree of separation between the two seasons, likely indicating that isotopic composition of floodwater samples of both seasons, which were collected directly from 5of13

6 Table 1. Summary of Isotope and Chemical Characteristics of Surface Water Samples Type Location Season d 18 O d 2 H SO 4 Cl SO 4 /Cl Baseflow All All 7.2 ± 0.7% 51 ± 6% 38.2 ± ± ± 2.2 Charleston All 7.5 ± 0.4% 55 ± 2% 27.7 ± ± ± 0.7 Highway 90 All 7.2 ± 0.7% 51 ± 6% 41.5 ± ± ± 1.2 Hereford All 6.9 ± 0.4% 48 ± 5% 37.0 ± ± ± 1.1 Palominas All 6.2 ± 0.6% 45 ± 6% 64.4 ± ± ± 1.8 Floodwater All All 7.9 ± 1.9% 57 ± 16% 46.7 ± ± ± 3.8 All Winter 8.6 ± 1.7% 61 ± 17% 63.5 ± ± ± 3.5 All Summer 7.5 ± 2.0% 54 ± 16% 32.7 ± ± ± 4.2 the river, are a mixture of seasonal precipitation and baseflow (a topic discussed in more detail later). SO 4 /Cl ratios are not markedly different between seasons (Table 1). The SO 4 /Cl ratio in floodwater is higher than in baseflow and, as with the isotopes, intermediate between ephemeral runoff (17.2 ± 16.6 [Goodrich et al., 2004]) and baseflow, indicating that floodwaters collected from the San Pedro are a mixture of the two sources. 5. Discussion [26] Determining the dominant sources of water in the riparian aquifer system helps identify the hydrologic processes that, if altered by land use changes, would have the most impact on the riparian ecosystem. In order to determine the relative contributions of water sources to the riparian system we used the results of these analyses to (1) identify and chemically characterize the possible sources contributing to riparian groundwater and baseflow; (2) qualitatively evaluate which sources contribute significantly to riparian groundwater and baseflow; and (3) quantify the contribution of each source and identify differences along the length of the river Potential Riparian Water Sources [27] Riparian groundwater and baseflow have five possible sources that fall into two categories: inflow of basin groundwater and local recharge of precipitation-derived floodwaters within the riparian corridor. Before we can quantify the relative importance of each source to the riparian waters, we must describe them chemically. To do so, we use two sets of analytes that respond to different processes; d 18 O and d 2 H vary with both elevation and seasonality of precipitation, and the SO 4 /Cl ratio changes in response to processes along the groundwater flow path. Because the processes affecting these analytes occur to different degrees for each water source, each source has a specific chemical signature (Table 2) Basin Groundwater Sources [28] Basin groundwater enters the riparian aquifer as lateral or upward flux from the regional aquifer. Considering the geology and hydrology of the basin, we recognized three possible basin groundwater sources: northward flow from Mexico (G1), eastward flow from the Huachuca Mountains (G2), and westward flow from the Mule Mountains (G3) (see Figure 1 for general locations of these waters). While the majority of our sampling focused on G2, limited data for G1 and G3 are available from Pool and Coes [1999]. G1 is typified by low SO 4 /Cl ratios and midrange isotopic values (Table 2 and Figure 4). The source of this water is precipitation that falls on the headwaters of the San Pedro River; the low elevation of this region, compared to the Huachuca Mountains to the west, results in heavier isotopic compositions. It should be noted that, while all wells used to describe the composition of G1 are located on the Arizona side of the international border, they are presumed to represent groundwater moving north from Mexico as other sources are unlikely to be present in notable fractions. The low SO 4 /Cl ratio, typical of all basin groundwater in the study area, is caused by a lowering of the amount of sulfate in the water away from recharge zones, the potential reasons for which will be discussed later. G1 seems to not be highly variable, although it is not well described due to a lack of data (n = 2 for anions, n = 3 for stable isotopes); one sample was obtained during the course of this study and the other two were described by Pool and Coes [1999]. [29] Groundwater that flows east from the Huachuca Mountains (G2) toward the San Pedro River is a mixture of mountain block recharge and mountain front recharge from the Huachuca Mountains. Because most development is on the west side of the basin, there is a large amount of chemical data available for characterizing G2. G2 groundwater has a low SO 4 /Cl ratio, and light isotopic values (Table 2). The light isotopic composition of this water is partly due to the higher elevation of the catchment and is Table 2. Potential Riparian Water Sources and Their End Member Compositions d 18 O d 2 H SO 4 /Cl Basin Groundwater a (Mexico, n = 3) G1 7.7±0.6% 55±4% 1.7±1.1 Basin Groundwater (Huachuca Mountains, n = 8) G2 9.6 ± 0.24% 67 ± 1% 1.2 ± 0.4 Basin Groundwater b (Mule Mountains, n = 2) G3 7.8 ± 0.4% 56 ± 0.4% 2.1 ± 1.7 Summer Runoff R1 6.3 ± 0.99% c 42 ± 8% c 17.2 d Winter Runoff R ± 2.8% c 76 ± 24% c 17.2 d a Data from Pool and Coes [1999] and this study. Data from Pool and Coes [1999]. c Averages of summer and winter precipitation from USGS data and from this study. d Walnut Gulch floodwater data from Goodrich et al. [2004]; SO 4 /Cl ranges from 5.84 to of13

7 Figure 4. Mixing plot for groundwater and baseflow samples using SO 4 /Cl ratios versus d 18 O values. In the riparian aquifer, wells vary from nearly the same chemistry as basin groundwater (G2) to similar to monsoon floodwater (R1) and fall along a calculated mixing line (10% increments are marked) between these end members. Dashed lines represent a range of potential mixing curves using the uncertainty ranges in G1 and R1 end members. These end members and other possible water sources are discussed in the text. A slash through a riparian groundwater datapoint indicates a well sample containing tritium greater than 1 TU whereas a datapoint with an X indicates a sample with tritium activity of 1 TU or lower. Other sources indicated on the figure are winter floodwater (R2), groundwater from the south (G1), and groundwater from the east (G3). also a result of the dominance of winter precipitation (about 70%) to mountain block and mountain front recharge in the Huachuca Mountains (discussed in detail by A. K. Wahi et al. (Geochemical quantification of recharge from a semiarid mountain range, submitted to Ground Water, 2007)). [30] G3 groundwater, representing westward flow from the Mule Mountains, is chemically similar to G1. It exhibits a low SO 4 /Cl ratio and mid-range isotopic values (Table 2). This isotopic composition is also indicative of a lowerelevation catchment area than that for G2. Water on this side of the river, like G1, is poorly characterized, using only two data points from Pool and Coes [1999]. However, for both samples it is clear that G3 differs significantly in isotopic composition from G Riparian Recharge Sources [31] During flood events, river stage increases, creating a local hydraulic gradient that slopes away from the river; this is particularly true during the summer when riparian evapotranspiration (ET) lowers the water table. When this happens, floodwater recharges to the riparian aquifer through bank infiltration. Geochemically this water can be divided into two sources, corresponding with the two rainy seasons: summer or monsoon riparian recharge (R1) and winter riparian recharge (R2). [32] Studies of summer ephemeral channel runoff indicate that the isotopic composition of ephemeral channel floodwater does not vary significantly or systematically from the isotopic composition of the precipitation from which it was derived [Goodrich et al., 2004]. Thus to characterize the summer and winter riparian recharge sources, we use the isotopic values determined for summer and winter precipitation at our gauge and the USGS gauges (Table 2). [33] For SO 4 /Cl ratios we cannot use our precipitation samples for end member values as sulfate is increased in floodwaters relative to precipitation due to dissolution of soluble sulfate-bearing surface salts following precipitation. The SO 4 /Cl ratios for these floodwater end members must therefore be determined directly from ephemeral channel samples. The SO 4 /Cl ratio in monsoon runoff has been shown to average 17.2 ± 16.6 in the Walnut Gulch Experimental Watershed, about 13 km northeast of the north end of the study area [Goodrich et al., 2004]. This number represents the volume-weighted average of 8 runoff events, and has a high standard deviation due to extreme high (52.4) and low (3.46) events. The extreme SO 4 /Cl ratio storms, however, have smaller discharge, and when volume weighted for the entire monsoon season the ratio is To provide a reasonable uncertainty in this end member we have used the average value with a range that excludes the high and low events (5.86 to 26.6). The Walnut Gulch watershed has no perennial flow, so we can assume runoff samples to be representative of runoff derived from precipitation. San Pedro River floodwater samples averaged 7.0 ± 4.2 and 7.6 ± 3.5 in summer and winter respectively, reflecting mixing between the higher values in ephemeral channel waters and the lower precipitation values, which average 4.4 ± 2.1 (n = 12) during the summer and 3.4 ± 2.2 (n = 6) in winter. As there has been no analysis of winter ephemeral channel runoff, and because we have no reason to suspect seasonal variability in SO 4 /Cl ratios, we use the same values for winter runoff as we use for summer runoff 7of13

8 (Table 2). The lack of a distinct seasonal pattern in San Pedro flood waters supports our use of the Walnut Gulch end member for both summer and winter precipitation SO 4 / Cl ratio. [34] R1 is isotopically heavy ( 6.3 ± 1.0% d 18 O; 42 ± 8% d 2 H), and has a high SO 4 /Cl ratio as discussed above. R2 exhibits the lightest isotope composition in the entire system ( 11.2 ± 2.8% d 18 O; 76 ± 24% d 2 H) (Figure 4). The mean SO 4 /Cl ratio set for floodwater is significantly higher than that of any of the basin groundwater sources. The elevated level of this ratio in all floodwater samples illustrates the value of this ratio as a flow path tracer (basin groundwater vs surface runoff and recharge); however the highly variable nature of these sources precludes the use of SO 4 /Cl in a quantitative mixing analysis for these sources. [35] The higher SO 4 /Cl ratio in floodwater sources relative to older basin groundwater sources indicates that one of two processes is occurring in the study area: sulfate may be reduced to hydrogen sulfide after a sufficient residence time in the aquifer; alternatively, the input of sulfate into the system may be higher today than in earlier times, due perhaps to open-pit copper mining or other industrial activities in the region. Waters with low SO 4 /Cl ratios also exhibit low radioisotope activities (r 2 value of 0.66 for a linear regression of SO 4 /Cl vs. 3 H), but many samples with low SO 4 /Cl ratios have detectable NO 3 or elevated dissolved oxygen, inconsistent with an environment where sulfate reduction would occur [Langmuir, 1997]. Lack of long-term precipitation or surface water samples for the region makes it challenging to directly evaluate the hypothesis that the amount of sulfate input into the basin has changed Qualitative Assessment of End-Members [36] Using SO 4 /Cl ratios and d 18 O values (Figure 4), we can see that almost all of the riparian wells and baseflow samples plot near a mixing line between G2 (Huachuca Mountain-derived groundwater) and R1 (summer runoff), indicating that these are likely the dominant sources to the riparian system. Before we can evaluate mixing between these two sources we must qualitatively analyze the possibility of input by the other three end-members, as well as the reasons for data points that lie off a mixing line that uses the average values of G2 and R1 as its end members. [37] If riparian groundwater or baseflow contained a significant proportion of R2 (winter runoff) it would have light d 2 H and d 18 O values and a high SO 4 /Cl ratio. A few riparian groundwater samples (Figure 4) are shifted slightly toward the R2 end member, but age tracer data indicates that, except for one sample, this water is quite old, suggesting a basin groundwater source. The slightly increased SO 4 /Cl ratio in these wells may instead be the result of the dissolution of a small amount gypsum or a similar sulfatebearing evaporite. Winter floodwater, therefore, does not appear to contribute a significant amount of recharge to the riparian aquifer. [38] Groundwater containing a significant proportion of G1 (Mexico-derived groundwater) or G3 (Mule Mountainderived groundwater) would have a moderate isotopic value ( 8%) and a low SO 4 /Cl ratio. If input of G1 were significant it would be expected to be greatest at Palominas, the southern end of the study area. The deepest riparian well at Palominas (59 m) is likely representative of G1; however, there is a sequence of low-permeability silt and clay layers present in this area that restricts communication between the deeper regional aquifer and the shallow riparian aquifer (D. R. Pool, personal communication). Shallow groundwater at Palominas falls within the range of R1 consistent with this losing reach of the river and indicates that G1 is not present in the shallow riparian aquifer at this location. However, a shallow well (5 m) at Kolbe (10 km downstream of Palominas) also seems similar to G1 (or G3). Three wells at Hereford (5 km downstream from Kolbe), ranging from shallow (25 30 m bls) to deep ( m bls), plot very close to the average G2-R1 mixing line or above it, indicating that G1 is unlikely to contribute significantly to the riparian aquifer north of this point. [39] In contrast, G3 may be a significant local input to baseflow along the length of the study area. For example, groundwater discharging at Lewis Springs (located between Highway 90 and Charleston), on the east side of the river, is chemically similar to G3 (the most likely source given its location). However, there are several factors that likely limit the G3 contribution over most of the study area: (1) the hydraulic gradient on the east side of the river is lower than on the west side, and also historically ran generally northward, trending northwest toward Lewis Springs in the northern part of the study area [Corell et al., 1996]; (2) the Mule Mountains are lower than the Huachucas, and thus receive less precipitation and therefore less recharge; and (3) the west side of the Mules faces southwest, reducing the amount of available water for recharge due to differences in net radiation. Ideally we would use a three-end member mixing model to quantify the contribution of this potential source; however, this is limited by the large uncertainty in the SO 4 /Cl ratio of the R1 end member. The qualitative assessment can be graphically demonstrated with the range of potential G2-R1 mixing lines (Figure 4), which are capable of explaining most all samples Quantification of Riparian Groundwater Sources [40] Using the end member values for G2 and R1, (Table 2) we are able to perform a two end member mixing model calculation to quantify the relative proportion of these sources in riparian groundwater. The wide range of stable isotope values in riparian groundwater demonstrates a great deal of variability in the importance of G2 vs. R1; this mixing analysis indicates that riparian groundwater ranges between 10 ± 7 and 85 ± 26% G2 (Figures 3 and 5; uncertainty estimates described in detail later). Given this large amount of variation it is critical to evaluate the temporal and spatial patterns of this variability to best understand the processes that control the variability in water source. [41] To evaluate annual and inter-annual variability in the riparian aquifer, we compared our data with samples collected by the USGS from 1994 to 1996 [Pool and Coes, 1999]. Several wells were sampled by the USGS during both the spring dry season and the monsoon season of the same year. Two shallow wells (located at Hereford and at Hunter, halfway between Hereford and Cottonwood, see Figure 1) completed in the post-entrenchment alluvium showed a slight decrease from the spring dry season to the monsoon season ( 0.3% d 18 O and 0.5% d 2 H). Seven deeper wells completed in the pre-entrenchment alluvium 8of13

9 Figure 5. SO 4 /Cl ratios for San Pedro River baseflow collected at four sites along the river over 2003 and At any individual site SO 4 /Cl ratios are steady throughout the year, until the monsoon season. Note ranges of different potential water sources on right side of graph (Table 2). showed a slight increase (0.4% d 18 O and 1.5% d 2 H) over the same time period. This change between seasons is nearly the same as the error of the analysis (0.1% d 18 O and 1.0% d 2 H), indicating that there is only limited annual variability in the riparian aquifer. For inter-annual variability, we can evaluate two wells (at Palominas and Highway 90) that were sampled repeatedly in the summer from 1994 to 1996 (by the USGS), with one well also sampled in 2004 (this study). The first well, which was sampled four times over 11 years, shows no significant change over the period of record. The second well, with two samples taken over two consecutive summers, shows a change of +0.57% d 18 O and +1.8% d 2 H over that year. With the limited data available, it appears that riparian groundwater composition is relatively stable over the ten year period sampled both in terms of the isotopic composition as well as the relative contribution of water sources. [42] In contrast, there is significant spatial variability within the riparian aquifer. Certain parts of the aquifer are dominated by monsoon floodwater, whereas other areas are dominated by basin groundwater (Figure 3). Given this clear spatial pattern one possible explanation is that the isotopic composition, and hence the riparian groundwater source, of a given reach reflects the hydrologic condition. Specifically, gaining reaches will be dominated by basin groundwater whereas losing reaches will be dominated by recharge of monsoon floodwaters, something previously noted by Pool and Coes [1999]. To test this we compared our results with an independent index of gaining and losing reaches. [43] Stromberg et al. [2006] used an objective method of assigning a condition class to different parts of the San Pedro River using vegetative and hydrologic indicators that relate to gaining and losing conditions. A condition class value of less than 2.5 indicates a losing reach of the stream, whereas 2.5 or greater indicates a gaining reach (Figure 1). If we compare the condition class values of Stromberg et al. to the chemistry of well samples in the same reach, we find that all riparian wells dominated by basin groundwater (>60%) are located in reaches with condition class values greater than 2.5. All wells located in reaches with condition class values less than 2.5 contain 70% or more monsoon recharge. The only deviations from a perfect correlation between condition class and chemistry are two very shallow wells (5 m) very close to the river bed (<10 m) that are located in gaining reaches but are dominated by monsoon floodwater (65 to 80%). This fits a pattern in the data where shallow wells in a given reach generally contain a higher percentage of monsoon floodwater than do deeper wells within the same reach. [44] A potential complication to a simple mixing analysis is the effect of evaporation on the stable isotopic composition of the recharged water. Because of the position of riparian groundwater samples on Figure 2b, it appears that riparian groundwater has undergone some evaporation. Using an evaporative enrichment line with a slope of 4 appropriate to the environment of the study area [Clark and Fritz, 1997], we can back out the effect of evaporation on the system. Extrapolating samples back along this slope to the mixing line between G2 and R1, the degree of isotopic enrichment due to evaporation can be estimated. The isotopic shift of the samples averages 0.6% d 18 O and 3% d 2 H, with maxima of 1.5% d 18 O and 6% d 2 H. Using the method presented in the work of Clark and Fritz [1997] this equates to about a 5% evaporative loss of riparian groundwater on average. If this evaporative effect were to occur after the mixing of the waters this would also imply a 15% overestimate in the amount of monsoon floodwater recharge in the quantitative mixing analysis on average with a maximum error of 36%. While potentially significant, this does not change the overall pattern of water sources or the importance of monsoon floodwater recharge for riparian groundwater in losing reaches. It is also important to consider that we use an end member composition based on precipitation and a significant amount of this evaporation 9of13

10 may occur during runoff generation prior to mixing within the riparian system. [45] Finally it is worth noting that a qualitative analysis of SO 4 /Cl ratios confirms the mixing quantification based on isotopes. All riparian wells together average 3.5 ± 2.1 (±1s). Wells in losing reaches average 4.7 ± 2.9, and wells in gaining reaches 2.7 ± 0.9 (Figure 3b). Wells that are similar isotopically to basin groundwater generally, though not always, have low SO 4 /Cl ratios, and wells that are isotopically similar to monsoon recharge generally have high SO 4 /Cl ratios Quantification of Baseflow Sources [46] Using the end-member isotope values for G2 and R1 (Table 2), we can quantify the sources of baseflow using data collected by the USGS at four sites from [Pool and Coes, 1999] and samples collected during the course of this project (Figure 3a). Palominas baseflow is dominated by monsoon precipitation and contains on average just 20 ± 28% basin groundwater. Hereford averages 31 ± 23% basin groundwater, and Highway 90 averages 42 ± 20%. At Charleston, basin groundwater input contributes an average of about 45 ± 19% of baseflow. [47] Spatially, we see a progressive increase in basin groundwater as the river passes through gaining reaches. This pattern in baseflow broadly mimics variability in the riparian groundwater (Figure 3), however baseflow composition exhibits a greater percentage of monsoon precipitation in almost all areas than co-located riparian wells (Figure 3). This would indicate that the riparian aquifer and the river are well connected in most locations. The higher values may be the result of upstream contributions from losing reaches (monsoon dominated) or the importance of monsoon floodwater in the near stream area even in gaining reaches (see discussion of riparian groundwater) both of which could keep baseflow values higher than the local groundwater. Finally, the low temporal variability in both d 18 O values and SO 4 /Cl ratios (Figure 5) at sites like Charleston may be indicative of a well-mixed groundwater source contributing to baseflow, whereas higher variability at Palominas may indicate floodwater dominance, a short flow path, or a local source for baseflow. [48] The prevalence of monsoon floodwater in baseflow relative to riparian groundwater, as well as the increasing contribution of basin groundwater to riparian groundwater with increased depth in riparian wells, indicates recharge of the shallow aquifer occurs during monsoon flood events. Throughout the rest of the year, this floodwater then discharges into the river and flows out of the study area. We observe an annual trend in baseflow where baseflow is most similar chemically to monsoon floodwater during and immediately after the monsoon season (Figure 5). The rest of the year sees baseflow chemistry trending steadily toward the chemistry of basin groundwater. The degree of this shift changes from site to site, with Palominas (potentially the greatest monsoon flood recharge) seeing the most extreme annual shift and Charleston the smallest (potentially the least monsoon flood recharge). [49] Finally, compared to riparian groundwater, evaporation has a potentially greater effect on the isotopic composition of baseflow samples regardless of season (Figure 2). Following the same analysis as was used on riparian groundwater, the average enrichment for baseflow is 1.0% d 18 O and 4% d 2 H, with a maximum of 2.5% d 18 O and 10% d 2 H. This evaporative shift would result in a 25% error in our mixing analysis on average, with the maximum evaporative shift introducing a potential error of up to 57%. The average evaporative shift indicates evaporation of about 7% of the water, although we cannot differentiate evaporation before recharge (as runoff) from that after discharge as baseflow. The maximum shift represents evaporation of about 14%. Evaporative shift is slightly higher in baseflow samples between March 1 and June 30 versus the rest of the year (8.3 ± 2.7% versus 5.3 ± 4.9%), although this difference is not significant Quantification of Uncertainty in Mixing Equations [50] We have already discussed uncertainty in quantification of baseflow and riparian groundwater sources related to the selection of proper end member waters, the possible effect of evaporation on our mixing analysis, and the problem of large uncertainty in end member chemistry. Uncertainty is also a function of the accurate quantification of end members values used in the mixing analysis as well as of the analytical uncertainty of the baseflow or groundwater sample. To evaluate the uncertainty of the mixing results, we followed the method of Genereux [1998], using the uncertainties in R1 (0.99% d 18 O, 8% d 2 H) and G2 (0.24% d 18 O, 1% d 2 H) reported in Table 2, as well as the analytical uncertainty associated with sample analysis (0.08% d 18 O, 0.9% d 2 H). These values are constant for all samples; however since the errors in the end members are not equal the resulting uncertainty depends on the location of the sample on the mixing line. [51] Uncertainty for all samples ranges from 7% for those samples close to the G2 end member to 48% for samples dominated by R1, averaging 15% in riparian groundwater and 21% in baseflow samples. Analytical uncertainty is responsible for less that 1% uncertainty in the mixing analysis and likewise the uncertainty in the G2 end member results in uncertainties of 1.5% in riparian wells samples and 0.15% in baseflow samples. In contrast the R1 end member for almost all samples is the dominant source of uncertainty, averaging 13% in riparian well samples and 21% in baseflow samples, roughly three-quarters of the uncertainty for riparian groundwater samples and virtually all of the uncertainty in baseflow samples. The R1 end member is based on precipitation samples collected during four summers; given the variable nature of precipitation it is likely that a longer isotopic record could reduce the uncertainty in this end member Implications [52] The results of this study have several implications both for understanding of the role of monsoon flood events on the hydrology and biogeochemical cycling of semiarid riparian systems as well as for the best management of this important area of biodiversity. Throughout the semiarid Southwest water managers are increasingly faced with balancing rapid population growth against riparian sustainability. Our methods can be applied throughout the Southwest and in similar climates worldwide. [53] First, the finding that monsoon floodwater recharge is a dominant baseflow water source year round (45 to 100%) 10 of 13