California's widespread and economically important vineyards offer substantial opportunities to understand the interface between hydrology and biogeochemistry in agricultural soils. The common use of native sulfur (S) as a fumigant or soil additive provides a novel way to isotopically differentiate among sulfate (SO42−) pools, allowing the estimation of water and SO42− budgets. The objectives of this study were (1) to characterize the near-surface hydrological flow paths in a vineyard during irrigation and storm events and (2) to determine how those flow paths affect the fate and transport of SO42− across seasons. Integrating hydrological theory with measurements of SO42− concentration and sulfate-S isotopic ratios (expressed as [SO42−] and δ34S, respectively) in inputs, soil water, and leachate provided a means of determining flow paths. Low [SO42−] and δ34S in leachate during 4-h irrigation events reflect minimal engagement of the soil matrix, indicating that preferential flow was the dominant path for water in the near surface. In contrast, high [SO42−] and δ34S values during 8-h irrigation and storm events reflect near-complete engagement of the soil matrix, indicating that lateral flow was the dominant pathway. Because hydrologic response and SO42− mobility are tightly coupled in these soils, the magnitude of water fluxes through the near surface controls S cycling both on and off site. These results indicate that preferential flow is an important loss pathway to consider in managing both water resources and water quality (reactive elements) in vineyard land use systems.
 Understanding the interactions between hydrological pathways and biogeochemical processes in soil is essential for improving management of water and nutrient resources in agricultural systems. Hydrological mechanisms modulate biogeochemical processes that sustain soil nutrient content and regulate the fate and transport of elements. Sulfur (S) is a reactive element that interacts with and affects other constituents depending on hydrologic conditions [e.g., Mayer et al., 1995; Alewell and Gehre, 1999; Alewell and Novak, 2001]. High losses of S as sulfate (SO42−) from terrestrial ecosystems can cause degradation of soil fertility and acidification of soils and surface waters [e.g., Adams et al., 1997; Likens et al., 2002; Lamontagne et al., 2006], and affect the cycling of heavy metals in wetlands [e.g., Bates et al., 2002].
 Vineyard agriculture presents a special case at the interface of hydrology and biogeochemistry. Vineyards require highly intensive water and nutrient management, and they occur under a wide range of topographic, geologic, and pedologic regimes. In one premier wine growing region, Napa Valley, California, the growth of grapes during the dry season (April through October) is facilitated by deficit (i.e., small amounts) drip irrigation. Growers must collect water during the preceding wet season (November through March) and store until needed in reservoirs, or pump groundwater. In many regions of the world, drip irrigation is known to increase soil salinity, which can lead to negative impacts on plants, including osmotic, nutritious, and toxic stresses [Badr and Taalab, 2007, and references therein]. However, this issue is not widely reported by Northern California wine grape growers; a larger emphasis is placed on minimizing irrigation amounts and losses to conserve limited water resources.
 The path water takes in the near-surface determines soil water content and residence time, important components of the regional hydrological budget. Past research has shown that vineyard soils experience the gamut of hydrologic responses during the growing [e.g., Oliveira, 2001; Grote et al., 2003] and dormant [Ramos and Martínez-Casasnovas, 2004] seasons. Major flow paths within the near-surface include (1) preferential flow, defined here as rapid water movement via cracks, macropores, and/or pipes to the subsurface, and (2) soil water movement, or slow-moving water flow through the soil matrix.
 Nutrient alterations in vineyards are intensive as well. Large quantities of quick-reacting elemental S (S0) are applied each growing season to kill the pest Uncinula necator, which causes powdery mildew. Elemental S, as micronized particles dispersed in water or fine dust, is coarsely broadcast by tractor or helicopter. On the soil surface, it quickly oxidizes to SO42− and moves into soil matrix, where it is subject to microbial transformations (E. S. Hinckley et al., High sulfur inputs in northern California vineyards: Short-term fates and long-term implications, manuscript in preparation, 2008). On the basis of the characteristics of dominant near-surface flow paths in vineyards, it is likely that SO42− carried in preferential flow is transferred directly from the soil surface to the subsurface, with minimal soil matrix interaction. In contrast, SO42− carried in soil water has been subject to microbial transformations, such as oxidation of sulfides, mineralization of sulfate (SO42−) from organic matter, and SO42− reduction.
 Stable isotopic ratios are a powerful tool for studying biogeochemical cycling and sources of S [Krouse and Grinenko, 1991; Bates et al., 2002]. Research in forested ecosystems has shown that the δ34S values of SO42− in soils and drainage waters are controlled by the isotopic contribution of sources (i.e., atmospheric deposition, mineral weathering, and net mineralization of organic S) and isotopic fractionation during microbial transformations in soil [Mayer et al., 1995; Alewell et al., 1999, 2000; Mitchell et al., 2001; Shanley et al., 2005; Mitchell et al., 2006]. The largest isotope fractionations are caused by SO42− reduction; microbially generated sulfides have much lower δ34S values than the initial SO42−, resulting in a pool of residual SO42− with high δ34S values. In contrast, immobilization, mineralization, and sulfide oxidation cause small changes in δ34S values. In vineyards, the δ34S of SO42− can be used as a tracer of applied S and water, providing that the isotopic compositions of these inputs differ from those of stored soil water and S.
 This study examines the contribution and timing of near-surface hydrological flow paths to total water and SO42− losses during irrigation and storm events in a Napa Valley vineyard. The aim of this research is to determine the interaction of flow paths and SO42− transport in a typical grape-growing location. This study is not intended to provide a spatially explicit, comprehensive investigation of water and nutrient losses from vineyards across the Napa Valley. The integrated approach includes applying hydrological theory to estimate flow velocities in the near surface, directly measuring water losses using zero-tension lysimeters, and using SO42− concentration and δ34S of SO42− values (expressed as [SO42−] and δ34S, respectively) of inputs and soil water to determine the flow paths contributing to water losses during irrigation events.
2. Site Description
 The study area, shown in Figure 1, is a 1.5 ha vineyard block in the Carneros region of Napa County, California (N 38°27′, W −122°33′). It is located at the base of a hillside, surrounded by other vineyard blocks and access roads. On the basis of surface topography, saturated subsurface flow is dominantly SW to NE. Soils are Bressa-Dibble complex [U.S. Department of Agriculture, 1978] with a clay loam texture to approximately 0.4 m depth, underlain by a sandy clay hardpan. The hardpan is consistent across the Carneros region and creates perched water tables during periods of saturation. Vines are Dijon 114 Pinot noir grafted on 101–14 rootstock, planted in rows perpendicular to the dominant slope with 1.8 m spacing, and were 14 years old at the time of this study. Each year, growers plant a dormant season cover crop (including Rosa, Trifolium, and Triticale spp.) between vine rows to maintain soil nutrients and decrease erosion.
 The average annual rainfall in this region of Napa County is 660 mm (WRCC data are available at http://www.wrcc.dri.edu/), falling primarily from October to April. Growers at the study area collect rainfall and divert streamflow from Old Sonoma Creek to store water in a surface reservoir for irrigation. From June through August, vines are drip irrigated weekly for 4 h, at a rate of 4 L h−1 vine−1. Following harvest (late August), growers apply an 8-h irrigation at the same rate to flush solutes below the vine rooting zone. Since 1993, the California Irrigation Management Information System (CIMIS) has maintained a weather station 6.5 km from the site (Carneros 109). Rainfall data from the CIMIS station and grower-reported irrigation data are used in this study (CIMIS data are available at http://wwwcimis.water.ca.gov/cimis/data.jsp).
 From April through June, growers at the study area apply S0 weekly to vines at a rate of 5–7 kg ha−1, increasing to 13–20 kg ha−1 midway through the growing season. Elemental S is the dominant source of S to the vineyard. Additional S sources include gypsum (CaSO4) and organic S (in compost) applications, and dissolved S in precipitation and irrigation water.
3.1. Sample Collection
 Zero-tension lysimeters were used to sample gravity-driven leachate during irrigation events, and tension lysimeters were used during periods of near saturation and saturation to sample soil water. Thirty-two 1 m soil pits were dug and instrumented with zero-tension lysimeters. A 0.1 m diameter PVC pipe was cut to 0.31 m and halved lengthwise to create the lysimeter body. The lysimeter was filled with pea stones and covered with plastic screening to reduce the accumulation of fine particles. Each end was capped and a 10 mm portal was drilled into one end, fitted with a small plastic connector, filled with glass wool, and connected by a plastic tube to the 2 L collection bottle (design modified from Jordan ). The lysimeters were installed 0.36 ± 0.01 m from the soil surface, directly under a vine, and angled 30° relative to the soil surface to allow water to run into the collection tube. In addition, half of the soil pits were instrumented with a tension lysimeter (Prenart®, Frederiksburg, Denmark) installed to the same depth. All instruments were installed one dormant season (six months) prior to sampling, which was sufficient time to settle into the soil profile, given the heavy rainfall and high soil biota activity. Figure 1 shows the locations of the instruments. Throughout this paper, samples collected in zero-tension lysimeters under both saturated and unsaturated conditions are referred to as “leachate” samples. Samples collected from tension lysimeters, that only collect water during saturated or near-saturated conditions in these clay-rich soils, are referred to as “soil water” samples.
 During 4-h irrigation events in 2005 and 2006, leachate was collected from the zero-tension lysimeters and irrigation water was collected directly from the drip emitters. During an 8-h irrigation event on 23 September 2005, leachate was collected in zero-tension lysimeters until the soil reached near saturation, at which point soil water was sampled in tension lysimeters. During storms from 2004 through 2007, rainwater and soil water from tension lysimeters were collected. Measurements of water collected during storms on 28 March 2005 and 5 December 2005 verified that both sets of lysimeters capture the same water chemistries under saturation. Both instruments could not be sampled during unsaturated conditions (i.e., the 4-h irrigation events), because adequate tension to draw samples could not be maintained. Table 1 includes a summary of the instrument inventory and sampling frequencies. All solution samples were filtered in the field using ashed (precombusted at 450°C for 4 h) GFF filters, collected in 60 mL HDPE bottles, brought to the laboratory, and frozen until analysis.
Table 1. Instrument Inventory and Measurement Frequencies
 Precipitation, irrigation, and leachate waters were analyzed for [SO42−] using an ion chromatograph (Dionex Corporation, Sunnyvale, California, USA.). Selected solutions collected from 2005 through 2006 irrigation and storm events were analyzed for δ34S-SO42− (referred to as δ34S). A bulk sample (four events) of precipitation was required to obtain sufficient material for analysis. All solution samples were prepared for isotopic analysis by precipitation of BaSO4 using BaCl2 crystals (procedure modified from Bates et al. ). Precipitates were filtered onto ashed glass fiber filters, washed with deionized water, dried at 60°C overnight, and then weighed in tin boats for analysis. The samples were analyzed on a Micromass Optima isotope ratio mass spectrometer at the USGS facility in Menlo Park, California. The data were corrected to CDT using standard material NBS-127 (at 21.3‰), along with two in-house standards, and are reported here in delta notation (δ34S) in parts per thousand (‰).
3.3. Hydrological Measurements
 Soil water content probes (ECH2O capacitance probes, Decagon, Pullman, Washington, U.S.A.) and tensiometers (Soil Moisture Systems, Tucson, Arizona, U.S.A.) were installed to 0.3 and 0.6 m depths in the clay loam and sandy clay horizons to measure soil moisture status and hydraulic head, respectively. Readings were taken during irrigation and storm events during 2005–2006 (see Table 1). Values are total head (h) (i.e., elevation head (z) plus pressure head (ψ)), relative to a local datum (24.4 m ASL).
 In addition to field readings, soil water retention was determined at the USDA-ARS laboratory in Riverside, California, using pressure plates. Samples from the clay loam and sandy clay horizons were taken at six locations in the study area and analyzed [Klute, 1986] for soil water content at 0.33, 0.5, 1, 5, and 15 bars of tension. Soil water characteristic curves for clay loam and sandy clay were generated using empirical estimates from Rawls et al. , Carsel and Parrish , and soil water content and retention data from the field and laboratory.
3.4. End-Member Mixing Model
 The degree to which water interacts with the soil matrix is reflected in the ultimate [SO42−] and δ34S of leachate (discussed in detail below), offering a means to distinguish between preferential and matrix flow paths. To provide a more rigorous framework to assess the contribution of different source waters, an end-member mixing model was applied. This model considers leachate collected in the zero-tension lysimeters to be a mix of irrigation water bypassing the rooting zone of the vines (preferential flow) and old soil water held in the matrix. Using this idealization of the system, the water mass balance can be written as
where Q refers to the volume of water (L). The subscripts “L,” “M,” and “I,” refer to leachate collected in the zero-tension lysimeters, matrix flow (as determined by soil water sampled by the tension lysimeters), and irrigation water (delivered via preferential flow to the lysimeters), respectively. Multiplication of [SO42−] with each term in (1) creates a mass balance for SO42− delivered to the zero-tension lysimeter on the basis of the matrix and irrigation water end-members as
Similarly, the mass balance for δ34S of SO42− can be expressed as
 Solving (2) for [SO42−]M and combining with (3) yields the following expression to partition water collected as leachate during irrigation events:
where FI is the fraction of irrigation water bypassing the rooting zone as preferential flow (i.e., FI = QI/QL).
4.1. Hydrologic Conditions During the Study Period
 Rainfall and pressure head readings reflect the different hydrologic conditions during the growing and dormant seasons. In water years 2004–2005 and 2005–2006 (Figures 2a and 2b), rainfall was 660 and 785 mm, respectively. Irrigation inputs during the growing season were 105 and 121.3 mm, respectively. During 2006–2007, the driest year of the study, measurements covered the period from 1 October 2006 to 1 March 2007 (Figure 2c). By March 2007, rainfall was 230 mm, which was only 47% of the rainfall at the same time in the previous 2 years. The pressure head readings show that the soil remained at saturation during most of the dormant season and was unsaturated during much of the growing season (Figure 3). Higher variability among sampling locations during the growing season was most likely due to the uneven wetting front of the drip irrigation at each tensiometer nest location.
 Conservative estimates of specific discharge (Darcy velocity) based upon published [Rawls et al., 1981; Carsel and Parrish, 1988] saturated hydraulic conductivity (Ksat) values for clay loam (6.39 × 10−7 m s−1) and sandy clay (3.33 × 10−7 m s−1), indicate that matrix water moves across the study area (SW to NE) in a timeframe of 10–100 years. The slow velocities at this site are largely determined by the shallow slope across the vineyard block and fine-textured soils. Using measured values of pressure head where total head is greater than elevation head, velocities were slightly slower. This discrepancy is most likely due to a higher water table relative to the ground surface, yielding an apparent decrease in the hydraulic gradient. In the vertical dimension, a maximum of 0.29 L of water can be collected in a zero-tension lysimeter during a 4-h irrigation event, assuming saturated conditions and an area of influence equal to the area of soil directly above a lysimeter.
 Both field and laboratory soil water content and retention data fall within the standard deviation of average soil water characteristic curves generated from published values [Rawls et al., 1981; Carsel and Parrish, 1988] (Figures 4a and 4b). The field data suggest that the upper limit of soil water content is less than that predicted by both the clay loam and sandy clay characteristic curves. This is likely due to the soil moisture probes underestimating soil water content under near-saturated and saturated conditions. Therefore, we employ the average soil water characteristic curves for the soil textures at the study area.
4.2. Evidence of Preferential Flow
 Field observations (see Figure 5) of soil cracking at the surface and within the near-surface during the growing season indicated that preferential flow was likely an important mechanism controlling the movement of water and solutes in the vineyard. These observations were substantiated by the delivery and accumulation of leachate in the zero-tension lysimeters from the onset of irrigation, revealing that a significant fraction of water moved at rates greater than those predicted. The presence of leachate in nearly all zero-tension lysimeters suggests that preferential flow paths are ubiquitous at the scale of an individual vine. Hence, distinguishing between locations where preferential flow is and is not present may only be possible with the application of dyes to a defined area [e.g., Jardine et al., 1989; Ghodrati and Jury, 1990; McIntosh et al., 1999], a method that is not practical in a working vineyard.
 Actual leachate volumes measured in zero-tension lysimeters were spatially variable (i.e., from <0.001 to nearly 1 L per event, or 0.006–6% of applied water), and often greater than 0.29 L (the maximum predicted). Temporal variability in leachate volumes collected in each lysimeter throughout the course of the growing season suggests that the locations and sizes of preferential flow paths are dynamic. Changes in flow paths at a particular location are likely caused by repeated passes on foot and tractor by workers, soil drying and cracking between irrigation events, water losses by plant uptake and evapotranspiration, and root growth and decay.
 Field observations and literature values reported for loam-type soils [Peacock et al., 1977; Mmolawa and Or, 2000] were used to estimate the wetted area below a drip emitter. The estimated wetted volume is shaped like a concave-down cone with a radius of 0.5 m. Assuming that irrigation water penetrates to the clay hardpan (0.4 m), the wetted volume below an emitter is 0.11 m3. On the basis of the average leachate volume collected by the zero-tension lysimeters (0.17 ± 0.02 L through 0.012 m3, the volume of soil above a lysimeter), the total leachate below an emitter scales to 1.6 ± 0.2 L or approximately 10% of inputs during 4-h events.
4.3. Leachate and Soil Water [SO42−] During the Growing and Dormant Seasons
 As shown in Figure 6, average [SO42−] in leachate and soil water changed significantly throughout the study period in response to hydrologic conditions. During the growing season, [SO42−] in leachate was elevated moderately above that of irrigation water (7.6 ± 0.9 mg L−1), and was consistent among irrigation events in 2005. During the postharvest irrigation (23 September 2005), [SO42−] was 12.9 ± 1.4 mg L−1 in leachate prior to soil saturation, and 41.0 ± 3.9 mg L−1 in soil water upon saturation. This increase is likely due to mobilization of SO42− stores contained in the soil matrix. In 2006, [SO42−] in leachate exhibited a similar increase following an accidental 19-h irrigation event on 31 July; [SO42−] changed from 26.6 ± 3.2 mg L−1 on 5 July to 46 ± 16 mg L−1 on 28 August. This increase was not statistically significant or uniform across the study area, but likely indicates some mobilization of matrix SO42−. Subsequent measurements of leachate on 5 September and 12 October 2006 were elevated above those prior to the irrigation on 31 July, but declined from those measured on 28 August.
 Dormant season measurements of soil water from October 2005 through April 2006 exhibited a slow decrease or “flushing” effect as [SO42−] declined from approximately 30 to 20 mg L−1. Creed et al.  reported a similar trend in a forested watershed where nitrate stores were seasonally washed from the upper soil horizons when the water table rose to the surface. In contrast, measurements in March and April 2005 and February 2007 had higher and more variable [SO42−]. The differences among sampling points may be due to the amount of water moving through the soil matrix at the times of measurement, or differences in SO42− production.
 Sulfate losses below the rooting zone can be scaled from the vine to the field using the estimate of wetted area below an emitter (0.11 m3), and assuming that zero-tension lysimeters sample a representative leachate volume and chemistry. Figure 7 shows that SO42− fluxes range from 0.06 to 0.23 kg ha−1 per irrigation event in 2005, which is less than the average load in irrigation water (0.53 kg ha−1 per event). In 2006, SO42− losses exceed those of irrigation water inputs, which may be due to activation of the entire soil matrix by the accidental, 19-h irrigation.
4.4. Partitioning Water Sources Using δ34S
 The δ34S of SO42− in water inputs (irrigation water and precipitation), leachate, and soil water exhibit distinct separation (Figure 8). Irrigation and rainfall are not significantly different, exhibiting δ34S values of 5.7 ± 0.3‰ and 5.5‰, respectively. Soil water shows a wide range of δ34S values (8.5 to 17‰) at low [SO42−]. At concentrations > 25 mg L−1, the values converge to an average δ34S value of 13.5 ± 0.1‰. Leachate collected during 4-h irrigation events has distinctly lower δ34S values, with an average of 7.7 ± 0.5‰, similar to the δ34S of irrigation water. In contrast, leachate and soil water collected during transitioning hydrologic conditions (e.g., the 2005 8-h irrigation) span the entire range of concentration and isotopic values expressed by the 4-h irrigation and storm events, with average δ34S values of 9.5 ± 0.9‰, approaching the average δ34S of soil water during complete saturation.
 The 8-h, postharvest irrigation in 2005 provided the opportunity to measure the transition from unsaturated to saturated conditions in the soil profile during initial soil flushing, and to confirm the expected transition in δ34S values from approximately 7.7‰ in leachate to 13.5‰ in mobilized soil water. The most complete data sets from three different sampling locations are depicted in Figure 9. Two of the three locations demonstrate the most common pattern among the sampling locations: low δ34S values during irrigation and high δ34S values upon saturation. The third location, however, was dominated by the high δ34S values of soil water throughout the sampling period. Water did not appear in the zero-tension lysimeter at this location until 2 hours later than the others. This lag may have been caused by the initial movement of water through the soil matrix, which subsequently came into contact with a preferential flow path, accelerating transport.
 An end-member mixing model was applied that used δ34S values and [SO42−] (data shown in Figure 8) to partition the sources of water in leachate samples collected during irrigation events. This approach solves the mass balance equations for [SO42−] and δ34S assuming only two flow paths with distinct isotopic signatures are responsible for delivering all leachate collected in a zero-tension lysimeter. Under the constraints of this model, the δ34S value and [SO42−] of leachate must lie between the prescribed end-member values. The model is sensitive, therefore, to the end-member values selected for δ34S of matrix and preferential flow, as well as the [SO42−] of preferential flow; the system of equations requires only three parameters, specifying [SO42−] found in matrix flow is not necessary.
 The δ34SM end-member was set at the 90th percentile value (i.e., the upper boundary of matrix water δ34S) measured in soil water (14.5‰). This value was selected to permit flow partitioning under the two–flow path simplification over most of the range of values shown on Figure 8, but avoid skewing the results by incorporating significant outliers. The natural choice for preferential flow end-member values would be the characteristics of the irrigation water; however, measurements show that [SO42−] of leachate from 4-h irrigation events increased above that of irrigation water without a commensurate increase in δ34S (see Figure 8). This result indicates that the SO42− concentration of water slightly increased as it moved down cracks to the lysimeter, most likely because of a small addition of SO42− accumulated during evaporation between irrigation events. Since there was not a significant trend in δ34SI over time, the mean value (5.7‰) was applied in the mixing model to represent the isotopic signature of preferential flow. The increase in [SO42−] during preferential flow is included in the mass balance as an offset (e), assumed to carry the isotopic composition of the irrigation water, given by
 The value of e was determined by averaging the difference between [SO42−]L and [SO42−]I for leachate samples with an isotopic composition close to that of irrigation water (taken as samples with δ34SL < 7‰). The [SO42−] in irrigation water changed over the course of the growing season, from about 3.6 mg L−1 in early July to 9.7 mg L−1 in mid-August (data not shown). This trend was included in the calculation of e.
 The percentage of irrigation water contributing to QL, calculated using the parameters described above, is presented in Table 2. The average percentage of water delivered to zero-tension lysimeters as preferential flow is 70 ± 5% of QL.Figure 10 shows how FI changes over the range of [SO42−] and δ34S values displayed in Figure 8.
Table 2. Parameters Used in the Two-End-Member Mixing Model and Resulting Values of the Fraction of Preferential Flow Contributing to Leachate
Volume (L) of leachate sampled in the zero-tension lysimeters during irrigation events.
Sulfate concentration (mg L−1) of leachate.
Isotopic composition of leachate (‰).
Sulfate concentration (mg L−1) of irrigation water, modified by the offset (e).
Fraction of preferential flow in leachate.
The δ34S matrix end-member was 14.5‰ for all samples.
6 Jul 2005
18 Jul 2005
15 Aug 2005
15 Aug 2005
22 Aug 2005
23 Sep 2005
23 Sep 2005
23 Sep 2005
23 Sep 2005
23 Sep 2005
23 Sep 2005
23 Sep 2005
23 Sep 2005
23 Sep 2005
23 Sep 2005
23 Sep 2005
5.1. Hydrological Flow Paths and SO42− Movement in the Near Surface
 Initial characterization of the hydrologic regime and soil physical properties of the near surface provides a framework to investigate the hydrologic flow paths present in vineyard soils. The use of drip irrigation by growers throughout California's dry season (Figures 2a–2c) creates a variably saturated near-surface environment, as reflected in the total head data (Figure 3). Between irrigation events, significant drying occurs, leading to the development of surface cracks, shown in Figure 5. Wetting and drying events throughout the growing season are followed by several months of rain (Figures 2a–2c) during the dormant season, thereby establishing conditions in which biogeochemical cycling and transport of S and other reactive nutrients occur year round. In nonirrigated, seasonally dry ecosystems, such as Northern California oak woodlands and grasslands, microbial transformations of reactive elements normally slow until the first winter rains [Davidson, 1991].
 The calculations of Darcy velocity through the soil matrix provide an upper estimate of soil water travel times within the study area. However, they do not take into account the development of preferential flow paths in the soil profile. As a result, soil texture–based velocity estimates are inherently slower than actual field conditions, especially for the surface and very near surface. The matrix flow estimates may be more appropriate for sandy clay (hardpan) than for clay loam because preferential flow paths develop to a lesser extent in the deeper layer. However, the extremely low permeability of the hardpan limits the flow of water between the two layers and likely induces ponding of water or lateral subsurface stormflow during prolonged hydrological events.
 Combining the physical analysis with the elemental and isotopic data leads to further understanding of the timing and relative magnitude of hydrological flow paths. The pattern of [SO42−] and δ34S in leachate and soil water collected during irrigation and storm events provides snapshots of seasonal SO42− dynamics, reflecting the relative engagement of the soil matrix. The influence of the soil matrix is lowest under 4-h irrigations, characterized by low [SO42−] (Figure 6) and δ34S values similar to those of irrigation water (Figure 8), and most pronounced during 8-h irrigations and storm events following dry periods. Initial flushing of micropores at these times mobilizes high [SO42−] (Figure 6) and δ34S values (Figure 8). Continued flushing of micropores during sustained storms (i.e., 28 December 2005 through 13 January 2006) causes [SO42−] of soil water to decline, despite continued engagement of the matrix.
 The insights gleaned from the qualitative assessment of the relative contributions of preferential and matrix flows are consistent with the results from the end-member mixing model (Figure 10 and Table 2) defined by the end-member values described above. Figure 10 shows the flow partitioning as a function of [SO42−] and δ34S in leachate for the two–flow path idealization of the hydrologic system. This type of analysis provides an index of soil matrix interaction based on leachate chemistry; the degree to which water flowing in the near subsurface interacts with the soil matrix is an important determinant of biogeochemical budgets. This approach suggests that preferential flow is a major factor to consider at the study area, and conceptualizing the system as such creates a starting point for studying the heterogeneity in the cycling of S and other reactive elements.
 While the δ34S of soil water provides the distinct separation from irrigation water that is necessary to partition flow paths, the specific cause of the δ34S values of soil waters remains elusive. The low [SO42−] and variable δ34S may be due to spatial heterogeneity of SO42− consumption due to microbial reduction [e.g., Alewell and Gehre, 1999]. Variability in δ34S may only be generated when the flux of water through the soil is low enough to draw water from the reaction sites (on soil particles). High water fluxes that cause preferential sampling of macropores may create apparent stability of δ34S values. Such scale-dependent isotope effects have been reported for nitrogen isotope studies [Houlton et al., 2006]. Discerning the degree to which S source signatures and microbial transformations determine soil water δ34S merits further study in vineyard soils.
5.2. Different Flow Regimes and Their Controls on the Fates of SO42− in the Near Surface
 The 4- and 8-h irrigations and storm events sampled in this study represent three flow scenarios with different consequences for the fate and transport of SO42− and other solutes in vineyards. Using the insights gleaned from the patterns in the SO42− concentration and δ34S values, as well as the results of the end-member mixing model, we explore the potential consequences for soil and water quality at local and watershed scales. Figure 11, a conceptualization of the flow scenarios in vineyards, is used throughout this section.
 The use of deficit irrigation amounts during regular, 4-h irrigation events creates a situation where most of the applied water and SO42− mobilized in the soil matrix remain on site (Figure 11a). However, preferential flow comprises approximately 70% of the leachate volume collected below the vine rooting zone (Table 2). If this water is not taken up by longer vine roots, it may eventually move into the hardpan. The small fluxes of SO42− transported with this water (e.g., Figure 7) may become bound to clay particles and resistant to further transport. In contrast, during 8-h irrigations and short, intense storm events, water likely saturates above the clay loam–sandy clay permeability contrast and causes some lateral subsurface transport (Figure 11b). With engagement of the matrix, soil-bound SO42− mobilizes and enhanced S cycling may take place on site.
 During sustained storms, such as the period sampled during the 2005–2006 dormant season, the dominant flow path is likely lateral subsurface transport leading to export of SO42− from vineyards, and saturated conditions where the slope is insufficient to cause flow (Figure 11c). Elevated SO42− export has potentially important in situ consequences, including high base cation losses, depletion of soil fertility, and acidification of soils and surface waters [e.g., Adams et al., 1997; Wilson et al., 1999; Likens et al., 2002].
 There may also be important implications for downstream systems. Bates et al.  showed that SO42− export from agricultural lands is an important control on mercury (Hg) methylation and mobilization in sediments in the Florida Everglades. Given the large inventory of elemental Hg in sediments of the San Francisco Bay, surrounding wetlands, and contributing rivers [Hornberger et al., 1999; Marvin-DiPasquale et al., 2003], a relic of gold mining activities in the mid-19th century, SO42− loading from inland sources may be of broad importance to understanding Hg cycling within the Bay region. Such potential links between S and other environmentally important biogeochemical constituents indicate the importance of quantifying SO42− export from vineyards at the watershed scale.
 This research provides a novel approach for evaluating the interactions between hydrologic response and SO42− transport by taking advantage of a “naturally applied” tracer introduced as part of normal management practices in an agricultural system. The combination of using the physical characteristics of the study area, SO42− solution chemistry, and a simple end-member mixing model demonstrate the tight coupling of near-surface hydrologic response and S biogeochemistry in vineyards. Preferential flow with low [SO42−] and δ34S values dominated leachate composition during 4-h irrigation events, whereas matrix soil water with high values was mobilized during 8-h irrigations and storm events. Both the qualitative assessment of the data and the application of a more rigorous mixing model approach suggest that efforts to regulate S outputs from vineyards or to model coupled hydrologic response and solute transport need to incorporate preferential flow.
 With respect to vineyard management, the maximum water loss estimated in this study (approximately 10% of applied water) represents a large fraction of growers' limited water resources. A common practice in grape growing is to apply drip irrigation from emitters that are approximately 0.25 m above the soil surface, thus causing water to take the path of least resistance through cracks. Simple modification of water delivery, such as decreasing the rate of the drip emitters and watering for a longer interval, may cause water to infiltrate directly into the vine rooting zone and enable growers to reduce irrigation amounts. This practice would be viable, as long as the postharvest 8-h irrigation and natural dormant season storms flush out salts that accumulate during the growing season.
 The three flow scenarios represented by the irrigation and storm events evaluated in this study lead to different consequences for water and SO42− budgets on and off site. Four-hour irrigation events create an on-site water loss issue for growers who need to minimize irrigation amounts. Eight-hour irrigation and short, intense storm events cause water and mobilized SO42− to pool at the permeability contrast, potentially creating conditions suitable for SO42− reduction. Long, dormant season storms likely generate lateral subsurface stormflow carrying mobilized SO42−, which may potentially affect biogeochemical cycling of S and other elements in downgradient ecosystems. Future studies should identify the mechanisms controlling S retention in vineyards, quantify the magnitudes of S loads transported to surrounding surface waters and wetlands, and explore the biogeochemical fates of SO42− within these receiving areas. This information is critical for growers who seek to improve their management practices and reduce environmental impacts at the local and regional scales.
 This research was funded by an EPA-STAR fellowship and a Geological Society of America grant, both to E. S. Hinckley, and funds from Stanford University to P. A. Matson. E. S. Hinckley gratefully thanks K. Grace and D. Zygielbaum at Robert Sinskey Vineyards (Napa, California) for site access, K. A. Lohse (University of Arizona, Tucson, Arizona) for helpful advice on study design and instrumentation, M. E. Rollog (USGS, Menlo Park, California) for stable isotope analyses, and P. Shouse (USDA-ARS Laboratory, Riverside, California) for soil-water content and retention analyses. Early drafts of this manuscript were improved with the helpful comments of T. D. Ahrens, B. A. Ebel, M. C. Long, P. A. Matson, and P. M. Vitousek. Three anonymous reviewers improved the quality of the final manuscript.