Overlying vegetation plays a key role in the hydrodynamics of desert vadose zones. To investigate the roles of vegetation and climate on vadose zone dynamics, eleven 5–10 m boreholes were drilled under ponderosa pine, juniper, grassland, and creosote vegetation communities along a transect ranging in elevation from 1470 to 2380 m in central New Mexico and were analyzed for soil water content and potential and environmental tracers. The results indicate that there has been no downward liquid movement past the root zone under the creosote sites over the past ∼20 kyr. There have been periodic, downward fluxes past the root zone under the grass and juniper sites (<0.4 mm yr−1) as well as preferential flow, but under ponderosa pine, downward fluxes are appreciable (2.3 mm yr−1). Fluxes were similar within vegetation communities, even at sites separated by considerable distance and elevation (and thus climate), but differed markedly over short distances across ecotones, demonstrating that vegetation ecology strongly modulates the influence of climate and that characteristic vadose zone hydrological regimes are associated with the various vegetation communities.
 Ecology has been identified as a likely significant influence on the hydrological characteristics of underlying arid vadose zones. Gee et al.  demonstrated through the use of lysimeters at desert sites that the mere presence or absence of vegetation greatly affects the underlying soil moisture fluxes. Significant water accumulation in the subsurface was observed at all the sites in their study when the vegetation was removed. At two sites with bare soil, subsurface water accumulation and deep drainage accounted for 50% of the annual precipitation. At the third bare soil site, elevated water storage persisted for more than three years even though rainfall during that time was below average. When vegetation (a mix of desert shrubs and grasses) was allowed to grow, water previously held in storage was quickly removed through evapotranspiration and no water penetrated below the root zone. Scanlon et al. [2003, 2005] found through long-term water potential monitoring (5–12 years) at natural sites that the penetration of wetting fronts under desert grass and shrub in response to seasonal fluctuations in precipitation was restricted to the upper 3 m of the profiles.
 A close linkage between vegetation and vadose zone dynamics was supported by Walvoord and Phillips , who found that areas with mixed shrub, creosote and grass vegetation had no water flux below the root zone and actually supported a small net upward flux across the groundwater table. In contrast, they determined that there was appreciable downward flux, about 3 mm yr−1, under juniper sites. Walvoord and Phillips  used water potential and chloride profiles to estimate the vadose zone moisture fluxes and computer modeling to infer the groundwater recharge and equilibrium conditions. The water potential became less negative going from shrubs to grass to juniper trees. Chloride concentrations were greatest under the shrubs sites and decreased going toward grass then juniper. Both the water potential and chloride accumulation indicated that downward liquid fluxes increased proceeding from creosote to grass to juniper. Further, computer modeling supported an absence of water flux below the root zone under the shrub and grass sites. Downward fluxes were simulated only under the juniper sites [Walvoord and Phillips, 2004]. All of the sites in their study, with the exception of the mixed shrub sites, were relatively close together, reducing the possibility that climate was the determining factor for the differences seen in the soil moisture fluxes.
Newman et al.  inferred deep downward fluxes under piñon-juniper woodlands at an elevation of 2140 m in the northern Río Grande Valley. These fluxes ranged from 0.002 to 0.11 mm yr−1 for the deepest sampled sections of the profiles. They found smaller downward fluxes under ponderosa pine, but this finding was attributed to the effects of a thick clay layer in the profile that restricted downward flow, rather than to the differences in vegetation [Newman et al., 1997].
 On the basis of these previous findings, we hypothesized that arid vegetation communities play a significant, and perhaps predominant, role in controlling deep soil moisture fluxes in arid vadose zones. The earlier studies described above were geographically scattered and employed a limited number of sample sites at each locality; they also made little attempt to control factors other than vegetation that might affect vadose zone fluxes. In order to more rigorously investigate this hypothesis, we designed a systematic and controlled study of vegetation communities within a small region, drilling multiple boreholes within each community.
 This study employed a series of replicated and controlled measurements that are more akin to typical scientific methodology in the field of ecology than hydrology, and the second section of this paper (after the introduction) describes the experimental design. The third section deals with the climate, hydrology, vegetation, and physical setting of the study and the fourth the field and laboratory methods. In the fifth, the variability and influence of field conditions that were controlled for (i.e., that were attempted to be maintained as constant as possible) are assessed. In the sixth section the vadose zone profiles are described and interpreted in terms of processes, history, and subroot zone fluxes. In the seventh the motivating hypothesis (that vegetation communities play a significant and perhaps controlling role in deep vadose zone fluxes) is evaluated, and the finding of the study are summarized in the eighth.
1.1. Vegetation and Soil Moisture Feedback
 While the studies cited above seem to indicate that vegetation may play a leading role in determining subsurface moisture fluxes, ecologists commonly posit that soil moisture is the primary factor in determining vegetation spatial patterns [e.g., Dick-Peddie, 1993]. This circularity has been addressed by Rodriguez-Iturbe , who examined the role of hydrologic mechanisms in ecological patterns. He concluded that soil moisture plays a dual role in being both a cause and a consequence of the type of vegetation. The signal characteristic of desert vegetation is adaptation to high water stress, but the dryness of desert soil is more a result of competition for water by desert vegetation than it is of climate alone (as evidenced by the lysimeter studies of Gee et al. ). It is the feedback between the availability of water from precipitation and the water use characteristics of the particular vegetation community that ecologically predominates in an area that determines the moisture state of the soil in the root zone. The hydraulic gradients and fluxes at the base of the root zone, in turn, control the moisture state of the deep vadose zone. In this sense, deserts are as much a creation of desert vegetation as desert vegetation is an adaptation to arid climate. Soil moisture is the key variable linking climate to plants, thus motivating our hypothesis that vegetation communities will be associated with characteristic soil moisture regimes.
 While numerous studies have been performed on soil moisture redistribution and plant water use efficiency at relatively shallow depths (<1 m) in arid and semiarid ecosystems, these do not directly bear on the question of why desert vegetation commonly maintains deep root zones (1–5 m) under nearly constant conditions of very negative water potential [Scanlon et al., 2005]. The issues of plant physiology mechanisms and adaptive significance of this behavior have been discussed by Seyfried et al. . They posit that access to even small amounts of deeper soil water may allow these plants to capture occasional deep wetting events that might otherwise pass below the root zone, and also to photosynthesize for longer periods, especially during seasonal or prolonged drought. To these hypotheses, we add that maintaining quite negative soil water tensions increases the propensity for soil water moving in preferential flow conduits to be drawn out laterally by capillary action into the soil matrix. This may permit plants to harvest additional water that, were the deep soil more moist, might pass through the root zone in the preferential flow conduits and be lost to plant use.
1.2. Surface Indicators of Groundwater Recharge Fluxes
 The principal focus of this study is the soil moisture regime of desert vadose zones (i.e., depth distribution of water potential and directions of fluxes) rather than groundwater recharge, per se. However, our results do have implications for estimating the distribution of diffuse groundwater recharge across the landscape. Within arid drainage basins, groundwater recharge shows a high degree of spatial variability [Hogan et al., 2004]. The distribution of recharge must be quantified in order to understand the groundwater hydrology. One approach to this problem is to estimate site-specific recharge fluxes based on measurements performed on vadose zone boreholes [Scanlon, 2004]. However, the determination of vadose zone fluxes for an entire basin via individual vadose zone profiles is generally prohibitively expensive and time consuming. If a surface feature that correlates to the vadose zone water fluxes could be found, this would substantially reduce the cost and effort. This surface indicator must have a strong linkage with deep vadose zone moisture fluxes and be readily observable on the surface. A corollary of the vegetation hypothesis described above is that the nature of the soil moisture state that characterizes different vegetation communities also plays a strong role in determining groundwater recharge. If so, understanding of the vegetation/recharge linkage might then be utilized to scale vadose zone results up from individual boreholes to groundwater recharge estimates at the basin scale.
2. Experimental Design
2.1. Selection of Study Sites
 The objective of this research was to examine the relative influences of ecology and climate on the hydrodynamics of underlying desert vadose zones. In order to isolate the influences of climate and ecology on soil moisture fluxes, study sites were chosen that had varying climatic and vegetation parameters, while other parameters were held as constant as feasible. Parameters we considered that could affect the vadose zone moisture fluxes, other than ecology and climate, were topography, pedogenic carbonate development, surface soil texture, and heterogeneous subsurface soils.
 The experimental design started with the selection of a transect area along which the climate changed gradually, due to an elevation change from 1470 to 2380 m, and also included all the ecosystems to be tested. After this transect area was chosen, sites within this area were identified based on the following criteria: sandy soils, relatively flat topography and relatively homogenous subsurface soils. In order to have access to the land and the site, additional criteria were considered: that the site was located close to an established road and that the land owners or managers were amenable to drilling on their lands. As a result of these constraints, the ecosystems to be tested were narrowed down to creosote, grass, juniper, and ponderosa pine. These ecosystems are prevalent in the area and constitute 55% of the total vegetation cover of New Mexico [Dick-Peddie, 1993]. The vegetation of the transect area is shown in Figure 1. On the basis of the criteria above, 11 drill sites were selected. Some characteristics of the sites are listed in Table 1 and their locations are shown in Figure 1.
Table 1. Summary of Site Locations and Physical Parametersa
 This study focused on the influences of ecological community on vadose zone hydrological characteristics, not on estimation of groundwater recharge. Natural variation in factors that we controlled to be uniform at our study sites (topography, carbonate development, soil texture, and moisture focusing) undoubtedly strongly affect vadose zone fluxes at sites where these factors vary. In particular, focused recharge under water courses or depressions is certainly very important for regional recharge [Hogan et al., 2004]. These factors deserve additional study, but were outside the scope of our investigation.
2.2. Aridity Index
 Aridity cannot be determined from precipitation alone. Both potential evapotranspiration and precipitation must be accounted for in order to characterize the climate, since they represent the principal source and sink terms for the hydrologic cycle [Thornthwaite, 1948]. The aridity index (AI) quantifies the degree to which a climate is wet or dry and consists of the ratio of potential evaporation to precipitation [Budyko, 1974].
 The aridity index was used in this study to characterize the extent to which climate influences the vadose zone moisture fluxes. Regions where the AI is greater than one are broadly classified as dry since the evaporative demand cannot be met by precipitation; regions with an AI of less than one are classified as wet [Arora, 2002]. Ponce et al.  further subdivided the AI values into the following climatic categories, 0.375 to 0.75 as humid, 0.75 to 2 as subhumid, 2 to 5 as semiarid, and 5 to 12 as arid. The transect area contains aridity values that range from 3 to 7, as shown in Figure 2, falling within the semiarid to arid regions of Ponce's climatic subdivisions. These divisions may correspond to changes in the vadose zone dynamics, since the division between basin floor nonrecharge areas and recharge areas presumably lies somewhere along the arid-semiarid gradient [Walvoord and Phillips, 2004].
3. Physical Setting
 There are three major groundwater basins in the transect area. The San Agustin Basin is a closed basin, and contains sites 10 and 11, with groundwater table depths between 110 and 120 m [Myers et al., 1994]. Sites 4, 5, 7, 8, and 9 are located in the La Jencia Basin, which is a partly closed groundwater basin with depths to the water table at the sites ranging from 50 to 110 m [Anderholm, 1987a]. The Albuquerque-Belen Basin, containing sites 1, 2, 3, and 6, with depths to the water table ranging from 40 to 90 m, is an open basin, which is drained by the Río Grande [Anderholm, 1987b]. The elevation along the transect varies from 1470 to 2380 m.
3.2. Modern Climate and Vegetation
 The eastern, low-elevation end of the transect is located on the Sevilleta National Wildlife Refuge (SNWR, operated by the U.S. National Fish and Wildlife Service, see http://southwest.fws.gov/refuges/newmex/sevilleta/history.html) and two weather stations are operated near our sites within the SNWR. Data from Station Number 44, (elevation 1500 m) located near sites 1 and 6, indicate a mean annual temperature of 15.4°C and annual precipitation of 238 mm (Long Term Ecological Research Station, Sevilleta National Wildlife Refuge, meteorology data, available at http://sevilleta.unm.edu/research/local/climate/meteorology/summaries/monthly/allyears/, hereinafter referred to as LTER, 2004). Located near sites 2 and 3, data from Station Number 45 indicates a mean annual temperature of 14.4°C and annual precipitation of 213 mm (LTER, 2004). The nearest weather station to the drill sites near Magdalena is a National Weather Service weather station (elevation 1550 m), which receives a higher mean annual precipitation of 463 mm and has a lower mean annual temperature of 11.5°C (National Climatic Data Center Station's historical listing for the National Weather Service's Cooperative Network, station network number 295353, available at http://www.wrcc.dri.edu/cgi-bin/cliMAIN.pl?nmmagd).
 The most common vegetation community within the lower elevations of the transect area is creosote. In the creosote ecosystem, Larrea tridentata (creosote bush) dominates strongly. The main grasses of this ecosystem are Pascopyrum smithii (western wheatgrass) and Muhlenbergia porteri (bush muhla), with Gutierre sarothrae (broom snakeweed) being the predominant weed. The ecosystem contains less than 1% Atriplex canescens (four-winged saltbush), Juniperus monosperma (one-seed juniper) and Prosopis glandulosa (honey mesquite).
 Grassland is common at low to intermediate elevations, particularly on relatively flat, sandy areas. The principal grasses in this ecosystem are Bouteloua gracilis (blue grama), Bouteloua eriopoda (black grama), Muhlenbergia porteri, and Hilaria mutica (tobosa). Opuntia imbricata (tree cholla) and Opuntia macrocentra (purple prickly pear) are common cacti. Several types of weeds are present, with Zinnia acerosa (desert zinnia), Heterotheca villosa (hairy golden aster), Salsola iberica (Russian thistle), Yucca baccata (banana yucca), Rhus trilobata (squawbush), and Gutierre sarothrae being common.
 Hillslopes and rocky areas at intermediate elevations are commonly covered by juniper “pygmy forest”. In the juniper ecosystem, the dominant juniper species is Juniperus monosperma, with less than 5% Pinus edulis (piñon pine) and less than 1% Pinus cembroides (Mexican piñon pine) and Juniperus osteosperma (Utah juniper). The prevalent grass is Bouteloua eriopoda. Other important plants in the ecosystem are Rhus trilobata, Yucca baccata, Heterotheca villosa, Gutierre sarothrae, Zinnia acerosa, and Linaria vulgaris (yellow toadflax). At slightly higher elevations, mixed piñon-juniper forest predominates, but this ecosystem was not investigated in our study.
 Mountain slopes and summits are covered by extensive forests of ponderosa pine. The trees in the ponderosa ecosystem consist of approximately 70% Pinus ponderosa (ponderosa pine), 20% juniper, and 10% Pinus edulis. Three types of juniper grow in this vegetation community: Juniperus osteosperma, Juniperus deppeana (alligator juniper), and Juniperus monosperma, with Juniperus monosperma being predominant. The dominant grass is Bouteloua eriopoda, and Gutierre sarothrae and Heterotheca villosa are the dominant weeds. The highest mountain summits are characterized by mixed conifer forest, but we did not include this ecosystem in our transect.
3.3. Relation of Drill Sites to Ecotones
 An ecotone is the portion of the landscape where two ecosystems meet, creating a boundary that contain characteristics of both ecosystems. We particularly attempted to quantify any differences in soil moisture fluxes on either side of ecotones because, inasmuch as the climate is essentially constant across ecotones, any differences are likely to be due to the differing ecosystems. Proximity to an ecotone was a factor in selecting sites along the transect. Two ecotones were of particular interest. The first was the grass and creosote ecotone; this ecotone is diffuse, and the ecology changes from grass to creosote patches several times between drill sites 1 and 6, which are 4 km apart. Grass site 6 is around 120 m from the closest creosote-dominated area, and creosote site 1 is around 330 m from the closest grass-dominated area.
 The second ecotone of particular interest was the grass and juniper ecotone. This is a fairly sharp ecotone and is approximately 340 m from grass site 5 and 1,140 m from juniper site 7. Sites 5 and 7 are separated by approximately 1,480 m. Grass site 4 is also about 300 m from a juniper-dominated area. The piñon-juniper mix ecosystem was not sampled and there was no juniper and ponderosa pine ecotone along the transect.
3.4. Disturbances of Plant Regimes
 At least since the arrival of the Spanish in the late 16th century in what was to become the State of New Mexico, humans have been impacting the vegetation of the state through grazing, recreation, and fire suppression [Scurlock, 1998]. Native Americans likely also influenced the environment, mainly through the setting of fires and engaging in agriculture along the Río Grande, which included water diversion [Scurlock, 1998]. All areas of the transect have been grazed, mainly by cattle and sheep. The SNWR is in the Río Grande Valley, but it has not been grazed since the Nature Conservancy acquired the land in early 1973. Grazing can cause the disappearance of valuable forage grasses, an increase in the amount of shrubs over large areas, and an acceleration of soil erosion. Hernandez et al.  predicted that vegetation recovery would take decades or more. Grazing was not intensive in the forested areas of the mountains until the early 20th century [Scurlock, 1998]. Changes in vegetation as a result of grazing and other pressures on the land may have an effect on the underlying vadose zone moisture regimes.
3.5. Paleoclimate and Paleoecology
 A major change in the climate of a region can affect its water balance and ecology. Paleoclimatic and paleoecological studies attempt to reconstruct such responses to climate change. Paleoenvironmental reconstructions have been performed at paleolake San Agustin which is located near the Magdalena sites. This reconstruction indicated that the climate typical of the full glacial period at the San Agustin site began about 26 ka and ended about 20.6 ka [Phillips et al., 1992]. The end of the glacial period at ∼15 ka marked the transition from a colder and wetter climate to one more similar to today's warm and dry climate [Plummer, 2002]. Another study in the San Agustin plains used pollen, diatoms, ostracodes, and radiocarbon analyses to reconstruct paleoclimate [Markgraf et al., 1984], and came to similar conclusions.
 The SNWR sites are located in the northernmost extent of the Chihuahuan desert. According to packrat midden paleocology reviewed by Van Devender , during the late Wisconsin stage (20–10 ka) a piñon-juniper-oak woodland covered the rocky slopes of the entire elevational gradient of the area now occupied by the Chihuahuan Desert. Xeric desert scrub communities formed in the early Holocene epoch (10 to 8 ka) after piñon and juniper departed [Van Devender, 1990]. However, it must be noted that preservation of packrat middens is biased toward rocky hillslopes and that evidence for the full glacial paleoecology of the flat basin floors is sparse.
4.1. Sample Collection and Field Observations
 A hollow-stem auger drill rig with a split-spoon sampler was utilized to obtain the soil cores at the sites. The drilling took place the week of 15 March 2004. No fluids were used in the drilling and hammer percussion advanced the sampler ahead of the rotating auger to prevent mixing of the soil samples. The split-spoon sampler was 1.52 m long with a 5.05 cm inner diameter and a 7.62 cm outer diameter. Immediately after the sampler was removed, it was opened and the soil was placed into plastic, sealable freezer bags to prevent water loss. At all sites the target depth for drilling was 10 m, but boulders or other impediments stopped drilling at shallower depth at several sites. Deeply weathered bedrock was encountered at ∼3 m at both ponderosa pine sites.
 After drilling, soil pits were dug at one site within each vegetation community: site 1 for creosote, site 5 for grass, site 8 for juniper, and site 11 for ponderosa pine. These pits were dug to provide soil descriptions, determine root density and distribution, as well as to determine if the peds (structured soil aggregates) taken from the soil cores were compacted during hammering of the core barrel. The compaction of the peds from the soil cores was found to be quite minimal [Sandvig, 2005]. The pits were dug to depths between 77 cm and 135 cm. Each soil horizon was described using the terminology developed by the Soil Survey Staff . The root density and distribution were determined through observations on the walls of the soil pits using a 10 by 10 cm square, where the roots inside this square were counted according to size classes. Three replicates were performed per 20 cm depth increment and averaged by size class. The soil profile descriptions and root densities are summarized in Figure 3. A tension infiltrometer of the type described by Ankeny et al.  (manufactured by Soil Measurement Systems) was used to determine the surface saturated hydraulic conductivity of each surface soil type found at the sites. Saturated surface hydraulic conductivity determines the maximum infiltration rate possible into a specific soil type.
4.2. Laboratory Analyses
 Soil samples were leached in deionized water to extract water soluble anions. This was accomplished by leaching approximately 100 g of soil with 100 g of 18MΩ deionized water. The concentrations of chloride, bromide, fluoride, nitrate, nitrite, phosphate, and sulfate were measured by ion chromatography. Soil water potential was measured using a water potential meter [Gee et al., 1992], the WP4 Dewpoint PotentiaMeter, made by Decagon Devices, Inc. For each sampled depth interval, bulk density measurements were taken for two peds, which were then averaged. The bulk density was determined by using the paraffin clod method, as outlined by Singer . The soil water content was determined by weighing the soil sample before and after oven drying at 105°C for at least 24 hours. Samples with obviously high clay contents were dried for several more days to make sure all water was released. Particle size distribution was measured for sections of the soil profile. Continuous sections of similar soil type in the soil profile were determined by appearance and texture. Particle size was determined according to the method of Janitzky . Intervals of similar calcium carbonate content were identified based on the color and bubbling intensity of acid dropped on the core and the calcium carbonate content was then determined using the method of Machette . Water content, carbonate content, and soil texture for all of the cores are presented in Figure 4. The electrical conductivity of the soil leachates was measured using a portable electrical conductivity (EC) probe.
5. Nonecological and Nonclimatic Influences on Soil Moisture Fluxes
 The experimental design was intended to hold as constant as possible all nonclimatic and nonecological influences on the vadose zone moisture fluxes of the transect. This section evaluates the extent to which the sites along the transect meet this criterion.
5.1. Infiltration and Surface Soil Texture
 As can be seen from Figure 4, most of the soil textures are sandy loam, loamy sand and sand. Therefore most of the soil in the profiles conformed to the experimental design criterion of being sandy soils. Loamy soils were found deep in the profiles of sites 3 and 11 and site 5 had sandy clay loam surface soils. The surface soils at site 5 contained 22% clay at the surface to 60 cm depth and 31% clay from 60 to 130 cm. High clay content decreases the infiltration capacity of soil. To determine if the high clay content of the surface soil at site 5 might cause the surface soil to have an anomalously low infiltration rate compared to the other sites, the saturated hydraulic conductivity of the surface soil found at the sites was measured with a disc infiltrometer. Results of the measurements are given in Table 2. All of the sites, except for site 5, had similar saturated hydraulic conductivities. Site 5 exhibited an anomalously low saturated hydraulic conductivity, less than one seventh the next highest value. With the exception of site 5, the surface soils at the other sites were representative of the soils of the surrounding area [Natural Resources Conservation Service, 2005].
Table 2. Infiltration Characteristics of Surface Soils at Representative Drill Sitesa
 Because the saturated hydraulic conductivity value is considered equivalent to the infiltration capacity of a soil [Hillel, 1998], infiltration of precipitation is less at site 5 than the other sites. A low infiltration rate combined with even a slight slope will result in an increase in runoff [Hillel, 1998]. The slope of site 5 is 2°, therefore runoff processes at this site probably dominate over infiltration at this site. The substantially decreased infiltration at this site affects not only the subsurface moisture fluxes but presumably also reduces the chloride input, rendering comparison with the other sites problematical. The difference in infiltration rate was so marked that site 5 was considered anomalous and not used in direct comparisons with the other sites.
 Most of the sites had slopes ranging from 1.0° to 2.5°, which were measured with a hand transit. Only site 10 had a steeper slope of 6.0°. However, this site had the coarsest surface soil texture and does not appear to have produced an anomalous amount of runoff. It was retained in the analysis.
5.3. Plant Roots
 Data on the rooting depths of desert vegetation are sparse. A literature review of the rooting depth of plants in water-limited ecosystems conducted by Schenk and Jackson  demonstrated that trees have an average depth of 5.8 m, shrubs 3 m and grasses 1.1 m. As these data pertain to this study, trees would be considered representative of ponderosa pine and junipers and shrubs representative of creosote. Data from Canadell et al.  agree with these numbers, with nonmesquite shrubs possessing an average rooting depth of 2.5 m and mesquite possessing an average rooting depth of 15 m. Ponderosa pine was cited as having an average rooting depth of 3.5 m [Canadell et al., 1996]. The depths of the soil pits used to determine root distribution were between 77 cm and 130 cm, therefore the maximum rooting depths of the plants could not be determined from our observations. As shown in Figure 3, most of the roots were found in the upper 40 cm of the pits under all vegetation types; this shallow root concentration may serve both for absorption of infiltration from small precipitation events and for plant stabilization. Large roots were found spread laterally under creosote bush at approximately 20 cm depth. The remaining fine roots were distributed almost evenly throughout the soil pit. Roots were present at the bottom of the pits in all vegetation types, which is consistent with the maximum depths in the literature sources cited above, therefore values estimated from those sources (1 m for grassland, 3 m for creosote, 6 m for juniper, and 3.5 m for ponderosa pine) were used in the evaluation of root effects on water distribution in the vadose zone.
5.4. Preferential Pathways
 Preferential flow paths are a very important mechanism for movement of water into and through the vadose zone in many hydrologic environments [McDonnell, 1990; Wilcox et al., 1997]. However, there is little evidence that preferential flow plays a significant role at depths of several meters in deep alluvial soils in arid environments [Phillips, 1994]. This study spans climates ranging from arid to semiarid, and thus preferential flow may become common at some point in this climate sequence. One approach to evaluating the possibility of preferential flow is to jointly examine matric potential profiles and chloride concentration profiles. Vadose zones that have evolved by the hydrological process described by Walvoord et al. [2002a] (which do not include any form of bypass flow) are expected to have moderate concentrations of chloride below the root zone, but quite negative matric potentials, due to downward propagation of negative potentials produced within the root zone. However, in some cases vadose zones are observed that are characterized by both relatively high values of matric potential and low chloride concentrations below the root zone. These may be indicative of wetting by unevaporated infiltration that has bypassed the root zone through preferential flow paths. Figure 5 shows idealized matric potential and chloride concentration profiles that indicate an area of possible preferential flow.
Figure 6 shows the chloride concentration and matric potential profiles for sites 4, 7, 8, and 9. At these sites, low chloride concentrations and corresponding high matric potentials in isolated areas of the soil profiles indicate the possibility of a preferential flow path. Additional evidence of preferential flow, such as macropores or abrupt changes in soil texture or carbonate content (discussed below), lend additional credence to the possibility of preferential flow in the area. Preferential flow also seems likely at the ponderosa pine sites, but recent flushing events have wiped out the possible evidence at these sites. Site 4, a grass site, has two regions of possible preferential flow. These regions are separated by an area of increased chloride and decreased matric potential levels. Sites 7, 8, and 9, which are all juniper sites, show areas in their profiles of possible preferential flow, starting at a depth of 4 m and then continuing to the bottom of the profile.
 Observations on rooting characteristics and previous studies on preferential flow help to explain some of our observations. Ponderosa pine have large and lengthy roots that probably make excellent preferential pathways, especially when decayed or burned. Newman et al.  observed preferential flow through macropores at a semiarid ponderosa hillslope in New Mexico. The pervasiveness of bypass flow in the juniper biome may also be explained by rooting characteristics. For example, a juniper tree was observed in the study area with roots that were exposed down to 3 m in an arroyo cutbank. The tap root and several other large roots of this juniper tree ran the length of the entire exposed 3 m deep section. This indicates that juniper trees have extensive deep rooting structures, though additional juniper root density studies would need to be conducted to confirm this. The juniper roots are larger in diameter than creosote or grass roots and probably extend to depths of ∼6 m. Numerous root holes were found in the soil pit dug at site 8 near a juniper tree, some containing decaying wood and others open [Sandvig, 2005]. The combination of frequent open root channels at depth, generally moister soils than in the lower-elevation biomes, and greater precipitation probably leads to a much greater role for preferential flow in juniper and piñon-juniper than grass and creosote.
 In contrast to the profiles under ponderosa and juniper, grassland and creosote profiles showed limited or no evidence of preferential flow. Grass roots are relatively shallow (∼1 m) and small. Creosote shrubs typically have an extensive network of moderate-diameter (>1 cm) roots, but these are nearly all in the top 50 cm. Creosote roots that extend to depth are generally smaller than 1 mm [Sandvig, 2005]. Work done by Martinez-Mesa and Whitford  on creosote and other desert shrubs demonstrated preferential flow along creosote roots, but only to ∼0.5 m depth The low chloride concentrations we observed in the first meter of the creosote profiles are compatible with preferential flow along creosote roots in the upper 0.5 m, but this flow does not affect the deep vadose zone fluxes, which are the focus of this study. Consistent with our observations, Bhark and Small  employed artificial rainfall experiments on desert grassland and creosote in New Mexico and documented considerable shallow variability in infiltration front depth, but did not observe deep preferential flow after excavation. Additional artificial rainfall experiments, especially in the juniper and piñon biomes in alluvial soils, might yield more definitive conclusions on this subject.
5.5. Pedogenic Carbonate Accumulation
 Calcium carbonate deposition is common in arid soils and may significantly affect water movement. Substantial amounts of calcium carbonate were found in all of the profiles, except sites 10 and 11, as shown in Figure 4. This was expected because sites 10 and 11 were the only sites shown to have significant and frequent downward liquid fluxes. Sites 3, 4, and 5 have carbonate levels that may restrict downward water flow through plugging of pores and cementation [Gile et al., 1981; Machette, 1985]. The location of high matric potential values just above a carbonate-rich layer in the profile of site 3 suggests that lateral flow is occurring above the second carbonate layer. The restriction of downward flow in this profile is further confirmed by the extremely dry soil (<1% gravimetric water content) just under the highly cemented layer. Aside from site 3, carbonate deposition does not appear to affect the matric potential or the chloride concentrations of the soil profiles.
6. Vadose Zone Profile Interpretation
 The most important tools in assessing vadose zone moisture fluxes and their relation to the climate and ecology of the area were the matric potential and chloride accumulation profiles. Nitrate and bromide concentrations also give insight to vadose zone processes.
6.1. Soil Water Potential
6.1.1. Osmotic Potential
 Inasmuch as the presence of ions increases the electrical conductivity of water, the electrical conductivity of the soil leachate can be converted into an approximate estimate of the osmotic potential. The details of this calculation are described by Sandvig . The profiles that had appreciable calculated values for the osmotic potential were sites 1, 2, 3, and 6, which are located in the SNWR, and site 7, which is located near Magdalena. These sites have fairly high levels of calcium carbonate and chloride salts, both of which could contribute to the osmotic potential. Even at these sites, the osmotic potentials were very small compared to the measured water potentials. The osmotic potential values were subtracted from all the water potential values to determine the matric potential.
6.1.2. Matric Potential
 The matric potential profiles for all of the sites are shown in Figure 7. The profiles from the more arid sites conform to the general conceptual model laid out by Walvoord et al. [2002a], in which deep drying fronts propagate below the root zone over long periods of time in response to glacial/interglacial climate change. When the data are grouped by ecosystem, a good correlation is observed between the sites within each ecosystem. The profiles show a systematic trend of increasingly negative average matric potentials going from grass to creosote to juniper to ponderosa ecosystems (Figure 8).
6.2. Environmental Tracers
6.2.1. Subsurface Nitrogen Accumulation
 Desert ecosystems have often been considered to be both water- and nutrient-limited systems [Smith et al., 1997]. However, Walvoord et al.  recently demonstrated that a large reservoir of bioavailable nitrogen has been accumulating throughout the Holocene below the root zone in desert ecosystems within the southwestern United States. Walvoord found 300 to 1000 kg ha−1 of total nitrogen below 1 m depth at sites in the Chihuahuan Desert. The Chihuahuan Desert sites sampled in this study showed lower concentrations of soil nitrogen below 1 m depth, ranging from 0.14 to 306 kg ha−1, with an average of 35 kg ha−1 [Sandvig, 2005]. The total soil nitrogen content for each site was calculated as the sum of nitrate and nitrite. Total nitrogen levels found in the Chihuahuan Desert reported in a study by Jackson et al.  were between 50 and 100 kg ha−1. Most, but not all, of the nitrogen levels found in this study are closer to that range. The nitrogen levels measured under the Chihuahuan Desert by Walvoord et al.  were the lowest of all of the deserts of the western US, so this result is not unexpected. The vegetation in the Chihuahuan Desert may be more effective utilizers of nitrogen than vegetation found in the other Southwestern deserts.
6.2.2. Chloride/Bromide Ratios
 The inferences derived from the Cl− concentration profiles that are described below require that the Cl− originate entirely from atmospheric deposition. Cl−/Br− ratios can be used as a tracer for the source of chloride. The Cl−/Br− ratio of atmospheric deposition usually ranges from 80 to 160 (mass/mass), whereas chloride from sedimentary sources usually has much higher ratios [Davis et al., 1998]. Small standard deviations of the ratio within a profile also support consistency in the source of the chloride. Some profiles contained samples with bromide concentrations below the analytical detection limit. All of the samples from sites 6 and 10 had bromide concentrations below the detection limit. Most of the sites have Cl−/Br− ratios between 80 and 160 and standard deviations less than 10, as shown in Figure 9. However, sites 5 and 8 had ratios less than 80 and site 11 ratios above 160. One relevant consideration is that the bromide concentrations at these sites were very close to the analytical detection limit of 0.1 mg L−1, which may have resulted in analytical bias. Factors in addition to analytical bias that may affect the Cl/Br ratios include bromide adsorption on minerals [Brooks et al., 1998] and organic matter [Gerritse and George, 1988], bromide uptake by vegetation growth and cycling by vegetation decay [Owens et al., 1985; Kung, 1990], and addition of chloride by brines or sedimentary minerals [Davis et al., 1998]. We were not able to identify any characteristics of vegetation or soil that were shared by the profiles with anomalous Cl/Br ratios. These potential mechanisms have been evaluated in detail by Sandvig , who concluded that there is no evidence that introduction of nonatmospheric chloride has affected any of the profiles.
6.2.3. Chloride Deposition
 Wet and dry chloride deposition data were obtained from three precipitation chemistry stations located within the SNWR. These data were extrapolated to the drill sites in this study based on site precipitation. The estimated total (wet and dry) chloride deposition rates ranged from 57 mg m−2 yr−1 at site 6 to 73 mg m−2 yr−1 at site 10. For more information on sampling methods and locations, data extrapolation procedures, and the chloride depositional rates estimated for each site, see Sandvig .
6.2.4. Chloride Profiles
 The chloride concentration profiles for each site are shown in Figure 10. The chloride profiles exhibit broad similarities within each ecosystem, with the exception of site 5, which is considered anomalous. There is a systematic trend between ecosystems (Figure 11), with creosote having the largest chloride accumulation within the profile, followed by grass, then juniper, then ponderosa pine.
6.3. Interpretation of Matric Potential and Chloride Profiles
 The creosote profiles are distinguished by very low matric potentials, −4 MPa or less, that persist to the bottoms of the profiles. These are accompanied by high chloride concentrations in the root zone that also persist to the profile bases. Previous modeling by Walvoord and Phillips  for a similar system indicated that a long period of drying (approximately 16 kyr was found to give a good match to the data) from a hypothesized initial, relatively wet, dilute condition was required to establish matric potential and chloride profiles such as these. The vadose zone under the creosote is apparently in a long-term drying transient, and the effects of this drying have propagated beyond the depth of sampling. There is no evidence of water infiltration below the base of the root zone over the past 10–15 ka, or perhaps longer.
 Compared to the creosote, the grass ecosystem exhibits more negative matric potentials in the root zone but less negative ones at depth, accompanied by significantly lower chloride concentrations. These characteristics could be the result of preferential flow as discussed above. Alternatively, they could result from flushing (or partial flushing) of the profiles during the Holocene, with the steeper curvature of the matric potential profile resulting from a shorter drying period. The shallow rooting depth and small root diameter of grasses do not favor the formation of deep macropores, and hence preferential flow pathways. However, shallow rooting depth would favor occasional penetration of wetting fronts past the base of the root zone. We thus favor the second of the explanations above. If preferential flow is the explanation, the amount must be very small in order to preserve the relatively negative potentials and high chloride concentrations.
 Preferential flow seems to be the most probable explanation for the relatively high matric potentials at depth under the juniper ecosystem and the relatively low chloride concentrations. This inference is strongly supported by the presence of root macropores in the juniper soil pit and the observed rooting structure of the juniper tree.
 The ponderosa profiles do not exhibit the characteristic “bulge” pattern associated with the Phillips  and Walvoord et al. [2002a] conceptual model. Instead, they show gradually increasing chloride and decreasing matric potential until bedrock is reached. This pattern is typical of vadose zones in relatively humid settings where downward percolation to the water table is the dominant hydrologic process. These profiles indicate that the ecohydrological limits of the arid/semiarid vadose regime, dominated by long-term, highly negative potentials at the base of the root zone and consequent upward water movement from the deep vadose zone, has been passed and that the ponderosa forest vadose zones operate in a fashion more similar to humid forests.
6.4. Chloride Mass Balance and Groundwater Recharge
 The groundwater recharge rate has frequently been estimated using the chloride mass balance (CMB) method. This method uses the deposition rate of chloride on the soil surface and the concentration of chloride measured in the soil water in the vadose zone to estimate the rate of groundwater recharge (also referred to as the “residual flux” since downward flux within the vadose zone is not necessarily the same as the flux across the water table) [Allison, 1987; Phillips, 1994]. The value of CCl (average chloride concentration) is best determined by plotting the cumulative chloride content (mass per unit volume of soil) with depth against the cumulative water content (volume of water per unit volume of soil) at the same depth. The data on such graphs usually plot in straight line segments whose slopes correspond to CCl for that depth interval. Figure 12 shows average cumulative chloride plotted against average cumulative water for each ecosystem. The average chloride accumulation time (i.e., total profile chloride inventory divided by total atmospheric chloride deposition rate) and average CMB groundwater recharge rate for each ecosystem is shown next to the corresponding curve.
 Although the CMB method has been commonly used for estimating recharge in desert environments, it can yield misleading results. The conceptual model implicitly assumes net downward infiltration to the water table. However, the data and modeling presented by Walvoord et al. [2002a, 2002b] and Scanlon et al.  have demonstrated that water fluxes have been upward for many millennia in many arid vadose zones. Under this circumstance the CMB method will calculate small, but nonzero, recharge (downward) fluxes when in fact there may be net upward flux from the deeper vadose zone or the water table. (However, we note that the CMB approach will still yield valid values for the length of time that chloride has been accumulating in the vadose zone.) In our study area the CMB method for estimating residual flux below the root zone is applicable at sites 10 and 11, because it is clear that downward fluxes predominate. The validity of employing the CMB method at the grass and juniper sites is uncertain and depends on whether the chloride inventory is in a steady state condition or transient. The grass and juniper sites may possibly be in a steady state or near-steady state condition in which the chloride concentration near the bottom of the profile is in equilibrium with the chloride input and thus represents the actual deep flux, but alternatively the chloride concentrations may be a result of strongly transient processes. Thus the CMB estimates of the residual flux through the soil matrix for the grass and juniper ecosystems should be considered limiting maximum estimates since chloride concentrations at the bases of the profiles may be a product of episodic, transient, introduction of dilute infiltration by preferential flow, rather than any true mass balance of the system. The residual flux for the creosote ecosystem was not calculated since the sites in this ecosystem give strong evidence of sustained upward flow and large chloride accumulation, therefore clearly violating the downward flow assumption.
 The apparent chloride accumulation times can yield insight regarding controls on leaching of chloride from the vadose zone. Phillips  observed that many vadose zone profiles in the southwestern United States contain chloride inventories corresponding to ∼15 kyr of atmospheric deposition. Subsequent studies have shown that this range can vary from ∼10 to ∼25 ka of accumulation, apparently depending mainly on the relative aridity of the study area [Phillips et al., 2004, and references therein]. This time interval corresponds to the end of the last global glacial period and the inception of Holocene aridity. The average chloride accumulation time under the desert scrub (creosote) sites investigated in this study is 22 kyr, which exceeds the time since the end of the last glacial period. This probably indicates that even under glacial climate, the lowlands were sufficiently arid to retain significant chloride in the vadose zone. In contrast, the chloride accumulation times of the grass and juniper ecosystems range from 5 to 10 kyr, postdating the end of glacial climate. This can be attributed to either episodic events that partially leached the profiles, or to preferential flow that removes small amounts of chloride on a regular basis. As discussed above, episodic leaching is probably more important under the grass and preferential flow under the juniper. Preferential flow through isolated sections of the profile would cause the calculations of time since the last complete flushing of the profile to be underestimated. Therefore, in the juniper sites, it has been at least 5 kyr since complete flushing of the profile and at least 10 kyr for the grass profiles.
 The chloride accumulation times of the two ponderosa pine sites are much shorter and differ significantly from each other (∼300 yr and 3.6 kyr). This difference is most likely the result of clay-rich layers at site 11 that are not present at site 10. Both sites are nevertheless considered representative of ponderosa pine areas, since clay-rich layers are common in soils in ponderosa pine areas [Newman et al., 1997]. It seems likely that chloride is removed from ponderosa vadose zones by a combination of preferential flow and fairly regular flushing events.
6.5. Basin-Wide Recharge Estimation
 If the CMB –based recharge rates for each ecosystem shown in Figure 12 are actually representative, it should be possible to estimate the total groundwater recharge for the entire transect area using the distribution of vegetation shown in Figure 1. This methodology will probably yield only a rough approximation of the actual recharge. The largest source of uncertainty is probably site bias: our study sites were selected to be on relatively level areas with deep, sandy soils. Much of the transect area consists of steep slopes with shallow, rocky soils; these probably yield more recharge than our study sites [Heilweil and Solomon, 2004; Wilson and Guan, 2004]. Other sources of uncertainty include low resolution of the vegetation map, the oversimplification of vegetation classification, and the limited number of sites drilled in this study. The piñon-juniper ecosystem was not sampled in this study, so the recharge rate for the juniper ecosystem was used to approximate piñon-juniper. For the transect area, our calculations indicate that groundwater recharge, averaged over the entire area, constitutes 0.1 percent of the precipitation. It is important to emphasize that this groundwater recharge estimate includes only diffuse recharge through the landscape and not focused recharge beneath arroyos and depressions.
7. Ecohydrological Influences on Vadose Zones
 The findings discussed above suggest broad correlations between the hydrological characteristics of vadose zones and the ecosystems that overlie the vadose zones. In this section we summarize our results in order to more easily draw connections with ecohydrological influences.
7.1. Systematic Intercomparison of Profile Characteristics
 Although the intended depth for the boreholes was 10 m, actual depths in some cases were less because of drilling difficulties. In order to compare depth-integrated properties of the boreholes, we extrapolated the chloride accumulation and matric potential data from all cores shorter than 9 m to a length of 9 m. Details of the extrapolation methodology are given by Sandvig . Depth-weighted sums of chloride content and matric potential were then computed to obtain the integrated properties of each profile.
7.2. Ecological and Climatic Influences on Vadose Zone Moisture Fluxes
 If vadose zone moisture fluxes were controlled solely by climate (as quantified by the aridity index in this study) then vadose zone moisture fluxes should also change gradually along the transect as the aridity index changes. This would be expressed as the integrated matric potentials becoming gradually less negative, and chloride accumulation time decreasing, as the aridity decreased. The integrated matric potential and chloride accumulation time for each site was plotted against the aridity index in Figures 13 and 14, respectively. The straight lines on these graphs represents the trends that would be expected if climate determined vadose zone moisture fluxes. A simple linear relation between climate and vadose parameters was used for purposes of illustration. This line was fitted to the data.
Figures 13 and 14 do show a general correlation of increasingly negative integrated matric potential and increasing chloride inventory with increasing aridity. However, the vadose zone properties do not vary smoothly with aridity, but rather in a stepwise fashion, with profiles within each ecosystem characterized by similar matric potentials and chloride inventories, whereas between ecosystems there are abrupt changes over small ranges of aridity. This pattern is especially notable when comparing the grass ecosystem, which has relatively constant matric potential and chloride inventory over a wide range of aridity index, with the adjoining creosote and juniper ecosystems.
 The hydrologic characteristics of arid/semiarid vadose zones can be most compactly evaluated by plotting integrated matric potential against integrated chloride (Figure 15). The position of a profile within this plot is an indication of the vadose zone regime. Profiles falling in the upper right corner of the graph (very negative integrated matric potentials and high chloride accumulations) are dry and static and have been maintained in this condition for a long period. They probably are characterized by upward flow from the deep vadose zone toward the base of the root zone and correspond to a typical arid vadose zone in the conceptual model of Walvoord et al. [2002a]. Conversely, profiles falling in the lower left corner of the graph (high matric potentials and low chloride accumulations) are relatively moist, with dynamic and presumably often downward flow. Such profiles have apparently been flushed regularly, resulting in a clearing of the chloride accumulation. In Figure 15, the sites within each ecosystem plot as distinct clusters, demonstrating systematic differences between these ecosystems. The sites from the creosote ecosystem plot in the persistent and static area of the graph, while the sites from the ponderosa ecosystem plots in the dynamic and downward flux area of the graph. The large difference in the position of these two ecosystems on the graph indicates a marked difference in the water dynamics of the vadose zones that underlie them. The grass and juniper ecosystems plot in between the creosote and ponderosa ecosystems, indicating that the vadose zone behavior at these sites have properties in between the extremes. As discussed above, these profiles may have been fully flushed in the past, but such events must have been highly episodic, on a thousands-of-years timescale. The plotting of the sites from the same ecosystems in distinct clusters demonstrates that there are strong ecological influences on the vadose zone hydrologic regimes.
7.3. Possible Influences of Vegetation Perturbations on Vadose Zone Characteristics
 The modeling studies of Walvoord et al. [2002a, 2002b] and Scanlon et al.  demonstrate that thick desert vadose zones equilibrate with their surface boundary conditions on timescales of greater than 103 years, with longer equilibration times associated with drier vadose zones. Recent fluctuations in ecotones could thus affect inferences based on the current vadose zone characteristics, because the vadose zones are not necessarily in (quasi) equilibrium with the current vegetation. Unfortunately, although evidence for historical vegetation changes in the region is abundant [Dick-Peddie, 1993; Scurlock, 1998], there are no site-specific data upon which to determine possible vegetation perturbations at our drill sites.
 Nevertheless, the likelihood of vegetation changes at our sites can be evaluated based on general patterns of vegetation change and their expected consequences. In the region of our study, the prevalent ecotone shifts over the past ∼150 years have been upward expansion of creosote into desert grassland and expansion of the juniper and piñon-juniper communities, both downward into grassland and upward into ponderosa pine [Dick-Peddie, 1993]. This pattern, first, implies that our grassland sites are probably unaltered; if vegetation replacement occurred, it would be grassland to other communities, not the reverse. However, the general pattern would permit the creosote sites to have recently been converted from grassland. This possibility is strongly contravened by the large chloride inventories of the creosote sites: 20 to 25 ka worth of chloride deposition. These contrast markedly with the 10 to 12 ka worth of chloride in the grassland sites (if the clay-rich site 5 is excluded). Recent conversion to creosote could not add chloride to the profiles. The much more negative matric potential values at depth (averaging approximately −5 MPa under creosote, compared to −2 MPa under grass) confirm the inference that our creosote sites have remained stable for 20 to 30 ka.
 Two of our juniper sites (8 and 9) are close to the middle of the juniper biome. They thus seem unlikely to have been affected by upward or downward migration of the ecotones. The site most likely to have been affected (7, which is close to the grassland ecotone) exhibits both chloride and matric potential profiles quite similar to sites 8 and 9 in the middle of the biome. The juniper sites contain 3–7 ka worth of chloride accumulation and the grass sites 10–12 ka. Although based on this evidence we cannot categorically exclude the possibility that our juniper sites have recently been converted from grass, it seems unlikely.
 The ponderosa pine sites do seem very likely to have been under stable vegetation for hundreds of years. First, there is no record of ponderosa expanding down into juniper, rather the record is the reverse. Second, our sites contained mature ponderosa that are probably close to 100 years old. In summary, we can be reasonably confident in stating that the creosote, grassland, and ponderosa pine sites have been under fairly constant vegetation over periods longer than 1000 years. The case for the juniper is less strong, but nevertheless, stability at sites 8 and 9, at least, seems likely.
 This paper reports the results of a systematic sampling of arid/semiarid vadose zones along a transect of decreasing aridity. The relative influences of climate and vegetation were evaluated by examining whether vadose zone characteristics changed smoothly and in parallel with the climate variation, or whether they exhibited irregularities associated with ecotones. While profiles within an ecosystem showed some variation, we observed marked, systematic differences between the ecological communities. These systematic differences indicate that vegetation ecology strongly modulates climate controls on vadose zone moisture fluxes. The observed vadose zone moisture regimes are the result of complex interactions and feedbacks between soil moisture, vegetation and climate. Consistent with previous work [Phillips, 1994; Scanlon et al., 2003; Walvoord and Phillips, 2004], the data gathered in this study support the scenario that, under creosote, there has been no downward water movement past the root zone since the last glacial termination. There do appear to have been small, episodic, downward fluxes past the root zone under the grass and juniper sites, as well as preferential flow. On the basis of independent observations, under grass ecosystems highly episodic (probably millennial timescale) and incomplete flushing by matrix infiltration below the root zone may be the dominant process, while under juniper relatively frequent preferential flow (probably largely through root holes) may be more important. The chloride mass balance method indicates maximum residual fluxes below the root zone of ∼0.1 mm yr−1 under grassland and ∼0.5 mm yr−1 under juniper. The data from the ponderosa pine sites are variable, but indicate frequent downward liquid fluxes below the root zone of a greater magnitude and frequency than the grass and juniper sites. The chloride mass balance method indicates that fluxes below the root zone are appreciable (∼2.3 mm yr−1) under the ponderosa pine ecosystem.
 Given the relatively large number of ecosystems investigated and small number of boreholes that could be drilled, sampled, and analyzed, our conclusions require further confirmation. Nevertheless, the vadose zone data we have collected strongly suggest that vegetation ecology plays a critical role in the hydrodynamics of arid/semiarid vadose zones, and that characteristic vadose zone hydrologic regimes are associated with the various vegetation communities. The differences between these regimes are dramatic. Over a twofold range in aridity index, the characteristic vadose zone behavior ranges from upward hydraulic gradients and nearly static conditions over periods of many millennia under creosote to dynamic downward water fluxes under ponderosa pine. Understanding the linkages responsible for these variations can advance hydrologic science in several directions. Improved understanding and quantification may lead to improved estimates of groundwater recharge. It can provide a baseline for testing hydroecological models [Rodriguez-Iturbe, 2000]. Inasmuch as soil moisture state provides the most important connection between hydrological and vegetation processes, models that cannot reproduce the observed vadose zone regimes are clearly not adequately simulating the hydroecological processes. Finally, better understanding of these linkages is clearly crucial to the ability to predict arid/semiarid vegetation changes in response to climate changes or other environmental fluctuations.
 This research was funded by the SAHRA (Sustainability of semi-Arid Hydrology and Riparian Areas) Science and Technology Center under the STC Program of the National Science Foundation, agreement EAR-9876800. The authors would like to thank Doug Moore (SNWR LTER) for providing the chloride deposition data, Hongjie Xie for his guidance in producing the GIS maps, Jan M.H. Hendrickx and Enrique R. Vivoni for technical assistance, the U.S. Forest Service and SNWR for permission to drill on their land, and the many graduate student volunteers, especially Setsuko Shindo, Huade Guan, and Alex Rinehart, for their field and lab assistance. We thank Enrique Vivoni, Jan Hendrickx, Kelly Caylor, Bridget Scanlon, and an anonymous reviewer for helpful and constructive commentary on the manuscript.