- Soil Science Society of America
Land use change from natural ecosystems to cropland influences groundwater recharge, including water quantity and quality. Soil core samples (0–11-m depth) from six boreholes beneath irrigated cropland and two boreholes beneath natural ecosystems, in Vertisols, were analyzed for particle size distribution, water content, and water-extractable Cl−. Chloride mass balance and numerical, one-dimensional unsaturated flow and transport modeling were used to assess average and transient recharge fluxes and to test matrix vs. preferential flow hypotheses. Water contents under irrigated cropland were significantly higher than under natural land with similar particle size distributions. Pore-water Cl− concentrations in deep vadose zones (>3 m) under irrigated cropland (900–2000 mg L−1) were similar to recent local groundwater Cl− and significantly lower than pore-water Cl− in deep vadose zones under natural land (3000–6000 mg L−1). Calibrated models' recharge rates through the soil matrix were much higher under irrigated cropland (90–230 mm yr−1) than natural ecosystems (1–3 mm yr−1) and were consistent with groundwater-balance estimates of average recharge (110–160 mm yr−1). In contrast, matrix recharge rates under natural ecosystems were considerably lower than those based groundwater balance (50–80 mm yr−1). While matrix unsaturated flow under irrigated cropland explains vadose zone and groundwater observations, under natural ecosystems preferential flow paths are required to support observations. Plowing and irrigation prevent development of crack networks and promote matrix percolation through the clay, which flushes salts from previously immobile vadose zone pore water. These phenomena may be applicable to similar land use changes in Vertisols globally.
- CMB, chloride mass balance
- ET, evapotranspiration
- LAI, leaf area index
- PET, potential evapotranspiration
An integrated analysis of deep vadose zone samples (∼10 m), unsaturated flow, chloride transport modeling, and historical groundwater data indicates predominantly preferential flow recharge under natural Vertisols and matrix flow under irrigated cropland. Land use change induces salt flushing from previously immobile matrix water.
Large areas worldwide undergo land use change from native vegetation to cropland. Many of these areas overlie aquifers, which are major sources of water in semiarid regions, including Israel (accounting for ∼65% of the total national fresh water consumption). Changing land use from natural lands to cropland in subhumid and semiarid regions is known to increase groundwater recharge. Increased recharge has been attributed to reduced evapotranspiration (ET) caused by shallower rooting depths associated with crops relative to the native vegetation and fallow periods when there is no vegetative cover (Falkenmark and Lannerstad, 2005; Scanlon et al., 2005, 2007a; Radford et al., 2009). Cultivation is often followed by degradation of groundwater quality due to mobilization of salts from the vadose zone, waterlogging, and leaching of fertilizers (Böhlke, 2002; Scanlon et al., 2007a; Khan and Hanjra, 2008; Silburn et al., 2009).
Irrigation may further impact recharge beyond the effects of cultivation alone. Irrigation plays a critical role in agricultural production globally and also in Israel, where ∼50% of cultivated land is irrigated (Israel Central Bureau of Statistics, 2004). The impact of irrigation on water resources depends, in part, on the source of irrigation water. Irrigation with surface water or imported water generally increases groundwater resources by raising water tables through enhanced recharge under irrigated cropland. In contrast, irrigation using local groundwater depletes groundwater resources, as seen in the U.S. High Plains (McGuire, 2009) and in the North China Plain (Liu et al., 2001). Soil profiles under irrigated cropland are often wetter than those under rainfed or natural lands with similar soil texture, as seen in the U.S. High Plains (Scanlon et al., 2005, 2010b; McMahon et al., 2006; Gurdak et al., 2008). Gravity is more effective at moving water downward in wetter unsaturated media; hence, matrix recharge fluxes through wetter profiles are generally higher. Irrigation can also degrade water quality by increasing the chemical load to the soil, depending on the source of irrigation water (fresh water, saline water, wastewater) and the amount of fertilizer applied. In addition, increased recharge under irrigated agroecosystems can mobilize preexisting salt inventories that accumulated during millennia in semiarid soils (Stonestrom et al., 2003; McMahon et al., 2006; Scanlon et al., 2010a).
The mechanism of deep percolation under natural lands in Vertisols (cracking clays) may differ significantly from that through coarser grained soils. The natural vegetation and the fracture network can maintain sufficiently dry conditions in the upper clayey matrix blocks to develop significant negative pressure heads and low hydraulic conductivities that result in low recharge fluxes through the soil matrix. If the clays overlie coarser grained material, deep matrix percolation requires water accumulation at the base of the fine-grained layer to overcome the capillary barrier at the interface between the fine- and coarse-grained materials, which may lead to the development of preferential breakthrough points into the underlying coarser soils (Heilig et al., 2003). Therefore, uniform matrix flow through Vertisols under natural vegetation in semiarid climates is likely to be small (Radford et al., 2009). Recharge through preferential flow paths in cracked clays should be considered (e.g., Kelly and Pomes, 1998; Liu et al., 2004), especially if relatively fresh groundwater is found under these clays (or existed prior to a significant change of land use to cropland).
Land use change from natural vegetation to cropland in Vertisols involves the removal of relatively deep-rooted native vegetation and the destruction of the soil crack network by tillage. In a Mediterranean climate, irrigation of summer (dry season) field crops may maintain sufficient water in the soil for the development of matrix percolation and suppress the development of deep crack networks. Fallow winters (wet season) provide favorable conditions for deep matrix percolation of rainwater under cropland. Therefore, we hypothesized that a significant change in the mechanism of deep unsaturated flow and groundwater recharge should follow the change from natural land to irrigated cropland in the studied Vertisols.
Local estimates of recharge fluxes can be used to compare recharge under different land uses. These recharge estimates are usually based on data obtained from vadose zone samples (e.g., Gvirtzman et al. 1986 [tritium, anions]; Scanlon et al., 2007b [Cl−, matric tension]; Rimon et al., 2007 [water content]). Recharge estimates from groundwater models are usually based on both regional model calibration to measured water levels in the aquifer and local land use and soil properties (e.g., Blandford et al., 2003). These groundwater-balance-based recharge estimates can integrate preferential recharge fluxes at different spatial scales that may not be represented in local vadose zone samples. Preferential flow paths that may be missed by small-diameter boreholes include earthworm channels, root networks, and crack networks as well as anthropogenically induced paths such as leaky irrigation systems (Hendrickx and Flury, 2001; Gurdak et al., 2008; Jarvis 2008). Direct vadose zone data from actual preferential paths are much more difficult to obtain than soil-core matrix data. Therefore, in this study, the evaluation of preferential flow in the deep vadose zone was indirect, i.e., it was based on observation of the soil matrix, which supports significantly lower recharge fluxes than those obtained from larger scale observations (e.g., groundwater data).
Rising water tables and groundwater Cl− concentrations in the southeastern part of the Israeli Coastal Aquifer have been a concern for several decades. Activation of a saline source from the underlying Saqiye Group and the contribution of salts from the eastern margins of the aquifer have been suggested by different studies (Rosenthal et al., 1992; Vengosh and Ben-Zvi, 1994; Gvirtzman, 2002; Avisar et al., 2004). The potential of the major land use change from natural lands to irrigated field crops as a source of both water and salts has not been evaluated using vadose zone data.
The objective of this study was to assess the impacts of land use change from natural land to irrigated cropland (field crops) on water recharging the aquifer through Vertisols that are typical of the southern Israeli Coastal Plain. In particular, the hypothesis of a changing recharge mechanism from preferential flow to matrix flow following the land use change was examined. A variety of approaches were used to achieve this objective including: (i) analysis of deep vadose zone samples for water content and Cl− concentration, (ii) Cl− mass balance models, (iii) calibration of transient unsaturated flow and transport models to measured water content and Cl− concentrations in the vadose zone, and (iv) assessment of vadose zone data and analysis within the context of larger scale groundwater head and Cl− measurements. An investigation of recharge dynamics, including matrix and mobile–immobile flow and transport simulations (van Genuchten and Wierenga, 1976; Šimůnek et al., 2003; Gerke 2006) completed the assessment. Findings from this study should be applicable to irrigated cropland systems in Vertisols generally.
Materials and Methods
The study area is in the southern portion of the Israeli Coastal Aquifer (∼1900 km2 area, Fig. 1 ). This aquifer supplies 20 to 25% of the water in Israel (400–500 × 106 m3 yr−1). The field study site is near Kibbutz Revadim, overlying the southeastern part of the aquifer (Fig. 1). The mean annual rainfall measured at Revadim (1950–2008) is 480 mm (Kibbutz gauge records). Precipitation falls almost entirely in the winter (0–3 mm of summer rain) between October and April, typical of a Mediterranean climate. The mean Cl− concentration in precipitation in this area is 15 mg L−1 (Herut et al., 2000; Gvirtzman, 2002). At present, the water table is about 30 m below the land surface.
The land near Kibbutz Revadim was first cultivated in the 1950s, mostly for rainfed crops (winter wheat [Triticum aestivum L.]). In 1976, a surface-water reservoir was constructed and irrigation began. In 1982, the use of imported treated wastewater for irrigation began and a mixture of treated wastewater (∼2/3) and surface water (∼1/3) has been generally used for field crop irrigation since 1982. The mean Cl− concentration in the irrigation water is 340 mg L−1 (5-yr mean, 2004–2008). Crops grown at the field site in recent years include corn (Zea mays L.), wheat, cotton (Gossypium hirsutum L.), sunflower (Helianthus annuus L.), and chickpea (Cicer arietinum L.). Irrigation rates range from ∼250 mm yr−1 (sunflower and chickpea) to 550 mm yr−1 (cotton and corn). The land is plowed to the 30-cm depth after the cotton or corn seasons and to shallower depths (disk plow) in other cases. Wheat is a rainfed winter crop grown in some years between irrigated summer crops.
Soils covering the Israeli Coastal Aquifer can be generally classified into three types: Grumosol clays (44% of area), Hamra loamy sand to loam (39%), and dune sands along the coast (17%, Fig. 1). Grumosols are brown to gray alluvial soils, classified as Vertisols, specifically Chromic Haploxererts (Soil Survey Staff, 1999). These Vertisols usually contain >35% montmorillonite-rich clay; therefore, they tend to develop desiccation cracks. Grumosols are fertile and used mainly for field crops. Typically, coarser textured loam, sand, and calcareous sandstone underlie the Grumosols over the Israeli Coastal Aquifer. Hamra soils are orange-red Mediterranean soils, dominated by fine sand and containing up to 26% clay (Ravikovitch, 1981).
Agricultural land covers ∼47% of the Israeli Coastal Aquifer outcrop (much of the remaining 53% is residential). The distribution of agricultural land and finer textured soils over the aquifer are correlated (Fig. 1). Approximately 60% of the agricultural land over the aquifer is on Grumosol soils (Fig. 1); therefore, together with general interest in unsaturated flow under natural vegetation vs. irrigated cropland in Vertisols, the evaluation of recharge under these soils is significant for understanding the impacts of agriculture on this aquifer.
Vadose Zone Sampling and Laboratory Analysis
Eight boreholes were drilled to depths of 6 to 11 m in Vertisols, and continuous soil core samples (43-mm diameter) were obtained using a GeoProbe direct push rig (Model 540UD, GeoProbe Systems, Salina, KS). Six boreholes were drilled under irrigated cropland from a 400-m2 area (centered at 31°46′28″ N, 34°49′10″ E) and two under adjacent (∼300 m away) uncultivated natural land (31°46′29″ N, 34°49′22″ E). The natural land was never cultivated, according to the farmer. Gravimetric soil water content (oven dried at 105° for 48 h) and Cl− concentrations of soil water extracts were analyzed at depth intervals of 0.4 to 0.6 m. Water extracts were obtained from a 2:1 water/soil gravimetric mixture that was shaken for 4 h, centrifuged (3000–7000 rpm) until the soil and water were visibly separated, and then filtered. Chloride concentrations were measured using either ion chromatography (Dionex ICS 2100, Dionex Corp., Sunnyvale, CA) or a coulometric-titration chloridometer (Labconco Corp., Kansas City, MO). The Cl− concentration in dry soil (mg kg−1) was calculated by multiplying the extract concentration by the extraction ratio; the Cl− pore-water concentration (mg L−1) is the dry soil concentration divided by the gravimetric water content and multiplied by the water density (1 kg L−1). Soil texture analyses (0.05-mm sieve for sand and hydrometer for clay) were conducted at an average depth interval of 1.2 m according to visible changes in the core for four out of the eight boreholes. Soil bulk density was estimated from the mass and volume of the soil core samples and their water content.
Vadose Zone Modeling
The Cl− mass balance (CMB) and numerical unsaturated flow and transport models were used to obtain average and transient recharge estimates, compare the travel times of Cl− under natural land and irrigated cropland, and simulate the evolution of Cl− profiles since the beginning of irrigation. Numerical simulations were also used to test hypotheses concerning the mechanism of recharge flow (matrix vs. preferential flow) under the natural Vertisols.
Steady-state groundwater recharge due to uniform flow in the vadose zone matrix was estimated using a CMB model (Allison and Hughes, 1983; Scanlon et al., 2007b):where R is the mean annual groundwater recharge flux (rate) [L T−1], P is the mean annual precipitation rate [L T−1], I is the mean annual irrigation application rate [L T−1], Cl is the steady-state approximation of the Cl− concentration [M L−3] with subscripts P and I referring to precipitation and irrigation water, respectively, and ClDUZ is the depth-weighted mean pore-water Cl− concentration [M L−3] in the deep vadose zone (below the root and cracking zone; operationally ≥3 m). Equation  assumes that the steady-state deep percolation rate through the vadose zone below the root and cracking zone is equal to recharge at the water table. Using the same steady-state piston flow assumption, the time required to accumulate Cl− in the vadose zone from the land surface to depth Z (tZ) can be estimated by summing the mass of Cl− to that depth and dividing by the Cl− surface input flux:where θ is volumetric water content [L3 L−3].
Transient recharge fluxes were estimated by calibrating one-dimensional numerical flow models of Richards' equation with root water uptake:where t is time [T], z is the vertical coordinate [L], h is pressure head [L], K is the unsaturated hydraulic conductivity [L T−1], and S is root water uptake [T−1], which is non-zero in a transpiring root zone.
The van Genuchten–Mualem formulation (van Genuchten, 1980; Mualem, 1976) was used for the water retention curves and unsaturated hydraulic conductivity functions of the different soil types (referred to hereafter as the hydraulic functions).
Solving the spatiotemporal fields θ(z,t), h(z,t), and K(h)(z,t) with Eq.  enables calculation of the temporal Darcian flux and pore-water velocity (flux divided by θ). The Darcian flux at the base of the simulated column was used as the estimated transient recharge water flux.
The pore-water velocity (v) links the flow model to the advection–dispersion transport model:where C is the concentration of a nonsorbing and nonreactive solute [M L−3] (Cl− in this study), D is the (longitudinal) hydrodynamic dispersion coefficient [L2 T−1], and v is the pore water velocity [L T−1]. This equation is solved simultaneously with Eq.  to give the spatiotemporal distribution C(z,t), enabling the estimation of transient Cl− profiles.
Equation  describes transport through a uniform porous medium, whereas in some cases the uniform medium approximation is invalid and a mobile–immobile transport scheme is required to match observations (Coats and Smith, 1964; van Genuchten and Wierenga, 1976):where the subscript m represents the mobile portion of the unsaturated medium and the subscript im the immobile portion (i.e., θm is the volume fraction of mobile water on a bulk soil basis), and α is a diffusion-like mass transfer coefficient between the mobile and immobile zones [T−1]. The spatiotemporal distribution of water content and water movement between the mobile and immobile fractions is based on similar treatment of Eq.  (dual-porosity flow model, Šimůnek et al. 2009). Unlike the matrix flow and transport models that were calibrated to field observations, there were no field observations to support a rigorous parameterization of a dual- porosity/mobile–immobile domain model. Following the hypothesis testing using the calibrated matrix flow models, a parameterization based only on a conceptual model of the mobile–immobile domain was used to explore potential preferential flow and transport scenarios.
The HYDRUS-1D finite element code (Šimůnek et al., 2009) was used for the transient flow and transport numerical modeling. The particle size distributions and bulk densities of the soil samples were used to estimate the parameters of the hydraulic functions using the ROSETTA Lite Library (Schaap et al., 2001), incorporated in the HYDRUS-1D system. Atmospheric boundary conditions were used at the top boundary of the unsaturated column, including daily precipitation, potential evaporation, irrigation, and leaf area index (LAI), which controlled root water uptake. The bottom boundary condition was free drainage (the water table is far below the bottom of the simulated column). Observations were used for initial conditions, and long periods were simulated to minimize the impact of the initial conditions (i.e., for estimating transient recharge during an 8-yr period, 32 yr were simulated for irrigated cropland and 64 yr for natural land, discarding the first 24 and 56 yr, respectively). Calibration of the matrix flow model entailed varying the LAI and root-depth distribution. A longitudinal dispersivity of one-tenth of the length of the simulated column was used throughout the transport simulations.
Eight years of automated gauge daily precipitation data from October 2000 to October 2008 and monthly daily-average pan evaporation (proxy for potential ET) values were used for the daily-resolution simulations. The longer period simulations (e.g., 32 and 64 yr) were produced by multiplication of this 8-yr time series. For long-term (250–2000 yr) annual-resolution simulations, we used stochastic realizations based on a lognormal distribution obtained from 58 yr of annual precipitation monitored at Revadim (1951–2008). In simulations of irrigated plots, application rates of 487 mm yr−1 were used for 100 d from the beginning of May, corresponding to corn.
Results and Discussion
Vadose Zone Profiles
Particle size distributions measured in soil samples from four representative boreholes indicated a prevalence of clay-over-sand profiles (Fig. 2a ). In all boreholes, a clay-rich layer (≥35% clay) exists in the upper 2.2- to 4.8-m zone and a sand-rich layer (≥90% sand) below the 6-m depth. Loamy layers 1 to 3 m thick lie between the clay and the sand. Bulk densities ranged from 1.35 g cm−3 in the upper clays to 1.60 g cm−3 in the deeper sands. The five-layer column used for modeling (Fig. 2a) typifies the profile. A road cut 1 km away from the investigated area revealed the profile beneath a sunflower field (Fig. 2a).
Water contents in cores from the six irrigated cropland profiles (blue lines, Fig. 2b) were significantly higher (Table 1 ) than those in cores from the two natural land profiles, both in the upper 2.2 m and below 6 m (Fig. 2b; Table 1). In these depth intervals, the particle size distributions in all cores were similar; therefore, wetter cores under irrigated cropland at these depth intervals were attributed to the different land use.
The Cl− content in the soil profiles under natural land was significantly higher than that under irrigated cropland at depths ≥3 m (the mean was three times higher; Table 1; Fig. 2c). The Cl− mass in the natural profiles accumulated in finer textured layers of the deep vadose zone (3–6 m), where the gravimetric water content was about four times higher than that in the sands deeper than 6 m (Fig. 2a, 2b, and 2c). Pore-water Cl− under the natural land was characterized by a steep increase in Cl− from the surface to ∼3-m depth (∼300–5,000 mg L−1, Fig. 2d) and relatively uniform Cl− throughout the remaining profile (mean = ∼4,700 mg L−1). Therefore, 3 m was chosen as the boundary between the root zone and the deep vadose zone for subsequent interpretations. Pore-water Cl− under the irrigated sites had relatively uniform lower concentrations (mean = ∼1300 mg L−1, Fig. 2d). These significant differences in deep pore-water Cl− between the two land uses (Fig. 2d; Table 1) were attributed to (i) evapotranspirative enrichment due to extensive root water uptake and enhanced evaporation caused by deep soil cracks in the upper clays beneath the natural land, and (ii) much lower root water uptake limited to the upper 1 m and an absence of large cracks under the irrigated land (Fig. 3 ).
Vadose Zone Models
Steady-State Chloride Mass Balance Recharge Estimates
Under the six irrigated locations, borehole depth-weighted (DW) mean pore-water Cl− concentrations from the ≥3-m depth (ClDUZ, Eq. ) range from 930 to 1880 mg L−1 (DW mean = 1280 mg L−1) vs. 4300 and 5040 mg L−1 (DW mean = 4700 mg L−1) under the two natural land locations. Implementing Eq.  and  (where P = 480 mm yr−1, ClP = 15 mg L−1, I = 487 mm yr−1 [irrigated], ClI = 340 mg L−1, and I = 0 [natural land]) resulted in the matrix recharge rates and Cl− travel times presented in Table 2 .
The significant differences between the raw measurements of pore-water Cl− under irrigated cropland vs. those under natural land (Table 1) translate to major differences in matrix CMB recharge fluxes and Cl− travel times (Table 2). While the estimated average recharge flux under irrigated cropland (142 mm yr−1, Table 2) agrees with recharge estimates from previous studies in this aquifer, negligible matrix recharge flux calculated under natural land (1.5 mm yr−1) does not correspond with any previous recharge estimates in this aquifer. For comparison, average recharge fluxes of 150 to 160 mm yr−1 were estimated by Gvirtzman et al. (1986) using vadose zone tritium data from cultivated (occasionally irrigated) Hamra soils in the central part of the aquifer (Fig. 1). An average recharge flux of 150 mm yr−1 for the entire aquifer was deduced from multiyear groundwater balance analysis (i.e., an average of 280 × 106 m3 of water recharging the aquifer annually due to precipitation and irrigation through a surface area of ∼1900 × 106 m2; Hydrological Service, 2007). This average recharge flux was used by Assouline and Shavit (2004) as representative of the aquifer and also by Vengosh and Ben-Zvi (1994) for water and Cl− budget calculations in a Vertisol area located 10 km southwest of the current research area. Shavit and Furman (2001) used a natural recharge coefficient of 0.174 (∼90 mm yr−1 recharge for mean precipitation) in a transient groundwater model of the same area (Fig. 1).
Calibrated Transient Matrix Flow and Transport Models
Transient flow and transport were simulated through the same five-layer soil column under both irrigated and natural conditions (Fig. 2a; Table 3 ). The calibration goal in these models was to match measured and simulated data of two types: (i) total column water storage at the time of sampling (October 2008, ∼2000 mm for irrigated cropland and ∼1400 mm for natural land), and (ii) deep vadose zone Cl− concentrations (October 2008; 1000–1500 mg L−1 for irrigated cropland and 3000–6000 mg L−1 for natural land; Fig. 2d) by changing the parameters affecting ET. Parameters of the hydraulic functions calculated from the measured particle size distribution and bulk density (Table 3) were not changed during the calibration.
Root depth distribution and LAI (controls of root water uptake) were adjusted until matches were obtained between measured and simulated water storage and deep Cl− concentrations. These adjustments were within reasonable boundaries, i.e., high LAI (1–7) for irrigated cropland during the growing season (corn as the main crop) and low LAI (0–1) at other times vs. a uniform LAI (1.2) for the natural land. Under natural conditions, the root-water-uptake sink term was also used to take into account evaporation caused by deep soil cracks. Therefore, low water uptake (simulated by low root density) was enabled to the base of the second clay layer (3.5 m, Fig. 2a) in agreement with the natural profile of Cl− that increased to this depth (Fig. 2d). Maximum root depth was finally set at 0.6 and 3.5 m under the irrigated cropland and natural land uses, respectively. Simulated water fluxes at the base of the column (10.2 m) of the calibrated models were the estimated daily recharge fluxes (Fig. 4 ).
The 8-yr average daily matrix recharge flux from the calibrated model under irrigated cropland (147 mm yr−1; Fig. 4) agrees with the CMB steady-state estimate (142 mm yr−1, Table 2). This similarity is encouraging because of the differences in the details and the quantity and type of data used in both models. The transient matrix recharge fluxes in the calibrated model show a nonlinear correlation between recharge and precipitation (e.g., compare recharge during and after the rainy year 2002–2003 to that in the average year 2005–2006; Fig. 4). Model results also show a lag of 4 to 12 mo between peak precipitation and maximum recharge, which is attributed to moisture accumulation above the sand (capillary barrier) prior to annual peak recharge (Fig. 2a and 4).
Initial calibration of the natural land model to the measured column water storage resulted in practically no matrix recharge (<0.1 mm yr−1). In this situation, Cl− from precipitation accumulates in the soil column and the model is not well calibrated to Cl− measurements. Only when the water storage in the soil column is allowed to increase (by decreasing root water uptake) to 15% higher than the measured water storage does the cumulative Cl− flux at the base of the column equal the cumulative flux of Cl− from precipitation. The average recharge rate at these water contents is 3.5 mm yr−1. We do not know whether the measured profile is close to steady state or not. Therefore, our calibrated model is a compromise between the two criteria (matching water storage and deep vadose zone Cl− concentration). The final model produced an increase of 10% in the column water storage and a slight increase in the average deep concentration (40 mg L−1 increase in each 8-yr period of simulation). Recharge rates in this simulation varied between 1.3 and 2.9 mm yr−1 with an average of 2.2 mm yr−1. This is in agreement with the small recharge estimated by the steady-state CMB method (1.5 mm yr−1, Table 2). The extra 10% of water storage in the calibrated model accumulated in the loamy third and fourth layers above the sand (Fig. 2a); this reflects the effective capillary barrier that exists under the natural land use. Matrix recharge was negligible, even after rainy years (e.g., 140% of multiyear average annual precipitation in 2002–2003); nevertheless, flow simulations of extremely rainy years (e.g., precipitation rates similar to 1991–1992, 220% of the multiyear average) produced significant matrix recharge also under natural land.
The initial Cl− concentration condition, which was used for the transport simulation under irrigated cropland, was similar to the measured Cl− profile for natural land (Fig. 5 ). Hence, the simulation shows the estimated evolution of the profile since the beginning of irrigated cropping (Fig. 5). Under irrigated conditions, the calibrated model shows that after 32 yr of cultivation and irrigation (2008), precultivation Cl− has been completely flushed from the top 10 m of the vadose zone. The calibrated model suggests that in the simulated interval, a steady-state Cl− profile at the current measured concentrations already existed in 2000 (Fig. 5).
Groundwater Levels, Salinization, and Aquifer-Based Recharge
Water-table elevation and Cl− concentration data from nearby wells have been collected by the Israeli Hydrological Service (after 1948) and the Department of Land Settlement and Water Commissioner of the British Government of Palestine (before 1948). These archived data are discussed in light of the vadose zone analysis of this study. Considerable rises in groundwater levels and Cl− concentrations are evident in a large region over the southeastern part of the Israeli Coastal Aquifer and are especially remarkable in the study area near Revadim (Fig. 6 ). Two types of groundwater data and interpretation are of special concern for our analysis: (i) recharge estimates that are based on the water balance in the aquifer; and (ii) groundwater Cl− concentrations prior to extensive cultivation in the research area.
An analysis of historical water table data was performed to quantify the water table rise in the study area. Differences in water table elevation between the beginning of the 1970s and the beginning of the 2000s in monitoring and pumping wells in the Revadim area were mapped (differences between the average head measured in wells from 2000–2005 and those measured from 1970–1975; Fig. 6b). All 16 wells that are perforated in the Coastal Aquifer surrounding the study area showed an increase in water table elevation during these 30 yr (Fig. 6b). The well at Revadim showed the largest water table rise during this period (16.7 m, Fig. 6b and 6c). The spatially averaged increase in water table elevation is 6.8 m (calculated with Thiessen polygons surrounding the wells, Fig. 6b).
For this study, we adopted recharge estimates based on groundwater balance calculations made by the Israeli Hydrological Service. These average annual recharge estimates are based on precipitation and irrigation recharge coefficients (a = R/P and b = R/I, respectively). Coefficients a and b were determined by multivariate regression of groundwater balance equations in which the annual change in water storage in an aquifer cell is equal to recharge due to precipitation, irrigation, artificial recharge, pumping, and flow to and from neighbor aquifer cells (or boundaries). The regression model (including a and b) was determined using 30 yr of water level, precipitation, irrigation, artificial recharge, and pumping data. In the region of the sampling site, these recharge coefficients range from 0.12 to 0.16 for both a and b (Hydrological Service, 2007). For average annual precipitation of 480 mm and annual irrigation of ∼500 mm yr−1, the estimated average annual recharge under the natural land is 50 to 80 mm yr−1, and under the irrigated land it is 110 to 160 mm yr−1. These groundwater-balance recharge estimates integrate all recharge fluxes from the vadose zone (matrix and preferential).
Groundwater Cl− concentrations from a few wells within 2 to 6 km of the sampled profiles prior to extensive cultivation (1930s−1940s) range from 180 to 330 mg L−1 (mean = 220 mg L−1; Department of Land Settlement and Water Commissioner, 1947; Fig. 6d). The Revadim well had a Cl− concentration of 250 to 300 mg L−1 at the beginning of the 1950s (Fig. 6c). Therefore, the mean groundwater Cl− concentration in the area, prior to extensive cultivation, was approximated to be 250 mg L−1.
Synthesis (Evaluation of Vadose Zone Results within the Context of Groundwater Data)
Comparison of the annual matrix-recharge estimates from CMB and a transient calibrated unsaturated flow and transport model (CUFTM) (Fig. 4) with the mean annual total recharge derived from groundwater balance unveils the dynamics of recharge under different land uses (Table 4 ). Whereas the mean annual recharge under irrigated cropland estimated from vadose zone data (CMB and CUFTM) and groundwater balance are similar, large differences are found in the same comparison for natural lands (Table 4).
Concentrations of Cl− in deep vadose zones under irrigated cropland (900–2000 mg L−1, Fig. 2d) correspond with groundwater Cl− concentrations in recent years (Fig. 6a and 6c). In contrast, deep soil water Cl− concentrations under natural land (3000–6000 mg L−1, Fig. 2d) are about 20 times higher than groundwater Cl− values (180–330 mg L−1) prior to extensive cultivation (Fig. 6d).
Recharge Mechanism in Vertisols under Natural Land
Matrix flow simulation of the extreme rainy year (1991–1992) produced significant recharge under uncultivated natural Vertisols; therefore, we investigated the possibility that aquifer recharge under natural lands occurs in response to extreme rainy years. For this purpose, stochastic realizations of 2000 yr of annual precipitation and potential ET (PET) were developed. The annual PET in these realizations was based only on the rainy season (October–April) to allow the possibility of recharge given an annual time step. Computed this way, the multiyear mean annual PET was 610 mm vs. ∼1500 mm. Flow and transport through the soil column under natural land conditions for these 2000 yr were simulated with HYDRUS-1D. Simulated recharge and Cl− profiles were compared with measured data. The mean annual recharge for the 2000-yr model is 4 mm yr−1, higher than the CMB estimate and the 8-yr calibrated model and much lower than the 50 to 80 mm yr−1 estimated from groundwater water balance calculations (Table 4). The Cl− profiles vary from more saline after drought periods to less saline after rainy periods (e.g., the profiles for 1600 and 1200 yr, respectively, in Fig. 7 ); nevertheless all simulated Cl− profiles are less saline than the measured profile under natural lands (Time 0, Fig. 7). Chloride concentrations at the base of the profile vary from 730 to 2800 mg L−1 (mean = 1600 mg L−1) in contrast to ∼250 mg L−1 in the groundwater prior to extensive cultivation (Fig. 6d and 7). This means that a multiyear mean matrix recharge of 4 mm yr−1 is too high to support the measured Cl− profile under the natural land and too low to produce the precultivation groundwater Cl− concentrations. Therefore, we conclude that episodic recharge through the soil matrix in response to extreme precipitation is not the primary recharge mechanism through the Vertisols under natural lands.
Under natural lands a mechanism in which water flows primarily through discrete fast pathways, bypassing most of the vadose zone, is simultaneously consistent with: (i) groundwater balance recharge estimates (Table 4), (ii) vadose zone Cl− concentrations (Fig. 2d), and (iii) groundwater Cl− concentrations prior to extensive cultivation (Fig. 6d). A dual- porosity/mobile–immobile flow and transport model (Šimůnek et al. 2003, 2009), in which the mobile fraction is highly conductive, was therefore developed using reasonable estimations for all parameters. The conceptual model for the conductive mobile fraction consists of vertical fractures in the top clay, some of which connect to underlying loams and sands. Therefore, the mobile zone in the model was characterized with the most conductive porous medium in the top fractured clay (saturated hydraulic conductivity Ks = 820 cm d−1), the least conductive in the intermediate loam (Ks = 330 cm d−1) and Ks = 500 cm d−1 was set at the bottom sands. Transfer coefficients between the mobile and immobile fractions (i.e., α for Cl− in Eq.  and ω for water [Šimůnek et al., 2009]) were set low (2 × 10−5 and 1 × 10−5 yr−1, respectively) to produce a low level of interaction reflecting the large mean distance between the mobile and immobile zones. Stochastic time series of rainfall and PET for 250 yr were used for the flow and transport model through the natural land (i.e., the root water uptake calibrated for the natural land profile). The resulting mean recharge flux and Cl− concentration (Fig. 8a ) as well as the immobile Cl− profile (Fig. 8b) are in reasonable agreement with the groundwater balance recharge estimate, precultivation groundwater Cl− concentration, and measured vadose zone Cl− profile (Table 4; Fig. 2d and 7).
The mobile–immobile flow and transport model (Fig. 8) was not calibrated to vadose zone measurements. Given that the model qualitatively produced vadose zone and groundwater measurements, however, some aspects of the recharge mechanism through preferential paths (with low water storage) can be examined. First, this mechanism enables percolation of fresh water to aquifers underlying surface clays that are exposed to high ET. Second, the relationship between high precipitation events and groundwater salinity dynamics is shown. In thick vadose zones in which matrix flow dominates (high water storage flow path), the volume of fresh percolating water in rainy years is still usually smaller than the volume of water stored in the vadose zone matrix. Therefore rainy years are followed by salinization of the groundwater due to flushing of salts that accumulated in the deep vadose zone during drier periods. This is similar to the mechanism proposed for an arid-region ephemeral lake system in Australia (Costelloe et al., 2009). In percolation through low-storage preferential paths, rainy years can result in fast recharge of fresh water and a reduction in the salinity of the groundwater (Fig. 8a). This study provides an improved understanding of groundwater measurements reported by Goldenberg et al. (1996), who investigated groundwater salinization trends before and after the extreme rainy year of 1991–1992 in the central parts (east–west) of the Israeli Coastal Aquifer (the “storage cells”). Their analysis showed that in the southern part of the aquifer that underlies Vertisols and sand dunes (Fig. 1), the 1991–1992 rainy season reversed the normal salinity trend from increasing to decreasing in most cases, while in cells overlain by Hamra soils (Fig. 1) that do not support preferential flow paths and have high unsaturated water storage relative to sand, a greater than usual increase in groundwater salinity was found after the 1991–1992 rainy season.
Effect on Recharge of Land Use Change
The results and analysis presented above suggest that natural vegetation and near-surface cracked clays result in high ET that precludes significant matrix recharge, yet these conditions enable rapid freshwater recharge through discrete preferential paths. Land use change from natural land to irrigated cropland on these Vertisols reduces ET during the rainy winter season (especially if no winter crop is grown), increasing the water available for recharge. Plowing and generally higher water contents found beneath irrigated cropland destroy the fracture network, precluding flow through preferential paths and recharge of fresh water. Following land use change, salts that accumulated in the vadose zone under natural lands are flushed through the soil profile, markedly increasing groundwater salinity. After the flushing period, a new regime of water and salt fluxes in the deep vadose zone develop, in which Cl− concentrations are relatively steady (Fig. 5). The salinity of the post-flushing recharging water under irrigated field crops depends mostly on the salinity of the irrigation water, while water fluxes are correlated with current and previous year precipitation (Fig. 4).
Summary and Conclusions
Vadose zone soil texture, water content, and Cl− concentrations were measured in soil cores from six boreholes under irrigated cropland and two boreholes under natural lands in Vertisols overlying the Israeli Coastal Aquifer. This region of the aquifer has undergone marked water table rises and increased Cl− concentrations during the last several decades. A clay-over-sand profile in the top 10 m of the vadose zone was found in all boreholes. Water contents in similar soil textures under irrigated cropland were significantly higher than those under natural land, while deep unsaturated Cl− concentrations were lower (900–2000 mg L−1 under irrigated cropland vs. 3000–6000 mg L−1 under natural land). Vadose zone CMB steady-state approximations of matrix recharge agreed with recharge estimates from one-dimensional transient unsaturated flow and transport models calibrated to water content and Cl− measurements, resulting in annual values ranging from 90 to 230 mm yr−1 under irrigated cropland. In contrast, these models estimated negligible matrix recharge (∼2 mm yr−1) under natural land (Table 4; Fig. 4). Groundwater balance mean annual recharge estimates under irrigated cropland (∼500 mm yr−1 irrigation) range from 110 to 160 mm yr−1, in agreement with vadose zone matrix estimates from this study, whereas mean annual recharge estimates under unirrigated land range from 50 to 80 mm yr−1, almost two orders of magnitude higher than vadose zone matrix recharge estimates. Similarly, deep vadose zone pore-water Cl− concentrations under irrigated cropland are consistent with recent local groundwater concentrations (800–2000 mg L−1, Fig. 2 and 6), whereas those under natural lands are more than an order of magnitude higher than local groundwater Cl− concentrations measured prior to extensive cultivation (180–330 mg L−1, Fig. 2d and 7).
While matrix recharge flow under irrigated cropland can explain both vadose zone and groundwater observations, only a dual-porosity medium, in which unsaturated flow occurs in highly conductive, low-storage, preferential paths, can explain the discrepancies between vadose zone and groundwater measurements under natural lands. Dual-porosity/mobile–immobile flow and transport simulations demonstrated a mechanism by which relatively fresh water can coexist in aquifers underlying natural lands in Vertisols that are subjected to high ET. The preferential recharge mechanism in uncultivated Vertisols explains the groundwater salinity response to extremely rainy seasons. Areas in the southern part of the Israeli Coastal Aquifer responded to the rainy 1991–1992 period mostly with a decrease in groundwater salinity (influenced by preferential recharge), whereas the central and northern parts of the aquifer, underlying sandy loam soils, responded with an increase in salinization (matrix recharge).
Land use change from natural land to irrigated cropland in the study area changes the primary recharge mechanism from percolation through low-storage preferential paths to more uniform matrix percolation. Low root water uptake in fallow fields during the rainy winters results in higher recharge fluxes through the matrix that flushes salt accumulations from vadose zones. If irrigation water is not local in origin, as is the case at the present study site, this land use change raises water tables and increases salinity in the underlying aquifer, as is evident in the southeastern parts of Israeli Coastal Aquifer. These processes should be of interest in other parts of the world where Vertisols are converted to cropland, especially under Mediterranean climates.
We thank Yagev Kilman from Kibbutz Revadim for providing rainfall and irrigation-water data and the agricultural history of the site. We also thank the Jackson School of Geosciences, University of Texas at Austin, for financially supporting the drilling and analytical costs for this study.
All rights reserved. No part of this periodical may be reproduced or transmitted in any form or by any means, electronic or mechanical, including photocopying, recording, or any information storage and retrieval system, without permission in writing from the publisher.
- Received August 30, 2010.