Nonpoint Source Pollutant Transfer across Deep Vadose ...kearney.ucdavis.edu/NEW MISSION-LIVE... · properties throughout a 16-m-deep, typical alluvial vadose zone consisting of unconsolidated,

2001-2006 Mission Kearney Foundation of Soil Science: Soil Carbon and California's Terrestrial Ecosystems

Final Report: 2006034, 1/1/2008-12/31/2009

1University of California, Davis, Land Air and Water Resources (LAWR)

2Daneil B. Stephens & Associates, Albuquerque, NM

3University of New Mexico

*Principal Investigator

For more information contact Dr. Thomas Harter ([email protected])

Nonpoint Source Pollutant Transfer across Deep Vadose Zones – A Multiscale Investigation to Inform Regulatory Monitoring, Assessment, and Decision Making

Thomas Harter1*, Anthony O’Geen1 and Peter Hernes1, Farag Botros2, Valerie Bullard1, Gary Weissman3, Jan Hopmans1

Research Highlights

• Soils and the sediments in the San Joaquin Valley that comprise the unsaturated zone between the water

table and the land surface are highly heterogeneous. We characterized the geologic and hydraulic

properties throughout a 16-m-deep, typical alluvial vadose zone consisting of unconsolidated, alluvial

deposits typical of the alluvial fans of the eastern San Joaquin Valley, California.

• Statistical analysis of field data confirm that lithofacies and other visual- and texture-based sediment

classifications explain a significant amount of the spatial variability of hydraulic properties within the

unsaturated zone. Geostatistical models can be used to describe hydraulic property variations within

lithofacies.

• A simple mass-balance approach to assessing nitrate leaching to groundwater provided long-term, field

scale average nitrate leaching results comparable to 2-D and 3-D vadose zone numerical stochastic

models that accounted – to various degrees – for the detailed local-scale heterogeneity within the vadose

zone.

• Neither the mass-balance approach nor detailed heterogeneous stochastic modeling with standard flow

and transport models explains the highly heterogeneous nitrate distribution found in detailed field work

nor can these approaches explain the low nitrate mass found within the deep unsaturated zone at the field

site. Research is ongoing to further investigate the role of preferential flow and denitrification in deep

vadose zones.

Objectives and Research Hypotheses

The overall objective of this (ongoing) project is to provide a rigorously upscaled modeling tool for basin-

scale assessment of nonpoint source (NPS) pollutant transport in the heterogeneous, alluvial, and often

deep vadose zones of California agricultural landscapes. Our principal research hypothesis is that flow

and transport in these unsaturated sediment systems is subject to highly non-uniform, localized

preferential flow and transport patterns that lead to accelerated solute transfer across the vadose zone with

potentially limited attenuation not captured by current deterministic or stochastic vadose zone models. To

test our hypothesis, we link core scale vadose zone information from two extensive deep vadose zone

drilling projects (one completed, one ongoing) to the effective vadose zone transport of NPS pollutants at

the field scale (orchard, field, corral, land application unit) and at the farm scale (farm, dairy) by using

geostatistical analysis and by applying a high resolution vadose zone flow and a transport model.

Specifically, our overall research objectives and research hypotheses are outlined here – the 2007-2009

Kearney Project addressed the first two research objectives:

Nonpoint Source Pollutant Transfer across Deep Vadose Zones – A Multiscale Investigation to Inform Regulatory Monitoring, Assessment, and Decision Making—Harter

2

Core Scale Description: Using spatially detailed drilling core log data (Figure 1A), determine

stratigraphy (with an emphasis on pedostratigraphic units in order to improve spatial correlation) and

develop geostatistical models of sedimentologic units comprising the deep vadose zones at various

field sites in the San Joaquin Valley.

o Research Hypothesis: The spatial variability of deep vadose zone sediments and their

associated hydraulic behavior can be captured with a combination of stratigraphic information

(facies description) and geostatistical models (intrafacies variability).

Field Scale Processes: Using this geostatistical and stratigraphic information, develop a high

resolution, three-dimensional field-scale flow and transport model of deep, heterogeneous unsaturated

zones to simulate salt, nitrogen, and carbon transport to groundwater (Figure 1B). Use the model to

define depth-dependent, effective travel time and attenuation of salt, nitrogen, and carbon transfer.

o Research Hypothesis: Under irrigated conditions, internal heterogeneity within

sedimentologic facies and non-uniform boundaries between these facies lead to preferential flow

patterns in the deep unsaturated zone thus greatly accelerating the transfer of nonpoint source

pollutants to the water table and providing significantly less attenuation than an ideal

homogeneous vadose zone.

Landscape Scale Decomposition (ongoing): Within the landscape/basin of the San Joaquin Valley

define and describe typical sedimentologic and soil stratigraphic patterns in the subsurface that are

representative of dominant surficial processes of the Valley and that can be depicted from a GIS, such

as proximity to basin alluvium, alluvial fans, and dissected fan remnants (Figures 1C, 1D, 1E).

o Research Hypothesis: A relatively small number (on the order of ten) of representative

stratigraphic-geostatistical scenarios defines the vast majority of deep vadose zones occurring

within the project area. For each scenario, a depth- and pollutant dependent transfer time and

attenuation factor can be determined using the high-resolution model.

Application (future work): Net field scale transfer of solutes will need to be fitted to a simple

transfer function model. We will identify key representative scenarios of deep vadose zone

stratigraphy in the San Joaquin Valley using sedimentologic and pedostratigaphic models in a GIS.

For each of these representative scenarios, we use the high resolution model to define the effective

field scale pollutant transfer to the water table for application across a basin-scale project area. We

then apply the effective field-scale transfer time and attenuation results to the landscape/basin and

provide a basin-wide, landscape scale assessment of NPS pollutant transport across the vadose zone

to groundwater. Two regional project areas were selected with a high density of dairies and intensive

agricultural production but with contrasting vadose zone properties: Tulare/Kings County with deep,

often loamy-textured vadose zones and Merced/Stanislaus County with shallow, predominantly sandy

vadose zones.

o Research Hypothesis: At the landscape/basin scale, the deep (> 10 m) vadose zones play a

significant role in the attenuation and travel time of a nonpoint source pollutants between the

source (at the land surface) and the point of groundwater extraction (monitoring well, domestic

well, public water supply well, irrigation well).

Background

Heterogeneity of Unsaturated Zone Processes: It has been recognized for some time that soils and the

sediments and rocks that comprise the unsaturated zone between the water table and the land surface are

inherently heterogeneous (c.f. Zhang, 2002; Harter and Hopmans, 2004). Numerous field studies have

been implemented to characterize spatial variability of soil moisture, soil water tension, soil water

hydraulic characteristics, and solute transport within the first 2 m below the land surface (Zhang, 2002;

Onsoy et al., 2005). But few of these field studies characterize spatial variability of unsaturated zone


14


3

properties below the root zone. We recently completed an extensive characterization and geostatistical

analysis of the geology, hydraulic properties, and nitrogen distribution in a 16 m deep vadose zone across

a nectarine orchard in Fresno County, California (Figures 1A and 1B; Onsoy et al., 2005; Harter et al.,

2005). We found that the deep vadose zone was characterized by several non-uniformly thick

stratigraphic facies that could be readily observed on continuous drilling cores. Significant variability of

hydraulic properties was observed between these explicitly defined facies, but also within individual

facies (Denton, 2004; Onsoy, 2005). The deep vadose zone was further characterized by very large

(several orders of magnitude) variability of nitrate-N concentrations. Most importantly, the total amount

of nitrogen mass contained within the deep vadose zone was smaller than would be expected based on the

annual leaching rate obtained from a field nitrogen mass balance. Denitrification in the deep vadose zone

was shown to be limited at this site, thus it cannot account for the relatively low nitrate levels observed at

the site. Our findings instead suggest that, given the highly variable soil texture, soil hydraulic properties

and nitrate concentrations observed at the site, preferential flow paths may lead to rapid, highly localized

nitrate transport to the water table leaving behind significantly less nitrogen mass than under uniform

flow conditions. The significant degree of stratigraphic layering enhances lateral flow and nitrate

exchange among adjacent coring locations. Water content, water fluxes, and solute fluxes are shown to be

highly transient throughout the vadose zone and even at the water table, under the irrigated, semi-arid

conditions typical for agricultural regions in semi-arid climates.

Upscaling Methods: Field measurements are typically obtained at a small core-scale and the physical flow

and transport equations have been defined at the laboratory bench scale. Translation of this information

into field-scale or watershed-scale effective representations of unsaturated flow and transport processes in

heterogeneous porous media is commonly referred to as “upscaling”. Stochastic methods in particular

have evolved as perhaps the most important tools among upscaling approaches (Dagan, 1989; Zhang,

2002). We have recently completed several extensive reviews of upscaling methods and stochastic theory

for unsaturated flow (Harter and Hopmans, 2004; Vereecken et al., 2006) and reactive solute transport

(Harter, 2002). This project is a significant link between our work on development and application of

stochastic methods and our farm and watershed/basin-scale systems work on tracking and assessing

nonpoint source pollution (Figure 2). As outlined in the above reviews, current upscaling methods are

limited to assessing flow and transport in relatively homogeneous unsaturated zones. They are most often

based on very restrictive assumptions about the system boundary conditions that drive flow and transport

process in the unsaturated zone and driven by root zone information. Furthermore, few attempts have

been made to apply such approaches above the field scale to farm or watershed scale applications with

deep (> 10 m) vadose zones.

Instead, large-scale applications most often depend on the a priori assumption that Richards equation for

unsaturated flow and the advection-dispersion equation for transport are valid at the large scale. These

models often use a mix of measured data and inverse modeling (parameter estimation) with “measured”

parameters obtained by linearly averaging core-scale measurements. These average parameter values are

assumed to represent large scales horizontally (on the order of 103 m – 104

m) and often the entire vadose

zone vertically (100 m – 102

m). The most commonly applied nonpoint source assessment tools are

typically not even based on a physical representation of the unsaturated zone: GIS-based map-and-overlay

methods for the vulnerability assessment of shallow groundwater are primarily driven by vadose zone

properties, which are represented in qualitative measures such as “low/intermediate/high soil

permeability” or “shallow/deep depth to groundwater.” Another class of tools applied to nonpoint source

pollutants are one-dimensional flow and transport models that often include extensive representation of

root-zone processes but are limited to effectively one-dimensional, homogeneous tipping-bucket or

Richards equation advection-dispersion equation based representations of the deep vadose zone (e.g.,

Johansen et al., 1984; Leonard et al., 1987; Wagenet and Hutson, 1987; Carsel et al., 1998). These have

been extensively used for assessing risks from potential leaching of agrochemicals to groundwater (e.g.,


4

Smith et al., 1991; Laroche et al., 1996; Close et al., 1999; Malone et al., 1999; Dust et al., 2000;

Rekolainen et al., 2000; Trevisan et al., 2000). For many applications, these simplified models may

perform adequately, but a thorough evaluation of these approaches to salt, nitrate, and carbon transport in

California’s deep and heterogeneous alluvial unsaturated zones is lacking. To evaluate the extensive

dataset we obtained from our orchard site, we recently completed the first phase of a multi-step modeling

approach including a fully heterogeneous, transient flow and transport model of the entire 16m deep

vadose zone at our Fresno County orchard site (Onsoy, 2005). We began our work by providing a

thorough mass-balance of the system over the past fifteen years (Onsoy et al., 2005). Based on the mass-

balance approach, a one-dimensional, homogeneous two-dimensional, and heterogeneous two-

dimensional analysis was implemented (Onsoy, 2005). However, Oliveira et al. (2006) demonstrated in

principle that the scaling approach we used in our work may lead to significant underestimation of

unsaturated flow and transport variability.

Our work expands upon these previous efforts in the following aspects:

use of fully three-dimensional instead of two-dimensional simulation domains;

use of a larger simulation domain size;

use of a fully non-linear, multi-parameter heterogeneity instead of linearized randomization of

spatial heterogeneity in the soil physical processes using the scaling factor approach; and

use of a high-resolution transport model specifically designed to handle large contrasts in

hydraulic properties without introduction of numerical dispersion.

Future work will apply this approach not only to the orchard site, but to other field sites and a variety of

scenarios representative for the broader range of unsaturated zone stratigraphy on alluvial fan systems in

California, and include salt and carbon transport in addition to nitrogen transport.

We are currently working on a regional scale assessment of groundwater quality impacts from nonpoint

source pollution (Figure 2). Also, as part of a CALFED-funded, so-called “Dairy Groundwater Project”

we drilled seventeen monitoring wells at three dairy facilities in Tulare and Kings County. Wells were

drilled to approximately 7 – 20 m below the water table, which was located at depths of 30 m to 40 m.

Complete, relatively undisturbed cores were recovered from the boreholes. Detailed sedimentologic

descriptions of the unsaturated zone stratigraphy were obtained from these cores and have been processed

in core logs. Core samples have been analyzed for pH, moisture, and major water quality parameters

including salinity, nitrogen, and carbon. Also as part of this project, we are developing a regional

nonpoint source pollution assessment modeling tool that is capable of tracking nonpoint source pollutants

from the water table to the groundwater extraction point, but currently does not include a vadose zone

component. This project will eventually provide the basis for including an upscaled vadose zone

component into the regional assessment model as part of a future project.

Rational and Significance

Nonpoint source (NPS) pollution of groundwater, particularly from pesticide and fertilizer use and from

salt mobilization has been a long-standing issue of California agriculture. Over the past decade, integrated

pest management, pesticide groundwater protection zones, advances in irrigation technology, adoption of

efficient irrigation systems, and better nutrient management in many cropping systems have provided

significant improvements over past practices with potentially important positive consequences for

groundwater quality. Yet, increased agricultural production, particularly in animal and animal feed

production (e.g., dairies) are putting continued pressure on the quality of California’s groundwater

resources.

Increasingly, state regulatory efforts focus on groundwater quality impacts from nonpoint or diffuse

sources. Implementation of the U.S. Clean Water Act, Section 303(d), has recently resulted in the

regulatory control of irrigated agriculture discharges to surface water and future renewals to the waste


5

discharge requirements (WDRs) for irrigated agriculture are expected to include regulation of discharges

to groundwater. Already, WDRs for dairies and other animal feeding operations (AFOs) require

substantial soil and groundwater monitoring not only in the animal production area, but also in the

irrigated crop production area receiving animal manure.

Groundwater degradation due to salt and nitrate loading is a dominant concern in the Central Valley, in

the Salinas Valley and other coastal valley regions. Pesticide leaching is currently controlled through the

Department of Pesticide Regulations, which maintains an extensive pesticide use monitoring program and

closely oversees management practices. Most recently, emerging contaminants associated with biosolids

from municipal wastewater treatment plants and emerging contaminants associated with animal manure

production and land application have also become potential regulatory targets.

A key obstacle to monitoring and assessing NPS emissions to groundwater is the often significant

thickness of the unsaturated zone. In many agricultural regions of Central and Southern California,

groundwater is found only at depth of more than 10 m – 20 m, and not infrequently at depths of over 50

m. At the field and landscape scale, the presence of such thick vadose zones raises a number of important

questions for proper NPS management and monitoring:

• How does the water quality obtained in soil water samples (within the top 2 m below the ground

surface) relate to the water quality of groundwater recharge at the water table? Can soil samples

be interpreted (and how) to provide a measure of recharge water quality?

• How relevant is soil survey information relative to the hydraulic and transport properties of the

sediment material between the root zone and the water table?

• How much attenuation of nitrate and organic contaminants (including emerging contaminants)

does the vadose zone provide?

• How does the heterogeneity of the alluvial sediments comprising the unsaturated zone affect

water flow and solute transport?

• What is the travel time of salts, nutrients, and organic contaminants to the water table relative to

the groundwater travel time from the water table to a groundwater monitoring or production well?

This project provided the conceptual basis for further developing tools that help agriculture,

environmental groups, planning and regulatory agencies at the local, state, and federal level to address

these questions through a rigorous multiscale investigation.

Approach

Outline: We completed a publication for Task 1 and Task 2. The Task 1 publication (core scale analysis)

was published in 2009 (see appendix 1). The Task 2 publication (field scale modeling) is in preparation

for submission to the Vadose Zone Journal (see appendix 2). For completeness, we also include a short

description of future tasks that are closely related to this project. Task 3 (landscape decomposition) and

Task 4 (application) will be part of future work to be implemented under separate funding. The project is

part of a more comprehensive effort to include vadose zone processes into regional scale nonpoint source

impact assessment of groundwater resources (Figure 2).

Research Task 1: Core Scale Data Analysis

• Task Summary:

o Analyze vertical distribution of sediment texture, sediment color, sediment facies distribution,

and related soil hydraulic properties in deep vadose zone soil cores;

o generate geostatistical models;

• Task Research Hypothesis: The spatial variability of deep vadose zone stratigraphy and its associated

hydraulic behavior can be captured with a combination of stratigraphic and geostatistical information.


6

• Ongoing Work: detailed core scale analysis of 17 deep vadose cores obtained in various locations on the

alluvial fans of the Kings, Kaweah and Tule Rivers.

• Key Results:

o We characterized the geologic and hydraulic properties throughout a 16-m-deep, alluvial

vadose zone consisting of unconsolidated, alluvial deposits typical of the alluvial fans of the

eastern San Joaquin Valley, California.

o The textural groups at the site range in grain size from clay to pebble and cover a wide

spectrum of silty to sandy sediments.

o The thickness of the beds varies from <5 cm for some clayey and silty floodplain material to

>2.5 m for large sandy deposits associated with buried former stream channels.

o Eight major geologic units (lithofacies) have been identified at the site.

o Multivariate analysis of variance and post-hoc testing show that lithofacies and other visual

and texture-based sediment classifications explain a significant amount of the spatial variability

of hydraulic properties within the unsaturated zone.

o Geostatistical analysis of hydraulic properties showed spatial continuity of within-lithofacies

variability in the horizontal direction in the range of 5 to 8 m, which is approximately an order of

magnitude larger than spatial continuity in the vertical direction. A low nugget/sill ratio is

obtained in the horizontal direction, indicating that 1- to 10-m sampling intervals are adequate for

detection of spatial structure in that direction.

o The existence of thin clay or silt layers within lithofacies units results in only moderate spatial

continuity in the vertical direction, suggesting inadequate sampling frequency for variogram

development in that direction.

(see Appendix 1 for complete Task 1 report)

Research Task 2: High Resolution, Stochastic Three-Dimensional Field Scale Modeling

• Task Summary:

o Develop a high-resolution, three-dimensional flow and transport model

o Investigate preferential flow conditions (when does it occur, when not?)

o Investigate value of root zone soil monitoring data relative to recharge water quality

o For each scenario from Tasks 1 & 3: use a 3D model to generate field scale breakthrough

• Task Research Hypothesis: Under irrigated conditions, internal heterogeneity within sedimentologic

facies and non-uniform boundaries between these facies lead to preferential flow patterns in the deep

unsaturated zone thus greatly accelerating the transfer of nonpoint source pollutants to the water table and

providing significantly less attenuation than an ideal homogeneous vadose zone.

• Key Results:

o Simple mass balance calculations were performed

o Six conceptually different 2-D and 3-D vadose zone numerical models were implemented

using varying degrees of hierarchical details of heterogeneity

o All methods resulted in a narrow range of estimated stored nitrate which was found to be

approximately four times larger than what was measured in the field

o This work raises concern about the applicability of Richards equation in deep unsaturated

zones under conditions of infiltration where gravity/pressure gradient dominates convective flux.

(see Appendix 2 to complete Task 2 report)


7

Research Task 3 (Ongoing): Soil Survey Analysis and Decomposition of Landscape/Watershed/Basin

• Task Summary:

o Determine a few representative deep vadose zone classes that can be used to decompose the

landscape/basin into major sedimentologic elements (e.g., proximal fan, distal fan, interfan,

incised valley fill, etc.) based on soil survey and borehole log analyses

o define key simulation scenarios of representative deep vadose zone types

o associate each soil survey unit with deep vadose zone type

o generate separate geostatistical models for each vadose zone type

• Task Research Hypothesis: A relatively small number (on the order of ten) of representative

stratigraphic-geostatistical scenarios defines the vast majority of deep vadose zones occurring within the

project area. For each scenario, a depth- and pollutant dependent transfer time and attenuation factor can

be found, or alternatively, a depth- and pollutant-dependent transfer function.

Task 1 defines site-specific vadose zone scenarios. The objective of task 3 is to provide a set of important

representative vadose zone scenarios that can be used to describe entire landscape/basins (Figure 1E).

These scenarios may include those obtained from the specific sites analyzed as part of Task 1.

Soils (the shallow-most part of the vadose zone) vary systematically across the landscape in response to

dominant near-surface processes. We will correlate the nature and properties of sediment and buried

paleosols in the vadose zone with contemporary soils at the surface in order to scale up the high density

core network to a basin scale. The spatial patterns of variability of contemporary soils are documented by

soil surveys. Many of these soils behave similarly in terms of their ability to accommodate groundwater

recharge and nonpoint source pollutant transport to groundwater. Information from digital soil survey

databases will be mined in order to repackage soils into aggregated soilscapes with similar hydrologic

function. Point measurements at deep cores will then be upscaled to area averages according to surface

soil counterparts (Heuvelink and Pebesma, 1999). Similarly, individual soils can be grouped based on the

nature of sedimentologic processes.

Geologically, the vadose zone and shallow groundwater aquifers of much of the San Joaquin Valley and

other agriculturally productive basins in California consist of alluvial sediments deposited as a series of

large alluvial fans and interfans along the mountain front and often reaching to and across the valley

troughs (Planert and Williams, 1995, Weissmann et al., 2005). Although other geologic features (e.g.,

lakebed deposits, basin-fill deposits) are important elements as well, we here focus our analysis on

alluvial fan unsaturated sediment systems (Figure 1E). Zhang (2005), following Belitz and Phillips

(1995), divided the alluvial sediments in the western San Joaquin Valley into three geostatistically distinct

categories: proximal (upper) alluvial fan, distal fan, and interfan deposits. Based on geologic

interpretation and information from well-logs, these three categories showed distinctly different patterns

of sediment distribution and of the volume proportions of finer- and coarser-textured materials. For

example, proximal alluvial fan deposits showed significant proportion of streambed-associate gravel- and

sand-facies, while interfan regions were composed predominantly of fine-textured flood-plain and mud

deposits. Weissman and Fogg (1999) and Weissmann et al. (1999) provide a detailed analysis of the

Kings River alluvial fan system, which is representative of other eastern San Joaquin Valley alluvial fan

systems. The shallow-most (top 100 m) of these sediment are comprised of a several depositional

sequences separated vertically by distinctly mature, laterally extensive paleosols. These paleosols

represent relatively long periods marked by the absence of depositional activity, which allowed for

extensive soil development. Weissmann and Fogg (1999) distinguished between three categories

(“systems tract type”) of alluvial sediments, each with its own geostatistical representation: open fan

deposits, incised valley fill deposits, and older Pliocene deposits. Paleosols provide distinct boundaries

between multiple sequences of open fan deposits, and between Pliocene deposits and overlying open fan

deposits (Figure 1C).


8

Dominant processes that form contemporary surface soils are linked to the character of buried sediment

and paleosols in the vadose zone (Figure 1D, Weissmann et al., 1999; O’Geen et al., 2003). We will

further quantify these factors by analysis of soils and deep cores that occur on dominant landforms and

that reflect potential differences in sedimentation such as basin alluvium, east-side alluvial fans, west-side

alluvial fans, the age/activity of fans, glaciation within the upper drainage basin of the fan source, and

basin subsidence rate (Weissmann et al., 2005). Based on the conceptual approach outlined in Figure 1,

we will construct a GIS database that extends into the sediment source areas (Coastal and Sierra Nevada

Mountains) to further characterize the sedimentologic environment. Factors such as contributing area,

lithology, slope, hydrography will be processed in a GIS in order to characterize alluvial fans in the

Valley. Heterogeneity of fans and associated sediment will be assessed in two ways: 1) the characteristics

of the source environment as discussed above, and 2) the degree of dissection/incision by ephemeral and

perennial streams. Digital aerial photography and digitized historic topographic maps will be use to assess

the degree of variability that may not be observable from the soil survey and land leveling.

As part of this analysis, we completed a map of the Central Valley showing the spatial distribution of

alluvial units within the Central Valley, based on a thorough review and spatial analysis of soil surveys in

the Central Valley (Appendix 3). This map provides a basis for classifying vadose zones in the Central

Valley. Future analysis will provide an extended landscape/basin scale classification of the unsaturated

zone sediments into systems tract types, as originally proposed by Weissmann and Fogg (1999),

Weissmann et al. (1999), and Weissmann et al. (2005). For each systems tract types, a sequence

stratigraphic and geostatistical model will be defined that may then be used as part of the scenario

modeling in Task 2.

Research Task 4 (Future Funding): Application to Landscape/Basin Scale

• Task Summary:

o integrate water, nutrient, carbon transfer functions into soil survey for project area

o incorporate vadose zone transfer time into regional dairy/NPS model

• Task Research Hypothesis: At the landscape/basin scale, deep vadose zones play a significant role in the

attenuation and travel time of nonpoint source pollutants between the source (at the land surface) and the

point of groundwater extraction (monitoring well, domestic well, public water supply well, irrigation

well).


9


10

List of Publications from this Project

Botros, F. E., T. Harter, Y. S. Onsoy, A. Tuli, J. W. Hopmans, 2009. Spatial variability of

hydraulic properties and sediment characteristics in a deep alluvial unsaturated zone. Vadose

Zone Journal 8:276–289 doi:10.2136/vzj2008.0087 (free public access)

Botros, F. E., T. Harter, Y. S. Onsoy, T. Ginn, J. W. Hopmans, 2010. Richards equation-based

modeling to estimate flow and nitrate transport in a deep alluvial vadose zone. In preparation for

submission to: Vadose Zone Journal.

List of Presentations from this Project

(ORAL) Harter, T., 2007. Patterns, connectivity, and effective properties in

heterogeneous/composite media, Complexity in the Oil Industry Workshop, Nathal, Brazil,

August 2008.

(ORAL) Harter, T., S. Onsoy, J. Hopmans, T. Ginn, 2007. Modeling nitrate transport in deep

alluvial vadose zones below an irrigated orchard. 2007 International Annual Meetings,

ASACSSA-SSSA, New Orleans, Louisiana, 4-8 November 2007.

(INVITED), Harter, T., 2007. Sustainability of groundwater: Understanding nonpoint source

contamination”, Stanford University, December 5, 2007.

(ORAL) Harter, T., S. Onsoy, M. Denton, F. Botros, J. Hopmans, 2008. Scaling factor analysis

in a hierarchical alluvial fan system. USDA-CSREES Regional Project W-1188 Meeting, Las

Vegas, January 2007.

(ORAL) Botros, F., S. Onsoy, and T. Harter, 2008. Long-term nitrate leaching in a deep alluvial

vadose zone: Flow and transport modeling, USDA-CSREES Regional Project W-1188 Meeting,

Las Vegas, January 2008.

(ORAL) Botros, F. and T. Harter, 2008. Modeling flow and nitrate transport in a deep alluvial

vadose zone: Different approaches for characterizing subsurface heterogeneity, Computational

Methods in Water Resources XVII International Conference, 6-10 July 2008.

(ORAL) Botros, F., and T. Harter, 2008. Modeling flow and nitrate transport in a deep alluvial

vadose zone: different approaches for characterizing subsurface heterogeneity. 2008 Joint

Annual Meeting GSA, SSSA, ASA, CSSA, Houston, TX, 5-9 October 2008.

(POSTER) Harter, T. and F. Botros, 2008. Effect of hierarchical, multi-scale heterogeneity on

long-term nitrate transport in a deep vadose zone, American Geophysical Union Fall Meeting,

San Francisco, 15 Dec 2008, Eos Trans. AGU 89(53), Fall Meet. Suppl., Abstract H13F-995.

(Oral) Harter, T., 2009. Effective properties in composite media, USDA Workgroup W1188

Annual Meeting, Tucson, AZ, Jan 4-7, 2009.

References

Ashby, S. F., and R. D. Falgout, 1996. A parallel multigrid preconditioned conjugate gradient algorithm

for groundwater flow simulations, Nucl. Sci. Eng., 124(1), pp. 145-159.

Bennett, G.L, Weissmann, G.S., Baker, G.S., and Hyndman, D.W., 2006, Regional-scale assessment of a

sequence-bounding paleosol on fluvial fans using ground-penetrating radar, eastern San Joaquin

Valley, California, Geological Society of America Bulletin, v 118, no. 5/6, p. 724-732, doi:

10.1130/B25774.1.


11

Belitz, K. and S. P. Phillips, 1995. Alternative to agricultural drains in California’s San Joaquin Valley:

Results of a regional-scale hydrogeologic approach, Water Resources Research, 31(8), 1845-

1862.Carsel, R.F., J.C. Imhoff, P.R. Hummel, J.M. Cheplick, and A.S. Jr. Donigian, 1998. PRZM-3: A

model for predicting pesticide and nitrogen fate in the crop root and unsaturated soil zones: Users

Manual for Release 3.0.

Chomycia, J.C., P.J. Hernes, T. Harter, and B.A. Bergamaschi, 2008. Land management impacts on

dairy-derived dissolved organic carbon in ground water. J. Env. Qual. 37(2), 333-343.

Close, M. E., J. P. C. Watt, and K. W. Vincent, 1999. Simulation of picloram, atrazine, and simazine

transport through two New Zealand soils using LEACHM. Aust. J. Soil Res., 37: 53-74.

Cortis, A., T. Harter, L. Hou, E. R. Atwill, A. I. Packman, P. G. Green, 2006. Transport of

Cryptosporidium parvum in porous media: Long-term elution experiments and CTRW filtration

modeling. Water Resour. Res. (in press).

Dagan, G., 1989, Flow and transport in porous formations, New York, New York, 465 pp.

Denton, M., 2004. Hydraulic characterization of a heterogeneous, deep vadose zone beneath an orchard

within the Kings River Alluvial Fan. M.S. Thesis, University of California, Davis. 2004.

Deutsch, C. V. and A. G. Journel, 1998. GSLIB Geostatistical Software Library and User’s Guide,

Oxford University Press, New York, NY. 369 pp.

Dust, M., N. Baran, G. Errera, J. L. Hutson, C. Mouvet, H. Schafer, H. Vereecken, A. Walker, 2000.

Simulation of water and solute transport in field soils with the LEACHP model. Agric. Water Manag.,

44:225-245.

Harter T., 2002. Stochastic analysis of reactive transport in heterogeneous porous media, in: Govindaraju,

R. S. (ed.), Stochastic Methods in Subsurface Contaminant Hydrology, American Society of Civil

Engineers, pp. 89-167.

Harter, T., 2003. Long-Term Risk of Groundwater and Drinking Water Degradation from Dairies and

Other Nonpoint Sources in the San Joaquin Valley, State Water Resources Control Board Contract #

04-184-555-0.

Harter, T. and J. W. Hopmans, 2004. Role of Vadose Zone Flow Processes in Regional Scale Hydrology:

Review, Opportunities and Challenges. In: Feddes, R.A., G.H. de Rooij and J.C. van Dam,

Unsaturated Zone Modeling: Progress, Applications, and Challenges, (Kluwer, 2004), p. 179-208.

Harter, T., Y. S. Onsoy, K. Heeren, M. Denton, G. Weissmann, J. W. Hopmans, W. R. Horwath, 2005.

Deep vadose zone hydrology demonstrates fate of nitrate in eastern San Joaquin Valley, California

Agriculture 59(2):124-132.

Heuvelink G. B. M. and E. J. Pebesma, 1999. Spatial aggregation and soil process modeling. Geoderma,

89:47 - 65.

Johansen N. B., J. C. Imhoff, J. L. Jr. Kittle, and A. S. Donigian, 1984. Hydrological Simulation Program-

Fortran (HPSF): User’s Manual for Release 8. EPA-600/3-84-066, US Environmental Protection

Agency, GA.

Jones J. E., and C. S. Woodward, 2001. Newton–Krylov-multigrid solvers for large-scale, highly

heterogeneous, variably saturated flow problems. Adv. Water Resour. 24, pp. 763–774.

Kollet, S. J., and R. M. Maxwell, 2006. Integrated surface-groundwater flow modeling: A free-surface

overland flow boundary condition in a parallel groundwater flow model, Adv. Water Resour. 29, pp.

945-958.

LaBolle, E. M., 2000. RWhet, Random Walk Particle Model of Simulating Transport in Heterogeneous

Permeable Media, Version 2.0, User’s manual and Program Documentation, Department of Land, Air

and Water Resources, University of California, Davis.

Laroche, A. M., J. Gallichand, R. Lagace, and A. Pesant, 1996. Simulating atrazine transport With HSPF

in an agricultural watershed. J. Environ. Eng. ASCE, 122: 622-630.

Malone, R. W., R. C. Warner, S. R. Workman, and M. E. Byers, 1999. Modeling surface and subsurface

pesticide transport under three field conditions Using PRZM-3 and GLEAMS. Trans. ASAE, 42:

1275-1287.


12

Minasny, B., J. W. Hopmans, T. Harter, S. O. Eching, A. Tuli, M. A. Denton, 2004. Neural networks

prediction of soil hydraulic functions for alluvial soils using multistep outflow data, Soil Science Soc.

of Am. Journal 68:417-429.

O’Geen, A.T., P.A McDaniel, J. Boll and C.K. Keller, 2005. Paleosols as deep regolith: Implications for

recharge in a Palouse climosequence. Geoderma 126:85-99.

Oliveira L. I., A. H. Demond, L. M. Abriola, P. Goovaerts, 2006. Simulation of solute transport in a

heterogeneous vadose zone describing the hydraulic properties using a multistep stochastic approach,

Water Resour. Res. 42 (5): Art. No. W05420, doi:10.1029/2005WR004580.

Onsoy, Y. S., 2005. Modeling Nitrate Transport in Deep Unsaturated Alluvial Sediments and Assessing

Impact of Agricultural Management Practices on Groundwater Quality. Ph.D. Dissertation, University

of California, Davis.

Onsoy, Y. S., T. Harter, T. R. Ginn, W. R. Horwath, 2005. Spatial variability and transport of nitrate in a

deep alluvial vadose zone. Vadose Zone J. 4:41-55.

Planert, M., and J. S. Williams, 1995. Ground Water Atlas of the United States: Segment 1 - California,

Nevada. U.S. Geological Survey Publication HA 730-B, Reston, Virginia.

Rekolainen, S., V. Gouy, R. Francaviglia, O. -M. Eklo, I. Barlund, 2000. Simulation of soil water,

bromide and pesticide behavior in soil with the GLEAMS model. Agric. Water Manag., 44: 201-226

Smith, S. J., and D. K. Cassel, 1991. Estimate nitrate leaching in soil materials. In R.F. Follett, D.R.

Keeney, and R.M. Cruse (eds.) Managing Nitrogen for Groundwater Quality and Farm Profitability.

Soil Sci. Soc. Am. Madison, Wisconsin.

Trevisan, M., G. Errera, G. Goerlitz, B. Remy, and P. Sweeney, 2000. Modeling ethoprophos and

bentazone fate in a sandy humic soil with primary pesticide fate model PRZM-2. Agric. Water

Manag., 44: 317-335.

Van Genuchten M. Th., 1980. A closed-form equation for predicting the hydraulic conductivity of

unsaturated soils. Soil Sci. Soc. Am. J., 44. pp. 892–898.

Vereecken, H., R. Kasteel, J. Vanderborght, and T. Harter, 2006. Upscaling hydraulic properties and soil

water flow processes in heterogeneous soils: a review. Vadose Zone Hydrology J. (submitted).

Wagenet, R. J., and J. L. Hutson, 1989. LEACHM: A process-based Model for Water and Solute

Movements, Transformations, Plant Uptake and Chemical Reactions in the Unsaturated Zone, Version

2.0., Vol. 2., New York State Water Resources Institute, Cornell University, Ithaca, NY.

Watanabe, N., T. Harter, and B. A. Bergamaschi, 2008. Environmental occurrence and shallow

groundwater detection of the antibiotic Monensin from dairy farms. J. Environ. Qual. 37:S-78–S-85

(2008). doi:10.2134/jeq2007.0371.

Weissmann, G. S., and G. E. Fogg, 1999. Multi-scale alluvial fan heterogeneity modeled with transition

probability geostatistics in a sequence stratigraphic framework. J. of Hydrol. 226, pp. 48-65.

Weissmann, G.S., S.F. Carle, G.E. Fogg, 1999. Three-dimensional hydrofacies modeling based on soil

surveys and transition probability geostatistics, Water Resour. Res., 35(6), 1761-1770.

Weissmann, G.S., Bennett, G.L., and Lansdale, A.L., 2005, Factors controlling sequence development on

Quaternary fluvial fans, San Joaquin Basin, California, U.S.A., in Harvey, a., Mather, A., and Stokes,

M., Alluvial Fans: Geomorphology, Sedimentology, Dynamics, Geological Society of London Special

Publication 251, p. 169-186.

Zhang, D., 2002. Stochastic Methods for Flow in Porous Media: Coping with Uncertainties. Academic

Press, San Diego, CA. 350 pp.

Zhang, H, 2005. Role of Heterogeneity In Flow and Solute Transport – Case Study In the Western San

Joaquin Valley. Thesis, Master of Science, University of California, Davis. 147 p.


13

Appendix 1: Task 1 – Core Scale Analysis

Published as:

Botros, F. E., T. Harter, Y. S. Onsoy, A. Tuli, J. W. Hopmans, 2009. Spatial variability of

hydraulic properties and sediment characteristics in a deep alluvial unsaturated zone. Vadose

Zone Journal 8:276–289 doi:10.2136/vzj2008.0087 (free public access)

Appendix 2: Task 2 – Field Scale Modeling

To be submitted:

Botros, F. E., T. Harter, Y. S. Onsoy, T. Ginn, J. W. Hopmans, 2010. Richards equation-based

modeling to estimate flow and nitrate transport in a deep alluvial vadose zone. In preparation for

submission to: Vadose Zone Journal.

This research was funded by the Kearney Foundation of Soil Science: Soil Carbon and California's Terrestrial Ecosystems, 2001-2006 Mission (http://kearney.ucdavis.edu). The Kearney Foundation is an endowed research program created to encourage and support research in the fields of soil, plant nutrition, and water science within the Division of Agriculture and Natural Resources of the University of California.

http://kearney.ucdavis.edu/

www.vadosezonejournal.org · Vol. 8, No. 2, May 2009 276

I in soil proper-

ties is the fi rst step in assessing vadose zone fl ow dynamics and

in predicting the fate of solute transport in soils. Variability in soil

properties is a critical element across wide areas of research includ-

ing the improvement of agricultural practices, environmental

protection in agricultural areas (Robert et al., 1996), environmen-

tal protection at potential waste discharge sites, land–atmosphere

interactions, and global climate change (Green et al., 2007). It

has been shown, theoretically and in fi eld experiments, that the

spatial variability of soil properties can signifi cantly impact the

amount of solute leaching in soils and that solute concentrations

may vary signifi cantly across short distances as a result of soil

heterogeneity (e.g., Lund et al., 1974; El-Kadi, 1987; Harter and

Yeh, 1996; Russo et al., 1997; Desbarats, 1998; Minasny et al.,

1999; Bagarello et al., 2000; Coutadeur et al., 2002). Th is may

lead to large amounts of solute being leached quickly in some

portions of the soil profi le, while others retain the solute for very

long periods of time.

Studies have recently focused on quantitatively assessing the

variability of soil physical properties between, within, and across

morphologically defi ned soil series taxonomic units (Makkawi,

2004; Iqbal et al., 2005; Herbst et al., 2006). Duff era et al.

(2007) conducted two mixed-model analyses and principal

component analysis to describe the fi eld-scale horizontal and

vertical spatial variability of soil physical properties and their

relations to soil map units in typical southeastern U.S. Coastal

Plain soils. Th eir results indicated that some of the soil physical

properties such as soil texture, soil water content (θ), and plant-

available water showed signifi cant horizontal spatial structure

and were captured by soil map units. Other variables such as

bulk density (ρb), total porosity (φ), and saturated hydraulic

conductivity (Ks) did not show much spatial correlation in the

fi eld and were unrelated to soil map units. Iqbal et al. (2005)

used geostatistical analysis and constructed semivariogram func-

tions of soil physical properties (e.g., ρb, Ks, and θ). Th ey used

the structured semivariogram functions in generating fi ne-scale

kriged contour maps and indicated that a sample spacing of 400

m provided an adequate measure to defi ne the spatial structure

of soil texture and a 100-m sampling range was adequate for

soil hydraulic properties and bulk density.

Most of these studies have focused on the variability in the

root zone horizons, which constitutes only the upper 1 to 2 m

of the soil. In many agricultural areas, particularly in arid and

semiarid regions, groundwater levels may reach up to 30 m deep

or more. Few studies have surveyed soil properties to such depths

Spa al Variability of Hydraulic Proper es and Sediment Characteris cs in a Deep Alluvial Unsaturated ZoneFarag E. Botros, Thomas Harter,* Yuksel S. Onsoy, Atac Tuli, and Jan W. Hopmans

Land, Air, and Water Resources, Univ. of California, Davis, CA 95616. F.E. Botros now at Daniel B. Stephens & Associates, Inc., Albuquerque, NM 87109 and also at Irriga on and Hydraulics Dep., Faculty of Engineering, Cairo Univ., Orman, Giza 12613, Egypt; Y.S. Onsoy, now at Kennedy/Jenks Consultants, San Francisco, CA 94107. Received 23 Apr. 2008. *Corre-sponding author ([email protected]).

Vadose Zone J. 8:276–289doi:10.2136/vzj2008.0087Freely available online through the author-supported open access op on.

© Soil Science Society of America677 S. Segoe Rd. Madison, WI 53711 USA.All rights reserved. No part of this periodical may be reproduced or transmi ed in any form or by any means, electronic or mechanical, including photocopying, recording, or any informa on storage and retrieval system, without permission in wri ng from the publisher.

A : HSD, honestly signifi cant diff erence; MANOVA, multivariate analysis of variance.

R

A

Sta s cal analysis and interpreta on of heterogeneous sediment hydraulic proper es is important to produce reliable forecasts of water and solute transport dynamics in the unsaturated zone. Most fi eld characteriza ons to date have focused on the shallow 2-m root zone. We characterized the geologic and hydraulic proper es of a 16-m-deep, allu-vial vadose zone consis ng of unconsolidated sediments typical of the alluvial fans of the eastern San Joaquin Valley, California. The thickness of individual beds varies from <5 cm for some clayey and silty fl oodplain material to >2.5 m for large sandy deposits associated with buried stream channels. Eight major geologic units (lithofacies) have been iden -fi ed at the site. Unsaturated hydraulic proper es were obtained from mul step ou low experiments on nearly 100 sediment cores. Mul variate analysis of variance and post hoc tes ng show that lithofacies and other visual- and tex-ture-based sediment classifi ca ons explain a signifi cant amount of the spa al variability of hydraulic proper es within the unsaturated zone. Geosta s cal analysis of hydraulic parameters show spa al con nuity of within-lithofacies vari-ability in the horizontal direc on in the range of 5 to 8 m, which is approximately an order of magnitude larger than spa al con nuity in the ver cal direc on. Low nugget/sill ra os suggest that 1- to 10-m sampling intervals are adequate for detec on of horizontal spa al structure. The existence of thin clay or silt layers within lithofacies units results in only moderate spa al con nuity in the ver cal direc on, however, sugges ng inadequate sampling frequency for hydraulic parameter variogram development in that direc on.

Harter et al., 2010, Final Report 16 Univ. of California Kearney Foundation


or monitored solute leaching to a deep water table (e.g., Onsoy et

al., 2005; Baran et al., 2007). Also, most of the intensively used

agricultural areas in semiarid regions around the world are located

in large to very large basins with surfi cial geology dominated by

continental, geologically young, unconsolidated deposits of typi-

cally very heterogeneous structure. Our current understanding of

the variability of hydraulic properties and their impact on solute

fate and transport below the root zone is therefore limited and

based on greatly simplifi ed models.

In this study, we characterized the variability of the geologic

(sediment) and hydraulic properties throughout a 16-m-deep,

alluvial vadose zone consisting of unconsolidated, alluvial

deposits typical of the alluvial fans of the eastern San Joaquin

Valley, California. In a novel approach, we used geologic char-

acterization to replace the soil series description common in

other spatial variability studies of soil hydraulic properties. Th e

study was implemented at a research orchard at the University

of California Kearney Agricultural Center. Th e Kearney site pro-

vides a unique, extensively sampled and characterized fi eld site

with a well-controlled, long-term fertilization research experiment

that was completed just before our intensive deep vadose zone

sampling campaign (Onsoy et al., 2005). Specifi cally, this study

(i) determined the hydraulic properties of the deep unsaturated

zone and their relationship to sedimentary facies and texture,

and (ii) statistically and geostatistically analyzed these hydraulic

properties. Th e results provide the basis for an analysis of fl ow

and solute transport in deep alluvial sediments, which will be the

main focus of a subsequent study.

Site Descrip on and Field ExperimentOrchard Experiment Overview

Details of the fi eld site characterization eff orts have been

described in Harter et al. (2005) and Onsoy et al. (2005). Briefl y,

the Kearney site, a former ‘Fantasia’ nectarine [Prunus persica (L.)

Batsch var. nucipersica (Suckow) C.K. Schneid.] orchard, is located

on the east side of the San Joaquin Valley (Fig. 1), approximately

30 km southeast of Fresno, CA, at the University of California

Kearney Research Center. Th e site is about 0.8 ha (2 acres) and

is located on the Kings River alluvial fan, a highly heterogeneous

sedimentary system consisting of coarse channel deposits, coarse

to fi ne overbank deposits, fi ne fl oodplain deposits, paleosols, and

fi ne eolian deposits. Sedimentary layers exposed to the surface for

a suffi cient amount of geologic time have developed soil profi les

with distinguishable horizons. Th e type of sedimentary layering,

the paleosols encountered, and the range of soil textural classes

present at this site are rather typical for many areas in the San

Joaquin Valley that have deep vadose zones (Weissmann et al.,

2002). Similar alluvial conditions are also found in the Salinas

Valley and in the desert basins of southern and southeastern

California. As in many surrounding areas, groundwater levels

at the Kearney site are signifi cantly deeper than the root

zone. Since 1970, water levels have fl uctuated between

approximately 11 and 20 m below the surface. In 1997

(the time of sampling), the unsaturated zone was

approximately 16 m thick.

Th e orchard was planted in 1975 and had four

cultivars of nectarines. A fertilization experiment

was conducted at the orchard with fi ve levels

of fertilization applied in a randomized com-

plete block design. Details of the orchard

geometry and the fertilization experiment

can be found in Johnson et al. (1995) and

Onsoy et al. (2005).

Core Sampling

During 1997, on completion of the

fertilizer experiment, three subplots were

selected for detailed sampling and inten-

sive data analysis (Fig. 2). Approximately

900 m of geologic material were obtained

from 62 continuous soil cores drilled to

the water table (?16 m), with 18 to 19

cores collected at each of the three sub-

plots (Fig. 2). An additional north–south

transect throughout the entire orchard,

consisting of six cores spaced 12 m apart,

F . 1. San Joaquin Valley shown on the map of California; solid box represents the study area.



was sampled to obtain estimates of heterogeneity at the scale of

the entire orchard.

While water content and NO 3 distributions were analyzed

from samples of all 62 boreholes (Onsoy et al., 2005), only 19

of the 62 boreholes were used for the analysis of soil hydraulic

properties and laboratory texture analysis. A complete sedimen-

tologic description by color, texture, grain size and roundness of

sands and gravels, and sediment structure was performed on all

cores. Texture was identifi ed using fi eld estimation methods of

the Soil Conservation Service (1994); a Munsell color chart was

used to identify color. Cross-bedding, mottling, clay coatings,

aggregate presence and size, and cementation or concretions

were identifi ed in the sedimentologic description. Individual

sediment beds were identifi ed and logged based on the aggregate

of these descriptors.

Laboratory MethodsHydraulic characterization was performed on 120 undis-

turbed core samples taken at various depths from the 19 core

locations. Hydraulic characterization included determination

of saturated hydraulic conductivity, grain size distribution, and

measurement of the dependence between unsaturated hydrau-

lic conductivity, moisture content, and soil water pressure.

Additional measurements such as bulk density and sand, silt, and

clay fractions were also included. Soil moisture was measured in

the fi eld on disturbed core samples taken adjacent to the undis-

turbed core samples.

Saturated hydraulic conductivity was

measured using the constant-head method

(Klute and Dirksen, 1986). Th e Division of

Agriculture and Natural Resources analytical

laboratory determined soil texture based on the

percentages by weight of sand, silt, and clay

(hydrometer method, ASTM, 1985). Bulk

density was obtained gravimetrically from the

undisturbed cores. Th e soil water retention and

unsaturated hydraulic conductivity relations are

basic elements necessary for the simulation and

prediction of fl ow and transport in the vadose

zone. A multistep outfl ow technique (Eching

and Hopmans, 1993) was used to determine

these relationships.

The principle of the multistep outflow

technique is to observe the water outfl ow from

an initially saturated soil core sample along

with soil water suction changes in that sample

at increasing steps of dryness. Th e method has

two components: (i) implementation of a labo-

ratory experiment, and (ii) computer analysis

of the laboratory experiment to determine

the hydraulic parameters of the unsaturated

hydraulic conductivity function and of the soil

water retention curve via inverse modeling.

Implementa on of Mul step Ou low Experiment

For the laboratory experiment, a 10-cm-

long, saturated, undisturbed sample was

placed into a pressure–suction chamber under

atmospheric pressure conditions. During the experiment, the air

pressure was increased in several discrete steps during the course

of several days (typical for sands) to several weeks (typical for

clays). Each stepwise increase in air pressure forces water to fl ow

out of the soil core sample until the soil water suction in the pores

matches the applied air pressure. Using high-precision instrumen-

tation, we monitored how quickly the soil pressure inside the core

changed in response to each pressure step and we monitored the

outfl ow rate from the core with time. Th e core was instrumented

with a tensiometer at its center measuring the soil water suc-

tion. A burette connected to the core captured the outfl ow. Soil

pressure and outfl ow were recorded automatically using pressure

transducers connected to the tensiometers and burettes, and the

data were sent to a computer. After completion of each experi-

ment, the measured data were converted into meaningful units

using laboratory-derived calibration curves (Tuli et al., 2001).

To streamline the implementation of the multistep laboratory

experiments, the 120 samples were arranged into 12 sets (or runs)

of 10 samples (or cells) running in parallel. Th e implementation

of a single set (10 parallel laboratory multistep experiments) typi-

cally took 3 to 6 wk including setup and take-down, depending

on the texture of the samples. Coarse-textured samples are typi-

cally faster to run than fi ne-textured samples due to their faster

response to pressure changes.

Th e multistep outfl ow experiments were successfully com-

pleted for 118 undisturbed cores. Due to a variety of experimental

F . 2. A three-dimensional view of 62 boreholes with their lithofacies descrip ons (SL1, sandy loam; C, clay; S1, predominantly sand; P1, paleosol hardpan; SL2, sandy loam with intercala ons; S2, sand; C-T-L, clayey, silty and loamy material; SL3, sandy loam to fi ne sandy loam; P2, paleosol hardpan). Thick lines around boreholes represent subplots.



complications and errors, however, the multistep outfl ow data for

21 soil cores were unusable, resulting in 97 viable samples for the

inverse modeling process.

Hydraulic Characteriza on: Inverse ModelingTo compute the hydraulic properties of the soil core, the mul-

tistep outfl ow experiment was emulated in computer simulations.

Hydraulic parameters of the computer model were calibrated to

the measured fl ow rate, moisture content, and soil water tension

data obtained during the experiment. From the inverse model-

ing, a set of hydraulic parameters for the soil water retention and

unsaturated hydraulic conductivity functions was obtained. Th e

optimization model solved the one-dimensional Richards equa-

tion of unsaturated fl ow. In its one-dimensional form with the

vertical coordinate, z [L], taken positive upward, the Richards

equation is written as

( ) 1h

K ht z z

⎡ ⎤⎛ ⎞∂θ ∂ ∂ ⎟⎜⎢ ⎥= + ⎟⎜ ⎟⎜⎢ ⎥⎝ ⎠∂ ∂ ∂⎣ ⎦ [1]

where θ is the volumetric water content (dimensionless), h is soil

matric head [L], K is unsaturated hydraulic conductivity [L T−1],

and t denotes time [T].

An existing fi nite element code, SFOPT, was adopted to

simultaneously optimize the soil-water retention, θ(h), and

unsaturated hydraulic conductivity, K(h), parameters, given our

particular experimental setup. Several models have been devel-

oped that describe θ(h) and K(h). We chose to use the soil water

retention function proposed by van Genuchten (1980):

( ) ( )re

s r

1mnh

S h−θ −θ

= = + αθ −θ

[2]

where Se (dimensionless) is the eff ective water saturation (0 ≤

Se ≤ 1), θs and θr (dimensionless) are the saturated and residual

water contents, respectively, and α [L−1], m (dimensionless), and

n (dimensionless) are empirical shape parameters, where m = 1

− 1/n. Substituting Eq. [2] in the capillary model of Mualem

(1976), van Genuchten (1980) derived the following unsaturated

hydraulic conductivity model:

( ) ( )2

1/s e e1 1

ml mK h K S S⎡ ⎤

= − −⎢ ⎥⎢ ⎥⎣ ⎦

[3]

where Ks is the saturated hydraulic conductivity [L T−1], Se and m

are the same parameters as used in Eq. [2], and l is a tortuosity–

connectivity coeffi cient (dimensionless), which was taken as 0.5

in this experiment and was not used in the inverse modeling pro-

cedure (Mualem, 1976). Th e only parameters used in the inverse

modeling process were, therefore, Ks, α, n, θr, and θs. Th ey were

simultaneously determined in the computer model with an opti-

mization algorithm using the Levenberg–Marquardt method.

Among the 97 samples, transient data were unavailable for

all the samples in Runs 7 and 8 (20 samples). Due to transducer

failure, seven samples in Run 4 also had unusable transient data

and thus the total number of data sets was reduced to 70 samples.

For the 27 samples with missing transient data, there existed

handwritten data for the equilibrium conditions between pres-

sure steps during the outfl ow experiment. Implementation of

the inverse modeling for these 27 samples and the remaining

70 samples with transient data was described in detail in Tuli et

al. (2001).

Sta s cal and Geosta s cal Analysis

After obtaining van Genuchten parameters for all soil sam-

ples, statistical and geostatistical analyses were performed on the

hydraulic parameters. Distribution goodness-of-fi t was tested

using a standard Kolmogorov–Smirnov test (Stephens, 1974).

Transformations were applied to those parameters found not to

follow a normal distribution (see below). Diff erences in popula-

tion means of the vector of soil hydraulic properties (hydraulic

conductivity, shape parameters, residual and saturated water con-

tent) among the various classes of sediments were tested using

multivariate analysis of variance (MANOVA; Bray and Maxwell,

1982). Multivariate ANOVA allows testing of group vector means

rather than testing group means of a single variable as in tradi-

tional ANOVA. Th e Wilks’ lambda multivariate test was used to

determine statistical signifi cance. Th e test generates an F value

for which the null hypothesis (that group means are not diff erent)

was here rejected at P values <0.05 unless otherwise indicated.

Multivariate ANOVA requires that the hydraulic properties are

normally distributed and assumes that variances across all sedi-

ment classes are similar (homoscedasticity). Homoscedasticity of

each hydraulic parameter across sediment classes was tested using

Levene’s test (Milliken and Johnson, 1984). Where the Wilks’

lambda test indicated that signifi cant overall diff erences existed

between sediment classes, two post hoc tests were performed to

determine which sediment classes were statistically similar and

which were statistically diff erent. Th e fi rst post hoc test is the

Tukey honestly signifi cant diff erence (HSD) test, which is used

for unequal sample sizes (between groups or sediment classes).

Alternatively, the Newman–Keuls (another post hoc test) is used

to make a pairwise determination of whether parameter means

between groups (sediment classes) are signifi cantly diff erent. In

determining whether diff erences are statistically signifi cant, both

post hoc test methods implicitly account for the fact that, in

the post hoc test, multiple pairwise comparisons are made. Th e

Newman–Keuls is a modifi cation of the HSD test that does not

require a homoscedastic data set. All statistical analyses were per-

formed using STATISTICA Version 6.1 (Statsoft, 2004).

For the geostatistical analysis of the spatial parameter distri-

bution, we computed standard semivariograms for the diff erent

van Genuchten parameters and the results were fi tted to a spheri-

cal semivariogram model. Th e software package GSLIB (Deutsch

and Journel, 1992) was used and the results were used to evaluate

the sampling spacing in the horizontal and vertical directions.

Results and DiscussionGeologic Forma on and Main Textures

Th e material obtained in the borehole cores is exclusively

composed of unconsolidated sediments. Th e sediments range

in grain size from clay to pebble and cover a wide spectrum of

silty, sandy, and loamy sediments in between. Th e colors of the

sediments range from grayish brown to yellowish brown, more

randomly to strong brown (no signifi cant reduction zones). Th e

thickness of individual sediment beds varies from <5 cm for some

clayey or silty fl oodplain and alluvial overbank materials to >2.5

m for large sandy deposits associated with buried former stream

channels. Both sharp and gradual vertical transitions are present

between texturally diff erent units. Based on fi eld descriptions of



texture, color, and degree of cementation, fi ve major sediment

units were identifi ed: (i) sand, (ii) sandy loam, (iii) silt loam to

loam, (iv) silt to silty clay, and (v) paleosol. Th e relative occur-

rence of each category as a percentage of the vertical profi le length

are 17.2% sand, 47.8% sandy loam, 13.8% silt loam to loam,

8.3% silt to silty clay, and 12.9% paleosol. Th e following is a brief

description of the main features of each sediment unit.

Th e sand (S) is quartz rich and contains feldspar, muscovite,

biotite, hornblende, and lithic fragments. Cross-bedding at the

scale of a few centimeters could be observed occasionally within

fi ne-grained sand, showing reddish-brown layers intercalated with

gray-brown ones. Th e dominant color of the sand is a light gray

to light brown as the brown hue increases with increasing loam

content. Th e mean thickness of sand layers is 1.7 m; however, it

can reach as much as 2.5 m. Very coarse sand and particles up to

pebble grain size (up to 1 cm) could be observed occasionally at

the bottom of sand units but were not present in all cores.

Sandy loam (sL) sediments have usually light olive to yellow-

ish brown color. Some of these sediments are considered to be

weakly developed paleosols because of their stronger red-brownish

color, root traces, and the presence of aggregates. Mean bed thick-

ness is 50 cm; however, it can be as much as 2 m thick. Th e

sorting is moderate to good. Th in clay or silt layers (0.5–1 cm)

occasionally occur in sandy loam units.

Silty loam and loam (tL-L) is usually slight olive brown to

brownish gray in color. Th e bed thickness is at a scale of a few

centimeters to tens of centimeters. Fine-grained sediments often

show sharp contacts between the units. Changes from one unit

to the next exist at small distances. Cross-bedding can more fre-

quently be observed within silty sediments than in fi ne sands.

Root traces and root-shaped mottles are quite common.

Silt and silty clay (T-C) is the fi nest textured sediments

observed in distinct layers. Th ese are distinguished from the

tL-L category by the apparent absence of sand. Th e main color

is brownish gray to olive brown. Fine, <1-mm-thick root traces

and mottles are also quite frequent in these fi ne-textured sedi-

ments. Much of these occur between 8- and 13-m depth where

we observed vertically and laterally quite heterogeneous and rela-

tively thin bedding of varying, mostly fi ne textures. In this region,

the fi ne-textured units are often thinly laminated to massive and

have a mean thickness of 12.8 cm, while the mode is only about

3 cm due to the thickness distribution being highly positively

skewed. Many of the sediments in this category have the appear-

ance of glacial rock fl our. A thick, clayey silt bed even extends to

50-cm thickness and was observed across the site in most of the

cores. Th e average spacing between these fi ne-textured strata is

56 cm, with 20% spaced <10 cm apart, 16% spaced 10 to 20 cm

apart, and 10% spaced 20 to 30 cm apart.

Paleosols (P) could be recognized in diff erent stages of matu-

rity and play an important role in the geologic interpretation of

the profi les. Th e paleosols separate distinctly diff erent sedimen-

tary deposition regimes or stratigraphic sequences (Weissmann

et al., 2002) and were generally observed across the site in all

boreholes. Th ey show a brown to strong brown, slightly reddish

color, exhibit aggregates, ferric nodules and concretions, and few

calcareous nodules, and are identifi ed in drilling logs as hard,

cemented layers. Th ey display a sharp upper and a gradual lower

boundary, as is typical for paleosols (Retallack, 1990). Clay con-

tent decreases downward in the paleosols. Th e thickness of the

paleosol horizons ranges from 50 cm to about 2 m.

Stra graphy and Main LithofaciesMajor stratigraphic units (lithofacies) observed in the core

profi les were used to construct a large-scale geologic framework

for the research site based primarily on the similarities in the

sequence of sediment units between individual profi les (Fig. 2).

Th e framework consists of a vertical sequence of lithofacies, where

each lithofacies is characterized by its textural, color, and sedi-

ment structure composition. Th e deepest parts of the cores from

15 to 15.8 m display a strong brownish colored, partly clayey

paleosol hardpan (P2). From a depth of 12 to 15 m below the

surface, the main textural units are sandy loam to fi ne sandy

loam (SL3). Coarse sand and gravel or fi ne-grained sediments are

occasionally right on top of the paleosol. Th ese sediments show

a remarkable wetness due to proximity to the aquifer water table.

Sediments between about 8- and 12-m depth are vertically and

laterally quite heterogeneous with relatively thin bedding (thick-

ness of centimeters to decimeters), consisting mainly of clayey,

silty, and loamy material (C-T-L). Between 6 and 8 m below the

surface, a sand layer (S2) is found with laterally varying thickness

averaging 1.7 m. A weak, mostly eroded paleosol is developed

on top of the sand unit. Sandy loam with intercalated sand and

clayey and silty material (SL2) is found at a depth about 4 to 6 m

below the surface. A 0.2-m- to >1-m-thick paleosol hardpan (P1)

occurs at a depth of about 3 to 4 m. Variable sedimentary struc-

tures, predominantly consisting of sand (S1), sit on top of the

hardpan and have a mean thickness of 0.75 m. Occasionally, the

P1 layer is covered by a very thin clay layer (C) with a thickness

of <0.25 m. Th is clay layer was not found in many of boreholes,

however, and was not included in any further analysis. Sandy

loam and subordinated loamy sand and loam (SL1) are present

from the top of that sandy layer to the land surface with an aver-

age thickness of 2.5 m (Fig. 3).

Stratigraphically, the sediments of this unsaturated zone pro-

fi le represent (from top to bottom), the Modesto, Riverbank, and

Upper Turlock formations, where the upper hardpan represents

the Riverbank Paleosol, separating the younger Modesto forma-

tion from the older Riverbank formation. Th e lower hardpan, just

above the water table, represents the Upper Turlock Lake Paleosol

(Weissmann et al., 2005). While spatially extensive, the absolute

elevations of the upper and lower boundaries of the individual

facies appear to vary signifi cantly in space, indicating a small but

signifi cant slope in the facies boundaries.

Equivalent to using soil horizons in root-zone studies, the

extensive and detailed sedimentologic data set was used to guide

the selection of undisturbed sediment core locations for later

hydraulic analysis with the multistep outfl ow experiments. Th e

selected core locations were distributed across the various facies

and were classifi ed by their respective facies membership. Clear

polyethylene terephthalate glycol acetate liners for the soil core

collection allowed an in-fi eld determination of the undisturbed

core sample location such that a core sample would be located

within a facies; however, a core sample did not necessarily consist

of a single massive sedimentary layer. Th e 10-cm cores, in some

cases, included multiple identifi able sediment layers, especially in

the fi ner textured facies. Where possible, undisturbed cores were

taken from relatively homogeneous, massive sediment structures.



Th e measured texture classifi cation represents the bulk 10-cm

core and was obtained only after the multistep experiment was

performed (laboratory classifi cation). In some cases, primarily for

samples representing the fi ne-textured facies, fi eld identifi cation

of texture and laboratory determination of texture diff ered some-

what. Laboratory analyses were generally higher in sand content

than fi eld descriptions (Table 1). It is thought that this refl ects

the presence of a signifi cant amount of very fi ne sand in the fi ne-

textured samples, but also the nonuniformity in the texture of

individual multistep outfl ow sediment cores.

To determine, whether sediment classifi cation accounts for

some of the observed spatial variability in hydraulic properties,

we statistically tested all three sediment classifi cation schemes

developed as part of this study (Table 1):

Classifi cation Type I is a fi eld description of the sediments as 1. described above. Th is classifi cation led to fi ve classes [sedi-ment fi eld (fi ve groups)]: sand (S), loamy sand and sandy loam (sL), silt loam and loam (tL-L), silt and clay (T-C), and paleosols (P).

Classifi cation Type II is a laboratory analysis of sediment tex-2. ture, which led to four classes [sediment lab (four groups)]: sand (S), loamy sand (lS), sandy loam (sL), and silt loam and loam (tL-L). Th e lab analysis did not identify any samples as only clay or silt.

Classifi cation Type III is the eight-lithofacies designation 3. [lithofacies (eight groups)] described above: SL1, S1, P1, SL2, S2, C-T-L, SL3, and P2.

Th e lithofacies classifi cation of the sediment samples could

also be simplifi ed by combining multiple, vertically separated

facies with similar properties into a single class (e.g., SL1, SL2,

and SL3 all become SL), which yielded a Classifi cation Type IV

[sediment (four groups)] with four groups or classes: S, SL, P,

and C-T-L.

Hydraulic Parameters

Inverse modeling to obtain the van Genuchten parameters

from the multistep outfl ow experiments was repeated for three

sets of initial parameter values. Repeated optimization provides

a measure to evaluate the uniqueness of the optimized param-

eter set. Th e three optimized parameter sets, corresponding to

the three sets of starting values, provided very similar albeit not

identical results (Fig. 4). We typically observed some small diff er-

ences in the simulated output between the three diff erent optimal

data sets. Th e fi nal van Genuchten parameters obtained through

optimization of either the transient or steady-state data were

selected specifi cally by comparing the mass balance error of the

computed fl ow simulation and the sum of the squared residual

of the measured vs. simulated data. Th e mass balance error for

most of samples was found to be <2%.

For 27 of the 97 samples, no transient data were available

from the multistep outflow experiments; instead, hydraulic

parameters other than Ks were fi tted by considering only the

steady-state water content and tension at the end of each step.

For these 27 samples only, the Ks values used for the statistical

and geostatistical analyses were obtained from direct measure-

ment of saturated fl ow such that a complete data set was available

for the analysis. Th e correlation between measured log Ks and

fi tted log Ks for the remaining 70 samples is signifi cant but not

strong (Pearson’s r = 0.31). More importantly, their means and

variances are not statistically diff erent (P < 0.05), thus avoiding

statistical bias.

Sta s cal Analysis of Total and Between-Facies VariabilityFor the 97 soil samples, the average Ks was 5.07 cm h−1 and

the variance of log Ks was 0.93, which is remarkably similar to

the value reported by Russo and Bouton (1992) at the Bet Dagan

trench site, which is 2.5 m deep. Optimized saturated hydraulic

F . 3. One of extracted typical core at the site. Small pictures show changes of lithofacies along the core (SL1, sandy loam; C, clay; S1, predominantly sand; P1, paleosol hardpan; SL2, sandy loam with intercalations; S2, sand; C-T-L, clayey, silty and loamy material; SL3, sandy loam to fine sandy loam; P2, paleosol hardpan).



T 1. Sample loca ons, sediment designa ons, and hydraulic parameters saturated hydraulic conduc vity Ks, the van Genuchten shape-fi ng parameters α and n, and residual and saturated volumetric water content (θr and θs, respec vely).

SampleLoca on

Ks α n θr θsField texture

(Type I)†Lab texture

(Type II)†Lithofacies(Type III)‡x y z

———————— m ———————— cm h−1 cm−1

1 13.72 53.64 2.81 3.057 0.031 7.469 0.069 0.322 S S S12 13.72 54.86 2.88 14.566 0.046 5.002 0.079 0.307 S S S13 25.91 54.86 6.69 37.793 0.046 4.611 0.055 0.339 S S S24 25.91 54.86 7.87 22.611 0.043 4.898 0.054 0.331 S S S25 56.39 42.67 6.71 1.203 0.008 3.933 0.148 0.369 S S S26 56.39 42.67 8.22 60.494 0.044 5.046 0.060 0.354 S S S27 59.44 42.67 2.88 24.502 0.053 5.243 0.122 0.300 S S S18 59.44 42.67 6.71 23.974 0.036 4.674 0.045 0.367 S S S29 56.39 54.86 6.85 22.729 0.026 4.998 0.155 0.399 S S S210 56.39 54.86 7.92 1.451 0.027 3.563 0.166 0.341 S lS S211 25.91 54.86 13.98 0.152 0.005 3.081 0.032 0.222 S S SL312 59.44 42.67 14.92 26.632 0.019 2.424 0.097 0.229 S S SL313 59.44 54.86 2.59 0.055 4.970 0.050 0.267 S S S114 62.48 42.67 6.85 0.034 3.833 0.074 0.284 S S S215 62.48 42.67 15.13 0.178 0.003 3.063 0.000 0.237 S lS SL316 65.53 54.86 2.73 0.048 3.933 0.079 0.237 S S S117 68.58 54.86 2.90 0.077 3.597 0.063 0.319 S S SL118 74.68 42.67 2.90 0.078 3.551 0.050 0.272 S S S119 74.68 54.86 2.83 0.061 3.910 0.060 0.281 S S S120 77.72 54.86 2.73 0.060 3.911 0.066 0.283 S S S121 22.86 54.86 10.66 0.028 0.007 2.760 0.258 0.309 sL sL C-T-L22 22.86 54.86 14.92 0.214 0.004 2.225 0.150 0.275 sL lS SL323 56.39 42.67 5.49 1.203 0.008 3.933 0.148 0.302 sL sL SL224 56.39 42.67 14.17 0.056 0.005 1.658 0.000 0.252 sL lS SL325 59.44 42.67 10.60 0.356 0.004 1.547 0.210 0.395 sL tL-L C-T-L26 59.44 42.67 14.01 0.416 0.008 2.346 0.042 0.269 sL lS SL327 62.48 42.67 5.94 0.074 0.006 1.983 0.000 0.329 sL sL SL228 62.48 42.67 13.17 1.375 0.015 2.720 0.032 0.263 sL sL SL329 77.72 42.67 5.70 0.018 0.002 1.400 0.064 0.329 sL tL-L SL230 77.72 42.67 9.14 0.241 0.013 1.479 0.001 0.346 sL sL C-T-L31 16.76 54.86 9.35 0.652 0.008 4.586 0.211 0.248 sL sL C-T-L32 16.76 54.86 11.28 0.260 0.006 1.932 0.131 0.256 sL lS C-T-L33 25.91 54.86 11.72 2.230 0.015 2.667 0.279 0.309 sL lS C-T-L34 56.39 54.86 5.12 0.254 0.011 2.717 0.155 0.248 sL lS SL235 56.39 54.86 11.82 1.439 0.011 1.724 0.138 0.345 sL sL C-T-L36 68.58 54.86 13.06 0.253 0.006 2.680 0.050 0.278 sL lS SL337 74.68 54.86 11.72 0.126 0.003 2.599 0.130 0.285 sL sL C-T-L38 77.72 54.86 11.72 0.370 0.007 4.569 0.151 0.267 sL lS C-T-L39 13.72 53.64 1.47 0.045 0.028 1.217 0.000 0.257 sL lS SL140 13.72 53.64 2.26 0.222 0.005 1.673 0.000 0.262 sL lS SL141 16.76 53.64 1.59 0.019 0.017 1.435 0.090 0.245 sL sL SL142 16.76 53.64 2.08 0.247 0.015 1.536 0.090 0.265 sL lS SL143 22.86 54.86 2.17 0.946 0.007 1.691 0.031 0.287 sL sL SL144 59.44 42.67 2.12 0.593 0.010 2.036 0.090 0.254 sL lS SL145 56.39 54.86 1.97 0.386 0.027 1.300 0.000 0.270 sL lS SL146 65.53 54.86 1.82 0.842 0.011 2.407 0.084 0.229 sL lS SL147 68.58 54.86 2.02 4.270 0.022 2.197 0.090 0.288 sL lS SL148 77.72 54.86 1.82 1.070 0.018 2.078 0.090 0.245 sL lS SL149 59.44 54.86 9.14 0.099 0.007 1.183 0.000 0.333 tL-L tL-L C-T-L50 59.44 54.86 9.45 0.121 0.006 1.784 0.090 0.245 tL-L sL C-T-L51 65.53 54.86 4.22 0.050 0.017 1.474 0.090 0.238 tL-L sL SL252 65.53 54.86 4.75 5.610 0.012 1.450 0.090 0.229 tL-L sL SL253 65.53 54.86 5.37 2.830 0.016 1.342 0.090 0.278 tL-L sL SL254 65.53 54.86 10.45 2.880 0.022 1.137 0.090 0.366 tL-L tL-L C-T-L55 65.53 54.86 12.04 3.560 0.021 1.136 0.061 0.328 tL-L sL C-T-L56 68.58 54.86 11.87 24.600 0.023 1.193 0.090 0.325 tL-L sL C-T-L57 74.68 42.67 10.45 15.400 0.016 1.128 0.058 0.471 tL-L tL-L C-T-L58 74.68 42.67 11.89 0.017 0.003 1.367 0.090 0.307 tL-L tL-L C-T-L59 16.76 53.64 6.64 0.212 0.005 2.407 0.172 0.362 tL-L sL SL260 16.76 53.64 8.22 1.616 0.007 1.870 0.268 0.408 tL-L tL-L C-T-L61 16.76 54.86 8.33 0.391 0.003 2.511 0.075 0.385 tL-L tL-L C-T-L62 22.86 54.86 5.02 0.793 0.007 3.136 0.122 0.296 tL-L sL SL263 22.86 54.86 8.90 0.014 0.002 5.042 0.000 0.354 tL-L tL-L C-T-L64 59.44 42.67 12.36 30.000 0.012 2.054 0.318 0.426 tL-L tL-L C-T-L

Table con nued.



SampleLoca on

Ks α n θr θsField texture

(Type I)†Lab texture

(Type II)†Lithofacies(Type III)‡x y z

———————— m ———————— cm h−1 cm−1

65 65.53 54.86 9.13 1.941 0.011 2.854 0.169 0.389 tL-L lS C-T-L66 68.58 54.86 5.73 9.554 0.009 5.298 0.164 0.333 tL-L lS SL267 77.72 42.67 12.80 0.204 0.001 2.153 0.000 0.348 tL-L tL-L SL368 74.68 54.86 12.79 12.246 0.009 2.638 0.311 0.386 tL-L tL-L C-T-L69 22.86 54.86 13.09 0.357 0.004 1.616 0.003 0.397 T-C sL C-T-L70 25.91 54.86 12.43 0.201 0.011 1.866 0.364 0.433 T-C tL-L C-T-L71 56.39 42.67 10.36 12.783 0.008 1.487 0.255 0.458 T-C tL-L C-T-L72 56.39 42.67 13.16 0.150 0.005 2.067 0.017 0.276 T-C sL SL373 59.44 42.67 9.03 1.069 0.011 1.422 0.166 0.300 T-C sL C-T-L74 59.44 42.67 12.92 2.115 0.004 3.143 0.245 0.348 T-C tL-L C-T-L75 59.44 54.86 10.48 30.000 0.011 2.908 0.369 0.443 T-C tL-L C-T-L76 62.48 42.67 10.20 0.565 0.003 2.533 0.339 0.408 T-C tL-L C-T-L77 77.72 54.86 10.50 0.008 0.007 1.709 0.396 0.426 T-C tL-L C-T-L78 13.72 53.64 3.43 2.732 0.012 1.976 0.158 0.251 P sL P179 16.76 54.86 3.10 0.251 0.005 2.810 0.165 0.230 P sL P180 25.91 54.86 2.73 0.338 0.003 1.206 0.008 0.344 P sL P181 25.91 54.86 3.19 0.179 0.007 1.501 0.138 0.328 P lS P182 56.39 54.86 3.19 24.187 0.011 2.720 0.168 0.303 P sL P183 56.39 54.86 3.85 0.481 0.006 1.122 0.000 0.244 P sL P184 62.48 42.67 4.06 2.557 0.012 2.474 0.143 0.297 P sL SL285 65.53 54.86 3.45 0.319 0.011 2.859 0.234 0.279 P sL P186 74.68 42.67 4.57 0.358 0.004 1.359 0.005 0.270 P sL SL287 77.72 42.67 2.88 0.141 0.006 2.138 0.139 0.283 P sL P188 56.39 42.67 15.07 1.283 0.011 2.460 0.144 0.279 P sL P289 56.39 42.67 15.43 0.025 0.002 1.256 0.017 0.281 P sL P290 59.44 42.67 15.53 0.221 0.008 1.840 0.151 0.239 P sL P291 59.44 54.86 15.24 0.471 0.007 2.096 0.081 0.255 P sL P292 62.48 42.67 15.53 0.157 0.008 1.954 0.153 0.237 P sL P293 65.53 54.86 15.58 3.095 0.016 1.718 0.117 0.315 P sL P294 68.58 54.86 15.54 0.163 0.008 2.466 0.133 0.275 P sL P295 74.68 42.67 15.38 0.751 0.009 2.359 0.120 0.293 P sL P296 77.72 42.67 15.48 1.040 0.007 2.579 0.075 0.238 P lS SL397 74.68 54.86 15.38 0.687 0.007 3.321 0.172 0.263 P sL P2

† S, sand; sL, sandy loam; lS, loamy sand; tL-L, silty loam to loam; T-C, clay and silt; P, paleosol.

‡ S, sand; SL, sandy loam; C-T-L, clay, silt, and loam; P, paleosol. Numbers dis nguish among similar lithofacies at diff erent depths.

T 1. Con nued.

F . 4. Example observa ons and param-eter op miza on for the mul step ou low experiment: suc on head (top) and total-ized ou low (bo om). The three solid lines correspond to three calibrated simula on results, where each automated calibra on was based on a diff erent star ng value for the calibra on parameters (rela vely low, interme diate, and high).



conductivity showed characteristics of a lognormal distribution,

with values varying across more than four orders of magnitude.

Saturated hydraulic conductivity values were found to range

between 60.5 cm h−1 (associated with a sand sample) and 0.0077

cm h−1 (associated with a silt loam sample). Th e van Genuchten

shape factors α and n also showed characteristics of a lognormal

distribution, with maximum and minimum values of α of 0.078

and 0.0013 cm−1, and were also associated with sand and silt

loam samples, respectively. Maximum and minimum values of n

were 7.47 and 1.12 and were associated with sand and hard pan

samples, respectively.

A Kolmogorov–Smirnov goodness-of-fit test confirmed

that the hydraulic parameters followed a lognormal distribution

(signifi cance level of 0.05). Th e hydraulic parameters θr and θs

were found not to follow either normal or lognormal distribu-

tions, with many data points spanning the whole range of both

parameters. Th e hypothesis of normality for the original or log-

transformed data was rejected by the Kolmogorov–Smirnov test

(signifi cance level of 0.1). Transformation of the θr and θs data

(Johnson and Kotz, 1970) was performed to produce normally

distributed variables. Hyperbolic acrsine (SU) transformation was

applied to the θr data while log-ratio (SB) transformation was

applied to the θs data (Carsel and Parrish, 1988). Th ese transfor-

mations are given as

( ) ( )SB: logY X A X B⎡ ⎤= − −⎣ ⎦ [4]

( ) ( )1 21 2SU: sinh ln 1Y U U U− ⎡ ⎤= = + +⎢ ⎥

⎢ ⎥⎣ ⎦ [5]

where U = (X − A)/(B − A) and

A and B are fitting parameters.

Parameters A and B were fi tted for

the θr and θs data independently

and the normality hypothesis for

the set of 97 transformed data

was signifi cant for both θr and

θs (Fig. 5).

Despite the lack of living

roots or agricultural modifi cation

of the measured sediments in the

deep vadose zone, similarities are

apparent between the range and

variability of the optimized sedi-

ment hydraulic parameters for our

site and those reported previously

in several studies investigating

root-zone hydraulic properties

only. Th e van Genuchten param-

eters Ks, α, and n were previously

reported to be lognormally

distributed by Hopmans et al.

(1988) at the Hupsel watershed.

Th e Bet Dagan trench data (Russo

and Bouton, 1992) also suggested

that α was lognormally distrib-

uted. Th e distribution type for these

parameters therefore appears to not

be aff ected by the absence of plant roots or agricultural practices

in these unsaturated sediments.

Th e hydraulic parameters for each sample are not entirely

independent of each other: hydraulic conductivity (log Ks) is sig-

nifi cantly correlated to log α and log n. Th e two shape parameters

of the van Genuchten model, log α and log n¸ are also signifi -

cantly correlated, as well as the two transformed bounds of the

water content, θr′ and θs′. Th ere are no signifi cant correlations

between the shape parameters and the water content parameters.

While signifi cant, the linear correlation coeffi cients (Pearson’s r) are relatively low (0.2–0.4, Table 2), except between log Ks and

log α, where r = 0.7. Th e latter confi rms fi ndings at a fi eld site

involving root-zone samples only, where signifi cant but generally

mild correlations were found between saturated hydraulic con-

ductivity and shape parameters (de Rooij et al., 2004).

We next consider the relationship between hydraulic param-

eter spatial variability and sediment characterization, which can

be obtained from fi eld characterization, possibly enhanced by

laboratory analysis of grain size distributions. Th e lithofacies

model constructed from the detailed fi eld geologic descriptions

(Classifi cation Type III) provides a framework for the macro-

scopic distinction of various sedimentary units that are—in the

framework of identifying macroscopic variability—equivalent to

the common distinction made between diff erent soil units iden-

tifi ed in soil surveys. Th e lithofacies concept, however, operates

within a three-dimensional sedimentary context rather than a

two-dimensional soil map. Similar to the approach applied by

Allen-King et al. (1998), we used the facies-based approach as a

tool to separate macroscopic, large-scale variability from micro-

scopic, Darcy-scale variability of sediment hydraulic parameters

F . 5. Measured correlograms and histograms of the van Genuchten shape-fi ng parameters α and n, residual volumetric water content θr, saturated volumetric water content θs, and saturated hydraulic conduc vity Ks (N = 97).



given by the van Genuchten model. In particular, we were

interested in determining the degree to which fi eld-identifi able

sedimentary structure constrained laboratory-derived hydraulic

properties commonly used for unsaturated fl ow modeling. While

visual inspection of soil properties has been found to be a weak

indicator of spatial variability in hydraulic properties within a

given soil (e.g., Nielsen et al., 1973), macroscopic lithofacies is

commonly used to distinguish hydraulically distinct groundwater

fl ow units (Davis et al., 1993; Weissmann and Fogg, 1999). Does

the spatial variability of hydraulic parameters between fi eld-iden-

tifi able sediment or lithofacies types explain a signifi cant portion

of the overall variability of the hydraulic parameters? To consider

the power of various sediment classifi cation schemes, we also

considered whether the various sediment classifi cation schemes

diff ered markedly in capturing the between-sediment variability

of the hydraulic parameters.

Th e mean, its standard error, and especially the standard

deviation of the hydraulic properties show signifi cant diff er-

ences between lithofacies, particularly for log Ks, log n, and log

α (Table 3, Fig. 6). Th e largest values of these parameters were

found in the S1 and S2 lithofacies. Th is outcome is consistent

with the grain size distributions, which indicates that the S1

and S2 facies contain coarse-textured sand materials and were

found to be signifi cantly diff erent from the other facies due to

the near-complete absence of any fi ne-textured material. In con-

trast, the lowest Ks values were

not surprisingly found in the

C-T-L facies. Consistent with

its fi ne-textured materials, the

C-T-L facies retains more mois-

ture than other facies, thus it has

the highest mean and variance of

residual water content. For satu-

rated water content, both the S2

and C-T-L facies have a higher

mean than the other facies, but

the latter also exhibits the larg-

est variation among all facies.

Th e SL3 has the smallest mean

for both residual and saturated

water content (Table 3).

Considering the sedi-

ment classification obtained

directly from the fi eld descrip-

tion (fi ve groups, Classifi cation

Type I), the results are mostly

consistent with those for the

lithofacies (Classifi cation Type

T 2. Correla ons (r) among soil hydraulic proper es logarith-mic saturated hydraulic conduc vity Ks and the van Genuchten shape-fi ng parameters α and n, and the transformed bounds of the residual and saturated volumetric water content (θr′ and θs′, respec vely). Values >0.2 are signifi cant at P < 0.05.

log Ks log α log n θr′ θs′log Ks 0.68 0.42 0.17 0.22

log α 0.68 0.36 0.00 −0.03

log n 0.42 0.36 0.17 0.02

θr′ 0.17 0.00 0.17 0.29

θs′ 0.22 −0.03 0.02 0.29

T 3. Groupwise means, standard devia ons (SD) of the hydraulic parameters logarithmic saturated hydraulic conduc vity Ks and the van Genuchten shape-fi ng parameters α and n, and the transformed bounds of the residual and saturated volumetric water content (θr′ and θs′, respec vely) obtained from the mul step ou low experiments, listed separately for two sediment classifi ca on schemes.

Group Nlog Ks log α log n θr′ θs′

Mean SD Mean SD Mean SD Mean SD Mean SD

All 97 −0.035 0.934 −1.979 0.386 0.371 0.197 0.989 0.605 −1.086 0.199Classifi ca on Type II: Sediment lab (four groups)†

S 18 1.046 0.625 −1.448 0.319 0.631 0.106 0.794 0.276 −1.088 0.198lS 22 −0.312 0.569 −2.005 0.272 0.361 0.162 0.886 0.587 −1.185 0.156sL 39 −0.299 0.727 −2.082 0.235 0.296 0.154 0.973 0.556 −1.145 0.159tL-L 18 −0.203 1.195 −2.253 0.346 0.285 0.175 1.343 0.829 −0.837 0.101

Classifi ca on Type III: Lithofacies‡SL1 11 −0.342 0.727 −1.788 0.324 0.264 0.136 0.609 0.427 −1.225 0.121S1 8 1.120 0.325 −1.282 0.117 0.665 0.105 0.776 0.197 −1.141 0.119P1 9 −0.139 0.734 −2.125 0.195 0.296 0.158 1.119 0.625 −1.152 0.168SL2 11 −0.339 0.866 −2.113 0.269 0.333 0.209 0.950 0.516 −1.127 0.183S2 8 1.098 0.635 −1.527 0.254 0.644 0.059 0.930 0.394 −0.930 0.094C-T-L 31 −0.158 0.984 −2.126 0.280 0.305 0.182 1.285 0.723 −0.940 0.179SL3 9 −0.400 0.778 −2.247 0.348 0.388 0.084 0.465 0.469 −1.251 0.176P2 10 −0.386 0.597 −2.128 0.235 0.330 0.116 1.102 0.382 −1.216 0.127

† S, sand; lS, loamy sand; sL, sandy loam; tL-L, silty loam to loam.‡ SL, sandy loam; S, sand; P, paleosol; C-T-L, clay, silt, and loam. Numbers dis nguish among similar lithofacies at

diff erent depths.

F . 6. Grouped means and standard error of the mean (95% confi dence interval) of the van Genuchten shape-fi ng parameters α and n, residual volumetric water content θr, saturated volumetric water content θs, and saturated hydraulic conduc vity Ks for Categoriza on Type III (lithofacies, top) and Classifi ca on Type I [sediment fi eld (fi ve groups), bo om].



III) characterization (Fig. 6): the S facies has the

highest log Ks, log n, and log α. Th e two water

content parameters are highest in the fi nest tex-

tured T-C facies, but also high in the tL-L facies.

Saturated water content is lowest in the paleosol

(P) facies. For the sediment classifi cation based

solely on a laboratory analysis of the grain size

distribution (Classifi cation Type II), the saturated

water content was slightly higher in the coarsest

(S) and fi nest textured groups (tL-L). For the

other parameters in this classifi cation, relative dif-

ferences among means are consistent with other

classifi cations (Table 3).

For the lithofacies classifi cation (Classifi cation

Type III), within-facies variations of the hydrau-

lic properties in the C-T-L facies were nearly

equal to those found within the entire data set.

Th is further confi rms that the C-T-L unit, which

forms the thickest stratigraphic layer, is also the

most variable unit at this site. Th e C-T-L unit

also represents the largest fraction of the data set

(31 samples, almost one-third of the entire data

set). When combining sedimentologically similar

lithofacies (Classifi cation Type IV), however, the

SL group is as large as the C-T-L group, while

the remaining third of the data is comprised of

16 S samples and 19 P samples.

One-way MANOVA shows that differ-

ences in mean hydraulic properties between the

various sediment classes (groups) are statistically

signifi cant for all four investigated sediment clas-

sifi cation schemes (Table 4). Th e signifi cance of

the sediment grouping does not rest in a single

sediment class alone. Post hoc tests using pairwise

comparison of grouped means for each hydraulic

parameter generally showed that there are two

to three, and—in the case of the lithofacies clas-

sifi cation especially for log α—as many as four

homogeneous groups across sediment classes, with

signifi cant diff erences in the parameter means

between homogeneous groups (Table 5). Many

of the parameters are associated with multiple

homogeneous groups. For log α, the two S facies

and three SL facies associate varyingly with each

other and with the two P and the C-T-L facies.

Importantly, the structure of the homogeneous

groups shown in Table 5 is not consistent between

parameters. For the lithofacies characteriza-

tion, each parameter has a distinctly diff erent

association of homogeneous groups (Table 5).

Considering the vector of all five hydraulic

parameters and their varying homogeneous

group associations, the post hoc tests support the

hypothesis that each of the identifi ed lithofacies

classes is characterized by its own unique vector

of hydraulic parameter means.

Sediment classifi cations with fewer classes

provide slightly simpler associations of homo-

geneous groups. Th e most persistent contrast in

T 4. Mul variate ANOVA test results using Wilks’ lambda mul variate test. Each row indicates a separate mul variate ANOVA test. Independent variable: sediment classifi ca on; dependent variables: log saturated hydraulic conduc vity, log van Genuchten shape-fi ng parameters α and n, and transformed bounds of the resid-ual and saturated volumetric water contents.

Classifi ca on scheme Wilks’ λ F df eff ect df error P

Sediment fi eld (fi ve groups), Classifi ca on Type I 0.184 9.768 20 292.8 <0.0001Sediment lab (four groups), Classifi ca on Type II 0.146 16.513 15 246.1 <0.0001Lithofacies (eight groups), Classifi ca on Type III 0.108 7.179 35 360.0 <0.0001Sediment (four groups), Classifi ca on Type IV 0.170 14.789 15 246.1 <0.0001

T 5. Post hoc mul variate tes ng for pairwise group diff erences in the mean of each of the hydraulic parameters log saturated hydraulic conduc vity Ks, log van Genuchten shape-fi ng parameters α and n, and transformed bounds of the residual and saturated volumetric water content, θr′ and θs′, respec vely. Columnwise entries with the same number do not show signifi cant pairwise (between sediment classifi -ca on) diff erence in mean (homogeneous groups). The nonequal N Tukey honestly signifi cant diff erence test was used to determine the sta s cal signifi cance of the diff erence in the mean (P < 0.05). For lithofacies, the results for the Newman–Keul test and the nonparametric Kruskal–Wallis test are shown.

Group log Ks log α log n θr′ θs′Sediment fi eld (Classifi ca on Type I)†

S 1 1 1 1 1, 2sL 2 2 2 1, 2 1P 2 2 2 1, 2 1tL-L 2 2 2 1, 2 2, 3T-C 2 2 2 2 3

Sediment lab (Classifi ca on Type II)†S 1 1 1 1 1lS 2 2 2 1, 2 1sL 2 2, 3 2 1, 2 1tL-L 2 3 2 2 2

Lithofacies (Classifi ca on Type III)‡S1 1 1 1 1, 2 1, 2S2 1, 2 1, 2 1 1, 2 2SL1 3 2, 3 2 1, 2 1SL2 3 3, 4 2 1, 2 1, 2SL3 3 4 2 1 1P1 2, 3 3, 4 2 1, 2 1, 3P2 3 3, 4 2 1, 2 1C-T-L 3 3, 4 2 2 2

Sediment (Classifi ca on Type IV)‡S 1 1 1 1, 2 1, 2SL 2 2 2 1 3P 2 2 2 1, 2 2, 3C-T-L 2 2 2 2 1

Lithofacies (Newman–Keuls)‡S1 1 1 1 1, 2 1S2 1 3 1 1, 2 2SL1 2 4 2 1, 2 1SL2 2 2 2 1, 2 1SL3 2 2 2 1 1P1 2 2 2 1, 2 1P2 2 2 2 1, 2 1C-T-L 2 2 2 2 2

Lithofacies (Kruskal–Wallis)‡S1 1 1 1 1, 2 1, 2S2 1 1, 2 1 1, 2 1SL1 1, 2 1, 3 2 2 2SL2 2 2, 3 2 1, 2 1, 2SL3 2 3 1, 2 2 2P1 1, 2 2, 3 2 1, 2 1, 2P2 2 2, 3 2 1, 2 2C-T-L 2 3 2 1 1

† S, sand; sL, sandy loam; lS, loamy sand; tL-L, silty loam to loam; T-C, clay and silt; P, paleosol.‡ S, sand; SL, sandy loam; C-T-L, clay, silt, and loam; P, paleosol. Numbers dis nguish among

similar lithofacies at diff erent depths.



hydraulic properties is that between the log Ks, log α, and log n

means of the S (or S1 and S2) class(es) and of the fi ner textured

classes. For the fi eld sediment classifi cation, for example, the S

facies for all three of these parameters is signifi cantly diff erent

from the four remaining otherwise homogeneous, fi ner textured

facies. Th e collapsed lithofacies classifi cation (four groups of

Classifi cation Type IV) also shows this association pattern.

Th e hydraulic properties were found to be only partially

homoscedastic across sediment groups (using Levene’s test), with

varying combinations of parameters that show inhomogeneous

variance between groups, depending on the selected sediment

classifi cation schemes. To control the lack of complete homo-

scedasticity, we also tested with the more robust Newman–Keuls

test. Th e latter is more conservative in highlighting diff erences

between groups. It results in slightly more homogeneous asso-

ciations than the Tukey HSD (less contrast, Table 5). We also

applied a completely nonparameteric test of signifi cance, the

Kruskal–Wallis test, which assumes neither a particular underly-

ing distribution nor homoscedasticity. Th at test provided similar

results to the Tukey HSD test, confi rming that the HSD test,

in this case, was not strongly sensitive to the lack of homosce-

dasticity. For the lithofacies Classifi cation Type III scheme, all

three tests confi rmed that each parameter associated diff erently

across sediment groups. Th is underscores the earlier observation

that the individual parameter group means within lithofacies are

all signifi cant in explaining some of the larger scale variability.

Similar results were obtained when applying the Newman–Keul

or Kruskal–Wallis tests to Classifi cation Types I and II. It was

therefore preferable to generate statistics of hydraulic properties

separately for each of these major lithofacies, even if they fall

within similar textural groups.

Geosta s cal AnalysisThis site exhibits exemplary

properties suitable for conducting

hierarchical geostatistical analysis of

hydraulic parameters (Barrash and

Clemo, 2002; Ye et al., 2005). A com-

plete hierarchical analysis, however,

requires an even more extensive data set

than the one available here. For example,

Barrash and Clemo (2002) conducted

hierarchical geostatistical analysis for

4699 porosity data at fi ve fl uvial deposit

units in the Boise Hydrogeophysical

Research Site. Ye et al. (2005) applied

the analysis to 1344 moisture content

data from fi ve layers at the Hanford Site

in the state of Washington. Onsoy et al.

(2005) used nearly 1000 core data to

determine the hierarchical (geo)statis-

tical distribution of water content and

NO3 at the Kearney site.

Cost and time constraints in

performing the multistep outfl ow exper-

iment did not allow us to have such an

extensive set of data as required for

hierarchical geostatistical analysis. Our

data set consisted of only 97 data points

for the eight diff erent facies. Variogram analysis was performed,

instead, on the hydraulic parameter data set as a whole since

there were not enough data points in each lithofacies to perform

hierarchical geostatistical analysis. Lithofacies boundaries were

considered to be texture discontinuities across which hydraulic

properties were uncorrelated. For the geostatistical analysis, only

data pairs with both points belonging to the same lithofacies

were considered.

Directional semivariograms were constructed with lag

intervals (h) appropriately assigned proportional to the average

horizontal and vertical (within-facies) sampling distance (Deutsch

and Journel, 1992). Maximum lag distances were set to no more

than one-half the maximum sampling dimensions (the horizon-

tal sampling domain is the entire orchard length, i.e., ?70 m).

Semivariance values with too few pairs (fewer than two pairs)

were discarded. Th e sample variograms were fi tted to the widely

used spherical model, which provided the most suitable fi t:

( ) ( )3

0 0

3 1, if 0

2 2

h hh c c c h a

a a

⎡ ⎤⎛ ⎞⎢ ⎥⎟⎜γ = + − − ≤ ≤⎟⎜⎢ ⎥⎟⎜⎝ ⎠⎢ ⎥⎣ ⎦

[6a]

( ) , if h c h aγ = ≥ [6b]

where c0, c, and a are the parameters corresponding to the

nugget, the sill, and the range of the semivariogram, respectively.

Parameters for the individual semivariogram models were obtained

by least squares optimization, separately for each direction.

Th e sill of the horizontal semivariogram was similar or close

to that of the vertical direction for all of the van Genuchten

parameters, suggesting that complete variability is gener-

ally observed across the vertical section of the facies (no zonal

anisotropy, Fig. 7). Spatial continuity in the horizontal direc-

tion, however, is much larger than that in the vertical direction.

F . 7. Semivariograms of the logarithms of saturated hydraulic conduc vity Ks and the van Genu-chten shape-fi ng parameters α and n in horizontal and ver cal direc ons. Solid lines represent the best-fi t spherical model to the data.



Horizontal spatial continuity for all parameters lies in the range

between 5 and 8 m. Vertical spatial continuity is in the range of

0.5 to 1.5 m but is also limited by the average facies thickness.

Th e nugget/sill ratio was calculated and used to determine

whether the sampling scheme was fi ne enough to capture the

spatial correlation of the parameters. Iqbal et al. (2005) suggested

that a range of c0/c between 0 and 25% represents a strong spa-

tial dependence, while a range between 25 and 75% represents

moderate spatial dependence, and weak spatial dependence is

suggested for ranges of c0/c >75% and indicates that the distance

between sampling points is not large enough to capture the spatial

continuity of the parameters.

Our analysis shows that the nugget/sill ratio of the horizontal

direction for all parameters was <25%, indicating adequate sam-

pling frequency in the horizontal direction for detection of spatial

structure. Moderate spatial dependence in the vertical direction is

observed, however, suggesting inadequate sampling frequency in

that direction. Th is was attributed to the occasional existence of

thin layers of clays within lithofacies units (see above and Fig. 3).

Similar conclusions were made by Ward and Gee (2002) at the

Hanford Site, where the vertical correlation structure for hydrau-

lic parameters was weak and attributed to the vertical distance

between measurement points (typically about 3 m) being larger

than the possible correlation length in the vertical direction, esti-

mated at 1 m.

ConclusionsUnderstanding the heterogeneity of fl ow processes in the

deep vadose zone is critical to assessing the fate and transport of

point and nonpoint source pollutants to aquifers in many semi-

arid and arid regions with a deep water table in unconsolidated

sediments. Th e deep vadose zone, however, lacks the descriptive

information of larger scale variability in soil properties that is

commonly available for (shallow) soil root-zone studies from

soil surveys.

In this study, extensive continuous sediment core information

was successfully used to characterize the textural and structural

variability throughout the thick unsaturated zone. Eight major

lithofacies units were identifi ed at the site. Subfacies structures

have also been identifi ed at the millimeter, centimeter, and deci-

meter scales, particularly in the fi ner grained sedimentary facies

units, which is believed to cause a high variability of hydraulic

parameters at these facies.

Our work shows that fi eld characterization of major lithofa-

cies provides a hierarchical framework of soil hydraulic property

variability equivalent to, for example, soil mapping units. Two

major sources of variability were identifi ed: large-scale variability

between the major, fi eld-identifi able sedimentologic facies, and

the smaller scale variability within individual sedimentologic

facies. Th e MANOVA results show that the means of the hydrau-

lic parameter vector diff er signifi cantly between individual facies,

ascertaining the signifi cance of individual fi eld-identifi able facies

as a source of hydraulic variability. Th e standard deviation for log

Ks, for example, is 0.93 overall and ranges from 0.6 to 1.0 within

individual facies. Th e spread between the means of log Ks is >1.5

for the lithofacies classifi cation.

Diff erent sediment classifi cation schemes for delineating

larger sedimentologic units provide somewhat diff erent results, but

we observed qualitative similarities in the structure of hydraulic

property means between these schemes. All schemes, including

the eight-group lithofacies scheme, provided classes that were

hydraulically signifi cantly diff erent from one another. Importantly,

sediment texture and structure (e.g., cementation) information

was signifi cant in selecting appropriate facies units. Furthermore,

signifi cant diff erences existed in the vector of hydraulic property

means even for texturally similar but geologically separated units,

therefore justifying a layered lithofacies characterization of the

vadose zone as the larger scale hierarchical level.

Within major facies, smaller scale variability was shown to

contribute signifi cantly to the overall hydraulic property variabil-

ity within the unsaturated zone. Saturated hydraulic conductivity

and the van Genuchten shape parameters α and n were lognor-

mally distributed, similar to other well-characterized soil and

aquifer sites. Th e data for θr and θs followed neither normal nor

lognormal distributions and were transformed using hyperbolic

arcsine and log-ratio transformations, respectively, for the statisti-

cal and geostatistical analyses, which are applied only to normally

distributed random variables. Vertical small-scale variability was

attributed to cross-bedding, thin layering, and intercalated clay

and silt beds observed especially in the fi ner textured facies. Th e

measurement method (multistep outfl ow experiments) has a sup-

port scale of 10 cm (core length), however, which may locally

exceed the observed thickness of both the cross-bedding observed

in the sand and the intercalated silt and clay layers. Hence, addi-

tional variability probably exists at a smaller observable scale

than that of the hydraulic characterization implemented here.

Horizontal continuity of hydraulic properties (the range of the

variogram) was approximately one order of magnitude larger than

the vertical range.

Our study provides a novel and extensive quantitative assess-

ment of hydraulic parameter variability in alluvial sediment

material and a framework for characterizing hierarchical spatial

variability in a deep vadose zone. We expect that similar properties

and similar variability exists in facies at other alluvial fan sites, at

least within the San Joaquin Valley, even if the facies assemblage

or facies sequences are diff erent (Weissmann et al., 1999).

Importantly, the data set provides an opportunity to apply

and test stochastic modeling and other upscaling methods (e.g.,

Harter and Yeh, 1996, 1998; Harter and Zhang, 1999). Highly

heterogeneous fl ow conditions are possibly prevalent at the site,

with strong fi ngering or preferential fl ow paths channeling much

of the water fl ow and solute transport through a relatively small

portion of the unsaturated domain (Onsoy et al., 2005, Harter

et al., 2005). Future work using stochastic simulation methods is

needed to further elucidate and fi eld validate such fi ndings.

ATh is work was performed with funding from the California Fertil-

izer Research Program and the University of California Kearney Foun-dation. We thank in particular Katrin Heeren, Adelphi-Consult GmbH, Berlin, who performed the geological analysis of the cores during drill-ing, and Gary Weissmann, University of New Mexico, for his invalu-able contributions in interpreting the site stratigraphy. We are thankful to Jim MacIntyre and Michelle Denton for implementing and inverse modeling the multistep outfl ow data. Geoprobe Systems in Salina, KS, generously provided direct push coring equipment for this project.

ReferencesAllen-King, R., R. Halket, D. Gaylord, and M. Robin. 1998. Characterizing the

heterogeneity and correlation of perchloroethene sorption and hydraulic con-

ductivity using a facies-based approach. Water Resour. Res. 34:385–396.



ASTM. 1985. D422-63: Standard test method for particle-size analysis of soils.

p. 117–127 In Annual book of ASTM Standards. ASTM, West Con-

shohocken, PA.

Bagarello, V., M. Iovino, and G. Tusa. 2000. Factors aff ecting measurement

of the near-saturated soil hydraulic conductivity. Soil Sci. Soc. Am. J.

64:1203–1210.

Baran, N., J. Richert, and C. Mouvet. 2007. Field data and modeling of water and

nitrate movement through deep unsaturated loess. J. Hydrol. 345:27–37.

Barrash, W., and T. Clemo. 2002. Hierarchical geostatistics and multifacies sys-

tem: Boise Hydrogeophysical Research Site, Boise, Idaho. Water Resour.

Res. 38(10):1196, doi:10.1029/2002WR001436.

Bray, J.H., and S.E. Maxwell. 1982. Analyzing and interpreting signifi cant

MANOVAs. Rev. Educ. Res. 52:340–367.

Carsel, R.F., and R.S. Parrish. 1988. Developing joint probability distributions

of soil water retention characteristics. Water Resour. Res. 24:755–769.

Coutadeur, C., Y. Coquet, and J. Roger-Estrade. 2002. Variation of hydraulic

conductivity in a tilled soil. Eur. J. Soil Sci. 53:619–628.

Davis, J.M., R.C. Lohmann, F.M. Philips, J.L. Wilson, and D.W. Love. 1993.

Architecture of the Sierra Ladrones Formation, central New Mexico: Dep-

ositional controls on the permeability correlation structure. Geol. Soc. Am.

Bull. 105:998–1007.

de Rooij, G.H., R.T.A. Kasteel, A. Papritz, and H. Fluhler. 2004. Joint distribu-

tion of the unsaturated soil hydraulic parameters and their eff ect on other

variates. Vadose Zone J. 3:947–955.

Desbarats, A.J. 1998. Scaling of constitutive relationships in unsaturated heteroge-

neous media: A numerical investigation. Water Resour. Res. 34:1427–1435.

Deutsch, C.V., and A.G. Journel. 1992. GSLIB: Geostatistical software library

and user’s guide. Oxford Univ. Press, New York.

Duff era, M., J.G. White, and R. Weisz. 2007. Spatial variability of southeastern

U.S. Coastal Plain soil physical properties: Implications for site-specifi c

management. Geoderma 137:327–339.

Eching, S.O., and J.W. Hopmans. 1993. Optimization of hydraulic functions

from transient outfl ow and soil water pressure head data. Soil Sci. Soc. Am.

J. 57:1167–1175.

El-Kadi, A.I. 1987. Variability of infi ltration under uncertainty in unsaturated

zone parameters. J. Hydrol. 90:61–80.

Green, T.R., T. Makoto, and H. Kooi. 2007. Potential impacts of climate change and

human activity on subsurface water resources. Vadose Zone J. 6:531–532.

Harter, T., Y.S. Onsoy, K. Heeren, M. Denton, G. Weissmann, J.W. Hopmans,

and W.R. Horwath. 2005. Deep vadose zone hydrology demonstrates fate

of nitrate in eastern San Joaquin Valley. Calif. Agric. 59:124–132.

Harter, T., and T.-C.J. Yeh. 1996. Stochastic analysis of solute transport in het-

erogeneous, variably saturated soils. Water Resour. Res. 32:1585–1595.

Harter, T., and T.-C.J. Yeh. 1998. Flow in unsaturated random porous media,

nonlinear numerical analysis and comparison to analytical stochastic mod-

els. Adv. Water Resour. 22:257–272.

Harter, T., and D. Zhang. 1999. Water fl ow and solute spreading in hetero-

geneous soils with spatially variable water content. Water Resour. Res.

35:415–426.

Herbst, M., B. Diekkruger, and H. Vereecken. 2006. Geostatistical co-regional-

ization of soil hydraulic properties in a micro-scale catchment using terrain

attributes. Geoderma 132:206–221.

Hopmans, J.W., H. Schukking, and P.J.J.F. Torfs. 1988. Two-dimensional steady

state unsaturated water fl ow in heterogeneous soils with autocorrelated soil

hydraulic properties. Water Resour. Res. 24:2005–2017.

Iqbal, J., J.A. Th omasson, J.N. Jenkins, P.R. Owens, and F.D. Whisler. 2005.

Spatial variability analysis of soil physical properties of alluvial soils. Soil

Sci. Soc. Am. J. 69:1338–1350.

Johnson, N.L., and S. Kotz. 1970. Distributions in statistics: Continuous uni-

variate distributions. Vol. 1. Houghton Miffl in Co., Boston, MA.

Johnson, R.S., F.G. Mitchell, and C.H. Crisosto. 1995. Nitrogen fertilization of

Fantasia nectarine: A 12 year study. UC Kearney Tree Fruit Rev. 1:14–19.

Klute, A., and C. Dirksen. 1986. Hydraulic conductivity and diff usivity: Labo-

ratory methods. p. 687–734. In Methods of soil analysis. Part 1. Physical

and mineralogical methods. 2nd ed. Agron. Monogr. 9. ASA and SSSA,

Madison, WI.

Lund, L.J., D.C. Adriano, and P.F. Pratt. 1974. Nitrate concentrations in deep soil

cores as related to soil profi le characteristics. J. Environ. Qual. 3:78–82.

Makkawi, M. 2004. Eff ect of porous medium heterogeneity on water fl ow: A

stochastic hydrofacies approach. Hydrogeol. J. 12:481–487.

Milliken, G.A., and D.E. Johnson. 1984. Analysis of messy data. Vol. I. De-

signed experiments. Van Nostrand Reinhold Co., New York.

Minasny, B., A.B. McBratney, and K.L. Bristow. 1999. Comparison of diff er-

ent approaches to the development of pedotransfer functions for water-

retention curves. Geoderma 93:225–253.

Mualem, Y. 1976. A new model for predicting the hydraulic conductivity of

unsaturated porous media. Water Resour. Res. 12:513–522.

Nielsen, D.R., J.W. Biggar, and K.T. Erth. 1973. Spatial variability of fi eld-mea-

sured soil-water properties. Hilgardia 42:215–260.

Onsoy, Y.S., T. Harter, T. Ginn, and W. Horwath. 2005. Spatial variability and

transport of nitrate in a deep alluvial vadose zone. Vadose Zone J. 4:41–51.

Retallack, G.J. 1990. Soils of the past: An introduction to paleopedology. Unwin

Hyman, London.

Robert, P.C., R.H. Rust, and W.E. Larson (ed.). 1996. Precision agriculture.

Proc Int. Conf., 3rd. ASA, CSSA, and SSSA, Madison, WI.

Russo, D., and M. Bouton. 1992. Statistical analysis of spatial variability in un-

saturated fl ow parameters. Water Resour. Res. 28:1911–1925.

Russo, D., I. Russo, and A. Laufer. 1997. On the spatial variability of parameters of

the unsaturated hydraulic conductivity. Water Resour. Res. 33:947–956.

Soil Conservation Service. 1994. A phosphorus assessment tool. Tech. Note

1901. SCS South Natl. Tech. Ctr., Forth Worth, TX.

Statsoft. 2004. STATISTICA 6.1 data analysis software system. Statsoft,

Tulsa, OK.

Stephens, M.A. 1974. EDF statistics for goodness of fi t and some comparisons.

J. Am. Stat. Assoc. 69:730–737.

Tuli, A., M.A. Denton, J.W. Hopmans, T. Harter, and J.L. MacIntyre. 2001.

Multi-step outfl ow experiment: From soil preparation to parameter esti-

mation. Pap. 100037. Hydrol. Progr., Dep. of Land, Air, and Water Re-

sour., Univ. of Calif., Davis.

van Genuchten, M.Th . 1980. A closed form equation for predicting the hydrau-

lic conductivity of unsaturated soils. Soil Sci. Soc. Am. J. 44:892–898.

Ward, A.L., and G.W. Gee. 2002. Vadose zone transport fi eld study: FY 2002

Test Plan. PNNL-13857. Pac. Northw. Natl. Lab., Richland, WA.

Weissmann, G.S., G.L. Bennett, and A.L. Lansdale. 2005. Factors controlling

sequence development on Quaternary fl uvial fans, San Joaquin Basin,

California, U.S.A. p. 169–186. In A. Harvey et al. (ed.) Alluvial fans:

Geomorphology, sedimentology, dynamics. Spec. Publ. 251. Geol. Soc. of

London, London.

Weissmann, G.S., S.F. Carle, and G.E. Fogg. 1999. Th ree-dimensional hydrofa-

cies modeling based on soil surveys and transition probability geostatistics.

Water Resour. Res. 35:1761–1770.

Weissmann, G.S., and G.E. Fogg. 1999. Multi-scale alluvial fan heterogeneity

modeled with transition probability geostatistics in a sequence stratigraph-

ic framework. J. Hydrol. 226:48–65.

Weissmann, G.S., J.F. Mount, and G.E. Fogg. 2002. Glacially driven cycles in ac-

cumulation space and sequence stratigraphy of a stream-dominated alluvial

fan, San Joaquin Valley, California, USA. J. Sediment. Res. 72:240–251.

Ye, M., R. Khaleel, and T.-C.J. Yeh. 2005. Stochastic analysis of moisture plume

dynamics of a fi eld injection experiment. Water Resour. Res. 41:W03013,

doi:10.1029/2004WR003735.


Appendix 2: Task 2 – Field Scale Modeling To be submitted: Botros, F. E., T. Harter, Y. S. Onsoy, T. Ginn, J. W. Hopmans, 2010. Richards equation-based-modeling to estimate flow and nitrate transport in a deep alluvial vadose zone. In preparation for submission to: Vadose Zone Journal.


1

RICHARDS EQUATION-BASED-MODELING TO ESTIMATE FLOW AND NITRATE TRANSPORT IN A DEEP ALLUVIAL VADOSE ZONE

Farag E. Botros 1, †, ‡, Thomas Harter 1, *, Yuksel S. Onsoy 1, §, and Timothy R. Ginn 2

(in preparation for publication)

1 Land, Air, and Water Resources, University of California – Davis, CA 95616 2 Civil & Environmental Engineering, University of California – Davis, CA 95616 † Also at Irrigation and Hydraulics Department, Faculty of Engineering, Cairo University, Orman, Giza 12613 – Egypt ‡ Now at Daniel B. Stephens & Associates, Inc. Albuquerque, NM 87109 § Now at Kennedy/Jenks Consultants, San Francisco, CA 94107 * Corresponding author email: [email protected] ABSTRACT

Heterogeneity is well known to exist at different scales; from microscopic scale to macroscopic scale. Incorporation of different types of heterogeneity in deep vadose zones is challenging because of the usual lack of information at such sites. In this paper, we used a characterization of geological and hydraulic properties throughout a 16 m deep, alluvial vadose zone consisting of unconsolidated, alluvial deposits typical for the alluvial fans of the Eastern San Joaquin Valley, California. We also used information gathered through a seven-year long field fertilization experiment to estimate the amount of nitrate stored within the vadose zone at the end of the experiment. Simple mass balance calculations were performed and six conceptually different 2-D and 3-D vadose zone numerical models were implemented using varying degrees of hierarchical details of heterogeneity. All methods resulted in a narrow range of estimated stored nitrate which was found to be approximately four times larger than what was measured in the field. This study raises concern about numerical dispersion encountered in Richards equation under conditions of infiltration where gravity/pressure gradient dominates convective flux.

1. INTRODUCTION

Fertilizers, salts, and pesticides continue to be a major source of nonpoint source pollution in agricultural areas. Our knowledge of vadose zone properties is essential to assess the long-term impacts of nitrate fertilizer management practices on groundwater quality especially in agricultural basins of California and similar semi-arid regions where irrigation return flow is a major component of recharge to local groundwater systems. The understanding of processes in the deep vadose zone is a critical factor in designing efficient nutrient management protocols (Ling and El-Kadi, 1998). These protocols aim to enhance agricultural production while protecting groundwater from nitrate contamination.

Considerable amount of research has been applied to study the nitrogen (N) mass balance in the root zone (e.g., Lafolie et al., 1997; de Vos et al., 2000; Allaire-Leung et al., 2001; Stenger et al., 2002). Nitrogen budgeting in the root zone has been widely used in agronomy to determine the fate of N in soils and the potential for N leaching to groundwater. It is commonly thought that the vadose zone below the root zone acts as a buffer zone where nitrate is naturally attenuated by denitrification before reaching groundwater and becoming a health concern. Contrarily, preferential flow paths in the vadose zone might occur and can expedite nitrate


2

leaching in the deep vadose zone and nitrate arrival to groundwater (Ünlü et al., 1990; Harter and Yeh, 1996; Simunek et al., 2003; Baran et al., 2007).

The effect of the vadose zone below the root zone on the estimation of nitrate leaching to groundwater is still not adequately quantified. A large degree of spatial variability in the deep vadose zone below the root zone is rarely characterized or mostly unaccounted for in most groundwater quality assessment studies. Among the main reasons for the limited vadose zone subsurface characterization are prohibitive experimentation costs and technical difficulties involved in sampling soil that is spatially variable and in monitoring nitrate transport at great depths over sufficiently long temporal and spatial scales. However, few studies have been guided by and compared to actual, extensive field data from a deep vadose site (Rockhold et al., 1996; Seong and Rubin, 1999).

We have recently described a deep vadose zone field research site that is ideally suited for testing various heterogeneity models and their associated transient, long-term flow and transport predictions against actually measured two- and three-dimensional profiles (Onsoy et al., 2005; Harter et al., 2005). Our project site is a former orchard on the alluvial fan of the Kings River in the San Joaquin Valley, Fresno, California. The orchard was subject of a long-term, 12-year field scale nitrogen fertilizer trial. Spatial variability of the site is well characterized to the water table at a depth of about 16 m and is considered typical for many alluvial basins in California (Botros et al., 2009). Data from the fertilizer experiment, site characterization, and description of the hydraulic parameters heterogeneity is found in Onsoy et al. (2005) and Botros et al. (2009) and are the foundation for the model development presented in this paper.

The intensive field sampling campaign employed by Onsoy et al. (2005) showed that the stored nitrate mass (calculated using kriged interpolation of nitrogen concentration of 1200 soil core sample) below root zone is much less than what is expected from a simple N and water flux 1-D mass balance analysis; given precipitation and evapotranspiration data along with irrigation and fertilization application and fruit nitrogen uptake records at the site. The main hypothesis was that, given the strong heterogeneity observed at the site, significant preferential flow paths may occur that facilitate nitrate movement through the deep vadose zone to the water table leaving little stored behind. The possibility of denitrification as a reason to explain the little stored N mass within vadose zone was not supported by the fact there was no significant variations in the N mass profile or 15N composition with depth.

We describe the development of six different conceptual modeling approaches using findings of statistical and geostatistical analysis of hydraulic parameters performed recently (Botros et al., 2009). Different levels of heterogeneity complexity are implemented in the different conceptual models. Simulations of both, two-dimensional and three-dimensional domains are implemented. The main objectives of these model runs are 1) to evaluate and discuss the impact of spatial variability observed at the site at two scales - lithofacies and local scale - on long-term dynamics and variations in water flow conditions, nitrate distribution, and nitrate mass in the deep vadose zone, 2) test the hypothesis that the stronger the heterogeneity implemented, the more preferential flow paths occur which might provide an explanation of the low observed nitrate mass in the deep vadose zone, and 3) test if Richards equation can provide preferential flow paths under conditions of infiltration where gravity/pressure gradient dominated convective flux.

This paper is structured as follows. Section 2 gives a brief description of the site and highlights the main findings of our sampling analysis. Section 3 provides a mathematical


3

background on the flow and solute transport in the unsaturated zone. Section 4 explains the main elements in the development of the different conceptual models. Section 5 then presents the results of these different conceptual models, Section 6 discusses the results and Section 7 summarizes the study and highlights main conclusions.

2. SITE DESCRIPTION

Details of the field site characterization efforts have been described in Harter et al. (2005) and in Onsoy et al. (2005). Briefly, the site is a former orchard of ‘Fantasia’ nectarines, about 0.8 ha (2 acres) located at the University of California, Kearney Agricultural Center (http://www.uckac.edu), on the Kings River alluvial plain, 30 km southeast of Fresno, California. As in many other surrounding areas, groundwater levels at the orchard have fluctuated between approximately 11 m and 21 m below the surface with the average thickness of the unsaturated zone approximately 16 m. The site elevation is 103 m above sea level. It has a semi-arid, Mediterranean climate.

From 1982 until 1994, a fertilizer experiment was conducted in a random block design to 14 different subplots at the site with application rates of 0, 110, 195, 280, or 365 kg N ha-1 yr-1 in several replicates. In 1997, three subplots with the 0, 110, and 365 kg N ha-1 yr-1 treatments were selected for sampling. For convenience, the three subplots are named throughout the text as “control”, “standard”, and “high”, respectively. Between July and October 1997, 62 undisturbed continuous soil cores were drilled to water table at depth of 15.8 m and soil samples were collected for the analysis of nitrate and hydraulic parameters distributions (Onsoy et al., 2005, Botros et al., 2009). Soil texture at the core scale was determined by the hydrometer method (Sheldrick and Wang, 1993); soil hydraulic properties were determined using the multi-step outflow method (Eching and Hopmans, 1993); and field soil water content was determined gravimetrically (Klute, 1986). Based on texture, color, and cementation encountered in the cores, eight, statistically significant, different stratigraphic units or layers were identified and are referred to as lithofacies (Botros et al., 2009).

The entire vadose zone at the site consists of unconsolidated sediments deposited on a stream-dominated alluvial fan. Textural groups range from clay, clayey paleosol hardpans to a wide range of silt and sand, occasionally coarse sand and gravel sediments. Lithofacies exhibit vertically varying thicknesses; similar sediment deposits are laterally continuous over the experimental site. Lithofacies units identified here are depicted in Figure 1 and they include from the ground surface towards the water table: upper Hanford sandy loam (SL1), sand (S1), shallow hardpan (P1), sandy loam unit (SL2), sand unit (S2), clay/clayey silt/clay loam (C-T-L), lower sandy loam (SL3), and deep hardpan, or paleosol (P2). Geologic formation and sedimentologic description and details on characteristics of each of these lithofacies are found in (Botros et al., 2009).

3. MATHEMATICAL DEVELOPMENT

Modeling flow and transport processes in porous media relies on the continuum approach which averages flux over a local volume of a porous medium, referred to as the representative elementary volume (REV) (Bear, 1972). This REV-averaged flux is then assigned to the center of the REV that serves as the mathematical definition of the spatial location of the flux.

3.1. Governing Equations for Flow


4

Based on the continuum concept and the REV approach, water flow in variably saturated media at the laboratory scale is governed by the classical Richards equation stated as below for one-dimensional flow:

WSzhhK

zt−⎥

⎦

⎤⎢⎣

⎡⎟⎠⎞

⎜⎝⎛ +∂∂

∂∂

=∂∂ )1)(θ (1)

where θ is volumetric water content [-], t is the time [T], h is the soil water matric head [L], z is vertical coordinate taken positive upward [L], and SW is a sink term which represents the volume of water removed per unit time from a unit volume of soil [T-1]. K(h) is the unsaturated hydraulic conductivity [L T-1].

A number of closed-form formulas have been proposed to empirically describe the dependence of hydraulic functions K(h) and θ(h) on pressure head (Brooks and Corey, 1964; Gardner, 1958; Haverkamp et al., 1977; van Genuchten, 1980; Russo, 1988). Among these relationships, Mualem (1976) and van Genuchten (1980) are widely used in modeling of unsaturated flow and are therefore used in this paper. They can be summarized as follows:

( ) ( )[ ]2/111 mme

les SSKhK −−= (2a)

( ) [ ] mn

rs

re hhS

−+=

−−

= αθθθθ 1 (2b)

where Ks denotes saturated hydraulic conductivity (LT-1), Se (-) is called effective water saturation (0 ≤ Se ≤ 1), θs and θr (-) are the saturated and residual water content, respectively, and α (L-1), m (-), and n (-) are empirical parameters dependent on soil type where m = 1-1/n and l denotes tortuosity/connectivity coefficient (-) which is found to have a value of 0.5 from the analysis of a variety of soils (Mualem, 1976). 3.2. Governing Equations for Solute Transport

The classic advection-dispersion equation for transport has been adopted to account for mixing and spreading of an inert solute during transient simulations. Advection-dispersion equation for a conservative trace is written as:

ci

i

jij

i

SxCv

xCD

xtC

−∂∂

−⎟⎟⎠

⎞⎜⎜⎝

⎛

∂∂

∂∂

=∂∂ θθ (3)

where C is the local concentration in the soil solution [M L-3], vi is the i-th component of water velocity [L T-1], Dij is the hydrodynamic dispersion coefficient tensor [L2 T-1] (i, j = 1, 2, 3), and Sc (M L-3 T-1) represents a sink term for solutes. Knowledge on water content and water velocity, v, is obtained from solutions of the Richards and Darcy’s equations. The second term on the right-hand side of Eq. (3), ii xCv ∂∂ / , is referred to as the advection term that describes the transport of solute traveling at the same velocity as water. The hydrodynamic dispersion coefficient Dij tensor, which describes the combined effect of mechanical dispersion and molecular diffusion, is given by (Scheidegger, 1960)

oijTji

TLij Dvvvv

D ++−= δααα )( (4)

where αL and αT are longitudinal and transverse dispersivities, respectively, v is the magnitude of pore water velocity, δij is the Kronecker delta (δij = 1 if i = j, and δij = 0 otherwise), and Do is molecular diffusion.


5

4. MODEL DEVELOPMENT HYDRUS (Simunek et al., 1996) was used to numerically simulate the movement of

water and nitrate in the unsaturated zone. The program solves the Richards equation (Eq. 1) and the advection-dispersion equation for transport (Eq. 2) using Galerkin-type linear finite element schemes with the mass conservative iterative scheme proposed by Celia et al. (1990). A detailed documentation on the software can be found elsewhere (Simunek et al., 1996; Rassam et al., 2003).

4.1. Model Domain and Boundary Conditions

Most of the drilled soil cores were clustered in three subplots at the field site. Lithofacies sequences along an E-W transect of the drilled soil cores were analyzed and were found to be relatively continuous across the site; however, variable lithofacies thicknesses are observed (Figure 1a). Three of these boreholes with different lithofacies thicknesses are combined together and were considered as a typical cross section of the heterogeneity at the site (Figure 1b) and therefore used in our numerical modeling simulations. The cross section has a width of 6.10 m and a depth of 15.8 m (depth to the water table). In 3-D models, the model domain is extended in the lateral horizontal direction and domain dimensions are assumed to be 6.10 m × 6.10 m × 15.80 m.

For the flow problem, atmospheric boundary conditions were assigned to the top of the domain prescribing daily values of water fluxes including precipitation, irrigation, and evapotranspiration. At the bottom of the domain, a constant pressure of h = 0 cm was imposed, representing the water table position at 15.8 m depth. For the transport problem, a second type (Neumann type) boundary condition was specified along the upper boundary to define N mass flux into the domain through fertilizer applications. A third type (Cauchy type) boundary condition was prescribed at the bottom boundary, where N mass was directed out of the domain at the water table. Vertical boundaries are assumed to be impermeable and were simulated with zero flux boundary conditions for both flow and transport simulations. 4.1.1. Precipitation and Irrigation Inputs

Daily values of precipitation from 1990 to 1996, are obtained from the Parlier station located within 1 km from the site. The station is managed by The California Irrigation Management Information System (CIMIS) (http://www.owue.water.ca.gov). Based on the 7 years of data, the annual precipitation average is 37 cm. In most years, essentially no precipitation was recorded between May and early October.

Throughout the experiment, flood irrigation, common for many orchards in the project area, was performed at the orchard usually from April through September. Dates of irrigation events are tabulated for the simulation duration of 1990 – 1996 (Table 1). The number of annual irrigation applications at the orchard varies from 9 to 17 with an average of 13 irrigations per year. Based on the site record, the amount of water applied, at each irrigation application, has an average of 13.44 cm. Irrigation duration is approximately 20 hours while irrigation frequency varies between few days to few weeks depending on crop growth stage. 4.1.2. Evapotranspiration

Evapotranspiration (ET) defines a combination of two separate processes whereby water is lost from the soil surface by evaporation and from the root zone by crop transpiration. Reference evapotranspiration, ETo, in which potential ET for a grass (or alfalfa) that does not suffer water stress, is predicted using climatologic data (DWR, 2001). Reference evapotranspiration is a climatic parameter and computed from weather data including solar


6

radiation, air temperature, air humidity and wind speed. Daily ETo data with grass as the reference crop is readily available from the local Parlier CIMIS meteorological station. Crop evapotranspiration is calculated through the estimation of the crop coefficient, which is dependent on crop characteristics, vegetative growth stage, canopy cover and height as well as soil surface properties (Doorenbos and Pruit, 1977). HYDRUS requires evapotranspiration pre-partitioned into two components; evaporation and transpiration. Evaporation is controlled by the water content and hydraulic gradient at the soil surface layer and it is essentially important in the early stages of crop development. Transpiration, on the other hand, is distributed over the root zone and can be limited to plant roots by soil water availability. A dual crop coefficient method (Allen et al., 1998) has been used to split evapotranspiration into the two separate components. Annual potential evaporation, averaged over 7 years, was calculated as 17 cm, which was about 15% of the annual potential evapotranspiration of 111 cm. 4.1.3. Nitrogen Application

Dates and amounts of fertilizer applications at the standard and high subplots are tabulated for the simulation period (Table 2). The nitrogen pulse was applied as NO3-N at the soil surface uniformly over the entire horizontal plane over a 1-day period. The vast majority of ammonium-nitrate was applied during late fall when trees were dormant. Ammonium-nitrate was expected to be nitrified and converted to nitrate rapidly (Ünlü et al., 1999). At our observation scale of multiple years, initial time needed for N transformations and their effects on root N uptake were considered negligible. In addition to nitrogen applied to the soil during fertilization, nitrogen was also added through irrigation water. Average NO3-N concentration in irrigation water was taken to be 4 g m-3 (Harter et al., 1999). Thus, each irrigation application supplied approximately 5.4 kg N ha-1.

4.2. Root Zone

Processes such as plant water uptake and plant nutrient uptake occur in the plant root zone. Approximately 90% of the tree roots in the nectarine orchard occur within the uppermost 1.8 m (Scott Johnson, personal communication, 2004). Thus, an average of 1.8 m was used to represent the depth of the tree root zone where the most root activities are confined. 4.2.1 Plant Water Uptake

The sink term, Sw in Eq. (1) represents the volume of water removed per unit time from a unit volume of soil due to plant water uptake. Feddes et al. (1978) defined Sw as

( ) ( ) Pw ShhS ⋅= γ (5) where the water stress response function γ(h) is a prescribed dimensionless function of the soil water pressure head (0 ≤ γ ≤1), and Sp is the potential water uptake rate [T-1] and is calculated as Tp / Lz where Tp (LT-1) is the transpiration obtained from partitioning ET data and Lz is the depth of root zone. 4.2.2. Plant Nitrogen Uptake

The majority of nutrient uptake occurs while soil solution moves to root surfaces by advective transport for plant water uptake (Dalton et al., 1975). This approach, known as passive uptake, incorporates a sink term, such as Sc in Eq. (2), to describe the rate of nutrient uptake as a function of the rate of transpiration. Nitrogen uptake was computed from the product of local water uptake and soil solution N concentration. This approach has been applied in practice to account for plant uptake of nutrients in numerical models (Vogeler et al., 2001).


7

4.3. Initial Conditions Initial distribution of water content or pressure head at the site was not readily available.

We established a pressure profile between the ground surface and the water table by applying the analytical method proposed by Rockhold et al. (1997). This method solves the Richards equation for one-dimensional, heterogeneous layered system under a steady state flux at the ground surface. We applied the method to our two-dimensional flow domain composed of eight horizontal layers. Boundaries of these layers were defined based on the average thicknesses of lithofacies identified at the field. At the site conditions, the steady state flux was estimated as 0.23 cm d-1 from a long-term (14 years) record of the site precipitation, evapotranspiration, and average annual irrigation. The pressure profile varies within the layers as a function of the soil hydraulic properties of each layer (Figure 2).

4.4 Unsaturated Soil Hydraulic Properties

The extensive deep vadose zone sampling campaign at the site provides one of only a few extensive dataset available to date, to the best of our knowledge, for evaluating a model with a more refined characterization of the subsurface heterogeneity. In an attempt to better simulate the subsurface geology and the corresponding soil hydraulic properties, six numerical models were developed to simulate transient flow and nitrate transport through the 15.8 m thick vadose zone. Three of these models were 2-dimensional (2-D) and the other three were 3-dimensional (3-D). The three modeling approaches in 2-D; namely the homogenous lithofacies model (HM-2D), heterogeneous lithofacies model using scaling factors (HetSF-2D), and heterogeneous lithofacies model using van Genuchten parameters (HetVG-2D), represent three different levels of heterogeneity complexity. Similarly, HM-3D, HetSF-3D, and HetVG-3D represent corresponding simulations in 3-D. The main emphasis is to investigate the relationship between the level of heterogeneity implemented in a model, the strength of simulated preferential flow, and the nitrogen content in the deep vadose zone. The other emphasis of these modeling approaches is to evaluate the impact of the spatial variability observed at the site at two scales: lithofacies scale (a few meters in thickness and tens of meters in lateral extent) and local scales (a few centimeters in thickness and a few decimeters in width), on water flow conditions, nitrate distribution, and N mass in the deep vadose zone. Results of the two-dimensional and three-dimensional simulations were compared to each other to test the effects of model domain dimensionality on flow and transport behavior. In all 2-D models, the longitudinal dispersivity (αL) is assumed to be 10 cm and the transverse dispersivity (αT) is assumed to be 1 cm while in 3-D models, and due to larger domain discretization, αL and αT are set to 20 cm and 2 cm, respectively. 4.4.1. Homogeneous Lithofacies Model (HM)

The first model conceptualizes each lithofacies of the eight main lithofacies as a homogenous unit with soil hydraulic parameters defined using the van Genuchten (VG) model and treated as constant deterministic values within each lithofacies. Mean VG parameters of each lithofacies can be found in Botros et al. (2009) and boundaries of each lithofacies are shown in Figure 1. This is considered our simplest model where only between-lithofacies heterogeneity was modeled and within-lithofacies or small-scale heterogeneity was neglected. In 3-D models, and due to the lack of information in the N-S transect, layers are assumed to be perfectly horizontal, along the 6.10 m, in that direction. 4.4.2. Heterogeneous Lithofacies using Scaling Factors (HetSF)


8

In contrast to the HM approach, small scale variability was taken into account in HetSF approach through scaling analysis. Scaling is a technique used to simplify the analysis of hydraulic parameter datasets in heterogeneous unsaturated sediments. It is based on the concept that the various hydraulic parameters, e.g., Ks, α, n, θs, θr, are all related to the pore size distribution and pore geometry. As the pore geometry varies with the type of sediment, the various hydraulic parameters vary accordingly. The scaling factor (λ) is a measure of that change in pore geometry and relates the actual hydraulic function derived for a sample to the scaled hydraulic function. Scaling factors were obtained for all soil samples and have been found to follow lognormal distribution (Figure 3) which is consistent with those reported in other studies (Warrick et al., 1977; Rao and Wagenet, 1985; Vachaud et al., 1988; Hopmans et al., 1988; Tseng and Jury, 1994; Braud et al., 1995; Zavattaro et al., 1999).

Previously, only few studies analyzed scaling factors with geostatistical tools (Jury et al., 1987; Russo, 1991; Zavattaro et al., 1999). In this study, geostatistical analysis of scaling factors has been implemented. Lithofacies boundaries were considered geostatistical discontinuities and no spatial correlation is assumed across lithofacies boundaries. Horizontal and vertical semivariograms are constructed and fitted to spherical covariance function. The horizontal and vertical spherical semivariograms constructed for scaling factors were used to define the degree of spatial variability. Scaling factors did not exhibit a significant horizontal spatial continuity (nugget effect) (Figure 4a) while vertical semivariograms showed a weak spatial continuity of the range of 1.5 m which is close to the average lithofacies thickness. The sill of the horizontal and vertical semivariograms is 0.39, which is close to the average scaling factor variances of different lithofacies, estimated as 0.35 (Onsoy, 2005).

Statistical and geostatistical analysis have been used to generate a random field of scaling factors that is used in our numerical simulation (Figure 5). Consistent with the lithofacies textures (Botros et al., 2009), high values of scaling factors coincide within the coarse-textured materials (S1 and S2), whereas low values were generated within other lithofacies that all have relatively small means. Due to the highly contrasting mean values of scaling factors, scaling factors vary by two orders of magnitude across the lithofacies interfaces, particularly between the S1 and P1 and between S2 and C-T-L. Spatial continuity in the E-W direction was also used to extend the generated random field in the N-S direction for the sake of modeling the 3-D simulation and the resulting random field is also shown in Figure 5. 4.4.3. Heterogeneous Lithofacies using van Genuchten Parameters (HetVG)

Oliveira et al. (2006) demonstrated in principle that when hydraulic parameters are all related to each other through a single variable such as scaling factor the resulting unsaturated flow and transport variability may be significantly underestimated. In the HetVG approach, and similar to HetSF approach, small scale heterogeneity has been taken into account. In HetVG approach, however, the assumption that all hydraulic parameters are all related to the pore size distribution and pore geometry and therefore to a single scaling factor was relaxed. Random fields of Ks, α, and n were generated based on statistical and geostatistical analysis of hydraulic parameters data (Botros et al., 2009). Much larger horizontal continuity than vertical continuity can be observed for all parameters in 2-D and 3-D simulation domains (Figure 6). Within-lithofacies variability of θr and θs was found to be small and was therefore neglected.

4.5. Space and Time Discretization

For 2-D simulations using the HM and HetSF approaches, the domain was discretized into square cells with dimensions Δx = Δz ≈15 cm. In the HetVG approach, and because of high


9

nonlinearity in the parameter values associated with this approach, the domain was discretized through cells with dimensions Δx = Δz =10 cm. In. 3-D simulation, all model conceptualization domains were discretized into cells with dimension Δx = Δy = Δz ≈ 30 cm. All simulations simulation ran for seven years (from January 1, 1990 through December 31, 1996). During the simulations, time step size Δt was automatically controlled according to the convergence history (e.g., the maximum change in pressure head and water content during each time step).

5. RESULTS

Relative water mass balance error below 2% was achieved for all models. Throughout the simulation, the heterogeneous lithofacies models (i.e., HetSF and HetVG) were computationally more expensive and produced slightly higher mass balance errors than the homogeneous lithofacies model. Nitrogen mass balance errors at the standard and high subplots fluctuated from near zero to 2% during the first four years and reached a stable condition around 0.2% in the HM and about 0.5% in both heterogeneous lithofacies models. Peclet number (Pe) and Courant number (Cr) criteria were used to control numerical oscillations and numerical dispersion during simulations. A commonly accepted stability criterion of Pe·Cr ≤ 2 was fulfilled throughout the simulation period (Simunek et al., 1996) for all cases. Very slight differences were observed between 2-D HM and 3-D HM approaches. This is expected because of the way the heterogeneity in the 3-D HM was generated as perfect horizontal layering in the N-S horizontal direction.

5.1. Plant Water and Nitrogen Uptake

In all models, simulated transpiration is smaller than the prescribed potential transpiration, especially during the growth period. The reduction is the result of limited root zone water availability for tree water uptake. Estimated total reduction at the end of the 7 year-period varies significantly between the three models in 2-D simulations and is estimated as 8 cm, 35 cm, and 50 cm for HM, HetSF, and HetVG, respectively. Corresponding values in3-D simulations are 11, 46, and 21 cm, respectively. Yet, all models show similar annual and seasonal trends. At the beginning of each growth period, water uptake by transpiration is low while evaporation is high. However, transpiration becomes the dominant process during the growth period (Figure 7).

Despite discrepancies in actual transpiration between models, plant N uptake estimated by all models agree well with each other, suggesting negligible effects of the model type and dimensionality on plant N uptake. The effect of fertilizer treatment is significant. Total N uptake estimated by the end of the simulation period at the standard subplot is clearly less than the amount at the high subplot (Figure 8). At the standard subplot, average annual plant N uptake, estimated by all models, is 69 kg ha-1. At the high subplot, average annual N uptake value is 113 kg ha-1. These model estimates are reasonably close to the 7-year average annual measured plant N uptake of 77 kg ha-1 and 98 kg N ha-1at the standard and high subplots, respectively (Onsoy et al., 2005).

5.2. Water and Nitrogen Flux at the Water Table

Slight differences between simulated water fluxes to groundwater were observed during most of the simulation time (Figure 9). These differences are believed to be a result of simplifications implemented in assigning initial conditions. These differences were observed early on the simulation and remain almost constant during the whole 7 years of simulation


10

(Figure 9). Under highly transient surface fluxes, water flux to the water table experiences seasonal and annual variations throughout the vadose zone despite its 14-m thickness below root zone. Each year, water flux peaks during the middle of summer and starts declining with increasing water loss at the surface from evapotranspiration. Annual peak fluxes vary from 0.36 cm d-1 in 1996 (with 9 irrigations) to 1.5 cm d-1 during 1992 (with 17 irrigations).

Similar to water fluxes, cumulative N fluxes at the water table show minor variations between different models. Results, however, vary significantly depending on the fertilizer treatments (Figure 10). At the end of the simulation, cumulative N flux at the water table of the high subplot is nearly three times that of the standard subplot. Close agreement between cumulative water flux to water table and total N flux curves reveals that N fluxes are mainly driven by the magnitude of irrigation fluxes prescribed at the surface. This has been also confirmed by field experiments (Biggar and Nielsen, 1978; Wagenet and Hutson, 1989; Troiano et al., 1993) and underscores the importance of budgeting irrigation water in irrigated agriculture to mitigate downward movement of surface applied agrochemicals.

Neither the fertilizer applications (standard or high) nor the model type (HM, HetSF, HetVG) nor model dimensionality (2-D or 3-D) showed a significant effect on the residence time of nitrate in the vadose zone or on the first breakthrough of nitrate at the water table. This can be attributed to the strong uneven layering effect relative to the influence of local-scale heterogeneity. The first arrival of N at the water table at both subplots occurs approximately 3.0 years after the beginning of the simulation or 2.8 years after N is first applied at the surface. Subtracting residence time within the root zone, the travel time in the 14-m deep vadose zone (below the root zone) is approximately 2.6 years.

5.3. Water Velocity Distribution in the Deep Vadose Zone

The flow velocity distribution of all six models after the irrigation season (Figure 11) and during the irrigation season (Figure 12) were plotted and the following observation were extracted. The flow velocity distribution in the HetVG approach showed only slightly more variability than that of the HetSF approach, which in turn showed only slightly more variability than that of the HM (Figure 11). In general, 3-D models showed more velocity variability than corresponding two-dimensional models. In all models, existence of inclined layer boundaries between two adjacent layers with much difference in their hydraulic characteristics has a great impact on directing the flow. This is most pronounced in HM between SL2; a fine-textured layer, and S2; a coarse-textured layer, where a high velocity region is created in the boundaries between these two layers (Figure 11a, d). This kind of layering is expected to cause potential preferential flow path within the sand layer (Kung, 1993). As expected the magnitude of velocity during the irrigation season (Figure 11) is much larger than that after the irrigation season (Figure 12).

5.4. Nitrogen Concentration Distributions in the Deep Vadose Zone

Because of the insignificant variability in flow velocity distribution with the hierarchical model heterogeneity (Figures 11 and 12), the concentration distribution did not show significant differences between simulations (Figures 13 and 14). Two major plumes corresponding to the two latest fertigation applications can be observed in the soil profile in all models. However, N concentrations in the heterogeneous models (i.e., HetSF and HetVG) exhibit more longitudinal spreading than that of the HM and the leading edge of N plumes becomes less smooth. N concentrations experience considerable local variability and vertical spreading, due to the local


11

scale heterogeneity implemented in heterogeneous models. Slight differences can be observed between concentration distribution of two-dimensional and three-dimensional simulations. Concentration also did not show difference in magnitude during or after the irrigation season, underscoring the role of dispersion on the distribution of N in all model simulations.

5.5. Total Nitrogen Mass in the Deep Vadose Zone

From 1990 to 1993, N mass at the deep vadose zone continues to increase at each subplot, resulting in a net accumulation of approximately 300 kg N ha-1 at the standard subplot and 800 kg N ha-1 at the high subplot. Starting in mid-1993, total N mass in the deep vadose zone values remain relatively the same at the standard subplot with small annual fluctuations (Figure 15). Model output for the total N mass in the deep vadose zone in the standard subplot after seven years of nitrogen management in 2-D model domain is 248 kg ha-1, 262 kg ha-1, and 301 kg ha-1 for the HM, HetSF, and HetVG, respectively. Corresponding stored N mass for the high subplot are 534 kg ha-1, 565 kg ha-1, and 638 kg ha-1. For 3-D model domain, corresponding values are 243, 256, and 287 kg ha-1 for the standard subplot and 539, 554, and 614 kg ha-1 for the high subplot. Neither of the different modeling approaches or different model dimensionality explained the low N storage in the deep vadose zone observed in the field estimated by Onsoy et al., 2005 as 36 kg N ha-1 and 87 kg N ha-1 for the standard and high subplots, respectively.

6. DISCUSSION

Despite local discrepancies between the model results, surprising similarities between these results suggest that the lithofacies effect and boundary controls override effects of local heterogeneity. In all models, each irrigation event creates similar large scale patterns of nitrate concentration in the profile, resulting in similar values of N mass flux to the water table is observed.

6.1. Numerical Modeling vs. Root Zone Mass Balance Approach

The root zone mass balance (MB) approach, (Martin et al., 1991) is a simple steady-state one-dimensional model that can be used for predictive purposes. Applying this concept, taking an average annual deep percolation from the root zone as 110 cm/year (Onsoy et al, 2005) and assuming an average field capacity of 25% (Martin et al., 1991), it results in an estimated residence time of nitrate in the 14-m deep vadose zone to be approximately 3.2 years, which is slightly larger than our model results of 2.6 years. The root zone MB approach yields total N mass in the deep vadose of 261 kg N ha-1 for the standard subplot and 478 kg N ha-1 for the high subplot, which is slightly smaller than our numerical modeling results.

6.2. Two-Dimensional vs. Three-Dimensional Simulation Domains

In these simulations, extending model dimensionality into third dimension did not prove to have much effect on model results except for HetVG-3D simulation which has noticeable higher N uptake (Figure 8) and lower N flux to the water table (Figure 10) than corresponding 2-D simulation. These differences could be explained by the fact that adding a third dimension with added heterogeneity to HetVG-2D simulation in lateral horizontal direction raises the probability of N mass being retained in a low conductivity zone; increasing chances of N uptake and decreasing cumulative N mass reaching water table with time.


12

6.3. Measured vs. Model Predicted Deep Vadose Zone N Mass Both, the root zone MB method and the numerical models severely overestimate the

measured N mass in the deep vadose zone. Apparently, under the highly heterogeneous geology of the alluvial sediments observed at the orchard site (Onsoy et al., 2005), neither the root zone MB approach with the commonly used assumption of uniform flow in the deep vadose zone nor the numerical layered models with uniform properties, or with local subsurface heterogeneity is adequate for explaining the observed fate of nitrate in the 14-m deep vadose zone.

From our field investigations, low N mass encountered in the vadose zone at the site was attributed to non-uniform flow conditions. Based on our close examination of the site subsurface textures, it was evident that wetting front instability might be present at the site (de Rooij, 2000). This wetting front instability had been shown by other field studies to create unstable, preferential flows in a variety of settings in both uniform (Glass et al., 1991; Wang et al., 1998, Wang et al., 2003a) and layered soil systems (Flury et al., 1994; Steenhuis et al., 1998; Wang et al., 2003b). Furthermore, extrinsic factors such as irrigation management and fertilizer applications were found to contribute much to variability in water fluxes and nitrate concentrations (Böhlke, 2002; Wang et al., 2004). Ritsema et al., 1998 estimated more than 80% of the infiltrating water was transported in preferential flow paths to the deep subsoil during single rain events. Field evidence also supported the existence of preferential flow due to repeated infiltration cycles (Glass et al., 1989) and the occurrence of fingers recurring in the same locations during successive irrigation events in a water repellent sandy field soil (Ritsema et al., 1998). All these conditions for preferential flow are present in our site; hence, preferential flow beyond that caused by heterogeneity in the soil hydraulic properties captured in Richards’ equation are a likely reason for the discrepancy between measured and simulated stored N mass.

We hypothesized that modeling small-scale heterogeneity leads to sufficient preferential flow that facilitates nitrate movement into groundwater thus significantly reducing the mass of N in the vadose zone. Yet, velocity variations in the heterogeneous models were only slightly larger than in the homogeneous model (Figures 11 and 12). Nitrate distribution patterns did not show strong preferential flow paths in the heterogeneous models (Figures 13 and 14).

The fact that none of the different model conceptualizations was able to simulate expected preferential flow could be attributed to the nature of Richards’ equation which is parabolic (diffusive), under conditions of strong capillarity and weak gravity (low velocity), and hyperbolic under conditions of infiltration (gravity/pressure gradient dominated convective flux) (Vogel and Ippisch, 2008). Under the latter conditions, Richards equation has numerical problems and grid discretization of an order less than 1 cm might be needed in sand layers to obtain solutions with reasonable accuracy (Vogel and Ippisch, 2008). Such resolution was not practical or even possible at the scale of the deep vadose zone at our site.

7. CONCLUSIONS

This study provides a detailed long-term, multi-year transient simulation of nitrate transport in a deep vadose zone. It is based on the Kearney field site, which offers a rich database for detailed geologic, hydraulic, and chemical characterization of the deep vadose zone stratigraphy that is typical for alluvial sediments in the eastern San Joaquin Valley, California. The site database provides a foundation for the development and validation of alternative modeling tools to assess the potential for nitrate leaching to groundwater in the presence of a deep, heterogeneous vadose zone at the site.


13

In conclusion, both the simple MB model and the numerical models used in this study were limited in their ability to estimate the low nitrate mass in the deep vadose zone. Preferential flow is believed to exist at the Kearney site. Overestimated dispersion was found to counterbalance the strong heterogeneity in the advective transport of nitrate, effectively preventing preferential transport to be simulated accurately. Yet, our hypothesis that a full accounting of heterogeneity in the simulation of flow based on Richards equation would explain the low nitrogen mass stored in the vadose zone, has been disproven. This is consistent with de Rooij (2000), Simunek et al. (2003), and Gardenas et al. (2006) who showed that Richards equation may be an inadequate model as it generally leads to relatively uniform flow and transport behavior. Much smaller grid resolutions may be needed or – alternatively - incorporating other conceptual models such as dual porosity or mobile-immobile flow domains may need to be included in the model to yield results that account for strong non-equilibrium preferential flow and transport. There was initially not enough evidence in our field measurements that supports using these types of models. Other conceptual elements to test could have included denitrification impacting more stagnant zones of water flow but not as much in preferential flow paths, nitrogen losses due to volatilization at the land surface, and a re-examination of the orchard nitrogen budget, particularly the N cycling through the leaf mass.

Regardless of shortcomings in explaining the low nitrogen mass in deep vadose zone, this paper confirms and illustrates these findings:

• The transient behavior of precipitation, irrigation, and evapotranspiration at the land surface affects not only the root zone, but rapidly affects moisture, suction, and water flux throughout a 16 m thick unsaturated zone. All of these are shown to be highly transient, even at the water table, under the irrigated, semi-arid conditions investigated here and typical for agricultural regions in semi-arid climates.

• Low irrigation efficiencies (on the order of 45% - 65%) contribute not only to significant leaching of fertilizer, but also to rapid transport of nitrate to groundwater. At the field site, irrigated at relatively low irrigation efficiencies, the travel time to groundwater through the 16 m thick vadose zone is predicted (based on the three models) to be as short as 2.5 – 3.5 years. By the same token, higher irrigation efficiencies would result in significantly longer travel times. It remains to be seen to which degree preferential flow or fingering of flow through heterogeneous alluvial systems counteracts such increases in solute travel times through the vadose zone.

• In situations where Richards equations is to be used under strong infiltration conditions using relatively large grid blocks, relatively uniform flow and transport behavior is expected regardless of level of heterogeneity incorporated.


14

REFERENCES Allaire-Leung, S. E., L. Wu, J. P. Mitchell and B. L. Sanden, 2001. Nitrate leaching and soil

nitrate content as affected by irrigation nonuniformity in a carrot field, Agric. Water Manage. 48, pp. 37–50

Allen, R. G., L. S. Pereira, D. Raes, and M. Smith, 1998. FAO Irrigation and Drainage Paper.

No. 56. Crop evapotranspiration. FAO of the United Nations, 300 pp. Baran, N., J. Richert, and C. Mouvet, 2007. Field data and modeling of water and nitrate

movement through deep unsaturated loess, J. Hydrol., 345, PP 27-37. Bear, J., 1972. Dynamics of Fluid in Porous Media. Elsevier, New York, 764 pp. Biggar, J. W., and D. R. Nielsen, 1978. Field monitoring of soil water constituents in the

unsaturated zone. In: L. G. Everett and K. D. Schmidt (eds.) Establishment of Water Quality and Monitoring Programs. Proc. Am. Water Resources Assoc., San Francisco, Ca. 12-14 June 1978. Tech. Publ. Ser. No. TPS79-1. Am. Water Resour. Assoc., Minneapolis, MN, p. 106-121.

Böhlke, J. -K., 2002. Groundwater recharge and agricultural contamination. Hyrogeology J., 10:

153-179. Bonilla, C. A., J. F. Munoz, and M. Vauclin, 1999. Opus simulation of water dynamics and

nitrate transport in a field plot. Ecol. Model., 122: 69-80. Botros, F. E., Th. Harter, Y. S. Onsoy, M. Denton, A. Tuli, and J. Hopmans, 2009. Spatial

variability of hydraulic properties and sediment characteristics in a deep alluvial unsaturated zone. Vadose Zone Journal, 8, 2: 276:289.

Braud, I., A. C. Dantas-Antonio, and M. Vauclin, 1995. A stochastic approach to studying the

influence of the spatial variability of soil hydraulic properties on surface fluxes, temperature, and humidity. J. Hydrol., 165(1-4): 283-310.

Brooks, R. H., and A. T. Corey, 1964. Hydraulic properties of porous media. Hydrol. Pap. 3.

Colo. State Univ., Fort Collins, 27 pp. Celia, M. A., E. T. Bouloutas, and R. L. Zarba, 1990. A general mass-conservative numerical

solution for the unsaturated flow equation. Water Resour. Res., 26(7): 1483-1496. Dalton, F. N., P. A. C. Raats, and W. R. Gardner, 1975. Simultaneous uptake of water and

solutes by plant roots. Agron. J., 67: 334-339. de Rooij, G. H., 2000. Modeling fingered flow of water in soils owing to wetting front stability: a

review.J. Hydrol., 231-232: 277-294.


15

de Vos, J. A., D. Hesterberg and P. A. C. Raats, 2000. Nitrate leaching in a tiled-drained silt loam soil, Soil. Sci. Soc. Am. J. 64, pp. 517–527

Doorenbos, J. and W. O. Pruitt, 1977. Crop water requirements. Rev. 1977. FAO Irrig. And

Drain. Paper 24, FAO of the United Nations, Rome, 144 pp. DWR. 2001. Crop water use. San Joaquin District Report, CA Department of Water Resources. Eching, S. O., and J. W. Hopmans, 1993. Optimization of hydraulic functions from transient

outflow and soil water pressure head data, Soil Sci. Soc. Amer. J., 57: 1167-1175. Feddes, R. A., P. J. Kowalik, and H. Zaradny, 1978. Simulation of Field Water Use and Crop

Yield. Simulation Monographs, Pudoc, Wageningen, The Netherlands, 188 pp. Flury, M., H. Flühler, W. A. Jury, and J. Leuenberger, 1994. Susseptibility of soils to preferential

flow of water: A field study. Water Resour. Res., 30(7): 1945-1954. Gardner, W. R., 1958. Some steady state solutions of unsaturated moisture flow equations with

application to evaporation froma water table. Soil Sci., 85(4): 228-232. Glass, R. J., T. S. Steenhuis, and J. -Y. Parlange, 1989. Mechanism for finger persistence in

homogeneous, unsaturated, porous media: theory and verification. Soil Sci., 148: 60-70. Glass, R. J., J. -Y. Parlange, and T. S. Steenhuis, 1991. Immesible displacement in pororus

media: stability analysis of three-dimensional, axisymmetric disturbances with application to gravity-drven wetting front instability. Water Resour. Res., 27(8): 1947-1956.

Harter, T., and T.-C. J. Yeh, 1996. Stochastic analysis of solute transport in heterogeneous,

variably saturated soils. Water Resour. Res., 32(6): 1585-1595. Harter, T., and T. C. J. Yeh. 1998, Flow in unsaturated random porous media nonlinear

numerical analysis and comparison to analytical stochastic models. Advances in Water Resources. 22 (2): 257-272.

Harter, T., Y. S. Onsoy, K. Heeren, M. Denton, G. Weissmann, J. W. Hopmans, W. R. Horwath,

2005. Deep vadose zone hydrology demonstrates fate of nitrate in eastern San Joaquin Valley, California Agriculture 59(2):124-132.

Haverkamp R, M. Vaclin, J. Touma, P. J. Wierenga, and G. Vachaud, 1977. A comparison of

numerical simulation models for one-dimensional infiltration. Soil Sci. Soc. Am. J., 41: 285-294.

Hills, R. G., P. J. Wierenga, D. B. Hudson, and M. R. Kirkland, 1991. The second Las Cruces

trench experiment: Experimental results and two-dimensional flow predictions. Water Resour. Res., 27(10): 2707-2718.


16

Hopmans, J. W., H. Schukking, P. J. J. F. Torfs, 1988. Two-dimensional steady state unsaturated

water flow in heterogeneous soils with autocorrelated soil hydraulic properties. Water Resour. Res., 24(4): 2005-2017.

Johnson, R.S., F.G. Mitchell, and C.H. Crisosto, 1995. Nitrogen fertilization of Fantasia

nectarine – A 12 year study. UC Kearney Tree Fruit Review Vol 1, 14-19. Jury, W. A., D. Russo, G. Sposito, and H. Elabd, 1987. The spatial variability of water and solute

transport properties in unsaturated soil: II. Scaling models of water transport. Hilgardia, 55: 33-56.

Klute, A (ed.), 1986. Methods of Soil Analysis, Part 1: Physical and Mineralogical Methods, 2nd

Edition, American Society of Agronomy, Soil Science Society of America, Madison, WI, 1188p.

Kung, K-J. S. 1993. Laboratory observation of the funnel flow mechanism and its influence on solute transport. J. Environ. Qual. 22:91-102. Lafolie, F., L. Bruckler, and A. M. de Corkborne, 1997. Modeling the water transport and

nitrogen dynamics in irrigated salad crops. Irrig. Sci., 17: 95-104. Ling, G., and A.I. El-Kadi. 1998. A lumped parameter model for nitrogen transformation in the

unsaturated zone, Water Resour. Res., 34(2), 203-212 Lu, Z., D. Zhang, and B. Robinson, 2007. Analytical solutions for one-dimensional steady state

flow in layered heterogeneous unsaturated soils under uncertainties, Water Resour. Res., 43, W09413, doi:10.1029/2005WR004795,

Miller, E. E., and R. D. Miller, 1956. Physical theory for capillary flow phenomena. J. Appli.

Phys., 27(4): 324-332. Mualem, Y., 1976. A new model for predicting the hydraulic conductivity of unsaturated porous

media. Water Resour. Res., 12:513-522. Nimmo, J.R., J.A. Deason, J.A. Izbicki, and Peter Martin, 2002. Evaluation of unsaturated zone

water fluxes in heterogeneous alluvium at a Mojave Basin Site. Water Resour. Res., 38(10), 1215, doi:10.1029/2001WR000735.

Onsoy, Y. S., 2005. Modeling Nitrate Transport in Deep Unsaturated Alluvial Sediments and

Assessing Impact of Agricultural Management Practices on Groundwater Quality. Ph.D. Dissertation, University of California, Davis.

Onsoy, Y. S., T. Harter, T. Ginn, and W. Horwath, 2005. Spatial variability and transport of

nitrate in a deep alluvial vadose zone, Vadose Zone J., 4:41-51.


17

Rao, P. S. C., and R. J. Wagenet, 1985. Spatial variability of pesticides in field soils: methods for data analysis and consequences. Weed Sci., 33(2): 18-24.

Ritsema, C. J., L. W. Dekker, J. L. Nieber, T. S. Steenhuis, 1998. Modeling and field evidence of

finger formation and finger recurrence in a water repellent sandy soil. Water Resour. Res., 34(4): 555-567.

Rockhold, M., R. Rossi, and R. Hills, 1996. Application of similar media scaling and conditional

simulation for modeling water flow and Tritium transport at the Las Cruces Trench Site, Water Resour. Res., 32(3), 595-609.

Rockhold, M. L., C. S. Simmons, and M. J. Fayer, 1997. An analytical solution technique for

one-dimensional, steady vertical water flow in layered soils. Water Resour. Res., 33(4): 892-902.

Roth, K., W. A. Jury, H. Fluhler, and W. Attinger, 1991. Transport of chloride through a vadose

field soil. Water Resour. Res., 27(10): 2533-2541. Russo, D., 1988. Determining soil hydraulic properties by parameter estimation: On the selection

of a model for the hydraulic properties. Water Resour. Res., 24: 453-459. Russo, D., 1991. Stochastic analysis of simulated vadose zone solute transport in a vertical cross

section of heterogeneous soil during nonsteady water flow. Water Resour. Res., 27(3): 267-283.

Scheidegger, A.L., 1960. The Physics of Flow through Porous Media. University of Toronto

Press. Toronto. Seong, K., and Y. Rubin, 1999. Field investigation of the waste isolation pilot plant (WIPP) site

(New Mexico) using a nonstationary stochastic model with a trending hydraulic conductivity field, Water Resour. Res., 35(4), 1011–1018.

Sheldrick, B. H., C. Wang, 1993. Particle size distribution. In: Carter, M.R. (Ed.), Soil Sampling

and Methods of Soil Analysis. Lewis Publishers, Boca Raton, FL, USA, pp. 499–512. Simunek, J., T. Vogel, and M. Th.van Genuchten, 1996. HYDRUS-2D Code for Simulating

Water Flow and Solute Transport in Two-Dimensional Variably Saturated Media. Version 1.0. US Salinity Laboratory USDA/ARS, Riverside, CA. and IGWMC-TPS 53. Golden, Colorado School of Mines

Simunek, J., N. J. Jarvis, M. Th. Van Genuchten, A. Gardenas, 2003. Review and comparison of

models for describing non-equilibrium and preferential flow and transport in the vadose zone. J. Hydrol. 272: 14-35.

Steenhuis, T., K. Vandelheuvel, K.W. Weiler, J. Boll, J. Daliparthy, S. Herbert, and K. -J. S.

Kung, 1998. Mapping and interpreting soil textural layers to assess agri-chemical


18

movement at several scales along the eastern seabord (USA). Nutrient Cycling in Agroecosystems, 50: 91-97.

Stenger, R., E. Priesack, and F. Beese, 2002. Spatial variation of nitrate-N and related soil

properties at the plot scale. Geoderma 105:259–275. Tartakovsky, D. M., S. P. Neuman, and Z. Lu, 1999. Conditional stochastic averaging of steady-

state unsaturated flow by means of Kirchhoff transformation. Water Resour. Res., 35(3): 731-745.

Tompson, A. F. B. and L. W. Gelhar, 1990. Numerical simulation of solute transport in three-

dimensional randomly heterogeneous porous media. Water Resour. Res., 26(10): 2541-2562.

Troiano, J., C. Garretson, C. Krauter, J. Brownell, and J. Huston, 1993. Influence of amount and

method of irrigation water application on leaching of atrazine. J. Environ. Qual., 22(2): 290-298.

Tseng, P. -H., and W. A. Jury, 1994. Comparison of transfer function and deterministic modeling

of area-averaged solute transport in a heterogeneous field. Water Resour. Res., 30(7): 2051-2064.

Ünlü K., D. R. Nielsen, and J. W. Biggar, 1990. Stochastic analysis of vadose flow: One-

dimensional Monte Carlo simulations and comparisons with spectral perturbation analysis and field observations. Water Resour. Res., 26(9): 2207-2218.

Ünlü, K., G. Ozenirler, and C. Yurteri, 1999. Nitrogen fertilizer leaching from cropped and

irrigated sady soil in Central Turkey. Eur. J. Soil Sci., 50: 609-620. Vachaud, G., M. Vauclin, and P. Balabanis, 1988. Stochastic approach of soil water flow through

the use of scaling factors: Measurement and simulation. Agric. Water Manag., 13: 249-261.

van Genuchten, M. TH., 1980. A closed form equation for predicting the hydraulic conductivity

of unsaturated soils. Soil Sci. Soc. Am. J. 44:892-898. Vogel, H.-J., and O. Ippisch, 2008. Estimation of a critical spatial discretization limit for solving

Richards’ Equation at large scales. Vadose Zone Journal, 7, 1: 112:114. Vogeler, I., S. R. Green, D. R. Scotter, and B. E. Clothier, 2001. Measuring and modeling the

transport and root uptake of chemicals in the unsaturated zone. Plant Soil, 231: 161-174. Wagenet, R. J., and J. L. Hutson, 1989. LEACHM: A process-based Model for Water and Solute

Movements, Transformations, Plant Uptake and Chemical Reactions in the Unsaturated Zone, Version 2.0., Vol. 2., New York State Water Resources Institute, Cornell University, Ithaca, NY.


19

Wang, Z., J. Feyen, and C. J. Ritsema, 1998. Susceptibility and predictability of conditions for

preferential flow. Water Resour. Res., 34(9): 2169-2182. Wang, Z., A. Tuli, and W. A. Jury, 2003a. Unstable flow during distribution in homogeneous

soil. Vadose Zone J., 2: 52-60. Wang, Z., L. Wu, T. Harter, J. Lu, and W. A. Jury, 2003b. A field study of unstable preferential

flow during soil water redistribution. Water Resour. Res., 39(4), 1-1 – 1-11. Wang, Z., W.A. Jury, A. Tuli, and D. Kim, 2004. Unstable flow during redistribution:

Controlling factors and practical implications. Soil Sci. Soc Am. J., 3: 549-559. Warrick, A. W., G. J. Mullen, D. R. Nielsen, 1977. Scaling field measured soil hydraulic

properties using a similar media concept. Water Resour. Res., 13(2): 355-362. Weinbaum, S. A., R. S. Johnson, and T. M. DeJong, 1992. Causes and consequences of

overfertilization in orchards. Proceedings of the Workshop Fertilizer Management in Horticultural Crops: Implications for Water Pollution Hort Technology, 2(1): 112-119.

Zavattaro, L., N. Jarvis, and L. Persson, 1999. Use of similar media scaling to characterize

spatial dependence of near-saturated hydraulic conductivity. Soil Sci. Soc Am. J., 63: 486-492.

Zhu, J. and B. P. Mohanty, 2002a. Analytical solutions for steady state vertical infiltration.

Water Resour. Res., 38(8), 1145, doi:10.1029/2001WR000398.


20

Table 1 Records of irrigation applications at the fertilizer subplots from 1990 to 1996. Variations reflect climatic variations with more irrigations following dry winters (e.g., 1992) and irrigation management decisions.

Irrigation # 1990 1991 1992 1993 1994 1995 1996

1 29-Mar 7-Feb 22-Apr 19-Mar 21-Mar 1-May 1-May

2 17-Apr 3-May 29-Apr 16-Apr 14-Apr 10-May 9-May

3 8-May 22-May 7-May 5-May 10-May 22-May 21-May

4 4-Jun 4-Jun 18-May 18-May 23-May 31-May 3-Jun

5 14-Jun 20-Jun 28-May 1-Jun 1-Jun 8-Jun 25-Jun

6 21-Jun 1-Jul 3-Jun 9-Jun 9-Jun 20-Jun 2-Jul

7 27-Jun 15-Jul 9-Jun 18-Jun 16-Jun 3-Jul 30-Jul

8 3-Jul 30-Jul 18-Jun 24-Jun 23-Jun 10-Jul 11-Sep

9 10-Jul 8-Aug 25-Jun 1-Jul 30-Jun 25-Jul 3-Oct

10 24-Jul 6-Sep 1-Jul 9-Jul 16-Jul 7-Aug

11 9-Aug 8-Jul 19-Jul 28-Jul 23-Aug

12 5-Sep 23-Jul 30-Jul 8-Aug 6-Sep

13 18-Sep 6-Aug 9-Aug 16-Sep 4-Oct

14 20-Aug 24-Aug

15 31-Aug 2-Sep

16 10-Sep 14-Sep

17 28-Sep


21

Table 2 Records of fertilizer applications at the fertilizer subplots from 1990 to 1996. N concentrations are evaluated based on the amount of surface water flux (either irrigation, I, precipitation, P, or both, I+P) at the orchard.

Year Dates of Fertilizer Application Subplot Applied N

(kg ha-1) Application

Method

28-Mar H 85 I

7-May H 85 I

4-Jun H 85 I 1990

17-Sep S/H 110 I

22-Mar / 23Mar H 85 R

2-May H 85 I

3-Jun H 85 I + R 1991

5-Sep S/H 110 I

20-Mar H 85 R

29-Apr H 85 I

28-May H 85 I 1992

9-Sep S/H 110 I

17-Mar H 85 I

3-May H 85 I

1-Jun H 85 I 1993

13-Sep S/H 110 I

21-Mar H 85 I

10-May H 85 I

1-Jun H 85 I 1994

16-Sep S/H 110 I

1996 11-Sep S/H 110 I

1S/H: Fertilizer applied both on the standard and high subplots; H: fertilizer applied on the high subplot.

2 I: Fertilizer applied by irrigation water; R: fertilizer applied by rainfall; I+R: fertilizer by irrigation and rainfall.

3 Fertilizer applied by rainfall in two consecutive days


22

West East

Figure 1 a) East-west lithofacies cross section identified at the fertilizer experimental site b) Typical cross section that is used in the numerical model.


23

-250 -200 -150 -100 -50 0

-1500

-1000

-500

0

P2

SL3

C-T-L

S2

SL2

P1S1

SL1

Initial heads (cm)

Dep

th (c

m)

Figure 2 Initial condition pressure profile established for the eight-layered soil domain under a steady flux of 0.23 cm d-1.


24

-5 -4 -3 -2 -1 0 1 2 30

5

10

15

20

25

30

35

40

45

ln (SF)

Freq

uenc

y

Figure 3 Histogram of log transformed of scaling factors. It shows near normal distribution.

0 5 10 15 20 25 30 350

0.1

0.2

0.3

0.4

0.5

0.6

lag distance (h)

γ ( h

)

0 0.5 1 1.5 2 2.5 3 3.50

0.1

0.2

0.3

0.4

0.5

0.6

0.7

lag distance (h)

Figure 4 semivariograms of ln (λ) in the horizontal and vertical directions.


25

Figure 5 distribution of ln λ in 2-D and 3-D model domains.


26

Log (Ks) Log(α) log(n)

Figure 6 distribution of log Ks, log α, and log n in all model conceptualizations. Top row: 2-D Simulations. Bottom Row: 3-D Simulations. Left to Right: HM, HetSF, and HetVG.


27

0 500 1000 1500 2000 2500 30000

100

200

300

400

500

600

700

800

Time (days)

Cum

. Trn

spira

tion

(cm

)

Potential2D Homogeneous2D Heterogeneous SF2D Heterogeneous VG3D Homogeneous3D Heterogeneous SF3D Heterogeneous VG

Figure 7 Comparison of daily values of potential and model simulated (actual) transpiration during the simulation period (1990 – 1996) in all models.


28

0 500 1000 1500 2000 2500 30000

100

200

300

400

500

600C

um. N

Upt

ake

(Kg

ha-1

)

Control Subplot

2D Homogeneous2D Heterogeneous SF2D Heterogeneous VG3D Homogeneous3D Heterogeneous SF3D Heterogeneous VG

0 500 1000 1500 2000 2500 30000

200

400

600

800

1000

Time (days)

Cum

. N U

ptak

e (K

g ha

-1)

High Subplot

Figure 8 Cumulative plant N uptake simulated at the standard and high subplots during the simulation period (1990 – 1996) in the all models.


29

0 500 1000 1500 2000 2500 30000

100

200

300

400

500

600

700

Time (days)

Cum

. Flu

x to

GW

(cm

)


Figure 9 Cumulative simulated water flux to the water table.


30

0 500 1000 1500 2000 2500 30000

50

100

150

200

250

300

350C

um. N

Flu

x to

GW

(Kg

ha-1

)

Control Subplot


0 500 1000 1500 2000 2500 30000

200

400

600

800

1000

Time (days)

Cum

. N F

lux

to G

W (K

g ha

-1)

High Subplot

Figure 10 Cumulative NO3-N flux at the water table of the standard and high subplots during the simulation period (1990 – 1996) in the homogeneous and heterogeneous lithofacies models


31

Figure 11 Velocity distributions in the deep vadose zone at the winter of year 1993, three months after the last irrigation application. Top row: 2-D Simulations. Bottom Row: 3-D Simulations. Left to Right: HM, HetSF, and HetVG.


32

Figure 12 Velocity distributions in the deep vadose zone in the summer of year 1994, middle of irrigation season. Top row: 2-D Simulations. Bottom Row: 3-D Simulations. Left to Right: HM, HetSF, and HetVG.


33

Figure 13 log Concentration for “standard” N treatment in the deep vadose zone at the winter of year 1993, three months after the last irrigation application. Top row: 2-D Simulations. Bottom Row: 3-D Simulations. Left to Right: HM, HetSF, and HetVG.


34

Figure 14 log Concentration for “standard” N treatment in the deep vadose zone at the summer of year 1994, in the middle of irrigation season. Top row: 2-D Simulations. Bottom Row: 3-D Simulations. Left to Right: HM, HetSF, and HetVG.


35

0 500 1000 1500 2000 2500 30000

50

100

150

200

250

300

350

400St

ored

N in

Dee

p VZ

(Kg

ha-1

)

Control Subplot


0 500 1000 1500 2000 2500 30000

200

400

600

800

1000

1200

Time (days)

Stor

ed N

in D

eep

VZ (K

g ha

-1)

High Subplot

Figure 15 Total NO3-N mass simulated in the deep vadose zone of the standard and high subplots during the period from 1990 to 1996 in the homogeneous and heterogeneous lithofacies models


Appendix 3: Soil survey derived geologic and geomorphic map of the surficial alluvial deposits in the Central Valley. The oldest geologic sediments within the mapping area are found in the “Foothills and Uplands” region. On the floor of the Central Valley, Laguna and Turlock Lake formations, and the Mehrten formation shape the older alluvial terraces that surround the incised alluvial fans of the newer Riverbank and Modesto formations. The most recent deposits (less than 10,000 years) are the Holocene deposits along the trough of the Central Valley. These are associated with the major streams and lakes in the Central Valley (Sacramento River, San Joaquin River, Tulare Lake, and others).



Documents

Nonpoint Source Pollutant Transfer across Deep Vadose ...kearney.ucdavis.edu/NEW MISSION-LIVE... · properties throughout a 16-m-deep, typical alluvial vadose zone consisting of unconsolidated,