Upload
others
View
7
Download
0
Embed Size (px)
Citation preview
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
Nonpoint Source Pollutant Transfer across Deep Vadose Zones – A Multiscale Investigation to Inform Regulatory Monitoring, Assessment, and Decision Making—Harter
14
Nonpoint Source Pollutant Transfer across Deep Vadose Zones – A Multiscale Investigation to Inform Regulatory Monitoring, Assessment, and Decision Making—Harter
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.,
Nonpoint Source Pollutant Transfer across Deep Vadose Zones – A Multiscale Investigation to Inform Regulatory Monitoring, Assessment, and Decision Making—Harter
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
Nonpoint Source Pollutant Transfer across Deep Vadose Zones – A Multiscale Investigation to Inform Regulatory Monitoring, Assessment, and Decision Making—Harter
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.
Nonpoint Source Pollutant Transfer across Deep Vadose Zones – A Multiscale Investigation to Inform Regulatory Monitoring, Assessment, and Decision Making—Harter
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)
Nonpoint Source Pollutant Transfer across Deep Vadose Zones – A Multiscale Investigation to Inform Regulatory Monitoring, Assessment, and Decision Making—Harter
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).
Nonpoint Source Pollutant Transfer across Deep Vadose Zones – A Multiscale Investigation to Inform Regulatory Monitoring, Assessment, and Decision Making—Harter
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).
Nonpoint Source Pollutant Transfer across Deep Vadose Zones – A Multiscale Investigation to Inform Regulatory Monitoring, Assessment, and Decision Making—Harter
9
Nonpoint Source Pollutant Transfer across Deep Vadose Zones – A Multiscale Investigation to Inform Regulatory Monitoring, Assessment, and Decision Making—Harter
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.
Nonpoint Source Pollutant Transfer across Deep Vadose Zones – A Multiscale Investigation to Inform Regulatory Monitoring, Assessment, and Decision Making—Harter
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.
Nonpoint Source Pollutant Transfer across Deep Vadose Zones – A Multiscale Investigation to Inform Regulatory Monitoring, Assessment, and Decision Making—Harter
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.
Nonpoint Source Pollutant Transfer across Deep Vadose Zones – A Multiscale Investigation to Inform Regulatory Monitoring, Assessment, and Decision Making—Harter
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.
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
www.vadosezonejournal.org · Vol. 8, No. 2, May 2009 277
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.
Harter et al., 2010, Final Report 17 Univ. of California Kearney Foundation
www.vadosezonejournal.org · Vol. 8, No. 2, May 2009 278
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.
Harter et al., 2010, Final Report 18 Univ. of California Kearney Foundation
www.vadosezonejournal.org · Vol. 8, No. 2, May 2009 279
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
Harter et al., 2010, Final Report 19 Univ. of California Kearney Foundation
www.vadosezonejournal.org · Vol. 8, No. 2, May 2009 280
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.
Harter et al., 2010, Final Report 20 Univ. of California Kearney Foundation
www.vadosezonejournal.org · Vol. 8, No. 2, May 2009 281
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).
Harter et al., 2010, Final Report 21 Univ. of California Kearney Foundation
www.vadosezonejournal.org · Vol. 8, No. 2, May 2009 282
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.
Harter et al., 2010, Final Report 22 Univ. of California Kearney Foundation
www.vadosezonejournal.org · Vol. 8, No. 2, May 2009 283
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).
Harter et al., 2010, Final Report 23 Univ. of California Kearney Foundation
www.vadosezonejournal.org · Vol. 8, No. 2, May 2009 284
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).
Harter et al., 2010, Final Report 24 Univ. of California Kearney Foundation
www.vadosezonejournal.org · Vol. 8, No. 2, May 2009 285
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].
Harter et al., 2010, Final Report 25 Univ. of California Kearney Foundation
www.vadosezonejournal.org · Vol. 8, No. 2, May 2009 286
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.
Harter et al., 2010, Final Report 26 Univ. of California Kearney Foundation
www.vadosezonejournal.org · Vol. 8, No. 2, May 2009 287
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.
Harter et al., 2010, Final Report 27 Univ. of California Kearney Foundation
www.vadosezonejournal.org · Vol. 8, No. 2, May 2009 288
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.
Harter et al., 2010, Final Report 28 Univ. of California Kearney Foundation
www.vadosezonejournal.org · Vol. 8, No. 2, May 2009 289
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.
Harter et al., 2010, Final Report 29 Univ. of California Kearney Foundation
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.
Harter et al., 2010, Final Report 30 Univ. of California Kearney Foundation
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
Harter et al., 2010, Final Report 31 Univ. of California Kearney Foundation
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
Harter et al., 2010, Final Report 32 Univ. of California Kearney Foundation
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
Harter et al., 2010, Final Report 33 Univ. of California Kearney Foundation
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.
Harter et al., 2010, Final Report 34 Univ. of California Kearney Foundation
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
Harter et al., 2010, Final Report 35 Univ. of California Kearney Foundation
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).
Harter et al., 2010, Final Report 36 Univ. of California Kearney Foundation
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)
Harter et al., 2010, Final Report 37 Univ. of California Kearney Foundation
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
Harter et al., 2010, Final Report 38 Univ. of California Kearney Foundation
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
Harter et al., 2010, Final Report 39 Univ. of California Kearney Foundation
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
Harter et al., 2010, Final Report 40 Univ. of California Kearney Foundation
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.
Harter et al., 2010, Final Report 41 Univ. of California Kearney Foundation
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.
Harter et al., 2010, Final Report 42 Univ. of California Kearney Foundation
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.
Harter et al., 2010, Final Report 43 Univ. of California Kearney Foundation
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.
Harter et al., 2010, Final Report 44 Univ. of California Kearney Foundation
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.
Harter et al., 2010, Final Report 45 Univ. of California Kearney Foundation
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.
Harter et al., 2010, Final Report 46 Univ. of California Kearney Foundation
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
Harter et al., 2010, Final Report 47 Univ. of California Kearney Foundation
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.
Harter et al., 2010, Final Report 48 Univ. of California Kearney Foundation
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.
Harter et al., 2010, Final Report 49 Univ. of California Kearney Foundation
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
Harter et al., 2010, Final Report 50 Univ. of California Kearney Foundation
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
Harter et al., 2010, Final Report 51 Univ. of California Kearney Foundation
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.
Harter et al., 2010, Final Report 52 Univ. of California Kearney Foundation
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.
Harter et al., 2010, Final Report 53 Univ. of California Kearney Foundation
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.
Harter et al., 2010, Final Report 54 Univ. of California Kearney Foundation
25
Figure 5 distribution of ln λ in 2-D and 3-D model domains.
Harter et al., 2010, Final Report 55 Univ. of California Kearney Foundation
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.
Harter et al., 2010, Final Report 56 Univ. of California Kearney Foundation
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.
Harter et al., 2010, Final Report 57 Univ. of California Kearney Foundation
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.
Harter et al., 2010, Final Report 58 Univ. of California Kearney Foundation
29
0 500 1000 1500 2000 2500 30000
100
200
300
400
500
600
700
Time (days)
Cum
. Flu
x to
GW
(cm
)
2D Homogeneous2D Heterogeneous SF2D Heterogeneous VG3D Homogeneous3D Heterogeneous SF3D Heterogeneous VG
Figure 9 Cumulative simulated water flux to the water table.
Harter et al., 2010, Final Report 59 Univ. of California Kearney Foundation
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
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 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
Harter et al., 2010, Final Report 60 Univ. of California Kearney Foundation
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.
Harter et al., 2010, Final Report 61 Univ. of California Kearney Foundation
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.
Harter et al., 2010, Final Report 62 Univ. of California Kearney Foundation
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.
Harter et al., 2010, Final Report 63 Univ. of California Kearney Foundation
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.
Harter et al., 2010, Final Report 64 Univ. of California Kearney Foundation
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
2D Homogeneous2D Heterogeneous SF2D Heterogeneous VG3D Homogeneous3D Heterogeneous SF3D Heterogeneous VG
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
Harter et al., 2010, Final Report 65 Univ. of California Kearney Foundation
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).
Harter et al., 2010, Final Report 66 Univ. of California Kearney Foundation
Harter et al., 2010, Final Report 67 Univ. of California Kearney Foundation