ASSESSING THE VEGETATION AND SOIL SHEAR TESTER (VASST)

AS A TOOL TO MEASURE VEGETATED SOIL SHEAR STRENGTH

BY

CASEY D. CAMPBELL

THESIS

Submitted in partial fulfillment of the requirements

for the degree of Master of Science in Agricultural and Biological Engineering
in the Graduate College of the
University of Illinois at Urbana-Champaign, 2014

Urbana, Illinois

Adviser:

Professor Prasanta Kalita

 ABSTRACT

The degradation of military training lands due to vehicular traffic presents a challenge for

land managers trying to optimize training capacity while reducing impacts. The Optimal

Allocation of Land for Training and Non-Training Uses (OPAL) Program aims to provide a

model for land managers to predict impacts of training under a range of land management

regimes utilizing soil, vegetation, and climatic data. In general, OPAL’s intended purpose is to

help land managers optimize training land carrying capacity.

As part of OPAL, the Vegetation And Soil Shear Tester (VASST) was developed as a

new method of in-situ measurement for shear strength of vegetated soils. During this research,

data on vegetation and soil parameters was collected concurrently with VASST measurements at

five geographic locations to determine if the VASST effectively measured soil shear strength.

Additionally, as part of this study, a Python program was developed to automate the process of

conditioning and analyzing VASST outputs to reduce human bias and dramatically decrease

processing times of raw data.

Evaluating the VASST against calculated soil shear strength as well as other common in-

situ strength measurements such as cone index, drop-cone value, and Clegg impact values,

confirmed that the VASST shear measurements were moderately to strongly correlated with one

another. An investigation into potential associations of soil parameters to strength measurements

obtained with the VASST was conducted. VASST measures were taken in high plasticity

(clayey) and non-plastic (sandy) soils where soil strengths were dictated by particle size,

aboveground biomass, and root weight. Results also indicated that soil moisture content

influenced VASST measured soil shear strength in high plasticity soils.

ii
 ACKNOWLEDGEMENTS

I would like to thank Heidi Howard of the United States Army Corps of Engineers

Engineer Research Development Center- Construction Engineering Research Laboratory

(ERDC-CERL) for giving me the opportunity to pursue a master’s degree and providing

guidance throughout my research. Also, I would like to thank Daniel Koch (ERDC-CERL) and

Andrew Fulton (ERDC-CERL) for providing me with valuable input and support during my

graduate studies and research. I also appreciate the many people within CERL and the Cold

Regions Research and Engineering Laboratory (CRREL) who provided direct assistance with

data collection, sample processing, as well as input on my research. In particular, I would like to

thank Kelly McDonald (CRREL), Taylor Leahy of the University of Illinois at Urbana-

Champaign (UIUC), Jeff Mifflin (CERL), Nick Wendling (UIUC), Paul Schumacher (UIUC),

and August Metzler (UIUC).

I thank Dr. Prasanta Kalita (UIUC) for serving as my advisor through my undergraduate

and graduate studies at UIUC. I greatly appreciate Dr. George Gertner (UIUC/CERL) who

provided me with advice on the statistical analysis of my research, as well as mentored me in

SAS programming procedures. I express my gratitude to Dr. Bhattarai (UIUC) for his feedback

on the presentation of my research and for serving on my committee. Thank you to Paul

Davidson (UIUC) for also serving on my committee.

Funding and support for this work was provided by ERDC-CERL under the Army 6.2

A896 program and as a part of the OPAL Work Package.

iii
 TABLE OF CONTENTS

CHAPTER 1: INTRODUCTION ............................................................................................. 1

CHAPTER 2: RESEARCH OBJECTIVES .............................................................................. 5

CHAPTER 3: LITERATURE REVIEW .................................................................................. 6

CHAPTER 4: MATERIALS AND METHODS .................................................................... 23

CHAPTER 5: COMPUTATIONAL APPROACH TO ANALYZING VASST DATA ........ 42

CHAPTER 6: RESULTS AND DISCUSSION ...................................................................... 60

CHAPTER 7: CONCLUSIONS ............................................................................................. 84

CHAPTER 8: RECOMMENDATIONS FOR FUTURE WORK .......................................... 87

REFERENCES ......................................................................................................................... 88

APPENDIX A: TEST AND SAMPLE QUANTITIES BY LOCATION ................................ 97

APPENDIX B: PYTHON CODE ........................................................................................... 101

APPENDIX C: PROC SGSCATTER RESULTS FOR CLASS VARIABLES ..................... 114

APPENDIX D: PROC SGPANEL RESULTS FOR CLASS VARIABLES .......................... 118

APPENDIX E: PROC CORR RESULTS............................................................................... 120

APPENDIX F: PROC LIFETEST RESULTS WITH BULK DENSITY .............................. 123

APPENDIX G: PROC LIFETEST RESULTS WITHOUT BULK DENSITY ...................... 129

APPENDIX H: ACRONYMS................................................................................................. 135

iv
CHAPTER 1: INTRODUCTION
1.1 Motivation

The Army’s mission requirements often include the use of heavy armored equipment,

some weighing up to 70 tons (Rostam-Abadi et al., 1993), on naturally vegetated soils within

training lands. Due to repeated traffic on training areas, land degradation is likely to occur. Land

degradation can lead to increased soil erosion, hazardous training conditions, decreased mobility

and realism, and costly land rehabilitation. It is important for the Army to understand the soil

mechanics leading to this type of land degradation in order to improve the quality of training

lands and optimize land rehabilitation (Guretzky et al., 2005).

Since the publishing of Army Regulation (AR) 210-21 in 1997, The United States Army

has been researching methods to reduce its environmental impact from repeated training

activities on training lands. The motivation for increasing environmental conservation is

addressed in AR350-19 which assigns responsibilities and provides policy and guidance for the

management and operation of ranges and training lands. AR350-19 supports U.S. Army long-

term sustainability and utility to meet the national defense mission with core programs such as

the Army’s Sustainable Range Program (SRP), Range and Training Lands Program (RTLP), and

Integrated Training Area Management (ITAM) Program. These programs enable the Army to

manage and maintain training lands to meet mission requirements while striving to minimize

negative effects of military activity on the environment (U.S. Department of the Army, 2005).

1
1.2 Background

The Mohr-Coulomb failure criterion is used throughout engineering practices to calculate

the shear strength of homogeneous materials. When dealing with engineered soil (non-

vegetated), Equation 1.1 describes soil shear strength.

𝜏 = 𝑐 + 𝜎 tan ϕ Eq.1.1

τ = shear strength

c = particle cohesion

σ = applied normal force

ϕ = the internal angle of friction

However, roots commonly present in vegetated soils have shown a measurable effect on soil

shear strength (Waldron, 1977; Abe and Ziemer 1991). Roots from vegetated soil are

unaccounted for in the Mohr-Coulomb failure criterion for homogenous materials.

One of the first studies on the shear strength of root permeated soil columns was

conducted by Endo and Tsuruta (1969). Endo’s research on the soil reinforcement of properties

of tree roots was done using a rigid box with a normal load applied at the soil surface and a

shearing load applied perpendicular to the normal load.

2
 Figure 1.1: Schematic of Endo and Tsuruta (1969) experimental apparatus (Endo, 1980)

Although the test was performed in-situ, it simulated a direct shear test commonly

conducted under laboratory conditions. The results of this study showed that roots in the soil

matrix increase soil shear strength. Later studies confirmed Endo’s findings of the reinforcing

effects of roots on soil (Waldron, 1977; Gray and Leiser, 1982; Gray and Sotir, 1996; Pollen and

Simon, 2005; Zhang, et al. 2010).

In 1977, L. J. Waldron proposed a modified Mohr-Coulomb failure criterion in order to

describe the effects of roots on soil shear strength. Results from Waldron’s testing showed that

soils containing roots can be analyzed as a composite material, with the roots acting as high

tensile strength fibers distributed in a lower tensile strength material (1977). The equation he

used to describe vegetated soil shear strength is Equation 1.2.

𝜏 = 𝑐 + ∆𝑠 + 𝜎 tan ϕ Eq. 1.2

The addition of ∆𝑠 to the basic equation accounted for the increase in shear strength due to the

reinforcing effect of roots and has been adopted as a common way of describing the shear

strength of vegetated soils.

Current research at the Engineering Research Development Center (ERDC) led to the

development of a device that tests soil strength by mimicking a direct shear test. The device,

known as the VASST, was designed and built as a tool to quantify soil mobility and

3
trafficability, with the reinforcing effects of vegetation present (MacDonald, et al. 2012). The

VASST is advantageous over previous methods due to its ability to perform relatively quick, in-

situ tests with minimal ground disturbance, while simulating horizontal forces at the soil surface

seen during vehicle trafficking.

1.3 Purpose

Many studies describing the effects of roots on shear strength have been disclosed in

previous studies. However, most research has addressed root reinforcement for slope stability

rather than trafficability. The VASST provides shear strength measurements analogous to

horizontal shearing occurring at the soil surface during typical vehicle maneuvers; providing a

perspective on the role roots play in laterally reinforcing the soil surface. By investigating the

role roots play in trafficability, the Army plans to incorporate belowground biomass into current

land carrying capacity models and the NATO Referenced Mobility Model (NRMM). The

addition of root influence on soil shear strength, as measured by the VASST, will aid the Army’s

conservation efforts by identifying training land conditions that may lead to accelerated land

degradation or areas which have a higher land carrying capacity for training maneuvers. As a

result, the Army will be able to more effectively utilize training lands by decreasing the down

time needed to rehabilitate vegetation and soil between training events.

4
CHAPTER 2: RESEARCH OBJECTIVES
The goal of this research was to develop a vegetated soil shear strength equation specific

to measurements taken with the VASST, based on the modified Mohr-Coulomb failure criterion.

The research focus was to develop a relationship between soil strength and potential of

vegetation rooting effects on soil trafficability using the VASST. The hypothesis was tested

using field-measured data to estimate vegetation rooting effects on soil trafficability.

These were the specific objectives:

1. Develop a model based on the modified Mohr-Coulomb failure criterion to describe

direct shear forces measured by the VASST tests.

2. Create an automated approach to computing critical soil shearing force resulting from

VASST tests.

3. Validate the model from Objective 1 using field-collected vegetation and soil data and

compare the VASST test results to in-situ measurements associated with soil

trafficability.

5
CHAPTER 3: LITERATURE REVIEW
In order to understand how the VASST may build on previous research of vegetated soil

strength, an in-depth literature review was performed. The history and past research presented in

this thesis served as valuable resources for the development of Mohr-Coulomb as it relates to

measurements obtained with the VASST.

3.1 Discovery of Mohr-Coulomb Failure Criterion

To fully understand the complexities of soil shear strength and how to calculate it, a review

of the Mohr-Coulomb failure criterion initial development was performed. The Mohr-Coulomb

failure criterion, as explained by Holtz et al. (2010) arose from the combining of ideas from

Christian Otto Mohr and Charles-Augustin de Coulomb. Mohr hypothesized that materials fail

when shear stress on a failure plane reached a critical value which depended on the normal

stress. This idea is shown in the following equation:

𝜏𝑓𝑓 = 𝑓(𝜎𝑓𝑓 ) Eq. 3.1

τ = shear strength

σ = applied normal force

With this function, Mohr had developed a method of predicting shear failure in homogeneous

materials.

In the 18th century, Coulomb was interested in the sliding friction characteristics of various

soils and observed both a stress-independent component and a stress-dependent component of

shear strength. The stress dependent component behaved similar to sliding friction; Coulomb

identified this stress as internal angle of friction, denoted as ϕ in Equation 3.2. The symbol c was

given to the independent component that seemed related to intrinsic cohesion.

6
 The combining of Mohr and Coulomb’s theories and observations led to the realization of

the Mohr-Coulomb Failure Criterion, shown in Equation 3.2, which is currently used to

determine the shear strength of engineered soils.

𝜏𝑓𝑓 = 𝑐 + 𝜎𝑓𝑓 𝑡𝑎𝑛 𝜙 Eq. 3.2

τ = shear strength

c= cohesion of soil particles

σ = applied normal force

ϕ = internal angle of friction

3.2 Adaption of Mohr-Coulomb Failure Criterion for Vegetated Soil

Although the Mohr-Coulomb Failure Criterion effectively estimates the shear strength of

engineered soil, it does not account for the reinforcing qualities of roots in the soil matrix. Plant

roots provide reinforcement to a soil matrix by increasing the soil matrix’s tensile strength, a

property that unvegetated soil lacks (Greenway, 1987). A reinforced soil matrix comprised of

roots transfers the stress to roots as a soil is loaded, thus increasing the soil shear strength

(Thorne, 1990).

There are several methods of testing the shear strength of the soil matrix with roots acting

as reinforcing members. These methods include direct shear test, triaxial load test, and the root

strength method. For the purpose of this research, the direct shear test will be reviewed in detail

as it shares the most similarities to the mechanical behavior of the VASST, mainly the

application of a load to the soil matrix in a single direction. The triaxial compression test will be

discussed only briefly to gain further insight on root-soil interactions under unconstrained failure

plane conditions.

7
3.2.1 Direct Shear Test Method

Pioneering the quantification of root influence on soil shear strength was a composite

study by Endo and Tsuruta (1969) and Endo 1980. Results were published after each phase of

the study. The study used 50 cm X 50 cm plots with 1 to 2 tree seedlings planted as yearlings.

After two growing seasons, the plots containing the 3-year-old trees were subjected to an in-situ

direct shear test at depths of 25 cm and 30 cm. In order to perform the test, the plots were incased

with a rigid box and surrounding soil was excavated. The plot, or soil column, had a normal load

applied to the soil surface. With the normal load applied, a horizontal load was gradually applied

to the box using a cable come-along. A strain gauge recorded the force which tended to gradually

increase to a certain maximum, then decrease thereafter. After failure had occurred along the

shear plane, all exposed roots were collected, and diameters of the cut root ends were measured.

The test apparatus is illustrated in Figure 1.1.

Endo (1980) concluded that the effect of roots in a soil matrix increase shear strength of

soil. Endo proposed that the cohesion variable in Mohr-Coulomb Failure Criterion is the sum of

the cohesion capacity of the soil itself and the effect of the root system. The composite report

stated that the total cohesion could be equated using Equation 3.3 or alternatively Equation 3.4.

𝐶 = 𝑐 + 𝑏𝑅 Eq. 3.3

𝑐 = cohesion from the soil particles

𝑏 = an experimental coefficient

𝑅 = the root weight per unit soil volume

8
 𝐶 = 𝑐 + 𝑏′𝐷 Eq. 3.4

𝑐 = cohesion from the soil particles

𝑏′ = an experimental coefficient

𝐷 = total cross-sectional area of roots at shearing plane

Waldron (1977) applied concepts from Endo and Tsuruta’s initial study to create a

laboratory experiment to measure the reinforcing effects of roots in a soil matrix. Waldron

created four types of engineered soils that were contained in rigid, 25.4 cm diameter cardboard

tubes. The tubes were planted with barley, alfalfa, and yellow pine and allowed to grow for 3

months, 12 months, and 16 months respectively. A direct shear test was performed at 15, 30, and

45 cm to determine the amount of reinforcement the roots provided to the soil matrix. Waldron

found that all species had reinforcing qualities in the soil matrix. The most important result of the

research was Waldron’s resulting modification of the Mohr-Coulomb Failure Criterion to

account for the reinforcing effect of roots. Waldron created a modified Mohr-Coulomb Failure

Criterion by adding Δs as a parameter to describe the effect of roots on shear strength, shown in

Equation 1.2. He found that roots’ effect on shear strength could be described with Equation 3.5.

𝛥𝑠 = 𝑇𝑟 (sin 𝜃 + 𝑐𝑜𝑠𝜃 tan 𝜙)(𝐴𝑅 ⁄𝐴) Eq. 3.5

Tr = average tensile strength per root area

θ = angle of distortion in the shear zone

𝐴𝑅 = cross-sectional area of roots at the shearing plane

𝐴 = total cross-sectional area of the shearing plane

Waldon’s (1977) model, shown in Equation 1.2, had some assumptions which severely

limited its applicability to the real-world scenarios of slope stability. The first assumption was

the limitation of the direct shear method, which only measured shear strength along a

9
predetermined plane, horizontal in this case, not the weakest shear plane. Also, the model

assumed that roots are flexible, linearly elastic, and uniform in size, type, and ordination, unlike

roots found in naturally vegetated soils. Furthermore, the tensile strain on the root was said to be

low enough that root length did not change during loading. Lastly, the assumption was made that

the soil’s internal angle of friction, 𝜙, does not change due to roots. However, this last

assumption was previously verified by Gray (1974) who reported that roots had little effect on

internal angle of friction.

Wu et al. (1979) studied the strength of tree roots and their relation to soil shear strength.

Wu focused on root decay’s influence on reducing soil strength after deforestation. Wu’s testing

procedures and initial findings validated previous studies by Endo and Waldron. Similar to

Waldron, cylindrical samples of vegetated soil were submitted to both saturated and unsaturated

direct shear tests under laboratory conditions. However, Wu et al. (1979) collected samples from

naturally vegetated soil, a key in capturing the natural root structures of plants. Wu et al. also

tested the reinforcing action of roots during in-situ tests analogous to Endo (1969). Results of the

in-situ test compared to the laboratory test showed some differences between test types on

unsaturated samples. However, the data was highly scattered and no conclusions could be drawn

about the difference in the in-situ and laboratory tests.

Further analysis by Wu et al. (1979) showed that Waldron’s model for describing the

effect of roots on shear strength, shown in Equation 3.5, could be simplified. After performing a

sensitivity analysis on 𝛥𝑠, it was found that the effect of roots in the modified Mohr-Coulomb

Failure Criterion was relatively unaffected by variations in θ in the range of 40 to 90 degrees and

𝜙 in the range of 25 to 40 degrees. With the value of 𝛥𝑠 between 1.0 and 1.3 for these ranges of

10
θ and 𝜙, Wu et al. (1979) selected 1.2 as a substitute for the angle term. The simplified way of

determining the increase in shear strength due to roots became

𝛥𝑠 = 𝑇𝑟 (𝐴𝑅 ⁄𝐴) ∗ 1.2 Eq. 3.6

Tr = average tensile strength per root area

𝐴𝑅 = cross-sectional area of roots at the shearing plane

𝐴 = total cross-sectional area of the shearing plane

The implication of this adjustment to the modified Mohr-Coulomb Failure Criterion, also called

the simplified perpendicular root model, is that the magnitude of reinforcement is solely

dependent on the root area ratio and strength of roots in the soil (Pollen and Simon, 2006).

Ziemer (1981) also used an in-situ direct shear method to determine the increase in soil

shear resistance due to the reinforcing effect of roots in vegetated sands of Northern California.

But unlike previous research, the shear plane was oriented perpendicular to the surface. The

device Ziemer used was very similar to that used by Endo and Tsuruta in 1969; the main

difference was that Ziemer’s device could be disassembled into plate sections. The plate sections

design was required to orient the box with the shearing surfaces perpendicular to ground surface

without disturbing the shearing surfaces. Once the shear box was in place, a mechanical jack was

used to apply a shearing force to the box at a speed of 1.27 cm/min and a proving ring was used

to measure the resistive force every second over a 7-minute period. Results of Ziemer’s test

found that 79% of the variation in soil shear strength could be accounted for by the dry weight of

live roots less than 17 mm in diameter. However, Ziemer reported that because root orientation

was not accounted for, his findings likely underestimated the true increase in soil strength due to

the reinforcing effects of roots. The Ziemer’s device was also successfully used by Wu et al.

11
(1988) across a range of soils including silty sand, silty sand to silty gravel, well-graded sand,

and lean clay to silt with similar results to Ziemer’s research.

Kato and Shuin (2002), developed an improved direct shear apparatus after reviewing

flaws in calculating the reinforcement behavior of root systems based on individual root strength,

mainly a group effect that emerges at high root densities. The direct shear device had several key

features to observe shear deformation as a result of root-soil interactions under controlled soil

moisture content. In order to observe shear deformation, acrylic beads were placed in vertical

lines in an engineered sand column. The beads were suspended with string as soil was being

placed in the column; the string was then removed once the beads were incased in sand. The

beads were then measured for relative changes in position after the test was performed. This led

to accurate deformation measurements throughout the soil column. Another key feature of the

new device was its ability to control soil water content of the sample being sheared. This was

accomplished by applying suction to the bottom of the sample using the magnitude of suction as

control for the moisture content. Using the modified direct shear apparatus, Kato and Shuin

showed the emergence of a group effect on shear strength as root area ratios increased.

3.2.2 Triaxial Compression Test

Zhang et al. (2010) used a triaxial compression test to observe the reinforcement qualities

of roots on loess soil samples. The roots were oriented in vertical (VR), horizontal (HR), and

vertical and horizontal, which was referred to in the study as cross (CR). All three orientations

were tested at soil water contents of 12.7% and 20% to understand the relationship between soil

water content and root reinforcement. To determine the shear strength of the samples, the

modified Mohr-Coulomb failure criterion, Equation 1.2, was used. The results of the triaxial

compression tests performed by Zhang et al. (2010) are presented in Table 3.1.

12
 Table 3.1: Observation of triaxial compression test (Zhang et al., 2010)
Samples 12.7% Soil Water Content 20.0% Soil Water Content

τ when σ=200 τ when σ=200

𝜙 C (kPa) 𝜙 C (kPa)

Plain Soil 27 29 131 22.3 18 100

HR 26.6 40 140 25.3 22 117

VR 23 64 149 23.7 23 111

CR 23 74 159 26.2 26 124

The most significant increase in soil shear strength compared to plain soil was the cross

rooted samples. The vertical and horizontal root orientations increased soil shear strength as

well, but because they only acted as tension members in one direction, the reinforcing quality

was limited. The data also validated the modified Mohr-Coulomb failure criterion by showing

that as the roots present in the soil matrix increase C, total cohesion, they have little effect on 𝜙,

the internal angle of friction. The study also showed that as soil water content increased, the

shear strength of soil decreased.

3.3 Effect of Soil Water Content on Soil Shear Strength

Roots, cohesion, and internal angle of friction are not the only soil matrix parameters that

affect soil strength, soil water content (SWC) also impacts on soil strength. The effects of SWC

are well documented and predate Waldon’s 1977 modified Mohr-Coulomb failure criterion. In

1936, Terzaghi proposed a modification to the Mohr-Coulomb failure criterion to reflect the

shear strength of soils under saturated conditions. Terzaghi (1936) found that the failure criteria

for saturated soils should be

13
 𝜏 = 𝑐 ′ + (𝜎𝑛 − 𝜇𝑤 ) tan ϕ′ Eq. 3.7

𝜏 is shear strength

𝑐 ′ is the effective cohesion

ϕ′ is the effective angle of internal friction

𝜎𝑛 being the total normal stress on the plane of failure

𝜇𝑤 is the pore-water pressure.

Although Terzaghi’s equation is still used today, most in-situ measurements are performed

in soils that are unsaturated. Fredlund and Morgenstern (1977) reported that two of three stress

state variables (𝜎𝑛 − 𝜇𝑎 ), (𝜎𝑛 − 𝜇𝑤 ), and (𝜇𝑎 − 𝜇𝑤 ) where 𝜇𝑎 is pore-air pressure and 𝜇𝑤 is

pore-water pressure, describe the shear strength of unsaturated soils. With the realization of these

stress states, Fredlund et al. (1977) described unsaturated soil shear strength as Equation 3.8.

𝜏 = 𝑐 ′ + (𝜎𝑛 − 𝜇𝑎 ) tan ϕ′ + (𝜇𝑎 − 𝜇𝑤 ) tan ϕb Eq. 3.8

ϕb = angle of relative change in shear strength compared to matric suction

ϕ′ = angle of relative change in shear strength compared to net normal

stress

Equation 3.8 can be rewritten as

𝜏 = 𝑐 ′ + (𝜎𝑛 − 𝜇𝑎 ) tan ϕ′ + (𝜇𝑎 − 𝜇𝑤 )𝛽 tan ϕ′ Eq. 3.9

𝛽 = tan ϕ′ ⁄tan ϕb = decrease in effective stress resistance as

matric suction increases

Peterson (1988) proposed 𝜏 = 𝑐 ′ + (𝜎𝑛 − 𝜇𝑎 ) tan ϕ′ + CΨ where CΨ is perceived as

cohesion due to suction. The influence of soil suction on shear strength is dependent on soil

water content. The cohesion component CΨ can be expressed as equal to (𝜇𝑎 − 𝜇𝑤 ) tan ϕb

making Petersons equation equivalent to Equation 3.8.

14
 Fredlund et al. (1996) theorized that, unlike previous models, a nonlinear model would

more accurately estimate the effect of SMC on soil shear strength. The equation, through

theoretical development, became:

𝜏 = 𝑐 ′ + (𝜎𝑛 − 𝜇𝑎 ) tan ϕ′ + (𝜇𝑎 − 𝜇𝑤 )[Θ(μa − μw )]𝜅 tan ϕ′ Eq. 3.10

Θ(μa − μw ) = normalized volumetric water content

The normalized volumetric water content was based on saturation, residual conditions, and 𝜅 as a

soil parameter based on soil type. Fredlund et al. (1996) concluded that a model using the soil-

water characteristic curve and saturated soil shear strength properties effectively predicts the

shear strength of unsaturated soil.

Although the effects of SMC on shear strength have been well documented and modeled,

most models do not account for the effects of vegetation. Kato and Shuin (2002) used a modified

direct shear test under varied soil moisture conditions to test the effect of vegetation on soil shear

strength. Study results showed that as root area ratios increased, a group effect emerged. The

observed group effect of roots increased as soil moisture decreased providing evidence of the

importance of SMC’s influence on vegetated soil shear strength. The SMC also played a key role

in shear deformations not accounted for by conventional models such as the modified Mohr-

Coulomb failure criterion.

Pollen (2007) showed that SMC affected mechanical properties of the soil and reinforcing

capabilities of roots. Various root diameters where tested for tensile strength using an in-situ root

pullout test. It was observed that roots could be described as failing one of two ways, breaking or

slipping. A breaking failure was when the tensile strength of the root was exceeded, causing

failure of the root fibers. Slipping failure occurred when the root was pulled from the soil matrix

as a result of a failure at the root-soil interface. Observations made during Pollen’s study showed

15
that as SMC increased, the threshold diameter for root pullout decreased. This information led to

the suggestion that as SMC decreases, the frictional bonds at the root-soil interface increases

resulting in more breaking failures. This conclusion echoed Peterson (1988) showing that

apparent cohesion of the soil increases as the soil matric suction increases.

However, Pollen (2007) was only observing the effects of SMC on the available tensile

strength of roots in the soil matrix. When applying an in-situ direct shear test similar to Endo

(1980), Fan and Su (2008) found that additional peak shear strength from root reinforcement

increases as SMC increases and calculated that roots can increase soil shear strength by up to

100% in soils having a degree of saturation between 80% to 85%.

3.4 Cone Index as a Measure of Soil Trafficability

Several methods has been developed to quickly measure the effect of key soil parameters

on overall strength; these methods measure soil strength in terms of soil trafficability. The

current standard for identifying terrain characteristics related to trafficability and mobility is the

empirical measurement of cone index (CI). CI is the measurement of a soil’s penetration

resistance (Wong, 2001). CI has led to recent developments in the NRMM as CI has a proven

correlation with vehicular performance, rutting, and slipping coefficients (Braunack, 1986;

Muller et al., 1990; Godwin et al., 1991; Shoop, 1993; Nam et al., 2010).

3.4.1 Penetrometer Types

CI values are derived from tests performed with a penetrometer. There are three basic

types of penetrometer: static cone, dynamic cone, and drop cone. These three types of

penetrometers are defined by their operation. Static cone penetrometers measure the force

applied to a metal cone as it is pushed at constant velocity through the soil. Dynamic cone

penetrometers apply a known amount of kinetic energy to a metal cone causing the cone to

16
penetrate the soil. Drop-cone penetrometers are dropped from a known height and the

penetration depth of the cone is measured (Godwin et al., 1991; Jones and Kunze, 2004). For the

scope for this thesis, static and drop-cone penetrometers were used for field measurements.

3.4.2 Static Cone Penetrometers

The static cone penetrometer is one of the most popular tools for measuring the

penetration resistance and is used extensively due to its simplicity, mobility, and measurement

speed (Shoop, 1993). It has been used to predict the trafficability of off-road vehicles (Freitag

and Richardson, 1968; Wismer and Luth, 1972).

The ERDC Geotechnical and Structures Laboratory (GSL) formerly Waterways

Experiment Station (WES) has long identified CI values as a valuable measurement for soil

strength used to determine soil trafficability. ERDC-GSL outlines the use of the static cone

measurements for trafficability purposes as being a 30-degree circular cone with a cone base area

of 3.23 cm2 which is slowly pushed through the soil with a force of up to 68 kg as measured by a

proving ring and deformation gauge (Meyer and Knight, 1961). However, a smaller cone option

is available. The smaller cone has the same apex angle, but the base surface area is 1.30 cm2. CI

is calculated based on the area of cone and the amount of force applied to the cone; the force is

measured incrementally as the cone passes through the soil profile. The standardization of the

static cone penetrometer is found in ASABE Standard 313.2 (1992).

3.4.3 Drop-Cone Penetrometers

The drop-cone, currently used by CERL, was developed as a tool capable of quickly

measuring penetration resistance, compared to previous penetrometers (Godwin et al., 1991).

The drop-cone consists of a 2-kg, 30 degree cone that is dropped from a height of 1 meter. The

depth of penetration into soil is then measured and used to describe penetration resistance. The

17
drop-cone, as described by Shoop (1993), applies a large force on the soil surface without the use

of large weights or hydraulic equipment; this is advantageous for performing in-situ penetration

tests.

When developing the drop-cone, Godwin et al. (1991) assessed the relationships between

drop-cone penetration depth and cone type, soil moisture content, and shear strength. The cone

types tested consisted of variations in mass and apex angle and were evaluated based on

penetration depth and sensitivity. Results showed that a 2 kg mass with a 30 degree apex angle

was most suitable for trafficability studies. Goodwin also showed the existence of a linear

relationship between drop-cone measurements and soil moisture content for sandy loam, clay,

and clay loam soils. Drop-cone and shear strength measurements as determined by an 18mm

diameter shear vane were found to have a linear relation at soil shear strengths above 20 kPa.

The drop-cone measurement was also found to be a good predictor of wheel rut depth for a range

of wheel systems.

3.5 California-bearing Ratio as a Descriptor of Soil Shear Strength and Trafficability

Similar to penetration resistance measurements of CI, the California-bearing ratio (CBR)

also uses the soil’s resistance to vertical loading to measure trafficability. CBR is a standardized

test to obtain a modulus of shearing resistance that is recognized by both the American Society

for Testing and Materials (ASTM) and the United States Military (Semen, 2006).

3.5.1 Standard CBR Test

According to ASTM Standard D4429 (2009), the standard CBR test is performed by

measuring the force required for the penetration of a cylinder with a 3-square-inch end surface

area through the soil surface. The force measurements are taken in 0.025 inch increments until a

18
0.500 inch penetration is reached. To determine the CBR index value, the force measurements at

0.100 and 0.200 inch depths are divided by values of 1,000 psi and 1,500 psi (Semen, 2006).

3.5.2 Clegg Impact Hammer

An alternate method of measuring the CBR of soils and aggregates is the Clegg impact

hammer which can be practically implemented in field and laboratory testing. The Clegg impact

hammer is an easy to operate, portable, and cost-effective means of measuring the Clegg impact

value (CIV) (Al-Amoudi et al., 2002).

One feature of the Clegg impact hammer similar to the drop-cone penetrometer is the

much larger area of soil tested compared to the static cone penetrometer. The advantage of the

large sample area is that variability from complex soil compositions is lowered between samples.

However, some researchers have found that large sample areas were too insensitive (Shoop,

1993). Also, similar to the drop-cone penetrometer, the Clegg impact hammer is limited to taking

measurements at the soil surface.

Although the Clegg impact hammer is typically used to assess the carrying capacity of

unsurfaced roads (Mathur and Coghlan, 1987), it can also be used on vegetated soils for

measures of trafficability. Koch et al. (2010) found that CIV and penetration resistance, as

measured by a static penetrometer, are highly correlated measurements in vegetated, clayey soils.

They also observed that CIV presents a lower variability in measurements compared to the drop-

cone penetration test.

An investigation of the correlation of the CBR and CIV was carried out by Al-Amoudi et

al. (2002). With tests conducted on 56 engineered samples of silty sands and silty gravels, they

found a general model was able to estimate CBR from CIV with a coefficient of determination

19
(R2) of 0.85. The general model, Equation 3.11, was similar to models reported in previous

literature.

𝐶𝐵𝑅 = 0.1691(𝐶𝐼𝑉)1.695 Eq. 3.11

3.5.3 CBR from CI

It was found that the CBR was also highly related to CI for soil types other than silty

sands and silty gravels. However, to better describe the CBR values of each soil type, the model

needed soil specific coefficients and exponents. Shoop et al. (2008) evaluated 562 cases of

coinciding CBR and CI measurements representing seven of the United Soil Classification

System (USCS) soil types. It was found that CBR and CI are highly correlated for fine-grained

soils but less correlated for course-grained soils. Similar to the model relating CIV to CBR

proposed by Amoudi et al. (2012), Shoop et al. (2008) found the form of the Equation 3.12 with

coefficients a and b per soil type listed in Table 3.2 could be used to describe the relationship

across a broader range of soil types.

𝐶𝐵𝑅 = 𝑎𝐶𝐼 𝑏 Eq. 3.12

Table 3.2: Coefficients and exponents for Equation 3.12 Shoop et al. (2008)

USCS Coefficients and Exponents

Soil Type Classification a b R2
All Soils 0.2985 0.5358 0.4715
Clay, High Plasticity CH 0.1264 0.6979 0.8516
Clay, Low Plasticity CL 0.1266 0.6986 0.8701
Silt, High Plasticity MH 0.0820 0.7174 0.7715
Silt, Low Plasticity ML 0.1111 0.7390 0.5193
Coarse-Grained SM + GP 1.1392 0.4896 0.3495
Fine-Grained CH, CL, MH, ML 0.1305 0.6776 0.7724
High Plasticity CH + MH 0.1460 0.6432 0.7741
Low Plasticity CL+ ML 0.1281 0.6984 0.7962

20
3.6 Development of the VASST

The VASST was designed to quantify shear strength of vegetated soil at the soil surface

(MacDonald et al., 2012). The motivation for the development of the VASST was to better

understand the shear strength relationship between roots and soil at the ground surface, the

section of the soil profile most susceptible to horizontal displacement. It was anticipated that the

VASST tests could better predict vehicle mobility and trafficability in vegetated soils when

compared to previously mentioned soil strength measurements (CI, CIV, and Drop-cone).

3.6.1 VASST Components and Operation

The VASST consists of the following major components: 1) tire analog/ shear surface, 2)

rails/tubes, 3) load cell, 4) linear position transducer, 5) winch, and a data acquisition unit. The

layout of the primary components of the VASST is shown in Figure 3.1.

Figure 3.1: Primary components of the VASST (Kelley et al., 2013)

21
 In order to test soil shear strength, the tire analog is inserted into the soil and pulled

horizontally through the soil via the winch. Data on the force being applied to the tire analog and

displacement of the tire analog is measured with the load cell and linear position transducer,

respectively. The action of the tire analog during a VASST test is shown in Figure 3.2. Data also

available from the VASST include time, temperature, equipment voltage, and event markers. All

data is recorded on a 10 milliseconds interval using the data acquisition unit. An output file is

then assembled from the sensor data and exported to a comma-separated values (CSV) file.

Figure 3.2: Tire analog during VASST test

3.6.2 Initial Test Results

MacDonald and Shoop (2013) tested the VASST alongside a Clegg impact hammer,

static cone penetrometer, Drop-cone penetrometer, and a shear vane in reconstructed sand and

clay. Tests of both vegetated and unvegetated sections were performed in each soil type. During

initial VASST testing, they found that soil shearing data obtained with the VASST followed

trends similar to common soil strength measures.

22
CHAPTER 4: MATERIALS AND METHODS
Materials and methods presented in this thesis were developed at the US Army Corps of

Engineers ERDC-CERL and ERDC-CRREL. Standardized laboratory testing was achieved using

ASTM Standards and ERDC-CERL testing protocols.

4.1 Site Selection and Typical Sampling Layout

The careful selection of sites was important to obtain results that would be meaningful

when used as part of the OPAL work package; this required a wide range of soil types to be

tested. Sampling protocol and layout was also important to maximize testing uniformity and

improve spatial correlation between the soil strength tests and measured soil parameters.

4.1.1 Test Locations

Geographic locations were chosen based on soil, eco-regions and training land usage.

Test locations were within Army installations in agreement with the OPAL program. Five

locations for the study were: Champaign, Illinois; Ft. Benning, Georgia; Ft. Bragg, North

Carolina; Ft. Hood, Texas; and Ft. Riley, Kansas.

4.1.1.1 Champaign, Illinois

Field testing was conducted at ERDC-CERL’s campus in Champaign, Illinois. An

existing monitored and controlled research site for trafficability studies was used. Soil at the

location was Catlin silt-loam (Soil Survey, 2014). Existing plots had five vegetation treatments:

bare soil, turf grass mixture, forbs mixture, native grasses mixture, and a mixture of

forbs/grasses.

23
4.1.1.2 Ft. Benning, Georgia

Fort Benning, located at roughly 32.36° N, 84.97° W, is within the Sandhills region

which contain deep, fluvial sands. Three sites were selected based on dominate soil type at Fort

Benning; soils of the test sites were Nankin sandy clay-loam, Troup loamy sand, and Lakeland

sand (USDA web Soil Survey, 2014). The vegetation in the three areas tested consisted primarily

of scrub oaks, longleaf pines, and sparse grasses (Dilustro, 2002).

4.1.1.3 Ft. Bragg, North Carolina

Fort Bragg, located at approximately 35.14° N, 79.00° W, is positioned on the inner

margin of the Atlantic Coastal Plain near the Sandhills region. Typical vegetation in the area

consists of a longleaf pine, oak, and wiregrass communities (Sorrie, 1997). The soils determined

to be representative of the area were Lakeland sand, Blaney loamy sand, and Candor sand

(USDA web Soil Survey, 2014).

4.1.1.4 Ft. Hood, Texas

Fort Hood, located near 35.14° N, 79.00° W, is in the boundary region of the Cross

Timbers and Prairies and Edwards Plateau eco-regions. The vegetation dominating the areas

tested consisted of tallgrasses associated with the Blackland Prairie such as Little Bluestem,

Indian Grass, Switch Grass, Eastern Gamagrass, etc. (Wootan, 2011). The area also contained

some mixed forest-shrub communities generally dominated by Ashe Juniper and Live Oak

(Directorate of Public Works Environmental Division, 2011). Soil studied at Ft. Hood included

Nuff Series very stony clay loam and Georgetown clay loam (USDA web Soil Survey, 2014).

4.1.1.5 Ft. Riley, Kansas

Fort Riley, located at 39.11° N, 96.82° W, is within the Flint Hills tallgrass prairie with

vegetation such as Big Bluestem, Indian Grass, and Switch Grass common (Kansas Sampler

24
Fountain, 2007). The area studied was an upland area with a soil taxonomy of Crete silty clay

loam (USDA web Soil Survey, 2014).

4.1.2 Selection of Sites

The test sites within each location were chosen based on the following characteristics:

representation of soils predominate on the installations, representative vegetation, and no history

of vehicle traffic within the plot. To quantify the predominate soils at each location, ArcGIS was

used to populate a list of soil types on the installation with information on area of coverage. The

three most common soils by area were then identified and accessed based on training utilization.

It was important for the purpose of this study that the sites had not been used for training

activities that would have created compaction and/or been destructive to the vegetation. This

criterion greatly narrowed down the site options greatly since the sites were within active

training areas. The final site selection criterion was vehicle accessibility, an important factor due

to the amount of equipment required to conduct field testing. A summary of the number of sites

and total number of samples and tests performed at each location is shown in Table 4.1.

Table 4.1: Sampling quantities per location

Sampling Quantities per Location

Soil Classification
Number of Sites

Above Ground
Below Ground
Penetrometer

Clegg Impact

Bulk Density
Drop-cone
Hammer

Biomass

Biomass
VASST
TDR

Location
Ft. Bragg, North Carolina 3 9 45 51 45 81 75 9 9 9
Ft. Benning, Georgia 3 9 54 81 45 45 68 9 9 9
Champaign, IL 5 13 61 65 39 130 102 13 13 13
Ft. Hood, Texas 2 5 25 35 25 45 40 5 5 5
Ft. Riley, Kansas 1 3 3 3 3 9 16 3 3 3
Totals: 14 39 188 235 157 310 301 39 39 39

25
4.1.3 Replication Layout

Replications established at each site were chosen based on similarities in vegetation and

soil characteristics as well as proximity to each other. For each replication, the tests performed

and samples taken are listed in Table 4.2.

Table 4.2: Tests performed and samples taken per replication

Soil Tests Soil Samples Vegetation Samples
VASST Soil Classification Sample Above Ground Biomass
Clegg Impact Test Bulk Density Below Ground Biomass
Cone Penetration Test
Drop-cone Penetration Test
TDR

These samples and tests were performed in a manner illustrated in Figure 4.1. Sample and test

quantities for each replication can be found in Appendix A.

Figure 4.1: Typical sampling layout for field data collection

26
4.2 Field Sampling Methods

Field sampling was conducted during 2012 and 2013 with one sampling event per location.

On location, sites were defined based on their soil classification, all in-situ measurements were

taken, and soil and vegetation samples were collected for laboratory testing.

4.2.1 Soil Classification

Soils were initially classified using the soil data and information from the National

Cooperative Soil Survey and the U.S. Department of Agriculture (USDA) and installation soil

surveys. Information obtained about each soil included the soil name, USDA textural soil

classification, distribution, and location. Soils were analyzed and classified using the USCS,

commonly used in engineering applications. The soil classifications by the USCS were

performed at ERDC-CERL on samples that were collected during field sampling; details of this

process are described in section 4.3.1.

4.2.2 Soil Moisture

Soil moisture was measured in-situ using a Fieldscout TDR 300 (Spectrum Technologies

Inc.) which measured SMC on a volumetric basis. The time domain reflectometry (TDR) probe

used for this study was equipped with 20cm probes and the standard calibration setting was used

as determined by the soil types tested. Figure 4.2 demonstrates the TDR being used to measure

in-situ soil moisture.

27
 Figure 4.2: Example of TDR test

The TDR measures soil moisture by injecting a step voltage in the soil in the form of a

pulse. The velocity of the pulse is measured across a known distance to determine the soil

moisture; slower pulse velocity indicates higher SMC. Evett (2003) explains the theory of TDR

systems when used to make measurements in porous materials by converting measurements of

time into length by using the relative propagation velocity factor, vp, a fraction of the speed of

light. The equation for calculating the relative propagation velocity factor is:

𝑣𝑝 = 𝑣⁄𝑐0 = (𝜀𝜇)−0.5 Eq. 4.1

𝑣 = propagation velocity of a pulse along the cable

𝑐0 = speed of light in a vacuum

𝜀 = permittivity

𝜇 = magnetic permeability of the dielectric material

28
The permittivity of soil is highly sensitive to water with soil type having a minimal influence on

permittivity. This means that 𝜀 is an effective indicator of soil water content.

4.2.3 Soil Strength Tests

Soil strength was considered the key component in determining the impact of

trafficability and was tested using several methods. Tests were conducted with the VASST, static

cone penetrometer, drop-cone penetrometer, and Clegg impact hammer.

4.2.3.1 Vegetation and Soil Shear Tester (VASST)

At each test site, the VASST described in section 3.6 was set up in two adjacent positions

with three to five tests per position. The test areas were prepared by clipping above ground

biomass and removing organic debris. The VASST was then anchored via eight anchoring spikes

inserted approximately 12 inches into the ground. Tests were performed with the tire analog

inserted to a depth of 1.5 inches into the soil surface, a depth recommended by MacDonald et al.

(2012). A load applied to the tire analog was increased until soil in contact with the tire analog

was displaced; this was verified by visual inspection. Figure 4.3 shows where the soil has been

displaced by the tire analog after two tests.

29
 Figure 4.3: Soil displacement after two VASST tests (tests indicated by arrows)

Concerns about a depth of 1.5 inches being adequate to test soil shear strength were

discussed at the beginning of this study. However, initial VASST testing at depths exceeding the

recommended depth proved to be beyond the capabilities of the VASST. At a depth of 2 inches,

in soil with high shear strength, the spikes deformed during the soil shearing, creating test data

that was not replicable.

4.2.3.2 Static Cone Penetrometer

The static cone penetrometer used was a Field Scout SC 900 Soil Compaction Meter

produced by Spectrum Technologies Inc. The penetrometer’s cone had a 30° apex angle and a

base area of 1.3 cm2. It was equipped with sonar to measure penetration depth, a load cell to

measure the force applied to the cone, and a data logger to record the average load applied over
30
2.5 cm distances. The SC 900 could have measured CI at depths of up to 45.0 cm, however,

measurements at that depth were unnecessary and impractical for the purpose of this study. An

example of the penetrometer being used is shown in Figure 4.4.

Figure 4.4: Example of static cone penetrometer test

In the field, the measurements using the SC 900 were taken adjacent to the VASST. The

CI values recorded per depth were average forces per area on the cone. When the datasets from

the SC 900 were uploaded to a computer, it was observed that the sonar had started

measurements before the cone made contact with the soil. This was corrected by removing zero-

force values at the start of the tests and adjusting the depth measurement so that the first positive

force was at the zero depth reading. The depths utilized were at 0.0, 2.5, and 5.0 cm and deemed

of the most interest due to the section of the soil profile tested by the VASST and drop-cone.

31
4.2.3.3 Drop-Cone Penetrometer

The drop-cone, built to the specifications of Godwin et al. (1991), had a 2 kg overall

mass and a 30 degree apex angle cone. The cone was attached to a graduated rod and dropped

through a poly-vinyl chloride (PVC) pipe from a height of 1 meter, illustrated in Figure 4.5.

Once dropped, the graduations on the rod were read and processed to determine the penetration

depth of the cone to the nearest millimeter. The drop-cone was performed at each replication

between three and ten times depending on the location; number of repeated measurements can be

found in Appendix A.

Figure 4.5: Example of drop-cone test

32
4.2.3.4 Clegg Impact Hammer

The Clegg impact hammer used for field testing was a 2.25 kg Clegg impact soil tester

Model 95049A, produced Lafayette Instruments. The Clegg impact hammer, based on a

modified Proctor hammer, contained a piezoelectric accelerometer and a digital output. Tests

were performed per standard test protocol, with the hammer dropped four consecutive times in

the same spot from a height of 45 cm, demonstrated in Figure 4.6; the peak CIV of the four tests

was recorded.

Figure 4.6: Example of Clegg impact test

33
4.2.4 Soil Classification Sample

Samples taken for soil classification procedures were obtained by first removing above

ground biomass and any organic debris on the soil surface. Then a spade was used to remove a

sample from the first 6 inches of the soil profile that was approximately 170 cubic inches in

volume. This was determined a sufficient amount of soil to perform both particle size analysis

for all soils and Atterburg limits for fine grained soils, (procedures described in Section 4.3.1).

4.2.5 Biomass Sampling

To quantify the vegetation at each location, both aboveground and belowground biomass

samples were taken at each plot.

4.2.5.1 Aboveground Biomass

U.S. Army Land Condition-Trend Analysis (LCTA) has laid out standards for accessing

vegetation on military lands via a line intercept method (Diersing et al., 1992). However, to

better suit the needs of this research, a modified Daubenmire frame or quadrat was used to

measure ground cover adjacent to the location of the VASST. The advantage of the quadrat

method over the methods described in LCTA was the area of sampling. The quadrat used for

vegetation sampling was 2500 cm2, a much larger sample area compared to the 60 cm2 sampled

per transect in the LCTA method.

The quadrate used in this study was constructed of 0.75 inch diameter PVC with sides

measuring 50 cm which created a sampling area of 2500 cm2. The quadrat also had a digital

camera mounted at a height of 1m that aided in the documentation of the sample. The vegetation

residing inside the frame was clipped at ground level and bagged in a pre-weighed paper bag,

dried in a laboratory oven. A description of the drying process is in section 4.3.2.1.

34
4.2.5.2 Belowground Biomass

The quantification of belowground biomass was of major importance in achieving the

goals of this research. However, quantifying belowground was difficult due to the root

architecture and distribution throughout the soil profile. A soil core method was chosen to

measure the amount of belowground biomass per volume of soil present near the soil surface.

The use of soil core sampling methods was the only option suitable for use on active military

lands due to speed of sampling and limited disturbance as not to interfere with future military

activities.

The core samples taken for belowground biomass measurements were 5 cm in diameter

and 25 cm deep. The samples were extracted using a split-core sampler driven into the ground

using an 8.6 kg slide hammer; both the spilt-core sampler and slide hammer were manufactured

by AMS, Inc. Once extracted, the soil core was placed in a sealable, gallon plastic bag, labeled

and transported back to CERL’s soils laboratory.

4.2.6 Bulk Density

Bulk density samples were taken with a 3 in. inside diameter by 3 in. deep ring as

standardized by the Natural Resources Conservation Service (NRCS) in 2001. However, unlike

procedures described in the NRCS (2001) method, the rings were driven into the ground using a

modified slide hammer developed at CRREL. The slide hammer was designed to deliver impacts

evenly across the edge of the ring ensuring the net force of the blow to be aligned with the

central axis of the ring, in turn reducing sample compaction and variability.

To take a bulk density sample, a ring was loaded into the slide hammer and driven into the

ground to a depth just over 3 inches. The volume of soil surrounding the sample was then

removed using a spade. Soil surrounding the wall of ring was removed by hand; soil extending

35
past the open ends of the ring was removed using a knife creating clean, uniform bulk density

sample. In order to transport the sample to the lab, it was placed in a sealed, quart plastic bag,

and labeled.

4.3 Laboratory Methods

Laboratory testing and analysis was required to obtain soil parameters and quantify

vegetation. All laboratory tests were conducted at CERL following ASTM Standards where

applicable. Tests not defined by ASTM Standards were performed using Army Corp. of

Engineers protocols.

4.3.1 Soil Classification

Soils were classified by USDA texture and the USCS. The tests used to describe the soils

by the two classification systems were particle size analysis and Atterburg limits respectively.

4.3.1.1 Particle Size Analysis

Particle size analysis (PSA) was performed on soil samples as a way of quantifying their

gravel, sand, silt, and fines content. ASTM Standard D422 (2007) procedures were followed for

all PSA samples.

4.3.1.1.1 Sieve Analysis

Sieve analyses were performed on the sandy soils from Ft. Benning and Ft. Bragg as a

method of classifying the soils by the USCS. The gradation of the particle sizes in the range of

4.75 mm (No. 4) to 75 µm (No. 200) was determined using ASTM Standard D422 (2007). The

samples were passed through three stacks of 8 inch sieves. Samples first passed through U.S.

Standard Sieve Sizes No. 4, 10, 16, 20, and 35, the second set was, 40, 60, 70, 100, 120, and 140

in size, and the final set 170, 200, and 325 in size.

36
 A sub-sample of approximately 200 g of air-dried soil was shaken through each sieve

stack for eight minutes via a mechanical sieve vibrating device. After each stack was shaken, the

contents remaining on each sieve were weighed to an accuracy of 10 mg. The soil fraction

passing through all the sieves in a stack was collected in a pan and loaded into the sieve stack

containing the next smaller series of sieves. The process was repeated for all three stacks

resulting in an accurate particle size gradation used to aid in the classification of coarse-grained

soils.

4.3.1.1.2 Hydrometer Analysis

Particle size distribution analyses of soils consisting primarily of silt and clay were

conducted using the ASTM Standard D422 (2007) hydrometer method. The hydrometer method

relied on larger particles settling out of a liquid solution faster than smaller particles. Rates of

sedimentation of various sized particles were captured with multiple hydrometer measurements

in a time series. The 151H hydrometers used for the study were manufactured to meet ASTM

Standard E100 (2010).

Concurrent with ASTM Standard D422 (2007), an air-dried sample weighing

approximately 50 g was stirred into a hexametaphosphate solution (40 g/L) and allowed to sit

more than 16 hours. The sample was then added to 500 ml of deionized water. The soil was then

mixed with a soil blender for 8 minutes and transferred to a 1000 ml graduated cylinder. The

total volume of the soil slurry was brought to 1000 ml with deionized water. Hydrometer and

temperature measurements were taken at 1, 2, 4, 15, and 30 minutes followed by measurements

at 1, 2, 4, and 24 hours.

37
4.3.1.2 Atterberg Limits

Atterberg limits were conducted on soils consisting primarily of silt and clay following

ASTM Standard D4318 (2010). The liquid limit and plastic limit were used to define each

sample’s plasticity index, a parameter used to define fine-grained soils by the USCS.

4.3.1.2.1 Liquid Limits

The liquid limit for each plot was found using a sample of soil that had particles larger

than Sieve Size No. 40 removed. The sample was hydrated and placed in an Atterberg device and

cut by a standard grooving tool, illustrated in Figure 4.7. The cup was then dropped repeatedly a

distance of 10 mm at a rate of two shocks per second. The number of blows used required to

close or “seal” the groove over a distance 13mm was counted. The sample was weighed and then

oven-dried at a temperature of 100°C for 24 hours to determine the SMC. This process was

repeated three times at various soil moistures to determine a linear relationship between the

number of blows and SMC. From the linear relationship, the SMC was found at which the gap

would close after 25 blows, a descriptor of liquid limit outlined in ASTM Standard D4318

(2010).

38
 Figure 4.7: Grooving of a liquid limit sample

4.3.1.2.2 Plastic Limits

The plastic limit for each plot was determined by ASTM Standard D4318 (2010). A

small amount of the extra soil prepared for the liquid limit test was used for the plastic limit test.

The sample weighed between 1 and 2 grams and was repeatedly rolled on a glass surface into a

rod that was 3.2 mm in diameter. Each time a rod of the specified diameter was formed, it was

broken into several threads and remolded allowing for the rod to be reformed by means

previously described. Once the sample’s SMC decreased to the point that a 3.2 mm rod could no

longer be formed, the sample was weighed, oven-dried at a temperature of 100°C, and reweighed

to determine the final SMC.

39
4.3.2 Biomass

The accurate quantification of biomass was critical to the scope of this research. Army

Corp. of Engineers protocols for quantifying biomass were used to achieve accurate

measurements.

4.3.2.1 Aboveground Biomass

Aboveground biomass was oven-dried at a temperature of 85°C in a convection

laboratory oven for a period of 48 to 72 hours, depending on the volume and approximate water

content of the vegetation. Once the mass of the vegetation was stable, indicating that no more

water was being removed, the vegetation was weighed to determine a mass of vegetation per

area.

4.3.2.2 Belowground Biomass

Belowground biomass samples were handled with great care so that root structures were

not degraded as soil was being removed from the roots. In order to recover as many roots from

the core samples as possible, the root washer and processing methods described in Fulton (2012)

were employed.

The first step of the washing process was to transfer each soil core into a mesh bag with

125 µm openings. The samples were then soaked in water for 1 to 3 days or until the soil cores

were thoroughly softened. Once the samples were softened, bagged samples were run through

the root washing apparatus described by Fulton (2012). After a large portion of the samples’ silt

and clay fraction had been removed, the roots and remaining soil particles were separated by

hand using a No. 18 sieve. The soil-free root samples were oven-dried at a temperature of 85° C

to determine the dry mass of the belowground biomass.

40
4.3.3 Bulk Density

The bulk density samples were placed in pre-weighed ceramic drying dishes and weighed.

The samples were then oven-dried in a convection laboratory oven at a temperature of 100°C for

24 hours or until all moisture was removed from the soil. After drying, the samples were allowed

to cool in a desiccator to reduce convention currents from the samples which can affect scale

readings. Samples were then weighed to determine the mass of soil contained in the known

sample volume.

41
CHAPTER 5: COMPUTATIONAL APPROACH TO ANALYZING VASST
DATA
The nature of this study resulted in a large and complex dataset that required simplification

and data-mining to produce meaningful results. Data on VASST measured soil shear strength

was derived from the data collected in the field; these measurements were tested against a

theoretical model. The theoretical model developed from literature was based on the modified

Mohr-Coulomb failure criterion (Waldon, 1977) and adjusted to describe the lateral shearing

action produced by the VASST.

5.1 VASST Raw Data Analysis

During operation, the VASST created large amounts of data for each test that had to be

refined to determine the breaking strength of the soil. A Python program was developed to

quickly identify the force resulting in shear failure for each sample tested.

5.1.1 Previous Data Analysis Method

When the VASST was initially being tested by MacDonald et al. (2012), the output CSV

file was imported into DIAdem, software that aids in the inspection, analysis, and reporting of

data. However, the program could only be used to view the VASST outputs in the form of graphs

and was not capable of identifying soil shear failure. Therefore, human input was required to

identify the point at which shear failure occurred. Furthermore, using DIAdem to individually

analyze datasets was a time consuming task. In order to remove human bias and reduce data

processing time, a Python program to analyze the VASST data was developed as part of this

thesis (Appendix B).

42
5.1.2 Algorithm for VASST Dataset Analysis Program

Creating a computational approach for analyzing raw data from the VASST was a

complex process. To simplify the procedure, the following process was implemented:

1. Directory setup/import raw data

2. Validate dataset/condition the data

3. Determine force maximums

4. Determine points of interest other than maximums

5. Check for special cases (root breaks and event markers)

6. Output relevant information and attributes pertaining to the shear failure

5.1.2.1 Directory Setup

The first action taken by the program was to specify three file directories: code, data, and

output. All three directories were user specified using the Tkinter package, a practical Graphical

User Interface (GUI) toolkit for Python. The first input was the directory of the supporting

program files which were sequentially executed from the main code during file processing. The

next directory specified by the user was the location of the data files; the location was used so

that all compatible data files in the directory could be processed without specifying each file

individually. Finally, an output directory was designated to contain all files created by the

program as outputs.

5.1.2.2 Importing Data / Compatibility Check

Once the data was located, a list of all CSV files in the directory was compiled. The list,

displayed in the Python shell, informed the user which CSV files were found in the directory.

Each file was then checked for size compatibility, meaning that files over 1 megabyte were not

processed. This was important for two reasons: the large size could have been an indication of

43
bad data or the program may not have ran depending on the capabilities of the computer. Once a

file’s compatibility was verified, the program opened the file using Python’s CSV module.

Processing was performed in alphanumerical order based on file name.

5.1.2.3 Data Conditioning

Conditioning of raw data collected from the VASST removed unused data and checked

for problems in the data sets. As the data were being read into a Python array, unnecessary

header information was removed. Since raw data from the VASST used the same number of

header rows, the rows were removed based on position rather than content.

If a second header was found within the dataset, it was an indication that multiple tests

were exported as a single CSV file as shown in Figure 5.1. Once the program identified the

existence of multiple tests in a single dataset, the program prompted the user to choose between

continue using the first test data or skip to the next file. If the user chose to continue using the

first test data, the program clipped the data so that only the first segment was used. If the user

chose to skip to the next file, the system continued to next file.

44
 Figure 5.1: Unedited dataset containing multiple tests

With the header removed and the data verified as being from a single test, the array

consisting of floating numbers was tested for equipment failures associated with the linear

position transducer and load cell. By comparing the maximum and minimum values recorded by

each of the two VASST components, failures could be detected. If the differences in the

maximums and minimums were less than 5 lb. or 0.25 in. respectively, the failure was reported

as an error and program skipped to the next file. Examples of both failures are shown in Figure

5.2 and 5.3.

45
 Figure 5.2: No load cell detected

Figure 5.3: No linear positon transducer detected

46
 The event marker was checked for failures resulting from the momentary switch short-

circuiting. A short-circuit displayed a marker signal throughout the entire test, as shown in

Figure 5.4. Similar to when multiple data segments were found, if an event marker failure

occurred, the program prompted the user to either continue with the current file or skip to the

next file. However, there were two continue options which allowed the user to decide whether

the field marker should or shouldn’t be displayed on the output graph.

Figure 5.4: Event marker short-circuit warning

Next, the data array was trimmed from the duration of the recording to contain only data

being recorded while a load was being applied as illustrated by Figure 5.5. Because data was

recorded at a higher precision than available from the sensors on the VASST, negative loads and

reverse movement were often indicated before the test started and after it ended; by removing

these readings, the dataset size was greatly reduced. Also, the VASST readings were limited to a

47
maximum position of just over 37.5 in.; thus load cell readings at a distance in excess of the

maximum position could be removed.

Figure 5.5: Untrimmed dataset with noise at beginning and end

At the end of the data conditioning, the program verified that a usable data array existed

in order to prevent errors in following steps. To determine whether the dataset was usable after

conditioning, the number of rows in the dataset had to contain a minimum of two points; the

minimum of two points was critical to ensure the existence of both maximum and minimum

forces and positions.

5.1.2.4 Start Point and Peaks in Force

As discussed in MacDonald et al. (2012), one primary point of interest was the first peak

in the displacement-force graph. The first peak was identified by comparing the fn force to fn+1

48
force where n was the row number. If 𝑓𝑛 > 𝑓𝑛+1 was true, then the force and position associated

with fn was recorded as a point of interest. After the first maximum was found, the next

maximum was a point of interest only if: 𝑓𝑛 > 𝑓𝑛+1 and 𝑓𝑛 > 𝑝𝑟𝑒𝑣𝑖𝑜𝑢𝑠 𝑚𝑎𝑥𝑖𝑚𝑢𝑚 𝑓𝑜𝑟𝑐𝑒 held

true. In order to avoid false maximums before the test began, the array of maximums was tested

against the dataset’s global minimum and an absolute start force value of 20 lb. If any of the

maximums had a force within 10 lb. or were within 5% of the global minimum or were less than

20 lb., they were discarded. An example of an array of maximums as determined by the program

is illustrated in Figure 5.6.

Figure 5.6: Example of complied maximums

A start point was added to the array containing maximums, referred to as points of

interest, using a logic search similar to the one implemented to find the maximums. The search

was conducted in the direction of descending displacement from the first maximum until the

49
force criteria 𝑓𝑛−1 > 𝑓𝑛 or 𝑛 = 0 was met. The addition of the start point to the array containing

points of interest aided in the calculation of soil displacement at shear failure.

5.1.2.5 Identifying Cases Requiring Human Input

Though the program was designed to automate the process of analyzing data obtained

from the VASST, three special cases were identified in which the program could not correctly

determine shear failure. For these special cases, the program prompted the user to identify the

point of interest that corresponded to the shear failure using a GUI.

5.1.2.5.1 Special Case 1: Rapid Change in the Force/Displacement Ratio

In some cases, especially when dealing with fine-grained soil, the point at which shear

failure occurred was not defined as a peak force. This was a result of the sheared soil pushing,

commonly termed plowing, intact soil forward. The occurrence of plowing was identified using

the force/displacement ratio, referred to as slope. To identify rapid changes in the slope, the

program created an array of the slopes using (𝑓𝑛+2 − 𝑓𝑛 )⁄(𝑝𝑛+2 − 𝑝𝑛 ) where f = force and p =

displacement. Once an array of slope values was created, each slope was compared to the

average of the previous slopes and the relative error was used to determine whether a sudden

slope change occurred. If a rapid slope change was identified before a maximum was reached,

the point was added to the array containing points of interest. Figure 5.7 demonstrates a sudden

change in slope before the first maximum, the red point on the graph.

50
 Figure 5.7: Example of special case 1, sudden slope change

5.1.2.5.2 Special Case 2: Inflection Points

Similar to rapid changes in slope, inflection points were identified as an indicator of shear

failure. Because the program looked for inflection points associated with shear failure, only

inflection points on positive force-displacement slopes shifting from decreasing slope to

increasing slope were calculated. The method used to find inflection points was a series of logic

tests: if 𝑓𝑛 < 𝑓𝑛+3 < 𝑓𝑛+5 was true, then test whether 𝑠𝑛+1 > (1.25 ∗ 𝑠𝑛+2 ) and ( 1.25 ∗

𝑠𝑛+3 ) < 𝑠𝑛+4 were true with 𝑠𝑛 being the slope between 𝑓𝑛 and 𝑓𝑛+1 . These logic tests were

chosen to limit the number of inflection points to just those that are likely due shear failure. If an

inflection point, as defined by the logic tests, occurred in the dataset at a displacement less than

the first maximum, all inflection points for the dataset are added to the points of interest array.

51
The user was then prompted to choose the point of shear failure. Figure 5.8 illustrates how

inflection points of interest are displayed on the graphical output.

Figure 5.8: Example of special case 2, inflection points of interest

5.1.2.5.3 Special Case 3: Root Breaks

Roots could cause false maximums if they broke before the soil had experienced

complete shear failure. A root break could be observed in the dataset as a series of 4 to 10

measurements that had the following pattern: a steep positive slope initially, followed by a steep

negative slope, and a return to a steep positive slope. When this pattern was detected anywhere in

the conditioned dataset, the program prompted the user to identify the point of interest that

corresponded to the shear failure. Figure 5.9 displays how the root break pattern appeared on the

graphical output.

52
 Figure 5.9: Example of special case 3, detected root breaks

5.1.2.5.4 Special Case 4: Event Marker Present

In the case of an event marker, the program automatically displayed the force-

displacement graph and prompted the user for input. This was done as a precaution, since event

markers may have indicated problems with the test or equipment. These issues may have resulted

in a false reading. Figure 5.10 demonstrates what an event marker was displayed as. However,

from visual inspection, the data used in the example appeared to be accurate. In this situation,

field notes from the operator were reviewed to explain the purpose of the event marker.

53
 Figure 5.10: Example of special case 4, event marker being triggered

5.1.2.6 Reporting Data

A combined CSV file containing relevant numeric information for all data sets within the

folder and individual images of the graphical output were created. The CSV file contained

numerical values that described key attributes from the data sets; attributes included were: data

directory, file name, force at initial soil failure, displacement required for soil failure, and slope

created by force/displacement. The CSV file of complied data was chosen to allow for efficient

data transfer to spreadsheets, databases, and statistical programs.

Images of each force-displacement graph were generated during the analysis and saved.

The graphs contained possible points of interest, as well as the point identified as the initial soil

failure. Although the images were not required for the analysis of the data, they offered an easy

way to verify that data reported in the CSV file accurately described the features of the data

collected from the VASST. Images were also produced to aid in further developments of the

program as more tests are conducted with the VASST. Figures 5.11 and 5.12 show examples of

54
the program’s graphical output for both the automated and manual decision making process

respectively.

Figure 5.11: Graphical output for program determined shear failure

55
 Figure 5.12: Graphical output for user determined shear failure

5.2 Development of Mohr-Coulomb Theory for Validation of VASST Measurements

The measurements taken with the VASST were explained theoretically with the modified

Mohr-Coulomb, Equation 1.2. It was observed that unlike the experimental apparatuses used by

Waldon (1977), Wu et al. (1979), and Endo (1980), the soil samples being sheared by the

VASST were not subjected to a vertical loading force. This lack of vertical loading, unique to the

VASST, led to the assumption that the normal force, 𝜎 in the modified Mohr-Coulomb equation,

would become negligible due to the mass of soil being the only vertical force acting at the

shearing plane. Under this assumption, the VASST was theoretically much less sensitive to

56
fictional forces present at the shearing plane compared to effects from cohesion and root

reinforcements on soil shear strength. Rohani and Baladi (1981) also stated that it can be

assumed that the tangent of the internal angle of friction is negligible, 𝑡𝑎𝑛ϕ = 0, for cohesive

soils. From Peterson (1988), the effects of soil moisture on shear strength could be accounted for

by changes in soil cohesion. Therefore, the modified Mohr-Coulomb equation hypothetically

developed for the forces present during VASST testing simplified to Equation 5.1.

𝜏 = 𝑐 + ∆𝑠 Eq. 5.1

Rohani and Baladi (1981) found that soil cohesion does not significantly contribute to shear

strength in granular soils. The lack of an applied normal force at the soil surface and insignificant

soil cohesion led to a further simplification of Equation 5.1 to form Equation 5.2 for granular

soils. This equation implied that root reinforcement within the soil matrix was the only parameter

affecting VASST measured shear strength at the soil surface in non-cohesive soils.

𝜏 = ∆𝑠 Eq. 5.2

To test the model developed for the VASST measurements in course grained soils, CBR

was determined using Equation 3.11, 𝐶𝐵𝑅 = 0.1691 ∗ (𝐶𝐼𝑉)1.695 (Al-Amoudi et al., 2002b).

The shear strength was then calculated from Equation 5.3, developed by Garcia and Thompson

(2004). To prove the theoretically developed equation, the VASST shear strength measurements

had to be highly correlated to the effects of roots in the soil matrix and insensitive to variations in

soil parameters.

𝜏 = 2.25 ∗ 𝐶𝐵𝑅 Eq. 5.3

For fine grained soils, CBR was estimated using Equation 3.12, 𝐶𝐵𝑅 = 𝑎 ∗ (𝐶𝐼)𝑏 (Shoop

et al., 2008), with coefficients from Table 3.1. Shear strength was again calculated using

57
Equation 5.3. This allowed for the comparison of the VASST measured shear force to the shear

force derived from CI.

5.3 Statistical Analysis

The entire statistical analysis for this research was performed using SAS/STAT(R) and

SAS/GRAPH software, Version 9.3 of the SAS System for Windows. (SAS Institute Inc., 2014)

5.3.1 Data Exploration

Collected field data was first used to generate several scatter plots in order to determine

which class variables would best describe changes in the VASST measurements due to soil type.

Scatter plots were produced using SAS 9.3 and the classes compared were soil plasticity (high,

medium, none), average particle size (coarse, fine), USCS soil type, and United States

Department of Agriculture (USDA) soil texture. It was determined that plasticity was able to

describe the behavior of the soil during VASST testing while maintaining a large enough sample

size per class to perform statistical analysis.

As a complement to the scatter plots, the same class variables were tested to determine

number of observations, means, standard deviations, minimums, and maximums for VASST

measurements. This information confirmed the viability of using plasticity as a class variable for

further analysis.

Once plasticity was determined to be an important descriptive class variable, box plots of

shearing forces were generated to confirm the differences in VASST measurements in soils with

different plasticities. Box plots contained several useful pieces of information such as 25th and

75th percentiles as well as the median and mean values for each class. Box plots were also used

to identify potential outliers in the data, points that lie outside of the box plot’s whiskers. If an

58
outlier was determined to be caused by human errors or equipment failures, the observation

related to the outlier was dropped from the data set.

To identify soil parameters related to VASST measurements, a correlation tool was used

to create a matrix of Pearson correlation coefficients. Additionally, the tool provided simple

statistics for each variable investigated. The Pearson correlation coefficients provided valuable

insight as to which soil and vegetation parameters had the largest impact on VASST

measurements per soil plasticity class. Furthermore, the tool was used to check for correlations

between soil shear strength, as calculated by methods described in section 5.2, and various

parameters measured by the VASST. The correlation results determined whether the VASST

could be used as a direct measure of soil shear strength.

Information from the exploratory statistics was applied to more suitable tests for

investigating soil and vegetation parameters that affect the measurements obtained using the

VASST. Survival analysis was chosen to carry out the statistical analysis as the dataset and

characteristics of shear failure share many qualities with clinical trials, experiments that often

investigate survival times. VASST induced shear failure, similar to clinical trials, was seen as a

measure of life expectancy but by applied force rather than event time. Survival analysis was

used to determine which soil and vegetation parameters have significant effects on soil strength

measured with the VASST. The survival analysis was conducted each level of soil plasticity

previously identified. Another advantage of using survival statistics was that unequal replications

were accounted for, a condition often occurring in clinical trials as well as this study. The

survival analysis provided valuable insight as to how the VASST performs in various soil

plasticities and how its measurements relate to the Mohr-Coulomb failure criterion.

59
CHAPTER 6: RESULTS AND DISCUSSION
A thorough investigation of the VASST’s capabilities as a tool for measuring vegetated

soil shear strength was conducted. The VASST measurements were compared to vegetation and

soil parameters to gain a better understanding of how the VASST’s performance varied under a

range of conditions. The existence of correlations between the VASST’s measurements, common

in-situ soil strength measurements, and calculated soil shear strength was also investigated to

determine its capabilities of measuring soil trafficability. Finally, survival analysis was

performed to determine which soil and vegetation parameters the VASST was sensitive to.

6.1 Descriptive Parameters of Soil and Vegetation Condition

Field and laboratory measured parameters were used to classify/quantify soil and

vegetation characteristics that would likely affect VASST testing results. The parameters were

categorized as either class variables or continuous variables. All SAS outputs from the statistical

analysis conducted during this research are found in Appendices C though G.

6.1.1 Class Variables

The statistical data set contained the following class variables: plasticity (high, medium,

and none), average particle size (coarse and fine), USCS soil type, and USDA soil texture. To

determine which class variable described the results by soil type, simple statistics and scatter

plots for each class variable were generated. These statistics were used to determine which class

variable to use. Scatter plots, found in appendix C, allowed for visual inspection of the data set to

identify groups of similar observations that could be accounted for by the class variables.

Similarly, the simple statistics provided summary tables that included number of observations,

60
mean, standard deviation, and maximum and minimum values associated with each level of each

class.

The scatter plots showed potential correlations between vegetation parameters, common

soil strength tests, and VASST measurements. The presence of a relationship between CI and

CIV as well as drop-cone values was visible in each scatter plot, a relationship that has been

documented in previous research. Also, the plots showed that trends exist between the VASST

force measurements and the various soil strength tests, the CIV in particular. Furthermore,

moderate correlations were visible between the VASST force and displacement measurements

and the vegetation quantification parameters: aboveground biomass, belowground biomass, and

root weight. It was also noted that soil moisture has the biggest effect on strength measurements

in high plasticity soil with little visible impact on non-plastic soils.

Comparing the simple statistics produced for each class variable tested helped determine

the appropriate variable to classify the behavior of different soil types during VASST testing.

When choosing the class variable to describe the soil type, it was important to capture how the

different soils failed without creating unequal class sizes.

The simple statistics by plasticity, shown in Table 6.1, indicated that non-plastic and

highly plastic soil classes contain nearly the same number of observations with very different

means and standard deviations. However, the medium-plasticity soils class only had 21 sets of

VASST measurements. The lack of observations increased the potential of Type II errors during

statistical analysis and was dropped from the remaining statistical analysis.

61
 Table 6.1: Simple statistics by plasticity

Simple statistics were also produced for average grain size to provide a summary of the

VASST measurements by coarse and fine particle sizes, shown in Table 6.2. When compared to

the classification by plasticity, the medium and high plasticity soils were combined in the fine

particle class. Although Tables 6.1 and 6.2 are very similar, observations made in the field

indicated that medium and high plasticity soils behaved quite differently when being sheared

with the VASST. Due to the differing behavior of these soils, it was determined that classifying

the soils by average grain size was inappropriate for the study.

Table 6.2: Simple statistics by average grain size

Although the methods described in the USCS were used to derive the class variables for

plasticity and average grain size, using USCS soil types created too many classes, as illustrated

62
in Table 6.3. The USCS classes accurately captured the presence of groups in the data set,

however the number of classes greatly and unevenly divided the observations. The small

observation numbers per class could have caused a Type II error when comparing multiple

continuous variables within a class. Therefore, the classification by USCS soil types was not

suitable for this study.

Table 6.3: Simple statistics by USCS soil type

The classes assigned by USDA soil texture also indicated the presence of groups in the

data set. Like the classes formed using the USCS soil type, the classes designated by USDA soil

texture created a large number of classes, as shown in Figure 6.4; this increased the risk of a

Type II error. Furthermore, Table 6.4 shows that sample sizes are very unequal.

63
 Table 6.4: Simple statistics by USDA texture

Based on the results from simple statistical analysis of each class variable, it was

determined that classifying the soils by plasticity was appropriate for the scope of this study.

Plasticity classes captured differences in VASST measurements without introducing many

classes which would have decreased the number of observations per class.

After simple statistics indicated plasticity as the proper class variable, box plots were

produced to visually inspect differences in the classification variables. The box plot of VASST

measured shear strength classified by plasticity, Figure 6.1, echoed the results of the simple

statistical analyses. The box plot showed that as plasticity decreased, variations in the VASST

measured shear strength decreased. Furthermore, the plot indicated the presence of outliers in the

64
high and non-plastic soils classes, likely a result of higher sampling rates when compared to

medium plasticity soils. Box plots for all class variables are found in Appendix D.

Figure 6.1: Box plots of VASST measured shear strength by plasticity class

Scatter plots classified by plasticity were also produced to verify the existence of trends

for vegetation parameters and soil strength measurements compared to VASST measured soil

strength. Additional, parameters from the VASST, force, displacement, and slope, were

compared to observe trends within plasticity.

Figure 6.2 illustrates how vegetation parameters form groups when compared to VASST

measured soil shear strength. General trends for high and medium plasticity soils showed that as

all vegetation measures increase so did VASST measured shear strength. However, there was

high variability in individual observations. VASST strength measurements in non-plastic soils

appeared less sensitive to root mass and only moderately sensitive to belowground biomass. On

65
the other hand, VASST shear strength measurements tended to increase as aboveground biomass

increased for all soil plasticity classes.

Figure 6.2: Vegetation parameters compared to VASST measured soil strength, classified by
plasticity

Similarly, it was noted that when comparing the VASST measured soil strength to other

common soil strength measurements, groups of observations coincided with soil plasticity levels.

Figure 6.3 illustrates the trends found between CI, CIV, drop-cone penetration, and VASST

66
shear strength measurements. The relationships showed that the VASST was more sensitive to

plasticity, as indicated by the strong emergence of plasticity groups, than the other three soil

strength measurements.

Figure 6.3: Common soil strength measurements compared to VASST measured soil strength,
classified by plasticity

67
6.1.2 Continuous Variables

Continuous variables were parameters measured using a continuous scale and included

measurements quantifying soil particle diameters, root weight, belowground biomass,

aboveground biomass, CI, SMC, CIV, drop-cone penetration, and bulk density. Variables also

included measurements obtained from the VASST (maximum force, displacement, and force-

displacement slope). These parameters were then placed into two categories, measures of soil

strength and variables affecting soil strength. This was done to compare the VASST

measurements against other measures of soil strength and test which soil and vegetation

parameters have the greatest impact on VASST measurements.

6.2 Correlations of the VASST Maximum Force Measurement to Common In-situ Soil

Strength Tests

The VASST maximum force measurements were compared to other in-situ measurements

of soil strength, CI, CIV, and drop-cone values, to investigate possible correlations between

measurements. Previous research (Koch et al., 2010; Shoop et al., 2008) showed strong

correlations between the mentioned common in-situ measurements and land trafficability,

research that led to the development of the VASST. Correlations between the VASST and

previously mentioned measurements are presented in Tables 6.5 through 6.8.

6.2.1 High Plasticity Soils

In high plasticity soils, the Pearson correlation coefficients, illustrated in Table 6.5,

showed that VASST maximum force measurements were strongly correlated with CIV and CI.

Strong correlations to other soil strength measurements that are used to quantify soil trafficability

suggest that the VASST can be used as a tool to quantify trafficability.

68
 Table 6.5: Pearson correlation coefficients for high plasticity soils

However, the drop cone was weakly correlated with the VASST when compared to

correlations between drop-cone values to CI and CIV. This alluded to differences in soil

parameter sensitivities between the VASST and drop-cone. Figure 6.4 presents the data used to

calculate the correlation coefficients for high plasticity soils and shows the distribution of data

for common in-situ measurements.

69
 Figure 6.4: Observations used for correlation coefficients of high plasticity soils

6.2.2 Medium Plasticity Soils

The medium plasticity soils, according to the Pearson correlation coefficients shown in

Table 6.6, suggested that VASST maximum force measurements were moderately correlated

with CIV, drop-cone values, and CI. However, p-values revealed that the confidence of this

correlation is weak. With only 23 samples used to predict the correlation coefficients, low

sample numbers for medium plasticity soils severely limited statistical rigor.

70
 Table 6.6: Pearson correlation coefficients for medium plasticity soils

The scatter plot, shown in Figure 6.5, emphasized the limitations of the medium plasticity

class. It revealed that only three observations were used for CIV, drop cone values, and CI

measurements. Unlike the high plasticity and non-plastic soils, the data was so limited that well-

distributed data was not possible within the medium plasticity class. Conclusions on the

VASST’s performance in medium plasticity could not be drawn with the current data set.

71
 Figure 6.5: Observations used for correlation coefficients of medium plasticity soils

6.2.3 Non-Plastic Soils

As with high plasticity soils, the Pearson correlation coefficients for non-plastics showed

that VASST maximum force measurements were moderately correlated with CIV and CI and

weakly correlated with drop-cone (Table 6.7). The correlation coefficients in non-plastic soil

compared to high plasticity soil indicates a limitation in the soil types the VASST can be used to

quantify trafficability. Many of the current trafficability models are based on CI; a stronger

72
correlation between the VASST maximum force and CI would be expected if the VASST was

measuring the same aspect of soil trafficability.

Table 6.7: Pearson correlation coefficients for non-plastic soils

The data used to calculate the correlation coefficients for non-plastic soils presented in

Figure 6.6, showed the distribution of data for common in-situ measurements. Similar to the

observations for high plasticity soils, the data points were fairly well distributed across the range

of the data with the exception of the VASST measurements that are left skewed.

73
Figure 6.6: Observations used for correlation coefficients of non-plastic soils

74
6.3 Comparison of VASST Measurements to Estimated Soil Shear Strength

Pearson correlation coefficients for the estimated soil shear strength, calculated for each of

the three plasticity classes were compared (Tables 6.8 through 6.10). Estimated soil shear

strength was moderately to strongly correlated to all of the VASST measured shearing

parameters for each plasticity class. Only soil displacement in high plasticity and non-plastic

soils was not correlated.

In highly cohesive soils, the VASST maximum force applied was strongly correlated

with the calculated shear strength, as shown in Figure 6.6. The correlation indicated that the

VASST’s applied force measurements were a potential indicator of in-situ soil shear strength of

high plasticity soils. Displacement distances were also correlated with the calculated shear

strength, likely a result of organic fibers increasing the tensile strength of the soil matrix. Due to

the elastic nature of organic fibers, they achieved their maximum tensile strength after being

stretched as described in Waldron (1977).

Table 6.8: Pearson correlation coefficients for high plasticity soils

75
 For medium plasticity soils, the Pearson correlation coefficients indicated that shearing

parameters were moderately to strongly correlated to calculated soil shear strength (Table 6.9).

However, the 21 observations that occurred in medium plasticity soils were limited when

compared to the 121 and 141 observations for highly plastic and non-plastic soils respectively.

The lack of observations was reflected by the limited confidence in rejecting the null hypothesis

(Table 6.9).

Table 6.9: Pearson correlation coefficients for medium plasticity soils

The analysis highlighted a lack of correlation between calculated soil shear strength and

displacement for non-cohesive soils measured by the VASST, as illustrated in Table 6.10. When

compared to cohesive soils, the correlation difference was a result of the VASST test spikes’

interaction with the two types of soil when a load was applied. As the spikes applied force to

cohesive soils, it was observed that the soil surrounding all five spikes was displaced, as a unit,

forming a single continuous shearing surface. On the other hand, as non-cohesive soils were

subjected to lateral forces via the spikes, soil particles moved around individual spikes; this

76
resulted in small, consistent displacement values. Therefore, the short, uniform displacement

values observed in non-cohesive soil did not relate to the calculated shear strength.

Table 6.10: Pearson correlation coefficients for non-plastic soils

6.4 Survival Analysis of VASST Measured Shear Strength

Survival analysis was performed on each plasticity class, of the data set, to identify key

soil and vegetation parameters influencing VASST soil shear strength. The limited number of

observations for medium plasticity soil did not result in an accurate depiction of the VASST’s

performance. Medium plasticity soil observations were not used in the final analysis.

The parameters initially analyzed as affecting the VASST measured shear strength were

root weight, belowground biomass, aboveground biomass, soil moisture, 30% finer particle

diameter, 60% finer particle diameter, and soil bulk density. Proc lifetest highlighted that the

addition of bulk density as a parameter caused confounding among the variables. This resulted in

no parameters being significant predictors of the VASST measured shear strength in high

plasticity soil. The significance of parameters in non-plastic soils was unaffected by the inclusion

of bulk density, so bulk density was removed from the survival analysis as a predictor. The

77
complete results of the survival analysis both with and without bulk density as a predicting

parameter, can be found in Appendices F and G.

6.4.1 High Plasticity Soils

The survival analysis for high plasticity soils showed that when covariates were

individually evaluated aboveground and belowground biomass were the only significant

variables affecting VASST force measurements (Table 6.11). However, when analyzing the

covariates in a forward stepwise sequence, as shown in Table 6.12, the significant parameters

became aboveground biomass, 60% finer particle diameter, root weight, and soil moisture

content.

Table 6.11: Rank test for the association of VASST shear force with covariates in high plasticity
soils

78
 Table 6.12: Stepwise test results of associated covariates in high plasticity soils

The significant effect of aboveground biomass on the VASST was the result of the

shallow depth at which the soil was being sheared. During field testing, it was observed that the

root crowns did not shear through their interior; in other words, the root crowns were displaced,

not broken. This caused an increased surface area of the shearing plane. The root crowns’ effect

on the area of the shearing plane was directly related to the amount of aboveground biomass and

had a large impact on the VASST force measurements. This observation was reflected in the data

with aboveground biomass being the leading contributor to soil shear strength near the surface.

The results implied that aboveground biomass had the most significant effect on lateral shearing

resistance of all the parameters tested in high plasticity soils.

Though the measurement of 60% finer particle diameter was not significant in the

individual association analysis, it was significant in the forward stepwise sequence. The

significance of the 60% finer particle diameter in the stepwise sequence was likely the result of

the parameter’s association with cohesion rather than particle size. As described in literature,

cohesion is the major contributing factor for soil shear strength in fine grained soils (Rohani et

al., 1981). It is also important to note why the 60% finer particle diameter was not significant in

the individual association test but became significant in the stepwise test. This difference
79
occurred because the 60% finer particle diameter described a soil parameter independent of

aboveground biomass. For the same reason, belowground biomass went from being significant as

an individual parameter to being insignificant when the parameters were analyzed as a group

(Table 6.12). This change in significance is due to the dependence of belowground biomass on

aboveground biomass.

As expected, the amount of roots, as measured by dry weight for this study, had a

significant effect on the soil shear strength obtained with the VASST. The roots acted as

reinforcing members within the soil matrix effectively adding tensile strength which resisted

shearing.

Lastly, SMC played a significant role in VASST shear strength measurements. The effect

was the result of water within the soil matrix acting as a lubricant between soil particles, thus

requiring less force to be displaced. Increased soil water content also reduced adhesion between

roots and soil particles. The reduction in adhesion caused roots to slip through the soil matrix

rather than be loaded to their ultimate tensile strength. Increased SMC resulted in reduced

cohesion between soil particles and adhesion between soil particles and roots, which caused a

decrease in the soil’s shear strength.

Concluding the results of the VASST testing in high-plasticity soils, the VASST was

affected by the same soil and vegetation parameters as the direct shear tests performed in

vegetated soils during previous studies. The forward stepwise survival analysis confirmed that

VASST performance was significantly associated with aboveground biomass, 60% finer particle

diameter, root weight, and soil moisture and implied that the VASST can be successfully used as

a device to measure in-situ soil shear strength of highly plastic, vegetated soils.

80
6.4.2 Non-Plastic Soils

The survival analysis for non-plastic soils showed that when covariates were individually

evaluated, 60% finer particle diameter, 30% finer particle diameter, aboveground biomass, and

belowground biomass were significantly associated with VASST force measurements (Table

6.13). In order for the theoretically developed equation, 𝜏 = ∆𝑠 (Equation 5.2), to be valid, the

only significant parameters in non-plastic soils should have been related to vegetation. Statistical

analysis of individual soil and vegetation parameters proved the influence of particle size on

VASST measurements and suggested that the Equation 5.2 does not hold true with the

assumption of a zero normal force at the shearing plane.

Table 6.13: Rank test for the association of VASST maximum force with covariates in non-
plastic soils

Field observations offered an explanation as to why the normal force was not zero. As the

test spikes were dragged through the soil, non-plastic soils behaved similarly to a high viscosity

fluid with increased internal pressures immediately forward of the spikes. These internal

pressures created normal forces acting on the individual soil particles and were present in the

direction parallel to the spikes’ motion. The normal forces among particles increased frictional

forces within the soil matrix, a soil property directly related to internal angle of friction.

81
 When reviewing the forward stepwise sequence of covariates in Table 6.14, the

significant parameters were 60% finer particle diameter, aboveground biomass, and root weight,

parameters also significant in describing high plasticity soils. But, unlike highly plastic soils, the

SMC was not identified as an associated covariate of the VASST shearing force.

Table 6.14: Stepwise test results of associated covariates in non-plastic soils

The significance of the 60% finer particle diameter in coarse grained soils was likely the

parameter’s association with internal angle of friction rather than particle size. The 30% finer

particle diameter was not identified as significant by the stepwise analysis because of strong

correlation between 60% and 30% diameters. This meant that only one parameter quantifying

particle size was needed to account for the effect of internal angle of friction associated with soil

shear strength.

Similar to high plasticity soils, the VASST force measurements in non-plastic soils were

affected by aboveground biomass. The shallow depth at which the VASST sheared the soil

explained this result. Once again, root crowns remained intact during testing, causing the surface

area of the shearing plane to increase. The increased surface area of the shearing plane directly

impacted the amount of applied force required for the VASST to shear the soil.

82
 Roots within the soil matrix, though not significant during individual associated covariate

analysis, were considered significant during forward stepwise sequence analysis. The result was

due to roots’ tensile strength increasing the tensile strength of the soil matrix as it was being

sheared. Comparing the roots’ individual effect in non-plastic soil to highly plastic soils, it was

observed that the roots had a larger effect on shear strength in high plasticity soil. Roots tended

to slip through the non-plastic soil, rather than break, indicating that the roots did not reach

ultimate tensile strength before shear failure occurred. This scenario mirrors root loading in high

moisture content, high plasticity soils.

Unlike VASST shear measurements taken in high plasticity soils, SMC did not

significantly affect soil shear strength in the non-plastic soils used for this study. The soils used

for non-plastic soil tests were sandy with negligible cohesion. SMC mainly reduced cohesion

which was assumed to be zero in sand as indicated in research performed by Rohani and Baladi

(1981).

Summarizing the results of the VASST testing in non-plastic soils, the VASST

measurements of vegetated soil shear strength could not be described using the theoretical

modified Mohr-Coulomb equation developed within this study. The assumption of a zero normal

force was inaccurate as proven by the significance of particle size within the survival analysis.

Although the developed equation did not produce an accurate description of the VASST

measured shear strength, the VASST should still be considered effective at measuring soil shear

strength. The significant parameters from the forward stepwise sequence test, 60% finer particle

diameter, root weight, and aboveground biomass, are important factors in the modified Mohr-

Coulomb proposed by Waldron (1977) when 60% finer particle diameter is considered as being

correlated to internal angle of friction.

83
CHAPTER 7: CONCLUSIONS
The VASST was tested at five geographic locations across five ecoregions. Soils ranged

from high plasticity inorganic clays to non-plastic sands. Soil strength measurements were taken

concurrent with the VASST to evaluate the use of the VASST for trafficability studies. Using a

static penetrometer, drop-cone penetrometer, and Clegg impact tester, tools commonly used for

in-situ trafficability quantification, VASST performance was quantified. Measurements of soil

and vegetation parameters were taken as well to describe the soil and vegetation characteristics;

these measurements included bulk density, soil moisture, and aboveground and belowground

biomass. Laboratory analyses of field samples quantified soil parameters for Atterburg limits and

particle size descriptors, 60% finer diameter and 30% finer diameter, textural analysis, and above

and belowground biomass.

VASST field data had to be interpreted to extract useful information on the force required

to shear the soil. Previous work done by McDonald (2012) visually interpreted the VASST data

one test at a time by plotting force versus displacement. Determining the force and displacement

at which the soil was initially sheared was also performed visually. To quickly analyze the

VASST data and ensure that analysis results were repeatable, a Python program was developed

to automate the process of extracting important test parameter data at the point of shearing.

Although the program was intended to be completely automated, the complexity of the data

caused the program to output erroneous results in a few instances. The special cases of inaccurate

shearing information were identified by the program but were complex in their solutions. The

final program was designed to be semi-automated with user selected shear for failure for special

cases identified. Force-displacement plots were automatically generated for each VASST test to

provide the user with a method of visual inspection.

84
 With data from field and laboratory testing compiled with VASST data outputted by the

Python program, statistical analyses were performed using SAS 9.3 to interpret how the VASST

could be implemented as a device to measure in-situ vegetated soil shear strength. Simple

statistical analyses and visual inspection of the data revealed that a class variable needed to be

implemented as the VASST performed quite differently in various types of soils. Due to a

limited number of observations, classification by USCS plasticity classes was used to describe

variations in the VASST performance in various soil types; the use of these classes also kept

sample sizes reasonable for high and non-plastic soils.

Once a descriptive class variable was established, the shearing force data obtained from

the VASST was compared against a static penetrometer, drop-cone penetrometer, and Clegg

impact tester, tools that have already been used to quantify land trafficability in previous

research. Strong correlations between VASST measured force and CI as well as CIV in highly

plastic and non-plastic soils suggested that the VASST can also be implemented to quantify

trafficability. The measurements of force, displacement, and force-displacement slope from the

VASST were also compared to τ, calculated shear strength, for correlations. It was found that

moderate to strong correlations exist between τ and VASST measured force as well as VASST

measured force-displacement slope for highly and non-plastic soils. No statements can be made

on the application of the VASST in medium and low plasticity soils due to a lack of observations

in this study.

An investigation into the soil and vegetation parameters which affect the measurements

of the VASST was carried out. The test procedure was performed within each soil plasticity class

and determined that in high plasticity soils, vegetation was a key dependent variable of force

measurements obtained with the VASST. On the other hand, the key dependent variables

85
affecting VASST measured shear force in non-plastic soils were more related to soil particle size

than biomass. However, biomass variables, aboveground and belowground, were still found to be

significant.

When dependent variables were analyzed in a forward stepwise manner, particle size

(described by 60% finer by diameter), root weight, and aboveground biomass were key

parameters affecting the soil shear strength as measured by the VASST in both highly plastic and

non-plastic soils. The exception was the significances of SMC in high plasticity soils which was

inversely related to shear strength but not significant in non-plastic soils.

Although the derived models (Equations 5.1 and 5.2) were over-simplified and failed to

describe the VASST’s measurements of soil shear strength, the VASST was sensitive to biomass

and soil parameters. These parameters affect soil shear strength and lead to the conclusion that

the VASST measurements were related to soil trafficability. However, the high number of

VASST tests within each sample revealed that the VASST measurements were highly variable.

This meant that one to two samples would be inadequate for describing soil trafficability. Also,

the VASST was a large, cumbersome apparatus that required assembly in the field. While it was

suitable for research purposes, the VASST could not be rapidly deployed, a key design flaw

when performing quick measurements on an as needed basis is desired.

86
CHAPTER 8: RECOMMENDATIONS FOR FUTURE WORK
After completing this study, the following are suggestions for further research into the

application of the VASST as a tool for measuring in-situ shear strength of vegetation and soil for

the purposes of quantifying land trafficability:

 A continuation of this research of this study to include observations for low plasticity and

more observations for medium plasticity soils would aid in describing the performance of

the VASST across a wider range of soil types. The addition of data for a wider range of

plasticity could also result in plasticity being implemented as a continuous variable in

future statistical analysis.

 Data on soil cohesion measured with a shear vane was not available at the time of this

study but would have been a good soil parameter to investigate. The modified Mohr-

Coulomb failure criterion is highly dependent on cohesion, and conclusions drawn in this

study about high plasticity soils are limited because cohesion was not directly measured.

 Further research on the effects of SMC on the VASST measured shear force should be

conducted. Because this study was focused on soil type, little variation in water content

was observed within soil type and location, a potential oversight in fully understanding

the soil shear strength data obtained from the VASST.

 Lastly, the VASST may be a useful tool to predict vehicle impacts, especially tracked

vehicles that produce a lateral shear at the soil surface during acceleration, deceleration,

and turning. Future research into the VASST’s accuracy in predicting vehicle impacts

could be best achieved by comparing VASST measurement to quantifiable impacts such

as rut depth and parameters that describe the vehicle creating the impacts.

87
CHAPTER 9: REFERENCES
Abe, K., & Ziemer, R. R. (1991). Effect of tree roots on shallow-seated landslides. (USDA

Forest Service General Technical Report PSQ-GTR-130). Washington, DC: U.S.

Government Printing Office.

Al-Amoudi, O. S. B., Asi, I. M., Wahhab, H. I. A., & Khan, Z. A. (2002b). Clegg hammer-

California-Bearing ratio correlations. Journal of Materials in Civil Engineering, 14(6),

512-523.

ASTM Standard D4318, 2010, “Standard Test Method for Liquid Limit, Plastic Limit, and

Plasticity Index of Soils,” ASTM International, West Conshohocken, PA, 2010.

doi: 10.1520/D4318, www.astm.org

ASTM Standard D 4429, 2009a, “Standard Test for CBR (California Bearing Ratio) of Soils in

Place.” ASTM International, West Conshohocken, PA, 2009, DOI: 10.1520/D4429-09A,

www.astm.org.

ASTM Standard D422, 2007, “Standard Test Methods for Particle-Size Analysis of Soils,”

ASTM International, West Conshohocken, PA, 2007, DOI: 10.1520/D0422-63R07,

www.astm.org.

ASTM Standard E100, 2010, “Standard Specification for ASTM Hydrometers,” ASTM

International, West Conshohocken, PA, 2010, DOI: 10.1520/E0100-10, www.astm.org.

ASABE Standard 313.2, 1992, “Soil Cone Penetrometer,” ASAE Standards. St. Joseph, MI, 611.

88
Braunack, M. (1986). The residual effects of tracked vehicles on soil surface properties. Journal

of Terramechanics, 23(1), 37-50.

Diersing, V., Shaw, R., & Tazik, D. (1992). US army land condition-trend analysis (LCTA)

program. Environmental Management, 16(3), 405-414. doi:10.1007/BF02400080

Dilustro, J. J., Collins, B. S., Duncan, L. K., & Sharitz, R. R. (2002). Soil texture, land-use

intensity, and vegetation of Fort Benning upland forest sites. Journal of the Torrey

Botanical Society, 129(4), 289-297.

Directorate of Public Works Environmental Division. (2011). Supplemental environmental

assessment for the repair of training area 41 hilltop access trail (HAT) on Fort Hood,

Texas. Unpublished manuscript.

Endo, T., & Tsuruta, T. (1969). The effect of the tree's roots upon the shear strength of soil.

(1968 annual report of the Hokkaido branch). Sapporo, Japan: Forest Experiment Station.

Endo, T. (1980). Effect of tree roots upon the shear strength of soil. Japan Agricultural Research

Quarterly, 14, 112-115.

Evett, S. R. (2003). Soil water measurement by time domain reflectometry. Encyclopedia of

Water Science, 894-898.

Fredlund, D. G., & Morgenstern, N. R. (1977). Stress state variables for unsaturated soils.

Journal of Geotechnical and Geoenvironmental Engineering, 103 (ASCE 12919)

89
Fredlund, D. G., Xing, A., Fredlund, M. D., & Barbour, S. (1996). The relationship of the

unsaturated soil shear to the soil-water characteristic curve. Canadian Geotechnical

Journal, 33(3), 440-448.

Fredlund, D., Morgenstern, N., & Widger, R. (1978). The shear strength of unsaturated soils.

Canadian Geotechnical Journal, 15(3), 313-321.

Freitag, D. R., & Richardson, B. Y. (1968). Application of Trafficability Analysis to Forestry (

No. AEWES-Misc-Paper-4-9595). Army Engineer Waterway Experiment Station.

Vicksburg, MS.

Fulton, A. (2013). Assessing belowground biomass changes following land management and

vehicle disturbance. (Unpublished Master of Science). University of Illinois Urbana-

Champaign, IL

Garcia, G. & Thompson, M. R. (2004) Subgrade Strength/Stiffness Evaluation (Project IHR-30,

Technical Note 6) University of Illinois Urbana-Champaign, IL

Godwin, R. J., Warner, N. L., & Smith, D. L. O. (1991). The development of a dynamic drop-

cone device for the assessment of soil strength and the effects of machinery traffic.

Journal of Agricultural Engineering Research, 48(2), 123-131.

doi: http://dx.doi.org/10.1016/0021-8634(91)80009-4

Gray, D. H., & Leiser, A. T. (1982). Biotechnical slope protection and erosion control. New

York, NY: Van Nostrand Reinhold Company.

90
Gray, D. H., & Sotir, R. B. (1996). Biotechnical and soil bioengineering slope stabilization: A

practical guide for erosion control. New York, NY: John Wiley & Sons.

Gray, D. H. (1974). Reinforcement and stabilization of soil by vegetation. Journal of the

Geotechnical Engineering Division, 100(6), 695-699.

Greenway, D. (1987). Vegetation and slope stability. Anderson M. G., Richards K. S. (Eds.),

Slope Stability. New York, NY: John Wiley & Sons Ltd.

Guretzky, J. A., Fehmi, J. S., & Anderson, A. B. Cattle Grazing and Tracked Vehicle Training

on Central and Southwestern U.S. Army Lands (CERL-05-33). Engineering Research and

Development Center. Champaign, IL: Construction Engineering Research Lab.

Holtz, R. D., Kovacs, W. D., & Sheahan, T. C. (2010). An introduction to geotechnical

engineering. Stark H., Rand K. (Eds.). Upper Saddle River, NJ: Pearson Education,

Limited.

Jones, D., & Kunze, M. (2004). Guide to sampling soil compaction using hand-held soil

penetrometers. Center for Environmental Management of Military Lands (CEMML), Fort

Collins, CO: Colorado State University

Kansas Sampler Foundation. (2007). Tall grass prairie national preserve and the flint hills,

chase county. Retrieved from

http://www.kansassampler.org/8wonders/8wondersofkansas-view.php?id=17

91
Kato, N., & Shuin, Y. (2002). Effects of root systems on the shear strain in a direct shear

apparatus. Proceedings of International Congress INTERPRAEVENT 2002 in the Pacific

Rim, Matsumoto, Japan. , 2 525-530.

Koch, D. J., Gertner, G. Z., Svendsen, N. G., Howard, H. R., Horner, D. A., & Sullivan, P. M.

(2010). Cumulative Interactions for Military Vehicle Impact Assessment. Pittsburgh,

Pennsylvania, June 20- June 23, 2010 1008489. doi: 10.13031/2013.35790

www.asabe.org

MacDonald, K. A., Coutermarsh, B. A., Shoop, S. A., & Burch, W. T. (2012). A new method to

measure vegetated soil shear strength using the vegetation and soil shear tester (VASST).

Proceedings of 12th European Regional Conference of the International Society for

Terrain-Vehicle Systems, Pretoria, South Africa.

MacDonald, K. A. & Shoop, S. A. (2013). Validation of the vegetation and soil shear tester

(VASST) with existing soil strength instruments. Proceedings of 7th Americas Regional

Conference of the International Society for Terrain-Vehicle Systems, Tampa, Florida.

Mathur, T., & Coghlan, G. (1987). The use of the Clegg impact tester in managing and designing

aggregate-surfaced roads. Transportation Research Record, 1(1106)

Meyer, M., & Knight, S. (1961). Trafficability of Soils: Soil Classification (AEWES-TM-3-240-

SUPPL-16). Army Engineer Waterway Experiment Station. Vicksburg, MS.

92
Muller, L., Tille, P., & Kretschmer, H. (1990). Trafficability and workability of alluvial clay

soils in response to drainage status. Soil and Tillage Research, 16(3), 273-287.

doi: http://dx.doi.org/10.1016/0167-1987(90)90101-I

Nam, J. S., Park, Y. J., & Kim, K. U. (2010). Determination of rating cone index using wheel

sinkage and slip. Journal of Terramechanics, 47(4), 243-248.

doi: http://dx.doi.org/10.1016/j.jterra.2010.02.002

Natural Resources Conservation Service. (2001). Chapter 4: Bulk Density Test. Soil Quality Test

Kit Guide (pp. 10). Washington, DC: United States Department of Agriculture.

Peterson, R. W. (1988). Interpretation of triaxial compression test results on partially saturated

soils. ASTM Special Technical Publication, (977), 512-538.

Pollen, N., & Simon, A. (2006). Temporal and spatial variability in the root-reinforcement of

streambanks: incorporating variations in soil shear strength and soil moisture into the rip

root model. Proceedings of the 2006 World Environmental and Water Resources

Congress, 87-91.

Pollen, N., & Simon, A. (2005). Estimating the mechanical effects of riparian vegetation on

stream bank stability using a fiber bundle model. Water Resources Research, 41(7): 25

doi: 10.1029/2004WR003801

Pollen, N. (2007). Temporal and spatial variability in root reinforcement of streambanks:

Accounting for soil shear strength and moisture. Catena, 69(3), 197-205.

doi: http://dx.doi.org/10.1016/j.catena.2006.05.004

93
Rohani, B., & Baladi, G. (1981). Correlation of Mobility Cone Index with Fundamental

Engineering Properties of Soil (Miscellaneous Paper SL-81-4). Army Engineer Waterway

Experiment Station. Vicksburg, MS.

Rostam-Abadi, F., Chaika, M., Lambrecht, S., & Namburu, R. (1993). M1, IPM1, and M1A1

(Through November 1990) Abrams Tank Lifting Provisions: Nonlinear Finite Element

Analysis of Front Lifting Eye (TARDEC-TR-93-13591). U.S. Army Tank-Automotive

Command Research, Development, and Engineering Center. Warren, Michigan.

Semen, P. M. (2006). A Generalized Approach to Soil Strength Prediction with Machine

Learning Methods (ERDC/CRREL-TR-06-15). Engineering Research and Development

Center. Hanover, NH: Cold Regions Research and Engineering Lab

Shoop, S. A. (1993). Terrain Characterization for Trafficability (CRREL-93-6). Engineering

Research and Development Center. Hanover, NH: Cold Regions Research and

Engineering Lab

Shoop, S. A., Diemand, D., Wieder, W. L., Mason, G., & Seman P. (2008). Predicting

California Bearing Ratio from Trafficability Cone Index Values. (ERDC/CRREL-TR-08-

17). Engineering Research and Development Center. Hanover, NH: Cold Regions

Research and Engineering Lab.

Soil Survey Staff, Natural Resources Conservation Service, United States Department of

Agriculture. Web Soil Survey. Available online at http://websoilsurvey.nrcs.usda.gov/

accessed 2/2/2014

94
Sorrie, B. A., Eerden, B. V., & Russo, M. J. (1997). Noteworthy plants from Fort Bragg and

Camp MacKall, North Carolina. Castanea, 62(4), 239-259. doi: 10.2307/4034041

Terzaghi, K. (1936). The shear resistance of saturated soils. Proceedings from 1st International

Conference on Soil Mechanics and Foundation Engineering, Cambridge, United

Kingdom. (Vol. 1, pp. 54-56).

Thorne, C. R. (1990). Effects of vegetation on riverbank erosion and stability. Thornes J. B.

(Ed.) Vegetation and Erosion (pp. 125-144). Chinchester, West Sussex: John Wiley &

Sons Ltd.

U.S. Department of the Army. (1997). Army ranges and training land program. (Army

Regulation 210-21). Washington D.C.: Headquarters, Department of the Army.

U.S. Department of the Army. (2005). The army sustainable range program. (Army Regulation

300-19). Washington D.C.: Headquarters, Department of the Army.

Waldron, L. (1977). The shear resistance of root-permeated homogeneous and stratified soil. Soil

Science Society of America Journal, 41(5), 843-849.

Wismer, R. D., & Luth, H. J. (1972). Off-road traction prediction for wheeled vehicles. (ASAE

Paper No. 72-619). St. Joseph, MI: ASAE.

Wong, J. Y. (2001). Theory of ground vehicles (3rd ed.). New York, NY: John Wiley & Sons

Ltd.

Wootan, Z. (2011, March/April). Blackland prairie harvest. The Reverchon Naturalist, pp. 1.

95
Wu, T. H., Beal, P. E., & Lan, C. (1988). In-situ shear test of soil-root systems. Journal of

Geotechnical Engineering, 114(12), 1376-1394.

Wu, T. H., McKinnell III, W. P., & Swanston, D. N. (1979). Strength of tree roots and landslides

on Prince of Wales Island, Alaska. Canadian Geotechnical Journal, 16(1), 19-33.

Zhang, C., Chen, L., Liu, Y., Ji, X., & Liu, X. (2010). Triaxial compression test of soil–root

composites to evaluate influence of roots on soil shear strength. Ecological Engineering,

36(1), 19-26.

96
APPENDIX A: TEST AND SAMPLE QUANTITIES BY LOCATION
Table A.1: Test and sample quantities at Champaign, Illinois

Location: Champaign, Illinois

Above Ground
Below Ground
Penetrometer
Classification

Clegg Impact

Bulk Density
Drop Cone
Hammer

Biomass

Biomass
VASST
TDR
Soil
Site Rep Soil Type
Inorganic Clays, High
1 1 Plasticity 1 1 5 3 10 9 1 1 1
Inorganic Clays, High
2 1 Plasticity 1 5 5 3 10 8 1 1 1
Inorganic Clays, High
2 2 Plasticity 1 5 5 3 10 7 1 1 1
Inorganic Clays, High
2 3 Plasticity 1 5 5 3 10 8 1 1 1
Inorganic Clays, High
3 1 Plasticity 1 5 5 3 10 7 1 1 1
Inorganic Clays, High
3 2 Plasticity 1 5 5 3 10 8 1 1 1
Inorganic Clays,
3 3 Medium Plasticity 1 5 5 3 10 7 1 1 1
Inorganic Clays, High
4 1 Plasticity 1 5 5 3 10 8 1 1 1
Inorganic Clays, High
4 2 Plasticity 1 5 5 3 10 8 1 1 1
Inorganic Clays, High
4 3 Plasticity 1 5 5 3 10 8 1 1 1
Inorganic Clays, High
5 1 Plasticity 1 5 5 3 10 8 1 1 1
Inorganic Clays, High
5 2 Plasticity 1 5 5 3 10 8 1 1 1
Inorganic Silts and
Organic Clays, High
5 3 Compressibility 1 5 5 3 10 8 1 1 1
Totals: 13 61 65 39 130 102 13 13 13

97
 Table A.2: Test and sample quantities at Ft. Benning, Georgia

Location: Ft. Benning, Georgia

Above Ground
Below Ground
Penetrometer
Classification

Clegg Impact

Bulk Density
Drop Cone
Hammer

Biomass

Biomass
VASST
TDR
Soil
Site Rep Soil Type
Poorly Graded Sand-
1 1 Clayey Sand 1 6 9 5 5 6 1 1 1
Poorly Graded Sand-
1 2 Clayey Sand 1 6 9 5 5 8 1 1 1
Poorly Graded Sand-
1 3 Clayey Sand 1 6 9 5 5 8 1 1 1
Poorly Graded Sand-
2 1 Clayey Sand 1 6 9 5 5 8 1 1 1
Poorly Graded Sand-
2 2 Clayey Sand 1 6 9 5 5 8 1 1 1
Poorly Graded Sand-
2 3 Clayey Sand 1 6 9 5 5 8 1 1 1
Poorly Graded Sand-
3 1 Clayey Sand 1 6 9 5 5 8 1 1 1
Poorly Graded Sand-
3 2 Clayey Sand 1 6 9 5 5 7 1 1 1
Poorly Graded Sand-
3 3 Clayey Sand 1 6 9 5 5 7 1 1 1
Totals: 9 54 81 45 45 68 9 9 9

98
 Table A.3: Test and sample quantities at Ft. Riley, KS

Location: Ft. Riley, Kansas

Above Ground
Below Ground
Penetrometer
Classification

Clegg Impact

Bulk Density
Drop Cone
Hammer

Biomass

Biomass
VASST
TDR
Soil
Site Rep Soil Type
Inorganic Clays, High
1 1 Plasticity 1 1 1 1 3 8 1 1 1
Inorganic Clays, High
1 2 Plasticity 1 1 1 1 3 7 1 1 1
Inorganic Clays, High
1 3 Plasticity 1 1 1 1 3 1 1 1 1
Totals: 3 3 3 3 9 16 3 3 3

Table A.4: Test and sample quantities at Ft. Bragg, North Carolina

Location: Ft. Bragg, North Carolina

Above Ground
Below Ground
Penetrometer
Classification

Clegg Impact

Bulk Density
Drop Cone
Hammer

Biomass

Biomass
VASST
TDR
Soil

Site Rep Soil Type

Poorly Graded Sand-
1 1 Silty Sand 1 5 5 5 9 6 1 1 1
Poorly Graded Sand-
1 2 Silty Sand 1 5 6 5 9 8 1 1 1
Poorly Graded Sand-
1 3 Silty Sand 1 5 6 5 9 10 1 1 1
2 1 Poorly Graded Sand 1 5 6 5 9 9 1 1 1
2 2 Poorly Graded Sand 1 5 6 5 9 10 1 1 1
2 3 Poorly Graded Sand 1 5 6 5 9 9 1 1 1
Poorly Graded Sand-
3 1 Silty Sand 1 5 5 5 9 9 1 1 1
Poorly Graded Sand-
3 2 Silty Sand 1 5 6 5 9 9 1 1 1
Poorly Graded Sand-
3 3 Silty Sand 1 5 5 5 9 5 1 1 1
Totals: 9 45 51 45 81 75 9 9 9

99
 Table A.5: Test and sample quantities at Ft. Hood, Texas

Location: Ft. Hood, Texas

Above Ground
Below Ground
Penetrometer
Classification

Clegg Impact

Bulk Density
Drop Cone
Hammer

Biomass

Biomass
VASST
TDR
Soil
Site Rep Soil Type
Inorganic Clays,
2 2 Medium Plasticity 1 5 7 5 9 7 1 1 1
Inorganic Clays,
2 3 Medium Plasticity 1 5 8 5 9 7 1 1 1
Inorganic Clays, High
3 1 Plasticity 1 5 6 5 9 10 1 1 1
Inorganic Clays, High
3 2 Plasticity 1 5 8 5 9 8 1 1 1
Inorganic Clays, High
3 3 Plasticity 1 5 6 5 9 8 1 1 1
Totals: 5 25 35 25 45 40 5 5 5

100
APPENDIX B: PYTHON CODE
1. # code_complier.py
2. # Developed for: United States Army Corp of Engineers
3. # Created on: 30-May-2013
4. # Created by: Casey Campbell
5. # Last Modified: 19-Feb-2013
6. # Usage: Import, Analyze, and Export VASST Data
7. # Description: This module is used to identify file
8. # directories, import raw data, excute logic
9. # based modules, and export relevant data
10. # from VASST CSV files
11. # ________________________________________________________________
12.
13. import numpy
14. import os
15. import CSV
16. import Tkinter, tkFileDialog
17.
18.
19. #get program dir
20. root=Tkinter.Tk()
21. root.withdraw()
22. prodirname=tkFileDialog.askdirectory(parent=root,initialdir="I:\School\Research\Modified
Mohr-Coulomb\VASST\VASST Code and Data",title='Select Program Directory')
23.
24. #get data dir
25. root=Tkinter.Tk()
26. root.withdraw()
27. datdirname=tkFileDialog.askdirectory(parent=root,initialdir="I:\School\Research\Modified
Mohr-Coulomb\VASST\VASST Code and Data\Raw VASST Data",title='Select Data Directory')
28.
29. #get output dir
30. root=Tkinter.Tk()
31. root.withdraw()
32. outdirname=tkFileDialog.askdirectory(parent=root,initialdir="I:\School\Research\Modified
Mohr-Coulomb\VASST\\VASST Code and Data\Outputs",title='Select Output Directory')
33.
34. #define check file size for compatibility
35. def getSize(fileobject):
36. fileobject.seek(0,2)
37. size = fileobject.tell()
38. return size
39.
40. #get names of all CSV files in directory
41. fileslist=numpy.array([],str)
42. os.chdir(datdirname)
43. for files in os.listdir('.'):
44. if files.endswith('.CSV'):
45. fileslist=numpy.concatenate((fileslist,numpy.array([files],str)))
46. print fileslist
47. numfiles=numpy.size(fileslist)

101
48.
49. #execute analysis
50. folder=numpy.array([])
51. filenam=numpy.array([])
52. loadm=numpy.array([])
53. posm=numpy.array([])
54. slopem=numpy.array([])
55.
56. filenumb=0
57. while filenumb<numfiles:
58. print 'Current Analyzing: ',fileslist.item(filenumb)
59. file = open(fileslist.item(filenumb),'rb')
60. if getSize(file)<1048576:
61. try:
62. execfile(prodirname+'/data_import.py')
63. execfile(prodirname+'/maxs.py')
64. execfile(prodirname+'/first_min.py')
65. execfile(prodirname+'/root_detection.py')
66. execfile(prodirname+'/inflection_pts.py')
67. execfile(prodirname+'/slope_change.py')
68. execfile(prodirname+'/decision_plotter.py')
69. print 'Done'
70.
71. #record current output
72. folder=numpy.concatenate((folder,numpy.array([datdirname],dtype=str)))
73. filenam=numpy.concatenate((filenam,numpy.array([fileslist.item(filenumb)],dtype=str)))
74. loadm=numpy.concatenate((loadm,numpy.array([firstloadpeak],dtype=float)))
75. posm=numpy.concatenate((posm,numpy.array([run],dtype=float)))
76. slopem=numpy.concatenate((slopem,numpy.array([slopepeak],dtype=float)))
77. filenumb=filenumb+1
78. except SystemExit:
79. filenumb=filenumb+1
80. print 'Done'
81. else:
82. print 'Error: File Size to Large'
83. filenumb=filenumb+1
84. print 'Done'
85. print''
86.
87. #create output arrays
88. outdat=numpy.vstack((folder,filenam,loadm,posm,slopem))
89. outdat=numpy.transpose(outdat)
90.
91. #write to file
92. f=open('%s\Output.CSV'%outdirname,'wb')
93. try:
94. writer=CSV.writer(f)
95. writer.writerow(('Data Location','File Name','Load at First Peak','Position of First
Peak','Slope to First Peak'))
96. for i in range(len(loadm)):
97. writer.writerow((outdat[i,0],outdat[i,1],outdat[i,2],outdat[i,3],outdat[i,4]))
98. finally:
99. f.close()
100. # data_import.py
101. # Developed for: United States Army Corp of Engineers

102
102. # Created on: 30-May-2013
103. # Created by: Casey Campbell
104. # Last Modified: 19-Feb-2013
105. # Usage: Import and Condition VASST Data
106. # Description: This module is used as part of the code_complier
107. # module to read CSV files to an array, check
108. # for errors in the raw data, and condition the
109. # raw data from VASST CSV files
110. # ________________________________________________________________
111.
112. from numpy import*
113. from pylab import*
114. from CSV import*
115. import sys
116.
117. #Load Data to ndarray
118. with open(fileslist.item(filenumb), 'rb') as f:
119. reader = reader(f)
120. count=0
121. for row in reader:
122. if len(row)>1:
123. if count==24:
124. DataA=row
125. if count>25:
126. Data=row
127. DataA=vstack((DataA,Data))
128. count=count+1
129. #Detect Multiple Segments
130. if count>25 and size(row)==1:
131. is_valid=0
132. print 'Warning: Multiple Segments Detected'
133. print '1. Use First Segment'
134. print '2. Skip to Next File'
135. print ''
136. while not is_valid:
137. try:
138. choice=int(raw_input('Enter choice [1,2]:'))
139. if choice==1 or choice==2:
140. is_valid = 1
141. except ValueError, e:
142. print 'Invalid Entry, Please Try Again'
143. if choice==1:
144. print 'Processing Continued'
145. if choice==2:
146. print 'Skipping to Next File'
147. sys.exit()
148. break
149. else:
150. count=count+1
151. DataA=DataA.astype(float)
152.
153. #Detect Loadcell
154. if max(DataA[:,4])--min(DataA[:,4])<=5:
155. print 'Error: No Loadcell Detected'
156. sys.exit()

103
157.
158. #Detect Stringpot
159. if max(DataA[:,5])-min(DataA[:,5])<=0.25:
160. print 'Error: No Stringpot Detected'
161. sys.exit()
162.
163. #Detect Markers
164. amarpos=array([])
165. amark=array([])
166. count=0
167. for row in DataA:
168. if abs(row.item(3))>=4:
169. mark=DataA[count,4]
170. marpos=DataA[count,5]
171. amarpos=concatenate((amarpos,array([marpos])))
172. amark=concatenate((amark,array([mark])))
173. count=count+1
174.
175. #Detect Marker Failure
176. if len(amark)==len(DataA):
177. is_valid=0
178. print 'Warning: Marker Failure'
179. print '1. Override Warning'
180. print '2. Accept Warning'
181. print '3. Skip to Next File'
182. print ''
183. while not is_valid:
184. try:
185. choice=int(raw_input('Enter choice [1,2]:'))
186. if choice==1 or choice==2:
187. is_valid = 1
188. except ValueError, e:
189. print 'Invalid Entry, Please Try Again'
190. if choice==1:
191. amarpos=array([])
192. amark=array([])
193. print 'Warning Overridden, Processing Continued'
194. if choice==2:
195. print 'Warning Accepted, Processing Continued'
196. if choice==3:
197. print 'Skipping to Next File'
198. sys.exit()
199. print ''
200.
201. #Removing Backward Motion
202. k0position=0
203. trDataA=ones(6)
204. for row in DataA:
205. k1position=row.item(5)
206. if k0position<k1position:
207. k0position=k1position
208. trDataA=vstack((trDataA,row))
209. trDataA=delete(trDataA,0,0)
210. DataA=trDataA
211.

104
212. #Remove Negative Loads
213. count=0
214. sizeDataA=size(DataA[:,4])
215. while sizeDataA>count:
216. if DataA[count,4]<=0:
217. DataA=delete(DataA,count,0)
218. sizeDataA=size(DataA[:,4])
219. else:
220. count=1+count
221. #Remove Loads at VASST stop
222. count=0
223. sizeDataA=size(DataA[:,4])
224. while sizeDataA>count:
225. if DataA[count,5]>=37.5:
226. DataA=delete(DataA,count,0)
227. sizeDataA=size(DataA[:,4])
228. else:
229. count=1+count
230.
231. #Detect No Positive Load+Motion
232. if len(DataA)<=2:
233. print 'Error: No Data Found'
234. sys.exit()

105
1. # maxs.py
2. # Developed for: United States Army Corp of Engineers
3. # Created on: 30-May-2013
4. # Created by: Casey Campbell
5. # Last Modified: 19-Feb-2013
6. # Usage: Determine Maximums from VASST Data
7. # Description: This module is used as part of the code_complier
8. # module to determine the force and position of
9. # increasing local maximums with respect to the
10. # start of a VASST test. Points of interest are
11. # stored in an array outside of the data array.
12. # ________________________________________________________________
13.
14. from numpy import*
15. from pylab import*
16.
17. count=0
18. countmax=0
19. loadmax=-10.0
20. aloadmax=array([])
21. apositionmax=array([])
22. loadgobmax=max(DataA[:,4])
23. loadgobmin=min(DataA[:,4])
24.
25. #find maximums/create array of them
26. for row in DataA:
27. loadk0=DataA[:,4].item(count)
28. positionk0=DataA[:,5].item(count)
29. if positionk0==max(DataA[:,5]):
30. break
31. try:
32. if loadmax<loadk0 and loadk0>DataA[:,4].item(count+1) or loadk0==loadgobmax:
33. aloadmax=concatenate((aloadmax,array([loadk0])))
34. apositionmax=concatenate((apositionmax,array([DataA[count,5]])))
35. loadmax=loadk0
36. count=count+1
37. except (count+1)==len(DataA):
38. break
39.
40. #remove false maximums at start
41. if max(aloadmax)>30.0:
42. sizea=size(aloadmax)
43. count=1
44. while count<sizea:
45. if 0.95<abs(loadgobmin/aloadmax[0]):
46. aloadmax=delete(aloadmax,0,0)
47. apositionmax=delete(apositionmax,0,0)
48. if abs(aloadmax.item(0)-loadgobmin)<=10:
49. aloadmax=delete(aloadmax,0,0)
50. apositionmax=delete(apositionmax,0,0)
51. if aloadmax.item(0)<=20:
52. aloadmax=delete(aloadmax,0,0)
53. apositionmax=delete(apositionmax,0,0)
54. count=count+1

106
1. # first_min.py
2. # Developed for: United States Army Corp of Engineers
3. # Created on: 31-May-2013
4. # Created by: Casey Campbell
5. # Last Modified: 19-Feb-2013
6. # Usage: Define the Start of a VASST Test
7. # Description: This module is used as part of the code_complier
8. # module to determine the start point of VASST
9. # tests. The start point is then added to an
10. # array with other points of interest.
11. # ________________________________________________________________
12.
13. from numpy import*
14. from pylab import*
15.
16.
17. DataB=DataA
18. #Remove Data After First Peak
19. DataB=flipud(DataB)
20.
21. for row in DataB:
22. if row.item(5)>=apositionmax.item(0):
23. DataB=delete(DataB,0,0)
24.
25. if size(DataB[:,4])<=1:
26. print 'Error: No Minimum Detected'
27. sys.exit()
28.
29. #Find Start Point
30. loadk0=DataB[0,4]
31. loadk1=DataB[1,4]
32. count=0
33. while loadk1<=loadk0:
34. if count+2==size(DataB)/6:
35. break
36. loadmin=DataB[count+1,4]
37. positionmin=DataB[count+1,5]
38. count=count+1
39. if loadk1==loadgobmin:
40. break
41. loadk0=DataB[count,4]
42. loadk1=DataB[count+1,4]
43.
44. #Add Start Point Min to Critical Point Data
45. aloadmax=insert(aloadmax,0,loadmin)
46. apositionmax=insert(apositionmax,0,positionmin)
47.
48. #Remove Data Previous of Start Point
49. DataB=flipud(DataB)
50. for row in DataB:
51. if row.item(5)<apositionmax.item(0):
52. DataA=delete(DataA,0,0)
53. DataB=delete(DataB,0,0)

107
1. # root_detection.py
2. # Developed for: United States Army Corp of Engineers
3. # Created on: 9-June-2013
4. # Created by: Casey Campbell
5. # Last Modified: 19-Feb-2013
6. # Usage: Check for Root Breaks in the VASST Data
7. # Description: This module is used as part of the code_complier
8. # module to identify if root breaking is present
9. # before soil failure that may lead to false
10. # points of interest being recorded.
11. # ________________________________________________________________
12.
13. loada=DataA[:,4]
14. posa=DataA[:,5]
15.
16. count=0
17. rbc=0
18. rbla=array([])
19. rbpa=array([])
20. slopea=array([])
21. maxcount=size(loada)/3
22.
23. while count<maxcount:
24. rise1=loada.item(count+1)-loada.item(count)
25. run1=posa.item(count+1)-posa.item(count)
26. slope=rise1/run1
27. slopea=concatenate((slopea,array([slope])))
28. slopeavg=average(slopea)
29. if slope<0 and slopeavg<=abs(slope*0.75):
30. rbc=1+rbc
31. rbla=concatenate((rbla,array([loada.item(count)])))
32. rbpa=concatenate((rbpa,array([posa.item(count)])))
33. count=count+1

108
1. # inflection_pts.py
2. # Developed for: United States Army Corp of Engineers
3. # Created on: 15-June-2013
4. # Created by: Casey Campbell
5. # Last Modified: 19-Feb-2013
6. # Usage: Identify Inflection Point in the VASST Data
7. # Description: This module is used as part of the code_complier
8. # module to determine major inflection point.
9. # Depending on the location of the inflection
10. # point, they may be points of interest
11. # ________________________________________________________________
12.
13. from numpy import*
14.
15. loada=DataA[:,4]
16. positiona=DataA[:,5]
17. timea=DataA[:,0]
18.
19. #find the slope a each pt
20. slopea=array([])
21. count=0
22. while count<size(loada)-1:
23. sfor= (loada.item(count+1)-loada.item(count))/(positiona.item(count+1)-
positiona.item(count))
24. slopea=concatenate((slopea,array([sfor])))
25. count=count+1
26. slopea=concatenate((slopea,array([0])))
27.
28. #find inflection points in positive slopes
29. loadb=array([])
30. positionb=array([])
31. count=0
32. while count<size(loada)-6:
33. if loada.item(count)<loada.item(count+3) and
loada.item(count+3)<loada.item(count+5):
34. if slopea.item(count+1)>slopea.item(count+2) and
slopea.item(count+3)<slopea.item(count+4):
35. if slopea.item(count+1)>1.25*slopea.item(count+2)and
slopea.item(count+4)>1.25*slopea.item(count+3):
36. inflecload=array([loada.item(count+3)])
37. inflecpos=array([positiona.item(count+3)])
38. loadb=concatenate((loadb,inflecload))
39. positionb=concatenate((positionb,inflecpos))
40. count=count+1
41.
42. #delete points within 0.625 inches of max points
43. count=0
44. while count<size(loadb):
45. if len(positionb)==0:
46. break
47. fastcount=0
48. while fastcount<len(apositionmax):
49. if len(positionb)<=count:
50. break
51. if abs(positionb[count]-apositionmax[fastcount])<=.125:

109
52. loadb=delete(loadb,count,0)
53. positionb=delete(positionb,count,0)
54. fastcount=fastcount+1
55. count=count+1
56.
57. count=0
58. while count<size(positionb)-1:
59. if len(positionb)==0:
60. break
61. if positionb.item(count+1)-positionb.item(count)<=.125:
62. loadb=delete(loadb,count+1,0)
63. positionb=delete(positionb,count+1,0)
64. count=count+1

110
1. #Developed for United States Army Corp of Engineers
2. # slope_change.py
3. # Developed for: United States Army Corp of Engineers
4. # Created on: 22-June-2013
5. # Created by: Casey Campbell
6. # Last Modified: 19-Feb-2013
7. # Usage: Critical Slope Changes in the VASST Data
8. # Description: This module is used as part of the code_complier
9. # module to identify a point of interest being a
10. # critical slope change before a maximum is
11. # reached.
12. # ________________________________________________________________
13.
14. from numpy import*
15.
16. loada=DataA[:,4]
17. positiona=DataA[:,5]
18. timea=DataA[:,0]
19.
20. #find the 2nd slope at each pt
21.
22. slopea2=array([])
23. countsc=0
24.
25. while countsc<size(loada)-2:
26. sfor= (loada.item(countsc+2)-loada.item(countsc))/(positiona.item(countsc+2)-
positiona.item(countsc))
27. slopea2=concatenate((slopea2,array([sfor])))
28. countsc=countsc+1
29. slopea2=concatenate((slopea2,array([0])))
30.
31. countsc=1
32. while countsc<size(loada)-2:
33. if DataA[countsc-1,4]>=2+DataA[0,4] and DataA[countsc-1,5]>=0.25+DataA[0,5]:
34. break
35. countsc=countsc+1
36.
37. avgslope=0
38. relerror=0
39. saloadmax=size(aloadmax)
40.
41. while DataA[countsc-1,4]<=aloadmax.item(saloadmax-1):
42. loadc=DataA[countsc-1,4]
43. positionc=DataA[countsc-1,4]
44. avgslope=(slopea2.item(countsc)+(countsc-1)*avgslope)/countsc
45. if avgslope>0 or avgslope<0:
46. relerror=(avgslope-slopea2.item(countsc))/avgslope
47. countsc=countsc+1
48. if relerror>=0.8:
49. break
50.
51. if DataA[countsc,4]<aloadmax.item(1)and DataA[countsc,5]<apositionmax.item(1):
52. if DataA[countsc,5]-DataA[0,5]>=0.125:
53. aloadmax[1]=DataA[countsc,4]
54. apositionmax[1]=DataA[countsc,5]

111
1. # decision_plotter.py
2. # Developed for: United States Army Corp of Engineers
3. # Created on: 30-May-2013
4. # Created by: Casey Campbell
5. # Last Modified: 19-Feb-2013
6. # Usage: Determine if a Human Decision is Required and
7. # Create and Save Graphical Data from the VASST
8. # Description: This module is used as part of the code_complier
9. # module to determine the point of shear failure.
10. # Human decisions may be required to identify the
11. # point of shear failure. This module also creates
12. # and saves all graphical VASST data.
13. # ________________________________________________________________
14.
15. from numpy import*
16. import Tkinter, tkSimpleDialog
17. from multiprocessing import Process
18. import sys
19.
20. #Determine if conditioned data set is complete
21. if size(aloadmax)<=1:
22. print 'Error: Conditioned Data Set too Small'
23. sys.exit()
24.
25. #Determine if marker is of interest
26. if size(amarpos)>0:
27. if (max(amarpos)-.125)<=min(apositionmax)or (min(amarpos)-
.125)>=max(apositionmax):
28. amarpos=array([])
29. amark=array([])
30. else:
31. print amark
32. print 'Notice: Marker(s) at', amarpos.item(0)
33.
34. #Determine number of points and how to handle them
35. numbcritpts=size(aloadmax)
36. if numbcritpts==2:
37. rise=aloadmax.item(1)-aloadmax.item(0)
38. run=apositionmax.item(1)-apositionmax.item(0)
39. slope=rise/run
40. firstloadpeak=aloadmax.item(1)
41. firstloadposition=apositionmax.item(1)
42. slopetopeak=slope
43.
44. if DataA[countsc,5]<apositionmax.item(1) or len(positionb)>0 and
positionb.item(0)<apositionmax.item(1) or size(amark)>0 or rbc>=1:
45. #Display graph in question
46. ion()
47. title('Force-Displacement: %s' %(fileslist.item(filenumb)),fontsize=16)
48. scatter(apositionmax,aloadmax,color='b',linewidths='2',label='maxs')
49. scatter(DataA[countsc,5],DataA[countsc,4],color='r',label='slope')
50. plot(DataA[:,5],DataA[:,4],color='b',label='data')
51. scatter(positionb,loadb,color='m',label='inflection pts')
52. plot(rbpa,rbla,color='g',linewidth=3,label='root breaking')
53. plot(amarpos,amark,color='r',linewidth=2,label='field marker')

112
54. #Stack/sort pts of user input
55. allptsload=hstack((aloadmax,loadb))
56. allptsposition=hstack((apositionmax,positionb))
57. allpts=transpose(vstack((allptsload,allptsposition)))
58. order=allpts[:,1].argsort()
59. allpts=transpose(take(allpts,order,0))
60. allptsload,allptsposition=vsplit(allpts,2)
61. #Prompt for user input
62. root=Tkinter.Tk()
63. root.withdraw()
64. ptselect=tkSimpleDialog.askinteger('Point Selection', 'Select First Peak')
65. scatter(allptsposition.item(ptselect),allptsload.item(ptselect),color='k',
marker='x',linewidths=10, label='max: user input')
66. ax=subplot(111)
67. box=ax.get_position()
68. ax.set_position([box.x0,box.y0,box.width*0.8, box.height])
69. legend(loc='center left',bbox_to_anchor=(1,0.5), prop={'size':10})
70. xlabel('Displacement in Inches',fontsize=12)
71. ylabel('Applied Force in Pounds',fontsize=12)
72. savefig('%s\%s.png'%(outdirname,fileslist.item(filenumb)),dpi=400)
73. close()
74. firstloadpeak=allptsload.item(ptselect)
75. firstloadposition=allptsposition.item(ptselect)
76. rise=(firstloadpeak-aloadmax.item(0))
77. run=(firstloadposition-apositionmax.item(0))
78. slopepeak=arctan(run/rise)
79. else:
80. firstloadpeak=aloadmax.item(1)
81. firstloadposition=apositionmax.item(1)
82. rise=aloadmax.item(1)-aloadmax.item(0)
83. run=apositionmax.item(1)-apositionmax.item(0)
84. slopepeak=arctan(run/rise)
85. title('Force-Displacement: %s' %(fileslist.item(filenumb)), fontsize=16)
86. scatter(firstloadposition,firstloadpeak,color='k',marker='x',linewidths=10,
label='max used')
87. scatter(apositionmax,aloadmax,color='b',linewidths='2',label='maxs')
88. plot(DataA[:,5],DataA[:,4],color='b',label='data')
89. ax=subplot(111)
90. box=ax.get_position()
91. ax.set_position([box.x0,box.y0,box.width*0.8, box.height])
92. legend(loc='center left',bbox_to_anchor=(1,0.5), prop={'size':10})
93. xlabel('Displacement in Inches',fontsize=12)
94. ylabel('Applied Force in Pounds',fontsize=12)
95. savefig('%s\%s.png'%(outdirname,fileslist.item(filenumb)),dpi=150)
96. close()

113
APPENDIX C: PROC SGSCATTER RESULTS FOR CLASS VARIABLES
Figure C.1: Scatter plot matrix of variables by plasticity

114
Figure C.2: Scatter plot matrix of variables by average grain size

115
Figure C.3: Scatter plot matrix of variables by USCS

116
Figure C.4: Scatter plot matrix of variables by USDA texture

117
APPENDIX D: PROC SGPANEL RESULTS FOR CLASS VARIABLES
Figure D.1: Box plots of VASST force measurements by plasticity

Figure D.2: Box plots of VASST force measurements by USCS

118
Figure D.3: Box plots of VASST force measurements by average grain size

Figure D.4: Box plots of VASST force measurements by USDA texture

119
APPENDIX E: PROC CORR RESULTS
Figure E.1: Proc corr results for high plasticity soils

120
Figure E.2: Proc corr results for medium plasticity soils

121
Figure E.2: Proc corr results for non-plastic soils

122
APPENDIX F: PROC LIFETEST RESULTS WITH BULK DENSITY
Figure F.1: Proc lifetest results for high plasticity soils
Parameters Affecting VASST Force Measurements with Bulk Density

The LIFETEST Procedure

Plasticity=High

Quartile Estimates
Percent Point 95% Confidence Interval
Estimate Transform [Lower Upper)
75 465.132 LOGLOG 443.079 494.536
50 362.699 LOGLOG 333.501 400.965
25 270.364 LOGLOG 237.250 292.142

Mean Standard Error

364.585 13.121

123
 Figure F.1 (cont.)
Summary of the Number of Censored and
Uncensored Values
Total Failed Censored Percent
Censored
137 137 0 0.00

Rank Tests for the Association of VM with Covariates

Univariate Chi-Squares for the Wilcoxon Test

Variable Test Standard Chi-Square Pr > Label
Statistic Error Chi-Square
Root_Wt -53.7449 54.6642 0.9666 0.3255 Root Wt
BGBio -51.9384 51.5964 1.0133 0.3141 Below Ground Biomass
AGBio 222.8 375.6 0.3519 0.5530 Above Ground Biomass
tdra 8.1571 28.9164 0.0796 0.7779 Soil Moisture
D30 0.00308 0.00337 0.8373 0.3602 30% Finer Diameter
D60 0.00668 0.0248 0.0725 0.7878 60% Finer Diameter
Bulk_Den 0.2237 0.3722 0.3611 0.5479 Bulk Density

Covariance Matrix for the Wilcoxon Statistics

Variable Root_Wt BGBio AGBio tdra D30 D60 Bulk_Den
Root_Wt 2988 2817 12491 -720 0 -1 -12
BGBio 2817 2662 11368 -637 0 -1 -11
AGBio 12491 11368 141070 -4706 0 -8 -89
tdra -720 -637 -4706 836 -0 0 -0
D30 0 0 0 -0 0 -0 -0
D60 -1 -1 -8 0 -0 0 0
Bulk_Den -12 -11 -89 -0 -0 0 0

Forward Stepwise Sequence of Chi-Squares for the Wilcoxon Test

Variable DF Chi-Square Pr > Chi-Square Pr > Label
Chi-Square Increment Increment
BGBio 1 1.0133 0.3141 1.0133 0.3141 Below Ground Biomass
D30 2 4.6885 0.0959 3.6752 0.0552 30% Finer Diameter
AGBio 3 6.8543 0.0767 2.1658 0.1411 Above Ground Biomass
Bulk_Den 4 8.7306 0.0682 1.8763 0.1708 Bulk Density
tdra 5 10.4986 0.0623 1.7679 0.1836 Soil Moisture
D60 6 12.9960 0.0431 2.4975 0.1140 60% Finer Diameter
Root_Wt 7 15.3887 0.0313 2.3926 0.1219 Root Wt

124
 Figure F.1 (cont.)
Univariate Chi-Squares for the Log-Rank Test
Variable Test Standard Chi-Square Pr > Label
Statistic Error Chi-Square
Root_Wt 7.1488 86.4994 0.00683 0.9341 Root Wt
BGBio 6.3140 81.0656 0.00607 0.9379 Below Ground Biomass
AGBio 708.8 639.2 1.2294 0.2675 Above Ground Biomass
tdra -14.6230 49.0493 0.0889 0.7656 Soil Moisture
D30 0.00562 0.00533 1.1112 0.2918 30% Finer Diameter
D60 -0.0188 0.0407 0.2130 0.6445 60% Finer Diameter
Bulk_Den 0.00897 0.5508 0.000265 0.9870 Bulk Density

Covariance Matrix for the Log-Rank Statistics

Variable Root_Wt BGBio AGBio tdra D30 D60 Bulk_Den
Root_Wt 7482 7003 29845 -2335 0 -3 -24
BGBio 7003 6572 26674 -2060 0 -3 -22
AGBio 29845 26674 408601 -15225 1 -21 -194
tdra -2335 -2060 -15225 2406 -0 1 1
D30 0 0 1 -0 0 -0 -0
D60 -3 -3 -21 1 -0 0 0
Bulk_Den -24 -22 -194 1 -0 0 0

Forward Stepwise Sequence of Chi-Squares for the Log-Rank Test

Variable DF Chi-Square Pr > Chi-Square Pr > Label
Chi-Square Increment Increment
AGBio 1 1.2294 0.2675 1.2294 0.2675 Above Ground Biomass
Bulk_Den 2 1.7946 0.4077 0.5651 0.4522 Bulk Density
D30 3 2.7940 0.4245 0.9994 0.3174 30% Finer Diameter
tdra 4 3.7499 0.4409 0.9559 0.3282 Soil Moisture
Root_Wt 5 4.0505 0.5422 0.3006 0.5835 Root Wt
BGBio 6 4.6895 0.5842 0.6389 0.4241 Below Ground Biomass
D60 7 7.5412 0.3748 2.8518 0.0913 60% Finer Diameter

125
 Figure F.2: Proc lifetest results for non-plastic soils
Parameters Affecting VASST Force Measurements with Bulk Density

The LIFETEST Procedure

Plasticity=None

Quartile Estimates
Percent Point 95% Confidence Interval
Estimate Transform [Lower Upper)
75 119.771 LOGLOG 112.489 131.656
50 96.688 LOGLOG 89.337 102.871
25 77.863 LOGLOG 72.505 82.466

Mean Standard Error

101.519 3.050

126
 Figure F.2 (cont.)
Summary of the Number of Censored and
Uncensored Values
Total Failed Censored Percent
Censored
141 141 0 0.00

Rank Tests for the Association of VM with Covariates

Univariate Chi-Squares for the Wilcoxon Test

Variable Test Standard Chi-Square Pr > Label
Statistic Error Chi-Square
Root_Wt -6.1632 5.6791 1.1778 0.2778 Root Wt
BGBio 53.5098 23.5356 5.1691 0.0230 Below Ground Biomass
AGBio 803.5 303.5 7.0119 0.0081 Above Ground Biomass
tdra 16.9234 19.1138 0.7839 0.3759 Soil Moisture
D30 -0.7363 0.2442 9.0908 0.0026 30% Finer Diameter
D60 -2.2581 0.5646 15.9968 <.0001 60% Finer Diameter
Bulk_Den 1.3187 0.6430 4.2058 0.0403 Bulk Density

Covariance Matrix for the Wilcoxon Statistics

Variable Root_Wt BGBio AGBio tdra D30 D60 Bulk_Den
Root_Wt 32.3 7.5 95.2 1.3 -0.3 -1.6 -1.1
BGBio 7.5 553.9 390.1 189.9 -0.2 -1.2 -0.8
AGBio 95.2 390.1 92084.7 55.1 1.9 10.0 -101.0
tdra 1.3 189.9 55.1 365.3 -2.6 -2.3 0.8
D30 -0.3 -0.2 1.9 -2.6 0.1 0.1 -0.0
D60 -1.6 -1.2 10.0 -2.3 0.1 0.3 0.0
Bulk_Den -1.1 -0.8 -101.0 0.8 -0.0 0.0 0.4

Forward Stepwise Sequence of Chi-Squares for the Wilcoxon Test

Variable DF Chi-Square Pr > Chi-Square Pr > Label
Chi-Square Increment Increment
D60 1 15.9968 <.0001 15.9968 <.0001 60% Finer Diameter
Root_Wt 2 28.5321 <.0001 12.5353 0.0004 Root Wt
AGBio 3 39.0752 <.0001 10.5431 0.0012 Above Ground Biomass
Bulk_Den 4 50.2210 <.0001 11.1458 0.0008 Bulk Density
BGBio 5 53.5116 <.0001 3.2906 0.0697 Below Ground Biomass
tdra 6 56.0864 <.0001 2.5748 0.1086 Soil Moisture
D30 7 57.0377 <.0001 0.9513 0.3294 30% Finer Diameter

127
 Figure F.2 (cont.)
Univariate Chi-Squares for the Log-Rank Test
Variable Test Standard Chi-Square Pr > Label
Statistic Error Chi-Square
Root_Wt -11.2311 8.4415 1.7701 0.1834 Root Wt
BGBio 100.2 50.1127 3.9960 0.0456 Below Ground Biomass
AGBio 1660.0 640.9 6.7094 0.0096 Above Ground Biomass
tdra 32.3441 33.1856 0.9499 0.3297 Soil Moisture
D30 -0.9282 0.4109 5.1039 0.0239 30% Finer Diameter
D60 -3.1869 1.1036 8.3392 0.0039 60% Finer Diameter
Bulk_Den 2.2561 1.0548 4.5753 0.0324 Bulk Density

Covariance Matrix for the Log-Rank Statistics

Variable Root_Wt BGBio AGBio tdra D30 D60 Bulk_Den
Root_Wt 71 7 186 -41 -0 -4 -3
BGBio 7 2511 10403 973 1 5 -14
AGBio 186 10403 410732 4470 47 158 -331
tdra -41 973 4470 1101 -7 -3 3
D30 -0 1 47 -7 0 0 -0
D60 -4 5 158 -3 0 1 -0
Bulk_Den -3 -14 -331 3 -0 -0 1

Forward Stepwise Sequence of Chi-Squares for the Log-Rank Test

Variable DF Chi-Square Pr > Chi-Square Pr > Label
Chi-Square Increment Increment
D60 1 8.3392 0.0039 8.3392 0.0039 60% Finer Diameter
AGBio 2 19.3436 <.0001 11.0045 0.0009 Above Ground Biomass
Bulk_Den 3 34.6927 <.0001 15.3491 <.0001 Bulk Density
Root_Wt 4 38.8941 <.0001 4.2014 0.0404 Root Wt
D30 5 42.0333 <.0001 3.1392 0.0764 30% Finer Diameter
BGBio 6 45.7060 <.0001 3.6727 0.0553 Below Ground Biomass
tdra 7 51.1677 <.0001 5.4617 0.0194 Soil Moisture

128
APPENDIX G: PROC LIFETEST RESULTS WITHOUT BULK DENSITY
Figure G.1: Proc lifetest results for high plasticity soils
Parameters Affecting VASST Force Measurements without Bulk Density

The LIFETEST Procedure

Plasticity=High

Quartile Estimates
Percent Point 95% Confidence Interval
Estimate Transform [Lower Upper)
75 465.132 LOGLOG 443.079 494.536
50 362.699 LOGLOG 333.501 400.965
25 270.364 LOGLOG 237.250 292.142

Mean Standard Error

364.585 13.121

129
 Figure G.1 (cont.)
Summary of the Number of Censored and
Uncensored Values
Total Failed Censored Percent
Censored
137 137 0 0.00

Rank Tests for the Association of VM with Covariates

Univariate Chi-Squares for the Wilcoxon Test

Variable Test Standard Chi-Square Pr > Label
Statistic Error Chi-Square
Root_Wt 110.2 59.3923 3.4434 0.0635 Root Wt
BGBio 114.2 56.4720 4.0865 0.0432 Below Ground Biomass
AGBio 1805.6 417.6 18.6965 <.0001 Above Ground Biomass
tdra -21.2920 29.3911 0.5248 0.4688 Soil Moisture
D30 -0.00119 0.00351 0.1144 0.7351 30% Finer Diameter
D60 0.0266 0.0255 1.0958 0.2952 60% Finer Diameter

Covariance Matrix for the Wilcoxon Statistics

Variable Root_Wt BGBio AGBio tdra D30 D60
Root_Wt 3527 3350 15745 -770 0 -1
BGBio 3350 3189 14600 -687 0 -1
AGBio 15745 14600 174377 -5214 0 -8
tdra -770 -687 -5214 864 -0 0
D30 0 0 0 -0 0 -0
D60 -1 -1 -8 0 -0 0

Forward Stepwise Sequence of Chi-Squares for the Wilcoxon Test

Variable DF Chi-Square Pr > Chi-Square Pr > Label
Chi-Square Increment Increment
AGBio 1 18.6965 <.0001 18.6965 <.0001 Above Ground Biomass
D60 2 59.1087 <.0001 40.4121 <.0001 60% Finer Diameter
Root_Wt 3 65.8502 <.0001 6.7415 0.0094 Root Wt
tdra 4 72.7190 <.0001 6.8688 0.0088 Soil Moisture
BGBio 5 72.7771 <.0001 0.0580 0.8096 Below Ground Biomass
D30 6 72.7862 <.0001 0.00916 0.9237 30% Finer Diameter

130
 Figure G.1 (cont.)
Univariate Chi-Squares for the Log-Rank Test
Variable Test Standard Chi-Square Pr > Label
Statistic Error Chi-Square
Root_Wt 174.3 91.6831 3.6160 0.0572 Root Wt
BGBio 175.9 86.1244 4.1715 0.0411 Below Ground Biomass
AGBio 2385.3 677.5 12.3943 0.0004 Above Ground Biomass
tdra -44.7797 51.0505 0.7694 0.3804 Soil Moisture
D30 0.00140 0.00559 0.0628 0.8021 30% Finer Diameter
D60 0.00270 0.0425 0.00402 0.9494 60% Finer Diameter

Covariance Matrix for the Log-Rank Statistics

Variable Root_Wt BGBio AGBio tdra D30 D60
Root_Wt 8406 7885 34318 -2532 0 -3
BGBio 7885 7417 30891 -2238 0 -3
AGBio 34318 30891 459038 -16618 1 -23
tdra -2532 -2238 -16618 2606 -0 2
D30 0 0 1 -0 0 -0
D60 -3 -3 -23 2 -0 0

Forward Stepwise Sequence of Chi-Squares for the Log-Rank Test

VariableDF Chi-Square Pr > Chi-Square Pr > Label
Chi-Square Increment Increment
AGBio 1 12.3943 0.0004 12.3943 0.0004 Above Ground Biomass
D60 2 33.5742 <.0001 21.1799 <.0001 60% Finer Diameter
Root_Wt 3 53.3908 <.0001 19.8166 <.0001 Root Wt
tdra 4 67.4105 <.0001 14.0197 0.0002 Soil Moisture
D30 5 69.4141 <.0001 2.0036 0.1569 30% Finer Diameter
BGBio 6 69.4332 <.0001 0.0191 0.8899 Below Ground Biomass

131
 Figure G.2: Proc lifetest results for non-plastic soils
Parameters Affecting VASST Force Measurements without Bulk Density

The LIFETEST Procedure

Plasticity=None

Quartile Estimates
Percent Point 95% Confidence Interval
Estimate Transform [Lower Upper)
75 119.771 LOGLOG 112.489 131.656
50 96.688 LOGLOG 89.337 102.871
25 77.863 LOGLOG 72.505 82.466

Mean Standard Error

101.519 3.050

132
 Figure G.2 (cont.)
Summary of the Number of Censored and
Uncensored Values
Total Failed Censored Percent
Censored
141 141 0 0.00

Rank Tests for the Association of VM with Covariates

Univariate Chi-Squares for the Wilcoxon Test

Variable Test Standard Chi-Square Pr > Label
Statistic Error Chi-Square
Root_Wt -6.1632 5.6791 1.1778 0.2778 Root Wt
BGBio 53.5098 23.5356 5.1691 0.0230 Below Ground Biomass
AGBio 803.5 303.5 7.0119 0.0081 Above Ground Biomass
tdra 16.9234 19.1138 0.7839 0.3759 Soil Moisture
D30 -0.7363 0.2442 9.0908 0.0026 30% Finer Diameter
D60 -2.2581 0.5646 15.9968 <.0001 60% Finer Diameter

Covariance Matrix for the Wilcoxon Statistics

Variable Root_Wt BGBio AGBio tdra D30 D60
Root_Wt 32.3 7.5 95.2 1.3 -0.3 -1.6
BGBio 7.5 553.9 390.1 189.9 -0.2 -1.2
AGBio 95.2 390.1 92084.7 55.1 1.9 10.0
tdra 1.3 189.9 55.1 365.3 -2.6 -2.3
D30 -0.3 -0.2 1.9 -2.6 0.1 0.1
D60 -1.6 -1.2 10.0 -2.3 0.1 0.3

Forward Stepwise Sequence of Chi-Squares for the Wilcoxon Test

VariableDF Chi-Square Pr > Chi-Square Pr > Label
Chi-Square Increment Increment
D60 1 15.9968 <.0001 15.9968 <.0001 60% Finer Diameter
Root_Wt 2 28.5321 <.0001 12.5353 0.0004 Root Wt
AGBio 3 39.0752 <.0001 10.5431 0.0012 Above Ground Biomass
BGBio 4 42.1703 <.0001 3.0951 0.0785 Below Ground Biomass
tdra 5 43.9207 <.0001 1.7504 0.1858 Soil Moisture
D30 6 45.8424 <.0001 1.9217 0.1657 30% Finer Diameter

133
 Figure G.2 (cont.)
Univariate Chi-Squares for the Log-Rank Test
Variable Test Standard Chi-Square Pr > Label
Statistic Error Chi-Square
Root_Wt -11.2311 8.4415 1.7701 0.1834 Root Wt
BGBio 100.2 50.1127 3.9960 0.0456 Below Ground Biomass
AGBio 1660.0 640.9 6.7094 0.0096 Above Ground Biomass
tdra 32.3441 33.1856 0.9499 0.3297 Soil Moisture
D30 -0.9282 0.4109 5.1039 0.0239 30% Finer Diameter
D60 -3.1869 1.1036 8.3392 0.0039 60% Finer Diameter

Covariance Matrix for the Log-Rank Statistics

Variable Root_Wt BGBio AGBio tdra D30 D60
Root_Wt 71 7 186 -41 -0 -4
BGBio 7 2511 10403 973 1 5
AGBio 186 10403 410732 4470 47 158
tdra -41 973 4470 1101 -7 -3
D30 -0 1 47 -7 0 0
D60 -4 5 158 -3 0 1

Forward Stepwise Sequence of Chi-Squares for the Log-Rank Test

VariableDF Chi-Square Pr > Chi-Square Pr > Label
Chi-Square Increment Increment
D60 1 8.3392 0.0039 8.3392 0.0039 60% Finer Diameter
AGBio 2 19.3436 <.0001 11.0045 0.0009 Above Ground Biomass
Root_Wt 3 32.0531 <.0001 12.7094 0.0004 Root Wt
BGBio 4 33.9652 <.0001 1.9122 0.1667 Below Ground Biomass
tdra 5 38.9261 <.0001 4.9609 0.0259 Soil Moisture
D30 6 39.5281 <.0001 0.6020 0.4378 30% Finer Diameter

134
APPENDIX H: ACRONYMS
Acronym Initial Components
AR Army Regulation
ASABE American Society of Agricultural and Biological Engineers
ASTM American Society for Testing and Materials
CBR California-bearing ratio
CERL Construction Engineering Research Laboratory
CI cone index
CIV Clegg impact value
CRREL Cold Regions Research and Engineering Laboratory
ERDC Engineering Research and Development Center
GSL Geotechnical and Structures Laboratory
GUI Graphical User Interface
HAT Hilltop Access Trail
ITAM Integrated Training Area management
LCTA Land Condition-Trend Analysis
NRCS Natural Resources Conservation Service
NRMM NATO Referenced Mobility Model
OPAL Optimal Allocation of Training Lands
PVC poly-vinyl chloride
RTLP Range and Training Lands Program
SRP Sustainable Range Program
TDR time domain reflectometry
UIUC University of Illinois Urbana-Champaign
USCS United Soil Classification System
USDA United States Department of Agriculture
VASST Vegetation And Soil Shear Tester
WES Waterways Experimental Station

135