SOIL BASED VEGETATION PRODUCTIVITY MODELING FOR A
NORTHERN MICHIGAN SURFACE MINING REGION
By
Dustin L. Corr

A THESIS
Submitted to
Michigan State University
in partial fulfillment of the requirements
for the degree of
MASTER OF ARTS
Environmental Design
2012

ABSTRACT
SOIL BASED VEGETATION PRODUCTIVITY MODELING FOR A NORTHERN
MICHIGAN SURFACE MINING REGION
By
Dustin L. Corr
The proliferation of mined landscapes and concern for the environmental impacts
associated with these lands have led to an increased interest in developing empirical predictive
models to quantitatively assess the vegetative productivity potentials of reconstructed soils (neosols). This research presents equations for a northern Michigan mining region based on data
derived from the National Resources Conservation Service. We employed principal component
analysis to develop models to predict the vegetative productivity of corn, corn silage, oats,
alfalfa/hay, Irish potatoes, red maple (Acer rubrum L.), white spruce (Picea glauca [Moench]
Voss), red pine (Pinus resinosa Aniton), eastern white pine (Pinus strobus L.), jack pine (Pinus
banksiana Lamb.), and lilac (Syringa vulgaris L.). Soil attributes that were examined in this
research include: available water holding capacity, moist bulk density, % clay, % rock
fragments, hydraulic conductivity, % organic matter, soil reactivity, % slope, and topographic
position. Five predictive equations based on land use have been developed and are described as
an all woody and crop equation, a xeric equation, an equation specific to jack pine, and two
semi-wet equations of varying conservativeness. The models were highly significant (p<0.0001)
and explained 87.93%, 74.52%, 65.33%, 91.79% and 87.68% of the variation in site productivity
of the respective land uses. These equations are intended to be used in efforts to assess the
vegetative productivity potentials of reconstructed soils on post-mined landscapes.

DEDICATION

This thesis is dedicated to Steven L. and Donna K. Corr

iii

ACKNOWLEDGMENTS

First and foremost, I would like to thank my advising professor Dr. Jon B. Burley for his
guidance throughout my study. His expertise in mining reclamation practices and analytical
techniques has provided insight that added invaluable depth to this study. I would also like to
take this opportunity to thank my committee members, Dr. Bert Cregg and Dr. Robert Schutzki.
Dr. Cregg taught me about the value of not only the data but the value in how that data is
analyzed in a way that can be used to improve our management of natural resources. I look
forward to working with Dr. Cregg in the future to add further value to this study and its
application to post-mined soils. Dr. Robert Schutzki has been a figure throughout my
undergraduate and graduate studies that has shaped the way I think about natural landscape
restoration more than anyone else. I have had the pleasure of working with Dr. Schutzki on a
variety of research projects in which he taught me about research methods that have been
instrumental in my success as a student and will undoubtedly help to guide me throughout my
career. I would also like to thank the faculty of the Michigan State University Landscape
Architecture Program. These academics have shaped the way I think, design, and live.
I would also like to thank my parents, Steven and Donna Corr, for their constant support
throughout my entire life. There personal sacrifices that made my education possible have not
gone unnoticed. I appreciate everything you both have done for me with all of my heart.

iv

TABLE OF CONTENTS

LIST OF TABLES…………..................................................................................................vi
LIST OF FIGURES…………...............................................................................................vii
1.1 Understanding Reclamation Terminology……………………………………………..….…1
1.2 Conditions of Reconstructed Soils…………………………………………………………..……2
1.3 Mine Reclamation Legislation…………………………………………………………………..….4
1.4 Michigan Reclamation Legislation…………………………………………………………….…8
1.5 Reclamation Evaluation Approaches………………………………………………………….10

INTRODUCTION

2.1 Methodology………………………………………...……………………………………………………..17
2.2 Study Area………………………………………………………………………………………………......26
2.2a Iron County……………………………………………………………………………………...29
2.2b Dickinson County……………………………………………………………………………..29
2.3 Variables Investigated………………………………………………………………………………..30

METHODS

3.1 Principal Component 1……………………………………………………………………………….34
3.2 Principal Component 2……………………………………………………………………………….36
3.3 Principal Component 3……………………………………………………………………………….37
3.4 Principal Component 4……………………………………………………………………………….41

RESULTS

4.1 Principal Component 1 Equation: Mesic Model……………………………...………….43
4.2 Principal Component 2 Equation: Northern Dry Forest Model…………..…..…45
4.3 Principal Component 3 Equation: Wet Mesic Model………………………………….46
4.4 Jack Pine Model………………………………………………………………………………….……….51
4.5 Interpretation of Models…………………………………………………………………………….54
4.6 Limitations of Study………………………………………………………………………….………...57
4.7 Additional Uses of Equations……………………………………………………………………...63

DISCUSSION

APPENDIX………………………………………………………………………………………………………………...65

LITERATURE CITED……………………………………………………………………………….………………72
v

LIST OF TABLES

Table 1.1

Example eigenvalue table. These results are strictly to
illustrate the methods of this study and do not reflect any
actual values derived from this investigation.……………………………………22

Table 1.2

Example eigenvalue chart. These results are strictly to
illustrate the methods of this study and do not reflect any
actual values derived from this investigation.………………………………..22-23

Table 1.3

Example Stepwise Maximum R Improvement acceptable
equation. These results are strictly to illustrate the methods
of this study and do not reflect any actual values derived
from this investigation.…………………………………………………………..25

Table 1.4

Independent variable soil parameter units and abbreviations
as they appear in the regression models and final developed
equations…………………………………………………………………………30

Table 1.5

Dependent variable crop and woody plant units and
abbreviations as they appear in the eigenvector analysis…………………….31-32

Table 2.1

Principal component analysis eigenvalues of the covariance
matrix for Iron County and Dickinson County, Michigan…………………….…33

Table 2.2

Principal component analysis eigenvectors for Iron County
and Dickinson County, Michigan dependent variables.
See Table 1.5 for explanations of variable abbreviations.
The number 1 attached to each variable indicates that that
variable has been standardized to a mean of zero and a
standard deviation of 1…………………………………………………………...34

Table A.1

Iron County Independent Variables……………....……………………..……65-66

Table A.2

Iron County Dependent Variables…………….….…………………..…...….66-67

Table A.3

Dickinson County Independent Variables…….….………………….…….…67-69

Table A.4

Dickinson County Dependent Variables…….………..……………………...69-70

2

vi

LIST OF FIGURES

Figure 1.1

Location of study area in Michigan.…..…………………………………………26

Figure 1.2

Locations of mines and mine exploration sites in the
Michigan’s western Upper Peninsula……………………………………………28

Figure 2.1

Best model found for the Stepwise Maximum
R-squared Improvement for Principal Component 1.
Refer to Table 1.4 for an explanation of independent
variable soil parameters…………………………………………………...……..35

Figure 2.2 Best model found for the Stepwise Maximum

R-squared Improvement for Principal Component 2.
Refer to Table 1.4 for an explanation of independent
variable soil parameters……………………………………………….……36-37

Figure 2.3 First model found for the Stepwise Maximum

R-squared Improvement for Principal Component 3
with all P values below 0.04. The number of terms in
this model (38 terms) was greater than the C(p) value
(33.95) and may be considered over specific. Refer to
Table 1.4 for an explanation of independent variable soil
parameters…………………………………………………………………….38-39

Figure 2.4 Second model found for the Stepwise Maximum

R-squared Improvement for Principal Component 3 with
all P values below 0.04. The number of terms in this
model (32 terms) was greater than the C(p) value (48.00)
and is not considered over specific. Refer to Table 1.4 for
an explanation of independent variable soil parameters……………………..39-40

Figure 2.5 Best model found for the Stepwise Maximum R-squared

Improvement for Principal Component 4. Refer to
Table 1.4 for an explanation of independent variable soil
parameters……………………………………………………………………….40

Figure 3.1 Best equation found for the Stepwise Maximum R-squared

Improvement for Principal Component 1. This equation is
best described as a mesic model………………………………………………………………44
vii

Figure 3.2 Best equation found for the Stepwise Maximum R-squared

Improvement for Principal Component 2. This equation is
best described as a northern dry forest model…………………………………………..46

Figure 3.3 First equation found for the Stepwise Maximum R-squared

Improvement for Principal Component 3. Comparison of
number of terms in the equation and C(p) value indicates
that this equation may be over specific………………………………………………47-48

Figure 3.4 Second equation found for the Stepwise Maximum R-squared

Improvement for Principal Component 3. Comparison of
number of terms in the equation and C(p) value indicates that
this equation is most likely not over specific. This is
considered the more conservative equation for this principal
component when compared to the equation in Figure 3.3.
This equation is best described as a wet mesic model…………………………48-49

Figure 3.5 “Behavior of major tree species on the combined ordination

of northern upland and lowland forests. 1, tamarack
(Larix laricina); 2, black spruce (Picea mariana); 3, white
cedar (Thuja ocidentalis); 4, hemlock (Tsuga canadensis); 5,
sugar maple (Acer saccharum); 6, white pine (Pinus strobus);
7, red pine (Pinus resinosa); 8, jack pine (Pinus banksina)”
(Curtis, 1959 pg. 180)……………………………………………………………………………...50

Figure 3.6 Plot of hypothetical values to illustrate potential findings of

principal component 2 and principal component 3 as separate
axes and a combined axis that shows the most reliable jack pine
axis…………………………………………………………………………………………………………51

Figure 3.7 Jack pine model derived from multiplying the second principal

component model by the most reliable model for the third
principal component………………………………………………………………………………..53

Figure 3.8 Best equation developed from the Stepwise Maximum

R-squared Improvement of Principal Component 2. This
equation has been selected as an example to illustrate
methods of interpreting vegetative productivity equations……………………….54

viii

INTRODUCTION
The proliferation of mined landscapes and concern for the environmental impacts
associated with these lands have led to an increased interest in developing empirical predictive
models to quantitatively assess the vegetative productivity potentials of reconstructed soils (neosols). Accurate modeling can be used to evaluate the conditions that are most influential to plant
growth in a given region and can aid in the creation of an optimum post-mine land use plan.
Landscape applications of this approach include agricultural uses as well as forested lands for
timber production, habitat creation, visual quality, carbon sequestration, and watershed
management. These equations provide an inexpensive alternative to many of the currently
utilized methods used to assess the potential of disturbed lands to reach productivity levels
greater than or equal to pre-mined conditions. This thesis examines the current knowledge base
associated with reclamation productivity equations, as well as investigates practical topics that
can lead to better understanding of the application of these equations. The research examines the
application of a methodology published by Burley and Thomsen (1987) in a multicounty mining
region in Michigan’s Upper Peninsula.

1.1 UNDERSTANDING RECLAMATION TERMINOLOGY
There are several terms that are used interchangeably throughout the collective body of
land reclamation literature. It is therefore important to address the nomenclature associated with
the renewal of disturbed land resources. Reclamation is most accurately used to describe
instances in which the post-mine land use is planned to be different from its pre-mine land use.
Rehabilitation and restoration are most often used to describe returning land to a previously
realized use. The term renewal has been used most often in instances in which there are some
1

type of built development associated with the project, or in cases in which a defined land use has
yet been established. Throughout this thesis reclamation will be used as an all-inclusive term
describing all activities that repair, redefine, or otherwise enhance disturbed lands, although
discuss will heavily focus on the establishment or reestablishment of vegetation on mined lands.

1.2 CONDITIONS OF RECONSTRUCTION SOILS
Numerous mining laws mandate that soil profiles be separately stripped and stockpiled
and subsequently reapplied during the reclamation process. There is inevitably some degree of
mixing of the soil profiles throughout this process, thus creating new soil textures, profiles and
patterns. These neo-soils are often vastly different from their pre-mining conditions and are
associated with unsuitable pH levels, increased bulk densities, greater percentage of rock
fragments, decreased water-holding capacity, lack of organic content and lack of microbial
activity (Bradshaw and Chadwick, 1980; Evanylo et al., 2005).
The pH of mined soils can change as newly exposed rock is weathered and dissolves.
Pyritic minerals oxidize to form sulfuric acid, which can rapidly lower the pH of a soil. For the
majority of plants, when soil pH is reduced below 5.5, plant growth is hindered due to metal
toxicities such as manganese or aluminum, phosphorus fixation, and reduced populations of Nfixing bacteria (Sheoran, 2010). Minerals containing carbonate, which is found in much of the
bedrock within Michigan’s Upper Peninsula, increases pH as they are weathered. Most plants
grow in pH levels ranging from slightly acidic to slightly alkaline, so breakdown of newly
exposed bedrock has the potential to greatly influence plant productivity.
Soil fertility is of major concern to post-mined soils. As previously mentioned, many
state laws mandate that soil horizons be stripped and reapplied in a specific order. Once these
2

horizons have been removed they are stockpiled until the land is reclaimed. While stockpiled,
nutrients may be leached from the soil, topsoil is eroded, and nutrient cycles are disrupted.
Reclamation specialists often plant cover crops on soil stockpiles to slow these processes, but
there is inevitably some degree of degradation when soils are disrupted. Macronutrients such as
N, P and K are commonly found to be deficient in reconstructed soils (Coppin and Bradshaw,
1982; Sheoran et al., 2008). In many cases, such nutrients are reintroduced through the
application of fertilizers to establish plant cover and initiate nutrient cycling.
One of the most influential edaphic factors that reclamation specialists must address
during reclamation efforts is the effect of increased bulk density of soils. Bulk densities are in
many cases drastically increased as heavy machinery involved in earth moving practices compact
soils. Compaction reduces the amount of pore space within the soil, thus limiting the amount of
water the soil is able to hold and make available to plants. Potter and others (1988) found that
reconstructed soils have greatly reduced hydraulic conductivity, largely due to increased
macropore space with increased bulk density. This reduction in pore space also reduces soil
aeration. Rooting of plants is also hindered by soils with high bulk densities. Many plants will
not penetrate soil profiles that are severely compacted. Sheoran (2010) suggests that three to
four feet of non-compacted media are needed to maintain available water contents adequate to
support plant growth through periods of drought. Darmody and others (2002) suggest that in
instances in which compaction caused by machinery cannot be avoided, that tillage of the first
120 cm can assuage the negative effects of highly compacted soils.
Microbial activity in reconstructed soils is greatly reduced in comparison to undisturbed
soils. Microbial activity has been found to decrease with depth and as the time soils are
stockpiled increase (Harris et al., 1989). Microbe populations are associated with maintaining
3

soil structure, decomposition of organic matter, as well as facilitating uptake of nitrogen and
phosphorus. Reduced pore space of compacted soils decrease the amount of spaces in which
these organisms survive (Edgerton et al., 1995). Nutrient content of soils is closely linked to soil
microbial activity. For instance, N-fixing microorganisms are needed to fix nitrogen in the soil
-

into nitrate (NO3 ). Without sufficient populations of these microorganisms soils will be
deficient of nutrients necessary to support vegetative growth.
Comparing the productivities of reconstructed soils to those of undisturbed soils is often
difficult due to the inherent differences between these soil conditions. Reconstructed soils are
often associated with soil attributes that are vastly different from pre-mined conditions making it
difficult for reclamation specialists to accurately predict the abilities of these soils to support
plant growth and to what degree. Potter and others (1989) suggest that it is possible to predict
post-mine productivity based on assessing soil attributes, however this practice is still being
refined.

1.3 MINE RECLAMATION LEGISLATION
Federal involvement in surface mine reclamation began with proposed legislation in the
1940s, which eventually led to the passage of Surface Mining Control and Reclamation Act of
1977 (SMCRA). Prior to the passage of the SMCRA there was no national regulatory system for
mine reclamation in the U.S. Many mined lands were abandoned after mining operations ceased
with no plans for reclamation of the disturbed landscapes.
Early mining law in the United States focused heavily on land ownership and mineral
rights and did little to address the reclamation of disturbed lands. As the federal government
became increasingly involved in leasing public lands to mining operators, the environmental
4

impacts and regulation of these lands became increasingly apparent to the scientific community.
In the early 1900s, ecologists, planners, and landscape architects throughout the United States,
and elsewhere, became increasing concerned with the unproductive and hazardous nature of
these lands and began raising public concern for ways to assuage the environmental impacts
associated with mined landscapes. However, the development of legislation in response to this
movement toward land reclamation was slow and prior to the SMCRA, made little effort was
made to make mine operators responsible for the reclamation of disturbed lands. In 1914,
Congress passed an Act that made one of the first moves toward making the mining prospector,
the responsible party for damages, to the disturbed lands. The Act focused on the disturbance of
productive agricultural land, mandating a separate disposal of the surface of the land to the
original agricultural landowner prior to the removal of any minerals. Furthermore, the Act
provided that the mining operation was responsible for compensation to the agricultural
landowner for any damages to the production of the land (Colby, 1945). Although primary
intent of this legislation was not to mitigate the environmental impacts of disturbances associated
with the mining practice in particular, it is used as an example to mark an important period in
mining law in which the scientific and political communities began to actively seek change in the
environmental policy of mined lands.
Legislation that directly addressed the reclamation of mined lands first occurred at the
state level. The oldest of such state laws, was passed by the state of Indiana in 1941. Several
additional states enacted similar laws in the years to follow. These early reclamation laws
focused primarily on establishing vegetative cover, without any major concern for slope
management of spoil piles. Dissatisfaction with the resulting topography from these early
reclamation efforts led to amendments in various state laws in the early 1950s, which required
5

lands to be returned to particular ranges of slope (LaFevers, 1977). Over the next two decades,
numerous states passed and amended regulatory legislation concerning post-mined lands.
However, inconsistencies among these state laws suggested need for national legislation to
establish minimal criteria for reclamation.
With growing political concern for environment protection in the 1960s, the federal
government began proposing legislation that would create a federal law to regulate mining
operations and reclamation of post-mined lands. The House of Representatives’ Bill 25, which
proposed such legislation, was voted in by congress but subsequently vetoed by President Ford
in May of 1975. However, the idea for federal surface mining regulation was reintroduced as
H.R. 2 to President Carter soon after taking office. The bill was quickly signed into law
thereafter on August 3, 1977 (Simpson, 1985).
In accordance with the National Environmental Policy Act of 1969, the federal Office of
Surface Mining (OSM) released an Environmental Impact Statement (EIS). The EIS released by
the OSM in 1979 outlined the purposes of the Act as to: “1) establish a national program to
protect society from the adverse impacts of coal mining; 2) where reclamation as required by the
Act is not feasible, to prohibit mining; 3) to require contemporaneous reclamation (no lag
allowed); 4) to balance coal production with the preservation of agricultural lands; 5) to assist
states in developing, administering, and enforcing a regulatory program; 6) to reclaim abandoned
mine lands; 7) to insure public participation in the development of regulations, standards and
programs within the SMCRA” (Burley, 2001 pg. 77). A full list of objectives is listed in the
official congressional Statement of Purpose of the SMRCA of 1977. The implicit intent of the
bill was not to prevent the surface mining of coal, but rather to set boundaries for these

6

operations that expressed the societal inclination that reclamation must be an integral component
of mining operations (Munshower and Judy, 1988).
Although the SMCRA developed a national regulatory system for mine reclamation, the
act imparted the primary responsibility of surface mining regulation and reclamation standards to
state governments. Congress authorized state governing bodies to administer surface mine
regulations as each deem fit so as to address the unique natures of their states’ various terrains,
climates, biological factors, geochemical factors, among other conditions that affect coal mining
areas (Workman, 1987). State regulations, therefore, vary considerably causing much debate
over the effectiveness of the SMCRA. Some critics argue that the SMCRA of 1977 and many
state mine reclamation laws overly emphasize water quality and erosion control, often
compromising site productivity, reforestation, carbon sequestration, and seeking alternative
productive land uses (Rodrigue and Burger, 2004). Other criticisms focus on the confusion the
duality of regulation that state and federal mandates can impose on mine operators (Lucas,
1987). Despite perceived flaws, the SMCRA of 1977 currently sets minimal standards for
surface coal mining operations and reclamation efforts in the U.S. All other additional mandates
are the responsibilities of the states to administer.
Although the SMCRA of 1977 requires mined lands to be reclaimed, the language that is
used leaves a significant amount of interpretation to be had in regards to what is considered
reclaimed. Section 515 of the SMCRA of 1977 requires mining operations to “restore the land
affected to a condition capable of supporting the uses which it was capable of supporting before
any mining, or higher or better use”. The inexplicit condition “higher or better use” can be
interpreted in many ways. When considering an alternative uses for post-mined lands there are a
variety of issues reclamationists explore to determine if the use is indeed “better”. Smith (2001)
7

presents a Canadian case study in which describes productivity calculations that assess the
productivity of a pre-mined land use of a forested landscape when compared to its post-mined
land use as a cattle ranch. However, comparisons among less comparable factors such as
environmental benefits (wildlife habitat, stormwater management, carbon sequestration, etc.),
economic benefits (crop and fiber production, tourism, job creation, etc.), and social benefits
(recreation, aesthetics, education, etc.) may all be considered when evaluating the
appropriateness of an alternative post-mined land use. Actually quantifying these factors is
difficult in many cases. Moreover, accurately predicting the probability that these land uses can
be brought to fruition can be particularly challenging and costly. Due to the intentionally vague
language of the SMCRA, states have formed their own laws that address the distinctive concerns
regarding mine reclamation.

1.4 MICHIGAN RECLAMATION LEGISLATION
Currently mining in the state of Michigan is regulated and enforced by the Michigan
Department of Environmental Quality (MDEQ). Regulation of mining operations and
reclamation is mandated under Parts 615, 625, 631, 632, 635, and 637 of Public Act 451, the
Natural Resources and Environmental Protection Act (NREPA). These parts regulate the mining
of oil and gas, mineral wells (mineral exploration), general mine reclamation rules, nonferrous
metallic mining, coal mining, and sand dune mining respectively. Although regulations of
mining types vary to some degree, due to the means of extraction and geological conditions
associated with the location of the mined material, all types require the submittal and acceptance
of proper permits including a reclamation plan before extraction begins.

8

MDEQ require mining companies to submit reclamation plans as part of the permit
žek et al., 2012, Beauchamp et al., 2006; Tripathy et al., 2008).
However, regionally based field testing should be conducted to verify the interaction of these
amenities with other soil properties.
The law further specifies that the reclamation plan will restore “the productivity of the
land prior to mining, based on the average yield of food, fiber, forage, or wood products
consistent with productivity of similar lands in this state under best management practices”, as
stated in Michigan Public Act 451 of 1994 part 635 section 324.63518.iii. Expected biomass
production, many times in the form of crop yields, is used to establish a predicted level of
productivity that the post-mined land can support. However, there is no specified, appropriate
method for predicting productivity levels within the law. In many cases, productivity of
reference areas is used to compare to post-mined lands. Doll and Wollenhaupt (1985) criticized

application process. Part 632 section 324.63527 [2.b] of the NREPA requires submission of a
reclamation plan that will effectively “restore the land affected to a condition capable of
supporting the uses that it was capable of supporting prior to any mining, or higher or better
uses”. The vague language “higher or better use” is echoed from the 1977 SMCRA. Although
this undefined language gives reclamationists flexibility when determining post-mine land uses,
it lacks identification of proper means of quantifying productivity. Additionally the law requires
that A and B soil horizons be stripped, stored, and reapplied separately, unless information can
provided that an available alternative would be more supportive to vegetative growth. The
flexibility in this section gives reclamationists the opportunity to explore different substrates to
be applied as growth media. A considerable amount of research has investigated the applications
of various alternative substrates as growth media or soil amendments (Zornoza et al., 2012;
Watts et al., 2012;

9

this method, describing it as unreliable and costly. Despite the perceived flaws of the methods
used, P.A. 451 Rule 9 of section 425 empowers “the supervisor of reclamation” to evaluate the
reclamation plan to determine if it adheres to the mandates of the act. It is the authority of this
person to assess the productivity of the land prior to mining, yet not acceptable evaluation
method is specified. It is therefore beneficial to explore accurate methods of quantifiably predict
vegetative productivity of post-mine soils based on observed crop yields within the mining
region.
Burley (2001) compares several state reclamation laws. In this Burley acknowledges the
applicability of a regression approach to many state mandates that require comparative
productivity level studies of pre-mined and post-mined lands. Although vegetative productivity
models, such as the one explored in this research, are not yet required by any state, these models
show great potential to accurately predict state mandated vegetative success rates.

1.5 RECLAMATION EVALUATION APPROACHES
Reclamation specialists are concerned with creating models to predict vegetative
productivity. Essentially, four general ecological model types have been developed that can be
applied to predict post-mining soil productivity to assess vegetation growth of agricultural crops,
rangeland plants, and woody plants (Le Cleac’h et al. 2004; Burley et al., 2001). Although some
states’ laws explicitly express the method to compare the productivity level of pre-mined and
post-mined soil productivity levels, each of these models has been used in past reclamation
efforts throughout the United States. Recognition and understanding of these model types gives
reclamationists a broader understanding of past and present reclamation practices and inherent
advantage and disadvantages of each described.
10

The first model is a heuristic method, known as the “reconstructing nature” approach.
Federal and many state laws mandate that mine operators strip and stockpile A, B and C
separately so that these horizons may be reapplied in the correct order during reclamation.
Although stratified stockpiling is supposedly an important and beneficial process, results are
extremely variable with limited ability to predict vegetative performance and have not been
validated by adequate research. Dancer and Jansen (1981) as well as McSweeney and others
(1981) discovered that replacement or alteration of claypan subsoils commonly found in southern
Illinois improved the chemical and physical properties of mined land. Here, mixtures of B and C
horizons showed greater vegetative growth compared to B horizon materials alone. Stripping
and reapplication of A horizon material is extremely important and a heavily used method in the
reconstruction of prime farmland as a post-mined land use. However, mandates of topsoil depths
to be respreads are dependent on the characteristics of the soil material and the crops to be grown
(Friendland, 2001). Mandated depths vary depending among state due to differences in edaphic
parameters including soil depth to bedrock or depth to toxic materials.
Post-mined soils are often vastly different from their pre-mining conditions (refer to
section entitled Conditions of Post-mined Soils for a description of the potential differences and
causes for these differences among post-mine and pre-mine soils). Recreating soil system with
similar productivity levels based on attempted recreation of pre-mine conditions is therefore,
difficult. The first model, although requiring minimal planning, has proven to be a poor
predictor of vegetative performance.
The second model is a statistical comparison method known as the “reference site”
approach. In this approach soils from an undisturbed site are compared with the same or a
similar, nearby mined site. If no statistical difference exists between undisturbed soils properties
11

and those of post-mined soils, the reclamation effort is deemed acceptable. Zipper and others
(2011) investigated the reforestation potentials of 25 sites in the Appalachian region of the
eastern United States and found that post-mined landscapes were comparable to nearby
undisturbed lands. The results of the study indicated that post-mine soils have the capacity to
support productive forest vegetation and identify conditions that affect growth. However, such
studies do not give a quantified prediction of vegetative productivity potential and are associated
with extensive and costly data collection (Burley, 2001). Doll and Wollenhaupt (1985)
described the reference site approach as “an unreliable and expensive means of evaluating
reclamation success”. Although this model is appropriate, and widely used to assess post-mine
growth potentials, it does little to predict post-mine vegetative productivity.
The reference site approach often determines successful restoration based on a
comparison of post-mining crop yields to those of similar, undisturbed reference areas. Once
adequate or comparable yields have been achieved, the land is considered reclaimed, thus
satisfying the obligations of the mining operation. Upon satisfaction of reclamation obligations,
the bond held by the governing body is released, withdrawing all financial connection to further
reclamation. Such post-reclamation evaluation methods to determine neo-sol productivities are
often conducted once the opportunities to cost effectively amend the soils are lost (Burley, 1987).
It is therefore, important to investigate methods to that are capable of accurately predicting
vegetative productivity of post-mined soils so as to give reclamation planners opportunities to
address identified soil factors that inhibit optimal growth prior to the establishment of a postmined land use.
The third model, the “sufficiency” approach, employs a series of expert derived tables
and charts. Here the soil is evaluated based on defined criteria and either accepted as sufficient
12

or rejected. This approach has been used to evaluate vegetative productivity potentials by
assigning point values to particular soils attributes, and is commonly used by foresters to predict
biomass production rates of various tree species. Foresters often investigate volume
measurements, plant indicators, and height growth indexes to evaluate sites (Woolery et al.,
2002). Baker and Broadfoot (1978, 1979) used the sufficiency approach to develop methods to
predict potential tree heights based on soil data published by the Natural Resource Conservation
Service (NRCS). The methodology, as defined by Baker and Broadfoot (1978, 1979), assigns
point values representing conditions that are good, fair, and poor for vegetative growth. These
values were weighed depending on importance to plant growth and totaled to determine the
predicted growth rates of a given tree specie. Neill (1979) used a similar index methodology.
Neill’s work led to the identification of most of the soil factors useful for the regression analysis
approach. Doll and Wollenhaupt (1985) later refined Neill’s index approaches and applied the
method to post-mine soils. However, the equation derived by Doll and Wollenhaupt is heuristic
and hypothetical (not statistically validated) and could therefore not be mathematically applied to
actual situations. More recent studies have been employed the index model with research
conducted in the Mississippi River Valley (Belli, 1998; Groninger, 2000). Similar to other
heuristic index models, these studies lack statistical reliability yet established fundamental
precedents leading to creation of more accurate statistical models. However, Potter and others
(1988) argue that this model is beneficial in cost-effectively assessing the amendment and
treatment needs of post-mine soils based on pre-mined soil conditions.
According to Burley and Thomsen (1987), index productivity models, do not address at
least six major points. 1) Soil attribute interactions are not accounted for. Soil attributes can
interact independently with crop production, grouped with other soil attributes interacting
13

collectively but independent from other soil attributes, or completely interdependent. For
instance, research shows that electric conductivity (EC), is a soil attribute that correlates with
many soil properties that exist in most agriculturally supportive lands such as clay content,
organic matter, and bulk density, but is traditionally used to quantify soluble salt contents
(Corwin and Lesch, 2005). However, the degree in which these attributes interact together to
affect plant growth is unknown. These attributes may be further complicated by secondary
interactions with additional attributes as well. In many existing index models, soil attributes are
all multiplied as though it were one interaction model despite a lack of statistical basis to do so.
2) Soil attributes may also exhibit non-linear responses such as squared terms, which are not
represented in index models. Squared terms represent soil attributes that increase or decrease
productivity to a certain point and then reverse its trend. These attributes are extremely
important when explaining instances such as when moderate amounts of available water content
aids in a plants ability to grow, yet when a point is reached in which soils are overly saturated,
plant growth is hindered. 3) Constants, also known as beta coefficients, are absent from index
models, yet it may be appropriate to consider slope constants and intercepts. 4) Interactions of
crop types are not reflected in index models. Soil attributes may affect various crop types
differently and therefore warrant separate equations to describe the predicted productivities.
This is discovered in statistical models by investigating what crop types covary together. If two
or more crop types covary they can be described by the same equation. 5) Regionality may be a
consideration when selecting soil attributes to investigate. Some attributes may not be
significant to include in certain regions while important to other models. Electric conductivity,
as an indicator of problematic soluble salt content is mostly associated with arid regions, such as
those found in the American southwest (Powers et. al, 1978). Electric conductivity may not be
14

beneficial to investigate if in a region that soil salinity is not as prominent such as the American
northeast. Researchers must consider regional differences that significantly affect edaphic
conditions when determining characteristics to investigate. 6) Finally, the significance of the
attributes is not considered. Soil attributes may vary in significance or not included at all in the
equation. For example, crop yields of specific plant groups may be reduced if available water
levels are too high or too low. Although available water is known to effect plant growth, when
investigated with other attributes such as bulk density and percent organic matter, available water
may not contribute to the accuracy of the model and therefore be omitted from the final equation.
In addition to these six points, Burley (1995) also describes his concern for restrictions of
variable exploration. Firstly, soils explored in this approach are often restricted to a limited
number of soil types. Secondly, vegetation variables are often restricted to one type that
describes all crop types. Other approaches may be further investigated for a broader range of
soils and a more focused group of plant types.
The fourth and final model describe here is the statistical “regression analysis” approach,
in which empirical evidence is used to develop equations that predict vegetation performance.
Similar to the reference site approach, the regression analysis approach is associated with
extensive data collection, which can take years to collect. Development of a reliable statistical
productivity equation requires a data set derived from growth of various species across all soil
types averaged over a ten-year period (Burley, 2001). For this reason many researchers have
pursued sufficiency models in lieu of investigating regression models. The Natural Resources
Conservation Service (NRCS), however, has compiled substantial county-based data that may be
applied to alleviate the costs associated with the data collection process necessary for the develop
of an accurate model employing the regression analysis approach. Similar studies have
15

employed (NRCS) data to reduce the costs associated with data collection need for statistical
analysis of vegetative productivity potentials (Le Cleac’h et al., 2004; Woolery et al., 2002;
Burley, 2001; Groninger et al., 2000, Burley et al., 1996; Burley, 1995a & b; Barnhisel and
Hower, 1994; Burger et al. 1994; Burley and Bauer, 1993; Barnhisel et al. 1992; Burley, 1991;
Gale et al., 1991; Burley, 1990; Burley and Thomsen, 1990; Burley et al., 1989). Studies by
Woolery and others (2002) and Groninger and other (2000) used the regression approach to
predict tree growth in southern Illinois. These studies used regression of a combination of
collected soils data and expect derived indexes. Although their methods are considered
scientifically accurate, the methodology described by Burley and Thomsen (1987) limit the
amount of theoretical information (derived indexes) by limiting the amount of qualitative
assumptions about variables explored.
Burley and Thomsen (1987) describes the methodology for establishing empirically
derived equations for predicting vegetative productivity. The first equation to use this
methodology was developed by Burley et al. (1989) and is used in this research to investigate
vegetative productivity on an area in Michigan’s Upper Peninsula.

16

METHODS

2.1 METHODOLOGY
Independent variables are representative of physical and chemical characteristics of each
soil investigated. Soil factors included in similar studies include topographic position, percent
slope, percent organic matter, bulk density, soil reactivity, percent clay, percent rock fragments
greater than 3 inches in diameter, hydraulic conductivity, and available water content. These
characteristics were selected based on the work of Neill (1979), and later refined by other
researchers Doll and Wollenhaupt (1985). Data for soils are available in published NRCS
county soil surveys, which provide these characteristics for nearly all soils included in the
surveys. Other soil characteristics may be investigated if data is available. Additionally, the
values for the soil characteristics must be available for the first 48 inches of each soil type. In
some cases, availability of soil characteristic values was limited by the depth of the soil type.
This was often due to shallow soils over bedrock or high water tables. The values that were not
included in the NRCS tables must therefore be interpolated based on other empirical evidence.
Soil productivity has been linked to soil depth. According to research conducted by Doll
and other (1984), soil depths affect vegetative productivity in different proportions. The research
describes a soil weighing method in which soil characteristics are assessed based on depth. This
method contributes 40% of the total crop yield to the first 12 inches of the soil profile. Inch 13
to inch 24 contribute 30% of the crop yield, while inches 25 through 36 account for 20% of the
yield. Finally, inches 37 through 48 contribute 10% of the crop yield value. This research
suggests that 100% of the crop yield value can be attributed to the first 48 inches of the soil
profile. Although it could be argued that this weighing method is flawed as being only an
17

estimate of the relationship between soil depth and plant productivity, there has been no other
method found that can provide a more accurate means of this affect. This methodology employs
this soil characteristic weighing method to gain a more accurate assessment of soil-plant
correlations.
To be statistically analyzed variables must be represented as a single value. Deriving
single values NRCS soil surveys is complicated by the way in which the data provided is
presented. Firstly, soil attributes are often given ranges of values as opposed a single value.
Bulk density, for example may be given a range of 1.0 g/cc to 2.0 g/cc for a given depth. This
range must be averaged so as to generate a single value, a value of 1.5 g/cc in this example.
Secondly, the chemical and physical soil characteristics provided in NRCS soil surveys are often
presented according to soil profile depths. For example, a characteristic such as pH may be
given for depths of 0 to 8, 8 to 25, 25 to 32, and 32 to 80 inches. These values must be
reorganized into appropriate depth ranges so they can be weighed according to the weighing
method used in this methodology.
Dependent variables in the vegetative productivity models are typically crop harvest data
and woody plant growth rates. Crop yield data should be collected over several growing seasons
including years of differing amounts of precipitation and temperature ranges, such as data
provided by the NRCS. Gaining data that has been collected over several seasons, gives a more
accurate understanding of crop yields in a typical year. It is important to note that crop yields
must be actual measured quantities and not derived from another index. Modeling of values
derived from other indexes will result in similar equations to the existing index, and thus result in
no further understanding of the data. Crop harvest data may be expressed in terms of various
units, most commonly bushels per acre, tons per acre, or pounds per acre. The NRCS presents
18

woody plant growth data in terms of feet of growth per year, although other growth rates or plant
volumes could potentially be used. Dissimilar units of measurement among vegetative types do
not necessarily pose a problem.
Units of measurement may vary among both independent variables and dependent
variables. For example, discrepancies in soil characteristic (independent variable) units include
g/cc for moist bulk density and percentages for characteristics such as for clay content and
organic content. Similarly, crop data can be presented as tons per acre or bushels per acres. It is
therefore necessary to standardize by a mean of zero and a variance of one all variable to
accurately compare variables with different units of measurement. If variables are not
standardized, variable with higher real number values will dominate the equation.
Furthermore, these larger real numbers may not even be greater quantities than their
smaller counterparts. Crop yields, for instance, may be presented as 75 bushels per acre of oats
and 13 tons per acre of corn silage for a given soil unit. In this example, 13 tons of corn silage is
a much larger quantity than 75 bushels of oats, but this is not reflected in the real number values
of the data. According to the United States Department of Agriculture (USDA, 1992), one ton of
oats contains 68.8944 bushels. By converting bushels of oats to tons of oats per acre, a value of
approximately 1.1 tons per acre of oats is found. The converted value of 1.1 is a much smaller
real number than 75, and now oats has a much smaller relative value when compared to the 13
tons per acre value of the corn silage. Such conversions are time consuming and in some cases
are not able to be made, such as converting units of g/cc to In/hr. Standardizing crop and soil
data alleviates the complications associated with different units and makes it possible to analyze
variables with different units. All variables must be standardized to be converted to z-scores of
the crop yield or soil characteristic sets of observations.
19

Once the data is standardized and organized it is ready to be analyzed. At this point it is
possible to create regression equations for each crop type. However, creating separate equations
may not be necessary. It may be possible to combine multiple crop types into one equation. By
investigating the multivariate relationship among crop types using Principal Component Analysis
(PCA), it is possible to identify crop types that covary and can therefore be combined into one
equation. PCA is used to determine the number of dimensions needed to explain the variance
across all crop types. Simply put, PCA is a dimension reduction tool. In this method each
dimension explains a set a crop types that covary together. Ideally, all crop types will covary
and investigators will be able to derive one equation that will explain all crops from one
dimension. If all crops do not covary the investigator may have to develop additional equations
to explain one or multiple significant dimension. For instance, if PCA shows that soybean and
corn silage covary together, while white pine (Pinus strobus L.) and red maple (Acer rubrum L.)
covary together, these four crop types can be described within two different dimensions.
Therefore, two separate equations will need to be developed to explain all four crops. In this
way PCA is extremely helpful to simplify large amounts of variable.
The PCA procedure begins with a selected set of dependent variable to be investigated
being organized into a covariance matrix. PCA of the matrix will generate a set of eigenvalues
that represent each dimension of the dataset. The maximum number of dimensions is equal to
the number of crop types (dependent variables) in the dataset. This means that if a total of seven
crop types are used, there can be no more than seven dimensions. The largest eigenvalue will be
the first value of the PCA results and can be no greater than the number of variables investigated.
Furthermore, the sum of all eigenvalues can equal no greater than the total number of variables.
For example, if there are three vegetative types used and the first eigenvalue is 2.6, the sum of
20

the remaining two eigenvalues must be equal to 0.4. The proportion of each eigenvalue of the
total dimension illustrates how much of the variance is explained by that eigenvalue. These
values can also be added to gain a perspective on how much a group of variables explains. For
instance, if the first five of nine eigenvalues represent 90 percent (a combined value of 8.1) of
the variance, the remaining four eigenvalues represent only 10 percent (a combined value of 0.9)
of the variance. In this way researchers are able to interpret the significance of eigenvalues so as
to more accurately make decisions on what is significant to the study.
Eigenvalues greater than 1.0 are considered to represent significant dimensions worthy of
further investigation. In smaller data sets, those consisting of less than 100 soil types,
eigenvalues greater than 0.8 should be selected. However, it is recommended that a data set
consists of approximately 100 independent variables for accurate modeling. Table 1.1 gives an
example of an eigenvalue table that examine 11 crop types. The results in Table 1.1, Table 1.2,
as well as in Table 1.3 were taken from preliminary trials of data used in this study, but were not
used in further investigations. These results are used strictly for illustrational purposes and do
not reflect any actual results used to develop equations in this study. The first four eigenvalues,
represented by Prin1, Prin2, Prin3, and Prin4, are significant enough to be considered for further
modeling analysis (>1.0). Additionally, notice the results give the relative proportion of the data
that each dimension explains. Collectively the first four eigenvalues explain more than 79% of
the variance.

21

Table 1.1 Example eigenvalue table. These results are strictly to illustrate the methods of this
study and do not reflect any actual values derived from this investigation.
Eigenvalues of the Covariance Matrix
Eigenvalue Difference Proportion
Prin1
Prin2
Prin3
Prin4
Prin5
Prin6
Prin7
Prin8
Prin9
Prin10
Prin11

3.46188048
2.55814253
1.86736487
1.33715712
0.68719124
0.51084934
0.35876282
0.31095943
0.27769709
0.13070932
0.12576630

0.90373795
0.69077766
0.53020775
0.64996588
0.17634190
0.15208652
0.04780339
0.03326234
0.14698777
0.00494302
0.12576630

Cumulative

0.2978
0.2200
0.1606
0.1150
0.0591
0.0439
0.0309
0.0267
0.0239
0.0112
0.0108

0.2978
0.5178
0.6784
0.7934
0.8525
0.8965
0.9273
0.9541
0.9779
0.9892
1.0000

Significant dimensions are then investigated by examining the eigenvector coefficients.
Table 1.2 investigates the four significant dimensions from Table 1.1. Eigenvector scores range
from 1.0 to -1.0. These scores indicate the strength of the association with the corresponding
principal component axis. Scores close to 1.0 or -1.0 are considered to be strongly associated,
while scores closer to 0 are considered weaker. Positive scores indicate a positive association
and negative scores indicate negative associations.

Table 1.2 Example eigenvalue chart. These results are strictly to illustrate the methods of this
study and do not reflect any actual values derived from this investigation.
Eigenvectors
Prin1
Corn
Corn Silage
Oats
Irish Potato
Alfalfa Hay

Prin2

0.383742
0.477603
0.328985
0.457793
0.363850

-.103134
-.186402
-.223164
-.099761
-.099289
22

Prin3
-.032841
-.000236
0.045024
-.372264
-.162625

Prin4
-.486517
-.390847
-.072393
0.385713
0.376985

Table 1.2 (cont’d)
Red Maple
White Spruce
Red Pine
Eastern White Pine
Jack Pine
Lilac

0.144508
0.220520
0.152413
0.226707
0.073695
0.157724

0.078080
0.027231
0.522985
0.482660
0.428711
0.433474

0.608105
0.435188
-.101128
0.222298
-.394658
0.239349

-.015604
0.455841
0.080511
0.108421
-.129633
-.267075

Interpretation of the eigenvector analysis allows researchers to understand the potential
for each principal to be translated into a vegetative productivity equation and which of the
dependent variable developed models will describe. It is important to understand how to
interpret these results so as to create models that will be most helpful for specific intent. By
understanding ways in which crop types covary, researchers are better able to derive equations
that will predict vegetative growth for particular planning objectives such as agriculture lands,
habitat creation, aesthetics, carbon sequestration, or growth for lumber production. It is
suggested that interpretation of the eigenvectors be conducted in order, as this is the order of
significance.
Eigenvector results are presented and can be interpreted in three main ways; all positive
numbers, some positive and some negative numbers, and one positive number with all other
numbers negative. All positive values, especially in the first principal component column
reveals that all crop types covary in this dimension, suggesting that one model could be derived
to explain all crop types. It is preferred that all crop types can be explained in one equation, as in
the example of principal 1 of Table 1.2. Alone, however, this example principal only explains
29.78% of the total variance (refer to Table 1.1). Therefore, it may be useful to investigate
additional models. Because the first four principals in Table 1.1 have eigenvalues greater than
1.0, further analysis of these principals is recommended.
23

Both positive and negative values within one eigenvector, indicates a predictive model
that describes soil characteristics that support two contrasting sets of crop types. The second
principal of Table 1.2, for instance, shows that the first six crops covary together, all having
positive signs, while the last five crop types, all having negative signs, covary together. Notice
that all crop types with a negative sign are agricultural field crops, and that all those with a
positive sign are woody plants. Therefore principal 2 may be investigated further as a field crop
versus woody plant model in which positive scores show favorable conditions for woody plants
and negative score for conditions that favor field crop production.
Eigenvectors with one positive number and all the remaining values are negative, can be
further investigated to derive equations for one crop type. In these types of vectors, the equation
will be suitable for predicting the vegetative success of the crop type with the positive value.
Burley (1990) derived a single crop equation for sugar beets in a similar study.
Once eigenvectors (dependent variables) have been identified and soil characteristic data
has been weighed, the data set is ready for regression analysis. Regression analysis is used to
examine the ability of soil attributes to predict the vegetative productivity of investigated crop
types. This procedure identifies main affects (ex. % clay), squared terms (ex. % clay x % clay),
and two variable interaction terms (ex. % Clay x pH) as they correspond to dependent variables.
Important variables are then entered into a stepwise regression procedure. The maximum Rsquared improvement technique is used to create a list of possible equations.
Selecting the strongest equation is determined by selecting an equation with the largest
explanation of the data, represented as an R-squared score, for equations presented that consist of
significant p-values. P-values less than or equal 0.05 are considered specific, while those less
than 0.01 are considered highly specific (refer to Table 1.3). Burley and Thomsen (1987)
24

recommend searching for two types of models when considering possible equations. The first
type of model is one in which all p-values are less than 0.01 and an R-squared value greater than
0.7. If this does not exist, it is suggested that an equation in which all variables have p-values
less than 0.05 and has an R-squared value greater than 0.7.
2

Table 1.3 Example Stepwise Maximum R Improvement acceptable equation. These results are
strictly to illustrate the methods of this study and do not reflect any actual values derived from
this investigation.
Maximum R-Square Improvement: Step 26
Variable SLCL Entered: R-Square = 0.7563 and C(p) = 10.0819

Source

DF

Model
Error
Corrected Total

12
81
93

Sum of
Squares
45409
14632
60040

Mean
Square
F Value
3784.04297
180.64020

Variable

Parameter
Estimate

Standard
Error

Type II SS

Intercept
SL2
BD2
AW2
TPSL
TPFR
TPPH
TPOM
SLCL
FRBD
FRHC
FROM
BDAW

-52.67096
-0.20017
70.25217
2826.95353
1.51899
1.44884
-1.80908
-4.26685
0.08558
-6.97459
1.01052
1.35934
-636.30437

19.35937
0.04212
13.36404
895.83163
0.27008
0.34309
0.40871
1.06019
0.04244
1.31855
0.14887
0.59059
165.40282

1337.13197
4080.71941
4991.81143
1798.86660
5713.91270
3221.41360
3539.23420
2925.92286
734.56634
5054.26848
8323.08235
956.97113
2673.36574

20.95

<.0001

F Value Pr > F
7.40
22.59
27.63
9.96
31.63
17.83
19.59
16.20
4.07
27.98
46.08
5.30
14.80

Bounds on condition number: 86.929, 4546.4

25

Pr > F

0.0080
<.0001
<.0001
0.0022
<.0001
<.0001
<.0001
0.0001
0.0471
<.0001
<.0001
0.0239
0.0002

Table 1.3 illustrates an example of significant stepwise regression equation. From a list
of potential equations, researchers are to select the equation with the least significant p-values
that explain the most variance of the dataset. This example explains 75.63% of the variance (Rsquare = 0.7563). Additionally, all 12 of the variables are considered significant (<0.05) and all
but two variables are considered highly significant (<0.0001). This equation would be
considered acceptable to be developed into an accurate equation.
The resulting equations from the process described here employs soil parameters to
predict a productivity index. The productivity index is a unitless value that indicates relative
productivity of a given soil. Typically index scores have ranged from five to negative ten, with
scores near five indicating highly productive soils and scores near negative ten being highly
unproductive (Le Cleac’h, 2004). These scores should be used in conjunction with eigenvector
values to indicate the appropriateness of a soil to plant species that are associated to varying
degrees with the equation employed.

2.2 STUDY AREA
Figure 1.1 Location of study area in Michigan.
IRON

DICKINSON
DICKIN
SON

26

County selection for compiling a data set should be based on availability of data provided
by the NRCS, or similar reliable source, and proximity to the site of interest. Soil surveys
developed by the NRCS are generated after years of data collection and released as a county
based document. Although this research uses entire counties to develop data sets, it is possible to
create models based on specific geological, climatological, social, or other identified boundaries.
The area of study is located in the western Upper Peninsula of Michigan. The geology of
the region is largely composed of sedimentary and igneous rock that is covered by glacial drift.
Both of these types of rock are important to human activity. Sedimentary rock is often
associated with extractable commodities such as petroleum, natural gas, salt, gypsum, and
limestone. Igneous rock in the region is associated with extractable minerals, most notably, iron
ore and copper. The mining of iron ore and copper accounts for the majority of mined surface
area in the region, most of which occurs in Marquette County, Michigan.
Marquette County is of particular interest to reclamationists as it currently hosts a largescale iron ore mining operations and has recently opened a copper mine. However, the NRCS
has not released the appropriate crop and woody species data necessary for the methodology
employed in this study. Burley (1995a) found that multicounty models could be developed that
could be applied throughout a region. Therefore, two adjacent counties that provided the
appropriate yield and growth rate information, Iron County and Dickinson County, were
selected. All three counties mentioned are within the Menominee Iron Range with the Marquette
range in Marquette County, the western range in Iron County and the eastern range in Dickinson
County. The similar geologies of the three counties of interest lend themselves well to the
creation of a regional model.

27

Figure 1.2 Locations of mines and mine exploration sites in the Michigan’s western Upper
Peninsula. For interpretation of the references to color in this and all other figures, the reader is
referred to the electronic version of this thesis.

Historically, all three counties have been mined. Figure 1.4 illustrates the locations of
mines and exploration sites. Michigan is divided into four major metallic mining ranges, the
Copper Mining District of Keweenaw, Houghton and Ontonagon counties; the Marquette Iron
Range of Marquette and Baraga counties; the Menominee Iron Range of Dickinson and Iron
Counties; and the Gogebic Iron Range of Gogebic county. Although not within the same range,
the similarities in mining histories and extraction methods of ores makes the counties in which
data was derived from, Iron and Dickinson counties, and other mined counties within the region,
most notably Marquette County, comparable.

28

2.2a Iron County
Iron County has a long history of mining that began when surveyors reported abnormal
compass readings in 1846. Soon after, iron ore exploration led thousands to the area to prospect
the land, of what is now know as the western Menominee range. Lumbering also grew rapidly as
an economic activity. Extensive iron formations were discovered near the Iron River and Crystal
Falls. When the Chicago and Northwestern Railways reached the eastern range in 1882, miners
had reportedly extracted 74,000 tons of ore. Mining steadily grew over the next forty years when
it finally reached peak production in the 1920s (Linsemier, 1997). According to Linsemier
(1997), the Sherwood was the last active mine in Iron County. Currently, approximately 90
percent of the Iron County is forest, which is largely dominated by sugar maple (Linsemier,
1997).
Iron County is located in the high plateau region with land formations resulting from
continental glaciation. Major bedrock types include the Michigamme slate and associated
formations including greywacke, greenstone, and quartzite deposits. Although outcrops are
common throughout the county, most of the area is covered by glacial drift. The landscape is
characterized by rolling ground moraines, end moraines, steep ice-compact features, and outwash
plains. Soils are predominately Spodisols, characterize as fine sands to sandy loams in texture
(Albert, 1995). Linsemier (1997) describes the taxonomic classification of each soil type
examined in this study.

2.2b Dickinson County
Similar to Iron County, the early history of Dickinson County is largely shaped by its
lumber and mining activities. The contemporary economic activities within the county are
29

heavily dependent upon the tree growth for lumber, pulpwood, and fuel (Linsemier, 1989). The
physiology of the county is the result of continental glaciation, dominated by moraines, till
plains, and outwash plains. More than 90 percent of Dickinson County is forested, with a similar
species composition as Iron County. Albert (1995) describes the soils of the county to be thin
layers of sandy to loamy sand soils on bedrock. A description of taxonomic classifications for
all soils investigated in this study are provided by Linsemier (1989).

2.3 VARIABLES INVESTIGATED
Independent variables employed in this study included nine soil parameters. The soil
parameters investigated in this study were available water holding capacity, moist bulk density,
percent clay, percent rock fragments, hydraulic conductivity, percent organic matter, soil
reaction, percent slope and topographic position. Units and abbreviations of these variables as
they are shown in the following results are included in Table 1.4.

Table 1.4 Independent variable soil parameter units and abbreviations as they appear in the
regression models and final developed equations.
Abbreviation
AW
BD
CL
FR
HC
OM
PH
SL
TP

Factor
Available Water Holding Capacity
Moist Bulk Density
% Clay
% Rock Fragments
Hydraulic Conductivity
% Organic Matter
Soil Reaction
% Slope
Topographic Position

30

Unit of Measurement
Inches/inch
g/cc
Proportion by weight
Proportion by weight of particles >3
inches
Inches/hour
Proportion by weight
pH
(Rise/Run)*100
Scale 1 to 5 where:
1 = Lowlands (Bottomlands)
2.5 = Mid-slopes
5 = Highlands (Ridge lines)

There were a total of 95 soils investigated between Dickinson and Iron Counties. 45
applicable soils were investigated from Iron County, while 50 applicable soils were examined
from Dickinson County. Soils that did not have crop or woody plant data were not included.
These soils were most often lowland mucks that did not provide adequate soil characteristic
information. Additionally, soil complexes were not included in the data because the proportions
of the soils that comprised the complexes that the crop and woody plants were grown on are
unknown. Most soils that comprise these soil complexes were included in the data set as
singular soil types.
Dependent variables consisted of five crop types and six woody plant species. Crops
investigated in this study include corn, corn silage, oats, alfalfa/hay, and Irish potatoes. Woody
plant species investigated were red maple (Acer rubrum L.), white spruce (Picea glauca
pine (Pinus banksiana Lamb.), and lilac (Syringa vulgaris L.). A list of dependent variables and

[Moench] Voss), red pine (Pinus resinosa Aniton), eastern white pine (Pinus strobus L.), jack

their associated units and abbreviations as they are shown in the eigenvector analysis are
included in Table 1.5. These specific plants were selected for investigation due to the
availability of information of these plants in both counties.

Table 1.5 Dependent variable crop and woody plant units and abbreviations as they appear in the
eigenvector analysis.

Agronomic Crops
Abbreviation
CO
CS
OA
IP
AH

Crop Type
Corn
Corn Silage
Oats
Irish Potatoes
Alfalfa Hay

Measured Average Yield
Bushels/acre
Tons/acre
Bushels/acre
Hundredweights/acre, 100lbs/acre
Tons/acre
31

Table 1.5 (cont’d)
Woody Plants

Abbreviation
RM
WS
RP
EP
JP
LI

Crop Type
Red Maple
White Spruce
Red Pine
Eastern White Pine
Jack Pine
Lilac

Botanical Name
Acer rubrum L.
Picea glauca [Moench] Voss
Pinus resinosa Aniton
Pinus strobus L.
Pinus banksiana Lamb.
Syringa vulgaris L.

32

Measured Average Yield
Feet/20 years
Feet/20 years
Feet/20 years
Feet/20 years
Feet/20 years
Feet/20 years

RESULTS

Table 2.1 illustrates the eigenvalues of the combined Iron County and Dickinson County
crop and woody plant data set. There were 4 principal component axes with eigenvalues greater
than 1.0. The eigenvalue for the first principal component axis contains 28.80 percent of the
variance in the crop and woody plant variables. The first axis contains the largest proportion of
the variance in the data set and is therefore the primary candidate for further investigation. The
eigenvalue for the second, third and fourth principal component axes contained 23.32 percent,
16.15 percent and 10.94 percent of the variance respectively. The first four principal
components together comprised 79.21 percent of the variance in the data set.

Table 2.1 Principal component analysis eigenvalues of the covariance matrix for Iron County
and Dickinson County, Michigan.
Eigenvalue
Prin1
Prin2
Prin3
Prin4
Prin5
Prin6
Prin7
Prin8
Prin9
Prin10
Prin11

3.16822981
2.56469741
1.77698412
1.20355081
0.57141461
0.49160125
0.39617155
0.31037446
0.27346041
0.12819264
0.11532292

Difference
0.60353241
0.78771328
0.57343332
0.63213620
0.07981335
0.09542970
0.08579710
0.03691405
0.14526776
0.01286972
0.11532292

33

Proportion Cumulative
0.2880
0.2332
0.1615
0.1094
0.0519
0.0447
0.0360
0.0282
0.0249
0.0117
0.0105

0.2880
0.5212
0.6827
0.7921
0.8441
0.8888
0.9248
0.9530
0.9779
0.9895
1.0000

Table 2.2 Principal component analysis eigenvectors for Iron County and Dickinson County,
Michigan dependent variables. See Table 1.5 for explanations of variable abbreviations. The
number 1 attached to each variable indicates that that variable has been standardized to a mean of
zero and a standard deviation of 1.

CO1
CS1
OA1
IP1
AH1
RM1
WS1
RP1
EP1
JP1
LI1

Prin1
0.403014
0.446947
0.388777
0.334832
0.333357
0.192615
0.255142
0.175800
0.280114
0.075413
0.214888

Prin2
-.129644
-.197810
-.286682
-.083064
-.108640
-.113110
-.003187
0.515927
0.454817
0.436752
0.408141

Prin3
-.165482
-.119725
-.041966
-.298629
-.177557
0.591965
0.464371
-.080397
0.229088
-.410653
0.205355

Prin4
-.453338
-.313282
-.030582
0.387424
0.455676
-.137687
0.436319
0.093462
0.092214
-.039082
-.331010

3.1 PRINCIPAL COMPONENT 1
As illustrated in Table 2.2 all the eigenvector coefficients for the first principal component were
positive suggesting that all crop and woody plant types investigated covary together. Dependent
variable values ranged from 0.447 to 0.193. The first principal component could be described as
an all crop and woody plant response axis. Values >0.4 or <-0.4 are considered to have strong
association with their principal components and are the most descriptive terms of the variables
investigated. There are two such variables in the first principal component: corn and corn silage.
These two terms represent the two most significant variable of the equation. In this way, the
equation could also be described as a corn-corn silage axis.
Figure 2.1 illustrates the best suitable model developed for the first principal component.
The model was found to explain 87.93 percent of the variance. The model is not over specific,
having 21 terms in the equation and C(p)=21.54. C(p) values that are close to the number of

34

terms in the equation are considered to “fit” well (Agresti, 2012), therefore further validating the
strength of this model.

Figure 2.1 Best model found for the Stepwise Maximum R-squared Improvement for Principal
Component 1. Refer to Table 1.4 for an explanation of independent variable soil parameters.
Maximum R-Square Improvement: Step 41
R-Square = 0.8793 and C(p) = 21.5359

Source

DF

Sum of
Squares

Model
Error
Corrected Total

20
73
93

258.95863 12.94793 26.60
35.53447 0.48677
294.49310

Variable
Intercept
TP
SL
CL
HC
AW
PH
SL2
FR2
AW2
TPCL
TPBD
TPHC
TPOM
SLOM
FRBD
CLHC
CLPH
BDOM
HCPH
AWPH

Parameter
Estimate
-0.87194
0.50466
-2.07312
0.80349
-0.96635
-2.37413
0.40641
0.39799
-0.26293
0.88193
0.36989
-0.59481
-0.48820
2.06655
-1.45369
-0.76847
-0.74717
-0.42867
-0.67422
-0.76090
-1.20130

Mean
Square F Value Pr > F

Standard
Error
Type II SS
0.16672
0.14014
0.17001
0.29730
0.22510
0.30757
0.12497
0.07666
0.03837
0.13862
0.11895
0.10137
0.12425
0.22987
0.30566
0.22587
0.20693
0.15428
0.15699
0.18229
0.23201

13.31485
6.31249
72.38096
3.55540
8.97071
29.00372
5.14788
13.12078
22.85670
19.70213
4.70698
16.76003
7.51535
39.34052
11.01033
5.63458
6.34626
3.75814
8.97763
8.48116
13.05026

F Value

Pr > F

27.35
<.0001
12.97
0.0006
148.70 <.0001
7.30
0.0086
18.43
<.0001
59.58
<.0001
10.58
0.0017
26.95
<.0001
46.96
<.0001
40.47
<.0001
9.67
0.0027
34.43
<.0001
15.44
0.0002
80.82
<.0001
22.62
<.0001
11.58
0.0011
13.04
0.0006
7.72
0.0069
18.44
<.0001
17.42
<.0001
26.81
<.0001

Bounds on condition number: 22.894, 3308.2
35

<.0001

3.2 PRINCIPAL COMPONENT 2
Table 2.2 shows the eigenvector coefficients for the second principal component. The
coefficients in the second eigenvector can be organized in three separate groups: positively
associated (positive values), negatively associated (negative values) and weakly associated
(values near zero). Positive coefficients include red pine, eastern white pine, jack pine, and lilac.
All of these species also had values that were considered be strongly associated (x>0.4, x<-0.4).
White Spruce was the only plant type to have a value near 0, indicating that it is not described
well by this model. Plant types with negative values were not as strongly associated as those
plant types with positive values. Negative values ranged from -0.083 to -0.287.
Figure 2.2 illustrates the best equation derived for the second principal component. The
equation was found to explain 74.52 percent of the variance. The model is not over specific,
having 13 terms and a C(p) value of 36.78.
Figure 2.2 Best model found for the Stepwise Maximum R-squared Improvement for
Principal Component 2. Refer to Table 1.4 for an explanation of independent variable soil
parameters.
Maximum R-Square Improvement: Step 24
R-Square = 0.7452 and C(p) = 36.7759

Source

DF

Sum of
Squares

Mean
Square

Model
Error
Corrected Total

12
81
93

177.86957
60.82255
238.69212

14.82246
0.75090

Variable

Parameter
Estimate

Standard
Error

Type II SS

Intercept
SL
FR

-0.42029
0.48772
0.87531

0.14323
0.09661
0.21882

6.46564
19.13846
12.01465

36

F Value

Pr > F

19.74

<.0001

F Value

Pr > F

8.61
25.49
16.00

0.0043
<.0001
0.0001

Figure 2.2 (cont’d)
AW
AW2
TPFR
TPCL
TPBD
FRHC
FRPH
FROM
CLHC
BDHC

-1.96851
0.77296
0.68607
0.63153
-0.33830
1.92499
-0.48216
1.29944
-0.83076
0.62243

0.19208
0.11849
0.19490
0.11230
0.10738
0.27429
0.19359
0.19656
0.16952
0.12067

78.86507
31.95257
9.30484
23.74723
7.45230
36.98520
4.65801
32.81805
18.03317
19.97710

105.03
42.55
12.39
31.63
9.92
49.25
6.20
43.71
24.02
26.60

<.0001
<.0001
0.0007
<.0001
0.0023
<.0001
0.0148
<.0001
<.0001
<.0001

Bounds on condition number: 8.1439, 462.11

3.3 PRINCIPAL COMPONENT 3
Table 2.2 illustrates the eigenvector coefficients associated with the third principal
component. This eigenvector can be divided into two main groups. Out of all plant species
investigated there were only two tree species that had positive values, red maple and white
spruce. Both red maple and white spruce showed values that were considered strongly
associated (x>0.4, x<-0.4), with values of .592 and .464 respectively. All other crop and woody
plant types had negative values. This suggests that this model can be best described as a red
maple-white spruce dimension.
Figure 2.3 illustrates the first equation derived for the third principal component, in
which all p values were less than 0.04. The equation was found to explain 91.79 percent of the
variance, which was the highest of all the best-fit models developed in this study. There were 38
terms in this model and had a C(p) value of 33.95. Although some researchers consider models
with C(p) values less than the number of terms to be over specific (Burley & Thomsen,1987),
others suggest that the relativity of the values is what should be considered (Agresti, 2012).

37

Depending on the interpretation of these terms, the model could be considered reliable. Because
of possible discrepancies of interpretations, a second, more conservative model was developed.
Figure 2.3 First model found for the Stepwise Maximum R-squared Improvement for
Principal Component 3 with all P values below 0.04. The number of terms in this model
(38 terms) was greater than the C(p) value (33.95) and may be considered over specific.
Refer to Table 1.4 for an explanation of independent variable soil parameters.
Maximum R-Square Improvement: Step 78
R-Square = 0.9179 and C(p) = 33.9479

Source
Model
Error
Corrected Total

DF
37
56
93

Sum of
Squares
151.70037
13.56251
165.26288

Mean
Square
4.10001
0.24219

F Value
16.93

Variable

Parameter
Estimate

Standard
Error

Type II SS

F Value

Intercept
TP
SL
FR
AW
PH
OM
TP2
SL2
FR2
CL2
BD2
HC2
TPFR
TPBD
TPHC
TPAW
TPPH
SLFR
SLBD
SLHC
SLAW
SLPH

-0.58174
-0.63534
0.94986
-1.98688
-0.47879
-0.31093
0.83128
0.46565
-0.23547
-0.56152
-1.66751
0.84355
-2.74580
-1.99879
-1.79813
-2.12934
-1.45257
-0.64835
0.59705
0.52308
0.55404
0.55188
0.21506

0.15337
0.16313
0.12365
0.32976
0.17128
0.11289
0.21698
0.14067
0.07126
0.08572
0.15338
0.08690
0.32562
0.26111
0.19132
0.34823
0.29009
0.13853
0.13604
0.09160
0.15350
0.16719
0.08479

3.48424
3.67348
14.29269
8.79217
1.89255
1.83737
3.55463
2.65374
2.64424
10.39341
28.62412
22.82194
17.22179
14.19194
21.39216
9.05563
6.07261
5.30537
4.66457
7.89863
3.15520
2.63872
1.55815

14.39
15.17
59.01
36.30
7.81
7.59
14.68
10.96
10.92
42.91
118.19
94.23
71.11
58.60
88.33
37.39
25.07
21.91
19.26
32.61
13.03
10.90
6.43

38

Pr > F
<.0001

Pr > F
0.0004
0.0003
<.0001
<.0001
0.0071
0.0079
0.0003
0.0016
0.0017
<.0001
<.0001
<.0001
<.0001
<.0001
<.0001
<.0001
<.0001
<.0001
<.0001
<.0001
0.0007
0.0017
0.0140

Figure 2.3 (cont’d)
FRCL
FRBD
FRHC
FRPH
FROM
CLHC
CLAW
CLPH
BDPH
BDOM
HCPH
HCOM
AWPH
AWOM
PHOM

-7.30507
-0.47675
-4.55352
0.53054
2.86700
-5.84236
-1.16284
0.72328
-0.52386
-0.66791
0.90940
1.58268
-0.48875
1.20690
-0.34027

0.57101
0.18432
0.61478
0.17869
0.53390
0.58049
0.38663
0.21485
0.14214
0.19598
0.19840
0.39057
0.23002
0.32471
0.10984

39.63747
1.62033
13.28637
2.13505
6.98382
24.53200
2.19076
2.74472
3.28952
2.81284
5.08856
3.97694
1.09348
3.34585
2.32418

163.66
6.69
54.86
8.82
28.84
101.29
9.05
11.33
13.58
11.61
21.01
16.42
4.52
13.82
9.60

<.0001
0.0123
<.0001
0.0044
<.0001
<.0001
0.0039
0.0014
0.0005
0.0012
<.0001
0.0002
0.0380
0.0005
0.0030

Bounds on condition number: 152.86, 44747

Figure 2.4 illustrates a second model for the third principal component. This equation
was found to explain 87.68 percent of the variance, which is comparable to the r-square value of
the first principal component. In contrast to the model shown in Figure 2.3, the number of terms
in this model shown in Figure 3.4 (38 terms) is less than the C(p) value (48.00). This equation
would not be considered over specific according to criteria described by Burley (1988), and is
therefore a more conservative model than that shown in Table R.5.
Figure 2.4 Second model found for the Stepwise Maximum R-squared Improvement for
Principal Component 3 with all P values below 0.04. The number of terms in this model
(32 terms) was greater than the C(p) value (48.00) and is not considered over specific.
Refer to Table 1.4 for an explanation of independent variable soil parameters.
Maximum R-Square Improvement: Step 60
Variable PH Removed: R-Square = 0.8768 and C(p) = 48.0039

39

Figure 2.4 (cont’d)

Source

DF

Sum of
Squares

Model
Error
Corrected Total

31
62
93

144.89768
20.36520
165.26288

Mean
Square
4.67412
0.32847

F Value
14.23

Pr > F
<.0001

Variable

Parameter
Estimate

Standard
Error
Type II SS F Value Pr > F

Intercept
TP
SL
FR
AW
TP2
SL2
CL2
BD2
HC2
AW2
PH2
TPFR
TPBD
TPHC
TPAW
TPPH
SLFR
SLBD
SLHC
SLAW
SLPH
FRCL
FRHC
FRAW
FRPH
FROM
CLHC
CLAW
BDOM
HCAW
HCPH

-0.64768
-0.56620
0.85356
-1.42155
-0.53829
0.67112
-0.24282
-1.51835
0.66391
-1.33439
0.57716
-0.23985
-1.90574
-1.63418
-2.13468
-1.53045
-0.80303
0.53827
0.47082
0.54800
0.61201
0.28039
-8.01308
-1.46864
2.05719
1.26966
0.63147
-5.90892
-1.46907
-1.13415
1.50420
1.08156

0.15896
0.16100
0.13144
0.26872
0.16094
0.13881
0.08060
0.14372
0.08253
0.32871
0.21442
0.09895
0.25036
0.19261
0.31306
0.26143
0.13070
0.14504
0.10204
0.16483
0.18416
0.09621
0.71664
0.37425
0.36236
0.17728
0.22656
0.58555
0.38887
0.19548
0.52412
0.12385

5.45334
4.06219
13.85171
9.19248
3.67441
7.67752
2.98123
36.66066
21.25649
5.41298
2.37983
1.93000
19.03170
23.64559
15.27275
11.25712
12.39936
4.52414
6.99323
3.63047
3.62761
2.78965
41.06728
5.05827
10.58654
16.84806
2.55187
33.44890
4.68798
11.05651
2.70549
25.05197

16.60
12.37
42.17
27.99
11.19
23.37
9.08
111.61
64.71
16.48
7.25
5.88
57.94
71.99
46.50
34.27
37.75
13.77
21.29
11.05
11.04
8.49
125.03
15.40
32.23
51.29
7.77
101.83
14.27
33.66
8.24
76.27

Bounds on condition number: 89.777, 18983
40

0.0001
0.0008
<.0001
<.0001
0.0014
<.0001
0.0037
<.0001
<.0001
0.0001
0.0091
0.0183
<.0001
<.0001
<.0001
<.0001
<.0001
0.0004
<.0001
0.0015
0.0015
0.0050
<.0001
0.0002
<.0001
<.0001
0.0070
<.0001
0.0004
<.0001
0.0056
<.0001

3.4 PRINCIPAL COMPONENT 4
Although the eigenvalue for the fourth principal component, as shown in Table 2.1 was
larger than 1.0 (1.2), the best model from the stepwise regression analysis was not significant
enough to be considered a relevant model. The best model found had a R-squared value of
0.3133. Because the model explained such a small portion (31.33 percent) of the variation in the
crop and woody plant axis, it was not considered for further analysis. No equation was produced
for this model.
Figure 2.5 Best model found for the Stepwise Maximum R-squared Improvement for
Principal Component 4. Refer to Table 1.4 for an explanation of independent variable soil
parameters.
Maximum R-Square Improvement: Step 4
R-Square = 0.3133 and C(p) = 105.8427

Source

DF

Sum of
Squares

Model
Error
Corrected Total

4
89
93

35.03746
76.80084
111.83830

Variable
Intercept
TP
BD
BD2
AW2

Parameter
Estimate
-0.42257
0.26187
-0.31102
0.18310
0.24152

Mean
Square
8.75936
0.86293

F Value Pr > F
10.15

Standard
Error
Type II SS
0.14536
7.29247
0.09767
6.20256
0.10184
8.04777
0.07441
5.22483
0.08266
7.36694

F Value
8.45
7.19
9.33
6.05
8.54

Bounds on condition number: 1.1179, 17.144

41

<.0001

Pr > F
0.0046
0.0087
0.0030
0.0158
0.0044

DISCUSSION
There were a total of 5 equations developed for the three principal component

investigated. Figures 3.1, 3.2, 3.3, 3.4, and 3.6 display the equations developed for all

significant principal components identified from the data of Iron and Dickinson Counties.
All equations developed are highly specific with p-values less than 0.0001, and are

considered statistically significant. As previously discussed, the significant eigenvector

values indicate plant species that each equation best describes and can be used to further
investigate similarities among dependent variables to categorize the equations.

Revegetation efforts in reclamation projects often involve an attempt to recreate a

biological community that previously existed on the site or to blend the reclaimed

landscape into adjacent lands. Therefore, it is beneficial to investigate possible way to
categorize developed equations to better describe the communities in which they are

intended to predict. Similar studies have categorized equations based on the plant species
the derived equations best describe. In a study of a three county North Dakotan coal

mining region, Burley (1995) investigated the potential to find a predictive equation to
describe the lowlands and the transitional zones that composed the study area. By

referring equations directly to a landscape type or post land use, reclamation specialist can
more effectively use such equations to aid in reclamation projects. Furthermore,

categorization is dependent upon the variable analyzed. For instance, Burley (1995a)

investigated biophysical variables to describe vegetative productivity across a spatial
region and found that the equations developed in the study could be categorized as a

geomorphology equation, a climatology equation, and a biological equation. Based on the
42

variables investigated in this study, the most appropriate method of categorization would
be to describe equations as community types.

John T. Curtis, a plant ecologist and botanist from the University of Wisconsin

conducted an in-depth inventory of the Wisconsin’s northern floristic zone (Curtis, 1959).
This series of studies gives a comprehensive analysis of a neighboring region that shows
similar geological and climatic (Linsemier, 1989; Linsemier, 1997) characteristics. Both
the study area of the Curtis studies and the study area of this investigation have been

classified by the International Union for Conservation and Nature and Natural Resources
(IUCN) (Udvardy, 1975) as being in the Great Lakes biogeographical province and by the
United States Department of Agriculture, Forest Service (McNab et al., 2005) as being

within the Laurentian Mixed Forest Province. Due to the strong, established similarities
between these study areas the data reported by Curtis is considered comparable and
applicable to the study area of this study.

4.1 PRINCIPAL COMPONENT 1: Mesic Model

Figure 3.1 presents the equation developed from the stepwise maximum r-squared

improvement for the first principal component. All plant species investigated in the first
principal component were found to covary together indicating that the model can be

described as an all crop and woody plant model. An analysis of eigenvector values implied
that the model could also be described as a corn-corn silage model. However, it may be

more beneficial to describe this equation according to its community type so as to connect
the model to a post-mine community rather than a particular set of plants. The plants

selected for this study are all adapted to growth in moderate soils that are typical in this
43

region. Moderate soils in the northern Upper Peninsula are often sandy to loamy, welldrained soils that are slightly alkaline to slightly acidic. These soils are well suited to
support a broad range of plant types and are typical of northern mesic forests. It is

important to note the distinction of these soils abilities to support growth and the actual
plant growth patterns that occur in natural conditions.

Figure 3.1 Best equation found for the Stepwise Maximum R-squared Improvement for
Principal Component 1. This equation is best described as a mesic model.
Plant = -0.872 + [((TP-3.121)*1.267-1)*(0.505)]
+ [((SL-8.00)*7.765-1)*(-2.073)]
+ [((CL-8.915)*4.138-1)*(0.803)]
+ [((HC-5.227)*4.302-1)*(-0.966)]
+ [((AW-0.122)*0.040-1)*(-2.374)]
+ [((PH-5.971)*0.671-1)*(0.406)]
+ [((SL-8.00)*7.765-1)* ((SL-8.00)*7.765-1)*(0.398)]
+ [((FR-4.571)*6.760-1)* ((FR-4.571)*6.760-1)*(-0.263)]
+ [((AW-0.122)*0.040-1)* ((AW-0.122)*0.040-1)*(0.882)]
+ [((TP-3.121)*1.267-1)* ((CL-8.915)*4.138-1)*(0.370)]
+ [((TP-3.121)*1.267-1)* ((BD-1.506)*0.084-1)*(-0.595)]
+ [((TP-3.121)*1.267-1)*((HC-5.227)*4.302-1)*(-0.488)]
+ [((TP-3.121)*1.267-1)*((OM-1.995)*1.201-1)*(2.067)]
+ [((SL-8.00)*7.765-1)* ((OM-1.995)*1.201-1)*(-1.454)]
+ [((FR-4.571)*6.760-1)*((BD-1.506)*0.084-1)*(-0.768)]
+ [((CL-8.915)*4.138-1)* ((HC-5.227)*4.302-1)*(-0.747)]
+ [((CL-8.915)*4.138-1)* ((PH-5.971)*0.671-1)*(-0.429)]
+ [((BD-1.506)*0.084-1)*((OM-1.995)*1.201-1)*(-0.674)]
+ [((HC-5.227)*4.302-1)* ((PH-5.971)*0.671-1)*(-0.761)]
+ [((AW-0.122)*0.040-1)*((PH-5.971)*0.671-1)*(-1.201)]

This points to an inherent concern when comparing the finds of Curtis’s and similar

naturalized community observation studies to the findings of this research. The study

conducted by Curtis was an inventory of tree stands in natural settings with the influence
of interspecies competition, while the soils surveys this research used to develop its data
44

set measured trees grow under more controlled circumstances that limited this interaction.

This could account for discrepancies between findings of these studies. Despite the manner
in which data sets were derived, the Curtis study gives useful insight into the tree species

that exist in a natural community setting, which revegetation reclamation efforts are most
likely going to attempt to replicate.

All species investigated in this study have unique ranges of soil conditions in which

they are able to grow. For instance, white spruce is usually associated with more wet soil
conditions, whereas red pine is most commonly associated with dry sites. However, the
range of both species as well as all others investigated reaches into more moderate

conditions. This common range of conditions that satisfies the growth needs of species

would imply that all species would covary in one axis, as was found for the first principal
component. The first model can therefore best be described as a mesic model.
4.2 PRINCIPAL COMPONENT 2 EQUATION: Northern Dry Forest Model

Figure 3.2 presents the equation developed from the stepwise maximum r-squared

improvement for the second principal component. The eigenvector analysis of this

principal component (refer to Table 2.2) found that four woody species were found to be

strongly associated with this axis; jack pine, red pine, eastern white pine, and lilac. Curtis
identifies five dominate species for the northern dry forest, the three most dominant of
which, in order of importance, are jack pine (Pinus banksiana Lamb.), red pine (Pinus

resinosa Aniton), and white pine (Pinus strobus L.). Additionally, Curtis identified red pine
as one of two “species which attain optimum importance” (Curtis, 1959 pg. 537) in this

community. This finding coincides the findings of the second eigenvector values, which
45

shows that red pine is the most strongly associated coefficient of the axis with a value of
0.516 (refer to Table 2.2).

Figure 3.2 Best equation found for the Stepwise Maximum R-squared Improvement for
Principal Component 2. This equation is best described as a northern dry forest model.
PLANT = -0.420 + [((SL-8.000)*7.765-1)*(0.488)]
+ [((FR-4.571)*6.760-1)*(0.875)]
+ [((AW-0.122)*0.040-1)*(-1.979)]
+ [((AW-0.122)*0.040-1)* ((AW-0.122)*0.040-1)*(0.773)]
+ [((TP-3.121)*1.267-1)*((FR-4.571)*6.760-1)*(0.686)]
+ [((TP-3.121)*1.267-1)*((CL-8.915)*4.138-1)*(0.632)]
+ [((TP-3.121)*1.267-1)*((BD-1.506)*0.084-1)*(-0.338)]
+ [((FR-4.571)*6.760-1)*((HC-5.227)*4.302-1)*(1.925)]
+ [((FR-4.571)*6.760-1)*((PH-5.971)*0.671-1)*(-0.482)]
+ [((FR-4.571)*6.760-1)*((OM-1.995)*1.201-1)*(1.299)]
+ [((CL-8.915)*4.138-1)*((HC-5.227)*4.302-1)*(-0.831)]
+ [((BD-1.506)*0.084-1)*((HC-5.227)*4.302-1)*(0.622)]

It is important to recognize that although jack pine was considered the most

important tree identified by Curtis, red pine was identified as reaching its maximum

growth potential in this community. This is most like accounted for by the influence of

interspecies competition, as previously discussed. Due to the strong connection between
the findings of Curtis and the eigenvector results of this study, the second principal
component equation would best be described as a northern dry forest equation.
4.3 Principal Component 3 Equation: Wet Mesic Model

Figures 3.3 and 3.4 show the equations developed for the third principal

component. Two separate equations were derived due to possible divergences in

statistical interpretations, as previously discussed. Figure 3.4 illustrates the second
equation developed from the stepwise maximum r-squared improvement, which is
46

considered to be more conservative than the equation shown in Figure 3.3. The equations
explain 91.79% and 87.68% of the variance. Considering the small amount of explanation
of the variance that is lost by selecting the second equation and its more conservative
nature, the second equation is considered to be the most reliable model for the third
principal component.

Figure 3.3 First equation found for the Stepwise Maximum R-squared Improvement for
Principal Component 3. Comparison of number of terms in the equation and C(p) value
indicates that this equation may be over specific.
PLANT = -0.582 + [((TP-3.121)*1.267-1)*(-0.635)]
+ [((SL-8.000)*7.765-1)*(0.950)]
+ [((FR-4.571)*6.760-1)*(-1.987)]
+ [((AW-0.122)*0.040-1)*(-0.479)]
+ [((PH-5.971)*0.671-1)*(-0.311)]
+ [((OM-1.995)*1.201-1)*(0.831)]
+ [((TP-3.121)*1.267-1)*((TP-3.121)*1.267-1)*(0.466)]
+ [((SL-8.000)*7.765-1)*((SL-8.000)*7.765-1)*(-0.235)]
+ [((FR-4.571)*6.760-1)*((FR-4.571)*6.760-1)*(-0.562)]
+ [((CL-8.915)*4.138-1)*((CL-8.915)*4.138-1)*(-1.668)]
+ [((BD-1.506)*0.084-1)*((BD-1.506)*0.084-1)*(0.844)]
+ [((HC-5.227)*4.302-1)*((HC-5.227)*4.302-1)*(-2.746)]
+ [((TP-3.121)*1.267-1)* ((FR-4.571)*6.760-1)*(-1.999)]
+ [((TP-3.121)*1.267-1)*((BD-1.506)*0.084-1)*(-1.798)]
+ [((TP-3.121)*1.267-1)*((HC-5.227)*4.302-1)*(-2.129)]
+ [((TP-3.121)*1.267-1)*((AW-0.122)*0.040-1)*(-1.453)]
+ [((TP-3.121)*1.267-1)*((PH-5.971)*0.671-1)*(-0.648)]
+ [((SL-8.000)*7.765-1)*((FR-4.571)*6.760-1)*(0.597)]
+ [((SL-8.000)*7.765-1)*((BD-1.506)*0.084-1)*(0.523)]
+ [((SL-8.000)*7.765-1)*((HC-5.227)*4.302-1)*(0.554)]
+ [((SL-8.000)*7.765-1)*((AW-0.122)*0.040-1)*(0.552)]
+ [((SL-8.000)*7.765-1)*((PH-5.971)*0.671-1)*(0.215)]
+ [((FR-4.571)*6.760-1)*((CL-8.915)*4.138-1)*(-7.305)]
+ [((FR-4.571)*6.760-1)*((BD-1.506)*0.084-1)*(-0.477)]
+ [((FR-4.571)*6.760-1)*((HC-5.227)*4.302-1)*(-4.554)]
+ [((FR-4.571)*6.760-1)*((PH-5.971)*0.671-1)*(0.531)]
+ [((FR-4.571)*6.760-1)*((OM-1.995)*1.201-1)*(2.867)]
+ [((CL-8.915)*4.138-1)*((HC-5.227)*4.302-1)*(-5.842)]
47

Figure 3.3 (cont’d)

+ [((CL-8.915)*4.138-1)*((AW-0.122)*0.040-1)*(-1.163)]
+ [((CL-8.915)*4.138-1)*((PH-5.971)*0.671-1)*(0.723)]
+ [((BD-1.506)*0.084-1)*((PH-5.971)*0.671-1)*(-0.524)]
+ [((BD-1.506)*0.084-1)*((OM-1.995)*1.201-1)*(-0.668)]
+ [((HC-5.227)*4.302-1)*((PH-5.971)*0.671-1)*(0.909)]
+ [((HC-5.227)*4.302-1)*((OM-1.995)*1.201-1)*(1.583)]
+ [((AW-0.122)*0.040-1)*((PH-5.971)*0.671-1)*(-0.489)]
+ [((AW-0.122)*0.040-1)*((OM-1.995)*1.201-1)*(1.207)]
+ [((PH-5.971)*0.671-1)*((OM-1.995)*1.201-1)*(-0.340)]

Figure 3.4 Second equation found for the Stepwise Maximum R-squared Improvement for
Principal Component 3. Comparison of number of terms in the equation and C(p) value
indicates that this equation is most likely not over specific. This is considered the more
conservative equation for this principal component when compared to the equation in
Figure 3.3. Best described as a wet mesic model.
PLANT = -0.648 + [((TP-3.121)*1.267-1)*(-0.566)]
+ [((SL-8.000)*7.765-1)*(0.854)]
+ [((FR-4.571)*6.760-1)*(-1.422)]
+ [((AW-0.122)*0.040-1)*(-0.538)]
+ [((TP-3.121)*1.267-1)*((TP-3.121)*1.267-1)*(0.671)]
+ [((SL-8.000)*7.765-1)*((SL-8.000)*7.765-1)*(-0.243)]
+ [((CL-8.915)*4.138-1)*((CL-8.915)*4.138-1)*(-1.518)]
+ [((BD-1.506)*0.084-1)*((BD-1.506)*0.084-1)*(0.664)]
+ [((HC-5.227)*4.302-1)*((HC-5.227)*4.302-1)*(-1.334)]
+ [((AW-0.122)*0.040-1)*((AW-0.122)*0.040-1)*(0.577)]
+ [((PH-5.971)*0.671-1)*((PH-5.971)*0.671-1)*(-0.240)]
+ [((TP-3.121)*1.267-1)*((FR-4.571)*6.760-1)*(-1.906)]
+ [((TP-3.121)*1.267-1)*((BD-1.506)*0.084-1)*(-1.634)]
+ [((TP-3.121)*1.267-1)*((HC-5.227)*4.302-1)*(-2.135)]
+ [((TP-3.121)*1.267-1)*((AW-0.122)*0.040-1)*(-1.530)]
+ [((TP-3.121)*1.267-1)*((PH-5.971)*0.671-1)*(-0.803)]
+ [((SL-8.000)*7.765-1)*((FR-4.571)*6.760-1)*(0.538)]
+ [((SL-8.000)*7.765-1)*((BD-1.506)*0.084-1)*(0.471)]
+ [((SL-8.000)*7.765-1)*((HC-5.227)*4.302-1)*(0.548)]
+ [((SL-8.000)*7.765-1)*((AW-0.122)*0.040-1)*(0.612)]
+ [((SL-8.000)*7.765-1)*((PH-5.971)*0.671-1)*(0.280)]
+ [((FR-4.571)*6.760-1)*((CL-8.915)*4.138-1)*(-8.013)]
+ [((FR-4.571)*6.760-1)*((HC-5.227)*4.302-1)*(-1.469)]
+ [((FR-4.571)*6.760-1)*((AW-0.122)*0.040-1)*(2.057)]
+ [((FR-4.571)*6.760-1)*((PH-5.971)*0.671-1)*(1.270)]
+ [((FR-4.571)*6.760-1)*((OM-1.995)*1.201-1)*(0.631)]
+ [((CL-8.915)*4.138-1)*((HC-5.227)*4.302-1)*(-5.909)]
+ [((CL-8.915)*4.138-1)*((AW-0.122)*0.040-1)*(-1.469)]
48

Figure 3.4 (cont’d)

+ [((BD-1.506)*0.084-1)*((OM-1.995)*1.201-1)*(-1.134)]
+ [((HC-5.227)*4.302-1)*((AW-0.122)*0.040-1)*(1.504)]
+ [((HC-5.227)*4.302-1)*((PH-5.971)*0.671-1)*(1.082)]

The eigenvector coefficients that were strongly associated with the third principal

component were red maple and white spruce (refer to Table 2.2). Both of these species are
considered to be well suited for growth in more lowland conditions or transitional zones
between wet lowlands and uplands. Curtis identified black spruce (Picea mariana [Mill.]

Britton, Sterns & Poggenb.) as the most dominant species in northern lowland forests. The
study showed that although white spruce (Picea glauca [Moench] Voss) was found in some
stands, numbers of observed trees were limited. Both species prefer comparable soils and
have similar North American distributions that cover portions of the northern United

States and much of Canada. These two species commonly occur together in more northern

limits of their distributions, white spruce is often out competed by black spruce in southern
range limits (Rook, 2002), which could account for the low occurrence of the species in the
areas surveyed. Although black spruce growth rates were not included in the available
data provided by the NRCS, due to the commonalities between white spruce and black

spruce these species have been considered comparable species for analytical purposes of
this model.

White spruce is known to grow poorly in soils with high water tables yet is not

considered a true wet lowland plant but rather is most commonly found in transitional

zones between swampy lowlands and uplands or in alluvial zones. Similarly, red maple

(Acer rubrum L.) is commonly associated with a wide variety of wet sites and transitional
49

zones. The strong associations of these species to the third principal component would
suggest that this axis could be best described as a wet mesic model or an alluvial zone
model.

Figure 3.5 “Behavior of major tree species on the combined ordination of northern upland
and lowland forests. 1, tamarack (Larix laricina); 2, black spruce (Picea mariana); 3, white
cedar (Thuja ocidentalis); 4, hemlock (Tsuga canadensis); 5, sugar maple (Acer saccharum);
6, white pine (Pinus strobus); 7, red pine (Pinus resinosa); 8, jack pine (Pinus banksina)”
(Curtis, 1959 pg. 180).
150

120

Importance Value

90

60

30

2

3

4

1

W

WM

5

6

DM
M
Compositional Index

8

7

D

One concern with naming the third principal component model the wet mesic model

is that Curtis found that jack pine (Pinus banksiana Lamb.) to show a spike in occurrence in
the wet mesic range as illustrated in Figure 3.5. However, the eigenvalues of this axis

indicate that jack pine was strongly negatively associated with the model with a value of -

0.411 (refer to Table 2.2). This strong negative association could cause some explanatory
concern for this model. However, because jack pine results showed a strong positive
50

association with the second principal component and a strong negative association with
the third principal component neither axis is considered to be truly descriptive of the

species. It is therefore necessary to develop an additional model that combines these two
axes to best predict the productivity potential of jack pine.
4.4 JACK PINE MODEL
An equation was developed specifically for jack pine due to the findings of the principal
component analysis. As illustrated in Table 2.2, the eigenvector values for jack pine in both the
second and the third principal components were either greater than 0.4 or less than -0.4 showing
a that this species is strongly associated with both axes. Because there are two models that could
be used to predict the vegetative productivity of jack pine, neither model alone is best suited as a
predictive model for this species. Therefore it is necessary to develop an additional model that is
most reliable to predict the vegetative productivity of jack pine.

Principal Component 2

Figure 3.6 Plot of hypothetical values to illustrate potential findings of principal
component 2 and principal component 3 as separate axes and a combined axis that shows
the most reliable jack pine axis.
6

Highly productive values
for principal component 2
equation

5
4

Highly productive values
for principal component 3
equation

3
2
1
0

0

1

2

3

4

Principal Component 3
51

5

6

Highly productive values
for the combined principal
2 and principal 3 equation
Low productivity values
for principal 2 or principal
3 equations

Figure 3.6 illustrates a plotting of hypothetical findings of equations that could be derived
from the second and the third principal components. The values plotted are not derived from any
developed equation, but rather are used to illustrate that combining the principal component 2
axis with the principal component 3 axis will result in a new axis that will most accurately
describe the productivity of jack pine. By using only one model, the reliability of the results are
not as strong, as shown by plot points near the x or y axes. By combining the two equations the
predictability of the model is not split between two axes, but rather combined into one that
specifically describes one plant, jack pine. Results from such a model would show trends more
similar to those of the combined principal 2 and principal 3 equations plot points. As this
illustration suggests, the most accurate equation to predict the productivity potential of jack pine
would be derived by combining the equations developed from principal component 2 and
principal component 3.
Figure 3.7 shows the best equation that describes jack pine. The equation is considered
to be highly specific and explains 65.33% of the variance. This model was developed by
multiplying the equation from each dimension that was strongly associated with jack pine. It
should noted that the second equation, that of principal component 3, was multiplied by a factor
of -1 to reverse the association with this species. This equation should be used to most
accurately predict the vegetative productivity of jack pine.

52

Figure 3.7 Jack pine model derived from multiplying the second principal component
model by the most reliable model for the third principal component.
PLANT = {-0.420 + [((SL-8.000)*7.765-1)*(0.488)]
+ [((FR-4.571)*6.760-1)*(0.875)]
+ [((AW-0.122)*0.040-1)*(-1.979)]
+ [((AW-0.122)*0.040-1)* ((AW-0.122)*0.040-1)*(0.773)]
+ [((TP-3.121)*1.267-1)*((FR-4.571)*6.760-1)*(0.686)]
+ [((TP-3.121)*1.267-1)*((CL-8.915)*4.138-1)*(0.632)]
+ [((TP-3.121)*1.267-1)*((BD-1.506)*0.084-1)*(-0.338)]
+ [((FR-4.571)*6.760-1)*((HC-5.227)*4.302-1)*(1.925)]
+ [((FR-4.571)*6.760-1)*((PH-5.971)*0.671-1)*(-0.482)]
+ [((FR-4.571)*6.760-1)*((OM-1.995)*1.201-1)*(1.299)]
+ [((CL-8.915)*4.138-1)*((HC-5.227)*4.302-1)*(-0.831)]
+ [((BD-1.506)*0.084-1)*((HC-5.227)*4.302-1)*(0.622)]}
* {[-0.648 + [((TP-3.121)*1.267-1)*(-0.566)]
+ [((SL-8.000)*7.765-1)*(0.854)]
+ [((FR-4.571)*6.760-1)*(-1.422)]
+ [((AW-0.122)*0.040-1)*(0.538)]
+ [((TP-3.121)*1.267-1)*((TP-3.121)*1.267-1)*(0.671)]
+ [((SL-8.000)*7.765-1)*((SL-8.000)*7.765-1)*(-0.243)]
+ [((CL-8.915)*4.138-1)*((CL-8.915)*4.138-1)*(-1.518)]
+ [((BD-1.506)*0.084-1)*((BD-1.506)*0.084-1)*(0.664)]
+ [((HC-5.227)*4.302-1)*((HC-5.227)*4.302-1)*(-1.334)]
+ [((AW-0.122)*0.040-1)*((AW-0.122)*0.040-1)*(0.577)]
+ [((PH-5.971)*0.671-1)*((PH-5.971)*0.671-1)*(-0.240)]
+ [((TP-3.121)*1.267-1)*((FR-4.571)*6.760-1)*(-1.906)]
+ [((TP-3.121)*1.267-1)*((BD-1.506)*0.084-1)*(-1.634)]
+ [((TP-3.121)*1.267-1)*((HC-5.227)*4.302-1)*(-2.135)]
+ [((TP-3.121)*1.267-1)*((AW-0.122)*0.040-1)*(-1.530)]
+ [((TP-3.121)*1.267-1)*((PH-5.971)*0.671-1)*(-0.803)]
+ [((SL-8.000)*7.765-1)*((FR-4.571)*6.760-1)*(0.538)]
+ [((SL-8.000)*7.765-1)*((BD-1.506)*0.084-1)*(0.471)]
+ [((SL-8.000)*7.765-1)*((HC-5.227)*4.302-1)*(0.548)]
+ [((SL-8.000)*7.765-1)*((AW-0.122)*0.040-1)*(0.612)]
+ [((SL-8.000)*7.765-1)*((PH-5.971)*0.671-1)*(0.280)]
+ [((FR-4.571)*6.760-1)*((CL-8.915)*4.138-1)*(-8.013)]
+ [((FR-4.571)*6.760-1)*((HC-5.227)*4.302-1)*(-1.469)]
+ [((FR-4.571)*6.760-1)*((AW-0.122)*0.040-1)*(2.057)]
+ [((FR-4.571)*6.760-1)*((PH-5.971)*0.671-1)*(1.270)]
+ [((FR-4.571)*6.760-1)*((OM-1.995)*1.201-1)*(0.631)]
+ [((CL-8.915)*4.138-1)*((HC-5.227)*4.302-1)*(-5.909)]
+ [((CL-8.915)*4.138-1)*((AW-0.122)*0.040-1)*(-1.469)]
+ [((BD-1.506)*0.084-1)*((OM-1.995)*1.201-1)*(-1.134)]
+ [((HC-5.227)*4.302-1)*((AW-0.122)*0.040-1)*(1.504)]
+ [((HC-5.227)*4.302-1)*((PH-5.971)*0.671-1)*(1.082)] * -1}
53

4.5 INTERPRETATION OF MODELS

Interpretation of models developed using this methodology is often difficult. It may

be most beneficial to select one equation for further analysis to demonstrate the methods
of interpreting such equations. The equation developed for the second principal

component has been selected as the example for equation interpretation due to its relative
simplicity in comparison to other equations in this study (refer to Figure 3.8). As

previously mentioned, the equation is best described as a northern dry forest or xeric

model. The woody plant species that are strongly, positively associated with this equation

grow well in well drained, sandy soils that are dry or seasonally dry, which is consistent
with the findings of the stepwise regression analysis.

Figure 3.8 Best equation developed from the Stepwise Maximum R-squared Improvement
of Principal Component 2. This equation has been selected as an example to illustrate
methods of interpreting vegetative productivity equations.
PLANT = -0.420 + [((SL-8.000)*7.765-1)*(0.488)]
+ [((FR-4.571)*6.760-1)*(0.875)]
+ [((AW-0.122)*0.040-1)*(-1.979)]
+ [((AW-0.122)*0.040-1)* ((AW-0.122)*0.040-1)*(0.773)]
+ [((TP-3.121)*1.267-1)*((FR-4.571)*6.760-1)*(0.686)]
+ [((TP-3.121)*1.267-1)*((CL-8.915)*4.138-1)*(0.632)]
+ [((TP-3.121)*1.267-1)*((BD-1.506)*0.084-1)*(-0.338)]
+ [((FR-4.571)*6.760-1)*((HC-5.227)*4.302-1)*(1.925)]
+ [((FR-4.571)*6.760-1)*((PH-5.971)*0.671-1)*(-0.482)]
+ [((FR-4.571)*6.760-1)*((OM-1.995)*1.201-1)*(1.299)]
+ [((CL-8.915)*4.138-1)*((HC-5.227)*4.302-1)*(-0.831)]
+ [((BD-1.506)*0.084-1)*((HC-5.227)*4.302-1)*(0.622)]

The equation is comprised of a combination of main-effect terms, squared terms,

and interaction terms. Main-effect terms are those associated with a single soil parameter.
Main-effect terms included in the second principal component equation are percent slope,

percent rock fragment, and available water content. The coefficient for the first main-effect
54

term, percent slope, is positive meaning that an increase in slope will add to the

productivity of the described species. Similarly, percent rock fragment also increases the

productivity of these plants. As expected, available water is negatively associated with the
model meaning that the plants of the xeric model grow well in dry soils. However, this
term is complicated by its interaction with the available water squared term.

Squared terms describe variables that increase or decrease productivity to a limit

then reverse their trends, forming parabolic trend lines. Available water content multiplied
by available water content is the only squared term in this equation. This term shows that

available water content increases productivity to a certain point then begins to decrease its
positive influence on growth after its peak. This implies that the species described the
equation can survive in conditions with low availability of water, but growth rates are

reduced when soils become overly saturated or extremely dry. Although the availability of
water soil parameter is only represented in this equation by a main-effect and a square
term, many other parameters found in other equations of this study are further
complicated by their relations with interaction terms.

The interaction terms are often much more difficult to draw intuitive conclusions

from when compared to main-effect and squared terms. Interaction terms are those terms
that multiply two different soil parameters together. For instance, the term that multiples

topographic position with percent clay is positively associated with the model. This means
that these plants grow well in steeply sloped soils with high clay content. This interaction

could possibly be interpreted as; soils that are steeply sloped may be less likely to erode if
they have strongly cemented clay-bond soils. However, no research has been found that

suggests such finds are related to productivity of the described plant species. This term is
55

further complicated by other terms that are associated with topographic position, such as
those that combine topographic positions with percent rock fragments and bulk density.

When analyzing the terms individually, all associations may not be intuitive. It is therefore
necessary to analyze all terms concurrently to fully comprehend the meaning of these
interactions. However, analysis of all terms can be very complicated, especially in
equations with many terms.

Some terms in the resulting equation may be considered tempering terms, there to

set limits upon other interactions within the equation. This could possibly be true for the
two terms; topographic position multiplied by percent clay, which show a positive

association, and topographic position multiplied by bulk density, which is negatively
associated with the model. It is difficult to draw definitive conclusions from such

interactions because one may think that soils with high percent clay would have high bulk

densities, yet the opposing trends of these terms contradict this assumption. In many cases

such contradictions cannot be intuitively explained as the terms are present in the equation

to set limits on one another.

An alternative rationalization for this specific contradiction may possibly be

explained by the presence of bedrock under shallow soils. Bulk density numbers can be

drastically increased in cases of shallow soils, in which the bulk density bedrock, having a

relatively high bulk density compared to soils, is measured within the first 48 inches of soil.
These high bulk density values could drastically change the overall bulk density value

reported for a soil and thus lower the resulting productivity score. This explanation would
also allow for the possibility for soils with high clay content to be along ridgelines.
56

Regardless of such speculation, all soil parameter interactions must be field tested

before any definitive conclusions are drawn from such equations. As illustrated in this
discussion it is difficult to decipher the actual edaphic interactions, as they relate to

vegetative productivity, that these equations describe. Complications associated with

analyzing such equations are exacerbated as the number of terms is increased and the
number of times a given soil parameter presents itself within the equation. Despite

difficulties, it is important for researches to review and analyze such equations to identify
issues that may be presented in developed predictive models.
4.6 LIMITATIONS OF STUDY
Although the models investigated in this study all explained significant amounts of the
variance, there are some factors that were not considered that could have strengthened the study.
Firstly, there is a general lack of consideration for post-mined nutrient conditions. Nutrients are
often leached from spoil piles greatly limiting the abilities of soils to support the growth of a
wide variety of vascular species, including the crop and woody plant species investigated in this
study. The loss of soil nutrients from disturbed soils makes it difficult to compare soils
measured in soil surveys that have an established A horizon and functioning nutrient cycles.
Reestablishment of nutrient contents throughout soil profiles that are similar to
concentrations found in undisturbed soils is often a lengthy process. Research by Ottenhof and
other (2007) suggests that soil organic matter as it relates to N, P, S, K, Ca, Na, and Mg
concentrations, is largely dependent upon the species grown. In this study, the reestablishment
of soil organic content in post-mined soils to pre-mined conditions ranged from estimates of 30
to 120 years. Acton and others (2011) identified soils organic carbon content as an indicator of
57

soil health, however carbon concentrations were found to be greatly limited to the first 10 cm of
soil on mined soils of up to 14 years since reclamation. Berg (1975) identified nitrogen as the
most limiting factor on mine tailings in Colorado. Often disturbed soils are planted with plants
associated with nitrogen fixation such as member of the legume family including clover,
soybean, or alfalfa (Berg, 1975) to initiate the nitrogen cycle in deficient soils. However,
establishment of a sustainable nitrogen system can take years. Although bringing soil nutrient
levels equal to or greater than pre-mined conditions throughout soil profiles may take a
considerable amount of time, bringing these concentrations to adequate levels to foster
vegetative growth may take much less time.
To initiate the processes post-mined soils are often planted with cover crops to avoid
nutrient loss, prevent erosion of topsoil, and help to initiate sustainable nutrient cycles. Until
plants are established reclamationists often fertilize areas to augment nutrient deficient soils.
When this occurs the availability of the nutrient is most likely not dependent upon its
concentration in the soil, but rather on the pH level of the soil as it pertains to nutrient
absorption. Brady and Weil (2002) related many other chemical and biological properties to pH
and considered it a master variable in determining nutrient availability. In this way, the equation
does account for at least a component of what is necessary for nutrient cycling, as it relates to a
plant’s ability to uptake nutrients from the soil. However, it may be beneficial to explore
variables that account for the concentrations of nutrients in the soil.
The most significant soil parameter investigated in this study that relates to soil nutrient
concentration is percent organic matter. Three major macronutrients, nitrogen, phosphorus and
potassium, which are generally deficient in post-mined soils (Coppin and Bradshaw, 1982;
Sheoran et al., 2008) have been related directly to soil organic content (Sheoran et al., 2010).
58

Although these factors have been correlated with soil organic content, exact concentrations of
žek and others (2003), as well as Whitford and Elkins (1986)
emphasized the importance of soil microorganisms to ecosystems and recommended proper soil
management during restoration to ensure biological health of the system. Again, there is no
variable that has been explored in this study that addresses the microfauna of soils directly.
Arshad and Martin (2002) identified soil respiration, as an indicator of microorganism health and
therefore of soil quality. Although soil respiration or other similar indicators have not been
utilized in this study they could be add to other studies for further investigation. A study by
Showalter and other (2007) found that microbial activity correlated with mine spoil pH, and that
pH correlated well with tree growth. pH is considered the only indicator of microorganism
activity used in the models in the research presented in this thesis.
The variables presented in this study are not a fully comprehensive set of soil quality
indicators but rather a set of variables experts that developed county soil surveys have identified
as significant soil parameters. Further exploration of possible variables should be conducted to
strengthen variance explanations. Furthermore, it is important to recognize that these equations
are regional and land use based. Regional factors may be included as the importance of soil
attributes that affect plant productivity are identified within various regions of study. For
instance, electric conductivity, as an indicator of soluble salts in the soil, was not included in this

individual nutrients have not been identified in the equations. It may be beneficial to investigate
variables to accurately assess effects of specific nutrients to account for further explanation of
the variance.
Microorganism populations in post-mined soils are often much smaller when compared
to undisturbed soils and restoration of populations takes time. Chodak and others (2009), Harris
(2003,2009), Jasper (2007),

59

study because salt content is not commonly considered a major limiting factor for plants in the
Upper Peninsula of Michigan. Instead, electric conductivity would more like be used to assess
soil quality in arid regions such as the southwestern United States (Scholl, 1986).
Some regional factors that have not explicitly included in this study are still represented
in the data set. For instance, there are many regions throughout the study that are characterized
by shallow soils and outcrops that would be expected to limit growth of larger species, however,
depth to bedrock was not one of the independent variables investigated in this study. Although
not represented with their own independent variable, bedrock depths were measured to be within
the first 48 inches of certain soil profile. The observed properties of the impermeable rock did
influence the soil attribute values associated with that soil type. For instance, the bulk density of
bedrock is considerably greater than all soils types in this region and would therefore produced
relatively high values for these soils. Similarly, the values of hydraulic conductivity, percent
clay content, organic matter content, and percent rock fragments would all be heavily influenced
by the presence of bedrock within the first 48 inches of earth.
Frost heave was identified by Linsemier (1989) as being a significant regional factor that
affects plant growth. Frost heave is generally associated with soil texture, depth of soil, and
depth of water table. As soils holding high amounts of water freeze and therefore expand and
shift, root systems are damaged and trees can be uprooted. The harsh winters and shallow soils
throughout the region makes frost heave an issue of major concern to the productivity of plants.
Although soil texture and depth to water table are attributes not directly expressed in the
equation as independent variables, I consider the measured properties used to create the data set
adequate for accounting for the influence of them. The texture of the soil is partially accounted
for by the parameters percent clay and percent rock fragments, although could be strengthen by
60

including other texture variables such as percent sand content. As previously mentioned, depth
to bedrock is represented in the equations as the values of bedrock with the first 48 inches of soil
affect attribute values. Depth to water table is largely unrepresented in this investigation. The
soil surveys used to derive soil data showed no crop or woody plant growth values for soils with
high water tables, commonly classified as mucks. It may therefore be beneficial to investigate
variables that account for the influence of high water tables.
The equations derived from this methodology are also land use based. The concept of
land use includes uses before, during, and the intended land use after mining. Land use prior to
mining as it relates to soil conditions is usually not a major concern for most reclamation
projects. However, rare instance in which industrial use could have contaminated soils prior to
mine could affect the reapplied A horizon of the post-mined soil. Soil contamination is most
likely to occur during the mining operations themselves. Soil contaminates can greatly influence
the soil quality and therefore the productivity potential of soils. Soil toxicity was not considered
a major variable to be considered in the development of the models due to the type of mineral
extraction that is currently being conducted in this region.
According to the United States Department of the Interior, U.S. Geological Survey, there
are currently no active mining operations in Iron County or Dickinson County (USGS, 2012), the
equations are to be applied on a regionally basis if to be applied directly to a mine reclamation
effort. The Tilden and Empire Mines are located in the adjacent, Marquette County. Both mines
are open pit iron mines. Iron mining is associated with relatively low occurrence of toxic
materials that persist in the soil when compared to the mining of many other materials, such as
gold, silver, nickel, or copper sulfide. The presented equations can only be applied to mining of
relatively inert substances such as sand, gravel, coal and iron when compared to mineral
61

extraction processes that result in more hazardous chemicals. However, the recent opening of a
copper and nickel mine in Marquette County may warrant the addition of further variable
investigation if being applied to this site. In September of 2011, the Kennecott Eagle Minerals
Company began underground copper and nickel mining operation. Bech et al. (1997) reported
the presence of Cu, Zn, Pb, Co, Ni, Cd, and As in soils surrounding copper mines. Although the
types of metal contaminates associated with copper mining depends on the ore being mined, it is
possible if some or all of these contaminants are present in post-mined soils from this operation.
Bech (1997) linked the availability of these elements to uptake by vegetation to soil parameters
suggesting that the exploration of such variables is beneficial. It may be beneficial to investigate
additional variables and redevelop the models if this study intends to be accurately applied to
mines that result in toxic materials.
The selection of crop and woody plant types in this study is also a limiting factor to its
use. Plant types were selected based on the NRCS data that was available. Although the
methodology presented in this study can be applied to any available data set, species are limited
in this case due to the data set being derived from a secondary source. Species are therefore
limited to those published in soil surveys. Increasing the number of species could have varying
affects on resulting equations. Firstly, it may give a more accurate indication on the community
type that the eigenvector describes. Investigating a broader array of plant types could also
develop equations for plant or community types that were not investigated in this study.
However, as previously stated, data collection for use in such equations is both time consuming
and costly. The methodology employed in this study presents a cost-effective process for the
development of empirical models to predict vegetative productivity potentials of post-mined
soils.
62

4.7 ADDITIONAL USES OF EQUATIONS
žek et al., 2012, Beauchamp et
al., 2006; Tripathy et al., 2008). The equations could be used to assess the influence of these
amendments as they change soil scores produced employing these models. By examining the
properties of a soil amendment and applying them to the equation as they proportionally
influence the properties of the treated soil, reclamationists are able to assess appropriate type and
quantities of given amendments. The addition of compost for instance, would be expected to
raise available water holding capacity, increase percent organic matter, and reduce the bulk
density of a soil. However, it is unknown how different quantities are expected to change the
productivity of soils and interact with other unaccounted for interactions that may not be

The models in this study can be used in two main ways. The primary intent of these
equations is to establish accurate models that predict growth potentials of crop and woody
species on reconstructed soils. However, these equations can also be analyzed as a means of
identifying post-mine soil treatments. For instance, cases in which plant species are described by
an equation that identifies bulk density as a major limiting factor to plant growth and the sampled
soil receives an insufficient score, reclamationists may recommend a tilling regime to reduce the
bulk density of the measured soil. This post-mine soil treatment would be expected to decrease
bulk density and therefore raise the soil index score. In this way the equations can be view not
only as a way to identify the applicability of a species or set of species to a soil, but also a way to
identify problematic conditions of the equation.
Similarly, the influence of soil amendments or alternative growth medias could be more
accurately predicted by the equations developed by this methodology. A considerable amount of
research has investigated the applications of various alternative substrates as growth media or
soil amendments (Zornoza et al., 2012; Watts et al., 201

63

apparent. The models developed in this study can be used to assess these interactions and assist
reclamationists in making amendment application decisions.

64

APPENDIX

65

Table A.1 Iron County Independent Variables
TP

IR1OCA
IR2OCB
IR3OCD
IR4PAA
IR5PAB
IR6PAD
IR7TRB
IR8TRD
IR9SOA
IR10CHA
IR11KAB
IR12KAD
IR13GAA
IR14FEA
IR15FEB
IR16FED
IR17PEB
IR18PED
IR19KEB
IR20KED
IR21ESB
IR22STA
IR23STB
IR24STD
IR25SAB
IR26SAD
IR27SUB
IR28SUD
IR29SUA
IR30MAA
IR31PEB
IR32PED
IR33SAB
IR34SAD
IR35ALB

1.5
1.5
2
2.5
2.5
2.5
2
3
1
1
2
2
1.5
2
2.5
2.5
3.5
3.5
2.5
3
2.5
1.5
2
3
3.5
4
3.5
4
3.5
1.5
2
2.5
2.5
3
2.5

SL

1.5
4
12
1.5
4
12
3.5
12
1.5
1.5
3.5
12
1.5
1.5
4
12
3.5
12
3.5
12
3.5
1.5
4
12
3.5
12
3.5
12
1.5
1.5
3.5
12
3.5
12
3

FR

0.138
0.138
0.138
4.5
4.5
4.85
2.85
2.85
7.2
5.583
0
0
0.479
0
0
0
9.233
9.233
12.252
12.252
10
5.013
5.013
5.013
8.383
8.383
4.366
4.366
3.934
4.646
5.65
5.65
11.05
11.05
1.5

CL

BD

10.842
10.842
10.842
9.583
9.583
9.583
11.238
11.238
14.25
6.35
8.591
8.591
12.967
12.541
12.541
12.541
6
6
8.25
8.25
8.6
10.675
10.675
10.675
8.267
8.267
6.263
6.263
7.058
7.238
8.429
8.429
8.5
8.5
16.113
66

1.54
1.54
1.54
1.536
1.536
1.536
1.488
1.488
1.533
1.482
1.484
1.484
1.534
1.532
1.532
1.532
1.505
1.505
1.624
1.624
1.478
1.416
1.416
1.416
1.601
1.601
1.541
1.541
1.541
1.568
1.607
1.607
1.384
1.384
1.596

HC

1.848
1.848
1.848
3.503
3.503
3.503
2.38
2.38
1.3
6.91
5.05
5.05
0.985
0.46
0.46
0.46
6.05
6.05
3.948
3.948
3.73
2.365
2.365
2.365
1.3
1.3
7.845
7.845
6.599
4.533
1.005
1.005
4
4
1.255

AW

0.148
0.148
0.148
0.139
0.139
0.139
0.154
0.154
0.152
0.107
0.114
0.114
0.19
0.19
0.19
0.19
0.099
0.099
0.092
0.092
0.101
0.205
0.205
0.205
0.118
0.118
0.127
0.127
0.142
0.108
0.146
0.146
0.128
0.128
0.175

PH

7.061
7.061
7.061
5.523
5.523
5.523
6.332
6.332
7.27
5.415
5.213
5.213
5.655
5.114
5.114
5.114
5.545
5.545
5.523
5.523
6.287
5.291
5.291
5.291
5.528
5.528
5.638
5.638
5.621
5.647
6.05
6.05
6.43
6.43
6.147

OM

1.5
1.5
1.5
1.25
1.25
1.25
2
2
2
2
1.25
1.25
3.5
1.5
1.5
1.5
2
2
1.25
1.25
1.75
2
2
2
2
2
2
2
2
2
2
2
2
2
3.5

Table A.1 Iron County Independent Variables (Cont’d)
TP

IR36MOA
IR37PEB
IR38PED
IR39MOA
IR40BEA
IR41STB
IR42LOB
IR43LOD
IR44MOA
IR45PEB

SL

FR

CL

BD

HC

AW

PH

OM

1
3.5
3.5
1
1
3
2.5
2.5
1
3.5

1.5
3.5
12
1.5
1
4
3.5
12
1
3.5

7.5
7.5
7.5
7.5
3.5
4.517
7.5
7.5
7.5
4.475

6.417
6.133
6.133
6.417
11
10.675
7.583
7.583
8.396
6

1.678
1.507
1.507
1.678
1.28
1.395
1.314
1.314
1.616
1.505

1.3
6.217
6.217
1.3
1.3
1.661
4.728
4.728
1.3
6.05

0.111
0.099
0.099
0.111
0.212
0.214
0.139
0.139
0.135
0.099

6.037
5.535
5.535
6.037
5
5.268
5.351
5.351
5.798
5.545

2.5
2
2
2.5
3
2
1.25
1.25
2.5
2

Table A.2 Iron County Dependent Variables
CO

IR1OCA
IR2OCB
IR3OCD
IR4PAA
IR5PAB
IR6PAD
IR7TRB
IR8TRD
IR9SOA
IR10CHA
IR11KAB
IR12KAD
IR13GAA
IR14FEA
IR15FEB
IR16FED
IR17PEB
IR18PED
IR19KEB
IR20KED

70
70
50
70
70
55
0
0
85
0
70
0
0
75
70
60
60
0
0
0

CS

11
11
8
11
11
9
14
0
14
0
13
0
0
12
11
10
10
0
10
0

OA

60
60
40
70
70
55
75
60
75
60
60
50
85
75
75
70
55
0
60
0

IP

0
0
0
0
0
0
350
0
0
0
0
0
0
0
0
0
0
0
250
0

AH

RM

0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
3
2.4

0
0
0
0
0
0
0
0
0
0
0
0
0
30.5
30.5
30.5
0
0
0
0
67

WS

0
0
0
0
0
0
0
0
0
20.5
0
0
20.5
20.5
20.5
20.5
0
0
20.5
20.5

RP

30.5
30.5
30.5
30.5
30.5
0
0
0
0
0
20.5
20.5
0
30.5
30.5
30.5
30.5
30.5
30.5
30.5

EP

30.5
30.5
30.5
30.5
30.5
0
0
0
0
30.5
20.5
20.5
30.5
30.5
30.5
30.5
30.5
30.5
30.5
30.5

JP

30.5
30.5
30.5
30.5
30.5
0
0
0
0
0
20.5
20.5
0
0
0
0
30.5
30.5
20.5
20.5

LI

11.5
11.5
11.5
11.5
11.5
0
0
0
0
11.5
11.5
11.5
0
11.5
11.5
11.5
11.5
11.5
11.5
11.5

A.2 Iron County Dependent Variables (cont’d)
CO

CS

OA

IP

AH

RM

WS

RP

EP

JP

LI

75
0
0
0
80
0
0
0
0
0
80
0
0
0
90
0
60
0
0
0
0
0
0
0
60

12
14
14
0
13
0
0
0
0
10
13
0
0
0
15
12
10
0
12
0
14
0
0
12
10

70
80
80
60
65
0
55
0
60
50
75
65
60
0
70
60
55
0
60
70
75
75
0
60
55

350
350
350
0
0
0
275
0
275
0
0
0
275
0
0
0
0
0
0
0
350
0
0
0
0

3.5
4
4
3
0
0
0
0
0
0
3.5
3.5
0
0
0
0
0
0
0
0
4
3.5
2.5
0
0

0
0
0
0
0
0
0
0
0
0
0
0
20.5
20.5
30.5
30.5
0
0
30.5
0
0
0
0
30.5
0

20.5
20.5
20.5
20.5
0
0
0
0
0
0
0
0
20.5
20.5
20.5
20.5
0
0
20.5
20.5
20.5
0
0
20.5
0

30.5
20.5
20.5
20.5
0
0
20.5
20.5
20.5
30.5
0
0
30.5
30.5
0
0
30.5
30.5
0
0
20.5
0
0
0
30.5

30.5
30.5
30.5
30.5
0
0
20.5
20.5
20.5
30.5
0
0
30.5
30.5
30.5
30.5
30.5
30.5
30.5
30.5
30.5
0
0
30.5
30.5

20.5
20.5
20.5
20.5
0
0
20.5
20.5
20.5
30.5
0
0
0
0
0
0
30.5
30.5
0
0
20.5
0
0
0
30.5

11.5
0
0
0
0
0
0
0
0
11.5
0
0
0
0
11.5
11.5
11.5
11.5
11.5
0
0
0
0
11.5
11.5

IR21ESB
IR22STA
IR23STB
IR24STD
IR25SAB
IR26SAD
IR27SUB
IR28SUD
IR29SUA
IR30MAA
IR31PEB
IR32PED
IR33SAB
IR34SAD
IR35ALB
IR36MOA
IR37PEB
IR38PED
IR39MOA
IR40BEA
IR41STB
IR42LOB
IR43LOD
IR44MOA
IR45PEB

Table A.3 Dickinson County Independent Variables
TP

DI1PEB
DI2PED
DI3PEF
DI4FEB
DI5FED

4
4.5
5
3.5
3

SL

3
12
26.5
3
12

FR

7.5
7.5
7.5
0
0

CL

10.265
10.265
10.265
12.2
12.2

BD

1.394
1.394
1.394
1.477
1.477
68

HC

3.825
3.825
3.825
0.64
0.64

AW

0.138
0.138
0.138
0.2
0.2

PH

5.82
5.82
5.82
5.05
5.05

OM

1.75
1.75
1.75
1.5
1.5

Table A.3 Dickinson County Independent Variables (cont’d)
TP

DI6KAB
DI7KAD
DI8KAF
DI9ESB
DI10ESD
DI11EMB
DI12EMD
DI13EMF
DI14PEB
DI15PED
DI16PEF
DI17DE
DI18ROA
DI19NAB
DI20NAD
DI21NAF
DI22ROB
DI23ROD
DI24ROF
DI25OCB
DI26OCD
DI27WAA
DI28MAB
DI29MAD
DI30MAF
DI31KI
DI32VIB
DI33VID
DI34VIF
DI35CHA
DI36ZIB
DI37ZID
DI38ZIF
DI39RUB
DI40RUD

4
4.5
5
3.5
4
3.5
4
5
3.5
4
5
1
3.5
4
4.5
5
4
4.5
5
3.5
4.5
1
4
4.5
5
1
4
4.5
5
1.5
4
4.5
5
4
4.5

SL

3
12
26.5
3
12
3
12
26.5
3
12
26.5
0
1.5
3
12
26.5
3
12
26.5
3
12
1.5
3
12
26.5
0
3
12
26.5
1.5
3
12
26.5
3
12

FR

0
0
0
2.625
2.625
2.5
2.5
2.5
4.683
4.683
4.683
0
0
8.5
8.5
8.5
0
0
0
1.275
1.275
0
3.417
3.417
3.417
0
0
0
0
4.563
0
0
0
0
0

CL

BD

5.817
5.817
5.817
8.879
8.879
15.417
15.417
15.417
5.5
5.5
5.5
5.8
5
7.675
7.675
7.675
5
5
5
9.775
9.775
6.55
7.625
7.625
7.625
5
2.983
2.983
2.983
6.3
4.75
4.75
4.75
4.437
4.437
69

1.489
1.489
1.489
1.484
1.484
1.624
1.624
1.624
1.522
1.522
1.522
1.515
1.468
1.414
1.414
1.414
1.468
1.468
1.468
1.556
1.556
1.401
1.375
1.375
1.375
1.45
1.579
1.579
1.579
1.491
1.595
1.595
1.595
1.443
1.443

HC

5.65
5.65
5.65
3.64
3.64
3.055
3.055
3.055
4.5
4.5
4.5
13
13
5.04
5.04
5.04
13
13
13
2.475
2.475
13
12.633
12.633
12.633
13
13
13
13
7.378
13
13
13
13
13

AW

0.087
0.087
0.087
0.105
0.105
0.13
0.13
0.13
0.09
0.09
0.09
0.064
0.079
0.125
0.125
0.125
0.07
0.07
0.07
0.12
0.12
0.082
0.083
0.083
0.083
0.05
0.08
0.08
0.08
0.103
0.08
0.08
0.08
0.061
0.061

PH

6.005
6.005
6.005
6.253
6.253
6.968
6.968
6.968
5.421
5.421
5.421
6.94
5.505
6.884
6.884
6.884
5.611
5.611
5.611
7.098
7.098
5.528
6.883
6.883
6.883
5.29
5.54
5.54
5.54
5.429
6.61
6.61
6.61
5.297
5.297

OM

1.25
1.25
1.25
1.75
1.75
2
2
2
2
2
2
8
1.5
2
2
2
1.5
1.5
1.5
1.5
1.5
3
1.75
1.75
1.75
9.5
1
1
1
2
1.5
1.5
1.5
0.75
0.75

Table A.3 Dickinson County Independent Variables (cont’d)
TP

DI41RUF
DI42UD
DI43SOB
DI44ENB
DI45ALB
DI46HE
DI47LOB
DI48LOD
DI49UVB
DI50UVD

5
2
1.5
1
3.5
1
3.5
4
4
4.5

SL

FR

26.5
0
2
0
3
0
3
12
4
12

CL

0
0
1.5
5
0.075
0
43.75
43.75
4.125
4.125

BD

4.437
0
13.354
14.65
16.637
31.3
8.083
8.083
18.125
18.125

HC

1.443
0
1.55
1.515
1.599
1.525
1.445
1.445
1.664
1.664

AW

13
0
1.3
3.189
1.255
0.432
1.263
1.263
0.92
0.92

PH

0.061
0
0.138
0.12
0.175
0.173
0.089
0.089
0.167
0.167

5.297
0
7.344
7.335
6.415
7.467
6.45
6.45
6.093
6.093

OM

0.75
0
2
5.5
3
3.5
2
2
2
2

Table A.4 Dickinson County Dependent Variables
CO

DI1PEB
DI2PED
DI3PEF
DI4FEB
DI5FED
DI6KAB
DI7KAD
DI8KAF
DI9ESB
DI10ESD
DI11EMB
DI12EMD
DI13EMF
DI14PEB
DI15PED
DI16PEF
DI17DE

CS

70
0
0
90
0
70
0
0
75
0
75
0
0
70
0
0
0

11
0
0
15
0
13
0
0
12
0
15
0
0
10
0
0
0

OA

60
50
0
80
70
60
55
0
70
55
75
70
0
60
55
0
0

IP

300
0
0
362.5
0
300
0
0
312.5
0
337.5
0
0
300
0
0
0

AH

RM

3
0
0
4
3.5
3
2.5
0
3.5
3.1
3.5
3.2
0
3
2.7
0
0

0
0
0
30.5
30.5
0
0
0
0
0
30.5
30.5
30.5
0
0
0
0

70

WS

20.5
20.5
20.5
20.5
20.5
0
0
0
20.5
20.5
20.5
20.5
20.5
0
0
0
30.5

RP

30.5
30.5
30.5
30.5
30.5
30.5
30.5
30.5
30.5
30.5
30.5
30.5
30.5
30.5
30.5
30.5
0

EP

30.5
30.5
30.5
30.5
30.5
30.5
30.5
30.5
30.5
30.5
30.5
30.5
30.5
30.5
30.5
30.5
30.5

JP

0
0
0
0
0
0
0
0
30.5
30.5
0
0
0
30.5
30.5
30.5
0

LI

11.5
11.5
11.5
11.5
11.5
11.5
11.5
11.5
11.5
11.5
11.5
11.5
11.5
11.5
11.5
11.5
11.5

Table A.4 Dickinson County Dependent Variables (cont’d)
CO

DI18ROA
DI19NAB
DI20NAD
DI21NAF
DI22ROB
DI23ROD
DI24ROF
DI25OCB
DI26OCD
DI27WAA
DI28MAB
DI29MAD
DI30MAF
DI31KI
DI32VIB
DI33VID
DI34VIF
DI35CHA
DI36ZIB
DI37ZID
DI38ZIF
DI39RUB
DI40RUD
DI41RUF
DI42UD
DI43SOB
DI44ENB
DI45ALB
DI46HE
DI47LOB
DI48LOD
DI49UVB
DI50UVD

CS

OA

IP

AH

RM

WS

RP

EP

JP

LI

70
70
0
0
60
0
0
70
50
70
70
0
0
0
50
0
0
70
70
0
0
0
0
0
0
80
0
90
0
70
0
80
70

10
11
0
0
10
0
0
11
8
14
13
0
0
0
8
0
0
11
10
0
0
0
0
0
0
14
0
14
0
11
0
15
14

60
70
65
0
50
45
0
70
65
60
60
55
0
0
40
0
0
65
60
55
0
0
0
0
0
80
0
85
0
70
65
65
60

300
312.5
0
0
287.5
0
0
312.5
0
0
300
0
0
0
262.5
0
0
0
300
0
0
0
0
0
0
0
0
0
0
312.5
0
362.5
0

3
3
2.6
0
2.8
2.2
0
3
2.5
3
3
2.7
0
0
2.5
0
0
3
3
2.7
0
2
0
0
0
3.5
0
4
0
3
2.7
3.8
3.8

0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
30.5
0
30.5
0
0
0
0

20.5
20.5
20.5
20.5
20.5
20.5
20.5
0
0
11.5
20.5
20.5
20.5
0
0
0
0
20.5
0
0
0
0
0
0
0
20.5
0
20.5
0
0
0
20.5

30.5
30.5
30.5
30.5
30.5
30.5
30.5
30.5
30.5
0
30.5
30.5
30.5
0
30.5
30.5
30.5
0
30.5
30.5
30.5
30.5
30.5
30.5
0
0
0
0
0
0
0
30.5

30.5
30.5
30.5
30.5
30.5
30.5
30.5
30.5
30.5
30.5
30.5
30.5
30.5
0
30.5
30.5
30.5
30.5
30.5
30.5
30.5
30.5
30.5
30.5
0
30.5
0
30.5
0
0
0
30.5

30.5
0
0
0
30.5
30.5
30.5
30.5
30.5
0
30.5
30.5
30.5
0
30.5
30.5
30.5
30.5
30.5
30.5
30.5
30.5
30.5
30.5
0
0
0
0
0
0
0
30.5

0
11.5
11.5
11.5
0
0
0
11.5
11.5
0
11.5
11.5
11.5
0
11.5
11.5
11.5
11.5
11.5
11.5
11.5
11.5
11.5
11.5
0
11.5
0
11.5
0
0
0
11.5

0

0

30.5

30.5

30.5

11.5

71

LITERATURE CITED

72

LITERATURE CITED

Acton, P., Fox, J.F., Campbell, J.E., Jones, A.L., Rowe, H., Martin, D., & Bryson, S. (2011).
Role of soil health in maintaining environmental sustainability of surface coal mining.
Environmental Science & Technology. 45(23), 10265-10272.
Agresti, A., & Franklin, C. (2012). Statistics: the art & science of learning from data (3rd ed.).
N.p.: Pearson College Division.
Albert, D. A. (1995). Regional landscape ecosystems of Michigan, Minnesota and Wisconsin: a
working map and classification. General Technical Report NC-178. St. Paul, MN: U.S.
Dept. of Agriculture, Forest Service, North Central Forest Experiment Station
Baker, J.B., & W.M. Broadfoot. (1978). A new technique of site selection for hardwoods.
Southern journal of Applied Forestry 2:42-43.
Baker, J.B., & W.M. Broadfoot. (1979). A practical field method of site evaluation for
commercially important southern hardwoods. U.S. Department of Agriculture, Forest
Service, Southern Forest Experiment Station General Technical Report SO-26.
Barnhisel, R. I., & BHower, J. M. (1994). The use of productivity index to predict corn yields
on restored prime farmland. In, International Land Reclamation and Mine Drainage
Conference and Third International Conference on the Abatement of Acid Drainage,
volume 3: Reclamation and Revegetation. United States Department of the Interior,
Bureau of Mines Special Publication SP 06-94: 48-56
R. I. Barnhisel, J. M. Hower, & L. D. Beard (Eds.) (1992). Development of a soil productivity
model for use in prime farmland reclamation. In, Prime Farmland Reclamation National
Symposium (pp. 205-211). N.p.: University of Illinois.
Beauchamp C.J., Camiré C., Chalifour F.P. (2006): Use of bark and combined paper sludge, for
the revegetation of bark-covered land. Journal of Environmental Engineering and
Science, 5: 253–261.
Bech, J., Poschenrieder, C., Llugany, M., Barcelo, J., Tume, P., Tobias, F. J., & Barranzuela, J.
L. (1997). Arenic and heavy metal contamination of soil vegetation around a copper min
in Northern Peru. The Science of the Total Environment, 203(1), 83-91.
Belli, K. L., Hodges, J. D., & Jeffery, C. G. (1998). Site productivity assessment for cherrybark
oak on minor stream bottoms in Mississippi: A comparison on methods. Southern
Journal of Applied Forestry, 22(1), 7-10.

73

Berg, W. A. (1975). Revegetation of land disturbed by surface mining in Colorado. In M. Wali
(Ed.), Practices and Problems of Land Reclamation in Western North America (pp. 7989). Grand Forks, ND: The University of North Dakota Press.
Brady, N.C., and R.R. Weil. (2002). The Nature and Properties of Soils. 13th ed. Prentice Hall
Inc., Upper Saddle River, NJ. 740 p.
Burger, J.A., Johnson, J.E., Andrews, J.A., and Torbert, J.L. (1994). Measuring mine soil
productivity for forests. International Land Reclamation and Mine Drainage Conference
and Third International Conference on the Abatement of Acid Drainage, volume 3:
Reclamation and Revegetation. United States Department of the Interior, Bureau of
Mines Special Publication SP 06-94: 48-56
Burley, J. B. (1990). Sugarbeet productivity model for Clay County, Minnesota. Journal of
Sugar Beet Research, 27(3&4), 50-57.
Burley, J. B. (1991). A vegetation productivity equation for reclaiming surface mines in Clay
County, Minnesota. International Journal of Surface Mining and Reclamation, 1-6.
Burley, J. B. (1995a). Vegetation productivity soil reclamation equations for the North Dakota
coal fields. Doctoral dissertation, University of Michigan
Burley, J.B. (1995b). A multi-county vegetation productivity equation for soil reclamation. In
Hynes, T.P. & Blanchette, M.C. (Ed.), In, Proceedings of Sudbury ’95-Mining and the
Environment. CANMET, Ottawa, (3): 1113-1112.
Burley, J. B. (2001). Environmental design for reclaiming surface mines. Lewiston, N.Y: Edwin
Mellen Press.
Burley, J. B., & Bauer, A. (1993). Neo-soil vegetation productivity equations for reclaiming
disturbed landscapes: a central florida example. In B. A. Zamora & R. E. Conally (Eds.),
The Challenges of Integrating Diverse Perspectives in Reclamation: Proceedings of the
10th Annual National Meeting of the American Society for Surface Mining and
Reclamation (pp. 334-347). Spokane, WA: ASSMR.
Burley, J. B., Fowler, G., & Polakowski, K. J. (1996). Construction vegetation productivity
equations by employing undisturbed soils data: an Oliver County, North Dakota case
study. In W. L. Daniels, J. Burger, & C. E. Zipper (Eds.), In, Proceedings Thirteenth
Annual Meeting of the American Society for Surface Mining and Reclamation: Success
and Failures (pp. 393-401). Blacksburg, VA: Powell River Project.
Burley, J. B., Fowler, G. W., Polakowski, K., & Brown, T. J. (2001). Soil based vegetation
productivity model for the North Dakota coal mining region. International Journal of
Surface Mining, Reclamation and Environment, 15(4), 213-234.

74

Burley, J. B., & Thomsen, C. H. (1987). Multivariate techniques to develop vegetation
productivity models for neo-soils. 1987 National Symposium on Mining, Hydrology,
Sedimentation, and Reclamation (pp. 153-161). Lexington: University of Kentucky.
Burley, J. B., & Thomsen, C. H. (1990). Application of an agricultural soil productivity equation
for reclaiming surface mines: Clay County, Minnesota. International Journal of Surface
Mining and Reclamation, 4, 139-144.
Burley, J. B., Thomsen, C. H., & Kenkel, N. (1989). Development of an agricultural soil
productivity equation for reclaiming surface mines in Clay County, Minnesota.
Environmental Management, 13(5), 631-638.
Colby, W. E. (1945). Mining law in recent years. California Law Review, 33(3), 368-387.
Coppin, N.J., & Bradshaw, A.D. (1982). The establishment of vegetation in quarries and openpit non-metal mines. Mining Journal Books, London, 112p.
Corwin, D. L., & Lesch, S. M. (2005). Apparent soil electric conductivity measurements in
agriculture. Computers and Electronics in Agriculture, 46(1-3), 11-43.
Curtis, J. T. (1959). The vegetation of Wisconsin: An ordination of plant communities. Madison:
Wisconsin Press.
Dancer, W. S., & Jansen, I. J. (1981). Greenhouse evaluation of solum and substratum materials
in the southern Illinois coal field. Journal of Environmental Quality, 10, 396-400.
Darmody, R.G., Dunker, R.E., & Barnhisel, R.I (2002). Reclamation of prime agricultural lands
after coal surface mining: The midwestern experience. 2002 National Meeting of the
American Society of Mining and Reclamation, Lexington KY.
Doll, E. C., Merril, S. D., & Halvorson, G. A. (1984). Soil replacement for reclamation of
stripmined lands in North Dakota. Bulletin of the North Dakota Agricultural Experiment
Station, 514.
Doll, E. C., & Wollenhaupt, N. C. (1985). Use of soil parameters in the evaluation of reclamation
success in North Dakota. ASSMR Second Annual Meeting. Denver, Colorado, 91-94.
Edgeton, D. L., Harris, J. A., Birch, P., & Bullock, P. (1995). Linear relationship between
aggregate stability and microbial biomass in three restored soils. Soil Biology and
Biochemistry, 1499-1501.
Evanylo, G. K., Abaye, A. O., Dundas, C., Zipper, C. E., Lemus, R., Sukkariyah, B., &
Rockett, J. (2005). Herbaceous vegetative productivity, persistence, and metal uptake on a
boisolids-amended mine soil. Journal of Environmental Quality, 34(5), 1811-1819.

75

Friedlander, J. (2001). Design principles for reclamation of agricultural lands. In J. B. Burley
(Ed.), Environmental design for reclaiming surface mines (pp. 155-183). Lewiston, NY:
Edwin Mellen Press.
Gale, M. R., Grigal, D. F., & Harding, R. B. (1991). Soil productivity index: predicion of site
quality for white spruce plantations. Soil Science Society of America Journal, 55, 1701-1708.
Groninger, J. W., Aust, W. W., Miwa, M., & Stanturf, J. A. (2000). Growth predictions for tree
species planted on marginal soybean lands in the Lower Mississippi Valley. Journal of
Soil and Water Conervation, 55(1), 91-95.
Harris J.A. (2003): Measurements of the soil microbial community for estimating the success of
restoration. European Journal of Soil Science, 54: 801–808.
Harris J.A. (2009): Soil microbial communities and restoration ecology: Facilitators or
followers? Science, 325: 573–574.
Harris, J. P., Birch, P., & Short, K. C. (1989). Changes in the microbial community and physiochemical characterists of top soils stockpiled during opencast mining. Soil Use
Management, 161-168.
Jasper D.A. (2007): Beneficial soil microorganisms of the Jarrah forest and their recovery in
bauxite Southwestern Australia. Restoration Ecology, 15: S74–S84.
LaFevers, J. R. (1977). Effect of legislative change on reclamation. In J. L. Thames (Ed.),
Reclamation and Use of Disturbed Land in the Southwest (pp. 18-22). Tuscon, AZ: The
University of Arizona Press.
Le Cleac'h, G., M. Salles, and J.B. Burley. (2004). Vegetation productivity model for Grand
Traverse County, Michigan. Abstract Booklet. 2004 Joint Conference, 21st National
Meeting of the American Society of Mining and Reclamation and The 25th West Virginia
Surface Mine Drainage Task Force. Morgantown, West Virginia, April 18-24, 2004,
ASMR, 3134 Montavesta Rd., Lexington, KY 40502, :34.
Linsemier, L. H. (1989). Soil survey of Dickinson County, Michigan. Natural Resources
Conservation Service: Forest Service.
Linsemier, L. H. (1997). Soil survey of Iron County, Michigan. Natural Resources Conservation
Service: Forest Service.
Lucas, K.R. 1987. OSMRE oversight of state primacy-is it working? Graves, D. (ed.) In:
Proceedings, 1987 National Symposium on Mining, Hydrology, Sedimentology, and
Reclamation. University of Kentucky :113-116

76

žek L.,

McNab, W.H.; Cleland, D.T.; Freeouf, J.A.; Keys, Jr., J.E.; Nowacki, G.J.; Carpenter, C.A., comps.
(2005). Description of ecoregional subregions: sections of the conterminous United
States. Washington, DC: U.S. Department of Agriculture, Forest Service. (5-8).

McSweeney, K., Jansen, I. J., & Dancer, W. S. (1981). Subsurface horizon blending: An
alternative strategy to B horizon replacement for the construction of post-mine soils. Soil
Science Society of America Journal, 45, 795-799.
Munshower, F., & Judy, C. (1988). Resource perspective and the surfae coal mine regulatory
system. In L. R. Hossner (Ed.), Reclamation of Surface Mined Lands (Vol. 1, pp. 19-39).
Neill, L. L. (1979). An evaluation of soil productivity based on root growth and water depletion
(Master's thesis).
Ottenhof, C.J.M., Faz Cano, A., Arocena, J.M., Nierop, K.G.J., Verstraten, J.M., van Mourik,
J.M. (2007). Soil organic matter from pioneer species and its implications to
phytostabilization of mined sites in the Sierra de Cartagena (Spain). Chemosphere.
69(9), 1341-1350.
Potter, K. N., Carter, F. S., & Doll, E. C. (1988). Physical properties of constructed and
undisturbed soils. Soil Science Society of America Journal, 1435-1438.
Power, J. F., Sandoval, F. M., & Ries, R. E. (1978). Restoration of productivity to disturbed land
in the northern Great Plains. In R. A. Wright (Ed.), Reclamation of Disturbed Arid Lands
(pp. 33-49). Albuquerque: University of New Mexico Press.
Rodrigue, J. A., & Burger, J. A. (2004). Forest soil productivity of mined land in the midwest
and eastern coalfield regions. Soil Science Society of America Journal, 68(3), 833-844.
Rook, E.J.S. (2002). Flora, fauna, Earth, and sky: The natural history of the north woods. In
Rook.org. Retrieved November 2, 2012, from
http://www.rook.org/earl/bwca/nature/trees/index.html

77

Sheoran, A.S., Sheoran V., & Poonia, P. (2008). Rehabilitation of mine degraded land by
metallophytes. Mining Engineers Journal 10 (3), 11-16.
Sheoran, V., Sheoran, A.S., & Poonia, P. (2010). Soil reclamation of abandoned mine land by
revegetation: A review. International Journal or Soil, Sediment and Water, 3 (2), 1-20
Showalter, J., Burger, J., Zipper, C., Galbraith, J., & Donovan, P. (2007). Physical, chemical, and
biological mine soil properties influence white oak seedling growth: A proposed mine
soil classification model. South J. Appl. For, 31, 99-107.
Simpson, J. W. (1985). The emotional landscape and public law 95-87. Landscape Architecture
Magazine, 75(3), 60-114.
Smith, J. A. (2001). Agriculture in the north: a Canadian case study. In J. B. Burley (Ed.),
Environmental Design for Reclaiming Surface Mines (pp. 172-183). Lewiston, NY: The
Edwin Mellen Press.
Sommers, L. M. (1984). Michigan, a geography. Boulder, Colo: Westview Press.
Tripathy S., Bhattacharyya P., Equeenuddin S.M., Kim K., Kulkarni H.D. (2008): Comparison of
microbial indicators under two wa- ter regimes in a soil amended with combined paper
mill sludge and decomposed cow manure. Chemosphere, 71: 168–175.
Udvardy, M. (1975). A classification of the biogeographical provinces of the world. In
UNESCO’s Man and the Biosphere Programme Project No. 8. (1-50). Morges,
Switzerland.
USDA. (1992). Weights, measurements, and conversion factors for agricultural commodities and
their products. In Agriculture Handbook Number 697 (p. 10).
USGS. (2012). Mineral resources on-line spatial data. In USGS.gov. Retrieved November 1,
2012, from http://mrdata.usgs.gov/mineplant/find-mines.php
Watts, D. B., Arriaga, F. J., Torbert, A. H., Busby, R. R., & Gebhart, D. L. (2012).
Noncomposted municipal soild waste byproduct influences soil and plant nutrients 5
years after soil reclamation. Soil Science, 177(8), 480-489.
Whitford, W. G., & Elkins, N. Z. (1986). The importance of soil ecology and the ecosystem
perspective in surface-mine reclamation. In C. C. Reith & L. D. Potter (Eds.), Principles
& Methods of Reclamation Science (pp. 151-182). Albuquerque: University of New
Mexico Press.
Workman, J.W. (1987). Making state primacy meaningful. Graves, D. (ed.) In: Proceedings,
1987 National Symposium on Mining, Hydrology, Sedimentology, and Reclamation.
University of Kentucky :107-108
78

Zipper, C. E., Burger, J. A., McGrath, J. M., Rodrigue, J. A., & Holtzman, G. I. (2011). Forest
restoration potentials of coal-mined lands in the eastern United States. Journal of
Environmental Quality, 40(5), 1567-1577.
Zornoza, R., Faz, A., Carmona, D. M., Kabas, S., Martinez-Martinez, S., & Acosta, J. A. (2012).
Plant cover and soil biochemical properties in a mine tailing pond five years after
application of marble wastes and organic amendments. Pedosphere, 22(1), 22-32.

79