to}... . .1 . f. v‘
. 1.... . o. a... ﬁgmfv!
.S ‘ . 0'.

t.
.J
‘ Kata! ,
u:
..

mum-"3%!

‘1 ‘
”ﬂaw.“
2.. a

 

_ .-.n
v x

; ‘

n... 91.5."

 

1.91%... , ‘
. , . . A 3,011.»?
new}: yww‘ ,LJubtﬁ . . . ‘

 

 

 

LIBRARY
I Michigan State .
University

This is to certify that the
dissertation entitled

SOIL AGGREGATE STABILITY: CROPPING SYSTEM
EFFECTS, SPATIAL VARIABILITY AND CONTRIBUTION TO
POTATO YIELD IN MICHIGAN

presented by

Edgar Allan C. Po

has been accepted towards fulﬁllment
of the requirements for the

Ph.D. degree in Crop and Soil Sciences

 

 

 

 

August 20, 2007

 

Date

MSU is an afﬁnnative-action, equal-opportunity employer

4 _.----n-o-o-o-o-o-g-.—o—n-o----~—-~-.-a--o-u—o-u-

.-unun—n-c-o-o-u-h-n-a-o-o-O--—-_ --.—.-.-n-n-a-c-u-u-u-o-n-o-.—-—.

 

PLACE IN RETURN BOX to remove this checkout from your record.
To AVOID FINES return on or before date due.
MAY BE RECALLED with earlier due date if requested.

 

DATE DUE DATE DUE DATE DUE

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

6/07 p2/CIRC/DaleDueindd-p1

SOIL AGGREGATE STABILITY: CROPPING SYSTEM EFFECTS, SPATIAL
VARIABILITY AND CONTRIBUTION TO POTATO YIELD IN MICHIGAN

By

Edgar Allan C. PO

A DISSERTATION

Submitted to
Michigan State University
In partial fulﬁllment of the requirements
for the degree of

DOCTOR OF PHILOSOPHY
Department of Crop and Soil Sciences

2007

ABSTRACT

SOIL AGGREGATE STABILITY: CROPPING SYSTEM EFFECTS, SPATIAL
VARIABILITY AND CONTRIBUTION TO POTATO YIELD IN MICHIGAN

By
Edgar Allan C. Po

Soil structure stability must be maintained to facilitate gaseous and water
exchange between the root and its external environment, ensuring optimum potato crop
productivity in the process. Seven potato rotation systems (2-year) with varying amounts
of carbon (C) inputs were evaluated over 3 years beginning in 2001, for their effects on
the dynamics of water stable soil aggregation (WSA) and C sequestration at a ﬁeld trial
located in Central Michigan on an irrigated Alﬁsol. The systems consisted of four main
crops (potato, snapbean, corn, wheat) and three winter cover crops (rye, hairy vetch, and
red clover). Regression and correlation analyses were conducted on the relationship of
WSA and digital imagery sources (i.e., in-ﬁeld and archived high-resolution imageries) to
potato yield in 2003 and 2004 from two commercial ﬁelds. Systems involving sweet corn
contributed 2-fold higher biomass than those with wheat or snapbean, and presence of a
legume in a cover crop system contributed signiﬁcantly higher amounts of C inputs (1.2
Mgha") compared to presence of rye alone (0.7 Mgha'l). In 2004, macro-WSAS declined
by 13% from 2001 levels for all systems except those with high C inputs, which
maintained macroaggregates. Residue C input was a moderate predictor of total soil C
(31% of variability explained), whereas macro and micro WSAs were signiﬁcant
predictors of total soil C, accounting for 58 and 72 % of observed variability,
respectively. Results from the two commercial ﬁelds’ stepwise and principal component

regression analyses explained an average of 63% and 54% of the yield variance.

The positive contribution of soil structure stability to yield indicates the need for
its management. Use of cereals as rotation crop with potato and leguminous cover crop
appeared to be an effective means to improve soil structure stability. In-season use of
digital imagery can potentially aid in pre-harvest yield mapping and monitoring of

important soil physical attributes.

ACKNOWLEDGMENTS

I would like to express my heartfelt thanks to my major professor, Dr. Sieglinde
Snapp, for the opportunity to collaborate on a research with practical impact to potato soil
resource management in Michigan; for access to ﬁnancial and personnel resources; and
for her patience along the way. For valuable technical advice and stimuli to think
wholistically, sincere appreciation goes to members of my graduate committee: Drs.
Alvin Smucker, Alexandra Kravchenko, Suzanne Lang and Richard Leep.

To Kitty O’neil and the rest of the resilient vegetable group over the years, for the
encouragement and support, my heartfelt thanks. For the excellent logistical support at
the Agronomy Crop Barns, my thank goes to Brian Graff and his staff.

Special thanks goes to Mark Otto for facilitating ﬁeld selection; availability of
historical data on the selected ﬁelds; and valuable advice along the way. Appreciation is
also extended to the Andersen Brothers for allowing me to use their ﬁelds for the study,
and the Michigan Potato Industry Council for funding part of the research.

I thank the Philippine government, the ﬁne folks at the Philippine American
Education Foundation - Fulbright, and supportive colleagues at Ateneo de Cagayan for
making my educational opportunity here in Michigan State University a reality.

Finally, a huge bear hug for my wife “Lillian”, for always being there even when
things looked a bit desperate; when everything was on the up and up; and everything in-
between. For my parents. Manuel and Flora; my siblings Abin, Josie and Fernando, I

appreciate the moral support.

iv

PREFACE

The three Chapters in this dissertation were written in the style required for
publication in the Soil Science Society of America Journal. Certain sections of the
materials and methods as well as some tables and ﬁgures were intentionally repeated,

hence will appear redundant.

TABLE OF CONTENTS

ACKNOWLEDGMENTS ................................................................................................. iv
LIST OF TABLES ........................................................................................................... viii
LIST OF FIGURES ............................................................................................................ x
CHAPTER 1: EVALUATION OF THE EFFECTS OF SEVEN POTATO SHORT
ROTATION COVER CROP SYSTEMS ON SOIL AGGREGATE STABILITY ........... 1
ABSTRACT ........................................................................................................................ 1
INTRODUCTION .............................................................................................................. 3
MATERIALS AND METHODS ........................................................................................ 6
Experimental Design and Treatments ............................................................................. 6
Carbon and Nitrogen Input ............................................................................................. 7
Water Stable Aggregates (WSA) .................................................................................... 8
Soil Properties ................................................................................................................. 9
Statistical Analyses ....................................................................................................... 10
RESULTS AND DISCUSSION ....................................................................................... 11
Entry Year Effects ......................................................................................................... 11
System Organic Inputs and Soil Response ................................................................... 11
System Effects and Aggregation ................................................................................... 14
Soil Aggregation and Carbon Dynamics ...................................................................... 15
CONCLUSION ................................................................................................................. 17
REFERENCES ................................................................................................................. 18
CHAPTER II: THE RELATIONSHIP OF SOIL PHYSICAL CHARACTERISTICS TO
YIELD IN INTENSIVELY MANAGED FIELDS .......................................................... 32
ABSTRACT ...................................................................................................................... 32
INTRODUCTION ............................................................................................................ 33
MATERIALS AND METHODS ...................................................................................... 40
Location ........................................................................................................................ 40
Climatic and Soil Conditions ........................................................................................ 40
Sampling Design ........................................................................................................... 41
Sampling for Water Stable Aggregates ......................................................................... 42

vi

Chemical and Mechanical Analyses ............................................................................. 43

In-situ Measurement ..................................................................................................... 43
Bulk density and Inﬁltration ..................................................................................... 44
Spectral Pictures ....................................................................................................... 45
Yield Monitor Data Collection .................................................................................. 45

Statistical Analysis ........................................................................................................ 46

RESULTS AND DISCUSSION ....................................................................................... 47

Spectral Response ......................................................................................................... 47

Yield Predictors ............................................................................................................ 49

F reduction test and principal components analyses ..................................................... 53

CONCLUSION ................................................................................................................. 55
REFERENCES ................................................................................................................. 56

CHAPTER III: SPATIAL BEHAVIOUR OF SOIL PROPERTIES AND THEIR
IMPLICATION FOR SAMPLING STRATEGIES IN A POTATO-BASED SYSTEM 72

ABSTRACT ...................................................................................................................... 72
INTRODUCTION ............................................................................................................ 74
MATERIALS AND METHODS ...................................................................................... 83
Location ........................................................................................................................ 83
Sampling Design ........................................................................................................... 83
Sampling for Water Stable Aggregates ......................................................................... 84
Chemical and Mechanical Analysis .............................................................................. 85
Moisture Content and Hydraulic Conductivity Calibration .......................................... 86
In-situ Measurement ..................................................................................................... 86
Bulk density and Inﬁltration ..................................................................................... 87
Spectral Pictures ....................................................................................................... 89
Yield Monitor Data Collection .................................................................................. 90
Statistical Analysis ....................................................................................................... 90
RESULTS AND DISCUSSION ....................................................................................... 92
Climatic and Soil Conditions ........................................................................................ 92
Spatial Distribution of Yield ......................................................................................... 92
DOQQ and Selected Soil Attributes ............................................................................. 95
Spatial Behavior of Selected Soil Attributes ................................................................ 96
CONCLUSION ................................................................................................................. 98
REFERENCES ............................................................................................................... 100

vii

LIST OF TABLES

Table 1.1 Mean values and standard errors for soil properties in the potato rotation trial at
the Montcalm Research Farm, Entrican, Michigan. ................................................. 22

Table 1.2 Rotation system and winter cover crop treatments used in a 2-year potato
experiment at Montcalm Research Farm, Entrican, MI. ........................................... 23

Table 1.3 Mean water stable aggregate, macroaggregates (Macro-Agg), total soil carbon
(TSC), mean weight diameter (MWD) and standard errors across different size
classes and potato rotation systems in 2001 at Montcalm Research Farm, Entrican,
Michigan. .................................................................................................................. 24

Table 1.4 Mean water stable aggregate, macroaggregates (Macro-Agg), total soil carbon
(TSC), mean weight diameter (MWD), carbon (C) input and standard errors for 7
short potato rotation systems in 2004 at Montcalm Research Farm near Entrican,
Michigan. .................................................................................................................. 25

Table 1.5 Sums of squares reduction test of full and reduced model for total soil carbon
as a function of carbon input and proportion of water stable aggregates from a potato
short rotation cropping system study at the Montcalm Research Farm near Entrican,
Michigan. .................................................................................................................. 26

Table 2.1 Basic statistics for potato yield, speciﬁc gravity and regressor variables
classiﬁed into chemical, physical and spectral factors taken from Fields 1 and 2 in
Montcalm County, Michigan in 2003 and 2004 respectively. .................................. 61

Table 2.2a Pearson correlation matrix of selected soil, plant and spectral properties of
ﬁeld 1 in Montcalm, Michigan. ................................................................................ 62

Table 2.2b Pearson correlation matrix of selected soil, plant and spectral properties from
ﬁeld 2 in Montcalm, Michigan. ................................................................................ 63

Table 2.3 Sums of squares reduction test of full and reduced model for potato yield as a
function of chemical, physical and plant spectral response variables from a potato
yield variability study at two commercial ﬁelds in Montcalm County, Michigan... 64

Table 2.4 Stepwise principal component regression analyses of chemical, physical and
plant response variableS‘l’. ......................................................................................... 65

Table 3.1 Basic statistics for potato yield, speciﬁc gravity and regressor variables

classiﬁed into chemical, physical and spectral factors taken from ﬁeld 1 and 2 in
Montcalm County, Michigan in 2003 and 2004 respectively. ................................ 104

viii

Table 3.2 Selected soil properties of ﬁeld 1 showing signiﬁcant correlations with digital
orthophoto quarter quad. ......................................................................................... 105

Table 3.3 Selected soil properties for ﬁeld 2 as a whole (whole), datapoints found on
Mancelona (Mb) and Gladwin (Ga) soil series sampled in fall 2004 showing
correlations with digital orthophoto quarter quad image taken in 1997. ................ 106

Table 3.4 Variogram proﬁle for soil structure related properties of ﬁeld 1 and 2 .......... 107

Table 3.5a Pearson correlation matrix of selected soil, plant and spectral properties of
ﬁeld 1 in Montcalm, Michigan. .............................................................................. 108

Table 3.5b Pearson correlation matrix of selected soil, plant and spectral properties from
ﬁeld 2 in Montcalm, Michigan. .............................................................................. 109

ix

LIST OF FIGURES

Figure 1.1 Total monthly precipitation at Montcalm Research Farm near Entrican,
Michigan for 2001 - 2004, and the 8-year average. Annual precipitation was 991,
827, 478, and 735 mm for 2001, 2002, 2003 and 2004 respectively. ....................... 27

Figure 1.2 Carbon inputs from 7 potato rotation systems (A) and averaged across systems
for pre-planned contrasts, SI and S4 for the low input category, S2 and S6 for the
medium category and S3, SS and S7 for the high C input category (B). The potato
rotation systems are described in the text, and C inputs included above- and below-
ground organic material averaged over two years (i.e., 2003 and 2004). Bottom,
middle and upper letters designate statistical signiﬁcance for main crop, cover crop,
and total C input across but not within systems. Columns with similar letters are not
signiﬁcantly different at 5% ...................................................................................... 28

Figure 1.3 Inﬂuence of seven short rotation management options for potato production
systems (A) and low (LC, 1.2 Mgha’l), medium (MC, 2.0 Mgha' ), and high (HC,
2.8 Mgha'l) C input (B) on the percent change of mean weight diameter (MWD,
mm) in 2004 compared to 2001 at the Montcalm Research Farm near Entrican,
Michigan. Bars with the same letter designation are not statistically different at S
5%. ............................................................................................................................ 29

Figure 1.4 Inﬂuence of seven short rotation management options for potato production
systems (A) and low (LC, 1.2 Mgha’l), medium (MC, 2.0 Mgha' ), and high (HC,
2.8 Mgha'l) C input (B) on the percent change of mean macroaggregates in 2004
compared to 2001 at the Montcalm Research Farm near Entrican, Michigan. Bars
with the same letter designation are not statistically different at S 5%. ................... 30

Figure 1.5 Plot of total soil carbon (C, %), carbon input, and water stable
macroaggregates (2-0.25 mm diameter) from a potato short rotation cropping system
study at the Montcalm Research Farm near Entrican, Michigan. Letters refer to
comparisons in water stable aggregates. Systems with similar letters are not
statistically different at 5%. l8] PBSBB = potato-bare, snapbean-bare; $2 PRSBR =
potato-rye,snapbean-rye; S3 PRSCB = potato-rye, sweetcom—bare; S4 PWWR =
potato-wheat-wheat-rye; SS PWWCC = potato-wheat-wheat/clover-clover; S6
PRVSBRV = potato-rye/vetch-snapbean-rye/vetch; S7 PRVSCRV = potato-
rye/vetch-sweetcom-ryevetch; .................................................................................. 3 1

Figure 2.1 Total monthly precipitation at Montcalm Research Farm (Entrican, Michigan)
for 2003, 2004, and the 8-year average ..................................................................... 66

Figure 2.2 Study ﬁeld 1 located at the Southeast quadrat of Vickeryville-Tamarack Roads
in Vestaburg, Montcalm County, Michigan showing the entire ﬁeld and the location
of grid points sampled. Soil types of the area is also shown: Mb (Mancelona loamy

sand, 0 — 2% slope), Mc (Mancelona loamy sand, 2 — 6% slope) and Nm (Newaygo
sandy loam with 2 to 6% slope). ............................................................................... 67

Figure 2.3 Study ﬁeld 2 located near the southeast quadrat of Tamarack-Bollinger Roads
in Vestaburg, Montcalm County, Michigan showing the the location of grid points
sampled by this study as well as soil types Ga (Gladwin loamy sand) and Mb
(Mancelona loamy sand). .......................................................................................... 68

Figure 2.4 Image acquisition, processing and analyses involving a Red-Green-Blue
(RGB, A) image and an Infrared RGB (B) being cropped (A1 and 3;) to remove
edge artifacts, composited into Idrisi for Windows (A2, A3, A4; B2, B3, B4),
unsupervised clustering into two groups was then performed on the infrared
composite image (B5) to produce an image with 0 being non—vegetated areas and 1
being vegetated areas (B6). Pixels having 1 as value in the image were then counted
and express as a percentage of the entire image pixels to represent percent vegetated.
Original image is in color. ........................................................................................ 69

Figure 2.5 Plot of actual and predicted potato yield from ﬁeld 1 fall of 2003 harvest data
using the signiﬁcant (P<0.0001) stepwise predictive equation of yield = -392.9 +
0.042(K) + 2.3359(Clay) + 8.7985(MWD) + 0.6178(Green Band
unadjusted)+1.307(Elevation, m) with adjusted R2 = 0.67 and an RMSE of 4.38. .. 70

Figure 2.6 Plot of actual and predicted potato yield from ﬁeld 2 fall of 2004 harvest data
using the signiﬁcant (P<0.0001) stepwise predictive equation of yield =
59.288+0.6997(250mm WSA)-89.259(Green by Red unadjusted ratio)+91.92(ECa)
with adjusted R2 = 0.60 and an RMSE of 5.9. .......................................................... 71

Figure 3.1 Digital Orthophotography Quarter Quad image detailing location of ﬁeld 2
superimposed by soil series map showing presence of Mancelona loamy sand (Mb)
and Gladwin loamy sand and Palo sandy loam (Ga) soil series. Original is in color.

................................................................................................................................. 110

Figure 3.2. A section of a digital orthophotography quarter quad (DOQQ) image detailing
location of ﬁeld 1 superimposed by soil series map showing presence of Mancelona
loamy sand with 0 to 2 percent slope (Mb), Mancelona loamy sand with 2 to 6
percent slope (Mc), and Newaygo sandy loam (Nm). ............................................ 111

Figure 3.3 Semi-variogram (y) plot for yield data taken from ﬁeld 2 during the fall of
2004 ﬁtted with exponential model (white line) showing a strong spatial dependence
among data taken within a range of approximately 50 m. ...................................... 112

Figure 3.4 Semi-variogram (y) plot for yield data taken from ﬁeld 1 during the fall of

2003 ﬁtted with exponential model (white line) showing a moderate spatial
dependence among data taken within a range of approximately 42 m. .................. 112

xi

CHAPTER 1: EVALUATION OF THE EFFECTS OF SEVEN POTATO SHORT

ROTATION COVER CROP SYSTEMS ON SOIL AGGREGATE STABILITY

ABSTRACT

Understanding processes that ameliorate structural degradation in sandy soils is
becoming of increasing importance as land values rise and cropping systems intensify.
Seven potato rotation systems (2-year) with varying amounts of carbon (C) inputs were
evaluated over 3 years for their effects on the dynamics of soil aggregation and C
sequestration. The 7 systems evaluated winter management (fallow, rye cover crop, rye-
hairy vetch cover crop mixture and red clover cover crop) and alternative crops rotated
with potatoes (snap bean, wheat and corn). Both entry points of the 2-year rotations were
included each year in the ﬁeld trial located in Central Michigan on an irrigated Alﬁsol.
The systems were compared individually and grouped into three categories to contrast C
inputs from cover crop and crop residues above- and below- ground. Average C input for
low (81 and S4), medium (S2 and S6), and high (SS, S3 and S7) input was 1.2, 2.0, and
2.8 Mgha'l respectively. Response variables included water stable aggregate (WSA) size
fraction 4 to 22 mm, <2 to 21 mm, <1 to 20.25 mm, <0.25 to 20.106 mm, <0.106 to
20.053 mm, macroaggregates (20.25 mm), microaggregates ($0.25 mm), mean weight
diameter (MWD), and total soil C. Systems involving sweet corn contributed 2-fold
higher biomass than those with wheat or snapbean, and presence of a legume in a cover
crop system contributed signiﬁcantly higher amounts of C inputs (1.2 Mgha'l) compared

to presence of rye alone (0.7 Mgha”'). Only the entry year inﬂuenced macro aggregates

in the ﬁrst year of the trial (2001), whereas both entry year and system inﬂuenced
aggregate size classes in the fourth year. In 2004, macro-WSAs declined by 13% from
2001 levels for all systems except those with high C inputs, which maintained
macroaggregates. Residue C input was a moderate predictor of total soil C (31% of
variability explained), whereas macro and micro WSAs were signiﬁcant predictors of

total soil C, accounting for 58 and 72 % of observed variability, respectively.

INTRODUCTION

Soil structure is an important crop production factor affecting numerous processes
critical to crop growth and ecological ﬁinction (Martin, 1953; Pachepsky and Rawls,
2003). Intensiﬁcation of cropping on Alﬁsols is becoming common in the Great Lakes
region, with high value, low residue crops grown quite frequently, and soil degradation is
becoming an urgent problem. Research is required to understand the role of root and
above-ground residues in promoting aggregation, C sequestration and soil structure
resilience (Rasse, et al., 2000) under continuous wetting and drying (Mikha, et al., 2005;
Park, et al., 2007), and freezing and thawing processes, in the face of frequent tillage and
harvest operations (Grandy et al., 2002). The ideal cover crop and alternate cash crop
rotation with potatoes would be able to maintain or enhance aggregate stability, soil C
and structural integrity.

Economic considerations and alternate land use could explain the increasing
preference for short crop rotation cycles among potato farmers. From 1950 to 1997 a
decline of 20% was observed in the amount of farmland hectarage in the US (Theolbald,
2001). Adoption of alternative crop management strategies becomes contingent on
observing improvements in soils over the shortest time period possible. A single
application of manure (24.5 Mg ha'l) in a Maine potato rotation trial resulted in
signiﬁcant formation of medium and large stable aggregates, whereas green manure crops
of oat and hairy vetch carried out over 5 years had no discemable effects on soil
aggregation (Grandy et al., 2002). Comparing a barley-forage rotation with a barley

monoculture, Bissonnette et al., (2001) noted a signiﬁcantly higher aggregation (as

measured by mean weight diameter) after 2 years in the rotation. Carter (1992) found
relatively rapid management-induced changes in the macro-aggregate size fraction > 1
mm in a tillage experiment.

Aggregates are formed by bridging actions among clay micelles, quartz surfaces,
organic colloids (Emerson, 1959), cations and anions (Hillel, 2004). The strength of these
bonds decreases as the size of the aggregate increases. Organic carbon (C) protection of
aggregates depends upon the degree of C encapsulation and subsequent contributions to
bonding strength, as well as the spatial location of C within the aggregate (Jastrow et al.,
1996). Carbon derived from recently added organic matter tends to be located on the
outer perimeter of aggregates (Kavdir and Smucker, 2005). Intra-aggregate locations of
soil C serve as hydrophobic deterrents to water dissolution of aggregates and contribute
to the binding of smaller aggregates into larger ones (Chenu et al., 2000). Therefore, the
effectiveness of different rotation cropping systems in improving soil structure can be
evaluated by looking at enhancements in the proportion of stable macro-aggregates.

The presence of legumes in a crop rotation scheme provides nitrogen (N)-
enriched residues, and this may support microbial activity that enhances polysaccharide
production and aggregation indirectly, but the mechanisms are not well deﬁned. Mazurak
et a1. (1954) found a high degree of aggregate stability in potato rotations that involved
alfalfa. Under laboratory conditions, alfalfa and clover inputs led to increased mean
weight diameter of aggregates in a silty-clay loam soil after only 9 days (Martens, 2000).
An equilibrium in the formation of new aggregate binding material may have occurred as
no further increases in aggregation were observed. Use of a combination of oat, peas, and

hairy vetch as a green manure resulted in a trend (P < 0.10) towards improved water

stable aggregate stability for a potato rotation system in Maine (Porter et al., 1999). In a
cotton production system the use of crimson clover improved aggregate stability by
100%, compared to a non-legume cotton system (Hubbs et al., 1998).

Winter cover crops are used by some producers of high value, irrigated crops such
as potatoes; however, farmers generally prefer cold-tolerant and productive cereal cover
crops in the Upper Midwest and rarely experiment with legume cover crops (Snapp et al.,
2005). A number of researchers have found that soil building and fertility is enhanced by
the presence of legumes and legume-cereal mixtures in row crop systems (Grifﬁn and
Hesterrnan, 1991; Honeycutt et al., 1996). In addition to a cover crop, the residues of the
main crop rotated with a potato crop are expected to inﬂuence soil aggregation and
carbon over time. Research has not fully resolved the question of relative beneﬁts of
quantity versus quality of residues, in terms of their effect on aggregate formation and
long-term soil productivity.

The objectives of this study were to 1) quantify above and below ground C inputs
from 7 potato systems with contrasting main crops (snap bean, corn and wheat) and three
winter cover crop management systems (fallow, rye, rye-hairy vetch, and red clover), and
2) determine the impact of the potato rotation systems and cropping system phase on the

formation of water stable aggregates and soil C dynamics.

MATERIALS AND METHODS

Experimental Design and Treatments

A trial was established April 2, 2001 at the Montcalm Research Farm (MRF;
85°10’32” longitude and 43°21 ’12” latitude; Montcalm County near Entrican, Michigan).
The soil type was a McBride sandy loam (coarse-loamy, mixed Eutric Glossoboralfs; and
coarse-loamy, mixed frigid Alﬁc Fragioorthods) (Ender, 2003). Table 1.1 shows soil
properties of the study site from samples collected and analyzed as described below.
Seven two year systems were evaluated, namely potato-bare, snapbean-bare (PBSBB,
SI); potato-rye, snapbean-rye (PRSBR, $2); potato-rye, sweetcom-bare (PRSCB, S3);
potato-wheat-wheat-rye (PWWR, S4); potato-wheat-wheat/clover-clover (PWWCC,SS);
potato-rye/vetch-snapbean-rye/vetch (PRVSBRV, S6); and potato-rye/vetch-sweetcom-
ryevetch (PRVSCRV, S7) (Table 1.2). The treatments were laid out in a randomized
complete block design with four replicates. To account for variability in seasonal
conditions, two entry-years were included for each system, doubling the total number of
systems present each year to 14. Plots were 5.49 m in length by 15.24 m in width.
Cultural practices and related ﬁeld operations for the main crops and cover crops of this
study followed Michigan State University recommendations, and were described in
Nyiraneza and Snapp (2007).

Monthly precipitation data recorded from the study area in 2001 to 2004 are
shown in Figure 1.1. The precipitation values of 991, 827, 478, and 735 mm for 2001,
2002, 2003 and 2004, respectively, indicate that 2003 was a dry growing season and

could potentially inﬂuence observed soil structure differences between initial

measurements taken in 2001, and those taken in 2004. However, precipitation was
supplemented with irrigation following standard potato production practice in Michigan,
diluting any moisture-related impact.

Carbon and Nitrogen Input

The annual amount of carbon that was incorporated into each of the systems in the
cropping cycle was calculated as the total oven-dry weight of above-ground biomass, less
any economic product that was removed. Above-ground biomass was collected using two
subsamples per plot from 0.25 m2 quadrats, oven dried at 65°C for 48 hours. Cover crop
biomass was measured just before incorporation on 25 April 2002, 23 May 2003 and 20
May 2004. Cash crop biomass was measured just before harvest on 11 July 2002, 24 July
2003 and 12 July 2004 for wheat; 26 August 2002, 20 August 2003, 17 August 2004 for
bean and corn and 3 Sept, 2002 for potato (only tubers were measured in 2003 and 2004
for potatoes, and the ratio of tubers to shoot used to estimate aboveground inputs).

To account for the below-ground C contribution, root biomass was measured for
cover crops in 2003 and 2004 by collecting a composite of ﬁve 2-cm diameter soil cores
per quadrat from the 0-25 cm depth (2 quadrats per replicate X 4 replicates). Soil samples
were wet sieved using a 4 mm mesh, roots were collected with tweezers and oven dried at
65°C for 48 to 72 hours (until no change in weight). This was expected to be an
underestimation of C inputs as root exudates and turnover of roots were not measured due
to methodological challenges. For snapbean (Phaseolus sp. L.)_. corn (Zea mays L.), and
wheat (Triticum aestivum L.), the shoot-root ratios used were set to 5:1, 5:1 and 4:1,
respectively (Bolinder et al., 1999). For potato, the ratio was 6.7: 1 , computed from Opena

and Porter (1999) and Porter et al. (1999). The average shootzroot ratio found here for

cover crops was 1.25:1, similar to the literature values for vegetatively growing cover
crops, such as a rye cover crop on a sandy soil in southern Michigan (Snapp et al., 2007).
To determine C inputs from total biomass a multiplier of 0.45 was used, based on average
C% in residues from this potato rotation trial (Nyiraneza and Snapp, 2007).

Water Stable Aggregates (W SA)

WSA samples were carefully collected using a trowel to preserve aggregates, 5
samples per composite representing a plot, from the 0-10 cm depth on the ﬁrst
(December 21, 2001) and fourth year (November 11, 2004) of the trial. Samples were
air dried, sieved through a nest of sieves to acquire a representative sample of soil
aggregates sized between 2 and 4 mm. Overall, approximately 80% of the soil
material by weight ended up in the less than 2 mm size range.

WSA was determined using a modiﬁcation of Rasse et al (2000) method by
placing approximately a single layer of 4 — 2 mm aggregates (~25g) on top of a nest
of 2, 1, 0.25, and 0.106 mm sieves in a water bath container attached to a
reciprocating machine with a vertical amplitude of S cm running at 35 rpm. Water
was then applied to the water bath until it contacted the 2 mm sieve facilitating
hydration of individual aggregates by capillary action for 10 minutes and reciprocated
for 10 minutes. Sand-free water stable aggregates were determined following the
methodology of Grandy et al (2002). For the rest of the text, referral to water stable
aggregate size fractions of 2, 1, 0.25, 0.106, and 0.053 mm would refer to aggregate
size ranges of< 4 to 2 mm, < 2 to 1 mm, < 1 to 0.25 mm, < 0.25 to 0.106 mm and <
0.106 to 0.053 mm, respectively. Data was also presented as stable macro-aggregates

(20.25 mm) and micro-aggregates (<0.25 mm). Mean Weight Diameter (MWD) was

computed as the summation of the average aggregate size remaining on each sieve,
multiplied by the percent of total sample represented by the respective aggregate class
as outlined by Kemper and Rosenau (1986).
Soil Properties

A representative soil sample collected above was sieved (< 2 mm) and ground to
powdery consistency using a Shatterbox Rotary Grinder (Spex Ind., Edison, NJ) for 2
minutes. Samples were prepared from the ground soil by weighing 60 to 70 micrograms
in a crimped tin capsules sent to the Stable Isotope Laboratory at the University of
California at Davis, and analyzed for total C and N by Europa Hydra 20/20 isotope ratio
mass spectrometer (Europa Scientiﬁc, Crewe, UK). Another representative subset of the
< 2 mm samples were sent to and analyzed by A&L Great Lakes laboratories (Fort
Wayne, IN) for chemical and texture analysis, including determination of available
phosphorous, exchangeable potassium, magnesium, calcium, soil pH, buffer pH, cation
exchange capacity (CEC), percent base saturation of cation elements and texture
components of sand, silt and clay (Ankerman and Large, 2001). Results of phosphorous
analyses were expressed as a weak Bray (P1) derived from Mehlich-3 through a
regression equation (Calhoun et al., 2002). Soils were sampled again on November 10,
2004 from the 0-20 cm depth using 2 cm diameter augers and 8 samples per composite to
represent a plot, to conduct a N mineralization potential aerobic assay. After wet-sieving
through a 6 mm sieve, a representative 10 g subsample was incubated at 60% water
holding capacity for 30 d at 25°C. NH4-N and NO3—N were determined in extracts by

colorimetric methods using an autoanalyzer (Lachat Instruments, Milwaukee, WI). Net N

mineralization was calculated using the difference between the inorganic N contents at
day 0 and at the end of the incubation period (N yiraneza and Snapp, 2007).
Statistical Analyses

Data analyses were conducted using SAS v9.1 (SAS 2006, Cary, NC) MIXED
procedure. Two-factor analyses of variance (ANOVA) were conducted with cropping
systems 81 through S7 as factor A, and the two entry years as factor B. The ﬁrst (2001)
and fourth (2004) year data were analyzed separately. The differences between 2001 and
2004 were calculated for each plot and also analyzed. Comparisons among the systems
and between the entry years were conducted using Fisher-protected least signiﬁcant
differences. The three groups, namely, low carbon input (LCI consisting of SI and S4),
medium carbon input (MCI consisting of S2 and S6), and high carbon input (HCI
consisting of S3, SS and S7), were compared via linear contrasts. Bonferroni adjustment
was used in testing signiﬁcance of the contrasts when the ANOVA was not signiﬁcant.

The contribution of water stable aggregates and biomass carbon input to soil C
prediction was evaluated through the sum of squares reduction test (Schabenberger et al.,
1999; Perez and Kogan, 2003). The test involved comparisons between the sums of
squares of single factor reduced model consisting of either water stable aggregates (i.e.,
macro or micro) or carbon input and the sum of squares of the full model involving all of

the afore-mentioned factors.

10

RESULTS AND DISCUSSION

Entry Year Effects

Incorporation of entry-year as a factor in the experimental design was effective in
quantifying the detrimental effects of potato production practices on soil quality. In 2001,
soil samples taken after the alternative cash crop of corn, wheat, or snap bean harvest had
58, 46 and 17% higher proportion of water stable aggregates in the larger size classes of 2
mm, 1 mm and the overall MWD, compared to aggregates after the potato main crop
(Table 1.3; P < 0.0001). A similar decline in aggregate size after potato harvest was
documented by Carter and Sanderson (2001).

Additional evidence of soil structure degradation with potato production was
documented in 2004 as alternative crops were associated with 11% compared to 9% for
the 1 mm size class aggregates after the potato crop (Table 1.4; P < 0.001). However, the
MWD was not different after the potato crop and the alternative main crops in 2004
(Table 4). After 4 years the alternative crops were not able to maintain the level of
aggregation and stability observed in year one of the study (Tables 1.3 and 1.4). Other
size fractions showing signiﬁcant effects of entry year in 2004 were the 0.25 mm and
0.053 mm WSA’s, with 10% reduction and 8% increase respectively for the alternative
cash crops over the potato entry year.

System Organic Inputs and Soil Response

The limited amount of biomass remaining in the ﬁeld in potato cropping systems

(Mazurak et al., 1954, Rees et al., 2002) necessitates the identiﬁcation of management

systems that enhance residue input and soil C assimilation from the residues. In our study

11

the average amount of C returned to the soil by the seven potato cropping systems was
2.08 Mg ha'1 per year (range of 0.93 Mg ha'I [SI] to 3.32 Mg ha'I per year [S7] - Figure
1.2A.). The amount of carbon contributed by S1 in this present study was 44% less than a
system consisting of continuous potato in a Canadian study (Angers et al., 1999).
Considering that S1 had a bare fallow, a low biomass legume as the alternative cash crop
(snap bean), and our site has low soil C (Table 1.1), all of these factors could have
contributed to lower growth and thus lower C input at our site than in the Canadian study.

The C input from systems had the following pattern of decreasing magnitude: S7
(potato-rye/vetch, sweetcom-rye/vetch) > S3 (potato-rye, sweetcom-bare) > SS (potato-
wheat/clover, wheat/clover-clover) > S2 (potato-rye, snapbean-rye) > S6 (potato-
rye/vetch, snapbean-rye/vetch), S4 (potato-wheat, wheat-rye) > S1 (potato-bare,
snapbean-bare). Planned contrast between low (S1 and S4), medium (S2 and S6) and
high (SS, S3, and S7) carbon input categories indicated the presence of highly signiﬁcant
differences, with low input having an average annual C of 1.164 Mg ha'1 while medium
and large inputs had 68% and 137% higher carbon returned to the soil, respectively
(Figure 1.2B).

Crop growth potential and residue generation has an impact on the amount of C
returned to the soil. All systems with corn had higher C contributions (S3 and S7),
followed by systems with wheat and rye (SS, S2, S6 and S4; Table 1.4). These results
were expected, considering that com has an efﬁcient C4 photosynthetic pathway, while
wheat together with rye are C3 plants, and research has indicated that com produces 70%

more biomass than wheat (Wilhelm et al., 2004).

12

The impact of a cropping system on soil C is dependent on many factors beyond
the quantity of C residues added, these include residue quality, the tillage intensity of
cropping systems and the soil environment. The proportion of labile to recalcitrant forms
of soil C is important, as annual contributions of residues have minimal impact on
systems dominated by high proportions of inert carbon (Paul, 2007); whereas in low C
soils, the addition of carbonaceous residues can increase total soil C in a relatively short
time frame (Shrestha, et al., 2002). Addition of 3.5 times the amount of carbon in S7
relative to SI did not, however, signiﬁcantly increase soil C in our experiment (Table
1.4).

As this was a cropping systems experiment, both the quantity and quality of
residues varied with treatment. Crop residues were mature with a low N concentration,
between 0.9% and 1.6% in 2002 (including snap bean residues). Cover crop residues
however were incorporated at vegetative stages and had higher N concentration,
particularly the legume-cereal mixtures. Weed biomass from S1 winter fallow was
determined using the cover crop protocol and was found to be <100 kg ha each year, so N
concentration was not determined for weed tissues. The rye cover crop in S2, S3 and S4
had a mean value of 2.2 N% (standard deviation 0.3), whereas the legume-cereal mixed
tissues in SS, S6 and S7 had a mean value of 2.7 N% (standard deviation 0.4). The
nitrogenous residues of S5, S6 and S7 could have a priming effect and been associated
with accelerated residue decomposition, resulting in limited soil C assimilation (Cope et
al., 1958; Rasse, et al., 2000). Presence of leguminous material ensures a relatively faster
biomass breakdown compared to biomass of wider C:N ratios. Inclusion of red clover in

the SS potato-wheat system showed 30% higher N mineralization potential (0.26 mg N g'

13

1 d'l) compared to S4 potato-wheat system (0.20 mg N g'1 d"; P = 0.05). No signiﬁcant
effects on N mineralization potential were found for other treatments, which averaged
0.25 mg N g“ d".

System Effects and Aggregation

After four years, the cropping systems signiﬁcantly affected the WSA size
fractions of 1 mm, >0.25 mm, >0.106 mm, >0.053 mm, and macroaggregates (20.25 mm)
(Table 1.4). Mean weight diameter (MWD) demonstrated improvements of 19.5 % in
2004 compared to 2001 levels, averaged across all treatments (Table 1.1). There was a
small change in MWD for the low C input systems, based on the planned comparison of
SI and S4 (low C inputs) with all other systems (Figure 3). 81 had a bare winter fallow
and demonstrated a trend consistent with no increase in MWD size after 3 years, although
this was not signiﬁcantly different than the other systems. It is interesting that S4 had low
tillage intensity, yet was still associated with a limited change in aggregate MWD; this
may be explained by the low C inputs associated with this system.

WSA increases above 2001 levels averaged 13%, 4% and 31% for aggregate
fractions 1, 0.106 and 0.053 mm, respectively (Table 1.1). In a Maine potato ﬁeld trial,
green manure involving hairy vetch increased the proportion of 1 to 2 mm, and 2 to 6.5
mm aggregate size fractions (Grandy et al., 2002), with no increase observed for small
macroaggregates (0.250 — 1 mm). The relatively high increase in the 0.053 mm size
fraction in this present study suggests an inherent weakness of aggregates formed by the
various systems. Continuous tillage of the soil may also have interfered with aggregate
formation. Angers et al. (1999) reported that increases in WSA were observed only after

the 6th year and not on the 10th year of a decade-long study. This discrepancy was

14

attributed to different initial moisture contents of the processed samples. The 10th year
samples were wetter than those analyzed during the 6‘h year. Samples in the present
Michigan study were air dried prior to wet sieving as opposed to ﬁeld moist samples used
by Angers et al. (1999), therefore initial soil moisture content appeared to have little
inﬂuence on the proportion of micro-aggregates observed among the seven treatments.

The amount of macroaggregates (diameter 2 0.25 mm) decreased for all
treatments after four years, except for S5. The 6% decline from an average of 55% in
2001 to 49% in 2004 for this present study was minimal compared to the decline in
macroaggregates measured in a potato rotation trial on Prince Edward Island, Canada.
Angers, et al. (1999) reported macro-aggregate percentages declined 20%, from 37% in
the third year to 17% in the 7th year. In their study, continuous potatoes for 9 consecutive
seasons showed the greatest decline in macroaggregates, as was observed in 81 in the
present study. The weak natural tendency by sandy soils to form larger aggregates during
constant soil tillage may have led to these macroaggregate declines. Under treatment SS
involving potato and wheat/clover species, the minimal tillage disruption compared to the
other rotation systems may have contributed to a modest 2% increase in macro-aggregate
development above the 2001 data, whereas the other systems experienced as much as a
15% decline. However, no signiﬁcant difference was observed on the delta macro-
aggregates among the seven systems (Figure 1.4).
Soil Aggregation and Carbon Dynamics

Relating the amount of total soil carbon to the amount of macro-, micro-
aggregation and residue contributed carbon input (Figure 1.5) showed an inconsistent

pattern of high macro-aggregation when total soil carbon was high as in S5 (with high

15

carbon input) while S4 also had high macro-aggregation, but the amount of carbon input
was not signiﬁcantly different from S1 (Figure 1.5). The high amount of aggregation in
S5 could be attributed to the reduce amount of tillage received as wheat was grown
across seasons and red clover being frost seeded in the spring (Snapp et al., 2004). Both
the carbon input and the aggregation contributed signiﬁcantly to total soil carbon level, as
determined based on the sum of square reduction test (Table 1.5). Residue C input was a
moderate predictor of total soil C (31% of variability explained), whereas macro and
micro WSAs were signiﬁcant predictors of total soil C, accounting for 58 and 72 % of
observed variability, respectively. Inclusion of macro-aggregation components and the
amount of carbon input in the full model was able to explain 83% of the observed

variability in total soil carbon.

16

CONCLUSION

The largest input of carbon was associated with the systems involving corn with
or without red clover showing three-fold higher C returned than the SI winter fallow
system. Overall, increasing C inputs through inclusion of wheat and red clover (SS) in a
rotation appeared to be an effective means to enhance macro-aggregation and improve
soil structure relative to SI. The difference between plots sampled for water stability
aggregation after potato and those sampled after an alternative cash crop, although still
signiﬁcant, declined from the 2001 level. Residue C input was a moderate predictor of
total soil C (31% of variability explained), whereas macro and micro WSAs were
signiﬁcant predictors of total soil C, accounting for 58 and 72 % of observed variability,
respectively. The three-parameter full model was able to account for 83% of the observed

variation in total soil carbon in 2004.

I7

REFERENCES

Angers, D. A., L. M. Edwards, J. B. Sanderson, and N. Bissonnette. 1999. Soil organic
matter quality and aggregate stability under eight potato cropping sequences in a
ﬁne sandy loam of Prince Edward Island. Can. J. Soil Sci. 79:411—417.

Ankerman, D., and R. Large. 2001. A&L Agricultural Laboratories agronomy handbook.
A&L Agric. Lab., Fort Wayne, IN.

Bissonnette, N., D. A. Angers, R. R. Simard, and J. Lafond. 2001. Interactive effects of
management practices on water-stable aggregation and organic matter of a Humic
Gleysol. Can. J. Soil Sci. 81:545—551.

Bolinder, M.A., D. A. Angers, M. Giroux, M. R. Laverdiere. 1999. Estimating C inputs
retained as soil organic matter from corn (Zea mays L.). Plant-and-soil. 215: 85—
91.

Bremner J M, C. S. Mulvaney C S. 1982. Total Nitrogen. In Methods of soil analysis:
Chemical and microbiological properties Part 2. Ed. A. L. Page et al. pp. 595-
694. AAmerican Society of Agronomy Inc., Soil Science Society of America, Inc.
Madison, Wisconsin, USA.

Calhoun, F.G., J .M. Bigham, and BK. Slater. 2002. Relationships among Plant Available
Phosphorus, Fertilizer Sales, and Water Quality in Northwestern Ohio. J. Environ.
Qual. 31 :38—46.

Carter, MR. 1992. Inﬂuence of reduced tillage systems on organic matter, microbial
biomass, macro-aggregate distribution and structural stability of the surface soil in
a humid climate Soil and Tillage Research. 23(4): 361-372.

Carter, M.R., and J.B. Sanderson. 2001. Inﬂuence of conservation tillage and rotation
length on potato productivity, tuber disease and soil quality parameters on a ﬁne
sandy loam in eastern Canada. Soil & Tillage Res. 63:1—13.

Chenu, C., Y. Le Bissonnais, and D.Arrouays. 2000. Organic matter inﬂuence on clay
wettability and soil aggregate stability. Soil Sci. Soc. Am. J. 64:1479-1486.

Chow, T. L., H. W. Rees, S. H. Fahmy, and J. O. Monteith. 2003. Effects of pulp ﬁbre on
soil physical properties and soil erosion under simulated rainfall. Can. J. Soil Sci.
83:109—119.

Cope, J. T., D. G. Stukkie, and A. E. Hiltbold. 1958. Effects of manure, vetch, and

commercial nitrogen on crop yields and carbon and nitrogen contents of a ﬁne
sandy loam over a 30-year period. Soil Sci. Soc. Am. Proc. 22:524-527.

18

Emerson, E. W. 1959. The structure of soil crumbs. J. Soil Sci. 10:235.

Ender, M. 2003. QTL analysis of genetic resistance to white mold (Sclerotinia
sclerotiorum) in common bean (Phaseolus vulgaris). Ph.D. diss. Michigan State
Univ., East Lansing.

Grandy, A.S., G.A. Porter, and M. S. Erich. 2002. Organic amendment and rotation crop

effects on the recovery of soil organic matter and aggregation in potato cropping
system. Soil Sci. Soc. Am. J. 66:1311-1319.

Grifﬁn, TS, and DB. Hestennan. 1991. Potato response to legume and fertilizer
nitrogen sources. Agron. J. 83:1004—1012.

Hillel, D. 2004. Introduction to Environmental Soil Physics. Academic Press.

Honeycutt, C.W.,W.M. Clapham, and SS. Leach. 1996. Crop rotation and nitrogen
fertilization effects on growth, yield and disease incidence in potato. Am. Potato
J. 73:45—61.

Hubbs, M. D., D. W. Reeves, and C. C. Mitchell. 1998. Measuring soil quality on the
‘old rotation’. p. 50-54. In T.C. Keisling, (ed) Proc. 2lst Annual Southern
Conservation Tillage Conference for Sustainable Agriculture. North Little Rock,
AR. 15-17 July 1998. Arkansas Agri. Exp. Stn. Special Report 186, Univ. of
Arkansas, Fayetteville, AR.

Jastrow, J. D., T. W. Boutton, and R. M. Miller. 1996. Carbon dynamics of aggregate-
associated organic matter estimated by carbon-13 natural abundance. Soil Sci.
Soc. Am. J. 60:801-807.

Kanll‘, Y. and A]. Smucker. 2005. Soil aggregate sequestration of cover crop root and
shoot-derived nitrogen. Plant and Soil 272: 263—276.

Kemper, W. D. and RC. Rosenau. 1986. Aggregate Stability and Size Distribution. p.
425-442. In A. Klute (ed). Methods of soil analysis: 1. Physical and mineralogical
methods 2nd ed. ASA and SSSA, Madison, Wisconsin.

Mikha, M.M., C.W. Rice, and GA. Milliken. 2005. Carbon and nitrogen mineralization
as affected by drying and wetting cycles. Soil Biol. Biochem. 37:339-347.

Martens, D. A. 2000. Plant residue biochemistry regulates soil carbon cycling and carbon
sequestration. Soil Biol. Biochem. 32:361-369.

Martin, P. 1953. Status Report on Soil Conditioning Chemicals. Soil Sci. Soc. Am. Proc.
17:1-9.

l9

Mazurak, A. P., V. T. Valassis, and L. C. Harris. 1954. Water-stability of aggregates
from potato plots as affected by different rotation systems under irrigation in
western Nebraska. Soil Sci. Soc. Am. Proc. 76:244-247.

Nyiraneza, J. and SS. Snapp. 2007. Integrated management of inorganic and organic
sources in an Alﬁsol enhance nitrogen efﬁciency and productivity. Soil Sci. Soc.
Am. J. 71 :xx-yy

Opena, G. B., and G. A. Porter. 1999. Soil management and supplemental irrigation
effects on potato: II. Root growth. Agron. J. 91 :426-431.

Pachepsky, Y. A., and W. J. Rawls. 2003. Soil structure and pedotransfer functions. Eur.
J.Soil Sci. 54:443—451.

Angers, DA, and MR. Carter. 1995. Aggregation and organic matter storage in cool,
humid agricultural soils. p.193-211. In M.R. Carter and BA. Stewart (ed.)
Structure and organic matter storage in agricultural soils. Advances in Soil
Science. CRC Press, Boca Raton, FL.

Park, E.J., W.J. Sul, and A.J.M. Smucker. 2007. Glucose additions to aggregates
subjected to drying and wetting cycles promote carbon sequestration and
aggregate stability. Soil Biol. Biochem. (accepted for publication).

Paul, EA. 2007. Soil microbiology and biochemistry. 3rd edition. Amsterdam, Academic
Press.

Perez, A. and M. Kogan. 2003. Glyphosate-resistant Loliul multiﬂorum in Chilean
orchards. Weed Science. 43: 12-19.

Porter, G. A., G. B. Opena, W. B. Bradbury, J. C. McBumie, and J. A. Sisson. 1999. Soil
management and supplemental irrigation effects on potato: 1. Soil properties,

tuber yield, and quality. Agron. J. 91 1416-425.

Rasse, D.P., A.J.M. Smucker, and D. Santos. 2000. Alfalfa root and shoot mulching
effects on soil hydraulic properties and aggregation. Soil Sci. Soc. Am. J. 64:725-
731

Rees, H. W., T. L. Chow, P. J. Loro, J. Lavoie, J. O. Monteith, and A. A. Blaauw. 2002.
Hay mulching to reduce runoff and soil loss under intensive potato production in

northwestern New Brunswick, Canada. Can. J. Soil Sci. 82: 249-258.

SAS Institute, Inc. 2006. Base SAS® 9.1.3 procedure’s guide, 2nd ed., Vols. 1-4. Cary,
NC.

Schabenberger, 0., BE. Tharp, J.J. Kells and D. Penner. 1999. Statistical tests for
hormesis and effective dosages in herbicide dose response. Agron. J. 91: 713-721.

20

Shrestha, R.K., J. K. Ladha, and R. D. Lefroy. 2002. Carbon management for
sustainability of an intensively managed rice-based cropping system. Biol. Fertil.
Soils. 36:215-223.

Snapp, S.S., K. Date, W. Kirk, K. O’Neil, A. Kremen and G. Bird. 2007. Root and shoot
tissues of Brassica juncea and Cereal secale promote a healthy potato rhizosphere.
Plant and Soil In press.

Snapp, 8.3., K. O’Neil, Quackenbush, R., and E. P0. 2004. Improving productivity and
soil quality in short potato rotations 2004 research and on-farm report. Southwest
MI Res. and Ext. Ctr. Michigan State Univ., MI.

Snapp, S. S., S. M. Swinton, R. Labarta, D. Mutch, J. R. Black, R. Leep, J. Nyiraneza,
and K. O’Neil. 2005. Evaluating Cover Crops for Beneﬁts, Costs and
Performance within Cropping System Niches. Agron. J. 97:322—332.

Theobald, D.M. 2001. Land-use dynamics beyond the American urban fringe. The
Geographical Review. 3: 544-564.

Wilhelm, W. W., J. M. F. Johnson, J. L. Hatﬁeld, W. B. Voorhees, and D. R. Linden.

2004. Crop and soil productivity response to corn residue removal: A literature
review. Agron. J. 96:1—17.

21

Table 1.1 Mean values and standard errors for soil properties in the potato rotation trial at
the Montcalm Research Farm, Entrican, Michigan.

 

 

Soil Properties 2001 2004
Elevation (m) 288.2i0.17
Texture (%)
Sand 75.4il.15
Silt 15.6i0.64

lay 8.9i0.62
Water stable aggregates (WSA, %)
Macroaagregates 55.2i2.28 48.6i1.08
< 4 — 2 mm 3.98i0.28 5.03i0.50
< 2 — 1 mm 9.10:0.43 10.3i0.45
< 1 — 0.25 mm 41.1il.07 36.1i0.69
< 0.250 — 0.106 mm 23.1i0.91 24.0i0.66
< 0.106 — 0.053 mm 18.3i0.48 24.1i0.59
Mean weight diameter (MWD,mm) o 77i0.02 0 92i0.04
Chemical profile (0-8 in depth)

Phosphorus (Brayll (ppm) 206i6.47 208i5.60
pH (buffer) 6.37i0.17 6.92i0.01
Calcium (ppm) 403i16.2 390i22.3
ccc (meq g*) 3.86:0.12 4.04:0.19
Potassium (ppm) 128i4.68 163:6.46
PH 5.75i0.16 6.13i0.04
Magnesium (ppm) 84.9:2.03 82.5:2.46
Soil organic carbon (%l O.68i0.04 O.78i0.06
Base saturation, %
Hydrogen 23.0i1.20 29.8i0.76
Potassium 7.94i0.30 10.8i0.42
Magnesium 17.5i0.63 18.0i0.62
Calcium 48.4i1.84 48.3i1.47

 

 

cuoO

 

ma ocm mmOPmuom umpwm Loum> Anew: + m>m cuou\oumuom >mum>mm mm

comm chm cmmm
ma Ugo mmoumDOm Hmpmm goum> huﬂmm + m>m mm:m\0umuom >mmm>mm mm

Loom: CH um>oHU omm Loﬂz omUmmmumpcﬂ

e some: .mmoomBOm emcee memes omm:3\oomoom omega mm
m moor: umuwm m>m .mmOUmoom mmowm ummcz ummL3\oumuom mzzm vm
Ha cuou umowm mumm .mmoomuom umuwm m>m cuoo\oumuom mommm mm

NH cmmm
cmmm chm ocm mmoumuom umuwm mum Qm:m\oumuom mmmmm mm

OH mem Gmmm
chm ocm Oompom umuwm Amumﬂv Boaamm chm\oomuom mmmmm Hm

AT.mHmmh N av ll
mGOﬂumuwmo aoaumuom

pawam unmawmmcmﬁ “duca3 mono can: awumhm OGHQQOHU

 

.=2
.SBCEm .Ecmm 583.8% EEoEoE a EoECmaxo BSoa Somm a S com: 35:53: no.5 8.50 SE? was 889? :orﬁom m; 033.

 

'Iable 1.3 h4ean \vater stable aggregate. niacroaggregates (hdacro-thg),t0tal 30H carbon (IS( L niean \veigln dianieter (h4\NT)) and Standard errors across diﬂereiu size
Classes and potato rotation systems in 2001 at Montcalm Research Farm, Entrican. Michigan.

 

 

 

 

Cropping Systemi Water Stable Aggregate Size Fractions, % TSC, % MWD,mm
>2 mm >1 mm >0.25 nmt >0.106 nmt >0.053 nmi Macro-Agg
51 PBSBB 3.3i0.6 9.1i1.1 42.5_+_1.7 24.5:1.7 19.7i0.8 53.1:3.4 0.63i0.06 0.77i0.06
S2 PRSBR 3.4i0.6 10.0i1.l 43.4i1.6 22.2i1.7 17.8i0.8 56.8i3.4 0.51i0.06 0.79i0.06
53 PRSCB 4.3i0.6 9.4i1.1 45.1i1.9 19.9i1.9 18.6i0.9 54.7i3.6 O.58i0.06 0.76i0.06
S4 PWWR 5.2i0.6 9.0i1.l 44.2i1.7 21.2i1.6 17.8i0.8 58.7i3.6 0.56i0.05 O.83i0.06
S5 PWWCC 3.4i0.7 9.0i1.1 42.2i1.6 23.5i1.7 19.0i0.8 55.3i3.4 0.63i0.06 0.78i0.06
56 PRVSBRV 3.7i0.6 8.1i1.1 45.1i1.7 22.2i1.6 19.5i0.8 58.1i3.6 O.61i0.06 0.81i0.06
S7 PRVSCRV 3.4i0.7 9.1i1.1 44.1i1.7 20.1i1.7 18.6i0.8 54.5i3.4 0.59i0.05 O.76i0.06
C Input LCI 4. i .5 9.0i0.8 43.3il.4 22.9i1.2 18.7i0.6 55.9i3.0 0.59i0.04 0.80i0.05
MCI 3.5i0.5 9.1i0.8 44.2i1.4 22.2il.2 18.6i0.6 57.5i3.0 0.56i0.04 O.80i0.05
HCI 3. _ .4 9.2i0.7 43.8i1.3 21.2il.1 18.7i0.5 54.9i2.8 O.60i0.04 O.77i0.04
Entry Year AP 3.0i0.4 7.4i0.7 44.6i1.3 22.6i1.05 I9.0i0.4 53.0i2.8 O.56i0.03 0.72i0.04
AACC 4.7i0.4 10.8i0.7 43.0i1.2 21.2i1.00 18.4i0.4 58.8i2.7 O.6li0.03 0.84i0.04
Analysis of Variance
Systeni ns: ns ns ns ns ns ns ns
HCI vs MCI ns ns ns ns ns ns ns ns
HCI vs LCI ns ns ns ns ns ns ns ns
MCI vs LCI ns ns ns ns ns ns ns ns
Entry Year *** *7kvk n3 IIS ITS *)<9< n5 ***
SystemTE—Entry Yea; ns ns ns ns ns ns ns ns
lSl PBSBB : potato—bare. snapbean—bare; S2 PRSBR : potato-rye,snapbean-rye; S3 PRSCB I potato-rye, sweetcorn-barc; S4 PWWR 7’ potato—wheat—wheat—rye; SS PWWCC > potato uh at ..'.. a: I . ‘1 \er; 86

PRVSBRV : potato-rye/vetch-snapbean—rye/vetch: S7 PRVSCRV : potato—rye/vetch—sweetcorn-ryevetch; HCl = high carbon input; MCI , medium carbon input; LC] : low carbon input; AP = after potato; AACC ~
after alternative cash crop. :ns not significant; * signiﬁcant at £0.05; ** signiﬁcant at $0.01; *** signiﬁcant at 30.0001

24

 

 

 

 

Shortpotaﬂ>rotantnisystcnisin 2004 athdorncalniliesearch Farninearlintncan,DAichigan.

 

Table I4 Mean water stable aggregate. macroaggregates (Macro—Agg). total soil carbon (TSC). mean weight diameter (MWD). carbon (C) input and standard errors for 7

 

Cropping SystemI Water Stable Aggregate Size Fractions, % TSC, % MWD, mm C Input,
>2 mm >1 mm >0.25 mm >0.106 mm >0.053 mm Macro-Agg Mgha'1
SI PBSBB 2.8i0.7 7.7i0.9 31.6i1.7 29.9i1.4 27.1i1.6 43.1i2.2 0.64i0.08 0.73i0.06 O.9i0.2
S2 PRSBR 3.7i0.6 8.0i0.9 35.8i1.7 24.1i1.5 27.3i1.6 47.5i2.2 0.61i0.07 0.81i0.05 2.0i0.2
53 PRSCB 4.0i0.7 11.7i0.9 36.6i1.7 20.1i1.5 22.7i1.6 55.8i2.2 0.68i0.07 O.84i0.06 2.8i0.2
S4 PWWR 4.2i0.7 12.6i0.9 37.0i1.7 20.9i1.4 23.7i1.6 55.4i2.2 0.63i0.07 0.88:0.06 1.4i0.2
55 PWWCC 5.1i0.6 12.8i0.9 35.1il.7 22.7i1.4 21.1il.6 56.2i2.2 0.77i0.08 0.86i0.05 2.2i0.2
S6 PRVSBRV 3.1i0.7 8.6i0.9 36.9i1 7 25.9i1.4 24.4i1.6 49.7i2.2 0.65i0.07 O.93i0.07 I.9i0.2
S7 PRVSCRV 4.2i0 7 11.0i0.9 39 7+ 7 22.2i1.4 22.1i1.6 55.7i2.2 O.66i0.07 O.88i0.06 3.3i0.2
C Input LCI 3.5t0 5 10.1i0 7 34. i 3 25.4i1.0 25.4i1.2 49.3i1.6 0.64i0.05 0.80:0.04 l.2i0.l
MCI 3.4- 5 08.3i0 7 36. 3 25.0i1.0 25.9i1.2 48.6t1.6 0.63i0.05 0.80i0.04 2.0i0.1
HCI 4.4, 4 11.8i0 6 37. 2 2 21.7i0.8 22.0, .1 55.9i1.3 0.70i0.05 0.86i0.03 2.8i .1
Entry Year AP 3.8i0.4 9.3i0.6 37.8i1 l 23.9i0.8 23.1i1.1 52.7i1.2 0.67i0.04 0.85i0.03 2.0i0.1
AACC 3.9i0.4 11.4i0.6 34.4il l 23.5i0.8 25.1i1.1 51 1i1.2 0.66i0.04 0.81i0.03 2.2i0.l
Analysis of Variance
System US: *** ** *** *J< tr* US US ***
HCI VS MCI I15 **~)< 1’18 * *‘k )(‘1' HS ITS **J<
HCI vs LCI ns * ns ** ** ** ns ns ***
MCI vs LCI ns * ns ns ns ns ns ns ***
Entry Year ns *** *** ns * ns ns ns ns
System x Entry Year ns ns Ns ns ns ns ns ns ns

 

'l‘Sl PBSBB : potato-bare, snapbean—bare: 82 PRSBR : potato-rye.snapbean-rye; S3 PRSCB : potato—rye, sweetcorn-bare; S4 PWWR 1 potato—wheat-wheat-rye; SS PWWCC : potato-wheat-wheat/clover—clover; S6
PRVSBRV : potato—rye/vetch—snapbean-rye/vctch; S7 PRVSCRV : potato-rye/vetch-sweetcom-ryevetch; HCl : high carbon input: MCI , medium carbon input; LCI ’ low carbon input; AP : after potato; AACC

after alternative cash crop. :ns not significant; * signiﬁcant at £0.05; ** signiﬁcant at $0.01; *** signiﬁcant at $0.000].

 

 

 

Table 1.5 Sums of squares reduction test of full and reduced model for total soil carbon
as a function of carbon input and proportion of water stable aggregates from a potato
short rotation cropping system study at the Montcalm Research Farm near Entrican,

 

 

 

Michigan.

Modelt R21 33 df MS Fob, 9

Full 0.83 1.56 37 0.04

Reduced
C input 0.31 6.36 42 0.15 22.81 0.01
Stable Macroaggregates 0.58 3.97 42 0.09 11.44 0.01
Stable Microaggregates 0.72 2.65 43 0.06 4.33 0.01

 

I Regression analyses with full model consisting of macro-, micro- aggregates and C input with reduced
model referring to single factor regression analyses involving each of the three independent factors; I R2 =
coefﬁcient of determination; SS = sum of squares; df = degrees of freedom; MS = mean square; Fob, =

observed F statistics; P = probability

.bo>coomm2 voom one Sam .83 Joom com SE mmk. can £5“ Kmm .50 we? cocﬁaﬁoﬁ 3:53.
Swab; Box-» 05 can .38 - Sow cow cmmEozz .cmoEcm So: Emu coaomom 818282 3 cosﬂamoocm .2558 go H z onE

Toomimi 88.8.1 meow -.+: 58+ REESE 88% @5395
5.6:

9.2 1:, :3. c3: .22 am: no“. :2.

con >oz poo new

 

O
on
d
J
m
2: I
d
T...
1?
a
1.
T.
a m
omp ..
I
I

room

 

«own

27

Annual Carbon Input. kgha"

Figure 1.2 Carbon inputs from 7 potato rotation systems (A) and averaged across systems
pro-planned contrasts, S1 and S4 for the low input category, S2 and S6 for the
medium category and S3, S5 and S7 for the high C input category (B). The potato
rotation systems are described in the text, and C inputs included above- and below-
ground organic material averaged over two years (i.e., 2003 and 2004). Bottom, middle
upper letters designate statistical signiﬁcance for main crop, cover crop, and total C
input across but not within systems. Columns with similar letters are not signiﬁcantly

for

and

3500-

3000 -

 

 

   
  
  

i

 

 

 

5?

§

2500*

2000‘

   

',:rl!

   

I
v

 
   
 
 

  

sit

    

 
 
 

Illlii
it

 

 

I!

 
 

If it

     

i“
nil"
i ll

   

- ,1

. a

     
   

Medum Input
Carbon Input

different at 5%.

28

A

O

L

 

 

A loan Weight [Jilin-tor, 'I.
b

 

Carbon Input

Figure 1.3 Inﬂuence of seven short rotation management options for potato production
systems (A) and low (LC, 1.2 Mgha"), medium (MC, 2.0 Mgha“), and high (HC, 2.8
Mgha") C input (B) on the percent change of mean weight diameter (MWD, mm) in
2004 compared to 2001 at the Montcalm Research Farm near Entrican, Michigan. Bars
with the same letter designation are not statistically different at S 5%.

29

axcm w
8 economic Emozmcﬁm co: o8 :ocmcwmmoc coco. 06% 05 53> mam .cmeoaz .cmombcm Soc Emu consumed 818282 05 cm Son 2
33950 voom E mocwwocwwmocomﬁ SSE mo omega “custom of co Amv SQE O A -mnwE Wm .UEV :wE use Assam—2 ON .03: BEBE
5.9%: N; .06 32 cam A<V mEocmzm cocoscoa 88cm com mcocao EoEowwceE couﬁoc tocm co>om mo 8:255 #4 23E

«:9: concco
or-

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

m?
n _ e-
n
. VI
m
or us. 0..
. _
, o
m
mEoﬂ>m 9.23050 350a om- V
. o—tl
9.
m m w- %
m
m m. n v-
1238-22,?» GVQSAOBEm avowed!» A>Emwl>mto Eaméru @3-9>Kv 65min! a

 

 

 

N
< _
or 05. 0.. . v

 

30

£896?
-EoocooBmtcoHoZobosﬁoQ H >m0m>~E hm m£30296-gonamcmtnocoiobocﬁoa n >mmm>~E cm uo>o.o-co>o_obmc:3-§o:3

-388 M 00325 mm moblaocsimoccsocsa H M325 em ”bantﬁoocooam 6.0-883 H momma mm 63-53%:mdbocﬁoq

n Mmmmm mm 639.8335 689-898 H mmmmm 5+ .e\em S “sesame bﬁocmzﬂm Ho: 8w B032 8:86 53> mEBmxm .mowmwocwwm

058m 883 E mcomtaanQ 9 been £834 .cwwanz dear—Em Soc 8.8m scammed 863202 65 3 33m 839? wcﬁaoco :ocﬂoc

tozm 988 a 80¢ C2386 88 36-8 mocewocwwmocome 293m holes» can .39: .888 6x. .8 cognac =8 38 mo SE m._ Semi
as. as

new a snuggle

m
0.. .(WLVV I
v 9. an
e. um - Z
t Wm
a em ’ A no
cvw . md
98 seam
‘
//

 

It

 

 
  

°o 1:) [‘08 1310.1.

 

 

 

 

31

CHAPTER II: THE RELATIONSHIP OF SOIL PHYSICAL

CHARACTERISTICS TO YIELD IN INTENSIVELY MANAGED FIELDS

ABSTRACT

Growers and crop management advisers need information on the major yield
determinants across the landscape in order to explain observed yield heterogeneity. A
study was conducted in two Michigan commercial ﬁelds, designated as ﬁelds I and 2 in
2003 and 2004, respectively, to evaluate the contribution of water stable aggregates
(WSA), soil nutrient status and crop spectral response to potato yield variability. The
study area was divided into grids with an effective area equal to 0.05 hectare. In 2003,
108 grid points were sampled from Field 1 to a depth of 10 cm for soil chemical
properties and WSA. Prior to potato harvest, plant spectral status was sampled using a
customized camera setup. The same was protocol was applied to the second ﬁeld but this
time only 100 grid points were sampled. Stepwise multiple regression was able to
account for 66% and 60% of the yield variance Field 1 and 2 respectively. Principal
component regression (PCR) analyses had a lower yield variance of 64% and 44% for the
same ﬁelds. PCR further conﬁrmed the signiﬁcant contribution of WSAs to potato yield
in both ﬁelds while spectral plant response was identiﬁed as a signiﬁcant principal
component for Field 1. The positive contribution of WSAs to yield indicates the need to
manage this aspect of the cropping environment, especially when the chemical
requirements of the plants have been comprehensively addressed. The positive role of

spectral measurement in predicting potato yield provides a useful monitoring tool.

32

INTRODUCTION

Farmers use various management tools to address pest, water, soil fertility and
related issues, but the underlying variability of soil physical properties are more
challenging to manage. Growing potatoes involves a host of variables that one has to
understand and properly manage to achieve consistently high yields. Proper fertilization
at the onset and during the cropping season; pest and disease management; irrigation
timing and schedule, are just a few of the complex factors that result in a bumper potato
harvest. Despite the wealth of information and expertise that go into producing a crop of
potatoes, variability in yield is still considerable (Verhagen et al., 1995).

Variation in growing conditions and plant response across a ﬁeld leads to
substantial within-ﬁeld yield differences. In the Netherlands, Verhagen et al. (1995) '
observed potato tuber yields across a ﬁeld ranged from 30 to 45 Mgha". At the Prince
Edward Island in Canada, Dehaan et al. (1999) reported an even greater potato yield
variability of 11 to S9 Mgha'l based on 7325 observations on an 18 ha ﬁeld. Cambouris
et al. (2006) identiﬁed two management zones within a ﬁeld with 5.9 Mgha'l potato yield
differences. Variability caused by erratic water management could mean potato yield
difference of 2 Mgha'l (Balchin, 1958). In other crops, within-ﬁeld variability for corn
(Zea mays L.) can range from 3.5 to 8.5 Mgha'l or higher an in wheat (Triticum aestivum
L.), yield can range from 2 to 6 Mgha'I (Mulla and Schepers, 1997).

A number of research strategies are available to deal with yield variability, and
usually involve multiple year data analysis. There are many studies dealing with yield

variability from the point of view of identifying and preserving observed stable patterns

33

of yield variability and managing identiﬁed zones, accordingly (Lark and Stafford, 1998;
Taylor et al., 2001; Dobermann et al., 2003; Pringle et al., 2003; Zaman and Schumann,
2006). Cambouris et al. (2006) used an electrical conductivity meter to identify potato
management zone within a ﬁeld that apparently looked uniform, but had considerable
yield variability of 5 Mgha'l. These studies do not attempt to achieve the same level of
productivity across the entire ﬁeld. On the other hand, studies exist that focus their
analyses across the landscape spatial variability without forming management zones
(deHaan et al., 1999; Kravchenko et al., 2005; Pettersson et al., 2006).

There are tremendous numbers of potentially important yield determinants.
Classifying factors of production into physical and chemical factors could allow the
complexity of crop production to appear more manageable from a producer's point of
view. The importance of various nutrients to growth and performance of crops is well
recognized. Nutrient status is a major production factor, and is usually the ﬁrst to be
addressed in a commercial ﬁeld setting, as crops respond markedly to nutrient inputs.
Despite such efforts to address all nutritional requirements of crops through best
management practices, considerable variation still exists in terms of distribution of yield
across the landscape.

Irmak et a1. (2001) used a decision support system CROPGRO-soybean to
improve yield prediction over a ﬁeld divided into 60 grids. This was a data intensive
model requiring information relative to climatic, basic soil information and chemical
nutrition, yet correlation between predicted and actual yield was very low at 16%,
averaged across the two years covered by the study. Analysis conducted on a per year

basis even showed a dismal I to 2% correlation coefﬁcient. To improve the regression

34

coefﬁcient, the authors classiﬁed the grid locations into the following categories: with
and without drainage problems; greater and lower than 2% slope; and a classiﬁcation
scheme based on the moisture retention capacity of soil. Impact of variable physical
characteristics of the growing environment was very apparent, with correlation
coefﬁcient ﬂuctuating from 90% and above, to 10%. Redulla et al. (2002) pointed out the
prominent inﬂuence of soil texture on potato yield compared to soil chemical properties.
Mackenzie et al. (2000) noted the effect of inadequate moisture on speciﬁc gravity
decrease and larger-sized of potatoes than when moisture was adequate. Paz et al. (1998)
indicated that as high as 69% of soybean yield variability was due to water stress.

Soil structure is an important link in the functioning of the soil-plant-atmosphere
continuum, and ultimately impacts yield performance of crops. Cambardella et al. (1996)
reported on the relationship between speciﬁc soil physical properties and yield variability
in a com-soybean rotation ﬁeld. Aggregate size distribution contributed signiﬁcantly as a
predictor of yield in ﬁve of the seven years of the study. The relationship with yield was
not consistent, with 3 out of the 5 years being positive, while negative for the remaining 2
years. An effective way to put these results in perspective is to look at the effect of
weather. However, details provided by the study on weather behavior were not available.
Arshad and Coen (1992) proposed aggregate stability as one soil physical property that
can serve as an indicator of soil quality status. This proposition is not difﬁcult to
understand, considering soil structure has an important role to play in such vital crop
requirements as moisture availability and aeration, as well as ionic exchange in the soil

colloid.

35

Growers and crop management advisers need improved information on the major
yield determinants across the landscape in order to explain observed yield heterogeneity.
For example, how much predictive value do soil nutrients have, as opposed to soil
physical and biological factors, in terms of the ﬁnal tuber yield and quality? Redulla et al.
(2002) examined the ﬁeld scale variability of soil physical and chemical factors and its
effect on potato yield variability of four ﬁelds in Southeastern, Washington. The Pearson
correlation coefﬁcient carried out between yield and soil properties resulted in the
majority of the pair-wise combinations not being statistically signiﬁcant. Point yield data
was signiﬁcantly correlated with sand and silt in only one ﬁeld. Yield in another ﬁeld
was signiﬁcantly correlated with clay, another component of texture. Stepwise regression
conducted on each of the four ﬁelds produced unique predictive equations. No uniform
set of variables was found to predict point yield, total count, and speciﬁc gravity across
the ﬁelds. In those four ﬁelds, soil texture components were the predominant predictor
variables compared to nutrient levels, with clay having a positive contribution, and sand,
a negative one. The authors surmised that available soil water content was inﬂuenced by
texture, and this could be the driving factor that accounted for soil texture effects on
potato yield, even in these irri gated systems (Redulla et al., 2002).

Topographic features may be closely related to the heterogeneous distribution of
factors of production. Potato management zones identiﬁed by Cambouris et al. (2006)
were based on a collection of soil water regime parameters including water holding
capacity, thickness of surface sandy soil and depth of water table. Also in potato, deHaan
et al. (1999) found a signiﬁcant reduction in yield, higher incidence of smaller tubers, as

well as loss of soil particles from areas where there was a considerable drop in elevation

36

compared to other areas, in a study conducted at Prince Edward Island in Canada. The
authors contend that elevation was not the direct causative factor for the observed
reduction in yield, but rather the associated soil degradation brought about by erosion.
Phiri et al. (1999) reported a minimal landscape position effect on nitrate content of the
soil in a study involving corn in a Southern Africa's steep watershed. Kravchenko et al.
(2000) noted the higher accumulation of organic matter in low-lying areas compared to
relatively higher positions in the landscape for a ﬁeld under com-soybean rotation.
Organic matter and phosphorus concentration were negatively correlated with elevation
of an area characterized as mixed tall-short grass prairie (Ovalles and Collins, 1986;
Aguilar and Heil, 1988).

Traditional soil sampling calls for a limited amount of samples mainly due to cost
and time constraints. Availability and utilization of technology that can increase the
number of samples if it is feasible and economic to take, will minimize interpolation error
between points, and enable a closer examination of the relationship with yield, as
generated by geo-position linked yield monitors. Apparent electrical conductivity (ECa)
meter and remote sensing are potentially two such technologies. Rhoades et al. (1989)
indicated three ways with which electrical conductivity values measured by an EC
instrument was made possible. Electrical current can ﬂow through the hygroscopically
held water molecule around the soil particles and a couple of molecular thickness beyond.
An electrical charge can be initiated through a continuous aqueous soil solution, as is the
case when EC is measured in the laboratory. The third pathway, in the absence of strong
hygroscopic properties (as may be the case in sandy soil), electrical conductance may still

be possible through movement of electrical charge over soil particles in direct contact

37

with each other. Rhoades et al. (1989) presented ECa as the summation of the apparent
electrical conductivity contributed by exchangeable cations present on, or integrated into
the clay mineral structure in addition to metallic minerals, a transmission coefﬁcient, the
electrical conductivity of the soil solution and the soil volumetric water content.

Kitchen et al. (2003) employed four analytical methodologies to determine the
role of apparent electrical conductivity (ECa) and topographic features that contributed to
yield variability in com, soybean and wheat. All four approaches demonstrated that ECa
and topographic soil features are important in accounting for yield variability. However,
the authors pointed out that in analyzing speciﬁc site year data, it is important to take into
consideration climate, crop type and speciﬁc ﬁeld characteristics to interpret yield and
electrical conductivity relationships. Kravchenko et al. (2003) used a similar electrical
conductivity sensor (Veris 3100), and found a strong spatial correlation between
electrical conductivity and soybean yield in Michigan. Cambouris et al. (1999) used a
geonomic electrical conductivity meter to probe a uniform ﬁeld for differences in soil
moisture regime and successfully identiﬁed two distinct management zones with varying
potato yield potential.

The use of remote sensing provides another source of ﬁne resolution data to
characterize yield potential. Hunt et al. (2005) used a consumer grade digital camera and
had variable success in relating biomass of corn, soybean and alfalfa to a normalized
green-red difference index with computed regression coefﬁcients of 88%, 39% and 47%,
respectively. Adamsen et al. (1999) conducted research in wheat and obtained a
correlation coefﬁcient of 96% with green-red ratio computed from digital imagery, with

the more traditional measure of vegetation, namely normalized difference vegetation

38

index (NDVI). The green-red ratio was also correlated (91%) with a leaf greenness
instrument, Minolta-SPAD meter (Ramsey, NJ).

The objectives of this study were to quantify the relative importance of chemical
and physical factors in predicting potato tuber yield. In particular, the contribution of soil
structure attributes of water stable aggregate and mean weight diameter, digital spectral

characterization and nutrient status were related to potato yield variation.

39

MATERIALS AND METHODS

Location

The study was conducted on two commercial potato production ﬁelds managed
by the same farm operator and designated as ﬁelds 1 and 2, and sampled 2003 and 2004,
respectively. The ﬁrst ﬁeld was a 43.5 hectares commercial production ﬁeld that had
been in a potato-based rotation for the past 60 years. Rotation crops included corn (Zea
mays L.), dry beans (Phaseolus sp. L.), and wheat (T riticum aestivum L.) Most recently,
the rotations of potato (Solanum tuberosum L.) cultivars grown in this ﬁeld included Pike
with cereal rye (Secale cereale L.) as a cover crop in the winter in 2001, and FL 1833
followed by wheat and rye as cover crops in 2003. It is located at 84°57' longitude and
43°26' latitude, Vestaburg, Montcalm County, Michigan, USA. The second ﬁeld was a
20.7 hectares commercial ﬁeld located at 84°52' longitude and 43°26' latitude. For the
past 30 years crop rotation practices were similar to those of the ﬁrst ﬁeld. When this
ﬁeld was sampled in 2004, it was planted with the FL 1879 potato cultivar and a rye
winter cover crop. Two years before, it was planted with Pike and a rye winter cover crop
followed by snap beans with wheat as winter cover crop the following year.
Climatic and Soil Conditions

Figure 2.1 shows 6 out of the 9 months of active crop growth period having a
higher precipitation in 2004 than in 2003. However, beginning July and up to September,
the amount of precipitation in 2004 was lower than in 2003. For both years, precipitation

in this time frame was below the 135- and 8- year average. A center pivot irrigation

40

system covered each ﬁeld, rotating 3600 in 24 hour and applying ~1.9 cm of moisture
each rotation.

Both ﬁelds were composed of Mancelona loamy sand soil series with 0-2% slope
(Mb) (Figures 2.2 and 2.3). A small section of ﬁeld 1 was a variant of the Mancelona
loamy sand (Mc, 2-6% slope), as well as Newaygo sandy learn (Nm, 2 to 6% slopes),
although this particular soil series was not present in the monitored section of the ﬁeld.
About three quarters of the ﬁeld 2 ﬁeld was composed of a complex of Gladwin loamy
sand and Pale sandy loam (Ga), with the remainder being Mb. Outside the study area, but
still within the ﬁeld’s boundary, was an Epoufette loamy sand and Ronald sandy learn
(Ec, 0-2%). Mb and Mc were characterized as well-drained, while Ga was somewhat
poorly drained (Schneider, 1960).

Overall soil characteristics showed that ﬁeld 2 had 50% more organic carbon and
44, 32, and 16% less potassium, magnesium and Bray-1 phosphorous, respectively, than
the ﬁrst ﬁeld. Field 2 had 8% more sand particles (823gkg'l), 18% lower amount of clay
(57.4 gkg'l) and 28% less silt (119.6 gkg'l), as well as higher amounts of
macroaggregates (20.25 mm aggregate size), 1 and 2 mm water stable aggregate
components, compared to ﬁeld 1 (Table 1). Average volumetric moisture content across
the grids was higher in in the ﬁrst ﬁeld (15.8%) than in the second ﬁeld (11.51%).
Sampling Design

Historical soil sample characterization data taken from the study area in previous
cropping seasons were used to formulate the sampling design. For ﬁeld I, the average
ratio of samples per hectare used historically by crop advisers was 1.7:] for the

determination of percent organic matter, phosphorus, potassium, magnesium, calcium,

41

pH, cation exchange capacity (CEC), and zinc. The study area was located at the center
of the ﬁeld where the slope was minimal, to minimize the effect of topography on the
potato yield. Prior geographically referenced sampling points for ﬁeld 2 had a ratio of 2.9
samples for every hectare (Otto, 2004 personal communication).

Field 1(Figure 2.2) was divided into 108 sections, each consisting of 16 rows
(spaced at 0.89 m apart) on the nerthem and southern sides, and approximately 9 m both
on the east and west sides from the center of the section, making the effective area equal
to 0.05 hectare. Thus, the ratio of sampling points per hectare was 20:1, compared to the
earlier ratio of 1.721. In a similar manner, a total of a hundred grid points were sampled
from ﬁeld 2 (Figure 2.3).

Sampling for Water Stable Aggregates

Soil samples for water stable aggregate analyses were taken a couple of days
before harvest on September 18, 2003 and September 11, 2004 for ﬁedds 1 and 2,
respectively. The composite samples were taken from a 4 meter radius around each grid
point to a depth of 10 cm using a trowel. Care was exercised not to unduly compact the
soil, and to protect the samples, cardboard boxes were used in transporting and storing
them. Extraneeus materials such as plant residues and gravel from the soil aggregate
samples were removed with the use of an 8 mm sieve, and the samples air-dried for a
number of days until they equilibrated with atmospheric water content. The soil sample
was sieved using a layer of sieves, with the top having a screen opening of 8, 6.2, 4, 2
mm and a catch pan at the bottom. Rocks, gravel and extraneous materials remaining on
the 8 mm were discarded. Soil aggregates that remained on top of the 6.2, the 4 mm, the 2

mm and those contained in the catch pan were placed in separate containers and labeled

42

as the 6.2-4.0 mm, 4.0-2.0 mm and < 2 mm sieve sizes. Results reported here referred to
water stable aggregate stability of the 4.0 to 2.0 mm size fraction. Water stable aggregate
analysis was previously described in P0 (2007).

Chemical and Mechanical Analyses

A representative subset of the samples that passed through the 2 mm sieve during
WSA sample preparation were ground to powdery consistency for 2 minutes using a
Shatterbox Rotary Grinder (Spex Ind., Edison, NJ). Samples for total carbon and nitrogen
analysis were prepared from the ground soil by weighing 60 to 70 micrograms to a tin
capsule. Crimped tin capsules were submitted to the Stable Isotope Laboratory at the
University of Califemia at Davis for total Carbon and Nitrogen analyses using a Europa
Hydra 20/20 isotope ratio mass spectrometer (Europa Scientiﬁc, Crewe, UK).

Another representative subset of the < 2 mm samples were sent to and analyzed
by A&L Great Lakes laboratories (Fort Wayne, IN) for chemical and texture analysis,
including determination of available phosphorous, exchangeable potassium, magnesium,
calcium, soil pH, buffer pH, cation exchange capacity (CEC), percent base saturation of
cation elements and texture components of sand, silt and clay. Results of phosphorous
analyses were expressed as a weak Bray (P1) derived from Mehlich-3 through a
regression equation (Calhoun et al., 2002).

In-situ Measurement

The volumetric soil water contents were measured from ﬁeld 1 and 2 prior to soil
sampling on September 17, 2003 and September 10, 2004, respectively using a Trime
time domain reﬂectemeter moisture probe (Ettlingen, Germany). Five measurements

were taken from each grid points and the average per grid point calculated to represent

43

soil moisture status for each measurement date. Relative soil health was measured using
the Veris mobile sensor platform (Veris Technology, Salina, KS) by apparent electrical
conductivity (ECa) and pH during the spring of 2004 (April 20, 2004) in for ﬁeld 2.
Brieﬂy, an electrical pulse is released to the soil through a pair of rotating discs separated
by ~57 cm that penetrates the soil to a depth of 6 cm. Voltage drops in the electrical
pulse traveling through the soil matrix were measured, and ECa computed to effective

depths of 30 cm (Sudduth et al., 2003).

Bulk density and Inﬁltration

Laboratory measurements were conducted on soil cores obtained using an
aluminum cylinder 7.62 cm in diameter and 7.62 cm inches in height, for bulk density
(Blake and Hartge, 1986) and inﬁltration (Li et al., 2005). The cylinders were randomly
installed within 6 transects running from East to West in ﬁeld 1. A mini-disc inﬁltremeter
from Decagon (Pullman, Washington) with a 6.0 cm suction was used to calculate the
unsaturated hydraulic conductivity at each of the grid point where the cylinders were
collected. Calculation for unsaturated hydraulic conductivity was adopted from Zhang
(1997) where the square root of time was obtained for each set of inﬁltration readings,
and then plotted on the x-axis with volume of water inﬁltrated on the y-axis. The
resulting point distribution was ﬁtted into a second order polynomial and a trend line
equation projected using Excel v7 (Microsoft, Redmond, Washington). The coefﬁcient of
the squared independent variable (square root of time) was utilized to compute the
hydraulic conductivity of the soil. After the hydraulic conductivity measurements were
completed, the soil cores were dried in a convection oven for 24 hours, weighed and bulk

density calculated using the core method outlined by Blake and Hartge (1986).

44

Spectral Pictures

An Olympus 340R (Melville, New York) digital camera was utilized to obtain
red, green, and infrared images at each of the grid points in ﬁeld I and 2 on August 31,
2003 and August 28, 2004, respectively. Two sets of images were taken per gridpoint,
where each consisting of a normal, automatically obtained image and an infrared image
using a pass through-ﬁlter. Using the camera without a ﬁlter in front of it resulted in a
picture with pixels composed of a combination of the red, green and blue parts of the
spectrum. Te extract the green, red and infrared parts of the spectrum, Paint Shop Pro 8
(Ottawa, Ontario, Canada) was used along with a macro program to facilitate the
processing of hundreds of pictures in a short time. The extracted green, red and infrared
band images were used as input to IDRISI for windows v. 1 (Worcester, Ma) to compute

the mean average value of the image pixels (Figure 2.4).

Yield Monitor Data Collection

Potatoes were dug from the ﬁeld using 1 8-row and 2 4-rows digger. A harvester
equipped with an electronic yield monitor (Harvest Master Yield Mapping System for
Bulk Crops, HM-500; HarvestMaster, Logan, UT) logging data every two meters (in ﬁeld
I) and every four meters in ﬁeld 2 on average, followed the tandem of diggers and loaded
tubers into trucks. A total of 1071 datapoints were generated for ﬁeld land roughly half
that quantity of 633 for ﬁeld 2. A tandem of 1 eight row and 2 four row diggers can scoop
up 16 rows as it moves from east to west and vice-versa. Therefore, for each grid point,
two passes of the harvester were required to cover the total of 32 rows representing each

ofthe 108 grids in ﬁeld 1 and 100 grids in ﬁeld 2.

45

Statistical Analysis

To test the hypothesis that physical factors explain more of the observed
variability in potato yield than chemical nutrient level in a commercially run operation,
correlations and multiple stepwise regressions were carried out on all soil variable data
collected as independent factors, and yield as the dependent factor. To minimize
multicollinearity, iterative sequences of computations were conducted through Proc Reg
of SAS v.9.1 (SAS 2006, Cary, NC) with the variance inﬂation factor (VIF) option
selected. At each loop, the variance inﬂation factor (VIF) was monitored and any
assumed predictive variable with a VIF of more than 10, an indicator of multicollinearity
(Gunst and Mason, 1980; Neter et al., 1989; Hair et al., 1995), was removed from the
succeeding computations. The iteration stopped when all assumed predictive variables
had VIF’s of less than 10. As an alternative approach to handling multi-collinearity,
principal component regression was performed using Proc Princomp of SAS.

The signiﬁcant contribution of physical, chemical and plant response was
evaluated through the sum of squares reduction test (Schabenberger et al., 1999; Perez
and Kogan, 2003). The ratios of the difference between the sum of squares of the reduced
model consisting of single factors of physical, chemical and plant response and the full
model involving all of the afore-mentioned factors divided by the difference in the
degrees of freedom and the mean square of the reduced model was used as observed F,
and compared to tabular F. Presence of a signiﬁcant observed F indicated the importance

of the full over the reduced model.

46

RESULTS AND DISCUSSION

Spectral Response

The physiological state of the potato crop can impact its interaction with light
energy. This interaction provides a feedback mechanism that moderates plant vigor and
acts as an independent parameter that inﬂuences potato yield. Although digital images
were taken at approximately the same time - August 28, 2004 and August 31, 2003 for
ﬁeld 1 and 2 respectively, ﬁeld 1 had more reﬂectance in both the unadjusted and
adjusted images for the green band (15 and 91%), compared with ﬁeld 2. On the other
hand, the amount of red and infrared reﬂectance was higher in ﬁeld 2 than in ﬁeld I for
the unadjusted images (92 and 387%, respectively). The higher level of red and infrared
in ﬁeld 2 could be attributed to a high proportion of non-vegetated area, with ﬁeld I
having 50% vegetation cover compared with 5% in the second ﬁeld (Table 2.1). The
seeming discrepancy in the vegetation cover between the two ﬁelds may be due to
differences in timing of planting and potato senescence. Field 2 was planted nearly two
weeks earlier (April 29, 2004 vs May 13, 2003) and harvested over a week later than the
other ﬁeld (October 3, 2004 vs September 24, 2003), and had less than ideal soil moisture
status towards the latter part of the growing season compared to ﬁeld I, which may have
accelerated senescence (Figure 2.1). The increased reﬂectance of the red portion of the
visible spectrum was expected. As plant integrity declines, so does the absorptive
potential of the biologically active far red phytochrome (PFR), as well as those of the
chlorophyll, as senescence exponentially increase over time (Lamb, 2000). The

considerable variation in the infrared values between ﬁeld 1 and 2 could be due to a

47

different cut-off ﬁlter used in ﬁeld 1, with a higher limit of 0.9 nm compared to 0.7 nm in
ﬁeld 2. With vigorous vegetation, infrared light reﬂection should be high and a
corresponding decrease observed as senescence started to set-in (Lamb, 2000). However,
no comparative infrared values were taken over the same infrared ﬁlter cut-off in both
ﬁelds and therefore interpretation was only conﬁned to relative differences among the
grid points in ﬁeld 1 and 2 at the end of bulking stage for this particular spectral band.
Average infrared values from digital orthophoto were taken at two different dates (April
13, 1999 [ﬁeld 2]; April 4, 1998 [ﬁeld 1]), yet showed a minimal difference between the
two ﬁelds.

Vegetation indices utilizing distinct bands of the electromagnetic radiation spectra
have been used to indirectly evaluate plant health and vigor (Seidl et al., 2004). In the
absence of a sophisticated spectral measuring device, the green band of a digital image
had been utilized as a proxy for infrared (Hunt et al., 2005), since both are strongly
reﬂected by a healthy plant. However, correlation analysis by Yang et al. (2000)
indicated a weak and variable correspondence between infrared and green spectral bands,
with the highest value obtained at 69DBH (49%) of grain sorghum and declining
thereafter to as low as —19% 14DBH.

As vegetation senesces and dries up, the red part of the electro-magnetic spectrum
is no longer absorbed, causing an increase in the reﬂectance of this spectrum to be picked
up by digital imagery. Adamsen et al. (1999) showed 10 to 30% more green light than red
light being picked up by a digital camera at 41 days before harvest (DBH) in wheat. At

the end of the image acquisition period (9DBH), the amount of red was equivalent, if not

48

greater than that of the green light. In both ﬁeld 1 and 2, green and green by red ratio,
respectively, positively explained potato yield.

Field 1, with a signiﬁcantly higher vegetative cover, had a 124% higher Green by
Red ratio compared to ﬁeld 2 (Table 2.1). At this late stage in the growing period (tuber
maturation phase, 24DBH/110DAP [ﬁeld 1]; 36DBH/12IDAP [ﬁeld 2]), the presence of
actively growing vegetative parts could be disadvantageous to yield, since it could divert
photosynthates from reallocating to the tuber from the shoot. However, in this case of
ﬁeld 1, the contribution of the unadjusted green band to yield prediction was positive
(Figure 2.5). This indicated the presence of high vegetation cover (Table 2.1) but not as a
result of new vegetative growth; rather, it was vegetation at its peak of photosynthate
production capability.
Yield Predictors

To allow meaningful interpretation of multiple regression equations on the
behavior of a dependent variable as a function of independent variables, each explanatory
variable must not be collinear or interdependent (Draper and Smith, 1981). Repetitive
removal of base predictive variables with unacceptable variance inﬂation factor (VIF) is a
means of addressing this collinearity challenge. VIF is an indicator of severe
multicollinearity if more than 10 (Gunst and Mason, 1981; Neter, et al., 1989; Hair et al.,
1995). In earlier studies, use of VIF has been used to reduce the total number of assumed
predictive variables by 50% (Kok and Veldkamp, 2001; Grifﬁth et al., 2002). There were
15 common variables in both ﬁeld I and ﬁeld 2 that did not contribute to
multicollinearity. A total of 23 predictor variables were selected from the original pool of

42 for ﬁeld 1, and 21 for ﬁeld 2 from a pool of 40. After the initial stepwise regression

49

analysis was conducted on the selected pool of predictor variables for each ﬁeld, the
analysis was repeated for a second time to determine if additional variables would
contribute signiﬁcantly to yield prediction. These additional predictor variables were
hypothesized to be important to yield prediction, and consisted of bulk density and
inﬁltration rate for ﬁeld 1, and apparent electrical conductivity for ﬁeld 2.

The ﬁnal stepwise regression equation for ﬁeld I explained 67% of potato yield
variability. Physical, chemical and spectral variables included in the equation were
aggregate stability (mean weight diameter), microelevation, soil texture (the amount of
clay), the unadjusted green band, and potassium level (Figure 2.5). Inclusion of
inﬁltration rate and bulk density in the pool of predictive variables did not affect the
variability explained, and neither one of these variables was included in the ﬁnal
regression equation.

A different situation existed for ﬁeld 2. Initially, the stepwise regression included
mean weight diameter, the red unadjusted band of the digital images, and the base
saturation portion of hydrogen all accounting for 44% of the observed potato yield
variability. Inclusion of ECa not only increased the explained variability to 60% (Figure
2.6), but it removed all of the identiﬁed yield predictors and replaced it with the green
and red unadjusted band ratio, and 0.25 mm water stable aggregate size fraction in the
predictive equation (Figure 2.6).

The inclusion of two different proxies of soil structure (i.e., mean weight diameter
and 250 mm water stable aggregate) in the yield predictive equation at both ﬁelds, show
the signiﬁcant contribution of soil structure to the positive dynamics of soil, plant and

environment continuum, resulting in enhanced potato yield. It is interesting that

50

parameter estimates for both soil structure proxies were positive. Field 2 was composed
of a soil series that were prone to drainage problems, and 2004 was a relatively wet year
(compared to 2003), yet moisture retention may have been a problem, as it was strongly
correlated with yield (Table 2.2). On the other hand, ﬁeld 1 had a higher clay content and
soil structure (i.e., mean weight diameter) was able to improve moisture ﬂow,
complementing the positive contribution of clay to adsorb water molecules. The year
2004 was associated with reduced precipitation, up to 50% of the long-term average for
some months (as low as 75% reduction compared to mean precipitation for September;
Figure 2.1). This occurred during the critical phase of tuber enlargement and maturation.
Thus, any mechanism that promotes the retention of moisture may contribute positively
to potato yield.

To demonstrate that soil structure was part of a complex system of interacting
variables that may show synergistic effects, examination of the correlation matrix (Table
2.) was undertaken. In ﬁeld I, the amount of clay signiﬁcantly contributed to the
predictive power of the stepwise regression model. Redulla et al., (2002) concluded as
well that the amount of clay was associated with well-aggregated soil, and positively
contributed to potato yield. On its own, soil structure indices 0.25 mm and MWD showed
minimal correlation coefﬁcient with yield, yet these characteristics contributed positively
to the ﬁnal yield prediction equation. In a crop modeling study using CropGro-Soybean,
soil structural condition contributed to a signiﬁcant increase in correlation coefﬁcient
between predicted and actual soybean yield. Without consideration of the underlying
drainage condition, the correlation coefﬁcient was at a low of 16%, but when the same

analysis was done on a subset of data-points from well-drained portions, the resulting

51

correlation coefﬁcient was raised signiﬁcantly to 98% (Irmak et al., 2001). Even in areas
with drainage problems, the correlation coefﬁcient was 59%, showing the signiﬁcant
impact of soil structure on yield.

The relative position of the potato crop in terms of micro-elevation across the
ﬁeld is important, under both excess and deﬁcient moisture conditions (Schneider et al.,
1997). Records for 2003 (Figure 2.1) showed a relatively higher precipitation compared
to 2004 towards the end of the potato growing season. This may have caused localized
ﬂooding conditions leading to detrimental effects to potato root and tuber growth, which
is highly sensitive to oxygen deprivation in saturated soil (Curwen, 1993). This could
potentially lead to a decline in ﬁeld 1 yield. However, the positive contribution of micro-
elevation to predicting potato yield in ﬁeld 1 indicated that plants located at relatively
higher micro-elevation sites were able to optimize growth and development. Such
optimization was manifested in more vigorous growth, indicated by green reﬂectance,
and overall had a positive contribution to yield. Excess moisture appeared to be mitigated
by the relative location of the potato plant at higher micro-elevation sites.

The role of potassium in the growth and development of tubers varies with the
intended market. Where the amount of solids is critical for processing potatoes,
potassium needs to be monitored closely, more so than when it is intended for the fresh
market. Potassium promotes moisture accumulation in the tubers resulting in decreased
speciﬁc gravity (Lang et al., 1999). The presence of high residual amounts of potassium a
couple of weeks before harvest could be an indicator of more than adequate intake by the
potato crop, leading to high absorption of moisture into the cellular structure of the

tubers, and contributed to higher tonnage. This scenario was very probable considering

52

that just before harvest in 2003, the amount of precipitation was sufﬁcient relative to ﬁeld
2 towards the end of the cropping season (Figure 2.1). No speciﬁc gravity measurements
were carried out in ﬁeld 1. However, for ﬁeld 2, speciﬁc gravity and K showed a
correlation coefﬁcient of 23% (P=0.03). In a Washington State study, Redulla et al.
(2002) found three out of four monitored ﬁelds had signiﬁcant correlation coefﬁcients
ranging from —27 to -41% for the relationship of K and speciﬁc gravity. Variability in
potassium levels across the ﬁeld was narrow yet inﬂuenced speciﬁc gravity values for the
harvested potatoes. This was indicated by the standard error of 2.2 ppm for a mean of
92.3 ppm (Table 2.1).
F reduction test and principal components analyses

The importance of physical, chemical and plant response to potato yield based on
the sum of square reduction test showed a signiﬁcant F statistic for the contribution of the
three factors to potato yield explaining 70% of variability (Table 2.3). Principal
component regression analyses identiﬁed six principal components explaining 64% of the
observed variability in ﬁeld 1 potato yield with stepwise F value to retain equal to 5%.
Examination of individual components making up the identiﬁed principal component
yield regressors did not produce a clear set of distinct factors that could explain potato
yield. Except for 15‘, 3rd and 5th (designated as aggregation, spectral response and
vegetation), the remaining principal components both have chemical, spectral and
textural parameters making deﬁnitive designation difﬁcult (Table 2.4). Field 2 on the
other hand, identiﬁed two principal components potato yield regressors designated as
nutrient status, aggregation and chemical nutrient potential all contributing signiﬁcant F

at 5% level. The principal component regression was able to explain 44% of potato yield

53

variability (Table 2.4). A principal component designated as plant response, was
identiﬁed but did not signiﬁcantly contribute to the regression coefﬁcient based on the

stepwise procedure.

54

CONCLUSION

The soil-plant-atmosphere continutun is a complex interaction of physical,
chemical and plant factors, but with proper analyses and understanding, can lead to
optimization of crop yield. Prudent baseline sampling of commercial potato ﬁeld enables
the farmer to supply critical quantities of nutrients needed for the normal growth and
development of the potato crop. However, signiﬁcant amount of variability in yield
persists, despite best management practices, which points out the need to investigate the
underlying growth environment, and vigor of the plant, to elucidate the observed
variability.

Yield predictive model derived from two potato ﬁelds in 2003 and 2004
demonstrated the signiﬁcant roles played by soil structure, speciﬁc spectral bands and
derived vegetative index, as well as the amount of potassium. The predictive regression
models were able to account for 66% and 60% of the yield variability for ﬁeld 1 and ﬁeld
2, respectively. The consistent positive contribution of soil structure characteristics points
to the need to manage this aspect of the cropping environment, especially when the
chemical requirements of the plants have been comprehensively addressed. The positive
role of spectral measurement in predicting potato yield provides a useful monitoring tool

as well.

55

REFERENCES

Adamsen, F. J., P.J. Pinter, E.M. Barnes, R.L. LaMorte, G.W. Wall, S.W. Leavitt, and
BA. Kimball. 1999. Measuring wheat senescence with a digital camera. Crop Sci.
39: 719—724.

Aguilar, R., and RD. Heil. 1988. Soil organic carbon, nitrogen, and phosphorus
quantities in northern Great Plains rangeland. Soil Sci. Soc. Am. J. 52:1076—1081.

Arshad, MA. and GM. Coen, 1992. Characterization of soil quality: physical and
chemical criteria. American J Alternative Agric. 7: 25-— 31.

Balchin, W. G. 1958. A Water Use Survey. The Geographical Journal, 124(4): 476-482.

Blake, GR, and K.H. Hartge. 1986. Bulk Density. In A. Klute (ed.). Methods of Soil
Analysis, Part 1. Physical and Mineralogical Methods: Agronomy Monograph no.
9 (2nd ed.), pp. 363-375.

Calhoun, F. G., J. M. Bigham, and B. K. Slater. 2002. Relationships among plant
available phosphorus, fertilizer sales, and water quality in Northwestern Ohio. J.
Environ. Qual. 31: 38—46.

Cambardella, C.A., T.S. Colvin, D.I. Karlen, S.D. Logedon, E.C. Berry, J.K. Radke, T.C.
Kaspar, T.B. Parkin, and DB. Jaynes. 1996. Soil property contributions to yield
variation pattern. p. 189-195. In P.C. Roberts et al. (eds). Proc. Third Int’l. Conf.
on Precision Agriculture, Madison, WI.

Cambouris, A. N., M.C. Nolin, B.J. Zebarth, M.R. Laverdiere. 2006. Soil management
zones delineated by electrical conductivity to characterize spatial and temporal

variations in potato yield and in soil properties. Am. J. Potato Res. 83(5): 381-
395.

Curwen, D. 1993. Water management. p. 67-75. In: Potato Health Management. R. C.
Rowe (ed.) The American Phytopathological Society, St. Paul, MN.

deHaan, K.R., G.T. Vessey, D.A. Holmstrom, J .A. MacLeod, J.B. Sanderson, and MR.
Carter. 1999. Relating potato yield to the level of soil degradation using a bulk

yield monitor and differential global positioning systems. Comput.Electron.Agric.
23 (2):]33-143.

Dobermann, A., J.L. Ping, V.I. Adamchuk, G.C. Simbahan, and RB. Ferguson. 2003.

Classiﬁcation of crop yield variability in irrigated production ﬁelds. Agron. J. 95:
1105-1120.

56

Draper, N.R. & H. Smith. 1981. Applied regression analysis. John Wiley & Sons, New
York.

Griffith, J. A., E. A. Martinko, J. L. Whistler and K. P. Price. 2002. Preliminary
comparison of landscape pattern—Normalized Difference Vegetation Index

(NDVI) relationships to central plains stream conditions. J. Environ. Qual.
31:846—859.

Gunst, R.F., and R.L. Mason. 1980. Regression analysis and its application: A data-
oriented approach. Marcel Dekker, Inc., New York.

Hair, J.R., R.E. Anderson, R.L. Tatham, and WC. Black. 1995. Multivariate data
analysis with readings. Prentice Hall, New Jersey.

Hunt, R.E., M. Cavigelli, C. Daughtry, J. McMurtrey, and C. Waltahll. 2005. Evaluation
of digital photography from model aircraft for remote sensing of crop biomass
and nitrogen status. Precision Agriculture. 6:359—3 78.

Irmak, A., J .W. Jones, W.D. Batchelor, and J .O. Paz. 2001. Estimating spatially variable
soil properties for application of crop models in precision farming. Trans.ASAE.
44 (5):]343-1353.

Kitchen, N.R., S.T. Drummond, E.D. Lund, K.A. Sudduth, and G.W. Buchleiter. 2003.
Soil electrical conductivity and topography related to yield for three contrasting
soil—crop systems. Agron. J. 95:483—495.

Kok, K. and A. Veldkamp. 2001. Evaluating impact of spatial scales on land use pattern
analysis in Central America. Agric. Ecosyst. Environ. 85: 205-221.

Kravchenko, A.N., and D.G. Bullock. 2000. Correlation of corn and soybean grain yield
with tepography and soil properties. Agron. J. 92:75—83.

Kravchenko, A.N., and D.G. Bullock. 2002. Spatial variability of soybean quality data as
a function of ﬁeld topography: 1. Spatial data analysis. Crop Sci. 42: 804-815.

Kravchenko, A.N., G.P. Robertson, K.D. Thelen, and RR. Harwood. 2005. Management,
topographical, and weather effects on spatial variability of crop grain yields.
Agron. J.. 97:514-523.

Kravchenko, A.N., K.D. Thelen, D.G. Bullock, and NR. Miller. 2003. Relationship
among crop grain yield, topography, and soil electrical conductivity studied with
cross-correlograms. Agron. J. 95:1132-1 139.

Lamb, D. W. 2000. The use of qualitative airborne multispectral imaging for managing

agricultural crops — a case study in south-eastem Australia. Aust. J. Exp. Agric.
40: 725—738.

57

Lang, N.S., R.G. Stevens, R.E. Thornton, W.L. Pan, and S. Victory. 1999. Potato nutrient
management for Central Washington. EB1871. Coop. Ext. Washigton State
Univ.,WA.

Lark, R.M., and J.V. Stafford. 1998. Information on within-ﬁeld variability from
sequences of yield maps: Multivariate classiﬁcation as a ﬁrst step of
interpretation. Nutr. Cycling Agroecosyst. 50:277—281.

Li, X.Y., A. Gonzalez and A. Sole-Benet. 2005. Laboratory methods for the estimation of
inﬁltration rate of soil crusts in the Tabemas Desert badlands. CATENA 60:255-
266.

McKenzie, R.C., S.A. Woods, L.E. Hingley, T.W. Goddard, J. Rodvang, and T. Harms.
2000. Site speciﬁc management of potatoes. p. xxx-xxx. In P.C. Robert (ed.) Proc.
of 5th Int’l. Conf. on Precision Agriculture and Other Resource Management, ,
Bloomington, MN. 16 — 19 July year. ASA/CSSA/SSSA, Madison, WI.

Mulla, DI, and J .S. Schepers. 1997. Key processes and properties for site-speciﬁc soil
and crop management. p. 1-18. In F.J. Pierce and E]. Sadler (ed.). The state of
site speciﬁc management for agric.

Neter, J., W. Wasserman, M. H. Kutner. 1989. p. In: Irwin, R. (ed.), Applied linear
regression models, 2nd ed. Homewood, IL. .

Ovalles, FA, and ME. Collins. 1986. Soil—landscape relationships and soil variability in
North Central Florida. Soil Sci. Soc. Am. J. 50:401—408.

Paz, J.O., W.D. Batchelor, T.S. Colvin, S. D. Logsdon, T.C. Kaspar, and D. L. Karlen.
1998. Analysis of water stress effects causing spatial yield variability. Trans.
ASAE. 41(5): 1527—1534.

Perez, A. and M. Kogan. 2003. Glyphosate-resistant Loliul multiﬂorum in Chilean
orchards. Weed Science. 43: 12-19.

Pettersson, C. G., M. Soderstrom, and H. Eckersten. 2006. Canopy reﬂectance, thermal
stress, and apparent soil electrical conductivity as predictors of within-ﬁeld
variability in grain yield and grain protein of malting barley. Precision Agric.
7:343 — 359.

Phiri, R.H., S. Snapp, and G.Y. Kanyama-Phiri, 1999. Soil nitrate dynamics in relation to

nitrogen source and landscape position in Malawi. AgroforestrySyst. 47
(1/3):253-262.

58

Po, EA. 2007. Potato cropping system effects on aggregate stability and spatial
distribution of chemical and physical factors affecting yield. PhD Dissertation,
Dept. Crop and Soil Sciences, Michigan State University, East Lansing, MI.

Pringle, M.J., A.B. McBratney, B.M. Whelan, and J .A. Taylor, 2003. A preliminary
approach to assessing the opportunity for site-speciﬁc crop management in a ﬁeld,
using a yield monitor. Agric. Syst. 76:273—292.

Redulla, C. A., J.R. Davenport, R.G. Evans, M.J. Hattendorf, A.K. Alva, and RA.
Boydston. 2002. Relating potato yield and quality to variability in soil
characteristics. Am. J. Potato Res. 79(5): 8.

Rhoades, J.D., N.A. Manteghi, P.J. Shouse, and W.J. Alves. 1989. Soil electrical
conductivity and soil salinity: New formulations and calibrations. Soil Sci. Soc.
Am. J. 53:433-439.

SAS Institute Inc. 2006. Base SAS® 9.1.3 procedures guide, 2nded. Volumes 1-4, Cary,
NC.

Schabenberger, 0., BE. Tharp, J.J. Kells and D. Penner. 1999. Statistical tests for
horrnesis and effective dosages in herbicide dose response. Agron. J. 91: 713-721.

Schneider, 1. 1960. Soil survey of Montcalm County, Michigan. USDA-SCS.

Schneider, S. M., R.A. Boydston., S. Han, R.G. Evans, and RH. Campbell. 1997.
Mapping of potato yield and quality. In J. V. Stafford (ed.). Precision Agriculture

’97: Proceedings of the First European Conference on Precision Agriculture, pp.
253—261.

Seidl, M. S., W. D. Batchelor, and J. O. Paz. 2004. Integrating remotely sensed images
with a soybean model to improve spatial yield simulation. 47(6): 2081-2090.

Sudduth, K.A., N.R. Kitchen, G.A. Bollero, D.G. Bullock, and W.J. Wiebold. 2003.
Comparison of Electromagnetic Induction and Direct Sensing of Soil Electrical
Conductivity. Agron. J. 95:472—482.

Taylor, R.K., G.J. Kluitenberg, M.D. Schrock, N. Zhang, J.P. Schmidt, and J .L. Havin.
2001. Using yield monitor data to determine spatial crop pduction potential.
Trans. ASAE 44:1409—1414.

Verhagen, A., H.W.G. Booltink, and J. Bouma. 1995. Site-speciﬁc management:
Balancing production environmental requirements at farm level. Agric. Sys. 49:
369-384.

59

Yang, C., J.H. Everitt, J.M. Bradford, and DE. Escobar. 2000. Mapping grain sorghum
growth and yield variations using airborne multispectral digital imagery.
Trans.ASAE 43(6): 1927-1938.

Zaman, Q., and A. W. Schumann. 2006. Nutrient management zones for citrus based on
variation in soil properties and tree performance. Precision Agric. 7:45—63.

Zhang, R. 1997. Determination of soil sorptivity and hydraulic conductivity from the disk
inﬁltrometer. Soil Sci. Soc. Am. 161: 1024-1030.

60

 

Table 2.1 Basic statistics for potato yield. speciﬁc gravity and regressor variables classiﬁed into chemical. physical and spectral factors taken from Fields 1 and 2 in

Montcalm County. Michigan in 2003 and 2004 respectively.

 

 

 

 

 

 

 

’1‘ Variables : 13731171] 7 buffer pH :Ca encalcium ; K 5 potassium ; Mg 2 magnesium ; Pl : phosphorus (Bray l) ; ECa 2 aparent electrical conductivity (shallow 0-30 cm); H : hydrogen ; CEC 2 cationic
exchange capacity (mqu‘l); Elev '7 elevation: MWD mean weight diameter, mm; MC : volumetric moisture content; k : hydraulic conductivity, mmhr'l; BD : bulk density; Vegarea = vegetated area, %;

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Yield ~ tonsha"; SG ' speciﬁc gravity : (l green : R a red ; 1R 7’ infrared .
+

+ _

x 2 mean; so. standard error.

61

 

 

 

 

 

Variables’r
ppm Base Saturation, 95
ClassificationFieldStatlC B_pH Ca K Mg Pl ECa Ca H K Mg pH CEC C, 96
’ 2' 6.87 502.8 167.7107 214.7 — 46.7 28.3 78.0 16.8 5.7
1 s.e. 0.01 11.1 4.7 1.9 0.2 — 0.6 0.8 0.2 0.3 0.0 0.1 0.01
r 6.54 575.5 92.3 72.614.00.18 48.1 37.3 4.2 10.4 5.5 5.9 1.05
Chgmigal L42n‘s.e 0.13 25.2 2.2 3.2 1.20.01 1.0 1.3 0.1 0.3 0.0 0.2 0.02
Aggregate size , um Texture Components , 96
Elev MWD 53 106 250 1000 2000 2250 Sand Silt Clay MC k BD
77’ ’77"—77 X 278.1 1.21 13.3 12.8 45.518.8 9.7 73.9 76.4 16.6 7.1 15.885.01.46
1 s.e. 0.3 0.02 0.2 0.2 0.3 0.3 0.3 0.3 0.2 0.2 0.1 0.1 3.90.01
‘77 r 251.7 2.18 7.5 6.4 26.222.5 37.4 86.1 82.3 12.0 5.7 11.5 —
Physical 2 EZZQ. 0.1 0.01 0.1 0.2 0.4 0.4 0.5 0.3 0.2 0.2 0.1 0.3 — —
# Spectral Band
Unadjusted Adjusted
R IR G R IR Vegarea
TTIIT’ ’ 99.0 32.451.549.316 2 47.2
0.5 2.0 1.5 1.4 1.2 1‘??7
190.0 160.1 4.8 9.1 8.6 4.6
Spectral 1.1 1.2 0.3 0.5 0.5 C.2
" "'” so
.0824
Dependent .0003

 

 

 

 

Table 2.2a Pearson correlation matrix ofselected soil. plant and spectral properties of ﬁeld I in l\4ontcalm. h'lichigan.

 

 

 

 

Field 1 c“ MC Elev K Clay 0.250mm MWD C.unadj Vegcover G/R,mdj R/IRunadj

Yield -0.049r1s 0.235 * 0.327** 0.595*** 0.474*** 0.l20ns 0.202* 0.654*** 0.056115 0.224* 0.423***
C 0.035ns -0.l_55ns 0.079ns 0.ll7ns 0.220“ —0.178ns —0.235* —0.067ns —0.l66ns -0.055ns

MC 0.210* 0.327** 0.223* 0.092ns —0.058ns 0.107ns 0.008ns 0.189ns 0.156ns

Elev 0.289** —0.075ns —0.023ns 0.003ns 0099* ~0.023ns 0.186ns 0.354**

K 0.279** 0.154115 0.064118 0.447*** 0.129ns 0.167ns 0.333**

Clay 0.320”? ~0.051ns 0.210ns 0.128ns 0.054ns 0.077ns

250mm e0.694*H 0.247* 0.037ns 0.146ns 0.27l**

MWD 0.054115 —0.02lns —0.ll3ns —0.l48ns

Gurmg" 0.070ns 0.490*** 0.587***
Veg-.714, 0.109ns —0.127ns

G/Runarlj —0.004HS

*** — significant at . 0.0001: ** — significant at “0.05: * — signiﬁcant at ”0.10: and ns -- not signiﬁcant

i C 7’ % total soil carbon : MC 7' volumetric moisture content : Elev : elevation ; K 7’ potassium ; Clay : "/6 Clay 0250111111 ' l — 2 0.250 sized water stable aggregate; MWD : mean weight diameter;

Gum, 7 (lrccn band unadjusted: chwc, - vegetative cover: G/RumJ : green red unadjusted band ratio; R/lRum, - red infrared unadjusted band ratio.

 

62

 

 

 

liable 2.2b Pearson correlation matrix ol‘selected soil. plant and spectral properties from field 2 in t\'lontcalm. Michigan.

 

 

 

 

 

Field 2 c+ MC Elev K Clay 0.250mm MWD Gunadj Vegcover G/Rmadj R/IRunadj ECa
livid w l7lns 0.488*** —0.150ns —0.034ns 0.049ns 0.2551040167439*** —o.405*** o.349** O.lOlns O.463***7 O.697***
C 0.575x** —0.194ns 0.280** -O.265** —o.203* —o.xi8* —0.511*** o.273** —0.139ns 0.127ns o 485***
MC 70.334+* o.232* 0.03lns —).1l5ns e0.32l** ~0.486*** 0.396*** 0.19lns o.321** O.781***
Elev —O.281** 0.093ns o.224* ~o.neans o.347*+ e0.l46ns 0.074ns 0.053ns —O.418***
K 0.122ns —O.326** 0.139ns —0.210* 0.122ns 0.045ns —0.039ns o.239*
Clay —.J.010ns 0.083ns 0.325H 0.040ns 0.425*** —0.040ns 0.025ns
250nm ~o.cz9k** 0.142ns 0.086ns o.234* 0.020ns 0.050ns
MWD 0.216* —0.300** —0.l51ns —0.l75ns —0.377**
cudji —o.427*** 0.589*** —O.371** —o.523*H
Veq,_ I 0.092ns O.4l4*** o.510***
c/pmﬁl —0.099ns 0.310**
R/IPMAH O.391**
.

** — significant at - 0.000 I: ** - significant at /‘0.05; * — significant at <0.10; and ns - not significant
’i‘ C ”/0 total soil carbon : MC ,
Gum, e Green hand unadjusted; chwm vegetative cover; G/RmdJ : green red unadjusted band ratio: R/lRWdI

63

' apparent electrical conductivity.

volumetric moisture content : Elev :’ elevation ; K 1 potassium ; Clay: 90 Clay ; 0.250mm , l - .7 0.250 sized water stable aggregate; MWD " mean weight diameter:
red infrared unadjusted band ratio: ECa

 

 

 

 

Table 2.3 Sums of squares reduction test of full and reduced model for potato yield as a
function of chemical, physical and plant spectral response variables from a potato yield
variability study at two commercial fields in Montcalm County, Michigan.

 

 

Model‘r R31” 33 df MS robs P
Full 0.70 8047.65 173 46.52
Reduced

chemical 0.62 10761.28 192 56.05 3.07 0.01
physical 0.63 10136.50 187 54.21 3.21 0.01
spectral 0.54 13204.59 197 67.03 4.62 0.01

1‘ Regression analyses with full model consisting of chemical. physical and potato spectral response with
reduced model referring to single factor regression analyses involving each of the three independent factors;
3‘ R2 = coefficient of determination: SS =—‘ sum of squares: df= degrees of freedom; MS = mean square; FobS =
observed F statistics: P = probability

64

.AmﬁocEEmQ E :39? m5 ESPEDwB

0o N 53> 323:; :m 5 36 v t08~>=owﬁc x33 83 mEocanco 39655 :23 5:30:09: 32:3 3:: 70:3; 2: cm; @2632: 335:3) x 6

mad use Ed 3 .3 :2823 :2282 can been 22st? 5:: womb—Em EocanoU 32055 332%; @325 nee/ton 263 mEocanoU BEBE;

3263838. .m of 2 :oczetEcU

 

 

 

+

.2 2an 303.832 m «Em _ 26: E 333.? 3:83. 2:6on 93 236.93 .2252? 529:6 .m can mm 3622: .0.
at T . ...-.. .7...
Amm.eiv :eim .Amm.e-c 5e eoN.e . mm.ee ee m H.
lnu.e_ mum ..am..v m ..-.. v. 1N.3 a .
we . a r
3m .cL mm . $26.0, .2 .37 g a. 9.2.3
3N6“. DEE .2m.©.... mu .Samju; emu mad 0.
Ame.ov mmommm> .lme.5v mo .ﬂee.el em eo.e m
Aem.oiv me .mm.ev eme
.lem.e-v mmeseo .lem.e-v mmesem .lem.ev mow .lem.ov ameeeexme me.e m
Aom.ev eze .Aom.ov gs .lem.el zetm ee.e N
Ammdiv mmommmummmoeome 20mg: SE $2.0 $3.0; Sea 2.0 H 4.
emceeum>
nmmeomeom> oCmouerH cesameaxm . mexm oe pause

 

emu—pat? omcommoe :83 was 306.23 .8282? mo $9.93 2232on 26:09:00 6&6:an om_.$q8m vN Esme

200 —

150«

100

Precipitation , II

 

 

 

a ,
L A I \‘
\‘A‘ °\.\ ’1

,--n 3‘ .
E” \ ﬁg’,

\ I

\ I

\
E

 

 

 

Month

 

for 2003, 2004, and the 8-year average.

66

.. WWiCt-opping Sgson :layrAvg —A-27(E:B—2004l
Figure 2.1 Total monthly precipitation at Montcalm Research Farm (Entrican, Michigan)

«a

4 -.~
.u. re...“

3.
f.
i

   

Figure 2.2 Study ﬁeld 1 located at the Southeast quadrat of Vickeryville-Tamarack Roads
in Vestaburg, Montcalm County, Michigan showing the entire ﬁeld and the location of
grid points sampled. Soil types of the area is also shown: Mb (Mancelona loamy sand, 0 —
2% slope), Mc (Mancelona loamy sand, 2 — 6% slope) and Nm (Newaygo sandy loam
with 2 to 6% slope).

s «app-w..- “A... .. V

67

      

 

.A . ”a. ‘
M.“ . , ‘
.ﬂ . . v—‘_
' " 14.4"

"am :11»! ' — '
Figure 2.3 Study ﬁeld 2 located near the southeast quadrat of Tamarack-Bollinger Roads
in Vestaburg, Montcalm County, Michigan showing the the location of grid points
sampled by this study as well as soil types Ga (Gladwin loamy sand) and Mb (Mancelona
loamy sand).

68

.830 E ﬂ owes: REwtO 628%? E683 E3952 9 22cm
owns: 655 2: mo owﬁcooeom a me $896 use 62:38 :2: 295 owns: 2: E “53> we _ wars: 26x5 .va wees Begum? wEB
w 28 mecca. woemeomgéo: memos o HEB oweg an 85on 9 am: owns: 0:89:00 eoaﬁﬁ 05 :o eogotoa :65 33 338w oz; 85
magma? 3223:qu Aem am .mm was a< as $5853 “8 RES BE 338988 £8th owuo 9682 8 Arm wee it 3380
memos Am: mOM 8535 an new owns: A< .mOMV “es—meooewéom m wEZo>E mombmcm use wcmmmoooa doEmEUom owes: VN 95mm.”—

msmm.= mvaw mm.“ H H

thvé vNme mad a :
HOMHHDHOHQ %uqmnuwhm uwﬁwahwnnﬂ uwlﬁqlhwlou mmouu

amewhﬂ uwacuwmwbinoz i a umuwhd 3595mm; u H mmeaow "Hanan. waﬁm

 

69

 

 

 

 

857 +
++
+ +
++ +-
80 " +
+
4+
4.!- + "H' + + +
. + +
,2 i- + + f + + +
a + + +4-
1 + +
3 70 + + + + +
a + +
2 ++ +
> 85‘ + + + +
_ + ‘4- + +
+
g + + + + + + ++ 4+ 4'
*9 ++ + + +++ .1.
0 80‘ + 'H' ++
< +.+
+ + +
+ +
+
55 - + +
50 ‘
i l i l I T l I i I i l
55 58 60 63 85 88 70 73 75 78 80 83 95

Predicted Yield, tonsha"
Figure 2.5 Plot of actual and predicted potato yield from ﬁeld 1 fall of 2003 harvest data
using the signiﬁcant (P<0.0001) stepwise predictive equation of yield = -392.9 +
0.042(K) + 2.3359(Clay) + 8.7985(MWD) + 0.6178(Green Band
unadjusted)+1.307(Elevation, m) with adjusted R2 = 0.67 and an RMSE of 4.38.

70

 

70

651

60‘

Actual Yield, tonsha‘I

35‘

30‘

55‘

50‘

45‘

40‘

 

25

 

 

-+ + + + 1 +
+ + +
+ ++‘*‘* +
+ +
+
+ +
+ +++ +
+++
+
i + ++
+¢++
+*+ *+
+ + 1;.
+1“ 4*
++¢+ “ll.- ++
+ + +
++
+
+
+ +
++
+-
+ +
l l 1 I l l l I I I
35 40 45 50 55 60 85 70 75 80

Predicted Yield, tonsha'l

Figure 2.6 Plot of actual and predicted potato yield from ﬁeld 2 fall of 2004 harvest data

using

the

signiﬁcant (P<0.0001) stepwise predictive equation of yield

59.288+0.6997(250mm WSA)-89.259(Green by Red unadjusted ratio)+91.92(ECa) with
adjusted R2 = 0.60 and an RMSE of5.9.

71

CHAPTER III: SPATIAL BEHAVIOUR OF SOIL PROPERTIES AND THEIR
IMPLICATION FOR SAMPLING STRATEGIES IN A POTATO-BASED

SYSTEM

ABSTRACT

Current farmer practice of sampling at low spatial resolution of one sample per
hectare or less is not sufﬁcient to characterize soil variability and its associated dynamics
that affect plant growth and development. A study was conducted in two Michigan
commercial potato ﬁelds, designated as ﬁeld 1 and 2, in 2003 and 2004, respectively, to
re-assess current recommendations on size of grids used in soil sampling. A high
resolution data source - digital orthoquad imagery was evaluated for its potential to
indirectly measure select soil properties. In this study, effective sampling grid size was
tested in relationship to spatial dependence of yield, which was shown to be 42 m and 50
m for ﬁelds 1 and 2, respectively. Semi-variogram analyses of soil quality surrogates of
water stable aggregates (WSA) showed moderate to strong spatial dependence, with
ranges that coincided to some degree with the observed ideal grid size for both ﬁelds
based on yield. Use of digital orthophoto (DOQQ) taken at a very high spatial resolution
of 1 m ground equivalent is a potential valuable tool for monitoring the levels of
important soil physical attributes such as amount of soil carbon, 0.25 and 1 mm WSA
size fractions. However, the usefulness of this resource should be put in context by

examination of soil survey data. A dramatic improvement in correlation coefﬁcient was

72

achieved through analysis carried out on the basis of soil type series, rather than across

series.

73

INTRODUCTION

Traditionally, agricultural ﬁelds have been managed based on the concept of
averages or median soil nutrient requirements (Franzen and Peck, 1993), or what is
known as whole ﬁeld management (Lowenberg-DeBoer and Aghib, 1999). This
practice leads to under-application of agricultural inputs in cases where the soil
nutrient reserves are low, and over-application where the native supply of nutrients is
above the mean value. Such conditions create concerns related to environmental
pollution, where over-application occurs (Wollenhaupt et al., 1995), and low crop yield
where under-application occurs (Lowenberg-Deboer and Swinton, 1997).

An alternative to whole ﬁeld management is site-speciﬁc management (SSM) or
precision agriculture (PA), where the amount of agricultural input is dictated according to
the actual soil supply and crop requirement. This takes into account variability in plant
growth and development requirements across a landscape, and heterogeneity in ﬁeld
micro-environment. Sensitivity analysis reported in a literature review done by
Bongiovanni and Lowenberg-Deboer (2004) have shown that targeted application that
reduces the amount of applied nitrogenous fertilizer to approximately half of
recommended doses on a whole farm basis resulted in modest proﬁtability above a
uniform fertilization strategy. Vetsch et al. (1995) reported a reduction of as much as an
average of 60 to 90 kg of nitrogen (ha") in optimum nitrogen rate from the conventional
single rate applied, in a study in Minnesota for com. This is consistent with considerable

over-fertilization associated with the common practice of non-targeted fertilization.

74

A review of current fertilizer recommendations by Hergert et al. (1997) concluded
that reﬁning knowledge and use of SSM will expose the problems associated with the
usual fertilizer management strategy of applying whole ﬁeld recommendation (plus 20%
to account for variability). Varying the amount of lime applied to a moderately acidic
ﬁeld planted with soybeans in Michigan improved overall yield, compared to a single rate
application for the entire ﬁeld (Pierce and Wamcke, 2000). Redulla et al. (2002),
sampling potato ﬁelds in Washington State at a square grid of 0.4 ha interval, concluded
that the consistent negative correlation between speciﬁc gravity and potassium across
three of the four ﬁelds they studied, could be a justiﬁcation for the use of variable rate
fertilization based on native potassium availability.

Identiﬁcation and mapping of stable features of the landscape relative to soil and
micro-environmental conditions that can impact crop performance, and managing those
stable features as separate units, is an example of SSM. SSM offers a convenient way to
harmonize ﬁeld operations. Through the use of management zones, yield variability for
corn and soybean compared to whole ﬁeld variance was reduced by 40% (Brock et al.,
2005). Hornung et al., (2006) compared annual yield performance of grains within
management zones identiﬁed through bare soil imagery, topographic position and
farmer’s perception of yield potential (color-based approach) on one hand, and the
delineation based on bare soil imagery, cationic exchange capacity, texture and yield
from previous year (yield-based approach). They found the color-based approach to be
more efﬁcient in identifying stable yielding potential than the yield-based approach. The
inclusion of the previous year’s yield together with independent variables that were

related to yield potential, may have a created a redundancy in the yield-based algorithm.

75

This may have reduced the effectiveness of the methodology and limited the ability to
identify stable zones. Developing empirical relationships between yield, optical and soil
properties from previous years, and carrying over the relationship to the current year as
input in the identiﬁcation of management zones, can potentially increase identiﬁed zone
yielding potential stability. Hornung et al. (2006) agreed that assigning weights to the
yield-based data-layer could have improved stable management zone identiﬁcation.
Identiﬁcation of management zones necessarily increases the amount of soil sampling
and related validation activities. Whether the resulting reduction in variability produces a
better economic return overall compared to whole farm management requires further
study (Brock et al., 2005).

The utilization of SSM raises a number of questions. What are the resources
available for its implementation? How much detail is required for the formation of
workable empirical yield models? Which variables in agricultural production are likely to
maximize the beneﬁts of SSM? Another important question is which statistical
approach is most effective to interpolate values between sampled locations?

To dissect the yield variability puzzle, consideration should be given to readily
available maps from the United States Department of Agriculture — National
Resources Conservation Service (USDA-NRCS) as guides for agricultural production
suitability. These maps range from one depicting the entire State of Michigan known
as a 5‘h order soil survey at a scale of 125,000,000; a 2nd order 1:190,080 county map;
and the high resolution 1St order map at a scale of 1:1000, depicting detailed resources

on a parcel of land (Cooper, 2004). Any of these maps can be used as input to a

76

Geographic Information System software, but no additional detail can be extracted
beyond the resolution that was present in the original material.

At ﬁrst glance, a 1St order survey map is the only one relevant to within ﬁeld
variability analysis. This map, represented on a letter size paper would correspond to
roughly 6 hectares on the ground, which may not be sufﬁcient for purposes of
mapping within ﬁeld variability. However, it serves as a starting point for more
detailed soil sample characterization. For agricultural production purposes, 1 to 2 soil
cores are commonly sampled and analyzed per hectare to obtain information on soil
characteristics prior to planting (Hergert, 1998). This sampling density and
subsequent soil analysis serves as a basis for fertilizer recommendation for a
particular yield target. At this density of sampling, this will provide information
resolution that is 600% greater than a lSt order map, but even this is not sufﬁcient for
most properties.

The majority of agricultural dealers sampled over the years from 2000 until 2006
(Akridge and Whipker, 2000; Whipker and Akn'dge, 2001, 2002, 2003, 2004, 2005 and
2006), indicated a higher proportion of surveyed dealers providing sampling services for
their customers, who preferred a grid sampling pattern followed by soil type as sampling
methodology. Of those that used grid sampling, a majority (ranging from 53 to 60%)
followed a 1 hectare sampling grid size. Ultimately, the inherent properties and
variability of the soil will dictate the intensity of sampling required to support detailed
mapping and understanding of crop response (Kravchenko and Bullock, 2000).

Micromanaging a production ﬁeld necessitates the presence of a detailed soil

property database, which in turn involves intensive sampling. Sampling to detect within

77

ﬁeld variability can be accomplished primarily in two ways: either by a) a grid of a
particular dimension which can be overlain on the ﬁeld of interest; or b) use of point data
collected based on some criteria determined prior to, or during the sampling process. In
the ﬁrst method, data is collected at each intersection of the horizontal and vertical grid
positions (Mueller et al., 2001). For the second method, utilization of prior information
from the ﬁeld of interest, such as previously conducted soil surveys or on-the-spot
determination of areas of interesting variability, can be used to guide the sampling
process (Pocknee et al., 1996). A combination of the two methods was illustrated by
Kravchenko and Bullock (2000), where sampling for topographic data involved using
semi-regular grids, with the further step of varying the distance between measurements
based on the complexity of the topography.

The dimension of the grid used to designate sampling point locations generally
varies with the type and number of factors of interest. These factors often include soil
texture, desired map quality (Hergert, 1998) and prior spatial structure of a particular
variable of interest (Flatman and Yfantis, 1996; Sadler et al., 1998), or limitations
imposed by harvest equipment (Homung, et al., 2006). Single grid cell area values of
0.0625 (Jaynes et al., 2003), 0.4000 (Pocknee et a1, 1996; Redulla et al, 2002) and 0.5000
hectare (Morgan et a1, 2002) were examples found in the literature.

Once point data is generated, the challenge is to use interpolation across the ﬁeld
to generate surfaces that are representative, and within acceptable constraints. A
promising approach involves a system of generating surfaces from a few points in the
ﬁeld, supplemented by the variability shown to occur from a densely sampled variable

that has a distribution related to the variable of interest. An example is provided by Heisel

78

et al. (1999), where co-kriging was used to improve estimation of a sparsely sampled
weed count variable using an intensively sampled variable of silt content. Results of their
study showed that this co-kriging approach reduced the variance by 11 %, compared with
a method relying on kriging alone. Colonna (2002) also utilized secondary information to
improve predictive ability of both simple kriging with varying means, and a co-kriging
interpolation methodology. These methods were used to improve the ﬁeld map of organic
matter distribution.

On the basis of applicability to end-user, simple kriging with varying means,
and ordinary kriging or lognormal ordinary kriging, are possibilities which could
greatly enhance predictive capability of interpolation methodology. However, these are
not available in commercial software packages (Colonna, 2002), or involve complicated
computations, including the formulation of a carefully chosen variogram model, as well
as appropriate log-transformation methodology (Kravchenko and Bullock, 1999).

Dense information may also be developed from spectral characteristics. An
example is provided by a study of a cranberry production site. Color Infrared imagery
was taken from an airplane and successfully used to predict cranberry vine density, areas
with drainage problems, yield and disease incidence in New Jersey (Pozdnyakova et al.,
2002). Visual representation of the area was made possible through kriging algorithms
derived from experimental variograms. In a wheat rotational system in Spain, Lopez-
Granados (2005) concluded that the use of kriging with varying local means, in
combination with spectral data from the blue band of the electro-magnetic spectrum,
offered increased precision in the prediction of organic matter, pH and potassium,

compared to simple regression and ordinary kriging with regression.

79

Yield monitor data provides an excellent source for mapping within ﬁeld
variability, as it relates yield data to physical location in the ﬁeld through global
positioning system (GPS) technology. Evaluating the stability of yield data over a
period of time could provide unique insights regarding persistent features of the
landscape that need to be managed through appropriate interventions. Doberman et al.
(2003) reported that there are two components of yield, namely the stable part and the
annual random part. If one can remove the random part, then one can maximize the
inherent productive capability of the soil. An example of a random effect could be
erratic precipitation which can be removed through the application of supplemental
irrigation. At the same time, this approach can reduce over- and/or under- application of
inputs under irrigated production environment.

The ﬁrst challenge in the determination of spatial yield stability is the
segregation of the production ﬁeld into grids or areas of interest. To accomplish this in
terms of yield potential, multi-year yield data are needed. Blackmore (2000) used three
years of yield monitor data to derive annual mean, overall mean and standard deviation
for areas of interest. He classiﬁed geospatially located data based on the performance of
an area as yielding higher or lower than the average, and whether the coefﬁcient of
variation was less than or greater than 30. Thus the ﬁeld was divided into the following
categories: a) higher yielding and stable, b) lower yielding and stable, and c) unstable
areas. Using this methodology, the resulting map can be used to identify the
problematic unstable areas and more directed research can be conducted on these areas.
One potential challenge is that the classiﬁcation scheme may not be sensitive enough to

come up with viable management sections to adequately address within ﬁeld variability

80

(Doberman et al., 2003). He suggested increasing the number of classes that is
developed through Blackmore's (2000) analysis to increase the utility of the map
developed as a tool for continuous variable rate management. Ultimately, the number of
classes developed is a function of area contiguity, practicality and convenience.
Practical considerations include the necessity to take into account farm machinery
requirements to negotiate across the ﬁeld to deliver targeted, speciﬁc area tillage and
inputs.

Further complicating stability analysis of within ﬁeld spatial variability pattern
is the challenge associated with calibration and accuracy of GPS- yield monitoring
equipment. Davenport et al. (2002) monitored two to four ﬁelds on a commercial potato
farm in southeastern Washington for three years, and found that the yield monitors
generally over-estimated potato yield compared to point estimates. This may have been
caused by the utilization of different yield monitors at different times, with the
corresponding calibration challenge across the study years. Furthermore, the very
nature of row crop agriculture presents a unique problem of location stability as the
location of rows from year to year varies. As a harvester operates across several rows,
even a split-second interval in data generation can translate to substantial differences in
yield reported, potentially throwing off comparison between years, since each year data
are averaged from different locations in the ﬁeld.

The objective of the study reported here was to determine the range of spatial
dependence of yield and soil structure related properties in Michigan potato production.

Speciﬁcally, the objective was to evaluate the potential of digital orthoquad imagery

81

high resolution data to correlate quantitatively these datasets with spectral data from a

consumer grade digital camera, and select soil properties.

82

MATERIALS AND METHODS

Location

The study was conducted on two commercial potato production ﬁelds managed
by the same farm operator and designated as ﬁelds 1 and 2, which were studied in 2003
and 2004, respectively. Field 1 is a 43.5 hectare commercial production ﬁeld that had
been in a potato-based rotation for the past 60 years. Rotation crops included corn (Zea
mays L.), dry beans (Phaseolus sp. L.), and wheat (Triticum aestivum L.) Most recently,
the rotations of potato (Solanum tuberosum L.) cultivars grown in this ﬁeld included Pike
with cereal rye (Secale cereale L.) as a cover crop in the winter in 2001 and FL 1833
followed by wheat and rye as cover crops in 2003. It is located at 84°57' longitude and
43°26' latitude, at Vestaburg, Montcalm County, Michigan, USA.

The second ﬁeld was a 20.7 hectare commercial ﬁeld located near ﬁeld 1 with the
following center geographic coordinates, 84°52' longitude and 43°26' latitude. This ﬁeld
had been in a potato-based rotation for the past 30 years. Crop rotation practices were
similar to those of ﬁeld 1. When this ﬁeld was sampled in 2004, it was planted with the
FL1879 potato cultivars and a rye winter cover crop. Two years before, it was planted
with potato (Solanum tuberosum L. var Pike) and a rye winter cover crop followed by
snap beans with wheat as winter cover crop the following year.

Sampling Design

Historical soil sample characterization data taken from the study area in previous

cropping seasons were used to formulate the sampling design. For ﬁeld 1, the average

ratio of samples per hectare used historically by crop advisors was 1.7:] for the

83

determination of percent organic matter, phosphorus, potassium, magnesium, calcium,
pH, cation exchange capacity (CBC), and zinc. The study area was located at the center
of the ﬁeld where the slope was minimal, to minimize the effect of topography on the
behavior of soil physical and chemical properties. Prior geographically referenced
sampling points for ﬁeld 2 had a ratio of 2.9 samples for every hectare (Otto, 2004
personal communication).

For ﬁeld 1, the study area was divided into 108 sections, each consisting of 16
rows (spaced at 0.89 m apart) on the northern and southern sides, and approximately 9 m
both on the east and west sides from the center of the section, making the effective area
equal to 0.05 hectare. Thus, the ratio of sampling points per hectare was 20:1, compared
to the earlier ratio of 1.721. The width of the sampling grid was chosen as to
accommodate 16 rows of potatoes on the north and south side of the grid center. Tuber
diggers normally involve 1 eight row and 2 four row machine diggers for a total of 16
harvested rows in one pass, moving from west to east or vice versa. The east and west
length was arbitrary and roughly included a total of six yield monitor data points. In a
similar manner, a total of a hundred grid points were sampled for ﬁeld 2, with the same
number of samples taken per hectare.
Sampling for Water Stable Aggregates

Soil aggregates are sensitive to rough handling, and therefore need to be taken
prior to major ﬁeld operations that have a high probability of disrupting the soil structure.
A soil sample for soil aggregates was taken September 18, 2003 and September 11, 2004
for ﬁeld 1 and 2, respectively. This sample was composited from ﬁve random locations

taken within each of the study area sub-sections, deﬁned by a circle with a radius of 4 m

84

around each grid point to a depth of 0-10 cm using a trowel. Care was exercised not to
unduly compact the soil, and to protect the samples, cardboard boxes were used in
transporting and storing them. Extraneous materials such as plant residues and gravel
from the soil aggregate samples were removed with the use of an 8 mm sieve, and the
samples air-dried for a number of days until they equilibrated with atmospheric water
content. The soil sample was sieved using a layer of sieves, with the top having a screen
opening of 8, 6.2, 4, 2 mm and a catch pan at the bottom. Rocks, gravel and extraneous
materials remaining on the 8 mm were discarded. Soil aggregates that remained on top of
the 6.2, the 4 mm, the 2 mm and those contained in the catch pan were placed in separate
containers and labeled as the 6.2-4.0 mm, 4.0-2.0 mm and < 2 mm sieve sizes. Results
reported here referred to water stable aggregate stability of the 4.0 to 2.0 mm size
fraction. Water stable aggregate analysis was previously described in P0 (2007).
Chemical and Mechanical Analysis

Samples that passed through the 2 mm sieve during WSA sample preparation
were ground to powdery consistency for 2 minutes using a Shatterbox Rotary Grinder
(Spex Ind., Edison, NJ). Samples for total carbon and nitrogen analysis were prepared
from the ground soil by weighing 60 to 70 micrograms to a tin capsule. Crimped tin
capsules were submitted to the Stable Isotope Laboratory at the University of California
at Davis for total Carbon and Nitrogen analyses using a Europa Hydra 20/20 isotope ratio
mass spectrometer (Europa Scientiﬁc, Crewe, UK).

Samples were analyzed by A&L Great Lakes laboratories (Fort Wayne, IN) for
chemical and texture analysis, including determination of available phosphorous,

exchangeable potassium, magnesium, calcium, soil pH, buffer pH, cation exchange

85

capacity (CEC), percent base saturation of cation elements and texture components of
sand, silt and clay . Result of phosphorous analysis was expressed as a weak Bray (P1)
derived from Mehlich-3 through a regression equation (Calhoun et al., 2002).
Moisture Content and Hydraulic Conductivity Calibration

A Trime moisture probe (Ettlingen, Germany) was used to measure volumetric
moisture content across the ﬁeld. Prior to using the unit, a calibration procedure was
undertaken by ﬁlling up a 19-liter container with soil from AB 12 ﬁeld to a bulk density
of approximately 1.60 g (cm3)", adding a known volume of water to bring moisture
content to 24%. A lOO-watt incandescent bulb was placed a couple of centimeters from
the soil surface to induce evaporation from the soil surface. The entire setup was placed
on top of a Sartorius weighing scale (Model No. 3862 MP8-1; Bradford, Mass). The
Trime sensor prong was centrally inserted to 15 cm and attached to a Compaq Ipaq
pocket PC (Palo Alto, California) interfaced with a customized program to communicate
with the Trime unit. The communication software was set to obtain volumetric moisture
readings every 30 minutes for a total of approximately 260 hours. At the end of the
calibration period, two soil cores were taken from the top 10 cm of the soil, oven dried at
105°C for 24 hours, and the bulk density computed. A one-is-to-one correspondence
between gravimetric and volumetric moisture content was plotted with a R2 of 0.99.
In-situ Measurement

The volumetric soil water contents were measured across 108 grid points at the
ﬁeld 1 and 100 grid points at the ﬁeld 2 site, prior to soil sampling on September 17,
2003 and September 10, 2004, respectively. The effects of time on changing soil water

contents were compensated for by taking single volumetric soil water content readings at

86

each grid point during a single pass across the ﬁeld. . This procedure was repeated 5
times, and the average per grid point calculated to represent soil moisture status across

time for each measurement date.

Bulk density and Inﬁltration

In addition to the volumetric moisture content data, laboratory measurements
were conducted on soil cores obtained using an aluminum cylinder 7.62 cm in diameter
and 7.62 cm in height, for bulk density and inﬁltration. The cylinders were randomly
installed within 6 transects running from East to West in ﬁeld 1 for a total of 18 cylinders
for each transect and each cylinder represent an area of 0.05 ha. The area chosen for
coring was cleared of vegetation, and with the use of a special ramming device, the
cylindrical core was rammed vertically through the soil. Extreme care was exercised to
minimize as much as possible compacting or disrupting the natural soil structure. Once
the top rim of the cylindrical core was ﬂush to the external soil level and the internal soil
level was also ﬂush with it, a spade was used to extract the soil core. A ﬂat knife was
used to scrape off any excess soil beyond the top and bottom of the cylindrical core and
then a special box was used to transport the cores to the laboratory.

A mini-inﬁltrometer from Decagon (Pullman, Washington) with a 6.0 cm suction
was used to calculate the unsaturated hydraulic conductivity at each of the grid point
using the collected soil cores. The process involved ﬁlling up the device under water and
then plugging the top end with a rubber stopper. An absorbent tissue was used to adjust
the water level within the cylinder by making contact to the plastic porous membrane at
the bottom of the inﬁltrometer cylinder. The center of the soil core was leveled carefully,

making sure that the least possible disruption to the soil structure was done. The

87

inﬁltrometer was placed vertically at the center of the core and readings on the side scale
were obtained every ten seconds. After approximately 30 ml of water was dispensed by
the setup, the inﬁltration was terminated. In most instances, the inﬁltration rate of these
sandy soils was very rapid, making it impossible to determine the water level at speciﬁc
moments in time with accuracy. To remedy the situation, a camera setup was used
whereby pictures were snapped every ten seconds and readings were later taken off the
pictures. A software, exifextracter (BR software; Eiksmarka, Norway), was used to
extract information imprinted by the camera on the digital picture, reﬂecting the time
when the picture was taken, as well as other camera parameters. Extracted information
were then tabulated in an excel spreadsheet to act as values for the X-axis of time. To
increase the accuracy of the readings taken of the pictures, GrabIt software (Datatrend
Software; Raleigh, North Carolina) was used for length measurements plotted on the Y-
axis. This is a novel approach that can provide precision for rapidly inﬁltrating soils such
as those used in the study.

Calculation for unsaturated hydraulic conductivity was adopted from Zhang
(1997) where the square root of time was obtained for each set of inﬁltration readings,
and then plotted on the x-axis with volume of water inﬁltrated on the y-axis. The
resulting point distribution was ﬁtted into a second order polynomial and a trend line
equation projected using Excel v7 (Microsoft, Redmond, Washington). The coefﬁcient of
the squared independent variable (square root of time) was utilized to compute the
hydraulic conductivity of the soil. After the hydraulic conductivity measurements were
completed, the soil cores were dried in a convection oven for 24 hours, weighed and bulk

density computed using the core method (Blake and Hartge, 1986).

88

Spectral Pictures

Availability of a detailed lSt order map is not widespread. In 1998 the National
Cooperative Soil Survey Services of the US federal government has completed mapping
Michigan with a 1m resolution per pixel Digital Orthophotography Quarter Quads. These
images were computer altered and generated in a way that distortion normally associated
with digital images taken at a resolution of 1:12000 or lower tend to distort areas away
from the center of the image. Furthermore, the spectral band present in a DOQQ are not
the usual red, green and blue, but in its place, spectral data obtained from infrared, red
and green were used respectively. The possibility of analyzing the spectral distribution of
infrared, red and green for its correlation with yield monitor data was undertaken.

An Olympus 340R (Melville, New York) digital camera was utilized to obtain
red, green, and infrared images at each of the grid points in ﬁelds 1 and 2 on August 31,
2003 and August 28, 2004, respectively. The camera was attached to a palm pilot
handheld (M100, Milpitas, California) with an installed freeware program that controlled
the camera settings, ensuring uniformity in procedure, as well as ﬁring up the camera
(Palmshot v. 1). Two image monitoring events were captured where each consisted of a
normal, automatically obtained image and an infrared image using a pass through-ﬁlter.
Using the camera without a ﬁlter in front of it resulted in a picture with pixels composed
of a combination of the red, green and blue parts of the spectrum. To extract the green,
red and infrared parts of the spectrum, Paint Shop Pro 8 (Ottawa, Ontario, Canada) was
used along with a macro program to facilitate the processing of hundreds of pictures in a
short time. The extracted green, red and infrared band images were used as input to

IDRISI for windows v. 1 (Worcester. Ma) to compute the mean average value of the

89

image pixels. The average green, red and infrared values were tabulated in Microsoft
Excel for windows, imported to Statistical Analysis Software where relevant correlation

analyses were calculated (SAS 2006, Cary, NC).

Yield Monitor Data Collection

Potatoes were harvested with two types of tuber harvesters, one with a 4 row and
the other an 8 row capacity. The harvesters were equipped with an electronic yield
monitor (Harvest Master Yield Mapping System for Bulk Crops, HM-SOO;
HarvestMaster, Logan, UT) logging data every two meters (in ﬁeld 1) and every four
meters in ﬁeld 2 on average. A total of 1071 datapoints were generated for ﬁeld 1 and
roughly half that quantity of 633 for ﬁeld 2. A tandem of 1 eight row and 2 four row
harvesters can scoop up 16 rows as it moves from east to west and vice-versa. For each
grid point therefore two passes of the harvester was required to cover the total of 32 rows
representing each of the 108 grids in ﬁeld 1 and 100 grids in ﬁeld 2.

Statistical Analysis

Prior to variogram analyses, soil and plant factors were examined for normality.
Variogram analyses were conducted using ArcGlS v. 9 (Redlands, Ca). Vector distance
and direction for each grid point and the rest of the grid points were computed for both
ﬁeld 1 and 2. For each ﬁeld, the resulting vector distances were classiﬁed into lag width
equivalent to the modal, mean, or the shortest distance between points. The maximum lag
distance was approximately equal to one-half the longest distance from any point, and the
number of lag interval was the quotient between the maximum lag distance and the lag
width. For points belonging to a speciﬁc lag width, the mean of the variance divided by

two was computed and designated as the semivariogram (Isaaks and Srivastava, 1989).

90

An experimental variogram was computed by plotting a two-dimensional graph of lag
distance on the x-axis and semivariogram values on the y-axis, which was then ﬁtted with
a single or a combination of mathematical curve function. For each variogram point a
minimum of 30 to 50 pairs were required as a rule of thumb (Joumel and Huijbregts,
1978) and this requirement was satisﬁed by the computed variogram for each of the two
ﬁelds. Visual examination of the resulting semi-variogram dictates the use of any of the
popular mathematical curve function of exponential, gaussian, and spherical, to mention a
few.

Spatial dependence was evaluated using Chang et al. (1999) where the nugget to
sill ratio of less than 0.25 serves as an indicator of strong spatial dependence, with ratios
equivalent to between 0.25 to 0.75, and those with greater than 0.75 as having moderate

to weak spatial dependence.

9]

RESULTS AND DISCUSSION

Climatic and Soil Conditions

Overall soil characteristics showed that ﬁeld 2 had 50% more organic carbon and
44, 32, and 16% less potassium, magnesium and Bray-1 phosphorous, respectively, than
the ﬁeld 1 (Table 3.1). Field 2 had 8% more sand particles (823 gkg'l), 18% lower amount
of clay (57.4gkg") and 28% less silt (119.6gkg'l), as well as higher amounts of
macroaggregates (2250 um aggregate size), 1 mm, and 2 mm water stable aggregate
components, compared to ﬁeld 1. Average volumetric moisture content across the grids
was higher in ﬁeld 1 (15.8%) than in ﬁeld 2 (11.51%). Fields 1 and 2 were both
composed of Mancelona loamy sand with 0-2% slope (Mb). A small section of ﬁeld 1
was composed of a variant of the Mancelona loamy sand (Mc, 2-6% slope) as well as
Newaygo sandy loam (Nm, 2 to 6% slopes), although this particular soil series was not
present in the monitored section of the ﬁeld. Field 2 was one-quarter composed by
Mancelona, and three-quarter Gladwin loamy sand complex, with Palo sandy loam at a
slope of 0-2% (Ga). Outside the study area but still within the ﬁeld 2 boundary, was
Epoufette loamy sand and Ronald sandy loam (Ec, 0-2%). Mc and Mb were well-drained,
Ga was somewhat poorly drained, while BC was poorly drained to very poorly drained,
dark colored soil (Schneider, 1960).
Spatial Distribution of Yield

Global spatial trend analyses conducted on yield monitor data derived from the
ﬁelds 1 and 2 indicated the presence of a spatial pattern roughly following the location of

the two soil series present in ﬁeld 2 (Mancelona (Mb) and Gladwin (Ga), Figure 3.1.) and

92

a slight tilt towards a northeast-southwest trend in ﬁeld 1, reﬂecting the location of the
Mancelona 0-2% slope (Mb) soil series sandwiched by Mancelona 2-6% (Mc) slope on
both sides (Figure 3.2.) These spatial trends justiﬁed a second order trend removal
conducted on the yield data prior to geostatistical analyses (Johnston et al. 2001). The
removal of the physical trend allowed modeling of yield as a ﬁmction of distance, un-
confounded by other factors.

The magnitude of spatial correlation observed among the yield monitor data
points was different for the two ﬁelds studied. For ﬁeld 2 (Figure 3.3.), the nugget to sill
ratio of 0.05 indicated a very strong spatial dependence, while for the ﬁeld 1, the ratio
was moderate at 0.41 (Figure 3.4.). Chang et al. (1999) indicated that a nugget to sill ratio
of less than 0.25 serves as an indicator of strong spatial dependence, with ratios
equivalent to 0.25 to 0.75, and those with greater than 0.75, as having moderate to weak
spatial dependence. The range of spatial dependence among the yield monitor data points
was approximately 50 and 42 meters for ﬁelds 2 and 1, respectively. At this level of
spatial dependence, current sampling grid size of 1 ha (Akridge and Whipker, 2000;
Whipker and Akridge, 2001; Whipker and Akridge, 2002; Whipker and Akridge,
2003; Whipker and Akridge, 2004; Whipker and Akridge, 2005; Whipker and
Akridge, 2006) do not cause redundancy in information but if sampling intensiﬁcation
need to be increased to improve accuracy of agricultural input recommendation maps,
grid size have to be increased from 4 to 5 times the current level.

Use of a 1:12000 map resolution Digital Orthophotography Quarter Quads
(DOQQ) images with a 1m per pixel resolution supplemented by satellite imagery, has

been successfully used to improve the accuracy of coarse resolution satellite data when

93

compared to previously published soil surveys when classifying soils based on drainage
characteristics (Peng et al. 2003). Correlation analysis of pixel values for the green, red
and near infrared from DOQQ’s with yield in the ﬁeld 2 resulted in non-signiﬁcant
probability values. However, subdividing the datapoints based on the underlying soil
series remarkably improved the correlation coefﬁcient, with the infrared part of the
DOQQ spectral reﬂectance for yield data recorded from a Mancelona soil environment
improving from —16.74% (P=0.0959) to 58.39% (P=0.0027). A slight improvement was
observed for yield data recorded on the Gladwin soil series from a non-signiﬁcant —
16.74% to —25.18% (P=0.0282) (Table 3.3). On the other hand, in ﬁeld 1, signiﬁcant
correlations with yield were obtained with correlation coefﬁcient equal to —21.22%
(P=0.0306) for the infrared band; -42.10% for red (P<0.0001); and —38.08% for green
(P<0.0001) parts of the electromagnetic spectrum captured by the DOQQ (Table 3.2) for
the entire ﬁeld. No signiﬁcant departure from the above trend was observed when the two
soil series in ﬁeld 1 were split and analyzed separately. This result was not suprising as
the two soil series in ﬁeld 1 were closely related, being modest slope variations of
Mancelona. Spatial analysis of DOQQ pixels in the ﬁeld 1 resulted in an ill-deﬁned semi-
variogram, indicating the inability to pick up subtle short range soil characteristic
changes that could otherwise shed light on observed variability in yield.

Coleman and Tadesse (1995) used the three spectral bands found in a DOQQ to
relate observed soil properties in a study involving 99,877 hectares of land encompassing
26 different soil series. Signiﬁcant correlations were observed between clay and the
infrared, red and green bands of the DOQQ (34, 38 and 39% for red, green and blue,

respectively, all signiﬁcant at 1%) and drainage classes. The authors concluded that due

94

to time difference and season when the imagery were taken, it was an inadequate
resource to use in differentiating surface soils for purposes of predicting soil properties
such as organic matter. In the present study, use of in-season, near ground digital imagery
also resulted in low correlation values for the amount of clay and the green spectrum
(21% at P3005 and 32.5% at P3001, for ﬁelds 1 and 2, respectively). The amount of
carbon, though, was signiﬁcantly correlated with the green spectrum (-23.5 at P<0.05 and
—5 1.1% at P<0.0001 for ﬁelds 1 and 2, respectively; Table 3.5). These ﬁndings illustrated
the potential to document variability of critical soil properties indirectly through digital
camera imagery.
DOQQ and Selected Soil Attributes

Further analysis of data taken from ﬁeld 2 indicated a very signiﬁcant
improvement in the correlation coefﬁcient of several soil physical characteristics, when
background soil series information was integrated into the analysis. By splitting the
dataset into two groups corresponding to the location of the background soil series,
improvement in the correlation coefﬁcient was observed. Looking at the infrared band, an
improvement in correlation coefﬁcient from 4% (P=0.63) to 53% (P=0.06) was observed
in the amount of clay, and from 11% (P=0.25) to 63% (P=0.02) for the 0.25 mm water
sable aggregate size fraction for the Mancelona soil group series (Table 3.3). Also
demonstrating a signiﬁcant improvement in correlation coefﬁcient were the following:
the 1 mm water stable aggregate size fraction, the ratio between micro and macro-
aggregates, apparent electrical conductivity and the percent amount of soil carbon.
However, no improvement was observed for the Gladwin soil series, with the amount of

clay actually dropping several folds in terms of correlation relationship compared with

95

whole ﬁeld values. A similar trend was observed in other soil quality parameters
mentioned previously. Examination of the other two bands in the DOQQ did show
improvement in the correlation coefﬁcient, but not as dramatic as observed in the infrared
part of the spectrum.
Spatial Behavior of Selected Soil Attributes

The spatial distribution documented here is consistent with the primary driver of
soil processes appearing to be dependent on soil series variability, in the sampled areas.
Results of geostatistical analyses of mean weight diameter (MWD) and the 2 mm water
stable aggregate size fraction showed strong autocorrelation in ﬁeld 1, with a nugget to
sill ratio of 0.009 and 0.22, respectively. In ﬁeld 2, the spatial structure was weak for the
MWD and the 2 mm aggregate size fraction, with a 0.82 and 0.74 nugget to sill ratio,
respectively (Table 3.4; Chang et al., 1999). The larger ﬁeld 1 was composed of three
different soil series, but the sampled area for this study was composed of one type and
would therefore be expected to exhibit uniformity in attributes, particularly in soil
structure, as opposed to ﬁeld 2 which was composed of two markedly different soil series
across the study area. Spatial uniformity in ﬁeld lwas previously observed in the
variogram parameters derived for yield (Figure 3.4). The Mancelona 2 — 6% slope soil
series surrounded the sampled Mancelona 0 — 2% in ﬁeld 1, and played a role in the
observed spatial structure MWD, and the largest component of water stable aggregate,
the 2 mm aggregate size fraction (Figure 3.2).

The other soil structure related properties of 1, 0.25, 0.106 and 0.053 mm size
fractions had moderate spatial dependence (Table 3.4), as the nugget to sill ratio was

within the range of 0.25 to 0.74 for all of these characteristics (Chang et al., 1999). The

96

observed range of spatial dependence ranging from 27 m to 54 m (Table 3.4) were
roughly within the identiﬁed range of spatial dependence for yield of 50 m and 42 m for
ﬁelds 2 and 1, respectively (Figure 3.3 and 3.4).

The strength of spatial dependence have implications on the accuracy of
interpolated maps as well the size of grids one follows when sampling for a host of soil
physical, chemical and plant properties. In the presence of strong, well-deﬁne spatial
dependence, accuracy of interpolated map is high compared to instances when spatial
structure is weak (Kravchenko, 2003). Sample size is critical in classical hypothesis
testing as it inﬂuence the degrees of freedom needed for pattern detection. Presence of
autocorrelation over-estimate effective sample size leading to increased probability of
having false positive results (type I error; Segurado et al. 2006). Sampling grid size have
to be adjusted to prevent autocorrelation and the range of spatial dependence obtained
from a ﬁtted theoretical variogram is a good guide. Between the prevailing grid size of 1
ha (Whipker and Akridge, 2006) and the grid size used in this study of 0.05 ha,
variogram analyses of yield data from ﬁelds 1 and 2 indicated middle value of 0.18 and

0.25 ha as effective grid size to prevent autocorrelation respectively.

97

CONCLUSION

The current farmer practice of sampling and representing a grid size of 1 hectare
is not sufﬁcient to reflect the true status of soil structure and its associated dynamics that
affect plant growth and development. The extent to which intensiﬁcation of the sampling
regime will be required is expected to vary with ﬁeld heterogeneity. In this study, a
recommended guide for ascertaining the extent of new sampling regime was found to be
the spatial dependence of yield. This was shown to be 42 m and 50 m for ﬁelds 1 and 2,
respectively. Semi-variogram analyses of soil quality characteristics, notably water stable
aggregates, were shown to have moderate to strong spatial dependence, with ranges
roughly coinciding with the observed ideal grid size for yield at both ﬁelds. Analyses of
data using traditional statistical methodology of simple or multiple regressions can satisfy
the assumption of independence among samples.

Use of digital orthophoto (DOQQ) undertaken at a very high spatial resolution of
l m ground equivalent was shown to have potential as a valuable tool for improving the
documentation of variability in soil physical attributes such as amount of soil carbon,
0.25 and 1 mm water stable aggregate size fraction. However, the usefulness of this tool
has to be put in context by examination of soil survey data. A dramatic improvement in
correlation coefﬁcient seemed to be possible only when analysis was done on a per soil
type series, and not across series. Furthermore, the degree of difference among the
various soil series present may also play a role in the great improvements observed, as

was the case in the ﬁeld 2 where the Gladwin soil series was markedly different from the

98

Mancelona. In ﬁeld 1, the difference between Mancelona on a 0 to 2% slope was not

much when compared to the Mancelona soil series on a 2 to 6%.

99

REFERENCES

Akridge, J.T., and L.D. Whipker. 2000. Agricultural services and enhanced seed
dealership survey results. Staff Paper No. 00-04. Farm Chem.Magazine and
Center for Agricultural Business. Purdue Univ., West Lafayette, IN.

Blackmore, S. 2000. The interpretation of trends from multiple yield maps. Comput.
electron. Agric. 26: 37-51.

Blake, G. R., and K.H. Hartge. 1986. Bulk density. p. 363 — 375. In A. Klute (ed.)
Methods of soil analysis. Part 1. 2nd ed. Agron. Monogr. 9. ASA, Madison, WI.

Bongiovanni,R., and J. Lowenberg-Deboer. 2004. Precision Agriculture and
Sustainability. Precision Agric. 5:359—3 87.

Brock, A., S. M. Brouder, G. Blumhoff, and B. S. Hoﬁnann. 2005. Deﬁning yield-based
management zones for com—soybean rotations. Agron. J. 97:1115—1128.

Calhoun, F. G., J. M. Bigham, and B. K. Slater. 2002. Relationships among plant
available phosphorus, fertilizer sales, and water quality in Northwestern Ohio. J.
Environ. Qual. 31: 38—46.

Chang, J., D.E. Clay,U.C. Carlson, D. Malo, and SA. Clay. 1999. Precision farming
protocols: Part 1. Grid distance and soil nutrient impact on the reproducibility of
spatial variability measurements. Precision Agric. 1:277-289.

Coleman, T.L., and W. Tadesse. 1995. Differentiating soil physical properties from
multiple band DOQ data. Soil Sci. 160:81—91.

Colonna, LA. 2002. Accuracy of spatial estimation methods for site-speciﬁc agriculture
at different sample densities. M.S. thesis. Purdue Univ., Lafayette, IN.

Cooper, T. 2004. Title [Online]. Available at
http://wwwsoilsagri.umn.edu/academics/classes/soi12125/doc/s8chpl .htm;
accessed 2004). Regents of the University of Minnesota.

Davenport, J.R., C.A. Redulla, M.J. Hattendorf, R.G. Evans, and RA. Boydston. 2002.
Potato yield monitoring on commercial ﬁelds. HortTechnol. 12 (2):289-296.

Doberman, A., J.L. Ping, V.l. Adamchuk, G.C. Simbahan, and RB. Ferguson. 2003.

Classiﬁcation of crop yield variability in irrigated production ﬁelds . Agron. J.
95:1105-1120.

100

Flatman, GT. and AA. Yfantis. 1996. Geostatistical sampling designs for hazardous
waste sites. p. 779-801. In L.H. Keith. (ed) Principles of environmental sampling.
2nd ed. Am. Chem. Soc., Washington, DC.

Franzen, D.W., and TR. Peck. 1993. Soil sampling for variable rate fertilization. p. 81—
91. In R.G. Hoeft (ed.) Proc. Illinois Fert. Conf., Springﬁeld, IL. 25—27 Jan. Univ.
of Illinois, Urbana-Champaign.

Heisel, T. and AK. Ersboll. 1999. Weed mapping with co-kriging using soil properties.
Precis. Agric 1(1): 39-52.

Hergert, G.W. 1998. A futuristic view of soil and plant analysis and nutrient
recommendations. Commun.SoilSci.PlantAnal. 29 (1 1/14): 1441—1454.

Hergert, G.W., W.L. Pan, D.R. Huggins, J.H. Grove, and TR. Peck. 1997. Adequacy of
Current Fertilizer Recommendations for Site-Speciﬁc Management. p. In F.J.

Pierce and W.W. Frye (ed.) The site-speciﬁc management for agricultural
systems. ASA, CSSA, and SSSA, Madison, WI.

Homung, A., R. Khosla, R. Reich, D. Inman, and D. G. Westfall. 2006. Comparison of
site-speciﬁc management zones: Soil-color-based and yield-based. Agron. J.
98:407-415.

Isaaks, E.H., and RM. Srivastava. 1989. An introduction to applied geostatistics. Oxford
Univ.Press, NY.

Jaynes, D.B., T.C. Kaspar, T.S. Colvin, and DE. James. 2003. Cluster analysis of
spatiotemporal corn yield patterns in an Iowa ﬁeld. Agron. J. 95: 574-586.

Johnston, K., J.M. Ver Hoef, K. Krivoruchko, and N. Lucas. 2001. Using ArcGIS
Geostatistical Analyst. .

Joumel, AG, and C.J. Huijbregts. 1978. Mining geostatistics. Academic Press, London.

Kravchenko, A.N. 2003. Inﬂuence of patial tructure on accuracy of interpolation
methods. Soil Sci. Soc. Am. J. 67:1564—1571.

Kravchenko, A.N., and D.G. Bullock. 2000. Correlation of corn and soybean grain yield
with topography and soil properties. Agron. J. 92:75-83.

Kravchenko, A.N., and D.G. Bullock. 1999. A comparative study of interpolation
methods for mapping soil properties. Agron J. 91 (3):393-400.

Lopez-Granados, F., M. Jurado-Exposito, J.M. Pena-Barragan, and L. Garcra-Torres.

2005. Using geostatistical and remote sensing approaches for mapping soil
properties. Europ. J. Agro. 23:279—289

101

Lowenberg-DeBoer, J ., and A. Aghib. 1999. Average returns and risk characteristics of
site speciﬁc P and K management: eastern corn belt on-farm trial results.
J .Prod.Agric. 12(2):276-282.

Lowenberg-DeBoer, J ., and SM. Swinton. 1997. Economics of site-speciﬁc management
in agronomic crops. p. 369-396. In F.J. Pierce and W.W. Frye (ed.) The ite-
speciﬁc management for agricultural systems. ASA, CSSAand SSSA, Madison,
WI.

Morgan, G.D., A.E. MacGuidwin, J. Zhu, and K. Larry. 2002. Population dynamics and
distribution of root lesion nematode (Pratylenchus penetrans) over a three-year
potato crop rotation binning. Agron. J. 94:1146-1155.

Mueller, T.J., F.J. Pierce, 0. Schabenberger, and DD. Wamcke. 2001. Map quality for
site-speciﬁc fertility management. Soil Sci. Soc. Am. J. 65:1547-1558.

Peng, W., D. B. Wheeler, J. C. Bella, and M. G. Krusemarka. 2003. Delineating patterns
of soil drainage class on bare soils using remote sensing analyses. Geoderrna 115:
261— 279.

Pierce, F.J., and DD. Warncke. 2000. Soil and crop response to variable-rate liming for
two Michigan ﬁelds. Soil Sci. Soc. Am. J. 64:774-780.

Pocknee, 8., BC. Boydell, H.M. Green, D.J. Waters, and CK. Kvien. 1996. Directed soil
sampling. p. 159-168. In P.C. Robert et al. (eds.) Proc. 3rd International

Conference on Precision Agriculture. Location. Speciﬁc Dates ASA, CSSA, and
SSSA, Madison, WI.

Pozdnyakova, L., P.V. Oudemans, M.G. Hughes, D. Gimenez. 2002. Estimation of
spatial and spectral properties of phytophthora root rot and its effects on cranberry
yield. Comput. Electron. Agric. 37:57-70.

Redulla, C.A., J.R. Davenport, R.G. Evans, M.J. Hattendorf, A.K. Alva, and RA.
Boydston. 2002. Relating potato yield and quality to ﬁeld scale variability in soil
characteristics. Am. J Potato Res. 79:317-323.

Sadler, E.J., W.J. Busscher, P]. Bauer, and D.L. Karlen. 1998. Spatial scale requirements
for precision farming: a case study in the southeastern USA. Agron J. 90 (2):]91-
197.

SAS Institute Inc. 2006. Base SAS® 9.1.3 procedures guide, 2nd ed., Volumes 1-4. Cary,
NC.

Schneider, 1. 1960. Soil survey, Montcalm County, Michigan. USDA-SCS in cooperation
with Michigan Agric. Exp. Stn.

102

Segurado, P., M. B. Araujo, and W. E. Kunin. 2006. Consequences of spatial
autocorrelation for niche-based models. J. Appl. Ecol. 43:433—444.

Vetsch, J.A., G.L. Malzer, P.C. Robert, and DR. Huggins. 1995. Nitrogen speciﬁc
management by soil condition: managing fertilizer nitrogen in corn. p. 465—474.

In P.C. Robert et al. (ed.) Site-speciﬁc management for agricultural systems. ASA
Misc. Publ. ASA, CSSA, and SSSA, Madison, WI.

Whipker, L.D. and J .T. Akridge. 2001. Precision agricultural services and enhanced seed
dealership survey results. Staff Paper No. 02-02. CropLife Magazine and Center
for Food and Agricultural Business,Purdue Univ., West Lafayette, IN.

Whipker, L.D. and J .T. Akridge. 2002. Precision agricultural services dealership survey
results. Staff Paper No. 02-02. CropLife Magazine and Center for Food and
Agricultural Business,Purdue Univ., West Lafayette, IN.

Whipker, L.D. and J .T. Akridge. 2003. Precision agricultural services dealership survey
results. Staff Paper No. 3-10. CropLife Magazine and Center for Food and
Agricultural Business, Purdue Univ., West Lafayette, IN.

Whipker, L.D. and J .T. Akridge. 2004. Precision agricultural services dealership survey
results. Staff Paper # 04-07. CropLife Magazine and Center for Food and
Agricultural Business. Dept. of Agric. Eco.s Purdue University, West Lafayette,
IN.

Whipker, L.D. and J .T. Akridge. 2005. Precision agricultural services dealership survey
results. Staff Paper # 05-11. CropLife Magazine and Center for Food and
Agricultural Business. Dept. of Agric. Eco., Purdue Univ., West Lafayette, IN.

Whipker, L.D. and J .T. Akridge. 2006. Precision agricultural services dealership survey
results. Staff Paper # 06-10. CropLife Magazine and Center for Food and
Agricultural Business. Dept. of Agric. Eco., Purdue University, West Lafayette,
IN.

Wollenhaupt, N.C., D.J. Mulla, and C.A. Gotway Crawford. 1995. Soil sampling and
interpolation techniques for mapping spatial variability of soil properties. p. 19-
54. In F.J. Pierce and E.J. Sadler (ed.) The state of site speciﬁc management for
agriculture. SSA and ASA, St. Louis, MO.

Zhang, R. 1997. Determination of soil sorptivity and hydraulic conductivity from the disk
inﬁltrometer. Soil Sci. Soc. Am. 161: 1024-1030.

103

 

’l'able 3.1 Basic statistics for potato yield. speciﬁc gravity and regressor variables classiﬁed into chemical. physical and spectral factors taken from ﬁeld I and 2 in
Montcalm County. Michigan in 2003 and 2004 respectively.

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

f 'Variablesr M
ppm ase Saturation, %
ClassificationFieldStatI B_pH Ca K Mg P1 ECa Ca H K Mg pH CEC C, 95
f 6.87 502.8167.7107.214.7 — 46.7 28.3 8.0 16 8 5.7
1 s.e. 0.01 11.1 4.7 1.9 0.2 — 0.6 0.8 0.2 0.3 0.0 0.1 0.01
f 6.54 575.5 92.3 72.614.0078 43.1 37.3 4.2 10.4 5.5 5.9 1.05
Chemical 2 s.e. 0.13 25.2 2.2 3.2 1.20.01 1.0 1.3 0.1 0.3 0. 0.2 0.02
Aggregate size, pm Texture Components, %
Elev MWD 53 106 2501000 2000 2250 Sand Silt Clay MC k BD
x 278.1 1.21 13.3 12.845.518.8 9.7 73.9 76.4 16.6 7.1 15.885.01.46
1 s.e. 0.3 0.02 0.2 0.2 0.3 0.3 0. 0.3 0.2 0.2 0.1 0.1 3.90.01
F 251.7 2.18 7.5 6.426.222.5 37.4 86.1 82.3 12.0 5.7 11.5 —
Physical 2 s.e. 0.1 0.01 0.1 0.2 0.4 0.4 0.5 0.3 0.2 0.2 0.1 0.3 — —
Spectral Band
Unadjusted Adjusted
G R IR G R IR Vegarea
X 102.1 99.0 32.4 51.5 49.3162 Z17.2
1 3.62. 0.5 0.5 2.0 1.5 1.4 16277730;
f 87.3190.0160.1 4.8 9.1 8.76747}
Spectral 2 s.e. 0.6 1.1 1.2 0.3 0.5 0.5 716127
Yiel SG
77 22.8 —
1 s.e. 0.2 —
)7 17.41.0824
Dependent 2 s.e. 0.3 0.0003J

 

 

 

 

 

 

 

1‘ Variables : BﬂpH : buffer pH :Ca : calcium ; K : potassium ; Mg : magnesium ; P1 ,, phosphorus (Bray l) : 15C,l : aparent electrical conductivity (shallow 0-30 cm); H e hydrogen ; CEC : cationic
exchange capacity (mqu’l); lilev 7’ elevation; MWD : mean weight diameter, mm; MC : volumetric moisture content: k ,7,- hydraulic conductivity, mmhr"; BD T bulk density; Vegarea 1’- vegetated area. %:
Yield ’- tonsha": SG 1 speciﬁc gravity ; G .: green ; R 2 red ; 1R : infrared.

i f : mean; so. : standard error.

104

 

Table 3.2 Selected soil properties of ﬁeld 1 showing signiﬁcant correlations with digital

onhophou3quanerquad.

 

Soil Property

Infrared

 

Microaggregates (%) 0
Volumetric Moisture Content 0
0.053 mm 0
0.250 mm 0
macrozmicro aggregate ratio —0.
Infiltration, mmhr'l —0.
Sand (%) —0.
Silt (%) 0.
Clay (%) 0.
OM (%) 0.
yield (Mgha'U -0
Cu —0
Ru 0.
Iru 0.
Ira 0.

2110 *
1297ns
2427**
2054 *
0801ns
2534**

.2122**
.1483ns

0403ns
0123ns
0173ns

 

.2004 *
.0398ns
.1972 *
.0161ns

Red Green
0.1787 * 0.1761 ns
0.0102 ns 0.0600 ns
0.2154 ** 0.2125 **

-0.1426 ns -0.1256 ns
-0.1983 * —0.1965 *
—0.1761 ns -0.1532 ns
—0.3433 ** —0.3339 **
0.3580 ** 0.3312 **
—0.0320 ns 0.0056 ns
0.4365*** 0.3984 **
-0.4210*** -0.3808***
—0.4419*** -0.4071 **
—0.1272 ns —0.1177 ns
0.2443 ** 0.2096 *
0.2614 ** 0.2104 *

 

*** - signiﬁcant at <40.0001; ** - signiﬁcant at (40.05; * - significant at <0.10; and ns - not signiﬁcant.

105

 

Table 3.3 Selected soil properties for ﬁeld 2 as a whole (whole). datapoints found on Mancelona (Mb) and Gladwin (Ga) soil series sampled in fall 2004 showing
dnrdaUOnsxvhhcﬁghalonhophouiquanerquadinmgetakenin1997.

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

VARIABLE INFRARED RED GREEN

WHOLE Ga ONLY Mb ONLY WHOLE Ga ONLY Mb ONLY WHOLE Ga ONLY Mb ONLY
Yield (tonshail) —0.167ns -O.252** 0.584** —0.023ns —0.025ns —0.122ns —0.061ns ~0.053ns —0.065ns
86 —0.189* —o.257** 0.597* 0.059ns 0.053ns 0.013ns 0.117ns 0.108ns 0.116ns
Cover (%) ~0.336** —0.351** —0.166ns 0.114ns 3.165ns —0.262ns 0.104ns 0.148ns —0.283ns
Volumetric MC —o.217** —o.230** 0.133ns —0.046ns —0.077ns 0.14ns —0.049ns i0f086ns 0.157ns
m”. 0.115ns 0.015ns 0.634** 0.048ns 0.124ns -0.265ns 0.143n_s_ 0.216** —0.165ns
ﬂﬂ —0.131ns —0.063ns —0.470ns 0.014ns —0.063ns 0.303ns —0.117ns —0.208* 0.237ns
Ratio —0.006ns 0.052ns —0.362ns 0.028ns —0.6€§H§7 0.430ns —0.100ns —0.182* 0.316ns
EC; —0.263** —0.3oo** 0.531* —0.088ns —0.l40ns —0.006ns —0.080nsg—0.143ns 0.073ns
Elevation, m 0.220** 0.252** 0.124ns —0.031ns —0.004ns —0.327ns 0.091ns g~0i.1’40ns —0.384ns
iPctCarbon —o.227** —0.184* —0.492* —0.151ns —0.185* —0.015ns —O.260** —0.285** —0.162ns
Sand (%) 0.137ns 0.175ns _:g.361ns 0.078ns 0.035ns 0.486* 0.102n37740;070ns o.502*
4s3i§44514_~44_4::720.157ns _;0.168ns —0.022ns -0.156ns —0.115ns —0.417ns —0.210** ~0.169ns —0.505*
Clay (%) 0.048ns -0.010ns 0.531* 0.166* 0.174ns 0.102ns o.231*+ 0.218** 0.242ns
Gu 0.436*** 0.465*** 0.174ns 0.056ns 0.010ns 0.471ns 0.189* 0.164ns 0.583**
Ru 0.386*** 0.436*** -0.314ns 0.09aH§770.083ns o 623** 0.088ns 0.078ns 0.643**
1ru 0.361** 0.402** —0.366ns 0.079ns 0.084ns 0.493* 0 062ns #0.078ns 0.431ns
Ga —0.185* —o.213** 0.165ns 0 067ns 470.140ns —0.443ns 0.040ns 0.098ns ~0.442ns
’Ra " —O.248** —0.268** —0.001ns 0.111ns _>0.179* —0.359ns 0.081ns ::0:135nsg -0.369ns
Ira “7 —0.182* —0.204* 0.102ns 0.10§g§:70.i78:4_—0.376ns 0.081ns_rpgia§gag—o.37ins

 

*** — signiﬁcant at (0.0001; ** — signiﬁcant at <0.05; * - signiﬁcant at <0.10; and us - not signiﬁcant

106

 

 

Table 3.4 Variogram profile for soil structure related properties of ﬁeld 1 and 2.

 

 

Field Properties Range (m) Nugget Sill N:S
1 2 mm 49.34 1.92 8.38 0.23
1 m 27.86 4.34 8.00 0.54

0.250 mm 36.10 3.31 10.50 0.31

0.106 mm 53.00 2.33 3.58 0.65

0.053 m 27.72 1.46 2.12 0.69

MWD 41.27 0.00 0.02 0.01

2 2 mm 30.23 17.11 23.05 0.74
1 mm 54.26 6.81 9.34 0.73

0.250 m 40.77 8.65 14.47 0.60

0.106 mm 35.66 0.40 1.12 0.36

0.053 mm 30.28 0.46 0.90 0.50

MWD 30.13 0.01 0.01 0.82

 

107

 

lafde 3.5a Pearson correhuion nrauix ofselccted soiL plantzuid spectralproperﬁcst>flield lin hiontcalnr.hAichigan.

 

 

ﬂ 1 c+ MC Elev K Clay 0 . 2 5 0mm MWD c;unadj Vegcover G/Runadj R/ IRmadj
Iield —0.049ns 0.235 * 0.327** 0.595*** 0.474“H 0.120ns 0.202* 0.654*** 0.056ns 0.224* 0.423***
C 0.035n5 —0.155ns 0.079ns 0.117ns 0.220* —0.178ns —0.235* —0.067ns ~0.166ns -0.055ns
MC 0.210* 0.327** 0.223* 0.092ns —0.058ns 0.107ns 0.008ns 0.189ns 0.156ns
Elt/ 0.289** —0.075ns —0.023ns 0.003ns 0.199* —0.023ns 0.186ns 0.354**
K 0.279** 0.154ns 0.064ns 0.447*** 0.129ns 0.167ns 0.333**
Clay 0.320** ~0.051ns 0.210ns 0.128ns 0.054ns 0.077ns
250mm —0.694*** 0.247* 0.037ns 0.146ns 0.271**
MWD 0.054ns ~0.021ns —0.113ns —0.148ns
Gnam 0.070ns 0.490*** 0.587***
Vest-m. 0.1.09ns —0.127ns
G/Rmmcj —0.004ns

 

*** — signiﬁcant at *0.0001; ** - signiﬁcant at -<0.05; * — signiﬁcant at /0.10; and us — not significant
”1‘ C ’ % total soil carbon : MC : volumetric moisture content ; Elev : elevation ; K : potassium ; Clay ' 0/0 Clay :
Gum, 1' Green band unadjusted; Vega,”r : vegetative cover; G/RmdJ : green red unadjusted band ratio; R/lRumdl red infrared unadjusted band ratio

108

(1.250mm : l - 2 0.250 sized water stable aggregate; MWD : mean weight diameter;

 

 

 

lable 3.5b Pearson correlation matrix of selected soil. plant and spectral properties from field 2 in Montcalm. Michigan.

 

 

Field 2 c1 MC Elev K Clay 0 . 250mm MWD c;unadj Vegcover G/Runadj R/ IRde Eca
Yield 0.171ns 0.488*** —0.150ns —0.034ns 0.049ns 0.255* —0.429*** —0.405*** 0.349** 0.101ns 0.463*** 0.697***
C 0.575*** ~0.194ns 0.280** —0.265H —0.203* —0.218* —0.511*** 0.273** —0.139ns 0.127ns 0.485***
MC —0.334** 0.232* 0.031ns —0.115ns ~0.321** —0.486*** 0.396*** 0.191ns 0.321** 0.781***
Elev —0.281** 0.093ns 0.224* —0.066ns 0.347** —0.146ns 0.074ns 0.053ns —0.418***
K 0.122ns —0.326** 0.139ns —0.210* 0.122ns 0.045ns —0.039ns 0.239*
Clay *0.010ns 0.083ns 0.325** 0.040ns 0.425*** —0.040ns 0.025ns
250mm e0.629*** 0.142ns 0.086ns 0.234* 0.020ns 0.050ns
MWD 0.216* —0.300** —0.151ns —0.175ns —0.377**
GWHM —0.427*** 0.589*** —0.371** —0.523***
Vegmuwr 0.092ns 0.414*** 0.510***
G/RMMA] ~0.099ns 0.310**
R/IRuam 0.391**

 

*** - signiﬁcant at ”0.0001; ** — signiﬁcant at <0.05; * - signiﬁcant at (10.10: and ns — not significant
- % total soil carbon : MC " volumetric moisture content ; Elev 7,. elevation ; K - potassium : Clay , % Clay ;

.1. (.

(1mm 7 Green band unadjusted: Vegcmcr f vegetative cover; G/Runadj : green red unadjusted band ratio: R/lRmW :' red infrared unadjusted band ratio; EC,l ’ apparent electrical conductivity.

109

 

0.250an : l — 2 0.250 sized water stable aggregate: MWD 7 mean weight diameter;

 

.830 5 mm :2:ch .motom mom 36V :82 Beam 2mm 98 team 552 52630 use 33: 9:8 x982 «co—3:32 mo 3:3er
@152? 92: motom :8 c3 pomOdEtomzm N Eon do 8582 wcmzﬁop owes: peso 3550 Easewoeosmonto 1.“wa _.m Eswi

.... latrtﬂ .7

.2.

23:53.1.955. :. L.t:...::.::::z.....t§.ﬂn§§ﬁu§3.3m3.3:33.333: .
.. r . v , ..

 

 

Figure 3.2. A section of a digital orthophotography quarter quad (DOQQ) image detailing
location of ﬁeld 1 superimposed by soil series map showing presence of Mancelona
loamy sand with 0 to 2 percent slope (Mb), Mancelona loamy sand with 2 to 6 percent
slope (Me), and Newaygo sandy loam (Nm).

111

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

0 1.95 3.90 5.85 7.80 9.75 11.7 13.65 15.60
Distance, 11 x 10 (m)

Figure 3.3 Semi-variogram (y) plot for yield data taken from ﬁeld 2 during the fall of
2004 ﬁtted with exponential model (white line) showing a strong spatial dependence
among data taken within a range of approximately 50 m.

-l

 

 

 

 

 

 

 

 

  

 

 

 

 

 

71:10
6.60
5.28
3.96 }",L‘ 5. _
2.64 '1‘, :11“
1.32 A” ]

 

 

 

 

 

 

 

 

0 2.2 4.4 6.6 8.8 11.0 13.2 15.4 17.6
Distance, 11 x 10 (m)

Figure 3.4 Semi-variogram (y) plot for yield data taken from ﬁeld 1 during the fall of
2003 ﬁtted with exponential model (white line) showing a moderate spatial dependence
among data taken within a range of approximately 42 m.

112

       

uiiiiiijijjijjjiiiji

2

S

[iii