‘33.:
1!.

u

I .

I ..

‘1.
124:,
:

ea
.1

 

 

 

 

git»:
..R3.Vzh:.r
. i
.LXh, ~
§‘* ‘ r
“:2
It: {.1
y’- I" =
gififinfiwmummfi fl
5. £3.“

 

u... .1. 7.
SE 1!... :5
mu..| .32.). .5 a:
. .9053 fr 138“)...
girlfh. 1339.. ‘3-
it. 31113.. 3 .I 1.
‘ ya: .4 gold“: 1 ’2‘;
It! 9. f. l"
AIL... Exnjurtnlel :..vl..:..xs
:4 £551.. 53...}... I .
a: . :x‘) 5...}. attvw-x. gt... |v
:1. 82.135.323. .nrts.«'._ It:
.31.;3uflnydg’tt f it;
.hr’sl.r,¢.. . Pia-an:

$.2-

zrnvifui a.
2 l. t... .
3.133.!!!

 

(3.3.1.15!
5.11:... .n. .
. . a:
. 20.5.13

. $9.18.
63.91...

. :b

:7. L5...

t3\.r.:..:.30) ti:
. . 3.5.3.5.

 

wfifi. :2 a %$§§y&gh‘

O

 

1pm.; . 15km Lmzufi

\l'

3: LL.“ ii.
.. .. k. t
v Liana... aka:

Illllllllllllllllll

2w? LIBRARY

Michigan State
University

 

 

 

 

This is to certify that the

dissertation entitled

Investigations on Statistical Analysis of Turfgrass
Rating Data, Localized Dry Spots of Greens, and
Nitrogen Application Techniques for Turf

presented by

Douglas Edward Karcher

has been accepted towards fulfillment
of the requirements for

 

Ph.D. degree in 11M Sciences

{item

Major profegsor

Datejl/Z’T/OO

MS U is an Affirmative Action/Equal Opportunity Institution 0-12771

 

v—‘I —‘v——.r "i——v

Mt

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

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

11/00 c-JCIRCIDatoDua.p65-p.14

INVESTIGATIONS ON STATISTICAL ANALYSIS OF TURFGRASS RATING
DATA, LOCALIZED DRY SPOTS OF GREENS, AND NITROGEN
APPLICATION TECHNIQUES FOR TURF

BY

Douglas Edward Karcher

A DISSERTATION

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

DOCTOR OF PHILOSOPHY

Crop and Soil Sciences

2000

ABSTRACT

INVESTIGATIONS ON STATISTICAL ANALYSIS OF TURFGRASS RATING
DATA, LOCALIZED DRY SPOTS OF GREENS, AND NITROGEN
APPLICATION TECHNIQUES FOR TURF

By

Douglas Edward Karcher

Three separate turfgrass research topics were addressed: 1) statistical
analysis of visual rating data, 2) localized dry spots (LDS) on putting greens, and
3) nitrogen fertilization using high pressure water injection cultivation (W lC) on
putting green turf. 1) Visual ratings are often used by researchers to assess
turfgrass quality. The resultant data are ordinal and often violate assumptions
necessary for analysis using ANOVA. Valid analysis of ordinal data is possible
with the proportional odds model (POM) and gives the researcher nearly the
same amount of information on treatment effects as ANOVA. A Rating Data
Analysis File Package was developed that allows researchers to analyze rating
data with the POM, without needing to program statistical software.

2) Consistent control of LDS is difficult. The effects of flutolanil (a fungicide), a
soil wetting agent, and WIC on LDS control, and the cause of LDS were
investigated. All treatments showed some control of LDS in 1998 or 1999, but
results were variable. Where characteristic LDS occurred, soil at the center of
the dry spot, and at a 1 cm depth, was most non-wettable. Slides buried into the

soil (1 cm beneath thatch) at the edges of the dry spots had the greatest amount

of fungal hyphae, consistent with the hypothesis that fungi may be involved in
development of LDS. 3) Subsurface nitrogen fertilization has increased nitrogen
use efficiency by crops in the food and forage industries. However, the
equipment needed for subsurface fertilization in turf was unavailable until the
recent development of WIC. Studies were conducted to compare putting green
turf responses between surface applications nitrogen and nitrogen injected via
WIC to 7.5 and 15 cm. An additional study was conducted to evaluate the effects
of several WIC nitrogen application methods on surface uniformity, since nitrogen
injection with traditional nozzles sometimes causes unacceptable striping of
putting green turf. Subsurface placement of nitrogen increased clipping yields
and nitrogen content in leaf tissues in 1997, and improved visual color and
general turf quality in 1997 and 1998 compared to surface applications. An
application of WIC on turf receiving surface applications of nitrogen did not
significantly affect any turf evaluation. Injecting nitrogen with a Toro HydroJect
3000® at a 7.5 cm by 2.5 cm spacing and with the roller washers turned on
significantly reduced turf striping following application compared to other nitrogen

injection techniques.

To all of the wonderful people at Michigan State University who enriched my

graduate education beyond all expectations

ACKNOWLEDGMENTS

I sincerely thank my major advisor, mentor, and good friend, Dr. Paul
Rieke. His balance of academic guidance and willingness to let me explore my
particular research interests made the graduate experience truly rewarding. In
my future endeavors I will aspire to work with the integrity and graciousness that
Dr. Rieke demonstrated during my tenure at MSU. I also want to thank the
remaining members of my guidance committee, Dr. Frank Dazzo, Dr. Eldor Paul,
and Dr. Oliver Schabenberger. Each contributed a unique aspect to my program,
which in turn, increased my breadth of scientific knowledge. I am grateful to the
Toro Company for funding my entire graduate education. Toro always held
professionalism in high regard and is a model company among those funding
university research. My experience at MSU would have been dull if I did not
have the good fortune to meet John Sorochan. John could not have been a
better friend and for that I thank him and will miss him dearly. To all the other
graduate students, faculty, staff, and other friends I made while at MSU, thanks
for the great fun. Finally, I want to thank my family, present and future, for their
unyielding support, even when it seemed like graduation day may never arrive. I
am excited to join Wendy in the Ozarks and look forward to spending the rest of

our lives together.

TABLE OF CONTENTS

List of Tables ................................................................................................... viii
List of Figures ..................................................................................................... x

CHAPTER 1: The Proportional Odds Model: A Statistically Sound Method

To Analyze Turfgrass Rating Data ............................................. 1
Introduction ................................................................................................ 3
Statistical Methods ..................................................................................... 6
Rating Data Analysis File Package ............................................................ 8
Example Usage ................ ...................................................................... 11

Experimental Design ..................................................................... 11
Generating SAS Code .................................................................. 11
Summary of Output ....................................................................... 17
Conclusions ............................................................................................. 21
Bibliography ............................................................................................. 23

CHAPTER 2: Investigating causes and cures for localized dry spots on

creeping bentgrass putting greens .......................................... 24
Introduction .............................................................................................. 26
Materials and Methods ............................................................................ 28

Experimental Area ........................................................................ 28
Treatment Design ......................................................................... 29
Treatment Evaluations .................................................................. 31
Statistical Analysis ........................................................................ 32
Results and Discussion ........................................................................... 33
Preventative Study ........................................................................ 33

vi

Curative Study .............................................................................. 42

Isolation Study .............................................................................. 46
Conclusions ............................................................................................. 49
Bibliography ............................................................................................. 50

CHAPTER 3: The Effects of Nitrogen Placement and Water Injection

Cultivation on Creeping Bentgrass Putting Greens ............... 52
Introduction .............................................................................................. 54
Nitrogen Injection - Management Practice Study .......................... 57
Application Uniformity Study ......................................................... 60
Statistical Analyses ....................................................................... 62
Results and Discussion ........................................................................... 63
Nitrogen Injection - Management Practice Study ......................... 63
Application Uniformity Study ......................................................... 75
Conclusions ............................................................................................. 81
Bibliography ............................................................................................. 83
APPENDIX A — “Turfratesas” ................................................................. 85

vii

Table 1.

Table 2.

Table 3.

Table 4.

Table 5.

Table 6.

Table 7.

LIST OF TABLES

Effects of WIC, SWA, and flutolanil on quality, wilt, and soil moisture of
a creeping bentgrass puttting green. ................................................... 35
Interaction effects of WIC and evaluation date on volumetric soil
moisture. Preventative study, 1998 ...................................................... 39
Effects of WIC, SWA, flutolanil, and sample depth on water drop
penetration time. .................................................................................. 41
Water drop penetration times as affected by WIC x depth interaction.
Preventative study, 1998. .................................................................... 42
Water drop penetration times as affected by flutolanil x depth
interaction. Preventative study, 1999. ................................................. 42
ANOVA and main effects of WIC, SWA, and flutolanil on turf quality and
soil moisture. Curative study, 1998. ..................................................... 44
Water drop penetration times and cumulative hyphae length as affected

by surface location and depth relative to LDS. Isolation study, 1998.. 47

Table 8. Effects of location x depth on water drop penetration time. Isolation
study, 1998. ......................................................................................... 49

Table 9. Summary of treatments comprising the nitrogen injection - management
practice study .......................................................... _. ............................ 58

Table 10. Treatments comprising the application uniformity study. ................... 61

Table 11. Treatment main effects, and nitrogen placement contrast for turf

quality, color, clipping yields, and nitrogen content. 1997. ............... 64

Table 12. Treatment main effects, and nitrogen placement contrast for turf

quality, color, clipping yields, and nitrogen content. 1998. ............... 65

Table 13 . Clipping yields as affected by management practice x nitrogen rate.

1997 and 1998. ................................................................................. 74

Table 14. Leaf nitrogen content as affected by management practice x nitrogen

rate. 1998. ......................................................................................... 75

viii

Table 15. Application method effects on turf color, quality, and stripe ratings.
1997 and 1998. ................................................................................. 77

Figure 1.
Figure 2.

Figure 3.

Figure 4.

Figure 5.

Figure 6.

Figure 7.

Figure 8.

Figure 9.

' LIST OF FIGURES

Rating data file created in MS Excel ................................................... 10

SAS code generating form, completed with information from nitrogen
application method study. This form is from the “Rating Data Analysis“
web page at "http://wwwmsu.edu/~karcherd/ratings”. ....................... 13

Waming message that appears in the Log window of SAS when errors
occur during maximum likelihood calculation. .................................... 14

Textbox used to supply a reduced model. In this case, the NSOURCE
x NRATE interaction term has been dropped from the full model ....... 15

Output generated from the “Rating Data Analysis“ web page form.
This code can simply be pasted into SAS v. 6.12 for expedient analysis
using the POM .................................................................................... 16

Tests of fixed effects produced by RDAFP. Both application method
(NSOURCE) and nitrogen rate (NRATE) effects were highly significant
when analyzed by the POM and ANOVA. .......................................... 17

Probability of each NSOURCE treatment to be rated into each quality
rating category, as well as mean separation tests from the POM and
ANOVA. .............................................................................................. 18

Probability distribution chart created automatically by RDAFP.
Cumulative probabilities are shown on the y-axis, whereas individual
category probabilities (greater than 5%) are labeled within each bar
section. ............................................................................................... 19

Textbox generated from checking the contrast option on the RDAFP
web page. Textbox input has identical syntax to the contrast statement
used in proc glm of SAS. .................................................................... 20

Figure 10. Contrast test from the RDAFP. Here, treatments receiving surface

nitrogen were significantly different (P < 0.001) than turf injected with
nitrogen ............................................................................................ 21

Figure 11. Quality rating probability distributions as affected by flutolanil x time.

Preventative Study, 1998 ................................................................. 36

Figure 12. Quality rating probability distributions as affected by time.

Preventative Study, 1998 ................................................................. 36

Figure 13.

Figure 14.

Figure 15.

Figure 16.

Figure 17.

Figure 18.

Figure 19.

Figure 20.

Figure 21.

Figure 22.

Figure 23.

Quality rating probability distributions as affected by SWA x flutolanil.
Preventative study, 1999. ................................................................ 38

Wilt rating probability distributions as affected by WIC, SWA,
flutolanil, and time. Preventative LDS study, 1999 .......................... 38

Quality rating probability distributions as affected by WIC, SWA,
flutolanil, and rating date. Curative study, 1998. ............................. 45

Quality rating probability distributions as affected by WlC x SWA.
Curative study, 1998. ....................................................................... 45

Images taken 30 seconds after water drop placement just beneath
the thatch layer at the inside, edge, and outside of a dry patch. ...... 47

Stained hyphae from a section of a buried slide at the edge of a dry
patch, immediately beneath the thatch layer. Image is magnified 250
times. ............................................................................................... 48

Color rating probability distributions for management practice and
nitrogen rate main effects. 1997. ..................................................... 66

Color rating probability distributions for management practice and
nitrogen rate main effects. 1998. ..................................................... 66

Color rating probability distributions for the management practice x
nitrogen rate interaction. 1997. ............................................ 69

Color rating probability distributions for the management practice x
nitrogen rate interaction. 1998. ....................................................... 69

Quality rating probability distributions for management practice and
nitrogen rate main effects. 1997. ..................................................... 70

Figure 24 Quality rating probability distributions for management practice and

Figure 25.

Figure 26.

Figure 27.

Figure 28.

nitrogen rate main effects. 1998. ..................................................... 70

Quality rating probability distributions for the management practice x
nitrogen rate interaction. 1997. ....................................................... 72

Quality rating probability distributions for the management practice x
nitrogen rate interaction. 1998. ....................................................... 72

Color rating probability distributions for application method effects.
1 997 ................................................................................................. 78

Color rating probability distributions for application method effects.
1998 ................................................................................................. 78

xi

Figure 29. Quality rating probability distributions for application method effects.
1997 ................................................................................................. 79

Figure 30. Quality rating probability distributions for application method effects.
1 998 ................................................................................................. 79

Figure 31. Stripe rating probability distributions for application method effects.
1 997 ................................................................................................. 80

Figure 32. Stripe rating probability distributions for application method effects.
1 998 ................................................................................................. 80

xii

CHAPTER ONE

THE PROPORTIONAL ODDS MODEL: A STATISTICALLY SOUND METHOD
TO ANALYZE TURFGRASS RATING DATA

ABSTRACT

A common objective of many turfgrass experiments is to evaluate the
effects of various treatments on turf quality. Analysis of variance (ANOVA) has
traditionally been used to analyze quality rating data. However, many data sets
resulting from turf quality ratings have ordinal outcomes, defined as the ranking
of a set of observed values. These data violate assumptions required for valid
statistical inference from ANOVA since they are not continuous. The
development of the proportional odds model (POM) allows for valid statistical
inference on treatment effects from ordinal rating data. The POM also estimates
treatment parameters and standard errors, making treatment separation tests
and contrasts possible. These options were not available with the traditional
statistical tests appropriate for ordinal data. Unfortunately, to use the POM to its
full potential a researcher had to be an experienced statistical software
programmer, making it unusable for many. The objective of the following work
was to develop a Rating Data Analysis File Package (RDAFP) that (i) analyzes
ordinal rating data in a statistically valid manner using the POM, (ii) outputs
nearly the same amount of information on treatment effects as ANOVA, and (iii)
has an intuitively simple user-interface, from data entry to the production of

output. An example quality rating data set from a 4 x 2 factorial randomized

complete block design was used to demonstrate how the RDAFP analyzes data
with the POM and outputs probability distribution charts into MS Excel. Complete
analysis of the quality rating data with the POM, comparison of the results to
ANOVA, and the production of probability distribution charts were possible with

minimal SAS programming knowledge needed.

INTRODUCTION

The majority of turfgrass research is funded by groups interested in
improving golf course, lawn care, or athletic field conditions. Therefore, an
objective of many turfgrass experiments is to examine the effects of treatments
on the functional or aesthetic quality of turf. This objective cannot be addressed
without an evaluation of turfgrass quality by the researcher. Historically, quality
evaluations have been done by visually rating plots on a scale of 1 to 9, where 1
= dead or brown turf, 6 = minimum acceptable quality (varying depending on the
intended use of the turf), and 9 = ideal turf (dark green, dense, and uniform). The
1 to 9 scale was probably first used because of its practicality. Nine rating
categories were usually adequate to distinguish quality differences observed
among the turf plots, statistical calculations with values from 1 to 9 were relatively
simple, and the results presented to the non-scientific community were
comprehensible.

Quality rating data have different characteristics from data such as
clipping yields that are obtained from an objective measuring device. Rating data
resulting from the 1 to 9 scale will only have nine possible values (1, 2, 8, 9).
Seventeen values are possible if half steps are used (1, 1.5, 8.5, 9). A typical
quality rating may result in less than five unique observed values, whereas a
clipping yield measurements usually result in a unique observed value for each
experimental unit.

Another property of quality rating values is that they are arbitrary, since

the values assigned to turfgrass plots are not from a standardized scale. An

alternate, but equally effective quality rating could be accomplished by using a
scale of “A” to “I” where “A” represented ideal turf, “I” represented dead turf and
“B through H” represented declining levels of turf quality intermediate to ideal and
dead. However, a scale of this sort certainly could not be used to evaluate
clipping yields. Clipping yields are measured with a standardized scale. For
example, an observed clipping yield of 17.6 grams can be precisely
comprehended by any turf researcher. It is obvious that quality rating data are of
a different type than clipping yield data.

Classical statistical texts define the type of data resulting from quality
ratings as ordinal (Freund and Wilson, 1993). Freund and Wilson define ordinal
data as, “... a ranking or ordering of a set of observed values. Usually these
ranks are assigned integer values starting with ‘1’ for the lowest value, although
other representations may be used.” In contrast, clipping yield data is
continuous, meaning that it can take on an infinite number of values within an
interval (Freund and Wilson, 1993). Of course, an infinite number of values is
limited by the precision of the measuring device.

Analysis of variance is a popular statistical tool because of the relatively
large amount of information obtained from the data compared to other statistical
analyses. Global hypothesis testing, treatment mean estimation, and treatment
mean separation tests can all be accomplished using ANOVA techniques. In
contrast, traditional statistical tests appropriate for ordinal data (Kruskal-Wallis
test, Friedmantest, or Spearrnan correlation) only test the global hypothesis of

treatment equality. The relative weakness of these tests, as well as the better

comprehension of ANOVA calculations by most turf researchers may account for
the frequent use of ANOVA for rating data.

Analysis of variance is only valid on continuous data, and only if the data:
1) result from a linear combination of the treatment effects and random error, 2)
error values are random and from a Gaussian distribution with mean = 0 and

variance = oz, and 3) data values are from independent and random samples

(Freund and Wilson, 1993). In addition to violating the continuous data
stipulation, rating data often violate the second assumption of ANOVA. Since
visual quality ratings usually lead to few unique outcomes (typical rating data '
may have a minimum value of “5” and maximum value of “8”), the error values do
not approximate a Gaussian distribution well. Furthermore, the analysis used to
analyze rating data should accommodate whatever rating scale is used by the
researcher. It would be impossible to use ANOVA if an “A” to “I” scale was used
to rate quality. Despite these statistical flaws, ANOVA has been used to analyze
turf rating data for decades.

McCullagh and Nelder (1980) described POMs capable of predicting
ordinal responses from independent variables. These models yield nearly the
same amount of treatment information as ANOVA. However, calculations of
treatment effects and standard errors are complex, and typically require
programming of statistical software.

Recently, Schabenberger et al. (2000) authored SAS® macros that
produce global hypothesis tests, treatment comparisons, and contrasts that

resemble ANOVA output. The complex SAS environment and macro

programming language may deter many turf researchers from using the macro.
The development of a simplified user-interface for this SAS macro may result in
more turf researchers using it to analyze rating data.

The objective of the following work was to develop a Rating Data Analysis
File Package (RDAFP) that (i) analyzes ordinal rating data in a statistically valid
manner using the POM, (ii) outputs nearly the same amount of information on
treatment effects as ANOVA, and (iii) has an intuitively simple user-interface,

from data entry to the production of output.

STATISTICAL METHODS

Ordinal data are traditionally analyzed by non-parametric methods that
only test global hypotheses of treatment equality. Logistic regression models,
first used in the 1940’s to analyze bioassay data (McCullagh and Nelder, 1989) ,
estimate the probability of a response based on predictor variables. Because the
model estimates probabilities rather than mean rating values, it is applicable
regardless of the rating scale used by the researcher. Logistic regression
models have gained popularity in the last 20 years, paralleling the refinement of
mathematical techniques used in their calculation. Kleinbaum (1994) presents
an overview oflogistic regression models with applied examples in a format
palatable to the non-statistician.

McCullagh (1980) described the POM, specialized for the analysis of
ordinal data. The POM involves parallel logistic regressions that estimate the

probabilities of an observation to fall into the ordered response categories, based

on the values of independent predictor variables. The RDAFP discussed later
uses the POM to analyze rating data.

The POM estimates a value, ranging from -oo to co, for each parameter in

the model. For rating data, the model parameters consist of the independent
treatment variables, their interactions, and all observed rating categories. For
example, a completely randomized design with treatment factors A (with 2 levels)
and B (with 3 levels), and observed rating values of ”5", “6", ”7", and "8" would

have values estimated for the following parameters: (11, a2, [31, [32, Ba, a181, a182,
a183, 0:281, a282, a283, n5, 1:6, 1:7, and 1:8. Most software packages will estimate

treatment and category effects as differences from a reference level. Therefore,

the parameter estimates for the first treatment levels (on, B1, and 0:181) and the
highest ranking rating category level (1:8) will be zero.

Parameter estimates are calculated by maximum likelihood techniques
(Shenton and Bowman, 1977). Maximum likelihood calculations result in
parameter estimates that best predict the observed values in the data set.
Calculations involve iterative, re-weighted, differentiation of likelihood functions
and become very complex with few model parameters. However, with the
development of powerful PC processors, maximum likelihood calculations have
become commonplace.

A latent variable, Z, represents a linear combination of the parameter
estimates for the treatment and rating category combination of interest. For
example, if a researcher was interested in the probability of a turf plot receiving

level 1 of factor A and level 3 of factor B being rated at best a 7, then 2 = a1 + [33

+ a1 53 + n7. Since the parameter estimates range from -oo to co, Z must also

share this range. However, probabilities are constrained between zero and one.

A logit-link function is used to transform 2 values into probability
predictions by the following equation: 1 / ( 1 + 6’) , where e is Euler's number
(2.178). This function has a range of zero to one, regardless of the value of Z.
Plugging the sum of the treatment parameter estimates of interest into the logit-
link function will result in a cumulative probability, the probability to be rated, at
best, in a given category. Individual category probabilities are calculated by
differencing cumulative probabilities for two adjacent rating categories. For
example, the probability of the treatment described above to be rated exactly a
“7” is calculated by:

{[1/(1+ e—(a1+B3+a183+1r7))]_[1/(1 + e—(a1+83+a183+1t6) )1}

A variance-covariance matrix for the parameter estimates can be
produced by maximum likelihood calculations in computer software programs.
From this matrix and the parameter estimates, statistical tests can be performed
on the equality between any combination of treatment levels. These tests give
information similar to the mean comparison tests and pre-planned contrasts often

used with ANOVA.

RATING DATA ANALYSIS FILE PACKAGE
Excel version 95 or later (Microsoft, 1995) and SAS release 6.12 (SAS
Institute, 1996) must be installed on the user’s PC to use the file package

described in this paper. The files needed to analyze, output, and graph rating

data are bundled in an installation program called 'RDAFP.exe". This program
can be downloaded from ”Rating Data Analysis File Package" web page at the
URL, http://wwwmsu.edu/~karcherd/rati@§ (Karcher, 2000).

Running “RDAFP.exe" will create a directory called “Ordinal Analysis" on
the C drive of the user's PC. The files, "PropOddsModel.sas“ (Schabenberger et
al., 2000), "turfratesas" (Appendix A), ”Rating Charts.xlt”, and ”readme.txt", are
all placed in the "Ordinal Analysis" directory. Additionally, a shortcut to "Rating
Charts.xlt“ is placed on the PC Desktop during installation.

“Rating Charts.xlt” is an MS Excel template that produces probability
distribution charts from data output by SAS. The “readme.txt” file is a text file
containing detailed instructions for RDAFP and covers installation through
interpretation of results. The “turfrate.sas” file (Appendix A) was created by the
author to run the Schabenberger et al. (2000) macro from easy to use web based
forms and the downloaded MS Excel template.

Once the installation program is completed, a data file needs to be
. created. Although data files can be created in either MS Excel or SAS, using MS
Excel simplifies the analysis process. If the data file is created in MS Excel,
variable names must be in the first row of the spreadsheet and the data values
must begin in the second row (Figure 1). There cannot be any blank rows within
the data when using MS Excel. Additionally, an MS Excel data file must be
saved in an Excel 95 file format (Figure 1). Finally, the MS Excel file must be
closed during analysis since the data cannot be imported into SAS if the file is left

open.

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

£392-92! Wise-3M “II-*- 63..- ' ' - . +1215!
1.0.-- 345331-! R i'.‘.’?,1‘f"‘ Q13]: ”.34 [I g&§'_w--:.:.®5
11...; ..—.,o .111nfigsm43x,xaag‘ljxeuéq .

m d I'

‘,_"____~ 8 L--.§- - o E . F G 11 I r ‘ J K 1. 15 N ,____,o -. _P_ c:-
L‘LPLOT BLOCK usounce m ouwrv -
‘2‘ 1 ‘ 3 1 ‘\

_3 2 1 4 2 3 ¥

4,_ 3 1 1 2 5 st

5 4 1 2 1 4 ' -

11 s 1 . 1 5 vanable names In 1 row

7 . 6 1 1 1 4

_e 7 1 3 2 a

9__ a 1 2 2 4

1o 9 2 1 2 5

11, 1o 2 2 2 s

12 11 2 1 1 3

13. 12 3 1 2 s 7

14. 13 ,3 1 1 -

15. 14 3 2 2

15 1s 2 2 1 4.... 1.11» agelglE-idua 1.61

17, 1s 2 4 1 ,

1a 17 2 3 2 “wt" I an I
13_ 1a 2 3 1

20 19 3 3 1 W 1
ill 20 2 4 2 We... I
22 21 4 1 2 ‘
23 22 3 4 2

24, 23 4 3 1

25, 24 4 4 2

ti- 2‘5 3 2 1

27_ E 3 4 1

2B 27 4 1 1 “W‘- l'“” 3

211 23 4 2 1

3) 29 4 ‘ 1 Sue-w [Mammalsumwvm l

_31_ 30 4 2 2 5

32 31 3 3 2 a 7

33 32 4 3 2 7

34

as. K

as

at, . save as Excel 95
39. .. . , no blank rows 1n data _
EJWI._ _-31I"

 

 

 

 

beef-"W ”T
‘-P¢::.B£1:1&?:1:L§ 983.13;é:$1§332§4._w- -,_-_____---..__._.__.-_-----,. -. ___._____-,. 2..-“-

Figure 1. Rating data file created in MS Excel.

The SAS code needed to analyze and generate probability charts from the
rating data can be generated by the RDAFP web page (Karcher, 2000). This is a
form based web page where the user answers a few questions regarding the
experimental design and treatment structure of the study that generated the
rating data. After answering all questions, clicking the ”Generate SAS Code"
button will generate code in a separate window titled, “SAS Code for Ordinal
Data Analysis” that is ready for pasting into the Program Editor window of SAS.

Copying the text from the web window into SAS is relatively simple: 1)
activate the output window by clicking it with the mouse pointer, 2) drag over the

code with the mouse to select, 3) press Ctrl+C to copy all of the code, 4) activate

10

the Program Editor window of SAS by clicking it with the mouse pointer, and 5)
press Ctrl+V to paste the code into the Program Editor window of SAS.

If probability distribution charts created in MS Excel are desired, "Shortcut
to Rating Charts” should be opened from the PC’s Desktop prior to executing the
SAS code. The SAS code is executed by activating the Program Editor window

in SAS and pressing the F8 key.

EXAMPLE USAGE
Experimental Design

The data used in this example are from a quality rating taken on a nitrogen
fertilization study. The objective of the study was to compare the effects of
nitrogen application method and rate on the quality of a ‘Penncross’ creeping
bentgrass (Agrostis palustn's Huds.) putting green. The application methods
included nitrogen injection using high pressure water injection cultivation (W lC)
(Murphy and Rieke, 1994) and traditional surface applications. The study was 4
x 2 factorial with four replications in a randomized complete block design. The
first factor was application method, having four levels: 1) surface sprayed N, no
WIC, 2) surface sprayed N, followed by WIC, 3) N applied via WIC to a 7.5 cm
depth, and 4) N applied via WIC to a 15 cm depth. The second factor was N
rate, having two levels: 1) 2.4 and 2) 4.8 g N rn‘2 application".
Generating SAS Code

Figure 2 shows the RDAFP web page for downloading the file package

and generating SAS code for analysis. Step #1 on the page instructs the user to

11

click the installation icon with the mouse pointer and download the file package.
After downloading and running the installation program, the form on the web
page can be used to generate SAS code.

Steps #2 through #6 on the form must be completed in order to generate
the proper SAS code. Step #2 defines the path on the researcher’s PC to the
MS Excel file containing the rating data, which in this case was
“C:\DATA FILES\RATINGS.xls” (Figure 1). Step #3 defines the treatment factors
as they are named in the MS Excel data file. The treatment factors in the data
file created for this study were named NSOURCE and NRATE (Figure 1). Step
#4 defines the blocking factor, if present, as it is named in the MS Excel data file,
which was BLOCK in this case (Figure 1). Step #5 defines the response
variable, as named in the MS Excel data file. The response variable was named
QUALITY in this‘situation (Figure 1).

Step #6 gives the user an opportunity to label the observed data values
from the rating. Possible labels for a typical quality rating scale are (1 =dead,
2=mostly dead, 3=severely flawed, 4=flawed, 5=slightly flawed, 6=acceptable,

=good, 8=excellent, 9=ideal). For this study, “3”, “4”, “5”, “6”, “7”, and “8” were
the only observed quality rating values. They were labeled here as
“SEVFLAWED”, “FLAWED”, “SLIFLAWED”, “ACCEPTABLE”, “GOOD”, AND
“EXCELLENT". Labels should not contain any spaces or special characters and
be relatively short in order to accommodate output. Labels provide the
researcher with an opportunity to describe the basis for rating the turf and are

typically more informative than arbitrary numbers.

12

 
   

51 Rating Date Andlpsm file Package Web Page ~ Netscape _ I

EieECRYQwfioWfieb

§§‘§i§~-R§flfh&fid.dlé§l

r 'iflx'ez l3

 

 

 

 

 

 

 

Home Search NW Print Seam; Shop 2:9;
; efit'smm 32 Locdiort , 3"wmdm
“HI,” ‘ _ ‘ j _
:2
1 Ratlng Data Analy81s Wade 5449
Ky Don, M
Installation
1. If you haven‘tdone so already, download endmnthe following installation progm; 5' '00"

 

W

2. Providethe path ofthemelflflle contdningyoudetx
jc: \mnxmmcs . x1:

 

3. Provide the neme(s) of the meble(s) containing the treatment fectoz(s)(sepexete each factor with a single space):
[NSOURCE 1mm:

 

4. Did youheve a block desigx?
(‘ No 6 Yes lfyes, provide the name of the ven'eble containing the blocking fectorlBLOCK

 

5. Provide the one ofthe variable contdningthe retingvelues:
locum»

 

6. Provide labels for the observed levels of the rating Mable 60m lowest to highest (separate each category with a single space):
jszmm mm SLIMUED ACCEPTABLE GOOD HELLER”?

 

W

7. Anelyasereducsdmodelaullnodelwithenintmctiontemsis default) P No 6 Yes
8. Main eflects yephedhExcelChuts 1" No 6 1r”

9. Lntmctionefi’ectyaphedinenExcelChut r"Nor”!es

10. Compare proportional odds model outputtoANOVA output (‘ No 6 Yes

ll. Testecontrest C Hoe Yes

12. Slice mintenctimsfi‘ect 6 No t‘ Yes

13, Pmyjd. .meSASW [movement ODDS noon. mmsxs

 

Generate SAS Cadet j

 

 

 

 

 

 

tail-0' (owner;

 

Figure 2. SAS code generating form, completed with information from nitrogen application method
study. This form is from the "Rating Data Analysis" web page at
"http://www.nsu.edu/~karcherd/ratings”.

Steps #7 through #13 are extra analysis options. Step #7 gives an option
to the user to define a reduced model. By default, a full model is used that
contains all treatment factors and all possible interactions. When the number of

treatment factors and observed rating categories is large relative to the number

13

of observations in the data set, a full model might result in errors during
maximum likelihood calculations. The following message (Figure 3) appears in

the Log window of SAS when maximum likelihood errors occur:

 

 

WARNING: There is possibly a quasicomplete separation in the sample points. The
laximum likelihood estimate may not exist.
WARNING: The LOGISTIC procedure continues in spite of the above warning. Results
shown are based on the last maximum likelihood iteration. Validity of the
model fit is questionable.

 

Figure 3. Warning message that appears in the Log window of SAS when errors occur during
nuudnnuulfikeflhoodrudcuhmfion.

Checking “Yes” in Step #7 will cause a text prompt to appear upon clicking
the “Generate SAS Code” button (Figure 4). The user may define a reduced
model in this text prompt if maximum likelihood errors occur when analyzing the
full model.

The example data set had 32 observations and 24 (4 BLOCK, 4
NSOURCE, 2 NRATE, 8 NSOURCE x NRATE, and 6 QUALITY) parameter
estimates in the full model. The full model resulted in maximum likelihood errors,
causing the error message in Figure 3 to be printed in the Log window of SAS.
Therefore, a reduced model was used by dropping the NSOURCE X NRATE
interaction term from the full model (Figure 4). No error messages resulted from

analyzing the reduced model.

14

 

 

 

ngthe Wt WW msu cdu [Javofmupl Application]

 

 

_._ . .
lPTlUfl REQUESTED: REDUCED MODEL
amplydes'sednodel
3 variable [NSOURCE NRATE}
1.2.... , 13:1 ___J m—

 

ting variable 60m lowest to lighest (sepm each category with a my: space):
‘31.: coo

Figure 4. Textbox used to supply a reduced model. In this case, the NSOURCE x NRATE
interaction term has been dropped from the full model.

 

 

 

Step #8 defines if probability distributions for treatment main effects are to
be graphed using the “Rating Charts” template in MS Excel. The default is “Yes”
and the template must be opened prior to executing code in SAS if “Yes” is
checked. Step #9 defines if probability distributions for an interaction term are to
be graphed in the “Rating Charts” template. The default for this option is “No”.
Interaction distributions can only be graphed after main effects analysis has been
executed in SAS. Following main effects analysis, it “Yes” is checked in Step #9,
a text prompt will appear for the user to define the interaction term for which
probability distributions are to be graphed in MS Excel.

Step #10 allows the user to compare results produced by POM analysis
with results obtained from ANOVA. The default for Step #10 is “No”. Checking
Step #10 will produce side-by-side tests of fixed effects and mean separation
tests from the POM and ANOVA.

Steps #11 and #12 give the user an opportunity to test treatment contrasts
and slice interaction terms by user-defined effects. These are the only options

that require SAS programming knowledge by the user. Checking these steps will

15

produce text prompts where the user must provide the proper “contrast” or “slice”
statement, using syntax identical to that of SAS’s glm procedure.

Clicking the “Generate SAS Code” button with the mouse pointer will
generate a new window that contains SAS code (Figure 5). The code shown in
Figure 5 is for the reduced model. This code must be pasted into the SAS
Program Editor window before POM analysis can take place. This is
accomplished by: 1) selecting the code below the horizontal rule with the mouse,
2) pressing Ctrl+C to copy the code, 3) clicking inside the Program Editor
window of SAS with the mouse pointer, and 4) pressing Ctrl+V to paste the code.
After pasting the code into the SAS Program Editor window, pressing the F8 key
will execute the “turfratesas” macro that uses the POM and all other files in the

RDAFP to generate output.

 

i )1 3 (it at a as i 7 Q - I
_ w FN/ad aw a... 5:
‘ Iqfi' Bantu-b ALMIWWMW‘T—jfl'mw

 

 

 

 

 

Figure 5. Output generated from the "Rating Data Analysis" web page form. This code can simply
be pasted into SAS v. 6.12 for expedient analysis using the POM.

16

Summary of Output

Executing the code in Figure 5 produced the output shown in Figure 6
through Figure 10. Figure 6 shows the comparison of fixed effects tests between
the POM and ANOVA. The output shows the degrees of freedom and computed
chi-square and F values used to determine the respective P-values for each
statistical test. In this example, nitrogen application method and nitrogen rate
significantly affected turf quality. The two statistical analyses produced
remarkably similar P-values for NSOURCE (P = 0.0001) and NRATE (P =

0.0003) effects.

 

_ . 71207011110991. 9905 110051. * -‘ ' ‘ 1 1
- 1 79070111101121. 0009 met 9119217379 ~

Analysis Comparison; Proportional Odds model (Pa) 35. «Analysis of Variance (ANOVA)
‘ " . DataSetzdata W - .. .

' Source of ~ 7011 7' I if 7911” “‘31.”.9011" . 12170119 mom

 

 

Variation ‘ 01' '9 phi-Sq- 1 , ~ Allle'rA F ' L: {’P58h16031 j'Sig. '. ' {Pet-‘1‘ '1 79819.

4.115009% a 1 20 9792 >. 7 21 77 911019649”;: 4... * . [9.0001] We

| 31191105- .. :1 10 97.49 . 18 99 309029920 "'”**‘ c.9903 . W"
P values for j Pvalues for ANOVA

 

 

 

 

 

 

‘7 proportional odds

 

._» an -v

 

 

Figure 6. Tests of fixed effects produced by RDAFP. Both application method (NSOURCE) and
nitrogen rate (NRATE) effects were highly significant when analyzed by the POM and
ANOVA.

Predicted probabilities for each treatment level to be rated into each
quality category and a comparison of mean separation tests between the POM
and ANOVA for NSOURCE are shown in Figure 7. Treatment #3, which

corresponded to nitrogen injected to a 7.5 cm depth, had the highest probability

17

(41%) to be rated as excellent. Conversely, Treatment #2, which corresponded
to surface applications of nitrogen plus WIC, had the highest probability (6%) to
be rated as severely flawed. From the treatment separation tests following
analysis by the POM , treatments #3and #4 (followed by A’s) were significantly
different from treatments #1 and #2 (followed by B’s). Examination of the
category probabilities for the treatment levels reveals that treatments #3 and #4
produced significantly higher quality than treatments #1 and #2. Similar results

were calculated by a post ANOVA LSD test.

 

 

rammfm 0096 110961“.
PBOFCHTFONAL 000$ ‘DDEL ANALYSIS
Predicred Category Probabilities If: BATIPIGCATEGDRYI'
- for NSOUBGE attests - '

967171.439695 * 61.1er WTABLE I 6000

 

066-: 396011606 FLAIfiBT'
>1 ' . 9 £0-090024‘5 0 40129 0.. 02001. ~ E*0. 04631 0.52714.
2 V ' 4 0.900125 . 0 99659 f0. 0641?» .0 16792 0.61693
. 6 1 . 0.626666; .9 66646 f'_ 0.66577 9.02796. 0.01177
4‘ I .2 9 0.957579, *0. 71964, 9421491“ ~ 9.91265 0.06525
. . -_  ~ - 4.110174
_ ' " . P0111 LS- AMA LS." P1311 lean. 116471
' _ .0837 EHGELLEHT lean teen Groups ;' Maps
‘1 v 1920 40623 9'44.- 67000 .7 2500909614 A f
23 169;. 11727 - 6061063- 46 69009000 . " in. : ”A
9 - 9.90061? ,. , 2 64906 «:4 760090001m . 46 B
4 a ._ .6 .

«i go. 00927.9, 1 6 16295; 9' 1.9.; 25900000

 

 

Figure 7. Probability of each NSOURCE treatment to be rated into each quality rating category,”
well as mean separation tests from the POM and ANOVA.

Category probabilities are easier to compare among treatments using a
probability distribution chart. The probability distribution chart shown in Figure 8

was created automatically in the “Rating Charts.xlt” MS Excel template.

18

Treatments with larger white bars were poorer in quality than treatments with
larger dark gray and black bars. Figure 8 demonstrates that analysis by the
POM yields a greater amount of information regarding treatment effects than the

arbitrary mean rating values produced by ANOVA.

Effect of NSOURCE on Visual Rating

E] SEVFLAWED U FLAWED I SLIFLAWED I ACCEPTABLE I GOOD I EXCELLENT

m
50
C

‘5
66

M

s-
O
G.)
O!)
N

u
C
U
8
O

9.4
O
>

'5

.‘E
2
E
=

U

3 (Al‘ 4 (A)
NSOURCE

‘Bars sharing a letter are not significantly different (P < 0.05).

 

Figure 8. Probability distribution chart created automatically by RDAFP. Cumulative probabilities
are shown on the y-axis, whereas individual category probabilities (greater than 5%) are labeled
within each bar section.

Treatments #1 and #2 corresponded to turf receiving surface applications
of nitrogen, whereas treatments #3 and #4 correspond to turf injected with
nitrogen. A contrast testing equality between the treatment groups would test the
effects of injecting nitrogen vs. surface applications of nitrogen on turf quality. A
hypothesis test comparing treatments #1 and #2 vs. #3 and #4 was

accomplished by checking “Yes” in Step #7 on the RDAFP web page and

19

inputting the appropriate contrast statement (Figure 9). The contrast statement
in the textbox has identical syntax to the contrast statement used in SAS’s proc

glm (minus a semi-colon).

 

lg the treatment factor(s)(separatc each factor with a single space):

 

 

variable ’ OFTEN REQUESTED: CUNTHAST TEST .
.- “Wmafimlseemdocunmi

 

rating va' ICONTRAST 'surfacevs. niected nitr9gen' NSOURCEl 1-1-1

 

. . ‘ _, J Camel l
mgvana‘t « ~. 4 1 , . . .. . . ‘ ispace):
ILE GOOD EXCELLENT

 

 

 

 

Figure 9. Textbox generated from checking the contrast option on the RDAFP web page. Textbox
input has identical syntax to the contrast statement used in proc glm of SAS.

Figure 10 shows the output resulting from the above contrast statement.
Whether nitrogen was applied on the surface or injected significantly affected turf
quality (P < 0.001 ). Caution must be exercised when interpreting contrast
results. A positive chi-square value means that treatments corresponding to
negative coefficients in the contrast statement had higher ratings. This results
because probabilities calculated from the logit-link function increase as 2 values
decrease. In this example, the negative coefficients correspond to treatments #3
and #4, which correspond to injected nitrogen. Since the chi-square value was

positive (20.59), these treatments had significantly higher ratings.

20

 

 

 

 

 

Figure 10. Contrast test from the RDAFP. Here, treatments receiving surface nitrogen were
significantly different (P < 0.001) than turf injected with nitrogen.

CONCLUSIONS

Analysis of turfgrass quality data with the RDAFP was a simple process.
The data file was created in MS Excel and the SAS code needed to run the
RDAFP was generated from an intuitive web based form. After generating SAS
code from the web and pasting it into the SAS Program Editor window, pressing
the F8 key executed the “turfrate.sas” macro. This macro accessed the other
files in the RDAFP to import the data from an MS Excel data file, analyze the
data using the POM, perform treatment separation tests, and output probability
distributions to the “Rating Charts” template in MS Excel. This occurred without
the need to program any SAS code. The only SAS knowledge needed was how
to paste code into the Program Editor window, and then press the F8 key to
execute the pasted code.

The RDAFP has potential to be a valid, user-friendly data analysis tool for
researchers in other agricultural sciences when data is acquired from subjective,

qualitative rankings. Examples include, but are not restricted to, disease ratings

21

on potatoes, insect damage ratings on tree leaves, and color brilliance ratings on
flowers. Several applications of the RDAFP also exist in the non-agricultural
sciences.

During the initial phase of the RDAFP creation, version 6.12 was the latest
release of SAS. Since then, versions 7.0 and 8.0 have been released, each
containing procedures (tlogisitic and genmod) capable of proportional odds
model analysis. An updated version of the RDAFP is under development that will

work with these procedures in the later versions of SAS.

22

BIBLIOGRAPHY

Freund, R. J., and W. J. Wilson. 1993. Statistical Methods. Academic Press,
Inc., San Diego, CA.

Karcher, D. K. 2000. Rating data analysis file package web page. Available at
http'J/www.msu.edu/~karcherd/ratings.

Kleinbaum, D. G. 1994. Logistic regression; a self learning text. Springer-
Verlag., New York, NY.

McCullagh P. 1980. Regression models for ordinal data. Journal of the Royal
Statistical Society, Series B 42:109142, London.

McCullagh P., and J. A. Nelder. 1989. Generalized linear models. 2nd ed.
Chapman and Hall. London.

Microsoft. 1995. Excel 95 for Windows. Microsoft Corp., Seattle, WA.

Murphy, J. A., and P. E. Rieke. 1994. High pressure water injection and core
cultivation of a compacted putting green. Agron J. 86:719-724.

SAS Institute. 1996. The SAS system for Windows. Release 6.12. SAS Inst,
Cary, NC.-

Schabenberger, O., D. E. Karcher, P. E. Rieke, J. D. Kelly, and J. C. Stier. 2000.
Treatment effects and comparisons in experiments with ordered
outcomes. Submitted to Crop Science.

Shenton, L. R., and K. O. Bowman. 1977. Maximum likelihood estimation in
small samples. Macmillan, New York, NY.

23

CHAPTER TWO

INVESTIGATING CAUSES AND CURES FOR LOCALIZED DRY SPOTS ON
CREEPING BENTGRASS PUTTING GREENS
ABSTRACT
Localized dry spots (LDS) on creeping bentgrass (Agrostis palustris

Huds.) putting greens are characterized by irregular patches of wilting turf that
resist wetting by irrigation or rainfall. Turf managers around the world have
attempted to control LDS, rarely with long term success. Previous research has
demonstrated the causal agent of LDS to be of fungal origin, possibly a
basidiomycete fairy-ring causing pathogen. Water injection cultivation (W lC), a
non-ionic surfactant blend soil wetting agent (SWA), and flutolanil fungicide have
all shown limited control of LDS. Combinations of these treatments may provide .
enhanced control of LDS. Additionally, if the causal agent of LDS is identified,
better strategies to control this problem may be developed. The objectives of this
research were (i) to evaluate the effects and interactions of WIC, SWA, and
flutolanil on the control of LDS on a sand based creeping bentgrass putting
green, and (ii) to more thoroughly define the spatial distribution of the most
affected area by LDS and (iii) to further evaluate the relationship of fungal
biomass and LDS activity. A preventative LDS study was conducted in 1998 and
1999 on a creeping bentgrass putting green with a coarse-textured root zone.
Treatments were arranged in a 2 x 2 x 2 factorial, with the factors consisting of:
WIC (tri-weekly or none), SWA (1.3 ml m'2 tri-weekly or none), and flutolanil (0.9

g a.i. m'2 tri-weekly or none). Treatment effects on turf quality, wilt, soil moisture,

24

and water drop penetration times varied between years. Wilting was reduced
with WIC in 1999, while flutolanil improved general turf quality during summer
stress periods in 1998. Flutolanil and SWA improved turf quality when applied as
curative treatments in 1998. The depth and surface distribution of water repellant
characteristics in randomly selected LDS patches were evaluated by direct
visualization using stereophotmicrography to locate where LDS symptoms were
most prominent. These studies indicated that water drop penetration times were
greatest at the inside and edge of dry spots, just beneath the thatch layer.

Buried slides examined by quantitative brightfield microscopy and computer
assisted image analysis revealed a significantly greater amount of growing
hyphae at the edges of the dry spots, consistent with the proposal that the cause

of LDS may be of fungal origin.

25

INTRODUCTION

Localized dry spots (LDS) routinely confound turfgrass management
because of unpredictable occurrence and difficulty of control. Localized dry
spots are defined as dry spots of turf and soil surrounded by more moist turf
conditions, which resist rewetting by normal irrigation or rainfall (Beard, 1982).
Most turf areas afflicted with LDS share common characteristics (Henry and
Paul, 1978; Wilkinson and Miller, 1978) including: 1) coarse-textured soils, 2)
only the surface centimeters of the soil are hydrophobic, and 3) water repellent
coatings on soil particles. Since the United States Golf Association introduced
putting green specifications (USGA, 1960), most new greens have been
constructed with sand based root zones, increasing the incidence of LDS
(Kamok et al., 1993). Continued use of sand topdressing as a putting green
management practice also contributes to the LDS problem.

Localized dry spot is not confined to any single geographic region. In a
' survey by York (1993) 86% of the greenskeepers in the United Kingdom
encountered LDS on their golf courses and 61% had LDS problems for at least
five years. In addition to reducing available water, LDS may contribute to other
poor soil physical properties. Dry putting green soils in Michigan became hard
and reduced the penetration of cultivation units (Rieke, 1974). Creeping
bentgrass putting greens (Agrostis palustris Huds.) with LDS in Ohio had 20%
lower infiltration rates than did healthy areas (Wilkinson and Miller, 1978).

Once LDS occurs on turf, complete eradication is difficult in a short period

of time. Only 3% of those experiencing LDS in the York (1993) survey achieved

26

complete control with wetting agent applications. Hydrophobic soils in northern
Michigan were improved only temporarily with application of high rates of a
wetting agent (Rieke and Beard, 1975). Management practices that completely
prevent LDS have not been firmly established. A survey of 10 Georgia golf
course superintendents experiencing LDS on creeping bentgrass putting greens
showed no correlation between standard management practices and LDS
severity (T ucker et al., 1990). Additionally, the survey showed no relationship
between LDS severity and chemical soil tests.

A specific causal agent of LDS has not been discovered, but the literature
suggests it may be of fungal origin (Bond and Harris, 1964). Organic coatings on
sand particles taken from LDS areas in Ohio had the presence of fungal mycelia.
The coatings were determined to be primarily fulvic acid compounds (Miller and
Wilkinson, 1977). Some basidiomycete fungi that cause fairy rings are known to
induce hydrophobic conditions that stress turf within the rings. (Smith et al.,
1988). Soil samples taken from within several fairy rings caused by Marasmius
oreades on golf course fairways in Norway had significantly low moisture content
(Smith, 1975). Similarly, inner zones of M. oreades fairy rings on Kentucky
bluegrass (Poa pratensis L.) turf in Saskatoon, Canada had significantly low soil
moisture contents (Smith and Rupps, 1978).

We hypothesize that the presence of fungi may contribute to LDS
formation by depositing organic coatings on sand particles. Flutolanil (Prostar®
fungicide manufactured by AgrEvo, Montvale, NJ) controls many fairy ring

causing fungi (Elliot and Hickman, 1998) and may be effective in preventing LDS.

27

Two other management practices that have reduced LDS severity are water
injection cultivation (W IC) with a Toro HydroJect 3000® (Karcher, 1997) and the
application of an effective soil wetting agent (SWA) (Gelemter and Stowell,
1998). Additionally, if we can identify the properties of the causal agent of LDS
and locate its growing front, better recommendations for effective control are
possible. The objectives of this research were (i) to evaluate the effects and
interactions of flutolanil, WIC, and SWA wetting agent applications on the control
of LDS on a sand based creeping bentgrass putting green, (ii) locate the zone of
maximum soil hydrophobicity in characteristic LDS, and (iii) further examine the

role of soil fungi in development of LDS.

MATERIALS AND METHODS
Experimental Area

Three separate experiments were designed to address research
objectives: A preventative study to compare the effects of WIC, SWA, and
flutolanil on the prevention of LDS, a curative study to compare the above effects
on curing turf that was severely afflicted with LDS, and an isolation study to
determine the precise location of activity and identify the characteristics of the
LDS causal agent.

Plots were established for the preventative study on a ‘Penncross’
creeping bentgrass putting green with a modified sand root zone (96% sand, 3%
silt, and 1% sand) in early April 1998. Plots were mowed at 4.0 mm five days per

week throughout the study. Nitrogen was applied at 30 g m'2 year'1 and irrigation

28

was applied only at severe drought to provide reasonable growth, but to
encourage the onset of LDS. Other nutrients were applied according to soil test
recommendations. In April 1999 the experiment was repeated on a ‘Penncross’
creeping bentgrass putting green with a modified loamy sand soil (modified fine-
loamy, mixed, mesic, Typic Hapludalf). The experiment was repeated in this
area because it had frequent occurrences of LDS in 1998.

For the isolation experiment, three individual patches showing pronounced
LDS symptoms adjacent to the preventative study were selected in 1998. Plots
were mowed five times per week at a 5.0 mm height and fertilized with 20 g N rn'2
year". Formation of LDS occurred despite this area being irrigated daily at rates
approximating evapotranspiration water loss.

All experimental areas were topdressed lightly with 100% sand
approximately every six weeks. The topdressing sand layer depth remained less
than 1.0 cm throughout the experiments. Each fall, all plots were core cultivated
at a 7.5 by 5.0 cm spacing to an approximate depth of 8 cm using 1.0 cm
diameter hollow tines. Cores were brushed, returning the soil to the plot area,
and the remaining thatch was removed. Fungicides were applied on a curative
basis.

Treatment Design
PREVENTA nvs STUDY

The preventative study had three treatment factors: WIC, SWA, and

flutolanil. Each factor had two levels, applied or not applied, yielding eight

treatment combinations (2 x 2 x 2 factorial). In both years, treatments were

29

applied on 21 -day intervals beginning in late April and ending in mid October.
Plots were drenched with 2.5 cm of water several hours prior to all treatments, in
accordance with the flutolanil label. Water injection cultivation treatments were
made first with a Toro HydroJect 3000® set at the closest hole spacing (7.5 by
2.5 cm). Flutolanil and SWA were applied next at rates of 0.9 g a.i. m'2 (WP) and
1.3 ml m'z, respectively, with a 002 powered plot sprayer. Following treatments,
plots were irrigated with 1 cm of water. Treatments were replicated 3 times in a
randomized complete block design.
CURATIVE STUDY

In August 1998, a turf area adjacent to the preventative LDS study, but
with separate irrigation control, had severe wilt symptoms with random dry
patches characteristic of LDS. A curative study was established with treatments
identical to the preventative study on this area. Treatments were applied tri-
weekly from August to October in 1998 for this study.
Isoumo~ STUDY

The treatment factors in the isolation study consisted of surface location
and depth, relative to the dry spots. Surface location was classified as the
center, edge, or outside of the dry spot. The edge of the dry spot was identified
as the border between turf with visual wilting symptoms from LDS and turf with
no visible LDS symptoms. Sampling of the outside of the dry patch was made 15
cm from the outside perimeter. Depth was classified in one cm increments from

just beneath the thatch layer to a 5 cm depth. Location and depth factors were

30

arranged in a randomized complete block design (blocked by dry spots) with
three replications.
Treatment Evaluations
PREVENTA TIVE AND CURA TIVE STUDIES

Treatments in the preventative and curative studies were evaluated for turf
quality and soil moisture content. Additionally, wilt and soil wettability were
evaluated in the preventative study. Quality and wilt were evaluated weekly, and
when symptoms were visible, respectively. The rating scale for qUality was from
1 to 9 (1 =dead, 2=mostly dead, 3=severely flawed, 4=flawed, 5=slightly flawed,
6=acceptable, 7=good, 8=excellent, 9=ideal), and for wilt was from 1 to 5 (1=no
wilt, 2=slight wilt, 3=moderate wilt, 4=significant wilt, 5=severe wilt). Wilt ratings
were typically done in the late afternoon when wilt symptoms were easiest to
detect. Volumetric soil moisture to a 15 cm depth was measured weekly on three
randomly selected locations per plot with a Trime®-FM portable time domain
reflectometry unit (manufactured by IMKO, Ettlingen, Germany). Soil wettability
was evaluated monthly by water drop penetration times. Three randomly
selected soil cores per plot were pulled and sectioned by depth into five one cm
increments. Within 48 hours of sampling, cores were sliced in half vertically with

a razor blade. A 100 pl water droplet was then placed on the flat surface of each

core section and the time elapsed until the droplet had completely penetrated the

soil surface was recorded.

31

ISOLATION STUDY

Treatment evaluations for the isolation study were designed to precisely
locate where the LDS causal agent affected soil wettability. Water drop
penetration times were measured on samples from the inside, edge, and outside
of three separate dry patches using the methods described above. Rossi-
Cholodny buried slides (Johnson and Curl, 1972) were inserted just beneath the
thatch layer at the inside, edge, and outside of three individual dry patches and
incubated for 21 days. The use of buried slides allowed for precise spatial
analysis of alive, active fungi capable of colonizing the slide. Following
incubation, slides were carefully extracted and stained with IactophenoI-aniline
blue-acid fuchsin. An AusJenaval brightfield microscope (25x objective) and
Panasonic WV185 Neuvicon camera were used to output fungal images to a
video monitor so that hyphae could be traced onto transparency overlays. All of
the hyphae occurring on a one-cm2 portion of the slide that was adjacent to the
soil immediately below the thatch layer were recorded. The hyphae were
digitized and analyzed by the cumulative hyphal length feature of the Image Tool
option in the CMEIAS software package (Liu et al., 2000).
Statistical Analysis

Quality and wilt rating were analyzed using the proportional odds model
that is incorporated into the Rating Data Analysis File Package (Karcher, 2000).
Treatment separation was done with pairwise chi-square tests of the treatment
least squared estimates. Probability distributions were constructed to represent

the odds of a treatment level to be rated in a particular category. These

32

distributions were constructed by inserting the appropriate combination of least
square estimates into the logit-Iink function.

Soil moisture, soil wettability, and hyphal length data were analyzed with
ANOVA. If treatment effects were significant, means were separated using LSD
at the 0.05 probability level. Where repeated measures were made on the same
experimental units, time was analyzed as a sub-plot factor of the experiment.
The best fitting covariance model among compound symmetry, first order auto-
regressive, and spatial exponential was used to fit correlations among time
points. The best fitting covariance model was determined by the highest
Akaike’s Information Criteron value (Littell et al., 1996). A log transform was
used to normalize the water drop penetration data, which originally were highly

right skewed because of several values near zero.

RESULTS AND DISCUSSION
Preventative Study

ANOVA results and the main effects of WIC, SWA, and flutolanil on turf
quality, wilt, and soil moisture are summarized in Table 1. The main effects of
WIC, SWA, and flutolanil on turf quality were not significant in either year.
However, there was a significant flutolanil x time interaction in 1998 and a
significant SWA x flutolanil interaction in 1999. .

The flutolanil x time interaction resulted from a significant flutolanil effect
on 5 out of the 20 rating dates in 1998 (Figure 11). On 7 May flutolanil slightly

decreased the probability to be rated high in quality, but on 17 July, 11 August,

33

18 August, and 18 September flutolanil significantly increased the probability to
be rated high in quality. The dates when flutolanil significantly increased quality
were dates when the plots, averaged over all treatments, were rated significantly
low in quality (Figure 12). In 1999, the probability of plots to be rated high in
quality decreased when flutolanil and SWA were both applied compared to the
application of either alone (Figure 13).

These results suggest that flutolanil may be effective in increasing turf
quality when environmental conditions are particularly stressful. On 11 August
and 18 September, 1998, the experimental area was drought stressed, resulting
in low overall quality. However, flutolanil significantly improved turf quality on
these rating dates. These results are to be expected if drought symptoms in turf
are partially caused by a fungal species susceptible to flutolanil. Adams (1989)
found that flutolanil had a high degree of fungicidal activity against Marasmius
oreades, a fungus known to cause fairy ring often expressing hydrophobic soil
conditions (Bayliss, 1911; Smith, 1975; Smith and Rupps, 1978).

The flutolanil x SWA interaction suggests that these products may be
slightly phytotoxic when applied together at the highest labeled rate of each.
Flutolanil has been reported to cause phytotoxicity when mixed with other

products (Gelemter and Stowell, 1997).

Table 1. Effects of WIC, SWA, and flutolanil on quality, wilt, and soil moisture of a creeping
bentgrass putting green.

 

 

 

 

 

 

1998 1999
Eff ct Quality Wilt Soil d f Quality Wilt Soil
e Parameter Parameter Moisture Parameter Parameter Moisture
likelihood likelihood 3 _3 likelihood likelihood 3 _3
estimate estimate m m estimate estimate m m
WIC
None -1.22 A1 5.22 A 26.6 A -1.03 A -0.95 A 26.6 A
Tri-Weekly -0.92 A 2.11 A 26.2 A -O.60 A 0.15 B 27.3 A
SWA
None -O.89 A 3.50 A 26.4 A -0.84 A -0.28 A 26.5 A
1.3 ml m'2 -124 A 3.33 A 26.5 A 4179 A -052 A 27.4 A
Flutolanil
None -0.80 A 3.76 A 26.5 A -O.84 A -0.57 A 26.6 A
0.9 g a.i. m'2 -134 A 3.57 A 26.4 A -0.78 A 43.23 A 27.3 A
ANOVA
Source of variation
Block 2 it! 2 *‘t 2 *t 2 3“ 2 81* 2 **
WIC (w) 1 NS 1 NS 1 NS 1 NS 1 " 1 NS
SWA (s) 1 NS 1 NS 1 NS 1 NS 1 NS 1 NS
w x s 1 NS 1 NS 1 NS 1 NS 1 NS 1 NS
Flutolanil (f) 1 NS 1 NS 1 NS 1 NS 1 NS 1 NS
w x f 1 NS 1 NS 1 NS 1 NS 1 NS 1 NS
5 x f 1 NS 1 NS 1 NS 1 ** 1 NS 1 NS
w x s x f 1 NS 1 NS 1 NS 1 NS 1 NS 1 NS
Time (t) 19 "'f 1 NS 16 *" 13 "'f 2 **"‘ 4 ***
w x t 19 NS 1 NS 16 **"' 13 NS 2 NS 4 NS
5 x t 19 NS 1 NS 16 NS 13 NS 2 NS 4 NS
wxsxt 19 NS 1 NS 16 NS 13 NS 2 NS 4 NS
f x t 19 “ 1 NS 16 NS 13 NS 2 NS 4 NS
wxfxt 19 NS '1 NS 16 NS 13 NS 2 NS 4 NS
sxfxt 19 NS 1 NS 16 NS 13 NS 2 NS 4 NS
wx sxfxt 19 NS 1 NS 16 NS 13 NS 2 NS 4 NS

 

‘, ", 1‘" Significant at the 0.05, 0.01, and 0.001 probability levels, respectively.
1* Within effects, means sharing a letter are not significantly different

~35

E I. .
Cl Poor DFlawed ISllghtly Flawed IAcceptable IGood I Excellent

Oi
.E
‘6
h
c»
c
E
o
D
2
'6
5'
3
cc
.0
9
o.
o
.2
E
3
E
=
0

No mm No FMOIInI'I

Flutolanil w w
7 May 981' 11 Aug. 98

Level of flutolanil x time

1 withln rating dates. probability distributions are simlficantty different (0.05).

 

Figure 11. Quality rating probability distributions as aflected by flutolanil x time. Preventative
Study, 1998.

 

Rating ggtggories
Poor I: Flawed ESIightly Flawed IAcceptable I Good I Excellent

.‘Hr
.‘ 7:}.
l

1

Hr.

Cumulatlve probablllty of recelvlng rating
----- El

 

 

 

 

Rating date

t Probability distributions followed the same letter are not significantly different (0.05).

 

 

 

Figure 12. Quality rating probability distributions as affected by time. Preventative Study, 1998.

36

The experimental area wilted enough for ratings on only two dates in 1998
and three dates in 1999. The lack of LDS formation on the experimental area
was probably due to the irrigation practices used to accommodate treatment
application. A tri-weekly application of 3.5 cm of water during the study may
have inhibited the onset of LDS, since a reported precursor to LDS formation is a
thorough drying down of the soil (Paul and Henry, 1973). When the experimental
area dried down, it usually did so in a uniform fashion rather than by forming
random dry spots characteristic of LDS. The only significant treatment effect on
wilt was WIC in 1999 (Figure 14). Averaged over all rating dates in 1999, WIC
treated turf had significantly lower probability to be rated high in wilt than
untreated turf. This result is consistent with results from previous WIC
experiments (Karcher, 1997). The 20 MPa water blasts from the HydroJect® unit
may partially remove hydrophobic coatings on sand particles. Additionally,
isolated channels created by the cultivation blast (Murphy and Rieke, 1994) may
allow for deeper rooting and subsequently, more total water available in the root

zone for WIC treated turf.

37

 

 

 

 

Rating cajgggries
El Slightly Flawed IAcceptable IGood I Excellent

100%.

90%
O)
E l
E 30% 1
O)
C
3 70%
0
0
9 60% ,
"5 .
g 50% ,
B
It!
8 40%
a l
g 30%
E l
E 20% .
3
O

10% ‘

0% , . ,
No Flutolanil (Aen nmmr (A) No Flutolanil (A) Flutolanil (s)
T Probability distributions followed by the same letter are not significantly different (0.05).

 

 

 

Figure 13. Quality rating probability distributions as affected by SWA x flutolanil. Preventative
study, 1999.

mm
D None El Slight I Moderate I Significant

Cumulative probability of receiving rating

Rafingdate

1' Within effects, probability distributions followed by the same letter are not significantly different (0.05).

 

Figure 14. Wilt rating probability distributions as affected by WIC, SWA, flutolanil, and time.
Preventative LDS study, 1999.

38

Volumetric soil moisture was measured on 17 dates in 1998, but only 5
dates in 1999 due to broken time domain reflectometry equipment. Time and
WIC x time effects were significant in 1998, whereas only time effects were
significant in 1999. There were significant differences between WIC treatments
on 5 of the 17 dates soil moisture was measured in 1998 (Table 2). Turf
cultivated by water injection had significantly greater soil moisture content on 20
May and 9 June, but significantly lower soil moisture content on the three
significant evaluation dates in August. These data are surprising when
considering the wilt reducing effect of WIC seen in 1999. These data suggest
that the reduction in wilt with WIC treatment is probably not caused by the simple
addition of water to the root zone. The HydroJect® adds approximately 1.5 mm

water to the turf at the closest hole spacing and with the roller washers off.

Table 2 . Interaction effects of WIC and evaluation date on volumetric soil moisture. Preventative
study, 1998.

 

 

 

 

 

Soil Nbisture
mermm 20May.98 91m98 3Aug.98 6Aug. 98 12Aug98
m3 m3
None 313+ 34.4 164 20.6 23.7
Tri-Weekly 335 37.0 10.4 16.9 21.1

 

T Within 0011mm, nuns are significantly difierent at the 0.05 probability level.

Water drop penetration times were recorded only in September in 1998
and monthly from June through September in 1999 (Table 3). Mishandled
samples from June through August in 1998 had to be discarded. Depth was the

only significant main effect in both years. Significant interaction effects included

39

WIC x depth in 1998, and flutolanil x depth, depth x time, WIC x flutolanil x time,
and WIC x SWA x flutolanil x time in 1999.

In 1998, the 3 and 4 cm depths had significantly longer penetration times
than depths 1, 2, and 5. Although the log(s) penetration times were statistically
significant, there were no practical differences in the untransformed means
among treatments. In 1999, the soil samples were significantly less wettable
near the surface and became more wettable with depth. However, all soil depths
showed some degree of non-wettability in both years. The 1999 data are fairly
consistent with those of Wilkinson and Miller (1978), who found that hydrophobic
conditions in sand based putting greens were restricted to the upper 2 cm of soil.
However, the 1998 data do not show this trend. Wilkinson and Miller evaluated
turf showing characteristic LDS symptoms, whereas the turf in this study dried
down uniformly. Uniform dry down conditions may result in hydrophobic soil at
depths greater than 2 cm. Bond (1968) reported hydrophobic conditions in sands
at depths of up to 0.5 meter.

Where no WIC applications were made in 1998, the 3 and 4 cm depths
were significantly less wettable than the 1 and 2 cm depth (Table 4). However,
there were no differences in wettability among depths where WIC was applied.
The 20 MPa blast of the WIC unit probably mixed the soil so that there were no
differences in penetration times among soil depths. Previous studies have

shown that WIC significantly mixes soil layers in putting green (Karcher, 1997).

40

Table 3. Effects of WIC, SWA, flutolanil, and sample depth on water drop penetration time.

 

Water drop penetration time

 

 

Effect ————— 1998 —————————— 1999 —————
df s log(s)l‘ (If s log(s)
WIC
None 5.7 0.22 A 30.2 2.69 A
Tri-Weekly 11.3 0.21 A 34.4 2.89 A
SWA
None 5.2 0.16 A 32.8 2.78 A
1.3 ml 111'2 11.8 0.28 A 31.8 2.80 A
Flutolanil
None 14.1 0.30 A 34.3 2.92 A
0.9 g a.i. m'2 2.9 0.13 A 30.3 2.66 A
Depth
1 9.5 0.10 B 56.1 3.71 A
2 9.7 0.13 B 48.7 3.46 A
3 10.0 0.32 A 32.3 2.92 B
4 10.8 0.35 A 15.2 2.17 C
5 2.6 0.18 B 9.2 1.70 D
ANQVA
Source of variation
Block 2 NS 2 NS
WIC (w) 1 NS 1 NS
SWA (s) 1 NS 1 NS
w x s 1 NS 1 NS
Flutolanil (f) 1 NS 1 NS
w x f 1 NS 1 NS
5 x f 1 NS 1 NS
w x s x f 1 NS 1 NS
Depth (d) 4 Mr 4 one
w x d 4 't 4 NS
8 x d 4 NS 4 NS
w x s x d 4 NS 4 NS
fit (I 4 NS 4 "
w x fx d 4 NS 4 NS
5 it fit d 4 NS 4 NS
w x s it fit d 4 NS 4 NS
Time (t) --- --- 3 ‘"
w xt --- --- 3 NS
5 xt --- --- 3 NS
w x s x t --~ --- 3 NS
fxt --- --- 3 NS
w it fit t m --- 3 "”"
s x f x t --- --- 3 NS
w x s it fit t --- --- 3 "
d xt --- ~-- 12 "'
w x d x t --- m 12 NS
s x d xt --- --- 12 NS
wxsxdxt --- --- 12 NS
fx (1 x t --- --- 12 NS
wxfxdxt --- --- 12 NS
sxfxdxt --- --- 12 NS
wxsxfxdxt --- ..- 12 NS

 

". ""'. "‘ Significant at the 0.05. 0.01, and 0.001 probability levels, respectively.
1 Data were normalized by a log(seconds) transformation for statistical analysis.

41

When no flutolanil was applied in 1999, soil became significantly more
wettable with depth, except from 4 to 5 cm (Table 5). Flutolanil application
increased the wettability of the 1 cm depth to where it was equal to the 2 cm
depth. The 1999 data suggest that flutolanil application improves soil wettability
near the turf surface, which would be expected if hydrophobic soil conditions
were partially caused by a fungus susceptible to flutolanil.

The other high order interactions with time imply that the effects of depth,
WIC x flutolanil, and WIC x SWA x flutolanil were inconsistent from month to

month.

Table 4. Water drop penetration times as afl‘ected by WIC x depth interaction. Preventative study,
1998.

 

 

 

Depth (cm)
WIC Treatement 1 2 3 4 5
108(5)
None 0.02 0.05 0.36 0.43 0.23
Tri-Weekly 0.17 0.21 0.28 0.27 0.13

 

1' Within columns LSD = 0.28, within rows LSD = 0.18, both at the 0.05 probability level.

Table 5. Water drop penetration times as affected by flutolanil x depth interaction. Preventative
study, 1999.

 

 

 

Depth (cm)
Flutolanil Treatment 1 2 3 4 5
log(s)
None 3.97 3.46 2.81 2.03 1.63
M a.i. m’2 3.45 3.47 3.02 2.31 1.76

 

t Within columns LSD = 0.56, within rows LSD = 0.49, both at the 0.05 probability level.

Curative Study
For the 1998 curative study, ANOVA results and the main effects of WIC,

flutolanil, and SWA on turf quality and soil moisture are summarized in Table 6.

42

The main effects of SWA and flutolanil on turf quality were significant in the
curative study. There was also a WIC x SWA interaction with regard to quality.
Time was the only significant effect with regard to soil moisture in the curative
study.

Averaged over all rating dates, both SWA and flutolanil application
significantly increased the probability of the turf to be rated high in quality
compared to application of neither (Figure 15). There are no data in the refereed
literature pertaining to initiating treatments on turfs that are already severely
affected with LDS. These results indicate that turf affected by LDS can be
improved by curative applications of SWA or flutolanil.

When plots received no WIC treatment in the curative study, quality
probability distributions were not affected by the addition of SWA (Figure 16).
However, plots receiving WIC treatment had a significantly greater probability of
being rated high in quality when SWA was also applied. The creation of
channels by WIC probably allowed better penetration of the wetting agent
through the thatch layer into hydrophobic soil areas. This may have made soil
more wettable and improved turf quality when WIC preceded SWA applications.
Wilkinson and Miller (1978) showed that a combination of core cultivation plus

wetting agent significantly improved turf quality over either alone.

43

Table 6. ANOVA and main effects of WIC, SWA, and flutolanil on turf quality and soil moisture.
Curative study, 1998.

 

 

Effect df Quail” df S9“
Parameter Mo1sture
likelihood 3 _3
estimate m m
WIC
None 0.75Af 28.8 A
Tri-Weekly 0.73 A 29.2 A
SWA
None 1.33 B 29.0 A
1.3 ml 111'2 0.15 A 29.0 A
Flutolanil
None 1.09 B 29.3 A
0.9 g a.i. m'2 0.38 A 28.8 A
ANOVA
Source of variation
Block 2 *** 2 NS
WIC (w) 1 NS 1 NS
SWA (s) 1 *** 1 NS
w x s 1 ** 1 NS
Flutolanil (t) l * 1 NS
w x f 1 NS 1 NS
8 x f 1 NS 1 NS
w x s x f 1 NS 1 NS
Time (t) 7 ** 2 ***
w x t 7 NS 2 NS
8 x t 7 NS 2 NS
w x s x t 7 NS 2 NS
f x t 7 NS 2 NS
w x f x t 7 NS 2 NS
5 x f x t 7 NS 2 NS
w x s x f x t 7 NS 2 NS

 

*, **, *** Significant at the 0.05, 0.01 , and 0.001 probability levels, respectively.
T Within effects, means sharing a letter are not significantly different.

 

Rating catggries
D Flawed El Slightly Flawed I Acceptable I Good I Excellent

_.
4

 

 

Cumulative probability of receiving ratan
8 i 3 § 3 § § § § § §

 

 

 

 

 

 

 

.
site as so $8.93 @089 $0

a” .4 9°” .8" a“ - 4" orig 91:30 4» .
«f $8 99“.: {.8 ff“ 9.. (Lb-£8 $8,). :68?
WIC SWA Flutolanil ——-———-Rat'ngdate

T Within treatment factors. probability distributions followed the same letter are not significantly different (0.05).

 

 

 

Figure 15. Quality rating probability distributions as afi'ected by WIC, SWA, flutolanil, and rating
date. Curative study, 1998.

W
E] Flawed El Slightly Flawed IAcceptabIe I Good I Excellent

Cumulative probability of receiving ratlng

No SWA (B): SWA (8) No SWA (C)

u rum

 

1 Probability distributions followed the same letter are not significantly different (0.05).

 

Figure 16. Quality rating probability distributions as affected by WIC x SWA. Curative study,
1998.

45

Isolation Study

Water drop penetration time and cumulative hyphal length ANOVAs and
main effect means are summarized in Table 7. Location and depth main effects,
as well as the location x depth effect, were all significant with regard to water
drop penetration time. Location relative to the center of the dry spot significantly
affected cumulative hyphal length at the 0.06 probability level.

Individual water drop penetration times ranged from 1 second to 21
minutes. The inside of the dry spot had significantly longer water drop
penetration times than the edge and outside. Additionally, the edge of the patch
was significantly less wettable than outside (Figure 17). There is no mention in
the refereed literature of evaluating differences in water drop penetration time
among locations, relative to the center of a dry patch.

The 1 and 2 cm depths were less wettable than all other depths, whereas
the 3 cm depth was intermediate, and the 4 and 5 cm depths were most
wettable. These results are similar to those reported previously (Wilkinson and
Miller, 1978; Tucker et al., 1990), where the upper few cm of the soil were the
most hydrophobic in sand based putting greens. The significant location x depth
interaction indicated that significant differences among locations only occurred at

depths of 1, 2, and 3 cm (Table 9).

46

Table 7. Water drop penetration times and cumulative hyphae length as afl'ected by surface location
and depth relative to LDS. Isolation study, 1998.

 

 

Water drop Cumulative
Effect _ df penetration time df hyphal length
5 log(s) urn hyphae cm'2

Location

edge 56 2.5 B1 10295 At

inside 284 3.5 A 3067 B

outside 3 0.8 C 7841 A
Depth (cm)

1 416 4.4 A --

2 128 3.6 A --

3 24 2.4 B __

4 3 0.6 C --

5 2 0.3 C --

ANOVA

Source of variation

Block 2 “ 2 NS

Location (I) 2 "" 1 §

Depth (d) 4 '" -- --

l x d 8 '" -— --

 

§, **, *** Significant at the 0.1, 0.05, and 0.001 probability levels, respectively.
1 Within effects, means sharing a letter are not significantly different (P S 0.05).
:1: Means sharing a letter are not significant at the 0.10 probability level.

 

Figure 17. Images taken 30 seconds after water drop placement just beneath the thatch layer at the
inside, edge, and outside of a dry patch.

47

Stained hyphae from the edge of a dry spot are shown in Figure 18. The
edge of the dry spot had significantly (P = 0.06) greater fungal biomass
(measured as the cumulative hyphal length) than the inside or outside. This is
further evidence that LDS may be caused by a fungal organism. If the causal
agent of LDS is of fungal origin, growth initiates at a central point and continues
radially outward, with the highest concentration of active viable fungi biomass at
the growing edge of a dry spot. This growth pattern is similar to many
pathogenic fungal species that are known to cause fairy ring and patch diseases
(Smiley et al., 1992). Although previous research has associated various fungal
species with LDS (Miller and Wilkinson, 1978; York and Baldwin, 1992), no

fungal species have been identified that are specific to LDS.

 

a “-11 ‘ ~. 0
W i “0‘ g ‘I ’f 5
. b l
.31 ~. ' '- ,
' ' A...” “l. ' “‘

Figure 18. Stained hyphae from a section of a buried slide at the edge of a dry patch, immediately
beneath the thatch layer. Image is magnified 250 times.

48

Table 9. Effects of location it depth on water drop penetration time. Isolation study, 1998.

 

Location relative to patch

 

 

 

 

Depth (cm) Inside Edge Outside
log(S)
l 6.91“ 4.7 1.5
2 5.7 3.7 1.5
3 2.8 3.2 1.0
4 1.1 0.8 0.0
5 1.0 0.0 0.0

 

1‘ Within columns LSD = 1.4, within rows LSD = 1.2, both at the 0.05 probability level.

CONCLUSIONS

Results from the preventative study indicated that LDS occurrence and ‘
control were highly variable, as treatment effects were inconsistent between
years. However, WIC, SWA, and flutolanil all showed some potential to alleviate
LDS symptoms. Water injection cultivation seemed to decrease visual wilting
symptoms, whereas flutolanil increased visual quality during general summer
stress conditions. Soil moisture analyses established that the effect of WIC on
wilt reduction was not simply by wetting the soil. Flutolanil and SWA appear to
provide some curative control of LDS, however combining both at the highest
labeled rate of each may result in phytotoxicity.

The causal agent of typical LDS symptoms appeared to affect the soil
immediately beneath the thatch, although hydrophobic soil was measured at
greater depths when dry down was uniform. A significantly greater accumulation
of hyphae was measured at the edges of the dry spots than the inside. This is

evidence that the causal agent of LDS may be of fungal origin.

49

BIBLIOGRAPHY

Adams, J. C. 1989. Control of turfgrass diseases in the United States with
flutolanil 50WP. Proc. of the Sixth Int. Turf. Res. Conf. 6:357-358.

Bayliss, J. S. 1911. Observations on Marasmius oreades and Clitocybe
gigantean as parasitic fungi causing fairy rings. J. Econ. Biol. 6:111-132.

Beard, J. B. 1982. Turf management for golf courses. Macmillen Publishing Co.
New York, NY. .

Bond, R. D. 1968. Water repellent sands. 9th Int. Congr. Soil Sci. Trans. 1:339-
347.

Bond, R. D., and J. R. Harris. 1964. The influence of the microflora on physical
properties of soils. Aust. J. Soil Res. 2:111-122.

Elliott, M., and B. Hickman. 1998. Control of necrotic fairy rings on berrnudagrass
putting greens. The 19th Hole. 7(1):6-7.

Gelemter, W., and L. J. Stowell. 1997. Evaluation of fungicides and wetting
agents for the management of localized dry spot and fairy ring. PTRI 1998
Turfgrass Research Report. 1998:22-28.

Henry, J. M. and J. L. Paul. 1978. Hydrophobic soils on putting greens. California
Turfgrass Culture. 28(2):9-11.

Johnson, L. F., and E. A. Curl. 1972. Methods for research on the ecology of soil-
bome plant pathogens. Burgess Publishing 00., Minneapolis, MN.

Karcher, D. E. 2000. The proportional odds model: a statistically sound method
to analyze turfgrass rating data. p. 1-23. In Investigations on statistical
analysis of turfgrass rating data, localized dry spots of putting greens, and
nitrogen application techniques for turf. Ph.D. diss. Michigan State
University. East Lansing, MI.

Karcher, D. E. 1997. Nitrogen fertilization and soil cultivation utilizing high
pressure water injection. M.S. thesis. Michigan State Univ., East Lansing.

Kamok, K. A., E. J. Rowland, and K. H. Tan. 1993. High pH treatments and the
alleviation of soil hydrophobicity on golf greens. Agron. J. 85:983-986.

Liu, J., F. B. Dazzo, O. Glagoleva, B. Yu, and A. K. Jain. 2000. CMEIAS: a
computer-aided system for the image analysis of bacterial morphotypes in
microbial communities. Microbial Ecology: 39(2) In Press.

50

Littell, R. C., G. A. Milliken, W. W. Stroup, and R. D. Wolfinger. 1996. SAS
system for mixed models. Cary, NC. SAS Institute Inc.

Miller R. H., and J. F. Wilkinson. 1977. Nature of the organic coating on sand
grains of nonwettable golf greens. Soil Sci. Soc. Am. J. 41 :1203-1204.

Murphy, J. A., and P. E. Rieke. 1994. High pressure water injection and core
cultivation of a compacted putting green. Agron J. 86:719-724.

Paul, J. L., and J. M. Henry. 1973. Non-wettable spots on greens. Proc. Calif.
Golf Course Superintendents Inst. 12:1 —5.

Rieke, P. E. 1974. Soil research - nitrogen carriers, potassium responses, and
hydrophobic soil studies. 44th Ann. Michigan Turfgrass Conf. Proc. 4:21 -
26.

Rieke, P. E., and J. B. Beard. 1975. Rewetting of a hydrophobic sand with
wetting agent and cultivation treatments. p. 100. In Agronomy abstracts,
Madison, WI.

Smiley, R. W., R. H. Demoeden, and B. B. Clarke. 1992. Compendium of
turfgrass diseases. 2nd ed. APS Press. St. Paul, MN.

Smith, J. D. 1975. Incomplete Marasmius oreades fairy rings. J. Sports Turf Res.
Int. 51:41-45.

Smith J. D., and R. Rupps. 1978. Antagonism in Marasmius oreades fairy rings. J
Sports Turf Res Int. 65:161-164.

Smith, J. D., N. Jackson, and A. R. Woolhouse. 1988. Fungal Diseases of
Amenity Turfgrasses. Chapman and Hall. London.

Tucker, K. A., K. J. Kamok, D. E. Radcliffe, G. Landry Jr., R. W. Roncadori, and
K. H. Tan. 1990. Localized dry spots as caused by hydrophobic sands on
bentgrass greens. Agron. J. 82:549-555.

USGA. 1960. Specifications for a method of putting green construction. USGA J.
Turf Manage. 13(5):24-28.

Wilkinson, J. F., and R. H. Miller. 1978. Investigation and treatment of localized
dry spots on sand golf greens. Agron. J. 70:299-304.

York, C. A. 1993. A questionnaire survey of dry patch on golf courses in the
United Kingdom. J. Sports Turf Res. Inst. 69:20-26.

York, C. A., and N. A. Baldwin. 1992. Dry patch on golf greens: a review. J.
Sports Turf Res. Inst. 687-19.

51

CHAPTER THREE

THE EFFECTS OF NITROGEN PLACEMENT AND WATER INJECTION
CULTIVATION ON CREEPING BENTGRASS PUTTING GREENS

ABSTRACT

Nitrogen is usually the most limiting nutrient in turfgrass growth. Food and
forage crops have benefited from receiving subsurface nitrogen fertilization.
Subsurface fertilization in turfgrass had not been feasible until the development
of high pressure water injection cultivation (W IC). Previous research showed
that applying nitrogen with WIC increased turf color ratings, clipping yields, and
nitrogen content compared to surface fertilization, but the effects of nitrogen
placement and cultivation were confounded. In addition, turf injected with
nitrogen exhibited occasional dark green striping that would be unacceptable in a
putting green situation. The objectives of this research were to evaluate (i) the
separate effects of nitrogen application method and WIC on growth, nitrogen
content in the plant tissue, turf color, and turf quality and (ii) the effects of several
alternative methods of injecting nitrogen on uniformity of color response. In a
nitrogen injection - management practice study, four combinations of nitrogen
injection, WIC, and surface fertilization were applied at 2.4 and 4.8 g N rn'2
application". In 1997, nitrogen injection increased turf color and quality ratings,
clipping yields and leaf nitrogen content, but in 1998 only increased turfgrass
color. In both years, WIC alone had no significant effect on any turfgrass

evaluations. In an application uniformity study, an application method providing

52

both injection and surface application of nitrogen provided superior color, quality,
and uniformity of putting green turf. If injecting nitrogen improves turf responses
compared to surface applications, less overall nitrogen may be needed to

maintain a healthy turf stand.

53

INTRODUCTION

Nitrogen is usually the limiting nutrient for turfgrass growth. It Is an
essential component of chlorophyll, other proteins, genetic material, and many
other plant substances. Turfgrass plants normally contain more nitrogen, 3 to 5
percent, than any other mineral nutrient. Responses to nitrogen by turf include
darker green color and increased growth and density. Too much nitrogen can be
detrimental by enhancing some diseases or causing osmotic burn (Emmons,
1995).

Turfgrass fertilization is traditionally accomplished through surface
applications. This is primarily due to the unavailability of equipment capable of
placing fertilizers below the soil surface without causing significant turf disruption.
Subsurface applications of nitrogen increase plant nitrogen use efficiency in the
food and forage crop industries (Mengal et al., 1982; Stecker et al. 1993).

Rapid suburban growth and golf course construction has increased
nitrogen fertilizer use in the turfgrass industry during recent decades. More
efficient nitrogen application methods on turfgrass, and the subsequent
conservation of nitrogen in the turfgrass industry, could save energy and reduce
risks of environmental pollution.

The introduction of water injection cultivation (WIC) (Murphy and Rieke,
1994) to the turfgrass management marketplace makes possible subsurface
placement of soluble materials in established turf. Although WIC was introduced
purely as a tool for soil cultivation, previous studies have concluded that injecting

soluble nutrients with WIC may be beneficial to turf (Miller, 1994).

54

Most literature available on fertilizer placement in turfgrass concerns
turfgrass establishment, i.e. fertilizer placement effects on seed germination or
sod establishment (Jackson and Burton, 1962; King and Skogley, 1969; King and
Beard, 1972) This work has shown minimal differences in turfgrass
establishment with regard to fertilizer placement. In contrast, Murphy and Zaurov
(1994) observed in a greenhouse study that perennial ryegrass (Lolium perenne
L.) receiving subsurface nitrogen injections (urea) had higher clipping yields,
greater root mass, higher nitrogen accumulation in plant tissues, and higher
water use rate efficiency than turfgrass receiving surface applications of nitrogen.

Studies examining the effects of injecting nitrogen with WIC on fairway
and putting green turfs were conducted in 1994 (Karcher, 1997). Treatments
included three rates of urea, either injected or surface applied. Plots injected
with urea had consistently higher clipping yields, nitrogen content in plant tissues,
and color ratings than plots receiving surface applications. One possible
explanation for these differences could have been as a result of ammonia
volatilization from surface applications, even though plots were irrigated shortly
after application. This hypothesis was tested by repeating the study in 1995
using ammonium nitrate as the nitrogen source, which is much less susceptible
to volatilization than urea. Results from the 1995 study were very similar to those
recorded in 1994. Clipping yields, nitrogen content in plant tissues and color
ratings were all increased by injecting ammonium nitrate. Plots injected with

nitrogen had a longer duration of a dark green turf response than plots receiving

55

surface applications during both years. Additionally, turf injected with nitrogen
was less susceptible to moisture stress than turf receiving surface applications.

These results suggest that by injecting nitrogen, a turfgrass manager may
be able to use less total nitrogen and increase water use efficiency when
compared to making surface applications. Plots receiving surface applications of
nitrogen in these studies were not subjected to WIC treatment with water alone.
Therefore, the effects of placement of nitrogen beneath the surface, and soil
aerification from WIC could not be separated.

In previous studies involving application of nitrogen with WIC, the turf
exhibited striping due to the nozzle alignment of the WIC unit on some dates.
Striping was most evident on closely mowed putting green turf, 5 to 14 days
following application. Turf striping occasionally reduced surface uniformity on
putting green turf to a level likely considered unacceptable by most turf
managers.

A group of studies were initiated in 1997 to compare the effects of surface
application and subsurface injection of nitrogen. The overall objective of these
studies was to determine if nitrogen application via injection is a practical and
improved means of fertilizing turfgrass. More specifically, they were to compare
injection and surface application of nitrogen by evaluating: (i) the separate effects
of nitrogen application method and WIC on growth, nitrogen content in the plant
tissue, turf color, and turf quality in a nitrogen injection - management

practice study and (ii) the effects of several alternative methods of injecting

56

nitrogen on uniformity of turf color and quality in an application uniformity

study.

MATERIALS AND METHODS
Nitrogen Injection - Management Practice Study
EXPERIMENTAL AREA

The nitrogen injection - management practice study was initiated in May
1997 at the Hancock Turfgrass Research Center (East Lansing, MI) on a one
year old ‘Penncross’ creeping bentgrass (Agrostis palustris Huds.) putting green
established on a root zone meeting USGA specifications (96% sand, 3% silt, 1%
clay) (Hummel, 1993). The experimental area was mowed at 4 mm and
maintained under typical putting green management practices. Pesticides were
applied on a curative basis and phosphorus and potassium were applied as
recommended from soil test values. Light sand topdressing applications were
made monthly with sand matching the texture of the root zone. Irrigation was
applied to approximate water loss due to average daily evapotranspiration.
TREATMENT STRUCTURE

This study contained two treatment factors, management practice and
nitrogen rate. There were four management practices: 1) surface applied
nitrogen without supplemental WIC, 2) surface applied nitrogen with
supplemental WIC using a standard, #53 nozzle (approximately 15 cm injection
depth), 3) nitrogen injected using a #56 nozzle (approximately 7.5 cm injection

depth), and 4) nitrogen injected using a #53 nozzle. There were two nitrogen

57

rates: 1) 2.4 g m2 application" and 2) 4.8 g rn'2 application". Combining the two
factors yielded eight individual treatments (Table 9). This treatment arrangement
allowed specific and separate analyses of the effects of nitrogen placement and
WIC. The effects of nitrogen placement were tested by contrasting treatments
#1, #2, #3, and #4 vs. #5, #6, #7, and #8. The effects of WIC were tested by

contrasting treatments #1 and #2 vs. #3 and #4.

Table 9. Summary of treatments comprising the nitrogen injection - management practice study.

 

Treatment No. Management Practice Nitrogen Rate

 

g rn'2 application"

1. N applied on surface with no WIC 2.4
2. N applied on surface with no WIC 4.8
3. N applied on surface plus WIC 2.4
4. N applied on surface plus WIC 4.8
5. N injected with #56 nozzle 2.4
6. N injected with #56 nozzle 4.8
7. N injected with #53 nozzle 2.4
8. N injected with #53 nozzle 4.8

 

The nitrogen source for all applications was ammonium nitrate. Nitrogen
applications were made once a month throughout the growing season. Fertilizer
injections and WIC were done with a HydroJect 3000® provided by the Toro Co.
of Minneapolis. Nitrogen injections were achieved by pumping dissolved
ammonium nitrate solution from a mounted tank to the intake line of the
HydroJect. The HydroJect was operated at the closest hole spacing
(approximately 7.5 cm x 2.5 cm). Surface applications were made using a 002
powered sprayer designed specifically for small plot applications. Approximately
5 mm of water were applied to the experimental area immediately following
nitrogen applications. On plots receiving surface applications of nitrogen plus

WIC, WIC was applied immediately following irrigation. Treatments were

58

replicated four times in a randomized complete block design. Individual plot
sizes were‘3.6 by 1.7 m.

Since the experimental area was extremely nitrogen deficient at the
beginning of the study, 38 g rn‘2 nitrogen was applied in 1997, whereas 24 g rn'2
was applied in 1998. Treatments were applied on 2 May, 28 May, 27 Jun., 31
July, 25 Aug., 25 Sep., and 12 Nov in 1997 and 9 May, 5 Jun, 10 July, 9 Aug.,
and 15 Sep. in 1998. The November 1997 application was a double rate late fall
application.

TREATMENT E VAL UA TIONS

Clippings were collected by mowing two passes lengthwise on each plot
with a greens mower once a week from May through October. Clippings were
dried at 60° C and weighed to determine yield. Clipping yields were evaluated on
22 dates in 1997 and 18 dates in 1998.

Plant tissue nitrogen content was determined from the dried clippings
using a Karsten Model 591 NIRS analyzer (Karsten Inc., Phoenix, AZ). Leaf
nitrogen content was evaluated on 15 dates in 1997 and 18 dates in 1998.

Turfgrass quality and color ratings were taken weekly throughout the
growing season. The rating scale for quality was from 1 to 9 (1=dead, 2=mostly
dead, 3=severely flawed, 4=flawed, 5=slightly flawed, 6=acceptable, 7=good,
8=excellent, 9=ideal) and for color was from 1 to 9 (1=tan, 2=greenish yellow,
3=yellowish green, 4=light green, 5=medium light green, 6=medium green,
7=medium dark green, 8=dark green, 9=extremely dark green). Turf quality and

color were evaluated on 21 dates in 1997 and 17 dates in 1998.

59

Application Uniformity Study
EXPERIMENTAL AREA

The application uniformity study was conducted in June through October
in 1997, and September through October in 1998 at the Hancock Turfgrass
Research Center on a 14-year old annual bluegrass (Poa annua L. reptans) turf
established on a fine-loamy, mixed, mesic Typic Hapludalf (68% sand, 19% silt,
13% clay). The experimental area was mowed at 4 mm and managed under
typical putting green management practices. Pesticides were applied on a
curative basis and phosphorus and potassium were applied as recommended
from soil test values. Light sand topdressing applications were made monthly
(96% sand, 3% silt, 1% Clay) and irrigation was applied to approximate water loss
due to average daily evapotranspiration.
TREATMENT STRUCTURE

Seven nitrogen application methods are evaluated in the 1997 study,
whereas nine were evaluated in 1998 (Table 10). Alternative application
treatments were chosen based on their potential to reduce the appearance of
green stripes following nitrogen injection. Nitrogen injections were made with a
HydroJect 3000. Nitrogen injections were achieved by pumping dissolved
ammonium nitrate from a mounted tank to the intake line of the HydroJect. The
HydroJect was operated at the closest hole spacing (approximately 7.5 cm x 2.5
cm).

Treatments #4 and #5 used experimental nozzles manufactured by the

Tom 00., with 2 orifices and 3 orifices, respectively. The orifices on these

60

nozzles were arranged to affect the largest volume of soil possible. Treatments
#6 and #8 involved turning on the roller washers of the HydroJect, which resulted
in approximately half of the nitrogen being applied on the turf surface. This was
determined by comparing the volume of water entering the roller washer to the
volume of water entering the HydroJect. The normal function of the roller
washers is to clean the rollers on which the unit rides during WIC.

The application rate for all treatments was 4.8 g N m'z. Treatments were

 

replicated four times in a completely randomized design. In 1997, treatments
were applied on 25 June, 15 Aug., and 25 Sep. The direction of nitrogen
injection was alternated 180 degrees between consecutive application dates.
Treatments were applied only on 10 Sep. in 1998. In both years, nitrogen was
applied on the experimental area at 3.6 g m2 every six weeks, from early May

until one month prior to treatment applications.

Table 10. Treatments comprising the application uniformity study.

 

Treatment N 0; Application Method

 

Surface application

Injected with #53 nozzle (approximately 15 cm depth)
Injected with #56 nozzle (approximately 7.5 cm depth)
Injected with 2 orifice prototype nozzle

Injected with 3 orifice prototype nozzle

Injected with #53 nozzle at half rate making two passes in perpendicular directions
Injected with #56 nozzle while surface roller washers on (using nitrogen solution)f
Injected with #56 nozzle at half rate making two passes in perpendicular directionsf

 

1
2
3
4
5.
6. Injected with #53 nozzle while surface roller washers on (using nitrogen solution)
7
8
9
T

Treatments were only applied in 1998 study.

Visual quality, color, and stripe ratings were taken weekly following
treatment applications. Quality and color ratings were taken in the same manner
described in the nitrogen injection - management practice study. A scale of 1 to

5 was used to evaluate turfgrass striping with 1 representing no discernible

61

striping, 2 representing barely discemible striping, 3 representing fairly
discernible striping, 4 representing easily detected striping, and 5 representing
obvious striping with sharp contrasting stripe borders. Ratings were taken on 13
dates in 1997 and 6 dates in 1998.

Statistical Analyses

Visual rating data for both studies were analyzed using the proportional
odds model that is incorporated into the Rating Data Analysis File Package
(Karcher, 2000). Treatment separation was done with pairwise chi-square tests
of the treatment parameter estimates. Probability distributions were constructed
to represent the odds of a treatment level to be rated in a particular category.
These distributions were constructed by inserting the appropriate combination of
parameter estimates into the Iogit-link function.

For the application uniformity study, maximum likelihood calculation errors
occurred using the full model. This was due to a relatively large ratio of model
parameters to experimental units. Therefore, a reduced model was used by
dropping the application method x rating date interaction term from the model,
and subsequently, only estimates of main effects were possible.

All other data were analyzed with ANOVA. If treatment effects were
significant, means were separated using LSD at the 0.05 probability level.
Where repeated measures were made on the same experimental units, time was
analyzed as a sub-plot factor of the experiment. The best fitting covariance
model among compound symmetry, first order auto-regressive, and spatial

exponential was used to fit correlations among time points. The best fitting

62

covariance model was determined by the highest Akaike’s Infomiation Criteron

value (Littell et al., 1996).

RESULTS AND DISCUSSION
Nitrogen Injection — Management Practice Study

Treatment main effects, nitrogen placement contrasts, and ANOVA results
are summarized in Table 11 (1997) and Table 12 (1998). The higher nitrogen
rate significantly increased the probability of the turf to be rated high in color and
quality, clipping yields, and leaf nitrogen content in both years. Management
practices effects were significant for all evaluations in 1997, but only color and
quality in 1998. The management practice x nitrogen rate interaction was
significant for all evaluations in 1997 and 1998, with the exception of leaf nitrogen
content in 1997. The nitrogen placement contrast was significant for all
evaluations in 1997, but only turfgrass color in 1998. Significant interactions
involving rating date resulted from the lack of treatment effects on a few rating
dates throughout the year. However, within years, treatment separation was
consistent on rating dates when treatment effects were significant.
COLOR FIA TINGS

Fertilizing turf with the high nitrogen rate increased the probability of being
rated as medium green or better by approximately 85% over the low rate in both
years (Figure 19 and Figure 20). Much previous research has demonstrated that
creeping bentgrass has a dark color response to increased rates of nitrogen

(Landschoot and McNitt, 1997; Powell et al., 1967, Brauen et al., 1975).

63

 

Table 11. Treatment main efl'ects, and nitrogen placement contrast for turf quality, color, clipping
yields, and nitrogen content. 1997.

 

 

I 1i 1' . .
Effect (1 f Co or df Qua ty df C 1pp1ng df Nitrogen
Parameter Parameter Yield Content
estimate estimate g 111'2 day" %
Management practices
surface N, no WIC 0.98 Bi 3.00 D 2.84 B 4.02 B
surface N, + WIC 0.31 C 2.08 C 2.69 B 4.04 B
injected N to 7.5 cm -5.00 A -1.72 A 4.40 A 4.41 A
injected N to 15 cm ~3.92 B -1.17 B 4.26 A 4.42 A
Nitrogen rate
2.4 g m‘2 app‘l 0.66 B 2.67 B 2.83 B 3.96 B
4.8 g m‘2 app" -4.47 A -1.57 A 4.27 A 4.48 A
Nitrogen placement contrast
surface 0.65 B 2.54 B 2.76 B 4.03 B
injected -4.46 A -1.44 A 4.33 A 4.41 A
ANOVA
Source of variation
BlOCk 3 Mn! 3 *t* 3 18* 3 #
Management practice (mp) 3 *** 3 *** 3 *** 3 ***
Nitrogen mm (m) 1 think 1 *** 1 *** 1 *8"!
mp X m 3 "a: 3 aural: 3 "at: 3 NS
Date (d) 20 *** 20 "* 21 *** l4 ***
mp x d 60 *** 60 *** 63 *** 42 ***
m X d 20 *** 20 tint! 21 *t* 14 tilt
mp x nr x d 60 *** 60 *** 63 *“ 42 **

 

1 Within effects, means sharing a letter are not significantly different.
*, **, *** Significant at the 0.05, 0.01, and 0.001 probability levels, respectively.

Table 12. Treatment main efl'ects, and nitrogen placement contrast for turf quality, color, clipping
yields, and nitrogen content. 1998.

 

 

Color Quality Clipping Nitrogen
Effec‘ df Parameter Parameter Yield Content
estimate estimate g 111‘2 day" %
Management practices
surface N, no WIC 0.16 Cl 0.12 C 5.24 A 4.57 A
surface N, + WIC -0.71 B -l.51 A 5.24 A 4.57 A
injected N to 7.5 cm -l.57 A -l.12 AB 4.98 A 4.53 A
injected N to 15 cm 074 B -0.88 B 5.10 A 4.51 A
Nitrogen rate
2.4 g m'2 app'1 1.02 B 0.76 B 4.18 B 4.38 B
4,3 g my2 app-1 -2.46 A -2.46 A 6.10 A 4.72 A
Nitrogen placement contrast
surface -0.28 B -0.70 A 5.24 A 4.57 A
injected -1.16 A -l.00 A 5.04 A 4.52 A
ANOVA
Source of variation
Block 3 NS 3 ** 3 *** 3 NS
Management practice (mp) 3 *** 3 *** 3 NS 3 NS
Nitrogen rate (nr) 1 *** 1 "** 1 *** 1 ***
mp X m 3 *** 3 *** 3 *1”: 3 It
Date ((1) 16 *** 16 *** l7 *** l7 ***
mp x d 48 *** 48 *** 51 NS 51 NS
m X d 16 *** l6 *** 17 *** l7 ***
mpxnrxd 48 NS 48 * 51 * 51 NS

 

f Within effects, means sharing a letter are not significantly different.
*, **, *** Significant at the 0.05, 0.01, and 0.001 probability levels, respectively.

65

B I' .
Elltgeen Bmedltgreen Imedgreenlmeddkgreen

a:
.E
..
E
u
.E
.2
o
o
9
._
o
a?
:5
in
.o
9
:1.
o
.2
..
2
3
E
3
0

Management Practices — N rate (9 / sq. m / app)

1- Within effects. bars sharing a letter are not simifiantly different (P < 0.05)

 

Figure 19. Color rating probability distributions for management practice and nitrogen rate main
effects. 1997.

W
Ellt green Elmed It green Imed green Imed dt green

Cumulative probability of receiving ratlng

\" N"
\V‘ \Y‘
\ 6‘2 a9

6“ e
6?: Q
of “€96

Management Practices —— N rate (9 l sq. m / app)

 

1 Within effects. bars sharing a letter are not simificantly different (P < 0.05)

Figure 20. Color rating probability distributions for management practice and nitrogen rate main
effects. 1998.

66

In both years, turf injected with nitrogen to a 7.5 cm depth had the highest
probability to be rated dark in color, followed by turf injected with nitrogen to a 15
cm depth, turf fertilized on the surface with nitrogen plus WIC, and turf fertilized
on the surface without additional WIC. The differences among treatments were
greatest in 1997. There is an extremely small amount of published work
regarding the effects of nitrogen placement in turf on color and quality. In
previous studies, Karcher (1997) found that injecting creeping bentgrass putting
green turf with urea or ammonium nitrate resulted in darker green turf compared
to surface applications. Injection to 7.5 cm probably concentrated the nitrogen
solution in closer proximity to the majority of the creeping bentgrass roots,
allowing for greater cumulative uptake compared to surface applications or
injection to a 15 cm depth. Previous rooting studies on creepingbentgrass turf
have determined the majority of the root mass is within 10 cm of the turf surface
(Cooper et al., 1998). Surface applications may have resulted in more microbial
immobilization compared to nitrogen injection since microbial populations are
greater near the turf surface than at soil depths of 7.5 cm or more. Thatch was
found to contain; up to 1600, 600, and 100 times as many bacteria, fungi, and
actimomycetes, respectively than soil on a creeping bentgrass putting green with
a sand based root zone (Mancino et al., 1993). Harper et al. (1996) discovered
that a significant amount of nitrogen applied to a perennial ryegrass pasture was
immobilized by microbes, and remobilized at insufficient rates to avoid nitrogen

stress by the grass.

67

The significant management practice x nitrogen rate interaction in 1997
resulted from nitrogen injected to a 7.5 cm depth increasing turf color compared
to the 15 cm depth at the low nitrogen rate, but not at the high rate (Figure 21).
These results suggest that nitrogen placement to an optimum depth is more
critical when fertilizing at relatively low rates. The significant management
practice x nitrogen rate interaction in 1998 resulted from WIC significantly
increasing turf color on turf receiving surface fertilization at the low nitrogen rate,
but not at the high rate (Figure 22).

QUALITY RA TINGS

Turf fertilized with the high nitrogen rate increased the probability of being
rated as acceptable or better by approximately 55% In .1997 (Figure 23) and 85%
in 1998 (Figure 24) over the low nitrogen rate. Similar results were reported in
previous studies examining the effects of nitrogen on creeping bentgrass quality

(Waddington et al., 1978; Christians et al., 1981).

68

Weenies

El It green Elmed It green I med green Imed dk green

c»
.E
..
us
..
I.»
.E
.2
o
0
cu
._
.—
o
:E‘
B
a
.o
o
h
a
a)
.2
..
r_e
3
E
:I
O

WNIE‘) tIIIINNICD) WNIE) WNIBC) mute) WNW anIDI "VINIAI
suface N. -WICT surface N. +WIC Injeaed N (7.5 cm) injected N (15cm)

Level of management practice x N rate

1 Bars sharing a letter are not significantly different (P < 0.05)

 

Figure 21. Color rating probability distributions for the management practice in nitrogen rate
interaction. 1997.

in ori

Elli green timed It green Imed green Imed dk green

Cumulative probability of receiving rating

IaaNlEtf WNW) lowNID) NohNIAB) INNIC) WNW IowNICI NohNIB)
surface N. -wrc surface N, +WIC Injected N (7.5 cm) injected N (15 an)

Level of management practice x N rate

1 Bars sharlng a letter are not simificantly different (P < 0.05)

 

Figure 22. Color rating probability distributions for the management practice x nitrogen rate
interaction. 1998.

69

W
Elsev flawed Cl flawed I sli flawed I accpetable I good

a:
E
..
as
b
c:
E
2
o
o
a
‘-
._
o
g
E
in
.o
9
a.
o
.2
..
S
3
E
3
O

6900‘
Management Practices —— N rate (9 / sq. m / app)

T Within effects. bars sharing a letter are not significantly difterent (P < 0.05)

 

Figure 23. Quality rating probability distributions for management practice and nitrogen rate main
efi’ects. 1997.

 

 

   

 

 

 

 

 

 

 

 

 

 

E I' .
Eltiawed Elsli flawed Iaccpetable Igood
m 100%
.E
E 90%
E”
E 80% .
8 70%
‘2 1
3 60%
g 50%
.8 40% , ,, ‘_ — fl
2 ._ ‘ "
D- 30% ' —— . ' —
g, . . 1 _ .
'a 20% fl ‘ >
2 ~. _ . ‘ ‘ ,.,».‘..,.,.,s
E 10% ' " _‘ -' Lfl
: - - ‘ - “ _
0 0% , . .. .
x \ \ \ K N
e +9 e to e e
0‘ \\ if“ $0 ,‘P‘ a,"
x 9 Q at
<5" G 0° (to
(<0 «P o
e 6
9 \AOdk \éeé

Management Practices _ N rate (9 / sq. m / app)

1 Within effects. bars sharing a letter are not simiflrantty ditterent (P < 0.05)

 

 

 

Figure 24 Quality rating probability distributions for management practice and nitrogen rate main
effects. 1998.

70

In 1997, turf injected with nitrogen to a 7.5 cm depth had the highest
probability to be rated high in quality, followed by turf injected with nitrogen to a
15 cm depth, turf fertilized on the surface with nitrogen plus WIC, and turf
fertilized on the surface without additional WIC. Each management practice was
significantly different from the others. More cumulative uptake of nitrogen in
1997 of turf receiving injected nitrogen to a 7.5 cm depth could likely have
produced more chlorophyll and increased turf density, improving turf quality.
However, in 1998, surface application of nitrogen plus WIC had quality equal to
the 7.5 cm injection treatment. Although turf injected with nitrogen had better
color than turf receiving surface applications, striping on several dates resulted in
equal quality among the application methods. Previous studies have shown
variable differences in quality between application methods due to striping of the
turf caused by injecting nitrogen (Karcher, 1997).

The significant management practice x nitrogen rate interaction in 1997
resulted from the 7.5 cm injection depth significantly increasing turf quality
compared to the 15 cm depth at the low nitrogen rate, but not at the high rate
(Figure 25). The significant management practice x nitrogen rate interaction in
1998 resulted from turf fertilized on the surface with nitrogen having quality equal
to both injected nitrogen treatments at the high nitrogen rate, but significantly
lower quality at the low nitrogen rate (Figure 26). These results suggest with
extended use at low nitrogen rates, injecting nitrogen to a 7.5 cm depth may
result in optimal quality, whereas at high nitrogen rates application method has

little effect on turf quality.

71

W
D sev flawed El flawed Isli flawed I aocpetable I good

or
E
..
a
5.
c»
.S
.2
a.)
O
o
_
._
o
E
5
(U
.n
o
a
a.
a:
.2
..
5
3
E
3
O

IWNiE') WMCD) |WME) WHNCBC) RIMS) HohNtA) IowaD) WNW
surface N, -WICt curiae N, 4,ch injected N (7.5 cm) injected N (15 cm)

Level of management practice x N rate

T Bars sharing a letter are not significantly different (P < 0.05)

 

Figure 25. Quality rating probability distributions for the management practice it nitrogen rate
interaction. 1997.

 

     

Rating cajggorigs
D sev flawed El flawed I sli flawed I accpetable I good

W I I -
E 90%
‘5: I I -
a, 80%
s I I -
g 70%
0
. I I -
z 60% .
° I I ,-
E 50%
s v I I -
.o 40% ' ..» .
9 I I ‘-
3 30% ‘
I I -
<3 20%
a I I -
3 10%
° ‘I -

Ioleoit victims) mule) mum Iowa) WNfBl mute) highNtB)
surface N, -WlC suface N, oWlC iriected N (7.5 cm) injected N (15 cm)

Level of management practice x N rate

 

 

f Bars sharing a letter are not slgnlficantty different (P < 0.05)

 

Figure 26. Quality rating probability distributions for the management practice x nitrogen rate
interaction. 1998.

72

CLIPPING YIELDS

Turf fertilized at the high nitrogen rate had 1.4 and 1.9 g m‘2 clay'1 more
clippings than the low rate in 1997 and 1998, respectively. A significant
production of creeping bentgrass leaf tissue in response to increasing nitrogen
levels has been previously reported (Carroll and Petrovic, 1991; Sheldrick et al.,
1990).

The significant management practice effect on clipping yields was due to
nitrogen placement rather than WIC (Table 11). In 1997, turf receiving
subsurface nitrogen placement averaged 1.6 g m'2 day'1 more clippings than turf
fertilized on the surface. However, this effect was not apparent in 1998, possibly
as the result of remobilized nitrogen in turf receiving surface applications. The
clipping yield effect seen in 1997 is similar to previous nitrogen injection studies -
(Karcher, 1997). in a greenhouse study, Murphy and Zaurov (1994) found that
applying nitrogen, potassium, and phosphorous at a 5 cm depth increased
perennial ryegrass clipping yields compared to surface fertilizations. More
cumulative uptake of nitrogen in 1997 by turf receiving injected nitrogen could
have produced more chlorophyll and increased turf growth compared to surface
applications. It is possible that nitrogen immobilized in 1997 was remobilized in
1998 causing turf receiving surface applications of nitrogen to produce clipping
yields equal to turf receiving injected nitrogen.

The significant management practice x nitrogen rate interaction in 1997
resulted from the 7.5 cm injection depth significantly increasing turf clipping yield

compared to the 15 cm depth at the low nitrogen rate, but the reverse effect

73

occurred at the high rate (T able 13). The significant management practice x
nitrogen rate interaction in 1998 resulted from turf fertilized on the surface with
nitrogen causing higher clipping yields than turf injected with nitrogen to a 7.5 cm

depth at the high nitrogen rate, but not at the low nitrogen rate.

Table 13 . Clipping yields as afl‘ected by management practice x nitrogen rate. 1997 and 1998.

 

Management practice

 

 

 

 

Nitrogen rate surface N, no WIC surface N, + WIC injected N to 7.5 cm injected N to 15 cm
3 m'2 day'1
1%
2.4 g my2 app" 2.3 Bi 2.1 E 3.7 c 3.1 D
48 g m'2 app'l 3.4 D 3.2 D 5.0 B 5.4 A
l 98
2_4 g 111'2 app'l 4.1 CD 4.0 D 4.3 C 4.3 C
4.8 g m'2 app'l 6.4 A 6.5 A 5.6 B 5.9 B

 

1* Within years, means sharing a letter are not significantly different (P < 0.05).

LEAF NITROGEN CONTENT

The high nitrogen rate increased nitrogen content in the leaf tissue 0.5 and
0.4 % in 1997 and 1998, respectively. Significant differences in leaf nitrogen
content among management practices in 1997 were the result of nitrogen
placement rather than WIC (Table 11). In 1997, injecting nitrogen resulted in a
0.4% increase in leaf nitrogen content compared to surface applications. A 0.4%
increase in leaf nitrogen content over surface applications is very similar to
previous nitrogen injection studies (Karcher, 1997). Murphy and Zaurov (1994)

reported significant increases in nitrogen accumulation in perennial ryegrass leaf

74

tissue when fertilizer was placed at 5 or 10 cm depths compared to on the
surface. There was no management practice effect in 1998.

A significant management practice x nitrogen rate interaction in 1998
resulted from surface applications significantly increasing leaf nitrogen content
compared to injecting nitrogen at the high fertilization rate, but not at the low

nitrogen rate (Table 14).

Table 14. Leaf nitrogen content as affected by management practice It nitrogen rate. 1998.

 

 

 

 

Management practice
Nitrogen rate surface N, no WIC surface N, + WIC injected N to 7.5 cm injected N to 15 cm
%
2.4 g rn'2 app'I 4.4 C 4.4 C 4.4 C 4.4 C
43 g m'2 app" 4.8 A 4.8 A 4.6 B 4.7 B

 

T Within years, means sharing a letter are not significantly different (P < 0.05).

Application Uniformity Study

Application method significantly affected turf color, quality, and striping in
both years (Table 15). Since parameter estimate calculation by maximum
likelihood was not possible with the full statistical model, no information was
available regarding the application method x rating date interaction.
COLOR RATINGS

Turf injected with nitrogen and with the roller washers turned on had the
highest probability of being rated dark in color during both years. In 1997,
_ (Figure 27), surface applications of nitrogen were equal to injecting with the roller

washers on, whereas in 1998 (Figure 28) a standard injection and injecting at a

75

half rate in two directions were equal to injecting with the roller washers on.
Surface application and injection with a 3 orifice prototype nozzle resulted in
significantly poorer color ratings than all other treatments in 1998.
QUALITY RA TINGS

Turf injected with nitrogen with the roller washers on had the highest
probability to be rated high in quality in 1997 (Figure 29) and 1998 (Figure 30).
The multiple orifice nozzles ranked relatively poor in quality among all treatments
across both years. Although turf receiving surface application of nitrogen had
equal quality to turf injected with nitrogen with roller washers on in 1997, turf
treated with surface applications had relatively low probability to be rated high in
quality in 1998.
STRIPE RA TINGS

Turf injected with the standard nozzle (15 cm depth) and the 7.5 cm depth
nozzle had the highest probabilities to be rated high in surface striping in both
years (Figure 31 and Figure 32). Turf injected with nitrogen and with the roller
washers on had an equal probability of striping to turf receiving surface
applications in both years. Turf fertilized with the multiple orifice nozzles had
more striping than turf fertilized with surface applications in 1997, but was rated

equal in striping to turf fertilized with surface applications in 1998.

76

Sac—o Eu m: 2N8: 22.5% 5.3 25v Sac—=28: 3.8? .cmEoEO “.29. 32.5 a
.God v at East? 358%.me .0: 2a 2.2 a menace. v.52: .8528 55$» 2.
2.2.8.22 .22.... £522.. .8... 2a .5... .8... 2.. a sesame 2.... .2. ..

 

 

.12.. m .12.. m .12.. m 12.. e .12.. N— .12.. mm 2mm—
...1. w .12.. w .12.. w .12.. c .12.. c .12.. 0 8502 c2323?
8:22? we 8.8m
2.300—
U 36 Q 9:. m vnd- .: l- -l 253026 N EEO: :8 WC 8.8%:
U de— m< mod- < mwfi- l- i- -l Eur—ma? + 3.30: Eu OS 8.8%:
m mud m< wed- < mad- U mad m No.7 m Rum 25:09.3 N BUOME
U 32: < m9? .4. nafi- D mofi < Cam- < 84 «.8533 + Boomfi
U 3.2 m _w.n Q 36 Q 8% m 8.7 m MNé Any—Nae: 02.3.5 9 302.5
U cod m Nvd U hwg- D mwé m mm._- m NWM GEO: nocto NV 8805
< mm..v Um OWN- < wmfi- < 5.0 U 3.9. m beam GENO: Eu Wt v0.8.2:
< 32v Q0 Sud- < 2%. m No; Um 8.7 m mw.m ”2682.5
0 No.3 m :6 Q wbfi O mmd— < Emu- ._.< OWN 885w
82.22 22.8.32
22:28 255.8 25:38 2.258 228:8 22:58
MOHBEHNAM huugnw hogaam uOHQCMme huughflnm uvugnm
~00
02.5% .3 5:2..0 .6 .530 .6 093m .6 .0330 .3 5.00 at am
M30— hoo—

 

 

 

 

 

dma can 83 .352: out? ES 35:25 :38 ~25 :e 8329 62:2: 553:9: .m— 023—.

 

 

 

 

 

 

R in c ri s
a, Dyellow green E! It green I med It green I med green I med dk green
.3: 100%
g .
ca 90% .
C l
'5 80% ‘
.5 ‘
8 70% I
'5 60% ‘
g 50% ‘
la? 40% ‘
'8 .
a 30%
.3 20% .
g 10% ' 5; V —— l
E 0% ‘
O
.2 2 \‘8 \e e .69“ \e
.06“, $0 1330 -de . é" 1&0
99 9° (‘0 c9 «3} \o\ ¢°
«3 <0 \cP 6‘ e
°\\°‘ 6" 08‘ ‘7’
x \ {L {5
e“? “°‘ \° 2
1 Bars sharing a letter are not significantly dillerent (P < 0.05).
3 Unless noted otherwise. injected treatements clone with standard nozzle (15 cm).

 

 

 

Figure 27. Color rating probability distributions for application method efiects. 1997.

 

E I' | .
Elyellow green It green I med It green I med green Imed dk green Idk green

   

g
I!
9... = = -
~% 80%
‘5 m E E E
E 50%
«m = = =
D .
e 30%
3 m, I I I
§ 10% I I I
E 0% I I I
3 x \ x x \ x x \ x
0 \r‘ \r \r \r Q“ s \o o Q
e . e \ o6 \ \o\
9"“ ”if 430 ($ch 69‘" of? d”? «94 «96$
9 {L 0.3"
{9&3 @9696 6‘69 fix“ (919° 069 «PG
‘° 6“ J“ 6!"
f9 \ \
#96) 9°
\(‘t \af

1 Bars sharlng a letter are not significantly different (P < 0.05).
: Unless noted otherwise. Injected “eaten-tents done with standard nozzle (15 cm).

 

 

 

Figure 28. Color rating probability distributions for application method efl’ects. 1998.

78

 

Ra ' ri s
D sev flawed Cl flawed I sli flawed I acceptable I good I excellent

«
i.

- 1
--47

Cumulative probability ot receiving rating

    

. {5
\985‘ \°\ ‘\ odkeb . *6 ($606

t Bars sharing a letter are not significantly different (P < 0.05).
t Unless noted otherwise. injected treatements done with standard nozzle (15 cm).

 

 

 

Figure 29. Quality rating probability distributions for application method efl'ects. 1997.

 

W
Dsev flawed Iflewed Isli flawed Iaooeptable I good Iexcellent

 

 

 

 

 

 

i

iiii

 

Cumulative probability of receiving rating
IIIIIIIIII

 

1 Bars sharlng a letter are not significantly ditlerent (P < 0.05).
1 Unless noted otherwise. injected treatements done with standard name (15 em).

 

 

 

Figure 30. Quality rating probability distributions for application method eflects. 1998.

79

 

W
D none [3 slight I medium I significant I distinct

Cumulative probability of receiving rating

 

{2.

t Bars sharing a letter are not significantly different (P < 0.05).
1. Unless noted otherwise, injected treatements done with standard nozzle (15 cm).

 

 

 

Figure 31. Stripe rating probability distributions for application method efl’ects. 1997.

 

 

E i | .
g Elnone leight Imedium
§ 100% 2
g 90% ;
i 80%
g 70%
'6 60%
g 50% f
‘8 40%
D
e 30% .
O.
a) 20%
>
:5 10%
g 0%
o e 490 {he go
§y¢1&@fq§\:§9f . 0°
©oé°&°« .9496) £959: ””05“ :96
966‘
”£926 9&6!» 119$ (pity <91? afiéfip es?»
‘6‘ (9 6o“) \
\é\ «660 Y5; Q" #96
\'

is“ s“

1 Bars sharing a letter are not significantly dtterent (P < 0.05).
1: Unless noted otherwise, injected treatements done with standard nozzle (15 cm).

 

 

 

Figure 32. Stripe rating probability distributions for application method effects. 1998.

80

The visual rating data suggest that putting green surface uniformity can be
maintained when applying nitrogen via WIC by leaving the roller washers on.
This technique applies nearly one half the nitrogen on the surface while injecting
the remaining nitrogen to a 15 cm depth. Across years, this application method
had equal surface uniformity and superior quality and color compared to surface

applications.

CONCLUSIONS

Although results were variable, improved turfgrass responses from
injecting nitrogen seemed to be the result of nitrogen placement beneath the
surface rather than WIC. This is probably the result of concentrating nitrogen in
close proximity to the majority of active creeping bentgrass roots, resulting in
increased uptakeand decreased opportunity for microbial immobilization.
Equality among management practices in 1998 with regard to clipping yields and
leaf nitrogen content may have resulted from remobilization of nitrogen from
1997 applications.

Subsurface nitrogen placement was most beneficial compared to surface
applications at the low nitrogen rate. The current trend in nitrogen fertilization of
putting greens is a low nitrogen rate. From the data here, subsurface fertilization
may benefit turf managers using low nitrogen rates. However, the nitrogen
injection process used in these experiments was relatively labor intensive and
time consuming compared to traditional application methods and would likely be

considered impractical by turf managers.

81

The unacceptable surface uniformity seen in previous nitrogen injection
studies can be remedied by injecting a half rate of nitrogen while applying the
remainder on the surface. This was accomplished by injecting nitrogen with the

HydroJect while using the roller washers.

82

BIBLIOGRAPHY

Brauen, S. E., R. L. Goss, C. J. Gould, and S. P. Orton. 1975. The effects of
sulphur in combinations with nitrogen, phosphorus and potassium on colour
and Fusarium Patch disease of Agrostis putting green turf. J. Sports Turf
Res. Inst. 51:83-91.

Carroll, M. J., and M. A. Petrovic. 1991. Wear tolerance of Kentucky bluegrass
and creeping bentgrass following nitrogen and potassium application.
HortSci. 26:851-853.

Christians, N. E., D. P. Martin, and K. J. Kamok. 1981. The interrelationship
among nutrient elements applied to calcareous sand greens. Agron. J.
73:929-933.

Cooper, R. J., Chunhua Liu, and D. S. Fisher. 1998. Influence of humic
substances on rooting and nutrient content of creeping bentgrass. Crop Sci.
38: 1 639-1 644.

Emmons, Ft. 1995. Turfgrass science and management. 2"d ed. Delmar
Publishers. Albany, NY.

Harper, L. A.; D. W. Bussink, H. G. vanderMeer, and; W. J. Corre. 1996.
Ammonia transport in a temperate grassland. I. Seasonal transport in
relation to soil fertility and crop management. Agron. J. 88:614-621

Hummel, N. H. 1993. Rationale for the revisions of the USGA green
construction specifications. USGA Green Section Record. 31 :7-10.

Jackson, J. E., and Burton, G. W. 1962. Influence of sod treatment and nitrogen
placement on the utilization of urea nitrogen by coastal berrnudagrass.
Agron. J. 54:47-49.

Karcher, D. E. 2000. The proportional odds model: a statistically sound method
to analyze turfgrass rating data. p. 1-23. In Investigations on statistical
analysis of turfgrass rating data, localized dry spots of putting greens, and
nitrogen application techniques for turf. Ph.D. diss. Michigan State
University. East Lansing, MI.

Karcher, D. E. 1997. Nitrogen fertilization and soil cultivation utilizing high
pressure water injection. M.S. thesis. Michigan State Univ., East Lansing.

King, J. W., and J. B. Beard. 1972. Postharvest cultural practices affecting the
rooting of Kentucky bluegrass sods grown on organic and mineral soils.
Agron. J. 64:259-262.

83

King, J. W., and C. R. Skogley. 1969. Effect of nitrogen and phosphorus
placement and rates on turfgrass establishment. Agron. J. 61 :4-6.

Landschoot, P. J., and A. S. McNitt. 1997. Effect of nitrogen fertilizers on
suppression of dollar spot disease of Agrositis stolonifera L. Int. Turfgrass
Soc. Res. J. 8:905-911.

Littell, R. C., G. A. Milliken, W. W. Stroup, and R. D. Wolfinger. 1996. SAS
system for mixed models. Cary, NC. SAS Institute Inc.

Mancino, C. F., M. Barakat, and A. Maricic. 1993. Soil and thatch microbial
populations in an 80% sand: 20% peat creeping bentgrass putting green.
HortSci. 28:189-191.

Mengal, D. B., D. W. Nelson, and D. M. Huber. 1982. Placement of nitrogen
fertilizers for no—till and conventional till com. Agron. J. 74:515-518.

Miller, C. M. 1994. Turfgrass response to fertilization and cultivation using high
pressure water injection. M.S. thesis. Michigan State Univ., East Lansing.

Murphy, J. A., and P. E. Rieke. 1994. High pressure water injection and core
cultivation of a compacted putting green. Agron J. 86:719-724.

Murphy, J. A., and D. E. Zaurov. 1994. Shoot and root growth response of
preennial ryegrass to fertilizer placement depth. Agron J. 86:828-832.

Powell, A. J., R. E. Blaser, and R. E. Schmidt. 1967. Physiological and color
aspects of turfgrasses with fall and winter nitrogen. Agron. J. 59:303-307.

Sheldrick, R. D., R. H. Lavender, and T. M. Martyn. 1990. Dry matter yield and
response to nitrogen of an Agrostis stolonifera-dominant sward. Grass
and Forage Sci. 45:203-213.

Waddington, D. V., T. R. Turner, J . M. Duich, and E. L. Moberg. 1978. Effect of
fertilization on Penncross creeping bentgrass. Agron. J. 70:713-718.

84

APPENDIX A - “TURFRATESAS”

%macro turfrate( excelin=, data=, factors=, block=, mode1=,
response=RATING, respnum=, category=,
contrast=, slice=, effect=, means=YES,
titIe=, summary=no, excelout=NO, anova=,

path=c:\OrdinalAnalysis );

%macro space(NUM);

%do t=1 %to &NUM;

%put;

%end;

%mend space;

%space(2);

%put TURFRATE NOTE: The macro 'turfrate.sas' has been initiated. Questions regarding the
application of this macro should be directed to karcherdemsu.edu.;

%space(2);

%put TURFRATE NOTE: The files 'PropOddsModel.sas' and 'pdnix612.sas' must be located in
the sasmacro directory for proper analysis!;

%space(2);

/* SET OPTIONS */
options ls=84 ps=64 nocenter notes;
/* CATEGORIES */

%if %length(&category)“=o %then %do;

%1et category= %upcase(&category);

%1et n = 0;

%do g=1 %to %length(&category);

%let char=%qsubstr(&category,&g,1);

%if &char=%str( ) %then %do;

%let n = %eval(&n+1);

%1et indan = &g;

%end;

%end;

%1et indo = 0;

%1et ind%eva1(&n+1)= %eval(%length(&category)+1);
%do g=1 %to %eva1(&n+1);

%let f = %eval(&g-1);

%1et first=%eva1(&&ind&f+1);

%let second=%eva1(&&ind&g-&&ind&f-1);

%1et cat&g=%substr(&category,&first,&second);
%end;

%let respnum a %eva1(&n+1);

%end;

%else %if %1ength(&respnum)‘=0 2then %do;
%do c=1 %to &respnum;

%1et cat&c= P&c;

%end;

%end;

%else %do;

%put ERROR: EITHER CATEGORY NAMES 0R CATEGORY NUMBER MUST BE PROVIDED;

f

85

%goto exit;

%end;

%put TURFRATE NOTE: RATING CATEGORIES:;
%do d=1 %to &respnum;

%put &&cat&d;

%end;

%space(2);

%if %length(&effect)=0 %then %do;

%include “8path\PropOddsModel.sas';
%include '&path\pdmix612.sas';

/‘ FACTORS AND MODEL STATEMENT '/

21et factors= %qupcase(&factors);

%let n = 0;

%do g=1 %to %length(&factors);

%let char=%qsubstr(&factors,&g,1);

21f &char=%str( ) %then %do;

21et n = %eval(&n+1);

%let indan = &g;

%end;

%end;

%let indo = 0;

%let ind%eval(&n+1)= %eval(%length(&factors)+1);
%do g=1 %to %eva1(&n+1);

%let f = %eva1(&g-1);

%let first=%eval(&&ind&f+1);

%let second=%eval(&&ind&g-&&ind&f-1);

%let var&g=%substr(&factors,&first,&second);
%end;

%let varnun = %eva1(&n+1);

21f %length(&model)=0 %then %do;
%let model=;

%do p=1 %to &varnum;

%let model=&model &&var&p;

%let sp0=0; /* find the spaces */
%let n=0;

%do q=1 %to %length(&model);

%let char=%qsubstr(&model,&q,1);
%if &char=%str( ) %then %do;

%let n = %eval(&n+1);

%let sp&n = &q;

%end;

%end; /* end find the spaces */
klet spkeval(&n+1)= %eval(%length(&model)+1); /' parse model terms */
kdo g=1 %to %eva1(&n+1);

%let f = %eval(&g-1);

%let first=%eval(&&sp&f+1);

%let second=%eval(&&sp&g-&&sp&f-1);
tlet term&g=%qsubstr(&mode1,&first,&second);
%let termnun = %eval(&n+1);

kend; /* end parse model terns */
%let r= &teranum;

%let 1=1;

86

*do %while (&r > 1);

%1et mode1=&model &&term&l.%str(')&&var&p;
%let l=%eval(&l+1);

%let r=%eval(&r-1);

%end;

%end;

%if %length(&block)‘=0 %then %do;
%let model= &block &model;
%end;

%end;

%put TURFRATE NOTE: MODEL STATEMENT:;
%put MODEL &&response = &model;
%space(2);

/* END MODEL STATEMENT '/
/* CONVERT EXCEL DATA TO SAS DATA */

%if %length(&excelin)“=0 %then %do;
%put TURFRATE NOTE: CONVERTING EXCEL 95 WORKSHEET INTO SAS DATA SET;
%space(2)
libname convert "apath';
proc access dbms=xls;
create convert.xlsa.access;
path = '&excelin';

scantype = yes;
getnames = yes;
assign = yes;
unique = yes;

create Convert.xlsv.view;

select all;
run;
data data; set convert.xlsv; run;
proc datasets library=convert; delete xlsv / memtype=view;

delete xlsa / memtype=access; run; quit;

%let data = data;
%space(2);
tput TURFRATE NOTE: END EXCEL CONVERSION. CONVERTED DATA SET = WORK.OATA.;
%space(2);
%end; /* end excel conversion */

/* SHORTEN NAME IF FACTOR LENGTH = 8 */
%do h=1 kto avarnua;

*if %length(&&var&h)=8 %then %let var&h._ = _%substr(&&var&h, 1 ,7);
%if %1ength(&&var&h “=8 %then %1et var&h._ = _&&var&h;
%end;

/' END SHORTEN NAME */
/* DATA SUMMARY */

%if %upcase(&summary)=YEs %then %do;

%put TURFRATE NOTE: DATA SUMMARY REQUESTED. SUMMARIZING DATA SET NOW.;
%space(2);

title 'PROPORTIOINAL 0008 MODEL';

titles 'Rating Data Summary';

87

title4 "Data = adata';

proc freq data=&data noprint;

tables &response / expected out=freq;

run;

%do y = 1 %to &varnum;

proc freq data=&data noprint;

tables &response*&&var&y / out=freq&y outpct;

run;
data freq; set freq freq&y;

label pct_row = 'Percentage of Treatment Ratings';
label pct_col = 'Percentage of Category Ratings‘;
run;

%end;

proc print data=freq label;
var &response

%do y = 1 %to avarnum;

&&var&y

%end;

count percent pct_row pct_col;
run;

%end; /* end data summary */

/* INVOKE POM */
/* Test for contrasts and slicing */

%if %length(&contrast)‘=0 %then %do;
%let contrast= contrast acontrast%str(;);
%end;

%if %length(&slice)‘=0 %then %do;
%let slice= lsmeans &slice%str(;);
%end;

%if kupcase(&anova)=YES %then %do;

%put TURFRATE NOTE: ANOVA COMPARISON REQUESTED. PERFORMING ANOVA NOW.;
%space(2); ‘

proc glm data=&data outstat=avtest noprint;

class afactors;

model aresponse = &model / ss3 ;

run;

%end;

%put TURFRATE NOTE: BEGINNING POM ANALYSIS.;
%space(2);

%pom(data=&data,
procopt=order=data,
stmts=%str(

class &factors;
model &response = &model / s;
lsmeans &model / diff;
&contrast
aslice
).

tit1e=POM &title,

options=datasets predlsm noprint );

88

/“r END POM '/

/* RESET OPTIONS */

options ls=84 ps=64 nonotes nocenter;
/‘ OUTPUT VARIOUS TESTS OF EFFECTS 1'/

data tests; set _tests;

label _effect_ = 'Source of Variation';
label df = 'Degrees of Freedom';

label chi2 = “Chi-Square Value';

label P = 'P > Chi—Square';

run;

%if %upcase(&anova)*=YES %then %do;
title2 ' ';

tit1e3 'Analysis of Variation';
title4 'Data Set = &data';

proc print data=tests

label;

var _effect_ df ch12 p;

run;

%end;

%if %upcase(&anova)=YES %then %do;
title2 ' ';

titles 'Analysis Comparison: POM (POM) vs. Analysis of Variance (ANOVA)';
title4 'Data Set = &data';

data avtest; set avtest;

rename _source_ = _effect_;

n = _n_;

run;

proc sort data=avtest; by _effect_; run;
proc sort data=tests; by _effect_; run;
data bothtest;

merge tests avtest;

by _effect_;

drop _name_ _type_ SS;

label DF = 'Degrees of Freedom';

label F = 'ANOVA F Statistic';

label PROB = 'ANOVA (Prob>F)';

label CH12 = 'POM Chi-Square Statistic';
label P = 'POM (Prob>Chi-Square)';

proc sort data=bothtest; by n; run;

proc print label data=bothtest (where=(P ne .)) noobs;
var _effect_ df chi2 f p prob;

run;

%end;

%if %length(&contrast “=0 %then %do;
titles 'Test of Contrasts';

title4 'Data Set = &data';

data contr; set _contr;

label contrast = 'Source of Variation';
label df = 'Degrees of Freedom';

89

label ch12 = 'Chi-Square Value';
label P = 'P > Chi-Square‘;

proc print label; run;

%end;

%if %length(&slice)‘=0 %then %do;
titles 'Test of Effect Slices';
title4 'Data Set = adata';

data slices; set _slices;

label df = 'Degrees of Freedom';
label ch12 = 'Chi-Square Value';
label P = 'P > Chi-Square';

proc print label; run;

%end;

/* OUTPUT PREDICTED PROBABILITIES */

%if %upcase(&means)=YES %then %do;

%space(2);

%put TURFRATE NOTE: PERFORMING MEANS SEPARATION TESTS FOR PROBABILITY PARAMETERS.;
%space(2);

titleS'Predicted Category Probabilities (P = RATING CATEGORY)‘;
proc sort data=_probs; by _effect_ _p1; run;

proc format; value cat

%do v=1 %to &respnum; .
&v = '&&cat&v'

%end;

; run;

kdo x=1 %to avarnum;

title4 “for &&var&x effects“;

data probs&x; set _probs;

if _effect_ = '&&var&x';

array names[&respnum] P1-P&respnum;
array _P[&respnum] _P1-_p&respnum ;
do i = 1 to &respnum;

names[i] = _p[i];

end;

drop _P1-_P&respnum i;

%do n=1 %to &respnum;

%if &&cat&n ‘= Pan %then %do;
label Pan = '&&cat&n' ;
%end;-

%end;

drop _effect_ ,

label level = 'Level of &&var&x';

run;

/* Prepare lettering of treatment groups */
data lsm&x; set _lsm;

if _effect_ = '&&var&x';
rename value = _lsmean_;

90

run;
data diffsax; set _diffs;

if _effect_ = '&&var&x';

rename level = &&var&x;

rename _level = &&var&x._;

-PT_ = P;
run;
%pdmix612(diffs&x , 1sm&x , sort=yes)
data msgrp; set msgrp;
keep level _lsmean_ msgroup;
run;
proc sort data=probs&x ; by level; run;
proc sort data=msgrp; by level; run;
data probs&x; set probs&x ;
merge probs&x msgrp;

by level;
label _lsmean_ = 'LS Mean' msgroup = 'Treatment Mean Groups';
run;

proc sort data=probs&x ; by descending _lsmean_; run;
proc print data=probs&x label; run;

proc sort data=msgrp; by level; run;

proc sort data=_probs; by level; run;

data msgrp&x; set _probs;

merge _probs msgrp;

by level;

run;

proc sort data=msgrp&x ; by descending _P1; run;

/* OUTPUT PROBS AND TREATMENT GROUPS TO EXCEL */

%if %upcase(&excelout)=YES %then %do;

%space(2);

%put TURFRATE NOTE: EXCEL OUTPUT REQUESTED. NOW ATTEMPTING T0 TRANSFER AND GRAPH
PROBABILITES.;

%space(2);

data label;
array P[&respnum] p1-p&respnum;

do i = 1 to arespnum;
P[i] = 1;

end;

format p1-p&respnum cat.;
drop 1;

run;

filename label dde 'excel|sas&x!r1c4zr1c100';
data label; set label;

file label;

put P1-P&respnum ;

run;

filename outfil dde 'excellsas&x!r2c1:r100c100';
data msgrp&x; set msgrp&x;

if _effect_ = '&&var&x';

file outfil;

effect = '&&var&x';

put effect level msgroup _p1-_P&respnum ;

91

run;
%end;
/' END EXCEL OUTPUT '/

%end;
%end;
/* END OUTPUT PREDICTED PROBABILITIES */

%end; /* if effects = O *l
/* OUTPUT INTERACTION EFFECTS */
%else; %if %length(&effect)‘=0 %then %do;

%space(2);

%put TURFRATE NOTE: EFFECTS OUTPUT REQUESTED FOR &EFFECT;

%put turfrate macro must be executed once prior to this request.;
%space(2);

/* Parse individual factors from interaction term 'I

%let effect=%upcase(&effect);

%let n = 0;

%do g=1 %to %length(&effect);

%let char=%qsubstr(&effect,&g,1);

%if &char=%str(') %then %do;

%let n = %eval(&n+1);

%let ind&n = &g;

%end;

%end;

klet indO = 0;

%let ind%eval(&n+1)= %eval(%length(&effect)+1);
%do g=1 %to %eval(&n+1);

%let f = %eval(&g-1);

%let first=%eval(&&ind&f+1);

%let second=%eval(&&ind&g-&&ind&f-1);

%let var&g=%substr(&effect,&first,&second);
%end;

%let varnum= %eval(&n+1);

/* Assign proper variable name lengths for diffs data set */

%do k=1 %to &varnum;

%1et var&k._ = _&&var&k;

%if *length(&&var&k)=8 %then %let var&k._ = _%substr(&&var&k, 1 ,7);
%if %length(&&var&k)‘=8 %then %let var&k._ = _&&var&k;

%let length&k = %length(&&var&k);

%end;

/* Prepare probs data set */

title3'Predicted Category Probabilities (P = RATING CATEGORY)‘;
data probsa; set _probs;

if _effect_ = 'Qeffect';

array names[&respnum] P1-P&respnum;

array _P[&respnum] _P1-_p&respnum ;

do i = 1 to &respnum;

92

names[i] = _p[i];
end;
drop _P1-_P&respnum i;

%do n=1 kto &respnum;

%if &&cat&n ‘= P&n %then %do;

label P&n = '&&cat&n' ;

%end; /* label category values */
%end; /* label all category values */

drop _effect_ ;
label level = “Level of &effect';
run;

/* Prepare lettering of treatment groups */
title4 “for all &effect effects';

data lsm3; set _lsm;

if _effect_ = '&effect';
rename value = _lsmean_;
run;

data diffsS; set _diffs;
if _effect_ = “&effect';

%let first=1;

%let lengthO=O;

%do h=1 %to avarnum;

%let i=%eval(&h-1);

%let second=%eval(&&length&h);

%let first=%eval(&first+&&length&i+1);
&&var&h = substr(level, &first, &second);
%end;

%let first=1;

%let lengthO=O;

%do h=1 %to avarnum;

%let i=%eval(&h-1);

%let second=%eval(&&length&h);

%let first=%eval(afirst+&&length&i+1);
_&&var&h = substr(_level, &first, asecond);
$end;

drop level _level;

_PT_ = P;

run;

data _diffs3; set diffsa;
keep _effect_

%do c=1 %to avarnum;
&&var&c

%end;

%do c=1 %to avarnum;
_&&var&c

%end;

93

run;

data diffsa;
merge _diffsa diffss;
run;

%pdmix612(diffsa, lsm3, sort=no)
data msgrp; set msgrp;

keep level _lsmean_ msgroup;

run;

proc sort data=probss; by level; run;
proc sort data=msgrp; by level; run;
data probs3; set probs3;

merge probsa msgrp;

by level;
label _lsmean_ = 'LS Mean' msgroup = 'Treatment Mean Groups';
run;

proc sort data=prob53; by level; run;
proc print data=probsS label; run;

data msgrp3; set _probs;
merge _probs msgrp;

by level;

run;

/* Output probs and letter groups to Excel */

%if %upcase(&excelout)=YES %then %do;
data label;

array P[&respnum] p1-p&respnum;

do i = 1 to arespnum;

P[i] = i;

end;

format p1-p&respnum cat.;

drop i;

run;

%let colnum=%eval(&varnum+3);

filename label dde 'excellinteraction&varnum.!r1c&colnum.:r1c100';
data label; set label;

file label;

put P1-P&respnum ;

run;

filename outfil dde 'excellinteractionavarnum.!r2c1:r100c100';

data msgrp3; set msgrp3;

if _effect_ = '&effect';

%let first=1;

%let length0=0;

%do h=1 %to &varnum;

%let i=%eval(&h-1);

*let second=%eval(&&length&h);

%let first=%eval(&first+&&length&i+1);
&&var&h = substr(level, &first, &second);
%end;

94

run;

data lsgrpa; set msgrp3;
file outfil;
effect = '&effect';

put _effect_

%do c=1 %to &varnum;
&&var&c

%put var&c = &&var&c;
%end;

msgroup _p1-_P&respnum ;
run;

%end; /' interaction excel out */
%end; /* two way interaction mean output *I

/* RESET OPTIONS */
options ls=84 ps=64 notes center;

%exit:
*put TURFRATE NOTE: EXITING MACRO;

%mend turfrate;

95

 

 

”ITIE'TTEEWJTTMMEHMETT