Analysis and modeling of cattle distribution in complex agro-ecosystems of south Florida

ANALYSIS AND MODELING OF CATTLE DISTRIBUTION IN COMPLEX
AGRO-ECOSYSTEMS OF SOUTH FLORIDA





















By

VIBHUTI PANDEY


A DISSERTATION PRESENTED TO THE GRADUATE SCHOOL
OF THE UNIVERSITY OF FLORIDA IN PARTIAL FULFILLMENT
OF THE REQUIREMENTS FOR THE DEGREE OF
DOCTOR OF PHILOSOPHY

UNIVERSITY OF FLORIDA

2007































Copyright 2006

by

Vibhuti Pandey


































To my parents and my loving wife









ACKNOWLEDGMENTS

I am greatly indebted to my supervisory committee chair (Dr. Gregory A. Kiker) for his

constant guidance, insight, encouragement, and most of all his enthusiastic and continuous

support and confidence in my research. His thorough and thoughtful coaching was unselfishly

tireless, and his enthusiasm has left me an everlasting impression. He always made himself

available and hence I was able to progress constantly in a sustainable manner. I would like to

acknowledge my thanks and appreciation to Dr. Chris J. Martinez for his help throughout the

development of my model. Without his guidance and support with programming, timely

completion of my research would have been impossible. His enthusiasm and helpful nature made

my research progress swiftly. I express my sincere appreciation to Dr. Kenneth L. Campbell for

his guidance during the first 3 years as my supervisory committee chair. He provided direction

that eventually helped me to identify the specific topic for my research. His advice on addressing

each of my technical problems and concerns has been invaluable as well. I am grateful to Dr.

Sanjay Shukla for his support and help during field work. Also, I would like to express thanks to

Drs. Michael Annable and Mark W. Clark who served on my committee and provided valuable

insights into my research.

I am greatly thankful to my lab-mates for their friendship and encouragement, and to the

staff of the Agricultural and Biological Engineering Department for their technical and moral

support. The department as a whole has been a wonderful working and learning environment.

Last but not the least, I would like to thank my family and friends for their relentless support and

advice throughout this endeavor.












TABLE OF CONTENTS



page

ACKNOWLEDGMENTS .............. ...............4.....


LIST OF TABLES ........._..... ...............8.._.._ ......


LIST OF FIGURES .............. ...............10....


AB S TRAC T ............._. .......... ..............._ 13...


CHAPTER



1 INTRODUCTION ................. ...............15.......... ......


Study Background .............. ......... ..... .. ..........1
Lake Okeechobee and Watershed Description ................. ........ ..... ...... ........ .......... .......1
Water Quality Best Management Practices (BMPs) for Lake Okeechobee Watershed.........1 8
Contribution to Information Required for Modeling and BMP Implementation .................19
Organization of This Dissertation............... ..............2

2 CATTLE BEHAVIOR DYNAMICS AND CURRENT MODELING APPROACHES ......22


Factors Influencing Cattle Distribution .............. ...............22....
Cognitive M echanisms .............. ...............23....
W ater Devel opment ............ ..... .._ ...............23...
Breed S el section ............ ..... .._ ...............24..
Seasonal Di stribution............... ..............2
Shade Structures .............. ...............25....
Social Behavior ..................... ...............27.
Cattle Location and Water Quality ................. ...............28.._._._ ....
Existing Modeling Approaches .............. ...............29....
Regression Model s ................... ...............29..
Habitat Suitability Index Models............... ...............30.
Mechanistic Models............... ...............35.
M etapopulation M odels.................. ... ........... .......3
Spatially Explicit-Individual Based Models .............. ...............38....
Numerical Fish Surrogate Model .............. ...............39....
Multi-Agent Systems ........._.___..... .___ ...............40.....
Cattle Tracking Techniques ........._._.. ....__... ...............41....
Summary ........._.___..... ._ __ ...............42.....

3 ANALYSIS OF GPS COLLAR DATA ............ ......__ ...............50.


Study Site: Buck Island Ranch .............. ...............50....












Summer Pastures .............. ...............5 1....
Winter Pastures ................. ...............52......__. .....

Hy drol ogic Data............... ...............52..
GP S D ata .............. ...............53....
Data Analy sis............... ...............54
Results and Discussion .............. ...............55....
Conclusion ............ ..... ._ ...............61...


4 DEVELOPMENT OF CATTLE MOVEMENT ALGORITHMS FOR ACRU2000 ............73


Habitat Suitability Index (HSI).................................. ........7
Model Design for Cattle Distribution in ACRU2000 ................. ............... ...._ ....73
Suitability Index for Cattle Distribution............... .. .............7
Preference Estimation Using Analytical Hierarchy Process .............. ....................7
Index for Heat Stress and Seasonal Distribution............... ..............7

Integration of HSI Model into ACRU2000 ........._ ..................... .... ........._ ...... 8
The Agricultural Catchments Research Unit (ACRU) Modeling System .....................81
Minimum Habitat Area for HSI Model in ACRU2000 ................. ........................86


5 MODEL RE SULT S ............_ ..... ..._. ...............9 1...


Testing Model Performance at Buck Island Ranch ........._.._.. ......._ ........_.._.......9
Calibration Results............... ...............92
Verification Results .............. ...............94....
Sensitivity Analysis .............. ........ ...............95
Hypothetical Scenario Model Testing .............. ...............96....
Summary ........._..... ...._... ...............100....

6 DISCUSSION AND CONCLUSION ........._._. ...._... ...............110...


GPS Collar Analysis .....__................. ...............110 .....
H SI M odel .............. ... ...............111.............

Management Implications ................. ...............113.............
Future Research Recommendation ....._.................. ...............113 .....
Herbivore Physiological Representation ................. ....._._ ....._._ .............1
Stream Routing Algorithm ..........._..._ ...............114..._.._ ......
Graphical User Interface ..........._...__........ ...............114....
Conclusion ..........._..._ ...............115.....__ ......


APPENDIX



A LIST OF NEW AND MODIFIED OBJECTS ..........._...__......_. ........... .........1


B HSI MODEL PROCESSES UNIFIED MODELING LANGUAGE (UML)
DIAGRAM S ................. ...............119......... ......


C SURVEY FOR DETERMINATION OF WEIGHTING FACTORS .............. .................122












D WEIGHTING FACTORS DETERMINED BY SURVEY ......____ ..... ... ._ ..............124


E RESULTS FROM SURVEY............... ...............126


F SENSITIVITY ANALYSIS RESULTS................ ...............13


LIST OF REFERENCES ............ ...... ..._. ...............137...


BIOGRAPHICAL SKETCH ........._.._.. ...._... ...............151....













LIST OF TABLES


Table page

2-1 Significant predictors of cattle behavior. .............. ...............48....

2-2 Coefficients in the seasonal grazing models. .............. ...............49....

2-3 Coefficients in the seasonal daytime resting models. ............. ...............49.....

3-1 Percent area of wetlands and ditches in summer and winter pastures. ........._.... .............69

3-2 Summary of climatological data during the study period. ................ ..................6

3-3 Summary of GPS collar data in the experimental pastures. ......____ ...... ....__..........70

3-4 Locations that are assumed to have presence of water. ..........._.....__ .............70

3-5 Mean percentage of daily time spent by cattle near water locations. ............. .................71

3-6 Mean percentage of daily time spent by cattle near water trough. ............. ...................71

3-7 Mean percentage of daily time spent by cattle in wetland. ....__ .............. ..... ..........71

3-8 Mean percentage of daily time spent by cattle in ditch. ............. .....................7

3-9 Mean daily distance traveled and mean daily MCP area by cattle .............. ..................72

5-1 Input parameters and their description used in the ACRU2000-HSI model. ..................108

5-2 Values of input parameters used in the ACRU2000-HSI model after calibration. ..........108

5-3 Input parameter values used in sensitivity analysis. .................... ...............0

5-4 Example of adjusted weighting factors used in sensitivity analysis. ............. ..... ........._.109

D-1 Summary of weightings of features as generated by the LDW program based on the
survey ........... ..... .._ ...............124..

D-2 Summary of weightings of three forage species as generated by the LDW program
based on the survey ........... ..... .._ ...............124.

F-1 Weighting factors used in sensitivity analysis. ......____ .... ... .__ .. ......__........3

F-2 Sensitivity of water, shade and forage weighting factors in warm season. .........._.........133

F-3 Sensitivity of the three vegetation species weighting factors in warm season. ...............134










F-4 Sensitivity of water, shade and forage weighting factors in cool season. ........................13 5

F-5 Sensitivity of the three vegetation species weighting factors in cool season. ...............136













LIST OF FIGURES


Figure page

1-1 Drainage Basins of Lake Okeechobee. ................ ...............21...............

1-2 Yearly average total phosphorus concentrations in the open-water (pelagic) region of
Lake Okeechobee............... ...............2

2-1 Average herbage yield of perennial grassesfrom year long access to water on
southern Arizona range. .............. ...............43....

2-2 The relationships between shrub habitat variables and suitability index values for
pronghom winter food quality.. ............ ...............44.....

2-3 The relationships between two variables of forage diversity and suitability index
values for pronghorn winter food quality.. ............ ...............44.....

2-4 The relationship between mean topographic diversity and suitability index values for
pronghom winter food quality. ............. ...............45.....

2-5 Graphical representation of the index ................. ...............45...............

2-6 Two performance suitability indicators expressed as functions of hydrologic
variables. ............. ...............46.....

2-7 A time series of values of a suitability indicator derived from time series of
hydrologic variable values. ............. ...............46.....

2-8 Creating a composite suitability indicator time series from multiple suitability
indicator time series. ............. ...............47.....

2-9 Three approaches to spatial ecology. ............. ...............47.....

2-10 Using GIS in metapopulation models. .............. ...............48....

3-1 Location of Buck Island Ranch and the Experimental Pastures. ..........._.. ......_.........64

3-2 Map displaying wetlands, ditches and water troughs in summer pastures. .......................64

3-3 Map displaying wetlands, ditches and water troughs in winter pastures. ..........................65

3-4 Example of rainfall and groundwater level data in summer pasture 3. ........._...._ .............65

3-5 Typical cattle movement in summer pasture 2 on June 11, 2001 ...........__... ................ 66

3-6 Average % time spent near water locations. .............. ...............67....











3-7 Average % time spent in shade structures. ............. ...............67.....

3-8 Typical MCP area in summer pasture 2 on June 31, 2001 ............_.. .. ...__ ...........68

4-1 Suitability index values of water features ................. ...............87........... ..

4-2 Suitability index values of shade area ................. ...............87........... ..

4-3 Suitability index values of forage consumption. ................ ............ ...................88

4-4 Goals hierarchy view in Logical Decisions for Windows@ software. ............. ................88

4-5 General structure of the ACRU (v 3.00) model ...._ ......_____ ...... .._ ........8

4-6 Configuration of multiple directional overland flows from source land segment to
adjacent land segments. ............. ...............89.....

4-7 Phosphorus cycle of the ACRU2000 model. ................ .....___.....___..........9

4-8 Nitrogen cycle of the ACRU2000 model ............... ...............90....___ ...

5-1 Land segment Discretization of summer pastures 4 and 5 for ACRU2000-HSI. ............102

5-2 Calibration results on SP4 in warm season ................. ...............103.............

5-3 Calibration results on SP4 in cool season. .............. ...............103....

5-4 Verification results on SP5 in warm season. ............. ...............104....

5-5 Verification results on SP5 in cool season ................. ...............104.............

5-6 Hypothetical scenario setup for ACRU2000-HSI model............... ...............105.

5-7 Total phosphorus results using ACRU2000-HSI model ................. .....___.............106

5-8 Total phosphorus results from various scenarios in ACRU2000-HSI model ..................1 06

5-9 Phosphorus budget of complete model domain using simulated results. ........................ 107

5-10 Phosphorus budget of top two model layers using simulated results. ............. ...... ......... 107

5-11 Total phosphorus retained within grazing cattle using simulated results. .......................108

B-1 PCalculateHabitatSuitabilityIndex UML diagram ................. .............................119

B-2 PForageConsumption UML diagram ................. ...............120...............

B-3 PDefecation UML diagram. .........____...... ..... ...............121....

C-1 Cattle preference of features in a pasture during summer. ............. ......................122










C-2 Cattle preference of features in a pasture during winter. ........._.__..... ..._._............122

C-3 Cattle preference of forage species in a pasture............... .................122

C-4 Example illustrating identification of the relative importance of one feature over the
other on the scale provided in the survey ......._..__ ........._._....... .........12

D-1 Range of weighting of features in warm and cool seasons ................. ......................124

D-2 Range of weighting of the three forage species. ....._._._ .... ... .__ ........_.........2

E-1 Simulation result on SP4 in warm season using weighting factors of researcher-1. ........126

E-2 Simulation result on SP4 in cool season using weighting factors of researcher-1. ..........126

E-4 Simulation result on SP4 in cool season using weighting factors of researcher-2. .........127

E-5 Simulation result on SP4 in warm season using weighting factors of ext. agent -1........128

E-6 Simulation result on SP4 in cool season using weighting factors of ext. agent -1. .........128

E-7 Simulation result on SP4 in warm season using weighting factors of ext. agent -2........129

E-8 Simulation result on SP4 in cool season using weighting factors of ext. agent -2. .........129

E-9 Simulation result on SP4 in warm season using weighting factors of rancher -1............130

E-10 Simulation result on SP4 in cool season using weighting factors of rancher -1..............130

E-11 Simulation result on SP4 in warm season using weighting factors of rancher -2............13 1

E-12 Simulation result on SP4 in cool season using weighting factors of rancher -2..............13 1









Abstract of Dissertation Presented to the Graduate School
of the University of Florida in Partial Fulfillment of the
Requirements for the Degree of Doctor of Philosophy

ANALYSIS AND MODELING OF CATTLE DISTRIBUTION IN COMPLEX
AGRO-ECOSYSTEMS OF SOUTH FLORIDA

By

Vibhuti Pandey

December 2006

Chair: Gregory A. Kiker
Cochair: Sanjay Shukla
Major Department: Agricultural and Biological Engineering

It is perceived that cow-calf operations in south Florida can be a substantial source of

phosphorus loading, to Lake Okeechobee. Spatial and temporal information of cattle location

within a pasture can be instrumental in estimating the deposition location of cattle fecal matter.

To address this issue, cattle position data were analyzed and a simplified distribution model was

developed. Cattle position data were acquired through GPS collars and a cattle distribution

model was developed and incorporated into a regionally tested hydrological/water quality model,

ACRU2000.

The GPS data were spatially and temporally analyzed to quantify the amount of time spent

by cattle near shade and water locations. The analysis revealed the prominence of seasonal

utilization of water troughs, ditches, and shade. Shade structures were utilized substantially

during the warm seasons. Wetland utilization was similar across cool and warm periods but was

variably distributed across times within periods. The analysis also revealed that there can be

significant differences in an individual cow' s behavior and utilization of water features.

The GPS analysis was instrumental in the identification of variables to be included in the

cattle distribution model. This distribution model was added as an add-on module within the









Java-based obj ect-oriented framework of the ACRU2000 modeling system. The algorithms are

composed of attractants of cattle (shade, water, and forage) and their weighting factors. The

algorithms were developed using the techniques of Habitat Suitability Index (HSI) and criteria

weighting was developed using the Analytical Hierarchy Process. The HSI model was integrated

with the current hydrology, nutrient, and vegetation modules within ACRU2000.

The HSI model was calibrated and verified on summer pastures of Buck Island Ranch,

Lake Placid, FL. Model verification revealed that its performance was in good agreement with

observed GPS data. Several Best Management Practice scenarios, designed to mimic fencing of

selected pasture areas, revealed that the phosphorus release from senesced biomass may be a

significant store amongst all other pools of phosphorus. The HSI model has enhanced the

capability of ACRU2000 to represent the spatial variability and nutrient effects of cattle

distribution within complex agro-ecosystems of south Florida.









CHAPTER 1
INTTRODUCTION

Study Background

The State of Florida has plentiful, diverse water resources that support a variety of

ecosystems, animals, food crops, industry, tourism, and recreation. However, rapid population

growth over the past 3 5 years is significantly affecting the quality of these systems. It is also

proj ected that Florida' s population will increase to 25.9 million in 2025 (US Census Bureau,

2005). Hence, Florida is facing a unique challenge of managing water quantity and quality with

the pressure of continuing population growth, accompanied with development and extensive

agricultural operations. The Florida Department of Environmental Protection (FDEP) is the

regulatory agency responsible for restoring and protecting the state's water quality. In its 2006

Integrated Water Quality Assessment Report, the FDEP documented increasing nonpoint source

pollution from urban stormwater and agricultural activities as a maj or environmental concern

(FDEP, 2006). Nonpoint source water pollution, sometimes called "diffuse" source pollution,

arises from a broad group of human activities for which pollutants have no obvious point of entry

into receiving watercourses. Because of its diffuse nature, nonpoint source pollution is much

more difficult to identify, quantify, and control than point source pollution. In south Florida,

especially in the Lake Okeechobee watershed, nonpoint source pollution from agricultural

operations is a matter of concern.

Lake Okeechobee and Watershed Description

Lake Okeechobee and its watershed are key components of south Florida's Kissimmee-

Okeechobee-Everglades ecosystem, which extends from the headwaters of the Kissimmee River

in the north, to Florida Bay in the south. Located in south central Florida, Lake Okeechobee

covers 1891 km2 (730 mi2) and functions as the central part of a large interconnected aquatic









ecosystem in south Florida. The lake is the second largest freshwater body located wholly within

the continental United States. Lake Okeechobee is a multipurpose reservoir providing drinking

water for urban areas, irrigation water for agricultural lands, recharge for aquifers, freshwater for

the Everglades, habitat for fish and waterfowl, flood control, navigation, and many recreational

opportunities (SFWMD, 1997). Under natural historic conditions, water flowed from Lake

Okeechobee to the Everglades. After events of heavy rainfall, water exited the lake' s littoral zone

by numerous small tributaries, and by a broad sheet-flow at the southeastern lake edge

(SFWMD, 1999). At that time, the lake bottom was composed primarily of sand that had low

phosphorus content. Conditions in and around Lake Okeechobee changed dramatically during

the last century, due to agricultural development in the watershed to the north of the lake, and

construction of the Central and South Florida (C&SF) Proj ect. Excess nutrient inputs from

agriculture and more efficient delivery of stormwater by the C&SF Proj ect have dramatically

increased in-lake total phosphorus concentrations.

The Okeechobee watershed is divided into six regions: Lower Kissimmee River (LKR) (S-

154, S-65D, and S-65E), Taylor Creek/Nubbin Slough (TCNS) (S-191), Fisheating Creek, Indian

Prairie/Harney Pond, the Lakeshore and the EAA (Figure 1-1). During the 20th century, much of

the land around Lake Okeechobee was rehabilitated to agricultural use. To the north, dairy farms

and beef cattle ranching became the maj or land uses. In the south, sugar cane and vegetable

farming increased rapidly. Associated with the land use changes were large increases in the rate

of nutrient inputs to the lake (SFWMD, 1999). The main sources of high nutrient loads in the

watershed are thought to be runoff from dairy barns and holding areas, direct stream access by

large numbers of dairy and beef cattle, and runoff from improved pastures.









The lake, a designated Class I water body (potable water supply), has been threatened

especially by high phosphorus levels, which have tripled since 1975 (Figure 1-2) causing large

algal blooms (SFWMD, 1999). The watershed has little relief, and the water table is near the soil

surface during the wet season. Before development this area was largely composed of wetlands

(Blatie 1980). During 1926 and 1928, flooding resulted in the loss of life and property, which

then resulted in the construction of a flood control levee (Herbert Hoover Dike) and a rim canal

around the lake to control flooding. Currently, all flows into and out of the lake are managed

through 140 miles of canals; control structures (gates, locks, and pumps); and levees, which were

completed in the late 1950s, as part of the Central and South Florida (C& SF) Flood Control

Proj ect.

The South Florida Water Management District (SFWMD), in conjunction with the United

States Army Corps of Engineers (USACE), regulates these structures and canals (SFWMD,

1997). This modified system has improved flood control and supplied irrigation water; however,

it negatively affected the water quality of Lake Okeechobee by expediting the delivery of

stormwater runoff to Lake Okeechobee.

Soils in the watershed (especially in northern regions) are mainly Spodosols which are,

sandy, low in clay content, low pH, low cation exchange capacity and low phosphorus retention

capacity. These soils (more than 90% sand) are characterized by high infiltration rates and poor

internal drainage due to low permeability of the Bh horizon. The Bh horizon contains large

deposits of Aluminum (Al) and Iron (Fe) along with organic matter and is known as the spodic

layer. When rainfall occurs, the soils can quickly become saturated. During much of the year, the

water table is located between the spodic horizon and the soil surface (Graetz and Nair, 1995).

Because of this restrictive layer, nutrient movement in Spodosols occurs through surface runoff









and also through subsurface flow (Campbell et al., 1995). Apart from landuse changes, soil and

hydrologic characteristics of the watershed have also facilitated in development of algal blooms

and other adverse impacts to water quality both in Lake Okeechobee and in downstream

receiving waters.

Consequently, in 1999, the FDEP initiated development of the Total Maximum Daily Load

(TMDL) of phosphorus for Lake Okeechobee. A TMDL is the maximum amount of a given

pollutant that a water body can absorb and still maintain its designated use. It was adopted by

rule in May 2001. The FDEP proposed a maximum annual load of 140 metric tons of phosphorus

to Lake Okeechobee to achieve an in-lake target phosphorus concentration of 40 ppb. The FDEP

is working in conjunction with other state agencies such as the Florida Department of

Agricultural and Consumer Services (FDACS), Water Management Districts (WMD), Soil and

Water Conservation Districts (SWCD), and U.S. Natural Resources Conservation Service

(NRCS) to support a healthy lake system, restore the designated uses of the lake, and allow the

lake to meet applicable water quality standards. These agencies are implementing a multifaceted

approach to reducing phosphorus loads by improving the management of phosphorus sources

within the Lake Okeechobee watershed through continued implementation of existing

regulations and Best Management Practices (BMPs) (FDEP, 2001).

Water Quality Best Management Practices (BMPs) for Lake Okeechobee Watershed

The state agencies responsible for water quality have recognized that mere implementation

of regulations will not be sufficient to achieve the targeted load reductions for the lake

(SFWMD, 1999). Other management strategies for this region are needed, including

development of non-enforceable guidelines and education of farmers and ranchers to adopt

BMPs to reduce the current pollution levels in surface waters. A step in this direction resulted in

development of the "Water Quality Best Management Practices for Cow-Calf Operations in









Florida" (FCA, 1999) by the Florida Cattlemen's Association (FCA). The BMPs included in the

manual include a variety of structural (e.g., fencing) and managerial (e.g., nutrient management)

BMPs. Although the BMPs developed represent the best efforts of the ranchers and state

agencies, limited information exists on the effectiveness of these BMPs. To address the

information gap, a study is currently underway by researchers at University of Florida, Institute

of Food and Agricultural Sciences (UF-IFAS) in conjunction with FDEP, FDACS, SFWMD, and

NRCS. The study is aimed at demonstrating and determining the efficacy of water quality BMPs

such as fencing and improved water management for reducing phosphorus loads to Lake

Okeechobee from cow-calf operations in the Okeechobee basin (UF-IFAS, 2002). Another

important factor governing the BMP implementation by the ranchers will be their economic

impact on ranch income. Unless a BMP is economically feasible for a rancher, its

implementation will be limited.

Contribution to Information Required for Modeling and BMP Implementation

Given the fact that cattle's presence in waters can be a source of direct loading of P, it is

important to quantify the time spent by them near/in waters so that an informative decision can

be made regarding water quality BMPs for ranches in the region of south Florida. It is also

essential that the current modeling systems should incorporate dynamics of localized grazing

pressure so that comprehensive representation of existing agro-ecosystems is accounted. Almost

all hydrological models lack representation of spatial distribution of cattle presence. A successful

model should not only represent hydrology and nutrients but also the dynamics of cattle

movement and behavior. If modeling systems are flexible and extensible they can be updated as

per the requirement of the system they are representing. One such modeling system is the

ACRU2000 which is available for expansion and incorporation of cattle distribution dynamics.

The specific objectives of this research are:









* Review cattle behavior studies, especially those associated with water quality impacts.

* Analyze cattle movement dynamics and their behavior using GPS Collar data.

* Construct a methodology to identify the spatial and temporal information of grazing cattle.

* Development of cattle distribution algorithms for the ACRU2000 modeling system.

* Test ACRU2000 on a site, which will best represent the condition for which the model is
being designed, i.e. agro-ecosystem of south Florida.

Organization of This Dissertation

Chapter 2 is a thorough review of two broad categories: cattle behavior studies and

existing modeling approaches that are used in defining ecological systems at various scales. In

Chapter 3, detailed analysis of GPS Collar data is presented. Results of various techniques have

also been presented that have been utilized to process GPS data. The following chapter (Chapter

4) includes model design, algorithm development and incorporation of cattle movement in the

ACRU2000 model. In Chapter 5 the cattle movement add-on module is tested and verified.

Finally Chapter 6 summarizes the effort of data analysis and modeling.



























































Lakewater Total Phosphorus
140

120-

-100 o



4 0 -



20


68 70 72 74 76 78 80 82 84 86 BB 90 92 94 96 98
Year


-f St. Lucie
*. anal






C-43 I, -
1- 10
PRIORITY
BASINS

L-18
North Newr
River
L-24
Mliamli Canal

DRAINAGE BASINS
OF LAKE OKEECHOBEE




Figure 1-1. Drainage Basins of Lake Okeechobee (SFWMD).


Figure 1-2. Yearly average total phosphorus concentrations in the open-water (pelagic) region of
Lake Okeechobee (SFWMD).









CHAPTER 2
CATTLE BEHAVIOR DYNAMICS AND CURRENT MODELING APPROACHES

High density animal operation is of interest as it can potentially be a cause of concern with

regards to its impact on the environment (Bottcher et al., 1995). Pastureland and dairies can

become an important source of diffuse or nonpoint source pollution if adequate practices are not

implemented or in cases when livestock are allowed to approach or enter surface waters. In

regions such as south Florida where cattle-ranching and dairy-farming are important agricultural

activities; there are concerns of increase in nutrient loadings from these agricultural lands.

Phosphorus loading from rangelands and its subsequent movement into the drainage waters

(Lake Okeechobee) is a maj or environmental concern in this watershed (Allen et al., 1982). The

primary source of phosphorus has been non-point source agricultural runoff, particularly from

beef cattle ranching and dairy farming, the two primary land uses in the Lake Okeechobee

watershed (Flaig and Reddy, 1995). Unlike dairy farms, beef cattle ranches are not yet treated as

sources of point source pollution due to lower animal stocking rates associated with cow/calf

production systems. Therefore, these ranches are not subject to any regulations from state and

federal agencies. A "voluntary" BMP implementation program exists for beef cattle ranches,

however, due to limited information on the effectiveness of these BMPs not many ranchers have

enrolled in the program.

Factors Influencing Cattle Distribution

Since cattle defecation is of maj or concern, it is evident that to develop a complete

understanding of the animal-plant-soil system in a ranch system, the spatial information of

grazing cattle will be crucial, which will aid in developing a comprehensive understanding of

ecological interactions. Various studies have examined the scope of improving pasture utilization

by increasing the distribution of cattle (Bailey et al., 1989a; Bailey et al., 1996; Ballard and










Krueger, 2005; Ganskoop, 2001; Marlow and Pogacnik, 1986; Owens et al., 1991, Schacht et al.,

1996; Smith et al., 1992, Sneft et al., 1985a, b).

Cognitive Mechanisms

In an invited synthesis paper Bailey et al. (1996) examined behavioral mechanisms that

produce large herbivore distribution patterns. It was reported that grazing distribution can be

attributed to biotic factors such as forage quality and abiotic factors such as slope. Abiotic factors

form cattle's conspicuous habit to graze "convenient areas" (Schacht et al., 1996). Selective

grazing of these convenient areas within pasture isolates area that do not get grazed or only

lightly grazed. This eventually causes reduction in the carrying capacity of grasslands and

efficiency of livestock enterprise (Anderson, 1967). Bailey et al. (1996) defined the foraging

process as an aggregate of two mechanisms: non-cognitive and cognitive. Non-cognitive

mechanisms do not require use of memory from large herbivores during foraging. Grazing

velocity and intake rate are examples of non-cognitive mechanisms that require little judgment

from the animal. Whereas, cognitive mechanism is a process of leaming and memory that have

shown to affect diet selection in selecting feeding sites. In earlier studies Bailey et al. (1989b, c)

demonstrated that large herbivores return to nutrient rich areas more frequently and generally

avoid nutrient poor areas. This is primarily because animals have an accurate spatial memory and

can associate food resource levels with the locations in which they were found.

Water Development

To ensure even pasture utilization, managers try to increase cattle' s uniformity of grazing

by changing these abiotic attributes of their pastures. Slope and distance to water have been

widely acknowledged as the two primary determinants of grazing patterns in large scale range

environments (Owens et al., 1991; Sneft et al., 1985b; Sneft et al., 1987; Schacht et al., 1996).

Areas that are steeper receive less use than those that are gentle (Mueggler, 1965), and locations









that are farther from water also receive less use than those that are near water (Valentine, 1990).

Development of water sources that are further than 1 km from existing water source usually

increases forage utilization and thereby increasing overall uniformity of grazing (Bailey, 2004).

Goebel (1956) increased the number of water developments from 9 to 52 over period of five

years and observed that this increase in water availability decreased concentrations of cattle in

overgrazed areas and increased use of areas which previously received little or no utilization.

During the growing season drainage channels and plant community near water locations gets

heavily grazed (Sneft et al., 1985b). Therefore, to increase distribution and also to lighten forage

over-utilization near water (Figure 2-1) some studies have controlled cattle's access to water

(Martin and Ward, 1970).

Water development has also been useful in protecting riparian areas thereby improving

stream water quality. Off stream water source has proven to decrease grazing pressure in the

riparian zone (Porath et al., 2002). Sheffield et al. (1997) reported that installation of off stream

water source reduced the average concentrations of total suspended solids, total nitrogen,

ammonium, sediment bound nitrogen, sediment bound phosphorous, total phosphorous and

stream bank erosion. In another study in Oregon, Miner et al. (1992) observed that cows reduced

their presence in the stream from 25.6 min/day to only 1.6 min/day (reduction of more than

90%),when off stream tank was made available.

Breed Selection

It has been reported that herbivores prefer gentle slopes near water (Mueggler, 1965).

Bailey et al. (2001) has suggested the use of breeds that originate from mountainous terrain

(Tarentaise) in rugged rangelands and use of breeds developed in gentler slopes (Hereford) in

rolling topographical rangelands. However, there can be some individuality associated with

regards to grazing on rugged terrain (Bailey et al., 2004). In a study conducted in the mountains










of Montana Bailey et al. (2004) compared the daily grazing patterns of cows that used steepest

slope and highest terrain to those that used gentler slopes and lower elevations. The authors

termed cows that spent more time grazing steeper slopes as "hill climbers" and those that used

gentler slope as "bottom dwellers". The study concluded that individual cows within a herd can

use different terrain.

Seasonal Distribution

A seasonal effect on cattle grazing behavior has also been reported in many studies

(Marlow and Pogacnik, 1986; Sneft et al., 1985b; Tanner et al., 1984). Typically during summer

(growing season), forage becomes mature and plentiful and there is more even grazing. Whereas,

during winter (dormant season) forage is not that palatable and hence there is more patchy

grazing. However, grazing distribution of weaning cows generally improves during late fall and

winter because of decreased water and nutrient requirements after weaning (Schacht et al., 1996).

A study conducted in northeastern Colorado used cluster analysis of forage-use to analyze the

consistent seasonal-grazing pattern and eventually construct a predictive model (Sneft et al.,

1985b). It was found that seasonal-grazing distribution was correlated with proximity to water

and site-quality indicators. Results of a 2-year behavior study in Montana also indicated seasonal

trend in cattle use of riparian and upland areas (Marlow and Pogacnik, 1986).

Shade Structures

In regions associated with high temperatures, another important factor in cattle distribution

and performance is availability of shade. At high temperatures, evaporative cooling is the

principal mechanism for heat dissipation in cattle (Blackshaw and Blackshaw, 1994). In order for

a cow to maintain a relatively constant body temperature with respect to its environment

homeostasiss), it must maintain thermal equilibrium via its developed heat-regulating

mechanisms. When the ambient temperature approaches or exceeds cattle's body temperature,









the cattle must increase their active cooling by evaporation of water from the respiratory tract or

from the skin by sweating (Lee, 1967). Failure to maintain homeostasis at high temperatures may

lead to reduced productivity or even death (Blackshaw and Blackshaw, 1994). Historically there

used to be a perception amongst producers that providing shade may reduce the time that cattle

spent grazing. However, recent studies have demonstrated that the amount of time cattle spent in

shade was related to environmental conditions and that shade seeking did not result in reduced

grazing time (Widowski, 2001). Also, by manipulating shade, cattle can be drawn to under-

utilized area of pasture (McIlvain and Shoop, 1971). Shade has also proven to bring financial

profits in ranching enterprise. In a 4 year study in Oklahoma it was quantified that shade

increased summerlong gain of yearling Hereford steers by a profitable 19 lb/head (McIlvain and

Shoop, 1971). The same study also concluded that "hot muggy days" (days when temperatures

were above 850 F and high humidity) reduced summerlong steer gains by 1 lb per day. In a

review paper on the effect of shade on production and cattle behavior Blackshaw and Blackshaw

(1994) reported that under high heat stress, Bos indicus breeds and their crosses have better heat

regulatory capacities than Bos taunts breeds. The authors attributed this difference due to

differences in metabolic rate, food and water consumption, sweating rate, coat characteristics and

color. The sweating of Bos indicus increases exponentially with rises in body temperatures;

whereas, in Bos taurus, sweating rates tended to plateau after an initial increase (Finch et al.,

1982) Therefore, Bos taurus must evaporate substantially more sweat than Bos indicus to

maintain normal body temperatures (Finch, 1986). Blackshaw and Blackshaw (1994) in their

review paper discussed some of the important physiological mechanisms that help cattle to cope

with heat stress:

* Evaporative Cooling
* Metabolic rate and tissue insulation










* Water consumption
* Cattle coat characteristics

Social Behavior

In mountainous terrain, cattle may form social groups (Roath and Krueger, 1982a).

Amongst these social groups cattle have been classified as leaders, followers and independents

with regards to movement during grazing (Sato, 1982). A dominance hierarchy exists in a herd

(Bennett et al., 1985; Bennett and Holmes, 1987; Broom and Leaver, 1978). Animals high in the

hierarchy have priority to feed, shelter, and water. Low-ranked animals maintain a certain

distance from dominant animals to avoid conflict. As subordinates get closer to dominant

animals, they may reduce their bite rate, stop feeding, relocate into areas of lower habitat quality

or wait their turn until the more dominant animals are satisfied and leave the area (Bennett et al.,

1985; Bennett and Holmes, 1987; Broom and Leaver, 1978). Therefore, management strategies

that involve social composition (e.g., herding, selective culling) can be used to relieve grazing

pressure on environmentally sensitive areas (Sowell et al., 1999).

Apart from the various factors mentioned above, there are plentiful other factors that may

be responsible in influencing cattle distribution dynamics. Schacht et al. (1996) have categorized

four techniques that can be employed for improving grazing distribution:

r Enticing the grazing animal to forage
Water placement
Salt and mineral placement
Supplemental feeding location
Rubs and oiler placement
Other methods (mowing, burning, shade etc.)
Pasture characteristics
Fencing
Pasture size
Pasture shape
r Grazing management strategies
Rotational Grazing
Stocking density










Flash grazing
Season of grazing
Livestock considerations
Class of livestock
Vegetation and terrain characteristics

Cattle Location and Water Quality

Considerable research pertaining to water quality impacts of grazing systems have been

well documented in the western states of USA (Belsky et al., 1999; Buckhouse and Gifford,

1976; Miner et al., 1992; Nader et al., 1998). Numerous studies have specifically targeted cattle

distribution patterns relative to water locations and riparian areas (Dickard et al., 1998; Gillen et

al., 1985; McIlvain and Shoop, 1971; Owens et al., 1991; Roath and Krueger, 1982b; Sneft et al.,

1985). Results from all the above studies have indicated water to be an influencing factor in

cattle distribution patterns.

In a cattle ranch system with stream there is concern of direct contamination within the

stream and significant impact on riparian areas. These impacts depend upon cattle behavior and

utilization of riparian vegetation (Marlow and Pogacnik, 1986). Cattle prefer to be closer to

water sources while grazing. This situation can lead to defecation, and eventually over

enrichment of the water bodies. High-density cattle activities near or on the stream banks can

result in rapid transport of manure to the streams (Bottcher et al., 1995). Apart from direct input

of nutrients into the stream, grazing near stream banks can also result in increased erosion of the

stream banks (Helfrich et al., 1998). Bowling and Jones (2003) listed four key potential impacts

grazing cattle can have on water quality:

Increased suspended sediment concentrations, due to the physical stirring up of the
bottom sediments when cattle are in the water, and due to increased sediment run off
from grazed foreshore areas.

Input of organic materials causing effects such as increased biological oxygen demand.










Increased nutrients by both direct deposition into the water or entrained in run off
entering the water body.

Increased fecal bacteria and potential pathogenic microorganisms, again through
defecation straight into the water, or in run off from nearby areas.

Cattle grazing and resting pattern will change with respect to water availability, climate,

presence of shade structures, and forage quantity and quality. Water seems to be the driving

force in attracting cattle towards in and around stream areas. Cattle wade into the shallow water

to graze on aquatic plants, to drink the water, and to wallow in it and remain cool on hot days

(Gary et al., 1983; Hagedorn et al., 1999). However, even when availability of water is not a

limiting factor still, cattle are known to spend significant time in grazing within the riparian area

due to availability of higher quality forage.

Existing Modeling Approaches

Model development is a crucial step in representing such a diverse ecosystem and it will

help define problems, organize our thoughts, develop an understanding of the data and

eventually be able to make predictions. There are various approaches for modeling population

response to environmental pattern. Following are some modeling methodologies that have been

widely utilized in the scientific community.

Regression Models

In some early model development effort pertaining to cattle's spatial distribution, Cook

(1966) used multiple regression equations to explain livestock spatial utilization patterns. The

same methodology was later used to predict spatial patterns of cattle behavior over an entire

landscape (Sneft et al., 1983, 1985a, 1985b). Data used to develop the regression model were

collected on the USDA-ARS Central Plains Experimental Range in northeastern Colorado during

1970-1973 (Sneft et al., 1983). Observations of cattle movement were made by following cattle

on foot for one 24-hour period during each month of the study period on two small paddocks, 1 1









ha and 22 ha. Over 60 independent variables were screened and seven were eventually

incorporated for analysis using stepwise multiple regression (Table 2-1).

In another observational study over a 2-year period (June 1980 through May 1982) at the

same site, researchers derived regression models of spatial patterns of grazing (Table 2-2) (Sneft

et al., 1985b) and resting (Table 2-3) (Sneft et al., 1985a).

It was concluded that even though mathematically the models are boundless (i.e. can be

applied to pastures of any size), it was noted that the models do not consider interactions among

variables. Hence, introduction of a complex herd structure might require more complicated

mathematical descriptions of spatial use. Sneft et al., (1983) also acknowledged that even though

the models are "fine-grained" in space, they are "coarse-grained" in time. This indicates that on

finer time scale, daily temperature variations, for example, may have an effect on the behavior of

cattle. These limitations have been overcome by using a different technique known as habitat

suitability index (HSI) modeling (Cook and Irwin, 1985; Schamberger et al., 1982).

Habitat Suitability Index Models

The U.S. Fish and Wildlife Services developed a methodology known as Habitat

Evaluation Procedures, a planning and evaluation technique that focuses on the habitat

requirements of fish and wildlife species (U.S. Fish and Wildlife Service, 1980). These

procedures were formulated according to standards for the development of Habitat Suitability

Index (HSI) Models (U. S. Fish and Wildlife Service, 1981). The HSI models are usually

presented in three basic formats: (1) graphic; (2) word; and (3) mathematical (Schamberger et al,

1982). The graphic format is a representation of the structure of the model and displays the

sequential aggregation of variables into an HSI. Following this, the model relationships are

discussed and the assumed relationships between variables, components, and HSI's are

documented. This discussion of model relationships provides a working version of the model and









is, in effect, a model described with words. Finally, the model relationships are described in

mathematical language, mimicking as closely and as simply as possible, the preceding word

descriptions.

HSI provides a probability that the habitat is suitable for the species, and hence a

probability that the species will occur where that habitat occurs. If the value of the index (Range

0 to 1) is high in a particular location, the chances of that species occurrence in that location are

high. For example, HSI of 0 would mean totally unsuitable habitat, whereas HSI value of 1

would mean optimum habitat. To determine habitat suitability, suitability indexes (SI) are

assigned to represent the degree in which the variable may contribute to species life requisites

(Hohler, 2004). The SI score is based upon empirical data, professional wisdom and at times,

inspired guesses (U.S. Fish and Wildlife Service, 1981).

Spatial location of herbivores has challenged many researchers who have tried to model

their distribution (Bailey et al., 1996; Coughenour, 1991; Pringle and Landsberg, 2004; Wade et

al., 2004). In an invited synthesis paper, Coughenour (1991) provided important insights into

models that integrate plant growth, ungulate movement, and foraging. A variety of modeling

approaches was discussed and HSI modeling was accredited of overcoming the limitations of

multiple regression models (application constraints). Bailey et al. (1996) developed a conceptual

model to demonstrate how cognitive foraging mechanisms can be integrated with abiotic factors

to predict grazing patterns of large herbivores. Abiotic factor multipliers were used in the

modeling systems which are similar to HSI models.

As an example of a typical HSI model, a step by step illustration of a HSI model

development is given by Allen et al. (1984) in a U.S. Department of Interior document. This

document is one in a series of publications that provides information on the habitat requirements









of selected fish and wildlife species. In this particular document, the HSI model was developed

for pronghorn (Antilocarpa amnericana) chiefly for application for the Great Basin and the Great

Plains region for winter weather. This simplistic model assumed the winter habitat characteristics

to be the most limiting conditions affecting pronghorn distribution. The model is based on the

assumptions that pronghorn survival and reproductive success are functions of winter food

availability. The model incorporates vegetation and topographic features that favor food

availability under mild snow conditions. After detailed review of literature describing the

relationship between habitat variables to the pronghorn's preference; the authors synthesized all

the information and identified six variables of interest:

Percent shrub crown closure (V1)
Average height of the shrub canopy (V2)
Number of species present (V3)
Percent herbaceous canopy cover (V4)
Amount of available habitat in winter wheat (Vs)
Slope of land (V6)

Each of these six variables has their respective suitability index relationships as shown in

the Figures 2-2 and 2-3 and synthesized in equation 2-1.


WFI = [SIVzx(SIV~xSIV~xSIV4 1/3] +SIVS (2-1)


Equation 2-1 accounts only for the forage factor towards the overall HSI where WFI is an

index representing the forage preference of the pronghorn' s diet. The geometric mean of the

three variable indexes (SIV2, SIV3 and SIV4) in the equation 2-1 is a compensatory function.

This function is used in multiplicative models so that partial compensation of the interacting

variables is accounted for (U. S. Fish and Wildlife Service, 1981). The three variable indexes are

assumed to have equal value, meaning that all three must be 1.0 (optimum) in order for this

function to be optimum. Also, a unit increase (e.g., increase an SI by 0.1) in the variable index is









assumed to have the greatest positive impact on the overall index (WFI). This relationship

(Equation 2-1) is combined with the suitability index of the sixth variable, slope of land (Figure

2-4), to calculate the combined food/cover index.


WFFI + SI F
FFFCI = 6 (2-2)



The HSI is equal to the WFCI as calculated in equation 2-2. Allen et al. (1984) extended

the application of the model for evaluation areas that may comprise several cover types. To

represent several cover types it was suggested to multiply the area of each cover type by its

respective WFI value, sum the products, and divide by the total area of cover types to determine

the area weighted WFI (equation 2-3) (Allen et al., 1984).




weighted FFI = (2-3)




where n is the number of cover types, WFli is the winter food index for individual non cropland

cover type, and Ai = area of cover type i.

A similar procedure was suggested to follow to determine the area weighted cover index

(CI) value (equation 2-4). Once both the weighted indexes are computed an overall HSI value is

determined by averaging the WFI and CI values.



C;I, A
weighted CI = '- (2-4)

i7=1









where Cli = cover index value for each cover type. A similar HSI model (equation 2-5) for elk

(Cervus elaphus nelsonii) has been documented by Thomas et al. (1988) for the Blue Mountain

winter ranges of Oregon and Washington. The authors made use of some published as well as

some unpublished data to derive a procedure for evaluating effectiveness of various habitat

variables.


HESRFC (HEs x HER x HEF x HEc)1/" (2-5)


where: HESRFC is the habitat-effectiveness index, allowing for the interaction of HEs, HER, HF,

and HEc, HEs is the habitat-effectiveness index derived from size and spacing of cover and

forage areas, HER is the habitat-effectiveness index derived from the density of the roads open to

vehicular traffic, HEF is the habitat-effectiveness index derived from the quantity and quality of

forage available to elk, HEc is the habitat-effectiveness index derived from cover quality, and

1/N is the Nth TOOt of the product taken to obtain the geometric mean. The mean reflects the

compensatory interaction of the N factors in the habitat-effectiveness model.

Similar to the pronghorn HSI model the geometric mean is also used in this model as a

compensatory function. The authors also incorporated graphical representation of the index

resulting from raising any product derived from (HES x HER x HEF x HEC) to the power of

1/N (1/4 in this case) (Figure 2-5).

In a more recent application, HSI technique has been utilized by the South Florida Water

Management District (SFWMD), for evaluating water management alternatives in the greater

Everglades ecological system, extending south of Lake Okeechobee in South Florida (Tarboton

et al., 2004). In their study, Tarboton et al. (2004) used conceptual ecological models to help

define water-dependent habitat suitability indices for select ecosystem indicator species and









landscape features. The first step in the process of defining habitat suitability functions was to

identify the indicator that would serve as a surrogate for the entire ecosystem. Six different

indicators were identified: three were landscape features and remaining three were fish, alligator

and wading birds. After identifying the indicator features and animals, the next step was to

determine the hydrologic variables, attributes, or characteristics that affect the selected indicator

feature and animals. Examples of hydrologic variables used are water depth, flow direction, and

hydroperiod. Once the specific hydrologic variables were selected for each feature or animal, the

next step was to identify the relationship between those variable (Figure 2-6) values and the

relative conditions of the indicator features or animals.

These functions were based on observed data and expert opinion. Once defined, these HSI

functions were combined with time series of hydrologic values to obtain an overall time series of

ecosystem habitat suitability values (Figure 2-7).

Eventually, based on time series values of multiple suitability functions, composite value

were obtained (Figure 2-8).

To obtain composite performance indicator values geometric means, weighted arithmetic

means, and maximum or minimum values were used. The methods selected for combining

different habitat suitability functions for the same ecosystem feature or species were determined

during the calibration procedure (Tarboton et al., 2004). The authors concluded that with this

approach they were able to link ecology to hydrology in a way that would make it easy for

anyone to understand, modify, test, and evaluate this linkage.

Mechanistic Models

Herbivore and plant dynamics have also been modeled utilizing classical predator-prey

relationship between two species in an ecosystem (Noy-Meir, 1975). In some cases researchers

have utilized an energy balance relationship to account for the balance between energy required









for herbivore body maintenance and the amount gathered by foraging (FAO, 1991; Hobbs and

Swift, 1985). Over time various simple as well as complex models have been developed which

along with other processes also attempt to describe animal responses to environmental inputs.

The SAVANNA ecosystem model (Coughenour, 1993) is a spatially explicit, process-oriented

modeling system developed to simulate ecosystems occupied by ungulate herbivores. The model

is composed of several submodels, which describe various processes and vary in complexity.

The herbivory submodel simulates forage intake by diet selection, forage abundance and forage

quality. An energy balance submodel simulates body weight of the mean animal of each species

based on differences between energy intake and energy spent. Smith (1988) described a detailed

mechanistic model in which they added a behavioral sub-model to simulate the ecology of an

arid zone sheep paddock in pastoral areas of south Australia. The spatial component was

included in the model by dividing the paddock into cells of 500*500 m2 and modeled on an

hourly timestep (movement of sheep while grazing is 500 m/hr). Movement of sheep was

determined by the state of its four physiological criteria: heat stress, thirst, hunger, and darkness.

Each of these criteria was defined in a hierarchy of trigger level conditions which determines the

dominant trigger and consequently determines where and at what speed animal movement will

take place.

Metapopulation Models

Levins (1969, 1970) defined metapopulation as "population of populations"; in which

distinct subpopulations (local populations) occupy spatially separated patches of habitat. In other

words metapopulation is a patchy distribution of population in which species exist in clusters that

are either isolated from one another or have limited exchange of individuals (Akgakaya et al.,

1999). It is a network of semi-isolated populations with some level of regular or intermittent









migration among them. In a review paper Hanski (1998) distinguished between three approaches

to large scale spatial ecology (Figure 2-9).

The approach of theoretical ecology assumes homogenous continuous or discrete (lattice)

space and the model does not incorporate any environmental heterogeneity. On the other hand,

landscape ecologists have developed models that are very descriptive of the complex real

environment. Hanski (1998) termed metapopulation models as a "compromise" where

landscapes are viewed as networks of idealized habitat patches in which species occur as discrete

local populations connected by migration. Metapopulation models are spatially structured so that

they incorporate information about habitat relationships and the characteristics of the landscape

in which the metapopulation exists (Akgakaya, 2001). RAMAS has been a popularly used model

which includes metapopulation dynamics integrated with GIS (Applied Mathematics, 2003)

(Figure 2-10).

Hanski (2004) pointed out that even though the idea of running simulation models using

metapopulation theory may seem tempting as it can be applied to any kind of population, it is

however prone to problems. Firstly, validating a complex simulation model will be virtually

impossible, and secondly, the simulation approach will yield specific results rather than more

general understanding. A good example of such a modeling approach has been exemplified in

Schtickzelle and Baguette (2004) where the researchers modeled the metapopulation dynamics

of the bog fritillary butterfly utilizing the abovementioned RAMAS/GIS. The model was

validated by comparing the predicted and observed distribution using the same empirical data

that were used to estimate model parameters. It is therefore important that modelers be careful in

the construction and parameter estimation of models using metapopulation theory. Hanski (2004)

has therefore, repeatedly emphasized that the classical metapopulation theory are most useful for









examining the dynamics of metapopulations living in highly fragmented landscapes. Such

landscapes are in which the suitable habitat for the focal species accounts for only a small

fraction of the total landscape area, and where the habitat occurs as discrete fragments.

Spatially Explicit-Individual Based Models

Spatially explicit population models are increasingly being used in modeling animal

populations and their movements (Dunning et al., 1995). These models can be simple as well as

complex. The extreme of simplicity in population models are the patch occupancy models that

are based on the number of occupied populations. On the other hand, the extreme of complexity

are the spatially explicit individual/agent based models, which describe spatial and habitat

information at the individual level. Logan (1994) has pointed out that complex systems need

complex solutions. The complexity of the processes involved in ecosystem, has compelled the

modelers to accommodate processes that vary across wide range of spatial and temporal scales

(Levin, 1992). Modelers of aquatic ecosystems have realized the constraint a limited spatial scale

simulation poses towards model accuracy and usefulness towards decision making. Individual-

based-model (IBM) is a relatively new approach in ecology. In an individual-based model, the

characteristics, behavior, growth, reproduction etc. of each individual is tracked through time.

This system is different than the commonly used modeling techniques where the characteristics

of the population were averaged together (Reynolds, 1999). These models provide ecologists

with an effective way to explore the mechanisms through which population and ecosystem

ecology arises from how individuals interact with each other and their environment. IBMs are

also known as entity or agent based models, and as individual/entity/agent-based simulations.

Similar to individual-based, agent-based models has also been utilized for simulating animals

with comprehensive and dynamic landscape structure (Topping et al., 2003). More recently

Ovaskainen and Hanski (2004) derived a stochastic patch occupancy (SPOM) model from an









IBM, where individuals obey the rules of correlated random walk. This unique and novel

modeling framework generates emigration and immigration events in a mechanistic manner and

avoids the need for particular assumptions about how the areas and connectivities of habitat

patches influence migration. It was concluded that in spite of being simplistic the SPOM

replicated the behavior of IBM remarkably well (Ovaskainen and Hanski, 2004).

Numerical Fish Surrogate Model

It is for this reason, Nestler et al. (2001) utilized a particle- tracking algorithm with

stimulus-response rules to develop a Numerical Fish Surrogate (NFS) system (Goodwin et al.,

2001), which creates virtual fish that are capable of making individual movement decisions

based on spatial physiochemical and biological information. The Numerical Fish Surrogate uses

a Eulerian-Lagrangian-agent method (Goodwin et al., 2006) for mechanistically decoding and

forecasting movement patterns of individual fish responding to abiotic stimuli. An ELAM model

is an individual-based model (IBM) coupling:

Eulerian framework to govern the physical, hydrodynamic, and water quality domains

Lagrangian framework to govern the sensory perception and movement traj ectories of
individual fish

Agent framework to govern the behavior decisions of individuals.

The modeling-philosophy behind ELAM is based upon two maj or theoretical approaches

that are coupled to represent the movement of fish (Nestler et al., 2005). Eulerian and Lagrangian

approaches are the two frameworks that have been integrated in ELAM. The former approach is

utilized by engineers to describe the physiochemical properties in hydraulics, while the latter

approach, used by biologists, is mostly centered on the stage development and movement

patterns of particles or individuals. The developers of ELAM hypothesize that by marrying these

two frameworks into a coupled Eulerian-Lagrangian (CEL) hybrid method, they can maintain the










integrity of individuals while concurrently simulate the physiochemical properties of the aquatic

ecosystem that affects fish' s movement. The hydrodynamic and water quality module of the CEL

hybrid model is CE-QUAL-W2 Version 3.0 (Cole and Wells, 2000), a 2-D laterally averaged

model developed at the U.S. Army Research and Development Center. The coupler in CEL

Hybrid model is based upon particle-tracking-algorithm that uses equations for computation of

forcing functions in the longitudinal and vertical directions.

The temporal scale of ELAM is exceptionally low, i.e. 2 sec time-step. Modelers argue that

to produce better fit of the fish' s movement in the vertical direction short time-steps are essential.

Given the fact that ELAM is an individual-based model, i.e., it tracks the behavior movement of

an "individual" fish at each time step, the mathematical computations become massively

demanding. It is for this reason that the model is currently run on U. S. Army Maj or Shared

Resource Center supercomputers. The computational infrastructure of the model (as of June

2004) handles simulations of 5,000 virtual fish in approximately 11 hr of run time (20,000 2-sec

time steps). More recently, Goodwin et al. (2005) have realized the involvement of substantial

run time associated with the mathematical computations of this model and have tried to increase

the computational efficiency by simulating more virtual fish in far less simulation times.

Multi-Agent Systems

Recently, several researchers have started to use multi-agent systems (MAS). MAS is

similar to agent-based modeling, but are more influenced by computer sciences and social

sciences (Bousquet and Page, 2004). MAS give more prominence to the decision-making process

of the agents and to the social organization in which these agents are embedded. Ferber (1999)

has defined a multi-agent system being composed of: environment, objects, agents, and relations

and operations. MAS has been effectively used in variety of cases, for example: modeling of

sheep's spatial memory (Dumont and Hill, 2001), prediction of duck population response to









anthropogenic cases (Mathevet et al., 2003) and predict the effects of alternative water

management scenarios in south Florida on the long-term populations of white-tailed deer and

Florida panther (Abbott et al., 1995).

Cattle Tracking Techniques

To develop comprehensive grazing management strategies to improve water quality in

watersheds consisting of beef cattle ranches, it is imperative to develop an understanding of

cattle's usage of water locations. This often involves observation of cattle movement in a pasture

setting. Earlier studies involved extensive field observations and in most cases observations were

limited to daylight only (Tanner et al., 1984). Research involving visual observations of the

cattle's position and its actions are prone to error as the observer can alter cattle behavior and

make visual errors. In such studies, observation periods are generally short due to its labor

intensity and concerns over observer fatigue. In subtropical regions such as south Florida, night

time observations can be critical because cattle exhibit bimodal grazing patterns (early morning

and evening) and with less adapted breeds of cattle spending a greater portion of the night

grazing as compared to day time (Bowers et al., 1995; Chase et al., 1999; Hammond and Olson,

1994). Global Positioning System (GPS) and Geographical Information System (GIS)

technology allow livestock grazing behavior and management to be evaluated with greater

spatial and temporal resolution (Ganskopp, 2001; Tumner et al., 2000; Ungar et al., 2005).

Animals can be tracked on a 24-hour basis using GPS receivers incorporated into collars.

Agouridis et al. (2004) tested GPS collars under static (open field, under trees and near fence)

and dynamic conditions to evaluate their accuracy for applications pertaining to animal tracking

in grazed watersheds. Their results indicated that the collars were accurate within 4 to 5 m,

deemed acceptable for most cattle operational areas. Collars can also record ambient temperature

and number of vertical and horizontal head movements. Head movements can be used to










determine grazing time and differentiate animal activity (resting or grazing) between location

fixes. Location and other programmed data are stored in the collar, and animals must be caught

and the collar removed to retrieve the data.

With more recent technical advancement, Cattle Traq LLC, an affiliate of American

Biomedical Group Inc. located in Oklahoma City, has developed software capable of monitoring

cattle and recording internal body temperature. Cattle Traq is an integrated system of microchips

located in ear tags, access control sensors and proprietary software (ABGI, 2005). It operates

with radio frequency waves sent from ear tags to software that decodes the signals and translates

them into usable information.

Summary

In their thorough review on grazing impacts on stream water quality in the southern region

of USA, Agouridis et al. (2005) credited the plentiful grazing studies of the western and mid-

western USA; but, also acknowledged that the differences between the arid west and the

southern humid region prohibit the universal transfer of research results. Models and concepts

developed elsewhere cannot be applied to the unique agro-ecosystems of the south-east (Platt

and Peet, 1998) such as south Florida. A limited number of grazing studies in the southern humid

regions (Tanner et al., 1984; Zuo, 2001) have provided valuable, yet incomplete information

with regard to the extent, if any, of water quality degradation by the grazing beef cattle in the

southeastern USA.

In recent times, with the advancement in computational power, researchers have exploited

new advanced computer-based technologies for the development of ecological simulation

systems. Primarily, the research in ecological model development has been greatly concentrated

in utilizing enhanced computer technology to incorporate the details of ecological phenomena.

The primary goal when building an ecological model should be to incorporate the knowledge and










understanding of a system's patterns and processes into a computerized tool that will simulate

the way in which the real system would behave under specific conditions. Simple models often

achieve this goal; they have simplistic assumptions, and can function with limited data.

However, they might neglect detailed aspects such as spatial heterogeneity and individual

variability. Alternately, complex models incorporate the details of ecological phenomenon but,

are often criticized because they are difficult to understand, parameterize, and hard to

communicate. Individual based models are good examples of complex modeling systems. These

models are useful classroom exercises to demonstrate effects at fine scale. Unfortunately, the

behavior rules at individual levels are poorly known and therefore modelers have to rely on

stochastic mechanisms. Another limitation of these models is the exorbitant computation demand

to represent large number of animals over large areas.



120

100






~ 0





100 200 300 400 500
Distance from water (yards)


Figure 2-1. Average Herbage Yield (lb/acre) of perennial grasses (1959-1966) from year long
access to water on southern Arizona range (Martin and Ward, 1970).










1.0 1.



S0.6 0.6

30.4 -1 t0.4



vn0.0 .I 0 6 12 18 24(in)
0 25 50 75 10) A 0 15.2 30.4 45.7 60.9(an)B






= t /Variable 3
30.4




I 2 3 4 5C


Figure 2-2. The relationships between shrub habitat variables and suitability index (SI) values
for pronghorn winter food quality. A) Percent shrub crown cover. B) Average height
of shrub canopy. C) Number of shrub species present per cover type (adapted from:
Allen et al., 1984).



1. II r X .


S0.6 0.6-
r Variable 4 variable 5
i 0.4 -1 C 0.4-

S0.2 0,2 -

vr0.0 vr0.0
U 25 50 75 1 0 A 0 25 50 75 100 B


Figure 2-3. The relationships between two variables of forage diversity and suitability index (SI)
values for pronghorn winter food quality. A). Percent herbaceous canopy cover B)
Percent of available habitat in winter wheat (adapted from: Allen et al., 1984).











* 1 9 1 1


A B C D.~.


1.0




0.6


0.4



0.2-


A) CI-2% slope; flat or

B) 3-8%8 slope; gently rolling
C) 9-25% slope; substantial
drainages, ridges, and/or
rims present
0) > 25% slope; mountainous


Figure 2-4. The relationship between mean topographic diversity and suitability index (SI)
values for pronghorn winter food quality (adapted from: Allen et al., 1984).


z
S0.8
TLI

ur.
S0.6


uJ .

e


.L 0.2


0.2 0.4 0.6 0.8 1.0
Value for (MEyx HER x HEF x HEC)


Figure 2-5. Graphical representation of the index (adapted from: Thomas et al., 1988).


















0 0



Hydrologic Variable Hydrologic Variable

Figure 2-6. Two performance suitability indicators expressed as functions of hydrologic
variables (adapted from: Tarboton et al., 2004).


Time, t


Figure 2-7. A time series of values of a suitability indicator derived from time series of
hydrologic variable values (adapted from: Tarboton et al., 2004).

























Time, t


Time, t


Tlkfe.l


Figure 2-8. Creating a composite suitability indicator time series from multiple suitability
indicator time series (adapted from: Tarboton et al., 2004).


T~heoretical
ecology


Metapopulation
ecology


Landscape
ecology


I
I
Ii '-(


...:.....i......... ..i.. ..L....~.....i....i
I--
I.....l.....i-~..1.....;.... i .


Figure 2-9. Three approaches to spatial ecology (adapted from: Hanski, 1998).











RAMAS/GIS
rmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmy
5 ~Patch
I HSI recognition PatchI


I Ia tutr


I I


SpatiI al
Population
Metapop.





Sensitivity Risk
analysis II analysis

111 1 11


Land-
GIS





Field studies


Demographic
data



Experiments


PVA
Reserve design
Wildlife management


Figure 2-10. Using GIS in metapopulation models (adapted from: Applied Mathematics, 2003).

Table 2-1. Significant predictors of cattle behavior (adapted from: Sneft et al., 1983).
Pasture Characteristic


Distance From


Cactus


Behavior Water Fence Corner Eley. Aspect Slope r
Freq.
Grazing + Travel S* S S S 0.50
Summer Resting S S S S 0.34
Winter Resting S S S 0.25
Bedding S S S S 0.20
Functional form 1 1 1 1 1
in models X X X X Cos(X +k) X X
*S denotes a pasture variable statistically significant in predicting the distribution of a given behavior at the 0.001
level.











Table 2-2. Coefficients in the seasonal grazing models (adapted from: Sneft et al., 1985b).
Independent variable
Proximity Oppo Agsm Eref Sihy Bogr rel.
Season Constant r
to water Freq. Freq. Freq. Freq. abund
Growing (Apr-Oct) 438.0 -.104 .316 .039 --4.30 .460
Dormant (Nov-Mar) 350.0 -.010 -.109 .014 .50 .269
Mathematical X1 x
express in model' X 2 3; 4q 5I X7
SSpecies symbols are defined in text
2 X1 = distance from stock tank (meters)
X2 to Xg = percent frequency to plant species
X6 = biomass of blue grama (Bogr) in community (g/m2)
X7 = biomass of all plant species in community, excluding pricklypear (g/m2)


Table 2-3. Coefficients in the seasonal daytime resting models (adapted from: Sneft et al.,
1985a).
Independent variable
Proximity Proximity to
Season Elevation Aspect Constant r
to water fence corners
Warm (Jun-Aug) 416.45 157.71 11.25 --1.89 .426
Cool (Sep-May) 408.80 106.60 5.34 k4 -1.51 .555
Mathematical 1 1 1
cos(X4 )C
express in model* X, X2 X3
* X, = distance from stock tank (meters)
X2 = distance from nearest fence corner (meters)
X3 = CICVation above 1646 m contour (meters)
k4 = 0.600 (cos(0.5236(month-12))
X4 = degrees deviation from due south









CHAPTER 3
ANALYSIS OF GPS COLLAR DATA

Prolonged hot summers in the region of south Florida can cause physiological heat stress

in cattle and drive them into shade and water-filled ditches and wetlands to cool down. The main

focus of this chapter is to quantify the amount of time spent by grazing cattle near or in water

locations (wetlands, ditches and water troughs) across seasons in a cow-calf production ranch in

south Florida. Partial data were used to conduct analysis on the impact of shade structures, total

distance traveled and total area utilization. This chapter is divided into five sections: description

of the study site, data collection, methodology utilized to quantify the time spent by cattle near/in

features of interest, results, and finally some conclusions from the analysis.

Study Site: Buck Island Ranch

The MacArthur Agro-ecology Research Center, Buck Island Ranch (BIR), Lake Placid,

Florida, USA (270 09'N, 81o 12'W) (Figure 3-1) is representative of the agro-ecosystem that

exists in the Okeechobee watershed. BIR (4, 168-ha) is a full-scale commercial cattle ranch

owned by the John D. and Catherine T. MacArthur Foundation and leased to Archbold

Biological Station and is located in the central portion of the Indian Prairie/Harney Pond Basin,

one of five maj or tributary basins of the Lake Okeechobee watershed. The ecology of this ranch

is composed of a mosaic of habitats that includes open grasslands, forests, and wetlands which

support a diverse and productive community of wildlife and plants (Arthington et al., 2006;

Swain et al., 2006). This ranch is representative of much of south Florida which was once a

native, subtropical, wet-prairie ecosystem. The ranch has been mostly drained and converted to

improved pasture; however, some patchy wetland areas still exist. For more than 10 years BIR

has been a platform of comprehensive interdisciplinary agro-ecological research (MAERC,

2005). The key goals of ongoing research efforts are to quantify the effects of various










management practices on surface water quality and protection of the natural biodiversity of

ranches while maintaining the economic viability of the ranching industry in central Florida.

The BIR is primarily a commercial ranch and therefore ranch operation management is

designed to support animal performance while optimizing the amount of beef production per unit

area of land. Cattle are rotated among the pastures to maximize the available forage for grazing

cattle. Cattle are stocked for longer periods in the improved pastures (typically during summer

season, May through October) and shorter periods in semi-native pastures (typically during

winter season, November through April). Two reasons necessitate this management strategy:

firstly, summer pastures are fertilized (NH4NO3 56 Kg N/ha) (Arthington et al., 2006) in spring

and, therefore, have better forage quantity and quality compared to winter pastures which have

never been fertilized (Swain et al., 2006). Secondly, winter pastures are less intensively drained

and as a result they are regularly flooded during the rainy season in summer. Rotation to winter

pastures provides time for the summer pastures to recuperate from active grazing.

This study was conducted over three years from 2001 to 2003 at BIR. Since the main

obj ective of this chapter is to quantify the time spent by cattle near water locations only the

physical description regarding ditches and wetlands is presented here. Detailed description of the

presence of pasture and wetland vegetation species has already been reported by Swain et al.

(2006); whereas, soil information has been reported by Capece et al. (2006).

Summer Pastures

Summer pastures (Figure 3-2) consist of eight (S1-S8), approximately 20 ha fields (range =

19.0 to 22.1 ha) with bahiagrass (Paspalum notatum) as the dominant forage species. These

pastures are located on soil types that for the area are considered relatively well drained. Pastures

S1 and S8 serve as control fields and were not stocked. The drainage ditch network in these

pastures is comprised of two orders of ditches: deep ditches (0.6 m deep) that run north-south









and receive flow from feeder ditches (0.3 m deep) that run east-west approximately every 30 m.

In the stocked pastures, the average total length of the ditch network is 6175 m (range = 5793.5

to 6864.8 m) and the average area of wetlands is 0.90 ha (range = 0.20 to 1.57 ha). Water troughs

were located at the north end of all stocked pastures (Figure 3-2).

Winter Pastures

Winter pastures (Figure 3-3) also consist of eight fields (W1-W8), that are slightly larger,

averaging 32.2 ha (range = 30.3 to 34. 1 ha). These fields consist of mixture of forage species but

were predominantly bahiagrass, and located on soil types that are considered poorly drained for

the area. All fields, except W4 and W7 which served as controls, were stocked during the period

of this study. Similar to summer pastures, winter pastures have a ditch network; however, W8

consists of an additional order (0.9 m deep) of ditch. In the stocked winter pastures, the average

total length of the ditch network is 4437 m (range = 6618.2 to 253 5.6 m) and the average area of

wetlands is 3.28 ha (range = 1.58 to 5.66 ha). Runoff from summer and winter pastures drains in

a collection ditch and is then conveyed into the Harney Pond Canal which discharges directly

into Lake Okeechobee. A summary of individual pastures, ditches and wetlands is provided in

Table 3-1.

Hydrologic Data

As part of the ongoing water quality study (Capece et al., 2006), on-site climatological data

and groundwater elevation data are collected for both summer and winter fields. All

experimental pastures are bermed so that surface water runoff from each pasture exits through a

single trapezoidal flume. This study utilized climatological information (Table 3-2) in

conjunction with groundwater level data (Figure 3-4) to estimate antecedent soil moisture

conditions and consequently determine the presence and level of water in ditches and wetlands.











The following criteria were utilized to determine the presence of water:

* A water table depth of 0.3 0.6 m was deemed to inundate wetlands, shallow and deep
ditches

* A water table depth of 0.6 0.9 m was deemed to inundate wetlands, and deep ditches

* A water table depth of 0.9 1.2 m was deemed to inundate wetlands

* A water table depth of below 1.2 m was deemed to inundate no features

GPS Data

Cattle position data was monitored continuously using GPS collars (GPS_2200, Lotek

Wireless Inc., Newmarket, Ontario, Canada). These collars are relatively lightweight (950 gm)

and primarily designed for use on smaller animals such as cattle, deer, wolves and bears. The

manufacturer reports that with differential correction deployed, accuracies of position reading

consist of errors that are less than 5 m. For the purpose of this study, data were recorded every 15

min during a 5-day period in spring (March), summer (June), fall (late August), and winter

(November or December) of each year. These periods were selected to be representative of

environmental extremes or expected seasonal differences in forage quality and to fit in the

standard animal handling routine of the ranch. Data collected included: collar identification,

latitude, longitude, temperature and time. A summary of the quantity of GPS Data is provided in

Table 3-3.

Figure 3-5 shows typical GPS collar data on summer pasture 2 collected on June 11, 2001.

The GPS point data have been joined by a line to illustrate the cattle's sequential movement.

Figure 3-5 represents a small portion of the large data set that was collected over the entire study

period (27,924 Total GPS Location Fixes).









Data Analysis

To analyze the data collected from the GPS collars ArcView@ (ESRI, Redlands, CA)

package was utilized. The first step in the analysis was to ascertain that the movement pattern

was not random. To accomplish that, a Nearest Neighbor analysis was performed. Developed by

Clark and Evans (1954) for work in the field of botany, Nearest Neighbor method computes the

ratio (R) of distance between nearest points and distances that would be expected on the basis of

chance. A freely available software that is an add-on extension to ArcView@ Animal Movement

extension (Hooge and Eichenlaub, 1997) was utilized to perform this statistical technique.

For cattle location analysis, all the fix data (latitude, longitude format) was converted to

UTM Cartesian coordinates (NAD 83, Zone 17N) for analysis with other features. The buffer

distance that was utilized for the features were: wetland = 5 m, ditch = 2 m, water trough = 20 m

and shade = 5 m. The extensive network of the ditches is a unique feature in these pastures as

they occupy a considerable area of the pastures. The buffer distance was assumed to be 2m on

each side of the line coverage to represent the narrow nature (2 m width) of this shallow ditch

system. A more flexible (5 m) buffer was utilized for the wetlands to capture the presence of

cattle in the transitional (ecotone) areas of the wetlands which can be wet or dry depending upon

moisture conditions. Cattle do not spend much time in actually drinking water (Wagon, 1963).

Therefore, to capture their presence near the trough the buffer for water trough was set to be at

20 m. Shade structures (5m by Sm) were present at the north end of all stocked summer pastures.

Apart from the shade structures and few patchy trees in SP5, there was complete absence of

natural shade. Winter pastures did not contain any shade structures as most of them had natural

shade from trees. The buffer used for shade structures was 5 m.

The data points that existed within the buffer zone were compared with total data points for

a day (typically 96). Consequently the data were converted into a percentage of time for a given










day. In addition, temporal dynamics with regards to the utilization of water features were

identified by categorizing hours of the day into 4 time zones: (a) Early Morning (12:00 am to

6:00 am) (b) Late Morning (6:00 am to 12:00 pm) (c) Afternoon (12:00pm to 6:00 pm) and (d)

Night (6:00 pm to 12:00 pm). Animal Movement extension was further utilized to compute total

distance traveled by collared cattle and a minimum convex polygon (MCP) home range.

Statistical analysis for comparison of mean percentage of time was performed using JMP

Statistical Software (SAS Institute, Inc., 2005). Tukey-Kramer' s Honest Significant Difference

(Tukey's HSD) test was performed. An alpha level of 0.05 was accepted as a nominal level of

significance and results were considered statistically significant when a P < 0.05 was obtained.

Results and Discussion

The test of nearest neighbor analysis for complete spatial randomness was performed for

all data. Values close to R = 1.0 indicate that the observed average distance is the same as the

mean random distance, suggesting that the spread of data is random. However, R values < 1.0

imply that the observed distance is smaller than the mean random distance, suggesting that data

is clustered. The average R value during summer was 0.51 (range = 0.80 to 0.13) and the average

winter R value was 0.47 (range = 0.74 to 0. 11), suggesting that the data was non-random and the

GPS fixes displayed more clustering during winter than summer months.

Climatological information (Table 3-2) and groundwater level data (Figure 3-4) were

utilized to make hydrologic judgment regarding the presence and level of water in ditches and

wetlands. This information was especially useful when making judgment regarding presence of

water in shallow or deep ditches. Table 3-4 summarizes the estimated presence and location of

water during different time periods of the study.









Since temperatures in summer and fall are often similar (Table 3-2) these two seasons were

grouped into one category of warm period. Accordingly, spring and winter was combined into a

cool period category. Average percentage of daily time spent by cattle near/in all possible water

locations (water trough, wetland and ditch) was relatively low (<15% of 24-hr period) compared

to the remainder of the pasture area, but was higher (P<0.01) during the warm than the cool

period (11.45 & 0.39%[(mean a s.e.; n = 215] vs. 6.09 & 0.69% [mean a s.e.; n = 160],

respectively). Statistical analysis was performed to compare means of percent utilization of

individual water features in different Seasons and also all water features within the same Season.

For example, wetland use was statistically the same and lower in all seasons except warm 2003

as indicated by the lowercase "b". In warm 2001 the use of all the three different water features

was not statistically different as indicated by the uppercase "A".

Wetland and ditches had similar, higher cattle presence compared to troughs (4.4110.35

and 5.2910.38 % vs. 1.9710. 18%, respectively) across periods (Table 3-5 and Figure 3-6). This

was not unexpected because wetlands and ditches buffer areas (approx. 20% of average pasture

area) was much larger than the buffer area of water troughs which were essentially a single point.

Utilization of the water sources differed within periods. This difference was mainly due to a

lower than average utilization of water sources in warm 2001. Unlike what was found in the

other years, utilization in warm 2001 was similar to what was found in cool 2001-02. Across

periods and years, cattle utilization of the different water features remained fairly consistent with

the exception of the ditch feature which showed higher use in warm periods and lower use in

cool periods for all years except 2001. This may have been related to the drier conditions that

occurred during the summer sampling period that year (Table 3-4). During the summer sampling

period in 2001, all ditch classes and wetland areas were dry. This observation supports the









hypothesis that cattle utilize water in ditch features for cooling in addition to possibly for

drinking or feed sources.

Water trough use was consistently higher in the warm periods than cool periods (Table 3-

5). Since troughs could only be used for drinking water, this observation supports what has been

observed in other studies (Goodwin and Miner, 1996; Kelly et al., 1955; Miner et al., 1992;

Sheffield et al., 1997) that cattle will preferentially use alternate and clean sources of water for

drinking. In contrast, the use of water-filled wetlands was fairly consistent regardless of periods

and did not differ, with the exception of an almost doubling of average utilization of wetlands in

warm 2003 (8.25% + 2. 11). The high wetland utilization during warm 2003 can be explained by

a single cow' s strong affinity towards wetland. During the warm 2003 period, data were

collected only from five collared cows during the summer season (no data were collected in fall,

Table 3-3). Amongst the five cattle, one displayed very high affinity towards wetland and

ditches. Average percent of time spent by this specific cow in the wetlands was 24.95%, which is

substantially higher than any other collared cow in any period. This individual cow entered the

wetland every day (all 5 observed days) during 8am to 9am in the morning and remained in the

wetland until 5pm to 6pm. Even if environmental factors are similar, differences in individual

cattle behavior have been previously reported as well (Bailey et al., 2004). If the data from this

individual cow are excluded, the average time spent in wetlands for the period of warm 2003

becomes 4.08%, which is similar to percent utilization in other periods.

There were two periods (spring 2002 and winter 2002) when cattle were stocked in

summer pastures instead of winter (Table 3-3). This occurred because of prescribed burning of

the winter pastures during spring 2002 and accidental burning during winter 2002. The rotation

of cattle due to fire events did allow the determination of whether cattle proximity to water









location was influenced by differences in summer vs. winter pastures (size, average depth of

water, forage differences, etc.) or driven by temperature. Spring 2002 and winter 2002

experienced 5.46% and 5.09% utilization of all water features respectively. This result was

consistent with cool season utilization of water features by cattle and demonstrated that water

usage in pastures was independent of pasture composition and forage quality.

The amount of time cattle spent near each water feature during a 24-h period was

investigated to identify any temporal dynamics associated with the use of these features. Water

troughs were generally not utilized during early mornings and night time regardless of periods

(Table 3-6). As expected, water trough usage was highest during afternoon times of all periods

with the exception of warm 2001. In warm 2001 cattle utilized the trough more during late

mornings than the afternoon. Warm 2001 had the highest maximum daily temperatures of the

whole trial (37.50C, table 2), and it has been acknowledged that increased water consumption is a

major response to thermal stress (Johnson and Yeck., 1964; McDowell, 1972). Drinking water

may have a direct comforting effect by cooling the reticulum as well as by reducing the thermal

load (Beede and Collier, 1986). Hence, it is possible that in periods of hot conditions such as

Warm 2001 the cattle utilized the trough earlier to mitigate their thermal stress. Data from late

morning as well as afternoon of remaining periods reveals that there was always higher presence

of cattle at the water troughs during warm periods as compared to cool periods. This observation

is in agreement with a previous study in which it was observed that in hot climates most water is

consumed by cattle during two 4-h periods: 7 a.m. to 11 a.m. and 4 p.m. to 8 p.m., which were

also the times when cattle grazed (Ittner et al., 1951). In early studies various researchers had

established that water intake of cattle is a function of forage consumption and ambient

temperature (Leitch and Thompson, 1944; Ritzman and Benedict, 1924; Winchester and Morris,










1956). Hence, during warm periods cattle utilized water troughs more during their two grazing

bouts.

Unlike cattle' s utilization of water troughs, the presence of cattle in wetlands appeared to

be similar across cool and warm periods but was variably distributed across times within periods

(Table 3-7). Wetland utilization was consistently lowest (0.015+ 0.04, P<0.05) in the early

morning hours and highest (1.5910. 18, P<0.05) in the afternoon hours regardless of period. Late

morning and night presence in wetlands was similar and intermediate to the other two times of

day, although there is a suggestion that period of year influenced the time of the day the cattle

started utilizing wetlands. Cattle presence was not recorded during late mornings in the two of

the warm periods; whereas, the data showed consistent utilization of wetlands during the same

time in the cold periods. The extraordinary use of wetlands during warm 2003 has been

explained in the previous section by the exorbitant use of wetland by one cow. As late morning

is a time when grazing activity normally occurs in Florida (Bowers et al., 1995; Chase et al.,

1999; Hammond and Olson, 1994), this data suggests that cattle were using wetlands for grazing

during the cool period but not during the warm period. Additionally, presence of cattle in

wetlands during warm period afternoon hours, when grazing does not normally occur (Bowers et

al., 1995; Chase et al., 1999; Hammond and Olson, 1994), also suggests that wetlands were used

for cooling and not grazing during the summer period. Since wetlands are expected to be the

deepest water containing feature in the landscape, it is reasonable to expect they would be used

for cooling. In contrast, since it is unlikely that cattle would not need to cool themselves during

the cool period, presence during the afternoon period in the cool season probably represents a

continuation of the morning grazing bout into the afternoon period due to lower forage

availability due to slower forage growth.









Unlike wetland presence, cattle' s presence in the ditches during all times of the day

exhibited a fairly consistent pattern of being higher during the warm periods and lower in the

cool periods (Table 3-8). The exception to this pattern was early and late mornings of the 2001

warm period, when cattle presence was similar during warm and cool periods. Cattle can utilize

the ditches for water as well as for higher quality of forage along the periphery of the ditches.

Generally lower presence of cattle in ditches during the cool period may reflect differences in

growth patterns of the forage species found in the ditches. Bahiagrass and bermudagrass were

the dominate forage species in the ditch areas and as warm season grasses, their growth rate

would be lower in the cool periods of the year. Lower growth rate and hence less forage

availability of these grasses in the cool season would explain both, lower cattle presence in the

ditches and higher cattle presence in the wetland areas, which contained more native forage

species. Unlike wetlands, though, there was no consistent pattern for time of day within warm or

cool periods. This suggests that cattle presence may not have been related to forage availability

or the need to regulate body temperature, and may simply reflect an artifact of pasture design

that necessitated a lot of ditches.

Partial data were used to analyze the utilization of shade structures in summer pastures.

The results are presented in a box plot format in Figure 3-7. Error bars represent standard

deviations. Summer 2001 was the driest season, wetlands and ditches are assumed to have no

water presence and hence highest use of shade during this season is expected. However, results

indicate that cattle did not use shade in summer 2002 and nominally in fall 2002. It is noteworthy

that this analysis was conducted using only partial data. Only two collared cows result was used

for shade analysis for summer 2002. The error bars in Figure 3-7 illustrate the high variability in

the use of shade. As mentioned before, it is possible that these two cows are not representative










of the herd behavior. Relatively low use during Fall 2002 could be attributed to high rainfall

during the five monitored days in this season.

Using the animal movement extension, two home range analyses were performed on the

entire data set. The first one was total distance traveled and the second one was Minimum

Convex Polygon (MCP). Total distance traveled is the sum of the length of polylines generated

by joining GPS location fixes. MCP is the smallest (convex) polygon which contains all points

which the cattle has visited. It should be kept in mind that the MCP will also contain a lot of

empty space that the animal never visited. Figure 3-8 shows a typical MCP area in SP2 during

the summer season of 2001. Both these analyses can be used in conjunction to get an

understanding of the area covered and effort made by grazing cattle. Seasonal means of these

two analyses is presented in Table 3-9. A seasonal pattern is evident in distance traveled by

cattle. Cattle traveled more during cool seasons and less during warm seasons. In terms of MCP

area covered by grazing cattle, both cool seasons were higher than warm seasons; however, Cool

2002-03 was the only season that was statistically higher than remaining seasons. Since forage

growth is slower in the cool season, it is likely that cattle have to travel greater distances to look

for palatable forage to meet their intake requirements and in doing so, they browse a greater

pasture area as well.

Conclusion

Beef cattle can utilize the water sources in south Florida to graze, to drink water, and to

keep cool. During these activities, urination and defecation can occur which can result in direct

contamination of watered locations. If BMPs are needed to minimize the impact of beef cattle

production on water bodies in south Florida, a better understanding of beef cattle utilization of

natural (wetland) and artificial (ditches and water trough) water sources is necessary. To quantify

the amount of time spent by grazing cattle near or in water locations GPS collars were used. The









GPS collars were successful in identifying, quantifying and eventually deriving pertinent

information regarding cattle utilization of water sources. Climatological information was used in

conjunction with observed groundwater level data to make hydrologic judgment regarding the

presence and level of water in ditches and wetlands.

The data illustrated that there was higher presence of cattle near water locations during

warm periods than in cool periods (11.45 f 0.39% vs. 6.09 f 0.69%). On a daily basis, cattle

utilization of all water sources (as determined by % time present) was relatively low (<15% in a

24-hr period). Cattle seemed to utilize water troughs in a fairly consistent manner, going to water

troughs earlier (late morning) and staying in the area longer during warm periods, compared to

cool periods when they went later (afternoon) in the day and for shorter periods of time. The

presence of cattle in the wetlands was generally well distributed across all periods as well as all

times (approx. 4% in a 24-hr period). Unlike water trough utilization, cattle utilized wetlands

considerably in the cool periods as well. This suggests that wetlands in Florida are used for

different purposes at different times of the year. During the cool periods, cattle were present in

wetlands when grazing would be expected to occur (late morning), indicating the need for feed

was the driving factor. In contrast, during the warm periods, cattle were present when grazing

was not an expected occurrence (afternoon), suggesting that cooling was the reason the cattle

were in the wetlands. Unlike wetlands, presence of cattle in ditches was generally higher in the

warm periods than the cool periods; though there was no consistent pattern for time of day within

warm or cool periods. This suggests that cattle presence in ditch areas may not have been

related to forage availability or the need to regulate body temperature, and simply reflect an

artifact of pasture design.









Another important factor this study identified was that there can be substantial variability

in individual cow behavior. This was recognized by an exceptionally high presence of cattle in

wetlands during the 2003 warm period, which was due to one individual's affinity towards

wetland. It is perceived that during this period this cow utilized wetland not only to drink water

but to cool itself by staying in water for extended hours. It is suggested that future studies deploy

multiple GPS collars on cattle to account for variability in the population distributions. Shade,

total distance traveled and MCP area also indicate seasonal utilization and browsing patterns in

grazmng.

The result findings may be useful from a ranch management perspective. Knowledge

regarding cattle' s preference of water location will be useful in developing a comprehensive

understanding of the pasture utilization. Information from this study is not comprehensive

enough to design appropriate management strategies to achieve targeted P load reductions.

Nevertheless, this study does provide useful information regarding cattle utilization of water

features in sub-tropical-humid pastoral environments of south Florida. From BMP

implementation perspective, information from this study can be utilized in conjunction with

other studies to suggest pertinent structural or managerial BMPs for this region. However, the

installation or use of one structural or management BMP will rarely be sufficient to solve the P

loading problem. Combinations of BMPs (BMP System) that control the same pollutant are

generally more effective than individual BMPs (Gilliam et al., 1997). In order for the BMPs to

be successful in the unique settings of subtropical agro-ecosystems of south Florida, they should

be strategically tailored to be site specific, effective, and cost efficient.




























Figure 3-1. Location of Buck Island Ranch and the experimental pastures.


SUMMER PASTURES


Figure 3-2. Map displaying wetlands, ditches and water troughs in summer pastures.



































Figure 3-3. Map displaying wetlands, ditches and water troughs in winter pastures.


GROUNDWATER ELEVATION
SUMMER PASTURE 3
MRain (cm) Ground Elevation -Ground Water Level








15 12






Figr 3-4 Exml ofrifl n rudae ee aain ume ate3


WINTER PASTURES


0 260 620 1.040 Melrs
I 1 1 1 t I r rI


Dit













--O


-I I / END





















O Early Morning (12am 6am)
O Late Morning (6am 12am)
O Afternoon (12am -6pm)

ANight (6pm 12pm)
Water Trough
Wetland
-Ditch
Fence

o 0 50 100 2 10 Meters
SI I I I I I I I I





Figure 3-5. Typical cattle movement in summer pasture 2 on June 11, 2001



















on Time Spent








Seasons

Figure 3-6. Average % time spent near water locations.


SHADE UTILIZATION


30


5 -






Figure 3-7.


SUMMER 2001 FALL 2001 SPRING 2002 SLIMMER 2002 FALL 2002 WINTER 2002 SLIMMEP 2003
Periods

Average % time spent in shade structures.










-0O


N


GPS Location on 31 June 2001
V~hter Trough
SMCP Area

SWetland
SFence

S0 50 100 2( 0 Meters
SI I I l I I I


Figure 3-8. Typical MCP area in summer pasture 2 on June 31, 2001










Table 3-1. Percent area (ha) of wetlands and ditches in summer and winter pastures.


Ditch
Length
(m)
4511.65
5878.47
6382.36
5598.73
6864.82
6202.32
5893.76
3463.35


2535.59
2843.65
4118.76
5243.47
4848.63
5656.41
6217.83
6618.18


Wetland
Area
(ha)
4.52
1.57
1.20
1.33
0.20
0.48
0.67
2.95


5.66
2.46
3.80
0.90
3.35
1.58
1.86
2.85


% Area
of
Wetland
20.51
8.26
5.88
6.49
0.95
2.46
3.49
14.53


17.03
7.86
11.30
2.64
10.37
4.93
6.15
9.42


% Area of
Buffered
Wetland **

11.47
8.47
8.83
1.53
3.49
5.15




21.28
10.38
14.48

13.59
6.92

13.41


% Area of
Buffered
Ditch **

12.37
12.50
10.93
13.11
12.73
12.27




3.05
3.63
4.90

6.00
7.05

8.75


Summer
Pastures

Sl*
S2
S3
S4
S5
S6
S7
S8*
Winter
Pastures
W1
W2
W3
W4*
W5
W6
W7*
W8


Area
(ha)
22.04
19.01
20.42
20.49
20.95
19.49
19.22
20.3


33.23
31.3
33.64
34.12
32.31
32.08
30.24
30.27


Amimal
Units

0
20
35
15
35
15
20
0


15
20
35
0
35
15
0
20


* Control Pastures (Not Stocked)
** Assumes a 5-m buffer around wetlands and a 2-m buffer around ditches


Table 3-2. Summary of climatological data during the study period.


Rainfall
Max During
Temp Study Peniod
(cm)
37.50 1.80
37.00 0.66
33.50 0.20
33.50 0.03
36.50 5.46
35.50 8.35


Av
Start Date End Date
Temp


Min
Temp

15.50
17.00
9.00
3.00
12.50
19.00
3.30
16.99
22.03


Season


Summer 2001
Fall 2001
Winter 2001
Spring_2002
Summer 2002
Fall 2002
Winter 2002
Spring_2003
Summer 2003


06/11/2001
08/27/2001
12/03/2001
03/04/2002
06/10/2002
08/26/2002
11/25/2002
03/03/2003
06/09/2003


06/15/2001
08/31/2001
12/07/2001
03/08/2002
06/14/2002
08/30/2002
11/29/2002
03/07/2003
06/13/2003


25.36
26.26
20.25
15.49
25.08
24.56
15.78
23.25
26.58


26.51
31.44
32.85


0.25










Table 3-3. Summary of GPS collar data in the experimental pastures. The number before the
parenthesis is the number of collars used within a pasture and number within
parenthesis is the average daily fixes during 5 day collection period in each season.
Summer Fall Spring Summer Fall Winter Summer
Summer Pastures 2001 2001 2002 2002 2002 2002 2003


Sl*
S2
S3
S4
S5
S6
S7
S8*

Winter Pastures
W1
W2
W3
W4*
W5
W6
W7*


4 (92)
4 (91)
4 (91)
2 (94)
3 (91)
3 (92)


Winter
2001
3 (95)
2 (96)
2 (96)

3 (95)
2 (73)


2 (87)
1 (95) 3 (85)
4 (91)
2 (94) 3 (95)
1 (95) 1 (96)
3 (84)


1 (95) 1 (96)
2 (95) 1 (96)
1 (93) 1 (89)
1 (81)
1 (96)
1 (96) 2 (84)


2 (96)
1 (96)
1 (94)

1 (94)


1 (96)
1 (95)
1 (96)


Spring
2002




2 (95)


1 (96) 1 (95)


Table 3-4. Locations that are assumed to have presence of water (water trough always contained
water) .


Water
Presence

W,DD
W,DD
W
W
W,D
W,DD
W,DD
W,D


Season

Summer
Fall
Winter
Spring
Summer


Start Date End Date
06/11/2001 6/15/2001


08/27/2001
12/03/2001
03/04/2002
06/10/2002


8/31/2001
12/7/2001
3/8/2002
6/14/2002


08/26/2002 8/30/2002


Winter
Spring
Summer
(W = Wetland, D


11/25/2002
03/03/2003
06/09/2003


11/29/2002
3/7/2003
6/13/2003


Both Shallow and Deep ditch, DD = Deep Ditches only)













SEASON* WATER TROUGH WETLAND DITCH j Season Mean
Warm 2001 2.7910.34; n = 130 a A 4.1210.98; n = 35 b A 3.8310.76; n= 35 b A 7.9010.68 c
Cool 2001-02 0.6410.15; n = 130 b C 3.4910.32; n = 130 b A 2.4210.34; n= 65 bB 5.5110.42 c
Warm 2002 3.6610.70; n = 60 aB 3.1310.42; n = 60 b B 9.9311.04; n = 35 aA 13.1611.16 b
Cool 2002-03 0.2410.09; n = 30 b B 4.0410.69; n= 30 b A 4.3310.55; n= 30 b A 8.6110.90 c
Warm 2003 2.6710.86; n = 25 a, b B 8.2512.11; n = 25 aA 9.4411.09; n = 25 a A 20.5912.19 a
Feature Mean 1.9710.18 B 4.4110.35 A 5.2910.38 A


Table 3-5. Mean percentage of daily time spent by cattle near water locations (mean & std error;
n = days).


* Warm = Summer + Fall; Cool = Winter + Spring
a b c: means within columns sharing a common letter are not significantly different (P>0.05)
A B C: means within rows sharing a common letter are not significantly different (P>0.05)


Table 3-6. Mean percentage of daily time spent by cattle near water trough (mean std Error).
Water Trough
Season Early Momning Late Momning Aftemnoon Night
(12am 6am) (6am 12pm) (12pm 6pm) (6pm 12pm)
Warm 2001 0.0010.00 b C 1.8110.25 aA 0.9510.15 b, c B 0.0110.01 aC
Cool 2001-02 0.1010.03 aA 0. 1110.04 b A 0.3210.11 cA 0. 11 0.04 aA
Warm 2002 0.0010.00 a, b B 1.4810.34 aA 2.1710.44 a A 0.00 0.00 a B
Cool 2002-03 0.0010.00 a, b A 0.1010.05 b A 0. 1310.08 c A 0.00 0.00 aA

Time Mean 0.0310.01 B 0.9710.11 A 0.9310.11 A 0.0410.01 B
a b c: means within columns sharing a common letter are not significantly different (P>0.05)
A B C: means within rows sharing a common letter are not significantly different (P>0.05)


Table 3-7. Mean percentage of daily time spent by cattle in wetland (mean a std Error).
Wetland
Season Early Momning Late Momning Aftemnoon Night
(12am 6am) (6am 12pm) (12pm 6pm) (6pm 12pm)
Warm 2001 0.0310.03 a B 0.0010.00 b B 1.6510.67 b A 0.63 f 0.17 a A, B
Cool 2001-02 0.1510.06 a B 1.0810.16 a A 1.0710. 14 b A 0.73 f 0.11 a A
Warm 2002 0.2410.13 a B 0.0010.00 b B 1.3610.26 b A 0.3210.09 a B
Cool 2002-03 0.0010.00 a B 1.3010.33 a A 1.4710.38 b A 0.8710.41 a A, B

Time Mean 0.1510.04 C 0.8110.11 B 1.5910. 18 A 0.6110.07 B
a b c: means within columns sharing a common letter are not significantly different (P>0.05)
A B C: means within rows sharing a common letter are not significantly different (P>0.05)










Table 3-8. Mean percentage of daily time spent by cattle in ditch (mean a std Error).
Ditch


Early Morning
Season
(12am 6am)


0. 1710.09 b C
0.2510.07 b B
2.2310.41 aA
0.5610.18 bB

1.0110.13 B


Late Morning Afternoon Night
(6am 12pm) (12pm 6pm) (6pm 12pm)
0.3310.17 c B, C 1.7610.58 a, b A 1.5510.30 a, b, c A, B
0.6710.13 c A, B 0.9310.14 bA 0.5610.18 c A, B
2.1810.35 a,bA 2.8610.38 aA 2.6510.44 aA
1.2810.27 b, c A, B 1.6110.31 a, b A 0.8710.20 b, c A, B

1.2110.12 A, B 1.6910.15 A 1.3710.13 A, B


Warm 2001
Cool 2001-02
Warm 2002
Cool 2002-03

Time Mean


a b c: means within columns sharing a common letter are not significantly different (P>0.05)
A B C: means within rows sharing a common letter are not significantly different (P>0.05)


Table 3-9. Mean daily distance traveled and mean daily MCP area by cattle (mean a std Error).
Distance Traveled MCP Area
Season
(meters) (acres)
Warm 2001 3179.17157.64 b 13.6910.22 b


Cool 2001-02 3994.68192.33 a


14.1510.49 b

13.8510.35 b

17.671.04a

13.3612.58 b


Warm 2002


3193.04182.16 b


Cool 2002-03 4331.051237.77 a


Warm 2003


2980.49198.58 b


a b c: means within columns sharing a common letter are not significantly different (P>0.05)









CHAPTER 4
DEVELOPMENT OF CATTLE MOVEMENT ALGORITHMS FOR ACRU2000

Habitat Suitability Index (HSI)

Modelers develop and use HSI models for land-use management plans because they are

simple to use and the outputs are generally in form of GIS-based maps, which are easy to

understand. These models are also preferred because they may be applied in an efficient manner

and are relatively inexpensive to operate (Schamberger and O'Neil, 1986). The first step in

developing HSI is to identify habitat variables. The second step is to develop suitability index

functions for each individual habitat variable. The final step is to combine these functions into an

equation for the HSI. In HSI modeling, animals get distributed in proportion to the habitat

suitability. More detail about HSI modeling methodology and some applications have been

discussed in Chapter 2.

Model Design for Cattle Distribution in ACRU2000

Limited information is available regarding cattle's preference in grazing systems of south

eastern USA. It has been elucidated in Chapter 2 that plentiful grazing studies have been

conducted in the western and mid-western USA; however, differences between the arid west and

the southern humid region prohibit the universal transfer of research results. The land is very flat

and the climate is warm and humid in the south Florida for most of the year. This is in contrast to

western regions where land is hilly and the temperatures are dry and extreme. The use of water

features is existent (Chapter 3); however, their utilization may be for different reasons.

Controlled as well as uncontrolled ranges in south Florida consist of abundant wetlands.

Therefore, accessibility to water is not a limiting condition, which may be the case in the west.

Hence, models developed elsewhere cannot be applied to the unique agro-ecosystems of the









south-east. Therefore, functions for individual habitat variable in a HSI model must be defined to

represent the distribution of cattle in ecosystems of south Florida.

Suitability Index for Cattle Distribution

The first step in the process of defining habitat suitability functions is to identify the

variables that would affect the distribution of cattle in a paddock system. Shade and water

features are the obvious attractants that dictate the distribution of cattle; hence they have been

included as variables for HSI computation. Water features such as wetlands and ponds may be

attractive for different reasons in different seasons. Depending on the presence or absence of

water, cattle may display a difference in the utilization of a wetland or pond. A dry wetland may

not be an attractive feature for hot or thirsty cattle; however, it may be luring for hungry cattle

that may graze in it for better quality of forage. Hence a dynamic suitability index is required for

features that may become devoid of standing water during dry periods and thereby changing their

attractiveness for grazing cows. The current hydrologic module in ACRU2000 simulates the

depth of water table for each land segment. The methodology used in Chapter 3, which

determines the presence of standing water in wetlands, is used here as well. Water table depth of

less than 0.6 m from the ground surface is deemed to have standing water in the wetlands and

ponds. Hence, the HSI for water is optimum (1.0) when the water table depth is less than 0.6 m

(2 ft). On the other hand, water table depth of more than 1.21 m (4 ft) from the ground surface is

deemed not to inundate wetlands and ponds. Therefore, HSI for water is minimum (0.0) when

the water table depth is more than 1.21 m. Thus, a dry day (i.e. when water table falls below 1.21

m) will have a different suitability index for a land segment with wetland than a wet day; thereby

making it very dynamic. The suitability index values for a water feature with respect water table

depth in between 0.6 m and 1.21 m is a linear relationship and is are illustrated in Figure 4-1 and

as:









HSI, (WT, ) = for WTDEP, > 1.21 (4-la)

HSI, (WT, ) = 1.67 -1.63 x WTDEP, for 0.60 < WTDEP, <1.21 (4-1b)

HSI, (WT, ) =1 for WTDEP, < 0.60 (4-1c)

where WTDEPr is the water table depth on day t; and HSIt(WTt) is the HSI of water feature on

day t.

Under extended warm humid conditions of southeastern USA when the ambient

temperature approaches or exceeds cattle's body temperature, the cattle will seek shade to cool

themselves. In a study conducted during summer in Louisiana, McDaniel and Roark (1956)

found that shade, either artificial or natural, increased the gains of cows and their calves. The

area of the shade will depend on the size of the herd. In an experimental study Clarke (1993)

tested the effects of shade on behavior, rectal temperature, and live weight gain. It was found that

2.5 m2 Shade/cow reduced rectal temperatures in both, zebu-cross steers and in Hereford steers.

Buffington et al. (1983) recommended at least 4.2 m2 Shade/cow but also agreed with Bond et al.

(1958) that 5.6 m2 Shade/cow was desirable. Alternately, Hahn (1985) only suggested 1.8-2.5 m2

shade/cow was required. For southeastern climatic conditions, a shaded area of 50 m2 was

considered representative of the agro-ecosystems of south Florida. Some isolated trees may also

attract a few cattle; however, a shade area of less than 30 m2 will be less attractive, and may

cause crowding (Buffington et al., 1983). Hence, the suitability index values linearly increase

from 30 m2 to 50 m2 (Figure 4-2) and as:

HSI, (SH) = 0 for SA < 0.0 (4-2a)

HSI, (SH) = SA x 0.02 for 0.0 < SA < 50.0 (4-2b)

HSI, (SH) =1 for SA > 50.0 (4-2c)









where SA is the shade area (m2); and HSIt(SH) is the HSI of shade on day t. Shade area is an

input from the user, and will remain constant throughout the simulation.

Herbivores eat to satisfy their need and desire for nutrients, the most prominent being

energy and protein (NRC, 1996; 2001). The mechanism via which herbivores satisfy their energy

and protein requirement is through consumption of available forage. The current version of

ACRU2000 is set up to simulate three functional forage groups: bahiagrass, floralta, and

panicum. These vegetation species represent functional groups that correspond to vegetation

found in uplands, transition zones and wetlands, respectively (Yang, 2006). The forage is also an

important factor that dictates herbivores movement and distribution. The HSI of forage

consumption (Figure 4-3) is based upon data published by Rayburn (1986), who summarized a

group of experiments and developed a more general relationship of relative intake (a proportion

of maximum or potential intake) to herbage mass. The forage suitability index is represented in

the model as:

HSI, (F ) = 0 for Wa,2,, < 150 (4-3 a)

HSI, (F~, ,)= Wa,~,, x 0.00066 +0. 1 for 150 < Wa,2,, < 1350 (4-3b)

HSI, (F,~,) = 1 for Wa,~,, > 13 50.0 (4-3 c)

where Wa,i~t is the aboveground biomass of species i on day t (Kg/ha); and HSIt(Ft) is the HSI of

forage on day t.

Preference Estimation Using Analytical Hierarchy Process

After development of a suite of suitability indices that are deemed influential in

herbivore's spatial location preference, the next step is to determine the relative importance of

parameters with one another. In case of limited literature availability a good strategy is to utilize

the technique of decision analysis to quantify the preferences of one variable over the other.










Analytical Hierarchy Process (AHP) is a mathematical tool within the field of multi-

criteria decision analysis that allows consideration of both qualitative and quantitative aspects of

decisions. AHP is especially suitable for complex decisions which involve the comparison of

decision elements which are difficult to quantify (Saaty, 1980). It is based on the assumption that

when faced with a complex decision the natural human reaction is to cluster the decision

elements according to their common characteristics. The AHP methodology involves building a

hierarchy (Ranking) of decision elements and then making comparisons between each possible

pair in each cluster (as a matrix). These pair-wise comparisons provide a weighting for each

element within a cluster (or level of the hierarchy) and also a consistency ratio (useful for

checking the consistency of the user-defined weights). This process requires the user to make

direct comparisons of the relative importance of the alternatives on the measure. In case of

herbivore distribution model there are three alternatives (Shade, Water, and Forage) that need to

be compared and evaluated. Within forage there are three types of forages: bahiagrass, panicum,

and floralta. Also, since the distribution is dominated by two seasons, two sets of preferences

need to be developed for all the alternatives. Logical Decisions@ for Windows (LDW) is a

decision analysis tool that helps define alternatives and variables (Logical Decisions, 2005).

Within LDW, AHP technique is available. To use this technique the user needs to pair-wise

compare two variables as part of the assessment process. The user enters the weight ratios for

each possible pair of variables in a matrix. This ratio describes the ratio of importance of a

variable as compared to the other. Within LDW there is also an option of printing the preference

assessment in a questionnaire format. This lets the user obtain a hard copy of the preference

assessment questions) being posed by LDW. The questionnaire asks the user to identify the

importance of one feature with respect to the other (e.g., Forage vs. Shade) on a scale of 1-9.









This is a useful feature of LDW where the questionnaire can be distributed to the participants in

the study who may not be readily available for direct questioning. For this research we utilized

this feature and distributed the questionnaire to many researchers, ranch managers, and extension

agents (Appendix B). This allowed us to acquire and incorporate a broad spectrum of expertise

and opinion into the herbivore distribution model. Once this information is entered into LDW

(Figure 4-4), it uses the AHP computation process to compute set of weights for the variables

(Appendix C). Model simulation results using weighting factors from the survey are shown in

Appendix D.

The model for herbivore distribution is given in equation 4-4:

HSI = [ (Shade WF Made IIIHSI) + (WaterWF WaterHSI) + {Forage WF ForageHSI) ] (4-4)

where ShadeWF and ShadeHSI are the weighting factor and habitat suitability index for shade,

WaterWF and WaterHSI are the weighting factor and habitat suitability index for water, and

ForageWF and ForageHSI are the weighting factor and habitat suitability index for forage. Since

ACRU2000 is capable of simulating multiple vegetation species, the forage factor in equation 4-

1 can be further expanded to include desired number of vegetation species (Equation 4-5).

(Veg, WF Veg, HSI) +(Veg,WF" eg,HSI)+.... .(Veg,,WF eg,,HSI) (4-5)

where VegWF and VegHSI are the weighting factor and habitat suitability index for the specific

forage species.

Index for Heat Stress and Seasonal Distribution

Summer heat stress has long been recognized as a factor that reduces both, the productivity

and reproductive efficiency of cattle in the Southeastern regions of the USA (Jordan, 2003). In

grazing systems cattle are exposed to varying amounts of solar radiation. This radiant energy can

come directly from the sun or indirectly from the immediate surroundings. During summer, this









radiant load may exceed metabolic heat production of cattle significantly. To cope with the hot

environment, cattle will strategize their behavior and physiology to relieve the total heat burden.

Cattle will seek shade, increase water intake and orient themselves away from direct sunlight.

These behavioral changes increase the tissue conductance to facilitate heat transfer from the

body core to the skin and eventually away from the skin by convection and radiation. There will

be increased sweating to increase evaporative loss as well as increased respiratory volume

(heavy breathing) (Blackshaw and Blackshaw, 1994). Cattle will reduce feed intake as an

immediate response to heat stress (Blackshaw and Blackshaw, 1994) and expenditure of energy

to maintain homeothermy (NRC, 1981). Heat stress is caused by environmental factors such as

air temperature, radiation, humidity, and wind velocity (Gwazdauskas, 1985). Over the years

researchers have created indexes that relate specific environmental characteristics to the

physiological variables of heart rate, respiratory rate and volume, sweating rate, and body

temperature (Blackshaw and Blackshaw, 1994). The two environmental parameters that have

been popularly used have been dry-bulb temperature and humidity. In a research conducted

during the summers of 1975-78 at the University of Florida' s Dairy Research Unit near Hague,

FL, Buffington et al. (1981) established that dry-bulb temperature and dewpoint temperature was

directly related to rectal temperature and respiration rate and inversely related to milk

production. They established that Black Globe-Humidity Index (BGHI) is a comfort index that is

based on the combined effects of dry-bulb temperature, humidity, net radiation, and air

movement, as can be seen in equation 4-6:

BGHI = tbg + 0.36.t, + 41.5 (4-6)

where: tbg = black globe temperature (oC) and tdp = dew point temperature (oC) calculated from

wet bulb temperature. When the BGHI is 75 or higher, milk yield and feed intake are seriously










depressed (Buffington et al., 1981). The black globe temperature, (tbg), iS measured by the black

globe thermometer, which usually consists of a 150 mm (6 inch) black globe with a thermometer

located at the centre. Since black globe temperature measurements are not available for BIR,

another index can be used which is based on relative humidity. Thom (1958) suggested a

temperature and relative humidity index (THI) to evaluate a cow's heat stress as:

THI = td (0.55(1 RH /100)(td 58) } (4-7)

where td = dry bulb temperature (oF) and RH = relative humidity (%).The heat stress is defined

as occurring whenever the THI exceeds 72 (Armstrong, 1994, De Dios and Hahn, 1993; Hahn,

1982; Hahn and Mader, 1997; Igono et al., 1991). However, most of the research has been

conducted in midwestern states on dairy cows with reduction in milk production due to heat

stress as the main concern. The cattle in south Florida are mostly beef cattle. More specifically,

the cattle on BIR are Brahman-crossbred (Arthington et al., 2006). Numerous studies have

documented that Brahman cattle have better heat regulatory capacities than other breeds

(Blackshaw and Blackshaw, 1994). This physiological advantage has been attributed to higher

respiratory rate (Finch et al., 1982; Kibler and Brody, 1952), lower metabolic rate (Kibler and

Brody, 1954; Vercoe, 1970; Worstell and Brody, 1953), less water consumption at higher

temperatures (Winchester and Morris, 1956), and thinner-brighter hides (Allen et al., 1970;

Finch, 1985; Finch et al., 1984; Hutchinson and Brown, 1969; Yeates; 1954). Keeping the better

heat resistance of Brahman cows in mind, the critical level of THI was set to 75. On a given day

if the index reaches 75, the weighting factor of shade and water increases by 20% and 10%,

respectively. Consequently, the forage weighting factor decreases by 30%. This change in the

weighting factors of shade, water, and forage account for the change in behavior the cattle will

exhibit during days when they experience thermal stress.









Analysis of cattle movement data in Chapter 2 clearly indicates that cattle display a

difference in their behavior in the two seasons of Florida. To represent the difference in behavior

in the two seasons in Florida, two entire sets of weighting factors have been developed with the

abovementioned variables, one for warm season and the other for cool. The warm season is

defined as March to October and the remaining four months are defined as the cool season.

Integration of HSI Model into ACRU2000

It has been demonstrated that HSI is a relatively swift way to utilize information either

from literature or even from opinions of experts to develop a model. Therefore, a similar

approach has been used to distribute the cattle in a modeling system, ACRU2000.

The Agricultural Catchments Research Unit (ACRU) Modeling System

The ACRU agrohydrological modeling system was originally developed in the Department

of Agricultural Engineering (now the School of Bioresources Engineering and Environmental

Hydrology) at the University of Natal by Schulze (1995). The developers of ACRU model

describe it as a multi-purpose and multi-level integrated physical conceptual model that can

simulate streamflow, total evaporation, and land cover/management and abstraction impacts on

water resources at a daily time step (Figure 4-5). The ACRU program code was developed in the

FORTRAN 77 programming language. As the model got developed and modified by researchers

around the world, the existing programming language (FORTRAN) posed problems with regards

to its compatibility to accommodate these newer versions.

It is for this reason the model was restructured entirely with an obj ect oriented

programming language: Java and was named ACRU2000 (Kiker et al., 2006). The advent of

ACRU2000 made the model more compliant with spatial hydrological aspects and addition of

newer modules became unproblematic (Campbell et al., 2001; Kiker and Clark, 2001; Kiker et

al., 2006; Martinez, 2006; Yang, 2006). ACRU2000 can operate either as a lumped small









catchment model with relatively homogeneous soil and land cover attributes, or as a distributed

cell-type model where complex catchments are separated into sub-catchments or land segments

(Figure 4-6).

The nutrient module within ACRU2000 modeling system was incorporated by Campbell et

al. (2001) (ACRU-NP), which borrowed the concepts used in GLEAMS (Knisel and Davis,

1999; Leonard et al., 1987). The nutrient module added capability in ACRU2000 to 1) simulate

nitrogen (N) and phosphorus (P) losses in surface runoff, sediment transport, and leaching, 2)

simulate N and P cycling in the soil-water-plant-animal system, and 3) simulate N and P mass

balances in the watershed system. However, ACRU-NP module was incapable of simulating

multidirectional lateral nutrient transport between multiple land segments through either surface

or subsurface water movement. The lateral nutrient transport component in ACRU-NP was

mainly designed for transporting nutrients dissolved in runoff and adsorbed in sediments through

single outflow from one land segment. Consequently, Yang (2006) restructured ACRU-NP and

added new components to enable multi-directional spatial transport of N and P through surface

runoff and lateral groundwater flow. In addition, a new conservative solute transport component

was also added to rectify the process of nutrient extraction and adsorption in which the ratio for

partitioning the nutrients between the water and soil phases was a function of clay content. Yang

(2006) compared the performance of the new conservative solute transport algorithm from

PMPATH (Chiang and Kinzelbach, 2005), which is an advective transport model using

groundwater pore velocities from MODFLOW (McDonald and Harbaugh, 1998) using a

hypothetical scenario. The comparison revealed that both models produced qualitatively similar

results. In addition, the ability of the modified nutrient model to predict non-point source nutrient

pollution, at BIR was evaluated. It was concluded that the model performed reasonably well.









Yang (2006) also added significant framework to the ACRU2000 modeling system by

adding a new vegetation component to enable multi-directional spatial simulation of

hydrological, chemical, and biological processes simultaneously in a daily time step. The

vegetation model is a simple model that avoids overwhelming data requirements, but is still

capable of capturing the vegetation dynamics. The model is based on the land segment system

developed by Yang (2006), where each land segment is initialized with one or multiple species,

which compete for light, water and nutrients. For each time step, plant growth is driven by

climate variables including solar radiation and temperature. The basic processes in this model are

light interception, conversion of light into dry matter production and allocation of dry matter

between aboveground and belowground dry matter. The impacts from the changes in hydrology

and nutrient concentrations are expressed in growth limiting factors. Yang (2006) accounts for

two types of pressure on the vegetation: lack or excess of water and nutrient. These stress factors

are combined to define a growth reduction factor that is used in the model to reflect the adverse

growth conditions causing the reduction of the potential dry matter production.

In the model the potential growth rate (AWot~i~t [kg/m2/day]) for each species i on day t is

calculated through a linear function of the absorbed light and a mean radiation-use efficiency

parameter as shown in Equation (4-8).

AWpo,l,, = RUE, xlabs,l,t xF,,(Tt) (4-8)

where RUEi is the average radiation use efficiency of species i [kg/MJ(PAR)] and is a summary

variable for all processes dealing with photosynthesis and respiration. Iabs,i,t and Fi~t(Tt) are the

light interception and temperature factor for assimilation by species i on day t, respectively. The

plant growth rate may be limited by N or P deficiency, water shortage or water logging during

different parts of the growing season:










AWred,i,t = AWpot,i,t x RFi,t (4-9)

ANi,t = AWpot,i,t x RFi,t x CNi,t (4-10)

Al,t =Apot,i,t x RFi~t x CPi,t (-1

where AWred,i~t is the reduced dry matter production rate of species i [kg/m2/day]; ANi~t and APi~t

are the N and P uptake rates corresponding to AWred,i~t [kg/ha], respectively; CNi~t and CPi~t are

the biomass N and P percent, respectively; RFi~t is a growth reduction factor of species i, which

integrates the limiting factors from water, N and P. RFi~t, is a unit less, species-specific growth

reduction factor with a value ranging from 0 to 1, which is obtained by taking the minimum

value of water stress, water logging, and N and P stress factors as shown in the following

equation:

RFi,t = min(Fw,,i,t, FWL,i,t ,FN,i,t, FP,i,t) (4-12)

where Fws,i,t is the water stress factor for species i; FwL,i,t is the water logging factor for species i;

FN,i,t is the N stress factor; and FP,i,t is the P stress factor. Plant senescence is assumed to start

when the daily sum of leaf' areas (LAlsum = C LAi,t, [m2 leaf/m2 grOund]) of all species on one


land segment exceeds the critical leaf area (LAler [m2 leaf/m2 grOund]), an input to the model.

The daily total senesced biomass (Ws,its [kg/ha]) and the corresponding N (Ns,i~t [kg/ha]) and P

and (Ps,i,t [kg/ha]) removed through the senesced biomass for each species on that land segment

are calculated as:

LAI, LAI,
Ws,i,t = Fracbs su c ) /SLAi (4-13)
LAICT


Ns,i,t = Ni,t x Ws,i,t / Wred,i,t (4-14)

Ps,i,t = Pi,t x Ws,i,t / Wred,i,t (4-15)









where Fracbs is the fraction of biomass above the critical level to senesce per day, which is

assumed to be a constant value in the model for all species. The senesced biomass and biomass

N and P decrease the amount of live biomass and its corresponding N and P pools:

Wred,i,t = red,i,t Ws,i,t= (4-16)


Ni~t = Ni~t Ns,i~t (4-17)

Pi,t = Pi,t Ps,i,t (4-18)

Currently the model simulates three perennial species including bahiagrass (Paspahtna

notatunt Fhigge~), floralta (Hentarthria altissinta), and panicum (panicunt rigiduhlnt) which are

believed to be the dominating forage species in south Florida. One of the outputs from this model

is aboveground biomass on a daily basis. This output is critical in the development of the cattle

distribution model.

Recent ACRU2000 model developers (Martinez, 2006; Yang, 2006) have enabled and

tested the model to be capable of simulating both hydrology and nutrient dynamics in field-scale

catchments. Pandey (2006) applied the distributed ACRU2000 modeling system to predict

hydrology and non-point source nutrient pollution, on a commercial beef cattle ranch (Pelaez

Ranch) in the Lake Okeechobee region. Pandey (2006) applied the model on the entire ranch by

finely discretizing the modeling domain into various sizes of 134 land segments. Thus,

ACRU2000 can be confidently used as a basis for coupling with an animal distribution

simulation model to form a more complete ecohydrological modeling system.

Once the HSI' s are computed for every land segment in the model's domain, they are

summed, and normalized (so that they sum to 1.0). The cattle in the population are then

distributed across their range in proportion to the distribution of the normalized HSI' s among

land segments. The redistribution occurs on a daily basis. After redistribution, the cattle consume









existing forage proportional to their population presence on land segments. It is assumed that

each cow will eat 16 Kg of forage per day. This amount is based on the typical cattle weight (640

Kg) (ASAE Standards, 2000) and forage consumption (2.5% of Body Weight) (NRC, 1996). The

total forage amount that gets consumed by grazing cattle is removed from the vegetation model

on a daily basis. The removal is based on the preference weighting assigned to individual forage

species. For example, if Vegl has a higher weighting factor than Veg2 and Veg3, Vegl will get

consumed more. This consumption will be in proportion to the weighting factors. According to

suitability index of forage consumption (Figure 4-3), the grazing herbivores will not "see" any

biomass that is less than 150 kg/ha. Once the forage biomass reaches that low level in a specific

land segment, the forage suitability becomes zero and thereby lowering the HSI for that land

segment. Lower HSI in turn allows less herbivores to be assigned and this allows forage to

recover in that specific land segment.

Cattle also defecate proportional to their population presence on land segments. It is

assumed that each cow will defecate 8.5 Kg/day (ASAE Standards, 2000). The cattle waste gets

applied to the top (litter) layer of the model. The waste is characterized into various nutrients

pools (Organic -P, Labile-P, Organic-N, Ammonium-N, Active-N, Organic Matter) (Figures 4-7,

4-8) based on the rates set by the nutrient model of ACRU2000.

Minimum Habitat Area for HSI Model in ACRU2000

Application of habitat suitability criteria requires that some specific spatial parameters be

defined. Minimum habitat area is defined as the minimum area of contiguous habitat that can

support a cattle population on a long term basis. In case of modeling cattle distribution within

ACRU2000-HSI, it is imperative the users maintain a minimum land segment of 0. 1 ha in order

for cattle distribution module to be able to perform reasonably. Maintaining the right spatial












scale is important in order for the suitability indexes to be realistic over temporal and spatial


variation.


Suitability Index of Water Table




08-


~06-


m04-


02-



152 121 06 03
Water Table Depth (m)



Figure 4-1. Suitability index values of water features.



I Suitability Index of Shade Area


08


006


~04


02


S0 50
I Shaded Area (m2)


Figure 4-2. Suitability index values of shade area.













Suitability Index of Forage


08-


i~06-





02-



0 150 1350 1650
Standing Aboveground Biomass (Kg/ha)



Figure 4-3. Suitability index values of forage consumption.


Figure 4-4. Goals hierarchy view in Logical Decisions for Windows@ software (Logical
Decisions, 2005).

















LI i


Figure 4-5. General structure of the ACRU (v 3.00) model (Schulze, 1995).


Neighbo~r 1


Neighbor


Neighbor 2


Figure 4-6. Configuration of multiple directional overland flows from source land segment to
adj acent land segments (adapted from: Yang, 2006).

































Groundwater


Denitrification Volatilization


Runoff

Percolation


Figure 4-7. Phosphorus cycle of the ACRU2000 model (adapted from: Knisel et al., 1993).


Runol. and
Percolation


Runo~fand
Percolatiocn


Figure 4-8. Nitrogen cycle of the ACRU2000 model (adapted from: Knisel et al., 1993).









CHAPTER 5
MODEL RESULTS

Testing Model Performance at Buck Island Ranch

A cattle distribution model (ACRU2000-HSI) was developed for the region of south

Florida in Chapter 4. The algorithms were developed using the procedure of Habitat Suitability

Index and criteria weightings were developed by processing expert opinion using the technique

of Analytical Hierarchy Process. The GPS data analysis in Chapter 3 was helpful in providing

insights into cattle's behavior in warm humid regions. However, the GPS data were not utilized

to create algorithms for the HSI model. The algorithms are composed of "attractants" of cattle

(shade, water, and forage) and their weighting factors. The attractants were determined based on

the features that exist in the landscape of this region. Weighting factors were determined using

surveys from experts and were then calibrated to obtain better results. Depending on the presence

or absence of water, cattle may display a difference in the utilization of a wetland or pond.

Therefore, the depth of water table, which is an output from the hydrologic model, was utilized

to determine presence or absence of water in wetlands. South Florida' s long hot and humid

summer can cause heat stress in grazing cattle. Cattle's change in grazing and resting pattern as a

result of heat stress in hot-humid environments of the southeast is also incorporated into the

model. To incorporate the difference in behavior in the two seasons of Florida, two sets of

weighting factors were developed.

After model development, the next crucial step is to verify the performance of the model.

The model should be verified and tested at a site that is representative of the region for which the

model is developed. It is also preferable to test the model using observed data. Since the GPS

data from BIR (described in Chapter 3) were available, they were utilized to verify the

performance of the HSI model. Table 3-3 is a summary of the sampling of the GPS data. During










any season the number of cows collared for the GPS study was not consistent. Since the GPS

collar data were available only from a few selected cows from the whole herd, for comparative

purposes, it is assumed that the collared cows are representative of the total cattle population in

the model domain. As there can be significant variability in an individual cow's behavior

(discussed in the conclusion section of Chapter 3), it was essential to select a pasture that had the

largest availability of GPS data. Consequently, SP4 and SP5 were selected for HSI model

application (Figure 5-1). The rationale for selecting these two pastures during the

abovementioned times is because of plentiful GPS data availability (Table 3-3) that qualifies the

data to be representative of the whole herd. For each collared cow, number of recorded "hits" in

a land segment were divided by the total number of hits and then multiplied by herd size to

convert the hits into number of cows.

The description of input parameters are given in Table 5-1 and their values that were used

for model calibration and verification are given in Table 5-2. The values in Table 5-2 are the

result of calibration of the HSI model input parameters that were obtained from the LDW

software using AHP technique. Since the water, shade, and forage parameters change during the

two seasons, it was essential to calibrate and verify the model during both warm and cool

seasons. Summer 2001 (June 11-15) and spring 2002 (March 4-8) were the two seasons selected

for testing of the HSI model. SP4 was selected for calibration and SP5 for verification.

Calibration Results

Figure 5-2 is the result of calibration on summer pasture 4 during warm season of 2001.

The box plot (Figure 5-2) shows the range of the observed GPS data in all land segments. Within

the box the dark dotted line and the light solid line represents the mean and median number of

cattle in each land segment, respectively. The light dotted line (running across all land segments)

illustrates the assumption of equal distribution of cattle (1.25 cattle per land segment) by the









ACRU2000 version. Land segments consisting of water trough, wetland, and shade are

abbreviated WT, WET, and S, respectively. The calibration results of the ACRU2000-HSI during

the warm season captures the overall dynamics of individual land segments. Under prediction of

the number of cattle in land segment 8 can be attributed to the lower forage biomass, especially

panicum (less than 150 kg/ha) which has the highest weighting factor amongst the three forage

species. There is a slight over prediction in land segment 3 due to higher biomass availability.

The variability in the GPS data is also noteworthy.

Figure 5-3 is the result of calibration on summer pasture 4 during cool season of 2002. The

calibration result for the cool season also captures the overall dynamics of land segments, with

exception to LS1 and LS2. A water trough and two shade structures are present in LS1. The

model is unable to represent the presence of two shade structures and it is possible that due to

more availability of shading area, more cattle are present. However, the simulation result is still

within the lower range of observed data. Cattle's presence is exceptionally high in LS2. This

high presence has been observed in the north section of all the summer pastures. Closer

examination of GPS data and personal communication with the ranch manager of BIR has

revealed that in this area cattle would stand or lie down, ruminate, for approximately 2 to 3

hours. Various studies indicate that cattle graze mostly in early morning and evening, and rest

mostly in the middle of the day (Bagshaw, 2001; Hafez & Bouissou, 1975; Martin, 1978; Sneva,

1970). A similar observation was made in an experimental study conducted to establish beef

cattle defecation frequency and distribution on hill country in New Zealand (Bagshaw, 2002). It

was observed that often cattle would rest between 11 am and midday. They would rest and

ruminate during this time on flat areas either at the top or middle of the Hield. In a Hield where

there was a large flat area at the top of the Hield next to a trough, cattle were observed to spend









the maj ority of their resting time in this area. Similar pasture setting exists at BIR where the

cattle display an affinity to rest in the northern section of all summer pastures. This high

presence is not recorded in the summer season because cattle spend most of the afternoon resting

and avoiding direct solar radiation under a shade.

The ACRU2000-HSI slightly over predicts the number of cattle in the southern land

segments. This is mainly due to the normalization of cattle population across all land segments.

The impact of under prediction in LS1 and LS2 is translated into slight over prediction in LS9-

12.

Verification Results

Figure 5-4 displays the result of verification of ACRU2000-HSI on summer pasture 5

during warm season of 2001. The stocking rate on this pasture was higher than SP4 (3 5 Cows).

There is also more variability in the observed GPS data in this pasture as compared to SP4. There

is slight over prediction in LS2 due to high initial biomass. Nevertheless, in all the land

segments, the model's performance is always within the range of observed GPS data. The

dynamics of water trough, shade, and wetland seems to be well represented by the model.

Figure 5-5 displays the result of verification on summer pasture 5 during cool season of

2002. Similar to warm season, the model is able to capture the dynamics of water trough, shade

and wetlands in the cool season as well. The phenomena where the cattle display an affinity to

rest in the northern section of the pasture is once again evident. The number of cattle recorded in

LS2 is higher than any other land segment (Figure 5-5). Additionally, there is slight over

prediction in the number of cattle in the land segments from mid field to the southern end of the

pasture (S6-12). The model results for most of these land segments (S6-9) are within the range of

observed data and overall the number of cattle prediction is better than the original model.









Sensitivity Analysis

In a typical modeling system, the model results are more sensitive to certain inputs

compared to others. This information is of essential use for future model users who may need to

calibrate the model for application on a different site. Therefore, it is important to perform a

sensitivity analysis to establish priorities in collecting and determining model parameters. An

analysis was performed to determine the sensitivity of model simulated cattle distribution to the

weighting factors.

The sensitivity analysis was performed using the six-year simulation (January 1, 1998

through December 31, 2003) on the experimental pasture at BIR. Since the model was already

parameterized for the SP5 (Figure 5-1), it was applied on the same pasture for this analysis.

Model sensitivity was determined for + 25, 50, 75, and 100% of the base input value (Table 5-3).

Summer 2001 (June) and spring 2002 (March) were the two seasons selected from the simulation

period. The sensitivity analyses were focused on the population of cattle in all land segments

during the two seasons. It is important to bear into mind that the ACRU2000-HSI model is

different from a typical process based modeling system where change in an input parameter will

result in an expected change in output. The ACRU2000-HSI model is dependent on the

hydrologic model for determination of water presence, nutrient model for rate of growth of

vegetation, and vegetation model for total biomass. In addition, as per model design, the three

weighting factors (water, shade and forage) must sum to 1.0. For example if warm season

weighting factor of water (WWFAC) is increased by 25% the other two corresponding variables

(i.e. warm season weighting factors of shade and forage, WSFAC, WOVRLFAC) must be

adjusted so that the sum of the three weighting factors equals 1 (Table 5-3). For this analysis the

adjustment of the two variables was carried out so as to maintain the ratio amongst the adjusted

variables. In Table 5-3, the sensitivity analysis is performed on WWFAC and WSFAC and









WOVRLFAC have been adjusted accordingly. Complete list of weighting factors values used in

sensitivity analysis and the corresponding adjustment in the weighting factors of other two

variables is given in Table F-1 of Appendix F.

Results of sensitivity analysis are given in Tables F-2 to F-5 of Appendix F. There is

considerable change in cattle population with change in shade and water weighting factors in the

warm period (Table F-2); especially in the land segments that consist of those features (LS 1 and

LS2). Even though water or shade availability do not exist in land segments apart from LS1 and

LS2, there is still change in cattle population in other land segments (LS3-12) due to change in

shade and water weighting factors. As explained before, this is due to corresponding change in

forage weighting factor which has to be adjusted so that all the three weighting factors sum to 1.

There is also considerable variation in presence of cattle in LS1 and LS2 with change in

weighting factor of forage in both, warm season as well as cool season (Table F-2 and F-4). This

drastic change in cattle population is due to the proportional change in shade and water

weighting factors (shade in LS1 and water in LS2). Since the base value of forage weighting

factor is higher in cool season (Table F-4), there is higher variability in cattle population in the

cool season as compared to warm season (Table F-2). A similar trend is observed with variability

in cattle population due to variation in weighting factor of individual forage species, more

variation in cool season (Table F-3) as compared to warm season (Table F-5).

Hypothetical Scenario Model Testing

Sufficient GPS data are not currently available from the region of south Florida to

quantitatively test the ACRU2000-HSI model towards BMP implications. Adequate data are

however, rarely available to make management decisions. This is the reason modeling is an

important tool which allows managers to envision future implications based on current decisions.









"In a decision-making context, the ultimate test of a model is not how accurate or truthful it is,

but only whether one is likely to make a better decision with it than without it" (Starfield, 1997).

Therefore, a hypothetical test was designed to evaluate the algorithms of the ACRU2000-

HSI model, coupled with the vegetation, hydrologic, and nutrient models. This scenario testing

determined whether the ACRU2000-HSI model can be utilized to determine the feasibility of a

BMP with phosphorus loading as an obj ective function. One of the obvious water quality-BMP

in a cow-calf operation is to exclude the cows from streams. By restricting cow's access to a

stream, direct deposition of the cattle feces in flowing water can be prevented. In its current state,

the ACRU2000 does not consist of a stream routing algorithm for water and P. Therefore,

fencing the cows away from streams or ditches cannot be defined in the model. However, a

similar scenario was designed to mimic restrictive access of cattle.

First of all, a basic set of simulation were made with the ACRU2000-HSI model to observe

change in P loading due to presence and absence of cows (Figure 5-7). The results from these

simulations came out to be counter intuitive. The P load in absence of cows was greater than in

presence of cows. These results warranted further investigation and more detailed scenario

simulations. Hence a new data obj ect called DCattleExclusion (Appendix A) was created to

accomplish the exclusion of cattle from user-specified land segments. This functionality allows

the user to specify the land segment from which the cattle are to be excluded. During simulation

the cattle are distributed only on land segments in which cattle exclusion option is turned off.

Primarily, it was important to see a difference in the P load prediction, if any, from the two

versions of the ACRU2000 model: one with the HSI algorithms and the other using equal

distribution of animal manure. Following the above stated basic run it was also crucial to see

whether the HSI additions within ACRU2000 modeling system have enhanced its capabilities to









make relevant management decisions. Three scenarios were tested on summer pasture 5 (Figure

5-1) at BIR using a 6 year (1998-2003) simulation time period. In the first scenario (Figure 5-6a)

cattle were excluded from the land segments that were close to the flume (LS7 LS12); in the

second scenario cattle were excluded from the land segments that were away from the flume

(LS1 LS6); and finally in the third scenario all cattle were stocked on the land segment that

adj oined the flume (LS 11) and they were excluded from all other land segments.

The three scenarios were designed to fence the cattle in various locations on the pasture to

observe any changes in the nutrient loading. In total, five set of simulations were made using the

ACRU2000 and the ACRU2000-HSI model and compared with observed P loading data (Figure

5-8). There is considerable difference in the nutrient predictions in the two versions of

ACRU2000. The ACRU2000-HSI' s prediction of TP is less than ACRU2000 and closer to

observed data (Figure 5-8). The ACRU2000 version assumed the animal manure to be

distributed equally amongst all land segments. The HSI version deposits manure on land

segments based on the number of animals assigned to specific land segments. However, in both

versions the total quantity of manure remains same; therefore, some difference was anticipated

yet the magnitude of difference required further investigation.

It should be noted that even though there was provision to include animal manure in case

of stocking in ACRU2000, there was no accountability of forage consumption by grazing cattle.

In the vegetation model, plant senescence is assumed to start when the daily sum of leaf areas of

all species on one land segment exceeds the critical leaf area. Each vegetation species senesces

biomass, N, and P in proportion to its leaf area. The daily total senesced biomass (Ws,i~t [kg/ha])

and the corresponding N (Ns,i~t [kg/ha]) and P and (Ps,i~t [kg/ha]) removed through the senesced

biomass are given in Equation 4-13 to 4-15 of Chapter 4. Since there was no consumption of the









vegetation by grazing cattle in ACRU2000 there was a high amount of nutrients being released

from senesced biomass. This seems to have been corrected by the ACRU2000-HSI model where

the cattle consume the vegetation as per their nutritional requirement.

Figures 5-7 shows that when the cattle are closer to the flume there is reduction in nutrient

load. This reduction can be explained by the change in quantity of senesced biomass that is

closer to the flume. When there are more cattle present near the flume they consume more forage

and hence reduction in senesced biomass. On the other hand, when cattle are away from the

flume there is increase in nutrient load due to increase in above ground biomass that senesces

and release nutrients. In the third scenario, when all the cattle are stocked on LS11 (land segment

that is adj acent to flume) there is a slight increase in nutrient loading. This increase can be

attributed to the exorbitant stocking rate (35 cows on 1.6 ha).

A budget of the ACRU2000-HSI modeling system was prepared to quantify various

"pools" of P using 6 years of simulation (Figure 5-9). It is evident that P from senesced residue

(two order magnitude higher than P from defecation) is the largest component. When the

ACRU2000-HSI model is turned on, the grazing cattle consume forage and reduce the amount of

senesced residue (Equation 4-16) which consequently reduces P load. Within the model, only the

top two layers (plant residue layer and soil surface layer) interact with surface runoff; therefore a

P budget with the top two layers of ACRU2000-HSI as a control volume was also computed

(Figure 5-10).

Apart from the P budget within the model domain it was also important to test the retention

of P within the cattle over time (Figure 5-1 1). With the exception of initial increase, the P

retained by cow remains within the bounds of 20-25 g. The P retained values correspond well to

the values published in literature (NRC, 1996). The initial jump in P retention can be explained










by the utilization of nutrient uptake algorithms in the vegetation model (Yang, 2006). The N and

P uptake algorithms in the vegetation model were adopted from GLEAMS (Knisel and Davis,

1999) with a slight modification to account for the nutrient uptake by multiple plant species in

one land segment. In the GLEAMS model the nutrient uptake is based on demand and supply of

nutrients. The P demand for species i at time t, DEMPi~t [kg/ha], is determined by the difference

between the dry matter P on two successive days as:

DEMPi~, = TDMPi,t TDMPi,t-1 (5-1)

Uptake of labile P, UPLPi~t [kg/ha], is estimated for each layer where transpiration, occurs using

UPLPi,,tt = CPLABWs,d,j x Ti~j~t (5-2)

where, CPLABWs,d~t, is the concentration of labile P. The total uptake of P is the sum over all

species i and all layers j where transpiration occurs. The P taken up is converted into the plant

biomass P:


Pu,Dt = IfUPL~Pi,;J (5-3)


where Pu,~t is the plant biomass P [kg/ha]. The amount of initial biomass will dictate the role of P

uptake during the initial phase of simulation. It is perceived that in some land segments there can

be high initial biomass of any of the three vegetation species (Bahiagrass, Floralta, and

Panicum). This will cause an increase in the supply of nutrients to support the growth of the

vegetation. Thus, during the initial stages, consumption of P enriched biomass is resulting in

higher P retention within the cow' s body. Over time, as the model equilibrates the high retention

"levels-off' to a more sustainable level.

Summary

The algorithms in the HSI model are composed of "attractants" of cattle (shade, water

trough, and wetland) and their weighting factors. The HSI methodology represents the dynamics