﻿
 UFDC Home myUFDC Home  |   Help
<%BANNER%>

# Mathematical Modeling of Citrus Greening

## Material Information

Title:
Mathematical Modeling of Citrus Greening
Physical Description:
1 online resource (78 p.)
Language:
english
Creator:
Stupiansky, Jillian C
Publisher:
University of Florida
Place of Publication:
Gainesville, Fla.
Publication Date:

## Thesis/Dissertation Information

Degree:
Doctorate ( Ph.D.)
Degree Grantor:
University of Florida
Degree Disciplines:
Mathematics
Committee Chair:
Pilyugin, Sergei S
Committee Members:
Keesling, James E
De Leenheer, Patrick
Hager, William Ward
Gillooly, James F

## Subjects

Subjects / Keywords:
citrus -- greening -- huanglongbing -- modeling -- psyllid -- roguing -- vector
Mathematics -- Dissertations, Academic -- UF
Genre:
Mathematics thesis, Ph.D.
bibliography   ( marcgt )
theses   ( marcgt )
government publication (state, provincial, terriorial, dependent)   ( marcgt )
born-digital   ( sobekcm )
Electronic Thesis or Dissertation

## Notes

Abstract:
Huanglongbing (citrus greening) is a bacterial disease that is significantly impacting the citrus industry in Florida and poses a risk to the remaining citrus-producing regions of the United States.  The work presented here centers around the development of a deterministic mathematical model that represents the spread of citrus greening disease within a single grove of trees.  The system models both the tree and insect populations and incorporates a control strategy of roguing infected trees and replanting healthy trees.  The basic reproductive number, R sub one, and its relation to the stability of the equilibria is discussed.  In particular, theorems regarding extinction of the disease when the R sub one 1 are proved.  Variations on the rates of roguing are examined to provide insight into the effect of the control strategy on the level of healthy trees that can be maintained.  A modification to the model is presented which allows for the possibility of the occurrence of a backward bifurcation.  The original model is then made more complex by incorporating a periodic parameter to account for recent biological findings in field studies, and results for this new system are proved.  Numerical studies are performed throughout to illustrate the theoretical findings.  In addition, simulations for a stochastic model are explored, as this type of model may allow for the future incorporation of the economic impact of the disease.
General Note:
In the series University of Florida Digital Collections.
General Note:
Includes vita.
Bibliography:
Includes bibliographical references.
Source of Description:
Description based on online resource; title from PDF title page.
Source of Description:
This bibliographic record is available under the Creative Commons CC0 public domain dedication. The University of Florida Libraries, as creator of this bibliographic record, has waived all rights to it worldwide under copyright law, including all related and neighboring rights, to the extent allowed by law.
Statement of Responsibility:
by Jillian C Stupiansky.
Thesis:
Thesis (Ph.D.)--University of Florida, 2013.
Local:
Electronic Access:
RESTRICTED TO UF STUDENTS, STAFF, FACULTY, AND ON-CAMPUS USE UNTIL 2015-08-31

## Record Information

Source Institution:
UFRGP
Rights Management:
Applicable rights reserved.
Classification:
lcc - LD1780 2013
System ID:
UFE0045402:00001

## Material Information

Title:
Mathematical Modeling of Citrus Greening
Physical Description:
1 online resource (78 p.)
Language:
english
Creator:
Stupiansky, Jillian C
Publisher:
University of Florida
Place of Publication:
Gainesville, Fla.
Publication Date:

## Thesis/Dissertation Information

Degree:
Doctorate ( Ph.D.)
Degree Grantor:
University of Florida
Degree Disciplines:
Mathematics
Committee Chair:
Pilyugin, Sergei S
Committee Members:
Keesling, James E
De Leenheer, Patrick
Hager, William Ward
Gillooly, James F

## Subjects

Subjects / Keywords:
citrus -- greening -- huanglongbing -- modeling -- psyllid -- roguing -- vector
Mathematics -- Dissertations, Academic -- UF
Genre:
Mathematics thesis, Ph.D.
bibliography   ( marcgt )
theses   ( marcgt )
government publication (state, provincial, terriorial, dependent)   ( marcgt )
born-digital   ( sobekcm )
Electronic Thesis or Dissertation

## Notes

Abstract:
Huanglongbing (citrus greening) is a bacterial disease that is significantly impacting the citrus industry in Florida and poses a risk to the remaining citrus-producing regions of the United States.  The work presented here centers around the development of a deterministic mathematical model that represents the spread of citrus greening disease within a single grove of trees.  The system models both the tree and insect populations and incorporates a control strategy of roguing infected trees and replanting healthy trees.  The basic reproductive number, R sub one, and its relation to the stability of the equilibria is discussed.  In particular, theorems regarding extinction of the disease when the R sub one 1 are proved.  Variations on the rates of roguing are examined to provide insight into the effect of the control strategy on the level of healthy trees that can be maintained.  A modification to the model is presented which allows for the possibility of the occurrence of a backward bifurcation.  The original model is then made more complex by incorporating a periodic parameter to account for recent biological findings in field studies, and results for this new system are proved.  Numerical studies are performed throughout to illustrate the theoretical findings.  In addition, simulations for a stochastic model are explored, as this type of model may allow for the future incorporation of the economic impact of the disease.
General Note:
In the series University of Florida Digital Collections.
General Note:
Includes vita.
Bibliography:
Includes bibliographical references.
Source of Description:
Description based on online resource; title from PDF title page.
Source of Description:
This bibliographic record is available under the Creative Commons CC0 public domain dedication. The University of Florida Libraries, as creator of this bibliographic record, has waived all rights to it worldwide under copyright law, including all related and neighboring rights, to the extent allowed by law.
Statement of Responsibility:
by Jillian C Stupiansky.
Thesis:
Thesis (Ph.D.)--University of Florida, 2013.
Local:
Electronic Access:
RESTRICTED TO UF STUDENTS, STAFF, FACULTY, AND ON-CAMPUS USE UNTIL 2015-08-31

## Record Information

Source Institution:
UFRGP
Rights Management:
Applicable rights reserved.
Classification:
lcc - LD1780 2013
System ID:
UFE0045402:00001

Full Text

PAGE 1

PAGE 2

c 2013JillianClaireStupiansky 2

PAGE 3

Tomyamazingparents 3

PAGE 4

PAGE 5

TABLEOFCONTENTS page ACKNOWLEDGMENTS .................................. 4 LISTOFFIGURES ..................................... 7 ABSTRACT ......................................... 8 CHAPTER 1BACKGROUND ................................... 9 1.1Introduction ................................... 9 1.2Preliminaries .................................. 12 1.2.1EquilibriaandSolutionsofDifferentialEquations .......... 12 1.2.2Matrices ................................. 14 1.2.3FloquetTheoryforPeriodicSystems ................. 15 2MATHEMATICALMODELWITHCONSTANTPARAMETERS .......... 18 2.1DevelopmentoftheModel ........................... 18 2.2Equilibria .................................... 19 2.3TheBasicReproductiveNumber R 0 ..................... 21 2.3.1Calculationof R 0 ............................ 21 2.3.2LocalStabilityoftheDisease-FreeEquilibrium ........... 23 2.4ExtinctionoftheDiseasewhen R 0 1 .................... 25 2.5PersistenceoftheDiseasewhen R 0 > 1 ................... 28 2.5.1PersistenceoftheDiseaseforFullSystem .............. 29 2.5.2StabilityoftheEndemicEquilibriuminaSpecialCase ....... 33 3ROGUINGASACONTROLSTRATEGY ..................... 37 3.1TransientBehavior ............................... 37 3.2DifferentRoguingRates ............................ 40 4MODIFICATIONOFTHEMODEL ......................... 44 5PERIODICITY .................................... 48 5.1PeriodicModel ................................. 48 5.2ExtinctionoftheDiseaseinPeriodicModel ................. 49 5.3IncorporatingthePeriodicFunctionasaSmallPerturbation ........ 53 5.4StabilityoftheDisease-FreeSolution .................... 60 5.5StabilityoftheEndemicSolutioninaSpecialCase ............. 64 6NUMERICALEXPLORATIONS ........................... 70 REFERENCES ....................................... 75 5

PAGE 6

BIOGRAPHICALSKETCH ................................ 78 6

PAGE 7

LISTOFFIGURES Figure page 2-1Flowdiagramforthegrove-scalecitrusgreeningmodel(2 # 1)-(2 # 6) ...... 18 2-2Simulationofsystem(2 # 1)-(2 # 6)with R 0 < 1 ................... 28 2-3Simulationofsystem(2 # 1)-(2 # 6)with R 0 > 1 ................... 29 3-1Simulationsofsystem(3 # 1)-(3 # 6)with R 0 > 1 anddifferentroguingrates ... 40 4-1Simulationsofsystem(4 # 6)-(4 # 6)with R 0 < 1 anddifferentinitialconditions 47 5-1Simulationofsystem(5 # 1)-(5 # 6)satisfyingCondition(5 # 7) .......... 53 5-2Simulationofsystem(5 # 1)-(5 # 6)satisfyingCondition(5 # 9) .......... 64 5-3Simulationofsystem(5 # 10)-(5 # 15)with S (0) < N 0 ............... 69 6-1Simulationofsingle-grovestochasticmodel .................... 71 6-2Timeaveragesofsimulationofsingle-grovestochasticmodel .......... 72 6-3Timeaveragesofsimulationoftwo-grovestochasticmodel ........... 73 6-4Simulationoftwo-grovedeterministicmodel .................... 74 7

PAGE 8

AbstractofDissertationPresentedtotheGraduateSchool oftheUniversityofFloridainPartialFulllmentofthe RequirementsfortheDegreeofDoctorofPhilosophy MATHEMATICALMODELINGOFCITRUSGREENING By JillianClaireStupiansky August2013 Chair:SergeiS.Pilyugin Major:Mathematics Huanglongbing(citrusgreening)isabacterialdiseasethathasasignicantimpact onthecitrusindustryinFloridaandposesarisktotheremainingcitrus-producing regionsoftheUnitedStates.Theworkpresentedherecentersaroundthedevelopment ofadeterministicmathematicalmodelthatrepresentsthespreadofcitrusgreening diseasewithinasinglegroveoftrees.Thesystemmodelsboththetreeandinsect populationsandincorporatesacontrolstrategyofroguinginfectedtreesandreplanting healthytrees.Initially,wewillexamineasysteminwhichallinteractionsoccurata constantrate.Inthiscase,thebasicreproductivenumber, R 0 ,anditsrelationtothe stabilityoftheequilibriaisdiscussed.Inparticular,theoremsregardingextinctionofthe diseasewhen R 0 1 andpersistenceofthediseasewhen R 0 > 1 areproved.Variations ontheratesofroguingareexaminedtoprovideinsightintotheeffectofthecontrol strategyonthelevelofhealthytreesthatcanbemaintained.Amodicationtothemodel ispresentedwhichallowsforthepossibilityoftheoccurrenceofabackwardbifurcation. Theoriginalmodelisthenmodiedtoaccountforseasonalvariationintherateof transmissionofthedisease,andresultspertainingtothestabilityofthisnewsystemare proved.Numericalstudiesareperformedthroughouttoillustratethetheoreticalndings. Inaddition,simulationsofastochasticmodelareexplored,asthistypeofmodelmay allowforthefutureincorporationoftheeconomicimpactofthedisease. 8

PAGE 9

CHAPTER1 BACKGROUND 1.1Introduction Huanglongbing (HLB)isavector-transmittedbacterialinfectionthatiscurrently affectingalltypesofcitrusinthestateofFlorida.InChinese,thenamemeans"yellow dragondisease"[ 20 ],butintheUnitedStatesitisfrequentlyreferredtoas citrus greening disease.Symptomsofthediseaseincludestuntedgrowthandpoorowering ofcitrustreesaswellasblotchymottlingandyellowingoftheirleaves.Inaddition,the fruitthatisproducedbyinfectedtreesismisshapenandsmallerthannormal,witha greentintandabittertaste[ 20 ],makingthefruitinedible.Thediseasehasseverely affectedthecitrusindustryinFlorida,thenation'slargestcitrusproducerandthesecond largestproduceroforangejuiceintheworld.ThepresenceofHLBhasalsobeen detectedinothersoutheasternstatesincludingTexas[ 26 ],andallcitrusproducing regionsoftheUnitedStatesareconsideredtobeatrisk.Becauseofitsimpactonmany sectorsoftheFloridaeconomyandtheimplicationsforthecitrusindustrynationwide, citrusgreeninghasbecomeanimportantissuetostudy. HLBwasrstdiscoveredinChinainthelate1800s[ 17 ].Differentstrainsofthe diseasehavealsocreatedproblemsinthepastinbothAfrica,around1930,andin Brazil,beginningin2004[ 17 ].Itisbelievedthatthevectorsthattransmitthedisease arrivedinsouthernFloridain1998[ 20 ],andcitrusgreeningbeganspreadingnorth throughoutthestatein2000[ 18 ].However,itwasnotuntil2005thatsymptomatic treeswererstdetected[ 17 ].Sincethen,thediseasehasbeenspottedinalmost everycountyasfarnorthasAlachua.Thediseasehasevenaffectedsomecitrusinthe Floridapanhandle,althoughsofaronlyatlargediscountstoresthathandleandship citrustrees[ 20 ]. AninsectknownastheAsiancitruspsyllid, Diaphorinacitri ,carriestheorganism thatcausescitrusgreening, CandidatusLiberibacterasiaticus (Las)[ 17 29 ].The 9

PAGE 10

PAGE 11

whichtreesmustberogued[ 17 ].Especiallyincommercialgroves,farmersarereluctant toeliminateahealthy-lookingtreethatisstillcapableofproducinggoodfruit.Oncea treeisrogued,anothercanbereplantedinitsplace.However,itispossiblethatthe remnantsoftreerootsinthesoilmayserveasareservoirforthedisease,asithasbeen observedthatrecentlyplantedtreesaremoresusceptible,creatinganothersetbackto thiscontrolstrategy[ 35 ].Anothermethodistotreatunhealthytreeswithantibiotics.This canbeveryeffective,butitisverycostly.Itisalsonotapermanentx;symptomswill returnaboutayearafterbeingtreated[ 20 ].Witheithercontrolstrategy,itisimportant toknowwhichasymptomatictreesareactuallyinfectedwiththedisease.Onewayis totestplanttissuewiththepolymerasechainreactiontechnique(PCR).However,it isnotfeasibleforeverysingletreeinagrovetobetestedthisway.Also,thedisease isnotevenlydistributedwithinagiventree,andthusthetestmaynotalwaysproduce accurateresults.Manyfarmersusescoutstovisuallyinspecttreesforpsyllids,butitis generallydifculttoseetheminisculeinsectsthatareonlythreetofourmillimetersin diameter[ 20 ]. Thereareafewcontrolstrategiesthataremorerealisticandeasiertoimplement, thoughlesseffective.Theseincludesprayinginsecticidesoverentiregrovesaswell aseliminatinggrafting[ 20 ].Itisalsoimportantforcitrusstocktobeinspectedand screenedbeforeitistransportedorsold,anditisverybenecialfornurseriesthatgrow citrustobescreened-in[ 17 ].AquarantinethatbeganJune17,2010,isdesignedto preventthetransportationofcitrusthroughoutthestateofFlorida[ 3 ]. Citrusgreeninghasbecomeverywide-spreadandisdifculttoeradicate.Because citrusissuchanimportantpartofFlorida'seconomy,thisdiseaseisanissueofgreat concern.ArecentstudybytheUniversityofFlorida'sInstituteofFoodandAgricultural Sciencesestimatesthatfrom2006to2011citrusgreeninghascaused\$3.63billion inlostrevenueandover6,000lostjobsinthestate.Theve-yearproductionlevelfor orangejuiceisestimatedtobe1.7billiongallonslessthanprojected[ 21 ].Ifthedisease 11

PAGE 12

continues,eventuallytherewillnotbeenoughcitrusproducedforthejuiceplantsto operate;manyplantshavealreadybeguntoimportextracitrusfromBrazilinorderto keepuptheirproductionlevels[ 35 ].Unfortunately,manybelievethatthedamagemay beirreversibleinthestateofFlorida.However,HLBhasalsobeendetectedinGeorgia, SouthCarolina,andLouisiana.InJanuary2012,theTexasDepartmentofAgriculture andUSDAconrmedthedetectionofthediseaseinacommercialgroveinTexas[ 26 ]. TherstincidencesofHLB-infectedtreeswerealsodiscoveredinCaliforniaandArizona in2012,andasiancitruspsyllidsalonehavebeendiscoveredinHawaii,Mississippiand Alabama.ItisclearthatresearchontheprogressionofthediseaseinFloridacouldbe instrumentalinpreventingthesameoutbreakinotherstates. 1.2Preliminaries Thefollowingdenitionsandtheoremswillbeusedthroughoutthisdissertation. 1.2.1EquilibriaandSolutionsofDifferentialEquations Webeginwiththebasicdenitionofstabilityofanequilibriumpoint,andalso provideresultspertainingtosolutionsofdifferentialequations.Thesefollowfrom Chicone[ 9 ]andSmithandWaltman[ 32 ]unlessotherwisenoted. Denition1. Let t ( x ) beasolutiontothedifferentialequation x = f ( x ) x \$ R with 0 ( x )= x .Anequilibriumpoint x 0 ofthedifferentialequationis stable ifforeach > 0 thereexists # > 0 suchthat | t ( x ) # x 0 | < forall t % 0 whenever | x # x 0 | < # .An equilibriumpoint x 0 is asymptoticallystable ifitisstableandthereexists # > 0 such that lim t !" | t ( x ) # x 0 | =0 whenever | x # x 0 | < # .Anequilibriumpoint x 0 is unstable ifit isnotstable. Denition2. Supposethat JxU isanopensubsetof R x R n .Afunction : JxU & R n givenby ( t x ) '& t ( x ) iscalleda ow ofthedifferentialequation x = f ( x ) ifforeach xed x 0 ( x ) ( x and t + s ( x )= t ( s ( x )) Denition3. Aset S ) R n iscalledan invariantset forthedifferentialequation x = f ( x ) x \$ R n ,if,foreach x in S ,thesolution t & t ( x ) ,denedonitsmaximal 12

PAGE 13

intervalofexistence,hasitsimagein S S isa maximalinvariantset if,foranyinvariant set T suchthat S T + = , S T Denition4. ([ 31 ])Let ( X d ) beametricspace,andlet ,+ = M X M iscalleda uniformrepeller under ifthereexistssome > 0 suchthat liminf t !" d ( t ( x ), M ) % forall x \$ X \ M Denition5. The stablemanifold ofanequilibriumpoint x 0 foranautonomousdifferentialequation x = f ( x ) withow t isthesetofallpoints x suchthat lim t !" t ( x )= x 0 Theorem1.1. ([ 14 ])Let S beacompactsubsetof X suchthat X \ S ispositively invariant.Anecessaryandsufcientconditionfor S tobeauniformrepelleristhatthere existaneighborhood U of S andacontinuousfunction P : X & R + satisfyingthe followingconditions: (1) P ( x )=0 ifandonlyif x \$ S (2)forall x \$ U \ S thereexists T ( x ) > 0 suchthat P ( T ( x ) ( x )) > P ( x ) Denition6. Let f : R xD & R n ,where D isanopensubsetof R n ,beavector-valued function, f =( f 1 f 2 ,..., f n ) .Thefunction f issaidtobeof typeK in D if,foreach i and all t f i ( t a ) f i ( t b ) foranytwopoints a and b in D satisfying a i b i forall i Theorem1.2. (Kamke'sTheorem)Let f beLipschitzcontinuouson R xD andoftype K Let x ( t ) beasolutionof x = f ( t x ) denedon [ a b ] .If z ( t ) isacontinuousdifferentiable functionon [ a b ] satisfying z f ( t z ) on ( a b ) with z ( a ) x ( a ) ,then z ( t ) x ( t ) forall t in [ a b ] .If y ( t ) iscontinuousanddifferentiableon [ a b ] satisfying y % f ( t y ) on ( a b ) with y ( a ) % x ( a ) ,then y ( t ) % x ( t ) forall t in [ a b ] Theorem1.3. (Gronwall'sInequality)Supposethat a < b andlet \$ and % be nonnegativecontinuousfunctionsdenedontheinterval [ a b ] .Moreover,supposethat \$ isdifferentiableon ( a b ) withnonnegativecontinuousderivative \$ .If,forall t \$ [ a b ] ( t ) \$ ( t )+ t a % ( s ) ( s ) ds 13

PAGE 14

then ( t ) \$ ( t ) e t a ( s ) ds forall t \$ [ a b ] 1.2.2Matrices Inthissectionwedenethespectralradiusofamatrixandgiveseveralresultsthat willbeusedtoanalyzestability.Inparticular,thePerron-FrobeniusTheorem,whichis ofgreatimportance,canbefoundhere.Wealsoprovideaspecialresultpertainingto Hurwitzmatrices.Unlessotherwisenoted,theseresultscanbefoundinBermanand Plemmons[ 5 ]andThieme[ 34 ]. Denition7. Let & 1 ,..., & n betheeigenvaluesofan nxn matrix A .Thenthe spectral radius is ( A )=max i ( | & i | ) Denition8. Let & 1 ,..., & n betheeigenvaluesofan nxn matrix A .Thenthe spectral bound is s ( A )= max i ( / ( & i )) Denition9. Asquarematrix A iscalled quasi-positive ifalloff-diagonalentriesare nonnegative. Denition10. Amatrix A is reducible ifthereexistsapermutationmatrix P suchthat P # AP isblocktriangular. A is irreducible ifitisnotreducible. Denition11. Amatrix A % 0 is primitive ifthereexists k \$ N suchthat A k > 0 Theorem1.4. (Perron-FrobeniusTheorem)(1)If A ispositive,then ( A ) isasimple eigenvalue,greaterthanthemagnitudeofanyothereigenvalue. (2)If A % 0 isirreducible,then ( A ) isasimpleeigenvalue,anyeigenvalueof A ofthe samemodulusisalsosimple, A hasapositiveeigenvector x correspondingto ( A ) ,and anynonnegativeeigenvectorof A isamultipleof x (3)If A isquasi-positiveandirreducible,then s ( A ) isaneigenvalueof A associatedboth withapositiveeigenvectorof A andapositiveeigenvectorofthetransposedmatrix A # Moreover,if x > 0 isavectorand u \$ R suchthat Ax % ux ,thereexistssomevector z > 0 andsomescalar & % u suchthat Az = & z andinparticular s ( A ) % u 14

PAGE 15

Theorem1.5. If A B arematricessuchthat 0 A B ,then ( A ) ( B ) Denition12. Asquarematrix A is Hurwitz ifalleigenvaluesof A havestrictlynegative realpart. Theorem1.6. ([ 9 ])Let A bean nxn matrixthatisHurwitzandlet 0 0 beanormon R n Thenthereareconstants k a > 0 suchthatforall v \$ R n andall t % 0 0 e tA v 0" ke \$ at 0 v 0 1.2.3FloquetTheoryforPeriodicSystems Toanalyzetheperiodicsystem( 51 )-( 56 ),wewillneedsomebackground informationonFloquettheory.ThesedenitionsandresultsarefromChicone[ 9 ] unlessotherwisespecied. Considerthelinearsystem x = A ( t ) x x \$ R n (11) where t & A ( t ) isacontinuous T -periodicmatrix-valuedfunction. Therstpropositionandtheorem,whilenotactuallyapartofFloquettheory,willbe usedintheanalysisofourperiodicsystem. Proposition1.1. (Liouville'sFormula)Supposethat t & ( t ) isamatrixsolutionofthe homogeneouslinearsystem 11 ontheopeninterval J .Thenforany t t 0 \$ J det ( t )=det ( t 0 ) e t t 0 tr A ( s ) ds Theorem1.7. ([ 6 ])Let X R beaconnectedintervaland C beasimplyconnected openregion.If f ( x & ) iscontinuousin ( x & ) andanalyticin & throughouttheregion ( X ) ,thesamewillbetrueof ( f / (& Denition13. An nxn matrix-valuedfunction X ( t ) on R iscalleda fundamentalmatrix solution ofthelinearsystem 11 ifitscolumnsformalinearlyindependentsetofvector 15

PAGE 16

solutions.Thefundamentalmatrixsolution X ( t ) isthe principalfundamentalmatrix solution if X (0)= I Theorem1.8. (Floquet'sTheorem)If X ( t ) isafundamentalmatrixsolutionofthe T -periodicsystem 11 ,then,forall t \$ R X ( t + T )= X ( t ) X \$ 1 (0) X ( T ). Inaddition,thereisamatrix B (whichmaybecomplex)suchthat e TB = X \$ 1 (0) X ( T ) anda T -periodicmatrixfunction t & P ( t ) (whichmaybecomplex-valued)suchthat X ( t )= P ( t ) e tB forall t \$ R .Also,thereisarealmatrix R andareal 2 T -periodicmatrix function t & Q ( t ) suchthat X ( t )= Q ( t ) e tR forall t \$ R Denition14. The Floquetmultipliers ofa T -periodiclineardifferentialequation,such as 11 ,aretheeigenvaluesoftheprincipalfundamentalsolutionevaluatedat t = T Thatis,theFloquetmultipliersaretheeigenvaluesof X ( T ) Theorem1.9. (1)IftheFloquetmultipliersoftheperiodicsystem 11 allhavemodulus lessthanone,thenthezerosolutionisasymptoticallystable. (2)IfatleastoneFloquetmultiplieroftheperiodicsystem 11 hasmodulusgreaterthan one,thenthezerosolutionisunstable. Nowconsiderthenonlinearsystem x = F ( t x ), x \$ R n (12) where F : R x R n & R n isa T -periodicsmoothfunctionsuchthat F ( t + T x )= F ( t x ) forall x \$ R n andall t \$ R .Let t ( x ) bethesolutionto 12 with 0 ( x )= x Denition15. If t ( x ) isdenedonthetimeinterval [0, T ] ,the Poincar ` emap isdened as P ( x )= T ( x ) 16

PAGE 17

Thesetof T -periodicsolutionsofthesystem 12 isinone-to-onecorrespondence withthexedpointsofthePoincar ` emap. Denition16. A T -periodicsolution y ( t ) ofthesystem 12 is locallyorbitallystable if foreachopenset V ) R n thatcontains { y ( t ):0 t T } ,thereisanopenset W ) V suchthateverysolution,startingatapointin W at t =0 ,staysin V forall t % 0 .The periodicsolutioniscalled locallyasymptoticallystablewithaphase if,inaddition, thereexists # > 0 suchthatforeach x 0 with | x 0 # y ( t ) | < # ,thereexists t ( x 0 ) \$ [0, T ) withthepropertythat ( t x 0 ) & y ( t + t ( x 0 )) as t &1 ThederivativeofthePoincar ` emapatapoint p is DP ( p )= " x T ( p ) .Thisderivative istheprincipalfundamentalsolutionat t = T ofthevariationalproblem W = F x ( t t ( p )) W W (0)= I (13) Theorem1.10. Thestabilityofa T -periodicsolution y ( t ) ofthesystem 12 isdeterminedbytheFloquetmultipliersoftheperiodicvariationalsystem 13 .Thatis,when ( DP ( p )) < 1 y ( t ) islocallyasymptoticallystable.When ( DP ( p )) > 1 y ( t ) is unstable. 17

PAGE 18

CHAPTER2 MATHEMATICALMODELWITHCONSTANTPARAMETERS 2.1DevelopmentoftheModel Wedevelopamodelthatrepresentsthespreadofcitrusgreeningdiseasewithina singlegroveoftrees.Thecitrustreepopulationisdividedintofourstages. S denotes susceptibletreesand R representsdeadtrees.Duetotheobserveddelayinthe appearanceofsymptomsofcitrusgreening,wesplittheinfectedtrees I intoan asymptomatic(latent)stage, I 1 ,andasymptomaticstage, I 2 .Forthepsyllidvector populationwelet V \$ and V + representtheuninfected/susceptibleandinfectedpsyllids, respectively.Weassumethatthegroveissubjecttoarogueandreplantdisease managementstrategy.Inthemodel,symptomaticanddeadtreesareroguedatarate andthecorrespondingspotsarereplantedwithnewtrees.Becausethesoilmay serveasareservoirforthedisease,weassumethataproportion f \$ [0,1] ofthenewly plantedtreeswillbehealthyandaproportion 1 # f willbecomeinfectedimmediately.A schematicforthemodelisprovidedinFigure 2-1 Figure2-1. Flowdiagramforthegrove-scalecitrusgreeningmodel(2 # 1)-(2 # 6) Weassumethatinfectedtreesprogressfromthelatentstatetothesymptomatic stateattherate ) .Theparameter \$ representsthedisease-associatedmortalityofthe symptomatictrees;thenaturaldeathrateofthetreesisneglected.Themortalityrateof thevectorsisgivenby ,whichisassumedtobethesameforuninfectedandinfected 18

PAGE 19

psyllids.Weassumethatthereisaxednumberofcontactsperunittimebetween treesandpsyllids.Thebitingrate ,givenasunits/time,accountsfortheproportion ofbitesthatresultinsuccessfultransmission.Thetotalnumberoftreesisgivenby N = S + I 1 + I 2 + R while N 0 representsthenumberofspacesinthegrove.Wewill assumethroughoutthat N N 0 .Thetreesareassumedtobereplantedataratethat isproportional,withconstant b ,tothenumberofemptyspacesinthegrove.Thetotal psyllidpopulationisrepresentedby V = V \$ + V + withlogisticgrowthrate r andcarrying capacity K .Asanexample,theinfectionrateoftreesisgivenbytheproductofthe bitingrateandtheprobabilitythatabiteinvolvesahealthytreeandaninfectedpsyllid. Withtheaboveassumptions,themodeltakesthefollowingform: I 1 = S N V + V # ) I 1 + (1 # f )( I 2 + R ), (21) I 2 = ) I 1 # ( \$ + ) I 2 (22) R = \$ I 2 # R (23) V + = V \$ V I 1 + I 2 N # V + (24) S = b 1 # N N 0 # # S N V + V + f ( I 2 + R ), (25) V \$ = rV 1 # V K # # V \$ V I 1 + I 2 N # V \$ (26) 2.2Equilibria Webeginouranalysisofthemodelbymakingthefollowingimmediateobservations. First,duetothespecicformofEquations( 21 )-( 26 ),themodeliswell-posed;thatis, allpositivesolutionsarewell-denedandremainpositiveinforwardtime.Second, addingEquations( 21 )-( 23 )and( 25 ),wendthatthetotaltreepopulationis governedby N = b 1 # N N 0 # 19

PAGE 20

Similarly,whenweaddEquations( 24 )and( 26 ),thetotalvectorpopulationsatises V = rV 1 # V K # # V Consequently,allpositivesolutionsof( 21 )-( 26 )areboundedandcanbeextendedto all t \$ (0,+ 1 ) .Additionally, N ( t ) & N 0 and V ( t ) & V # =(1 # r ) K (weassumethat < r sothat V # > 0 )exponentiallyfastas t & + 1 Nowwelookforsteadystates ( I # 1 I # 2 R # V # + S # V # \$ ) ofsystem( 21 )-( 26 ).Clearly, N = N 0 and V = V # atequilibrium.Trivially, x 0 =(0,0,0,0, N 0 V # ) isanequilibrium whichwedesignateasthedisease-freeequilibrium(DFE).Tondendemicequilibria, weset( 21 )-( 26 )equaltozeroandobtain I 2 = ) \$ + I 1 R = \$ I 2 V + = V # ( I 1 + I 2 ) N 0 V # + ( I 1 + I 2 ) Thuseachinfectedstatecanbeexpressedintermsof I 1 .If I 1 =0 weseethat I 2 = R = V + =0 andweareinthecaseofthedisease-freeequilibrium.Hencethereexistsan endemicequilibriumifandonlyif I 1 > 0 .If I 1 + =0 ,( 25 )yields S = N 0 f ) I 1 + N 2 0 f V # 2 \$ 1 # + 1 \$ + % % Recallthat N 0 = S + I 1 + I 2 + R ,sowecancompute N 0 = N 0 f ) I 1 + N 2 0 f V # 2 ( 1 # + 1 \$ + % ) + 1+ ) # I 1 Thereforethenecessaryconditionforexistenceofapositive(endemic)equilibriumis: I 1 > 0 2 N 2 0 f V # 2 ( 1 # + 1 \$ + % ) < N 0 2 2 ( 1 # + 1 \$ + % ) N 0 V # f > 1. (27) Thelastinequalityiscloselyrelatedtothebasicreproductivenumber R 0 whichwe discussinthefollowingsection.Infact, I 1 > 0 ifandonlyif R 0 > 1 ;thatis,theendemic equilibriumexistsifandonlyifthebasicreproductivenumberisgreaterthanone. 20

PAGE 21

2.3TheBasicReproductiveNumber R 0 Thebasicreproductivenumber R 0 representstheaveragenumberofsecondary infectionsthatresultfromtheintroductionofasingleinfectedindividual(atreeora vector)intoasusceptiblepopulation.Therearevariousapproachesforcalculating R 0 whichleadstosomeambiguityinitsdenition.Forinstance,vandenDreisscheand Watmoughdene R 0 tobethespectralradiusofthenextgenerationmatrix FV \$ 1 [ 36 ].In Subsection 2.3.1 wecalculate FV \$ 1 foroursystem( 21 )-( 26 )andnd R 0 = 1 # f 2 + & (1 # f ) 2 4 + 2 V # N 0 1 ) + 1 \$ + # (28) WeprovebelowinSubsection 2.3.2 theexistenceofanequivalentthresholdcondition involvingthequantity T 0 = 2 V # N 0 1 ) + 1 \$ + # +1 # f (29) Thequantity T 0 providesthefollowingbiologicalinterpretation.Supposethatasingle infectedtreeinstage I 1 isintroducedintoacompletelysusceptiblegrove.Theaverage numberofsecondaryinfectionsresultingfrompsyllidcontactduringthe I 1 and I 2 stages ofthetreeare,respectively, 2 V # N 0 ) and 2 V # N 0 ( \$ + ) Thetreewillnecessarilyberoguedineitherthe I 2 or R stageandwillonaverage produce 1 # f newlyinfected I 1 trees.Thustheexpectednumberofsecondaryinfections isexactly T 0 .InTheorem 2.1 ,weprovetheequivalenceof T 0 and R 0 aswellasa conditionforlocalstabilityoftheDFE. 2.3.1Calculationof R 0 Mathematically, R 0 hasbeendenedtobethespectralradiusofthenextgeneration matrix[ 36 ].Thatis,werewritethevectoreldof( 21 )-( 26 )as x i = F i ( x ) # V i ( x ), i =1,2,3,4,5,6, 21

PAGE 22

where F i ( x ) representstherateofnewinfectionsappearinginstate i and V i = V \$ i # V + i where V + i istherateofindividualsenteringstate i byallmeansotherthananew infectionand V \$ i istherateofindividualsleavingstate i .Wenowconsideronlythe 4 x 4 matrixcorrespondingtothestates I 1 I 2 R V + ,asthisdenitionfor R 0 dependsonlyon transitionsintoandoutoftheinfectedcompartments.Welet F = ( F i ( x j ( x 0 ) ( V = ( V i ( x j ( x 0 ) ( 1 i j 4, where x 0 istheDFE.Thus F isthematrixofderivativescorrespondingtonewinfections intheinfectedcompartmentswhile V isthematrixofderivativescorrespondingtoall othermodesofenteringorexitinganinfectedcompartment.Thenthebasicreproductive numberisdenedas R 0 = ( FV \$ 1 ). Wecomputethenextgenerationmatrix FV \$ 1 forsystem( 21 )-( 26 ).Calculationgives F = ) * * * + 0 (1 # f ) (1 # f ) & V 00 00 00 00 & N 0 & N 0 00 , V = ) * * * + ) 000 # )\$ + 00 0 # \$' 0 000 , and V \$ 1 = ) * * * + 1 # 000 1 \$ + % 1 \$ + % 00 \$ % ( \$ + % ) \$ % ( \$ + % ) 1 % 0 000 1 . 22

PAGE 23

Therefore FV \$ 1 = ) * * * + 1 # f 1 # f 1 # f & V 0 000 0 000 & N 0 \$ 1 # + 1 \$ + % % & N 0 ( \$ + % ) 00 . Clearly, FV \$ 1 hastwozeroeigenvalues.Theremainingeigenvaluesaredeterminedby & 2 # (1 # f ) & # 2 V # N 0 1 ) + 1 \$ + # =0. Thus,thespectralradiusof FV \$ 1 is: ( FV \$ 1 )= 1 # f 2 + & (1 # f ) 2 4 + 2 V # N 0 1 ) + 1 \$ + # = R 0 2.3.2LocalStabilityoftheDisease-FreeEquilibrium Theorem2.1. Forthesystem( 21 )-( 26 )and R 0 and T 0 asdenedin( 28 )and( 29 ), respectively: 1. R 0 < 1 ifandonlyif T 0 < 1 2. T 0 < 1 ifandonlyifalleigenvaluesoftheJacobianmatrixofsystem( 21 )-( 26 ) evaluatedattheDFEhavenegativerealparts. Proof. ItisshowninSubsection 2.3.1 that R 0 isthelargestpositiverootof p ( & )= & 2 # (1 # f ) & # 2 V # N 0 1 ) + 1 \$ + # Thismeansthat 0= R 2 0 # (1 # f ) R 0 # ( T 0 # (1 # f )), whichrearrangesto T 0 # 1=( R 0 # 1)( R 0 + f ). Since R 0 + f ispositive,thesignsof T 0 # 1 and R 0 # 1 mustbethesameandtherst assertionfollows. 23

PAGE 24

Toprovethesecondassertion,wecomputetheJacobianmatrixofsystem ( 21 )-( 26 )attheDFE: J ( x 0 )= ) * * * * * * * + # )' (1 # f ) (1 # f ) & V 00 ) # ( \$ + )0 000 0 \$ # 000 & N 0 & N 0 0 # 00 # b N 0 # b N 0 + f # b N 0 + f # & V # & N 0 0 # & N 0 # & N 0 02 # r 0 # r . Weseethat J ( x 0 ) hasablocktriangularformandtherefore # & N 0 and # r are eigenvaluesof J ( x 0 ) .Itfollowsfromourassumption < r thatthesetwoeigenvalues arenegative.Welet J ( x 0 ) betheupperleft 4 3 4 submatrixof J ( x 0 ) .Thentheremaining eigenvaluesaredeterminedbythecharacteristicequationof J ( x 0 ) : p ( & )=( & + )( & + )( & + ) )( & + \$ + ) # 2 V # N 0 ( & + )( & + \$ + + ) ) # )' (1 # f )( & + )( & + + \$ ). Anyroot & of p ( & ) withRe ( & ) % 0 isalsoarootof q ( & ) denedby q ( & )= p ( & ) ( & + )( & + )( & + ) )( & + \$ + ) =1 # 2 ( & + \$ + + ) ) V # N 0 ( & + )( & + ) )( & + \$ + ) # )' (1 # f ) ( & + )( & + ) ) Weobservethat q ( & ) ismonotoneincreasingin & when & > 0 .Thereforesincethe leadingcoefcientof p ( & ) ispositiveweknowthat p ( & ) hasnopositiverealrootsif andonlyif q (0) > 0 .Thereforeallrealeigenvaluesof J ( x 0 ) arenegativeifandonlyif q (0) > 0 ,whichisequivalentto 2 V # N 0 1 ) + 1 \$ + # +1 # f < 1; thatis,allrealeigenvaluesof J ( x 0 ) arenegativeifandonlyif T 0 < 1 .Toprovethatall complexeigenvaluesof J ( x 0 ) havenegativerealpartswedene G ( & )=1 # q ( & ) and 24

PAGE 25

suppose p ( & )=0 withRe ( & ) % 0 andIm ( & ) + =0 .Then G ( & )=1 where G ( & )= 2 V # N 0 ( & + )( & + ) ) + 2 ) V # N 0 ( & + )( & + ) )( & + \$ + ) + )' (1 # f ) ( & + )( & + ) ) Weclaim | G ( & ) | < G ( Re ( & )) .Indeed, | G ( ) | " 2 V # N 0 | + || + # | + 2 # V # N 0 | + || + # || + \$ + % | + #% (1 # f ) | + % || + # | < 2 V # N 0 ( Re ( )+ )( Re ( )+ # ) + 2 # V # N 0 ( Re ( )+ )( Re ( )+ # )( Re ( )+ \$ + % ) + #% (1 # f ) ( Re ( )+ % )( Re ( )+ # ) = G ( Re ( )), wherethesecondinequalityisstrictsinceIm ( & ) + =0 .Thereforewehave T 0 < 1 2 q (0) > 0 2 G (0) < 1 ,whichimplies G ( Re ( & )) < 1 since G isdecreasingin & .Then | G ( & ) | < G ( Re ( & )) < 1 givesacontradiction. Inadditiontoprovingtheequivalenceofthethresholdconditions T 0 and R 0 Theorem 2.1 alsostatesthattheDFEislocallyasymptoticallystablewhen R 0 < 1 ,while R 0 > 1 indicatestheDFEisunstable. 2.4ExtinctionoftheDiseasewhen R 0 1 WenowproveaglobalstabilityresultfortheDFEinthecase R 0 1 Theorem2.2. If R 0 1 ,thenallnonnegativesolutionsof( 21 )-( 26 )convergetothe DFE (0,0,0,0, N 0 V # ) Proof. Suppose limsup t !" I 1 ( t )= m > 0 .Thenforevery > 0 thereexists + 1 > 0 such that I 1 ( t ) m + forall t % + 1 .Substituting,wehavethat I 2 ( t ) ) ( m + ) # ( \$ + ) I 2 ( t ) forall t % + 1 .Thenthereexists + 2 > + 1 suchthat I 2 ( t ) ) ( m + ) \$ + + 25

PAGE 26

forall t % + 2 .Thismeansthat I 1 ( t )+ I 2 ( t ) m +2 + ) ( m + ) \$ + forall t % + 2 .Substitutingagain,wegetthat R ( t ) \$ ) ( m + ) \$ + + # # R ( t ) forall t % + 2 .Thusthereexists + 3 > + 2 suchthat R ( t ) \$ ) ( m + ) \$ + + # + forall t % + 3 .So I 2 ( t )+ R ( t ) ) ( m + )+ \$" +2 forall t % + 3 Wehavethat N ( t )= b (1 # N N 0 ) ,so N ( t ) & N 0 as t &1 .Thusthereexists + 4 > 0 suchthat N 0 # " N ( t ) forall t % + 4 .Wealsohavethat V ( t )= rV (1 # V K ) # V ,and thus V ( t ) & V # as t &1 .Sothereexists + 5 > 0 suchthat V # # " V ( t ) forall t % + 5 Wenowsubstitutethistogetthat V + ( t ) m +2 + # ( m + ) \$ + % N 0 # # V + ( t ) forall t % + 2 + 4 .Thenthereexists + 6 > max { + 2 + 4 } suchthat V + ( t ) \$ m +2 + # ( m + ) \$ + % % ( N 0 # ) + forall t % + 6 .Nowwecanrewrite 26

PAGE 27

I 1 ( t ) S ( t ) N ( t ) & ( m +2 + ( m + ) # + \$ ) + ( N 0 \$ ) ( N 0 \$ )( V \$ ) +(1 # f ) \$ # ( m + )+ \$' % +2 % # ) I 1 ( t ) N ( t ) \$ I 1 ( t ) N ( t ) & ( m +2 + ( m + ) # + \$ ) + ( N 0 \$ ) ( N 0 \$ )( V \$ ) +(1 # f ) \$ # ( m + )+ \$' % +2 % # ) I 1 ( t ) & ( m +2 + ( m + ) # + \$ ) + ( N 0 \$ ) ( N 0 \$ )( V \$ ) +(1 # f ) \$ # ( m + )+ \$' % +2 % # ) I 1 ( t ) # & ( m +2 + ( m + ) # + \$ ) + ( N 0 \$ ) ( N 0 \$ )( V \$ )( N 0 + ) I 1 ( t ) forall t % + 3 + 5 + 6 Thenthereexists + 7 > max { + 3 + 5 + 6 } suchthat I 1 ( t ) & ( m +2 + ( m + ) # + \$ ) + ( N 0 \$ ) ( N 0 \$ )( V \$ ) +(1 # f ) \$ # ( m + )+ \$' % +2 % ) + & ( m +2 + ( m + ) # + \$ ) + ( N 0 \$ ) ( N 0 \$ )( V \$ )( N 0 + ) + forall t % + 7 .RecallfromSection 2.3 thequantity T 0 whosethresholdbehavioris equivalenttothatof R 0 .Nowas & 0 ,theinequalitybecomes I 1 ( t ) % 2 ( m + m # + \$ ) N 0 V +(1 \$ f ) # m # + % 2 ( m + m # + \$ ) N 2 0 V = m # % 2 ( 1 + 1 # + \$ ) N 0 V +(1 \$ f ) # # + m N 0 % 2 ( 1 + 1 # + \$ ) N 0 V # = mT 0 1+ m N 0 ( T 0 \$ (1 \$ f )) = mT 0 1 \$ m N 0 (1 \$ f )+ m N 0 T 0 Notethat m =limsup t !" I 1 ( t ) limsup t !" N ( t )= N 0 ,sotheconstant 1 # m N 0 (1 # f ) isalwayspositive.Thismeansthat mT 0 1 \$ m N 0 (1 \$ f )+ m N 0 T 0 isanincreasingfunctionof T 0 therefore I 1 ( t ) m 1+ f m N 0 < m 27

PAGE 28

since R 0 1 implies T 0 1 .Thus limsup t !" I 1 ( t ) < m ,acontradiction.So m =0 Then limsup t !" I 2 ( t )=limsup t !" R ( t )=limsup t !" V + ( t )=0 followsaswell fromtheinequalitiesobtainedthroughouttheproof.Thismeansthatwemusthave lim t !" S ( t )= N 0 and lim t !" V \$ ( t )= V # .Soallnonnegativesolutionsconvergetothe DFE. InFigure 2-2 ,wepresentanumericalsimulationofsystem( 21 )-( 26 )with f = 0.75 r =1.5 b = N 0 = ) = = V # =1 \$ = =1.5 =0.25 andinitialcondition ( I 1 (0), I 2 (0), R (0), V + (0), S (0), V \$ (0))=(0.5,0.25,0,0.5,0.25,0.5) toshowtheextinction ofthedisease.Withthesehypothesizedparametervalues R 0 4 0.716 ,andweseefrom thesimulationthatthesolutionconvergestothedisease-freeequilibrium. Figure2-2. Simulationofsystem(2 # 1)-(2 # 6)usingMATLABode45solverwith f =0.75 r =1.5 b = N 0 = ) = = V # =1 \$ = =1.5 =0.25 ,and initialcondition ( I 1 (0), I 2 (0), R (0), V + (0), S (0), V \$ (0))=(0.5,0.25,0,0.5,0.25,0.5) 2.5PersistenceoftheDiseasewhen R 0 > 1 Wenowturntothecasewhen R 0 > 1 .Wearenotabletoanalyticallyestablish thelocalstabilityoftheendemicequilibriumbutsimulationssuggestthatitisstable 28

PAGE 29

wheneveritexists.InFigure 2-3 ,wepresentanumericalsimulationofsystem ( 21 )-( 26 )with f =0.5 r =1.5 b = N 0 = \$ = = ) = = = V # =1 andinitialcondition ( I 1 (0), I 2 (0), R (0), V + (0), S (0), V \$ (0))=(0.1,0,0,0,0.9,1) Withthesehypothesizedparametervalues R 0 =1.5 andtheendemicequilibriumis ( I # 1 I # 2 R # V # + S # V # \$ )=(0.267,0.133,0.133,0.286,0.467,0.714) .Weseefromthe simulationthatthesolutiondoesindeedconvergetotheendemicequilibrium. Figure2-3. Simulationofsystem(2 # 1)-(2 # 6)usingMATLABode45solverwith f =0.5 r =1.5 b = N 0 = \$ = = ) = = = V # =1 ,andinitialcondition ( I 1 (0), I 2 (0), R (0), V + (0), S (0), V \$ (0))=(0.1,0,0,0,0.9,1) 2.5.1PersistenceoftheDiseaseforFullSystem Thoughstabilityoftheendemicequilibriumisnotproved,weareabletoprove stronguniformpersistenceofthediseaseinthecase R 0 > 1 .Taking N = N 0 and V = V # weconsiderthelimitingsystemoftheinfectedcomponents: I 1 = 1 # I 1 + I 2 + R N 0 # V + V # # ) I 1 + (1 # f )( I 2 + R ), (210) I 2 = ) I 1 # ( \$ + ) I 2 (211) R = \$ I 2 # R (212) V + = 1 # V + V # # I 1 + I 2 N 0 # V + (213) 29

PAGE 30

withsolutionsrestrictedtothefeasibleregion # = { ( I 1 I 2 R V + ):0 I 1 + I 2 + R N 0 ,0 V + V # } ,apositivelyinvariantsubsetof R 4 + Theorem2.3. Suppose R 0 > 1 .Thenthereexists 0 > 0 suchthat liminf t !" d ( x ( t ), ( R 4 + ) > 0 forallpositivesolutions x ( t ) of( 210 )-( 213 ). Proof. Itsufcestoconsider 0 < f 1 becausetheresultfollowsfromTheorem 2.4 when f =0 .Werstprovethatthereexists 1 > 0 suchthat limsup t !" I 1 ( t )+ I 2 ( t )+ R ( t ) % 1 forallpositivesolutionsof( 210 )-( 213 ).Considerthematrix A ( )= ) * * * + # ) (1 # f ) (1 # f ) & V (1 # N 0 ) ) # ( \$ + ) 0 0 0 \$ # 0 & N 0 (1 # &' V N 0 ) & N 0 (1 # &' V N 0 )0 # for > 0 .Observethatas & 0 A ( ) & J ( x 0 ) where J ( x 0 ) istheupperleftfourbyfour submatrixcontainedintheJacobianmatrixofthesystem( 21 )-( 26 ).Bycontinuityof eigenvalues,theeigenvaluesof A ( ) convergetotheeigenvaluesof J ( x 0 ) .Since R 0 > 1 byTheorem 2.1 thereexistsaneigenvalueof J ( x 0 ) thathaspositiverealpart.Therefore thereexists 1 > 0 suchthat A ( 1 ) hasaneigenvaluewithpositiverealpart. Bywayofcontradiction,suppose X ( t )=( I 1 I 2 R V + )( t ) isapositivesolutionand limsup t !" I 1 ( t )+ I 2 ( t )+ R ( t ) < 1 .Weseefrom( 213 )that limsup t !" V + ( t ) < &' 1 V N 0 Thusthereexists + > 0 suchthatforall t > + I 1 % V # 1 # 1 N 0 # V + # ) I 1 + (1 # f )( I 2 + R ), I 2 % ) I 1 # ( \$ + ) I 2 R % \$ I 2 # R V + % N 0 1 # *" 1 N 0 V # # ( I 1 + I 2 ) # V + 30

PAGE 31

Wedene s = t # + andlet Y ( s )=( i 1 i 2 r v + )( s ) \$ R 4 + bethesolutionof Y = A ( 1 ) Y withinitialcondition Y (0)= X ( + ) .ApplyingKamke'sTheorem(Theorem 1.2 ),we concludethat X ( s ) % Y ( s ) forall s % 0 Let & betheeigenvalueof A ( 1 ) withthelargestrealpart.Since A ( 1 ) isquasi-positive, byPerron-Frobenius(Theorem 1.4 )thereexistsanonnegativeeigenvector v of A ( 1 ) witheigenvalue & .Choose k sufcientlysmallsothat kv Y (0) entrywise.Thenfor Z ( s )= ke ( s v wehave Z = A ( 1 ) Z and Z (0) Y (0) ,whichimpliesthat Z ( s ) Y ( s ) for all s % 0 .Thenthereexistsacomponent Z i ( s ) suchthat lim s !" Z i ( s )= 1 .Therefore lim s !" Y i ( s )= 1 ,whichimpliesthat lim s !" X i ( s )= 1 ,acontradiction.Therefore limsup t !" I 1 ( t )+ I 2 ( t )+ R ( t ) % 1 forallpositivesolutions x ( t ) of( 210 )-( 213 ). Nowlet M bethemaximalinvariantsetin ( R 4 + # .Suppose x \$ M .Let x ( t )= ( I 1 I 2 R V + )( t ) bethesolutionto( 210 )-( 213 )with x (0)= x .Supposethat V + ( t 1 ) > 0 forsome t 1 % 0 .Since x \$ ( R 4 + thereexistsan i \$ { 1,2,3 } suchthat x i ( t 1 )=0 Suppose I 1 ( t )=0 forall t % t 1 .Then ( I 2 + R )= # ( I 2 + R ) sothereexists t 2 > t 1 suchthat ( I 2 + R )( t 2 ) < N 0 .Then I 1 ( t 2 ) > 0 andcontinuity impliesthatthereexists t 3 > t 2 suchthat I 1 ( t 3 ) > 0 .Thenif I 2 ( t 3 )=0 wehave I 2 ( t 3 )= ) I 1 ( t 3 ) > 0 so I 2 ( t ) becomespositive.Similarly, R ( t ) eventuallybecomes positive.Butthenthereexists t 4 > t 3 suchthat x ( t 4 ) \$ R 4 + whichcontradictsthe invarianceof M .So V + ( t )=0 forall t % 0 .Then V + ( t )=0 impliesthat I 1 ( t )= I 2 ( t )=0 forall t % 0 .If f =1 weconcludethat M = { (0,0, R ,0):0 R N 0 } .If f < 1 Equation ( 210 )impliesfurtherthat R ( t )=0 forall t % 0 andhence M istheorigin.Weshow nextthat M isauniformrepellerineithercase. Firstlet f < 1 and < 1 3 .Let x ( t ) beanonnegativenonzerosolutionof ( 210 )-( 213 )with || x (0) || < .Bytheargumentaboveif x (0) \$ ( R 4 + thesolution eventuallybecomespositivesoitsufcestoconsiderpositivesolutionsonly.Then 31

PAGE 32

limsup t !" I 1 ( t )+ I 2 ( t )+ R ( t ) % 1 impliesthatthereexists T > 0 suchthat || x ( T ) || > ByTheorem1ofFonda[ 14 ]thereexists > 0 suchthat liminf t !" d ( x ( t ),0) % forall nonnegativenonzerosolutions x ( t ) Nowlet f =1 .Supposethat limsup t !" V + ( t ) " 2 forsomepositivesolution x ( t ) and 2 > 0 .Thenthereexists + > 0 suchthatfor t > + and y =( I 1 ( t ), I 2 ( t ), R ( t )) T y " 2 + Ay where = V # ,0,0 # and A = ) * * + # ) 00 ) # ( \$ + )0 0 \$ # . Clearlytheprincipaleigenvalueof A isnegative;byPerron-Frobenius(Theorem 1.4 ) thereexists & > 0 andapositivelefteigenvector v suchthat vA = # & v .Dene \$ ( y )= v y .Then \$ ( y )= v y " 2 ( v )+ vAy = 2 ( v ) # & \$ ( x ), fromwhichitfollowsthat limsup t !" \$ ( y ) " 2 ( v ) & andhencethereexists C > 0 suchthat limsup t !" I 1 ( t )+ I 2 ( t )+ R ( t ) C 2 .If 2 < 1 C wegetacontradiction.Therefore limsup t !" V + ( t ) % 1 C Toshow M isalsoauniformrepellerinthiscase,suppose < 1 2 C and x ( t ) isa solutionsuchthat x (0) \$ # \ M and d ( x (0), M ) < .If x (0) \$ ( R 4 + thenthereexists i \$ { 1,2,4 } suchthat x i (0) > 0 .Wewillshowthat x ( t ) \$ R 4 + forsome t > 0 .Suppose I 1 (0) > 0 .Then I 2 (0), V + (0) > 0 ,sothereexists t > 0 suchthat I 2 ( t ), V + ( t ) > 0 .Then R ( t ) > 0 ,sobycontinuitythereexists t 1 > t suchthat R ( t 1 ) > 0 .Then x ( t 1 ) \$ R 4 + Similarly,ifwesupposeeither I 2 (0) > 0 or V + (0) > 0 ,thereexists t > 0 suchthat x ( t ) \$ R 4 + .Soitsufcestoconsiderpositivesolutions.Now limsup t !" V + ( t ) % 1 C 32

PAGE 33

impliesthatthereexists T > 0 suchthat d ( x ( T ), M ) % || V + ( T ) || % 1 2 C > .ByTheorem 1.1 thereexists > 0 suchthat liminf t !" d ( x ( t ), M ) % forallsuchsolutions,in particularforpositivesolutions. Thereforefor 0 < f 1 thestablemanifoldofMdoesnotintersect R 4 + .Thusby Theorem4.3in[ 15 ]theowisuniformlystronglypersistent;thatis,thereexists 0 > 0 suchthat liminf t !" d ( x ( t ), ( R 4 + ) > 0 forallpositivesolutions x ( t ) of( 210 )-( 213 ). 2.5.2StabilityoftheEndemicEquilibriuminaSpecialCase Inthecasewhere f =0 thesystemismoretractable.Inthelimitingcase N = N 0 and V = V # ,theequationsofthemodelaregivenby: I 1 = S N 0 V + V # # ) I 1 + ( I 2 + R ), (214) I 2 = ) I 1 # ( \$ + ) I 2 (215) R = \$ I 2 # R (216) V + = V \$ V # I 1 + I 2 N 0 # V + (217) S = # S N 0 V + V # (218) V \$ = V # # V \$ V # I 1 + I 2 N 0 # V \$ (219) Thissystemhastheuniqueendemicequilibrium ( I # 1 I # 2 R # V # + S # V # \$ )= N 0 ) + )' N 0 ( \$ + )( ) + ) \$) N 0 ( \$ + )( ) + ) V # + ,0, V # \$ # where V # + satises V # + = & N 0 ( I # 1 + I # 2 ) + & N 0 V ( I # 1 + I # 2 ) < V # and V # \$ = V # # V # + > 0 .Inthiscaseweareabletoprovethefollowingtheorem concerningglobalasymptoticstabilityoftheendemicequilibrium. Theorem2.4. Allpositivesolutionsofthesystem( 214 )-( 219 )with S (0) < N 0 convergetotheendemicequilibrium ( I # 1 I # 2 R # V # + S # V # \$ ) 33

PAGE 34

Proof. Firstweobservethat S ( t ) isdecreasingandboundedbelowby0.So S ( t ) & S as t &1 forsome 0 S < N 0 .Nowlet M ( t )= I 1 ( t )+ I 2 ( t )+ R ( t ) .Sincethetotal populationis N 0 wehavethat M ( t ) & M as t &1 for M = N 0 # S > 0 .Hence I 1 ( t )= M + h ( t ) # I 2 ( t ) # R ( t ) where h ( t ) & 0 as t &1 .Thenweconsiderthe subsystem: I 2 = ) ( M + h ( t ) # I 2 # R ) # ( \$ + ) I 2 (220) R = \$ I 2 # R (221) Letting I 2 = #% ( \$ + % )( # + % ) M and R = \$# ( \$ + % )( # + % ) M ,weobservethat 0= ) ( M # I 2 # R ) # ( \$ + ) I 2 0= \$ I 2 # R Wethenperformashiftbydening i 2 = I 2 # I 2 and r = R # R .Then( 220 )-( 221 )can bewrittenas x = Ax + f ( t ) where x = ) + i 2 r , A = ) + # ( ) + \$ + ) # ) \$ # , f ( t )= ) + ) h ( t ) 0 . Usingvariationofparameters,wendthat x ( t )= t 0 e ( t \$ s ) A f ( s ) ds + e tA x 0 Notethatthematrix A isHurwitzsincebotheigenvalueshavenegativerealparts,so e tA & 0 as t &1 .Thisalsomeansthatthereexist C \$ > 0 suchthat | e ( t \$ s ) A f ( s ) | Ce \$ \$ ( t \$ s ) | f ( s ) | forall t % s % 0 .Let > 0 begiven.Since f ( t ) & 0 ,thereexists + > 0 suchthat | f ( t ) | < '\$ 2 C forall t % + .Forthisxed + ) 0 | e ( t \$ s ) A f ( s ) | ds e \$ t ) 0 Ce \$ s | f ( s ) | ds .Sothereexists T % + suchthat e \$ \$ t ) 0 Ce \$ s | f ( s ) | ds < 2 forall 34

PAGE 35

t % T .So / / / / t 0 e ( t \$ s ) A f ( s ) ds / / / / t 0 Ce \$ \$ ( t \$ s ) | f ( s ) | ds = ) 0 Ce \$ \$ ( t \$ s ) | f ( s ) | ds + t ) Ce \$ \$ ( t \$ s ) | f ( s ) | ds Ce \$ \$ t ) 0 e \$ s | f ( s ) | ds + "\$ 2 t ) e \$ \$ ( t \$ s ) ds < 2 + 2 (1 # e \$ \$ ( t \$ ) ) ) forall t % T < Thus x ( t ) & 0 as t &1 .Hence I 2 & I 2 and R & R as t &1 .Therefore I 1 & % # + % M whichwedeneas I 1 .Now I 1 + I 2 > 0 andwehave I 1 ( t )+ I 2 ( t )= I 1 + I 2 + g ( t ) forsome g ( t ) & 0 as t &1 .Wecanthenexpress( 217 )as V + = N 0 ( I 1 + I 2 + g ( t )) # N 0 V # ( I 1 + I 2 + g ( t ))+ # V + = g 1 ( t ) # g 2 ( t ) V + wherethefunctions g 1 ( t ) and g 2 ( t ) approachthepositivelimits & N 0 ( I 1 + I 2 ) and & N 0 V ( I 1 + I 2 )+ ,respectively,as t &1 .Itthenfollowsthat V + approachesthepositivelimit V + = & N 0 ( I 1 + I 2 ) + & N 0 V ( I 1 + I 2 ) Hence V + > 0 andsothereexists + > 0 and > 0 suchthatforall t > + S "# N 0 V # S andtherefore S ( t ) S ( + ) e %& N 0 V ( t \$ ) ) wheretherightsideconvergesto0as t &1 .Therefore S =0 ,whichimpliesthat M = N 0 .Thus I 1 = I # 1 I 2 = I # 2 R = R # ,and V + = V # + 35

PAGE 36

Thuswehaveprovedthatinasimpliedspecialcase,theendemicequilibriumis globallyasymptoticallystable. 36

PAGE 37

CHAPTER3 ROGUINGASACONTROLSTRATEGY Themodelwehavedevelopedincorporatesacontrolstrategyofroguingand replantinginfectedtrees.Modelsforotherplantvirusdiseaseshaveincludedroguing, suchasthoseforbananabunchytop[ 1 ]andcitrustristezavirus[ 13 ].Chanand Jeger[ 8 ]developedamodelincludinghealthy,latentlyinfected,infectiousand post-infectiousplantsandinvestigatedthepopulationdynamicswithandwithout roguing.Analysisoftheirmodelwithroguingshowedthatthebasicreproductivenumber andtheequilibriumhealthypopulationdidnotdependonwhetherroguingwasdone inthepost-infectiouscategory.Howdodifferentroguingmethodsaffectourmodel? Toanswerthisweconsideracontrolmethodwhereroguingofthe I 2 and R treesis doneatrates 1 and 2 ,respectively.Thecasewhere 2 =0 isanalyzedinSection 3.1 whereweconcludethatthediseaseistransientandapositivepopulationofhealthy treeswillremainindenitely.Whilewedon'thaveacompleteanalyticalunderstanding ofthemodelwhen 1 > 0 and 2 =0 ,simulationssuggestthatthelevelofremaining healthytreesislargerwhenroguingisperformedin I 2 thanwhenthereisnoroguingat all.Section 3.2 focusesonthecasewhere 2 > 0 .Inthiscasethediseaseisnolonger transientandanendemicequilibriumwillexistif R 0 > 1 .However,wewillshowthatthe equilibriumhealthypopulationoftreesisincreasingwithrespecttoboth 1 and 2 3.1TransientBehavior Inthemodelwehavebeenconsideringthusfar,wehaveaccountedforroguing ofsymptomaticaswellasdeadtrees.However,itisworthwhiletoexaminethecases whenonlyroguingofsymptomatictreesoccursorwhenthereisnoroguingatall.We showthatineitherofthesesituationsthediseaseistransient;thatis, lim t !" I 1 ( t )=lim t !" I 2 ( t )=lim t !" V + ( t )=0. 37

PAGE 38

Thesystemthatweanalyzenowisthesameasouroriginalsystem,exceptthatitdoes notincludearoguingtermintheequationfor R ,andthusalsodoesnotaccountforthe replantingofarogueddeadtreeineitherthe S or I 1 equations.Weallowforroguingof I 2 treesatrate ,whichcanbezero.Notethatsincedeadtreesarenolongerroguedand replantedpossiblyasinfectedtrees,the R compartmentisnotconsideredtobeinfected. I 1 = S N V + V # ) I 1 + (1 # f ) I 2 (31) I 2 = ) I 1 # ( \$ + ) I 2 (32) R = \$ I 2 (33) V + = V \$ V I 1 + I 2 N # V + (34) S = b 1 # N N 0 # # S N V + V + fI 2 (35) V \$ = r ( V \$ + V + ) 1 # V \$ + V + K # # V \$ V I 1 + I 2 N # V \$ (36) Theorem3.1. Forallnonnegativesolutionsofthesystem( 31 )-( 36 )andforall % 0 thediseaseistransient. Proof. Forthissystem lim t !" N ( t )= N 0 and lim t !" V ( t )= V # asshownfortheoriginal systemintheproofofTheorem 2.2 .Herewehave R = \$ I 2 .Since I 2 % 0 and I 2 ( t ) isbounded,anapplicationofBarbalat'slemma[ 16 ]impliesthat lim t !" I 2 =0 sothat R doesnotincreasewithoutbound.Thissimilarlyforces lim t !" I 1 =0 asaresultof Equation( 32 ).Thus lim t !" V + =0 aswellbecauseofEquation( 34 ).Soallinfected compartmentseventuallybecomeextinct. Theorem3.2. Inthesystem( 31 )-( 36 ), lim t !" S = S > 0 forany % 0 Proof. IntegratingEquation( 33 ),wehavethat R ( t ) # R (0)= \$ t 0 I 2 ( + ) d + .Since R (0) and R ( t ) arebothbounded,theirdifferenceisbounded,andthus t 0 I 2 ( + ) d + is bounded.Similarly,Equation( 32 )givesthat I 2 ( t ) # I 2 (0)+( \$ + ) t 0 I 2 ( + ) d + = 38

PAGE 39

) t 0 I 1 ( + ) d + ,andthus t 0 I 1 ( + ) d + isboundedaswell.Let > 0 .Since N ( t ) N 0 forall t lim t !" N ( t )= N 0 ,and lim t !" V ( t )= V # ,thereexists T > 0 suchthat N 0 # " N ( t ) N 0 and V # # " V ( t ) V # + forall t % T .Since % 0 byEquation ( 35 )wehavetheinequality S %# S N V + V %# S ( N 0 # ) V + ( V # # ) Therefore S ( t ) % S (0)exp # & ( N 0 \$ )( V \$ ) t 0 V + ( + ) d + # forall t % T .Bywayof contradiction,assumethat S =0 .Thenbytheaboveinequality, lim t !" t 0 V + ( + ) d + = 1 .Notethat V V I 1 + I 2 N V V I 1 + I 2 N 0 \$ forall t % T .Thenforall t % T t 0 V \$ ( + ) V ( + ) I 1 ( + )+ I 2 ( + ) N ( + ) d + T 0 V \$ ( + ) V ( + ) ( I 1 ( + )+ I 2 ( + )) N ( + ) d + + t T ( I 1 ( + )+ I 2 ( + )) N 0 # d + T + N 0 # t T ( I 1 ( + )+ I 2 ( + ) ) d + T + N 0 # t 0 ( I 1 ( + )+ I 2 ( + ) ) d + whichisbounded.ByEquation( 34 ),wehave V + ( t ) # V + (0)= t 0 V ( ) ) V I 1 ( ) )+ I 2 ( ) ) N d + # t 0 V + ( + ) d + .Sincetherstintegralisboundedbythepreviousinequality,itfollows that lim t !" V + ( t )= #1 ,acontradiction.Thus S > 0 Theorems 3.1 and 3.2 tellusthatwhenroguingofdeadtreesdoesnotoccur, thediseasewilleventuallydieoutleavingonlyhealthyanddeadtrees.Thepresence ofremaininghealthytreesisafeaturesimilartothatofthestandardSIRepidemic model.Thisleadstoaquestionaboutthebenetsofroguingandreplanting.Isit worthwhiletorogueinfected,symptomatictrees,orshouldnotreesberogued?In eithercasethediseaseistransient;thedifferenceisthenumberoftreesthatremain healthy.Weusesimulationstocomparethevaluesof S ,thenumberofunaffected trees,ineachsituation.Considerthehypothesizedparametervalues N 0 = b = = \$ = ) = =1 and r =1.5 inbothcases,with =1 and f =0.5 forthesystem 39

PAGE 40

includingroguingand =0 forthesystemexcludingroguing.Webeginwithonly susceptibletreesandasmallnumberofinfectedpsyllids;theinitialconditionsare ( I 1 (0), I 2 (0), R (0), V + (0), S (0), V \$ (0))=(0,0,0,0.1,1,0.9) .Thisresultsin S = 0.2750 withroguingand S =0.2192 withoutroguing,somoretreesareunaffected whenroguingoccurs.Withallcombinationsofparametervaluesthatweexplored,the simulationssupportthenotionthatmorehealthytreeswillremainwhensymptomatic treesareroguedthanwhennotreesarerogued. A B Figure3-1. Simulationsofsystem(3 # 1)-(3 # 6)usingMATLABode45solverwith r =1.5 N 0 = b = = \$ = ) = = V # =1 ,initialcondition ( I 1 (0), I 2 (0), R (0), V + (0), S (0), V \$ (0))=(0,0,0,0.1,1,0.9) ,and(A): =1, f =0.5 ,(B): =0 3.2DifferentRoguingRates Nowweconsider 1 > 0 and 2 > 0 ,inwhichcasethediseaseisnolongertransient andanendemicequilibriumwillexistwhen R 0 > 1 .Substituting 1 and 2 appropriately insystem( 21 )-( 26 ),thereproductivenumberisrecalculatedtobe: R 0 = 1 # f 2 + & (1 # f ) 2 4 + 2 V # N 0 1 ) + 1 \$ + 1 # (37) Forourmodel,asforChanandJeger'smodel[ 8 ], R 0 doesnotdependontherateof roguingofthedeadtrees.Althoughthediseasewillbemaintainedinthecasewhen R 0 > 1 ,itmaybethat S # ,theendemicequilibriumhealthytreepopulation,ishighereven 40

PAGE 41

inthepresenceofdiseaseascomparedtothecasewithnoroguingwherethedisease diesout.Thatis,theremaybeatradeoffbetweenallowingthediseasetopersistand maintainingaprotablelevelofhealthytrees.Thereforeitisworthconsideringthe dependenceof S # on 1 and 2 .Indeed, S # isdeterminedtobe: S # = fN 2 0 ) + 1+ V & \$ 1+ \$# % 2 ( \$ + % 1 + # ) % fN 0 & + 1 # + 1 \$ + % 1 \$ 1+ \$ % 2 % . UnliketheresultofChanandJegerweseethattheequilibriumhealthypopulation dependsonwhetherroguingisdoneinboththeinfectious,symptomaticstageandinthe post-infectiousstage.Fromthisexpressionwecandeterminethat S # increasesas 1 and 2 increase,asshownbelow. Proposition3.1. Let R 0 > 1 .Thentheendemicequilibriumlevelofhealthytrees S # = fN 2 0 ) + 1+ V & \$ 1+ \$# % 2 ( \$ + % 1 + # ) % fN 0 & + 1 # + 1 \$ + % 1 \$ 1+ \$ % 2 % isincreasingwithrespectto 1 and 2 Proof. Werstprovethat S # increasesas 1 increases.Wehave S # = fN 2 0 ) + 1+ V & + V \$# &% 2 ( \$ + % 1 + # ) fN 0 & + 1 # + 1 \$ + % 1 \$ 1+ \$ % 2 % = fN 2 0 0 a + b x + # c + d x 1 = fN 2 0 ax 2 +( a ) + b ) x cx 2 +( c ) + d ) x + d ) ( where x = \$ + 1 a =1+ V & b = V \$# &% 2 c = fN 0 & + 1 # ,and d =1+ \$ % 2 .Notethat x a b c and d arepositivefor 1 % 0 .Afterdifferentiatingandsomesimplicationwedetermine ( S # (' 1 = fN 2 0 ( ad # bc ) x 2 +2 ad ) x + d ) ( a ) + b ) ( cx 2 +( c ) + d ) x + d ) ) 2 ( % fN 2 0 x [( ad # bc ) x +2 ad ) ] ( cx 2 +( c ) + d ) x + d ) ) 2 ( Thereforeitsufcestoshowthat ( ad # bc ) x +2 ad ) > 0 .Recallingtheexpressionfor R 0 (Quantity 37 )weobservethat,similartotheproofofTheorem1, R 0 > 1 ifandonlyif 2 \$ 1+ # \$ + % 1 % V # N 0 f ) > 1. 41

PAGE 42

Usingthisinequalitywecalculate ( ad # bc ) x +2 ad ) = 1+ V # + \$ 2 # V # N 0 f )\$ 2 2 ( ( \$ + 1 ) +2 ) 1+ V # #" 1+ \$ 2 # > 1+ V # + \$ 2 # \$ 2 1+ ) \$ + 1 #( ( \$ + 1 ) +2 ) 1+ V # + \$ 2 + V # \$ *' 2 # > 1+ V # # \$) 2 ( \$ + 1 ) ( ( \$ + 1 )+ 2 \$) 2 = 1+ V # # ( \$ + 1 )+ \$) 2 > 0. Itisprovedsimilarlythat S # isincreasingwithrespectto 2 .Indeed,wehave S # = fN 2 0 0 1+ V & + V \$# & ( \$ + % 1 + # ) % 2 fN 0 & + 1 # + 1 \$ + % 1 + \$ ( \$ + % 1 ) % 2 1 = fN 2 0 0 g + h % 2 j + k % 2 1 = fN 2 0 g 2 + h j 2 + k ( where g =1+ V # h = V # \$) ( \$ + 1 + ) ) j = fN 0 + 1 ) + 1 \$ + 1 and k = \$ \$ + 1 Notethat g h j ,and k arepositive.Hence ( S # (' 2 = fN 2 0 gk # hj ( j 2 + k ) 2 # anditsufcestoshowthat gk # hj > 0 .Since R 0 > 1 weusetheinequalityaboveagain toobtain gk # hj = 1+ V # # \$ \$ + 1 # V # \$) ( \$ + 1 + ) ) fN 0 + \$ + 1 + ) ) ( \$ + 1 ) # = \$ \$ + 1 # V # \$ N 0 f ) 2 ( \$ + 1 + ) ) = \$ \$ + 1 2 3 1 # V # N 0 f ) 2 \$ 1+ # \$ + % 1 % 4 5 > 0. 42

PAGE 43

Theseresultsforvaryingroguingmethodsdependonthenovelfeatureofour modelwhichistheinclusionofthepositiveprobability 1 # f thatareplantedtreewill immediatelybecomeinfected.Thatis,thetreesinthe R stageareapotentialsourceof infectionduetothesoilandremainingrootsystembeingareservoirforthedisease. 43

PAGE 44

CHAPTER4 MODIFICATIONOFTHEMODEL Informulatingtheoriginalmodel,weassumedthatpsyllidsareequallyattracted toalltypesoftrees,includingthedeadones,whichmaybebiologicallyunrealistic.We nowconsideramodiedversionofourmodelinwhichweassumethatthepsyllidsare notattractedtodeadtrees.Thatis,theprobabilitythatatreebittenbyavectorisa susceptibletreeisdenedtobe = S S + I 1 + I 2 Thenewmodelisthengivenby: I 1 = *! V + V # ) I 1 + (1 # f )( I 2 + R ), (41) I 2 = ) I 1 # ( \$ + ) I 2 (42) R = \$ I 2 # R (43) V + = V \$ V (1 # ) # V + (44) S = b 1 # N N 0 # # *! V + V + f ( I 2 + R ), (45) V \$ = r ( V \$ + V + ) 1 # V \$ + V + K # # V \$ V (1 # ) # V \$ (46) whichwestudyawayfromthesingularity S = I 1 = I 2 =0 .Thesystem( 41 )-( 46 )has thesameDFEasouroriginalsystem( 21 )-( 26 ).Italsoholdsthat N = N 0 and V = V # attheequilibriumof( 41 )-( 46 ).Howeverinthismodiedsystemwehavethepossibility fortwopositiveendemicequilibria.Wedenethefollowingquadraticfunction: g ( )= )\$ ( \$ + ) 2 # + ) + ) fN 0 # + ) N 0 f ( + V # ) 2 whereeasycalculationshowsthat g ( # )=0 foranyendemicequilibriumvalue # InTheorem 4.1 ,weprovetheexistenceofthetwopositiveequilibriagivencertain parametervalues.Additionally,computationrevealsthatthenextgenerationmatrix of( 41 )-( 46 )isidenticaltothenextgenerationmatrixof( 21 )-( 26 )calculatedin 44

PAGE 45

Subsection 2.3.1 .Thereforeforthemodiedsystemwealsohave R 0 = 1 # f 2 + & (1 # f ) 2 4 + 2 V # N 0 1 ) + 1 \$ + # Itiseasytoshowthat R 0 > 1 when f =0 andthereexistsaunique,globallyattractive endemicequilibrium;theproofissimilartothatofTheorem 2.4 .However,inthecase f > 0 ,wehavethefollowingresult.ForuseinTheorem 4.1 wedene crit = \$ + 2 \$ + ) ) + f N 0 # Theorem4.1. Twodistinctpositiveendemicequilibriaof( 41 )-( 46 )existifandonlyif 1. crit < 1 and 2. 2 f N 0 1 ) + 1 \$ + # < V # < crit 1 # 2 + ( + ) ) 2 ') fN 0 # # 1 ( Furthermore, 2 impliesthat R 0 < 1 Proof. For \$ (0,1) wehavethat isanequilibriumvalueifandonlyif g ( )=0 .Soit sufcestoshowthat g hastworootsin (0,1) ifandonlyif(i)and(ii)hold.Observethat g (0) > 0 andthatthecriticalvalueof g ( ) isgivenby crit .Then g willhavetworootsin (0,1) ifandonlyif g (1) > 0 0 < crit < 1 ,and g ( crit ) < 0 .Itistrivialthat 0 < crit .We have g (1)= ) N 0 V # f # 1+ ) \$ + # Then g (1) > 0 ifandonlyif 2 f N 0 1 ) + 1 \$ + # < V # asincondition(ii).Notethatitistrivialfromthisinequalitythat(ii)implies R 0 < 1 .We provenowthat g ( crit ) < 0 ifandonlyiftheright-handinequalityof(ii)holds.Indeed, g ( crit ) < 0 2 ) fN 0 ( V # + ) 2 < ) fN 0 + + ) # crit # )\$ ( \$ + ) 2 crit 45

PAGE 46

2 V # < 2 ) fN 0 '" ) fN 0 + + ) # crit # )\$ ( \$ + ) 2 crit ( # 2 V # < 2 ) fN 0 crit ) fN 0 1 # 2 # + + ) 2 ( # 2 V # < crit 1 # 2 + ( + ) ) 2 ') fN 0 # # 1 ( Wenextexaminethestabilityoftheequilibriainthecasethatmultiplepositive endemicequilibriaexist,whichiswhen R 0 < 1 .Considerthehypothesizedparameter values f = =0.1 =0.01 r =1.5 ,and b = N 0 = \$ = ) = = V # =1 ,whichresult inthevalue R 0 =0.9199 .WendthattheDFEaswellasoneoftheendemicequilibria willbestable,asalleigenvaluesoftheJacobianevaluatedateachoftheseequilbria havenegativerealparts.However,theotherendemicequilibriumisunstable;forthis particularequilibrium,theJacobianhasapositiverealeigenvalue. Simulationsshowthat,dependingontheinitialconditionsusedincombinationwith theaboveparametervalues,thesolutions ( I 1 ( t ), I 2 ( t ), R ( t ), V + ( t ), S ( t ), V \$ ( t )) will convergetooneofthetwostableequilibria(Figure 4-1 ).First,withtheinitialcondition (0,0,0.5,0,0.5,1) ,convergenceistotheDFE (0,0,0,0,1,1) .Ifinsteadweusethe initialcondition (0,0,0.9,0,0.1,1) thesolutionconvergestotheendemicequilibriaat (0.01,0.01,0.977,0.081,0.003,0.919) 46

PAGE 47

A B Figure4-1. Simulationsofsystem(4 # 6)-(4 # 6)usingMATLABode45solverwith f = =0.1 =0.01 r =1.5 b = N 0 = \$ = ) = = V # =1 ,andinitial conditions(A): ( I 1 (0), I 2 (0), R (0), V + (0), S (0), V \$ (0))=(0,0,0.5,0,0.5,1) (B): ( I 1 (0), I 2 (0), R (0), V + (0), S (0), V \$ (0))=(0,0,0.9,0,0.1,1) 47

PAGE 48

CHAPTER5 PERIODICITY 5.1PeriodicModel Recentevidencesuggeststhattransmissionofthediseasemaynotoccurata constantrateasoriginallyassumed.Itisbelievedthatpsyllidsprefertolayeggsand feeduponnewush,whichonlyoccurstwiceayear[ 20 ].Duringtheoffseason,it ispossiblethatthepsyllidsstillfeedontheleavesofcitrusplantsbutperhapsdon't transmitthediseaseaswell,orthattheyfeedandbreedonorangejasmineinstead. Thuswealterouroriginalmodeltoincorporateacontinuousperiodicbitingfunction ( t ) torepresentthisuctuatingtransmission.Thetimeaveragefor ( t ) overtheperiod T is representedby 5 6 = 1 T T 0 ( t ) dt Theincreasedcomplexityofthismodelcanprovideamorerealisticrepresentation ofthedisease,butcanalsobedifculttofullyanalyze.Withthisperiodicity,ourmodel nowbecomes: I 1 = ( t ) S N V + V # ) I 1 + (1 # f )( I 2 + R ), (51) I 2 = ) I 1 # ( \$ + ) I 2 (52) R = \$ I 2 # R (53) V + = ( t ) V \$ V I 1 + I 2 N # V + (54) S = b 1 # N N 0 # # ( t ) S N V + V + f ( I 2 + R ), (55) V \$ = rV 1 # V K # # ( t ) V \$ V I 1 + I 2 N # V \$ (56) Forthenon-periodicsystem( 21 )-( 26 ),wewereabletoperformmuchofour analysisusingthebasicreproductivenumber R 0 .Wewerenotabletodene R 0 explicitlyforthisperiodicsystem,butwearestillabletondsomeconditionsthat willguaranteeextinctionorpersistenceofthediseaseincertaincases,aswellas stabilityandinstabilityofthedisease-freesolution. 48

PAGE 49

5.2ExtinctionoftheDiseaseinPeriodicModel Werstexaminethepossibilitythatthediseasebecomesextinct.Asinthe non-periodicsystem,thetotaltreepopulationagainsatises N = b 1 # N N 0 # whilethetotalvectorpopulationsatises V = rV 1 # V K # # V So N ( t ) & N 0 and V ( t ) & V # =(1 # r ) K (assuming < r )as t & + 1 .The disease-freesolution(DFE)fortheperiodicsystem( 51 )-( 56 )istheconstantsolution x 0 =(0,0,0,0, N 0 V # ) .Weprovethatthediseasecanbecomeextinct. Theorem5.1. Allnonnegativesolutionsofthesystem( 51 )-( 56 )convergetotheDFE (0,0,0,0, N 0 V # ) ifthefollowingconditionholds: ( \$ + + ) ) T 2 e 2 T ( + # ) 5 6 2 ( \$ + ) N 0 V # f ( e T # 1)( e # T # 1) < 1. Proof. Suppose limsup t !" I 1 ( t )= m > 0 .Thenforevery > 0 thereexists + 1 > 0 such that I 1 ( t ) m + forall t % + 1 .Substituting,wehavethat I 2 ( t ) ) ( m + ) # ( \$ + ) I 2 ( t ) forall t % + 1 .Thenthereexists + 2 > + 1 suchthat I 2 ( t ) ) ( m + ) \$ + + forall t % + 2 .Thismeansthat I 1 ( t )+ I 2 ( t ) m +2 + ) ( m + ) \$ + 49

PAGE 50

forall t % + 2 .Substitutingagain,wegetthat R ( t ) \$ ) ( m + ) \$ + + # # R ( t ) forall t % + 2 .Thusthereexists + 3 > + 2 suchthat R ( t ) \$ ) ( m + ) \$ + + # + forall t % + 3 .So I 2 ( t )+ R ( t ) ) ( m + )+ \$" +2 forall t % + 3 Wehavethat N ( t )= b (1 # N N 0 ) ,so N ( t ) & N 0 as t &1 .Thusthereexists + 4 > 0 suchthat N 0 # " N ( t ) forall t % + 4 .Wealsohavethat V ( t )= rV (1 # V K ) # V ,and thus V ( t ) & V # as t &1 .Sothereexists + 5 > 0 suchthat V # # " V ( t ) forall t % + 5 Wenowsubstitutethistogetthat,forall t % + 2 + 4 V + ( t ) ( t ) m +2 + # ( m + ) \$ + % N 0 # # V + ( t ). Thenbyvariationofparameterswehave V + ( t ) t 0 ( s ) Le ( s \$ t ) ds + V + (0) e \$ t where,as & 0 L = m +2 + # ( m + ) \$ + % N 0 # & ( \$ + + ) ) m ( \$ + ) N 0 Sinceweareinterestedinthisquantityas t &1 ,wehave V + ( t ) L t 0 ( s ) e ( s \$ t ) ds Forany t ,thereexistsaninteger K suchthat t \$ ( KT ,( K +1) T ] .Sowecanset t =( K +1) T # / forsome / \$ [0, T ) .Then 50

PAGE 51

V + ( t ) Le \$ t t 0 ( s ) e s ds Le \$ t ( K +1) T 0 ( s ) e s ds = Le \$ t 0 T 0 ( s ) e s ds + 2 T T ( s ) e s ds +...+ ( K +1) T KT ( s ) e s ds 1 = Le \$ t T 0 ( s ) e s ds + e T T 0 ( s ) e s ds +...+ e KT T 0 ( s ) e s ds ( = Le \$ t \$ 1+ e T +( e T ) 2 +...+( e T ) K % T 0 ( s ) e s ds = Le \$ t e ( K +1) T # 1 e T # 1 # T 0 ( s ) e s ds bysumofageometricseries = Le 1 # e \$ ( K +1) T e T # 1 # T 0 ( s ) e s ds since e \$ t = e \$ ( K +1) T e Le T 1 e T # 1 e T T 0 ( s ) ds = Le 2 T e T # 1 T < ( t ) > as t &1 Turningtotheequationfor I 1 andsubstitutingseveralpreviousinequalitieswehave that,forall t % + 3 + 5 I 1 ( t ) ( t ) V + ( t ) V # # + M # ) I 1 ( t ) where,as & 0 M = ) ( m + )+ \$" +2 # (1 # f ) & ) m (1 # f ). Usingvariationofparametersagaingivesusthat I 1 ( t ) t 0 ( s ) V + ( s ) V # # + M # e # ( s \$ t ) ds + I 1 (0) e \$ # t Sinceweareinterestedinthisquantityas t &1 ,wesubstituteourboundfor V + ( t ) and have 51

PAGE 52

I 1 ( t ) LTe 2 T < ( t ) > ( V # # )( e T # 1) t 0 ( s ) e # ( s \$ t ) ds + M t 0 e # ( s \$ t ) ds LTe 2 T < ( t ) > ( V # # )( e T # 1) T < ( t ) > e 2 # T e # T # 1 # + M ) (1 # e \$ # t ) (similartoabove) LT 2 e 2 T ( + # ) < ( t ) > 2 ( V # # )( e T # 1)( e # T # 1) + M ) as t &1 Nowas & 0 ,thefollowinginequalityholdsas t &1 : I 1 ( t ) ( \$ + + ) ) mT 2 e 2 T ( + # ) ( \$ + ) N 0 V # ( e T # 1)( e # T # 1) + m (1 # f ). Theaboveinequalitywillyield I 1 ( t ) < m when ( \$ + + ) ) T 2 e 2 T ( + # ) 5 6 2 ( \$ + ) N 0 V # f ( e T # 1)( e # T # 1) < 1. (57) Thatis,whenCondition 57 holds,thesuppositionthat limsup t !" I 1 ( t )= m > 0 is contradicted.ThustheDFEisgloballystablewhenCondition 57 issatisedandthe diseasebecomesextinct. Notethatas T & 0 ,theleft-handsideofCondition 57 convergestothequantity ( \$ + + ) ) 5 6 2 ( \$ + ) N 0 V # f ) Thisquantityislessthanoneifandonlyif T 0 (Quantity 29 ofthenon-periodicsystem) islessthanone.Weprovedasimilarresultinthatcasethatextinctionofthedisease occurswhen T 0 1 InFigure 5-1 ,wepresentanumericalsimulationofsystem( 51 )-( 56 )with f = 0.75 r =1.5 b = N 0 = ) = = V # =1 \$ = =1.5 T =0.25 ( t )= + sin( 2 + t T ) with =0.25 (so 5 6 =0.25 )andinitialcondition ( I 1 (0), I 2 (0), R (0), V + (0), S (0), V \$ (0))= (0.5,0.25,0,0.5,0.25,0.5) toshowtheextinctionofthedisease.Withthesehypothesized parametervaluesCondition 57 issatised,andweseefromthesimulationthatthe solutionindeedconvergestothedisease-freestate.Theparametervaluesusedhere 52

PAGE 53

arethesameasusedinthesimulationinFigure 2-2 ,whichshowstheextinctionofthe diseaseforthenon-periodicsystem( 21 )-( 26 ). Figure5-1. Simulationofsystem(5 # 1)-(5 # 6)usingMATLABode45solverwith f =0.75 r =1.5 b = N 0 = ) = = V # =1 \$ = =1.5 T =0.25 ( t )= + sin( 2 + t T ) with =0.25 ,andinitialcondition ( I 1 (0), I 2 (0), R (0), V + (0), S (0), V \$ (0))=(0.5,0.25,0,0.5,0.25,0.5) 5.3IncorporatingthePeriodicFunctionasaSmallPerturbation Werstrepresenttheseasonaluctuationofthediseasetransmissionwiththe continuousperiodicperturbation ( t )= + f ( t ) ,where > 0 issmalland f ( t ) isa T -periodicfunction.Sincethisfunctionrepresentsthetransmissionrate,weassume that ( t ) % 0 andthatitstillincorporatestheprobabilitythatabitesuccessfullyresultsin transmissionofthedisease. Asinthenon-periodiccase,weconsidertheinfectedcompartments I 1 I 2 R V + andexaminethe4x4JacobianattheDFE x 0 =(0,0,0,0, N 0 V # ) ,whichisgivenby: J ( x 0 t )= ) * * * + # )' (1 # f ) (1 # f ) & ( t ) V ) # ( \$ + )00 0 \$ # 0 & ( t ) N 0 & ( t ) N 0 0 # . 53

PAGE 54

Werstconsiderthecasewhen f ( t ) hasmean0overperiod T andexaminethe system X ( t )= J ( x 0 t ) X ( t ) .WerewritetheJacobianmatrixas J ( x 0 t )= A 0 + A 1 ( t ) ,where A 0 = ) * * * + # )' (1 # f ) (1 # f ) & V ) # ( \$ + )00 0 \$ # 0 & N 0 & N 0 0 # , A 1 ( t )= ) * * * + 000 f ( t ) V 0000 0000 f ( t ) N 0 f ( t ) N 0 00 . Thematrix e A 0 t willbeusedthroughoutouranalysis.Foranyquasi-positivematrix A ,the exponentialmatrix e A isalwaysnonnegative.Forourparticular A 0 ,wehavethat e A 0 t is strictlypositive. Lemma1. Forthematrix A 0 above, e A 0 t isstrictlypositivefor t > 0 Proof. Since A 0 isquasi-positive,thereexists r > 0 suchthat D = rI + A 0 % 0 .Then D k % 0 forallintegers k % 1 .Bydenition, e Dt = 6 k =0 1 k ( Dt ) k .Forourparticular A 0 ,by calculatingtherstfewpowersof D ,wendthat D 3 > 0 (andthus D k > 0 forall k > 3 aswell).So e Dt > 0 .But e Dt = e ( rI + A 0 ) t = e rt e A 0 t forall t > 0 .Since e rt > 0 ,itfollows that e A 0 t > 0 forall t > 0 Substitutingfor J ( x 0 t ) ,weanalyzethesystem X ( t )=( A 0 + A 1 ( t )) X ( t ) where A 0 isquasi-positiveand A 1 ( t ) hasmean0overperiod T .Letting X ( t ) bethe principalfundamentalmatrixsolution,wehave X (0, )= I .TheFloquetmultipliersof thissystemarerepresentedby & i ( ) ,withdominantFloquetmultiplier & ( ) (whichisthe 54

PAGE 55

dominanteigenvalueofthematrixsolution X ( T ) ).Weneedboth X ( T ) and & i ( ) to beanalyticin inordertoexpressthemusingaTaylorseriesexpansionforouranalysis. Lemma2. Theprincipalfundamentalsolution X ( t ) ofthesystem X ( t )=( A 0 + A 1 ( t )) X ( t ) anditssimpleeigenvalues & i ( ) areanalyticin Proof. Forasystem X ( t )= f ( t X ( t ), ) ,byTheorem1.3inChicone[ 9 ], X ( t ) isanalytic ifthefunction f isrealanalytic.Inourcase,thefunction f isleftmultiplicationbythe matrix J ( x 0 t )= A 0 + A 1 ( t ) ,whereeachcomponentisanalyticin .Thus X ( T ) is analyticin Thismeansthatthecharacteristicpolynomial p ( & ) of X ( T ) isalsoanalytic in .Any & i thatisaneigenvalueof X ( T ,0) satises p ( & i ,0)=0 .Thusbyimplicit differentiation,wehave ( p (& ( & i ,0) d & i d + ( p (" ( & i )=0. Notethat p "( ( & i ,0) + =0 foranysimpleeigenvalue & i .Sothereexistsananalyticfunction & i ( ) suchthat & i (0)= & i and p ( & i ( ), )=0 forsufcientlysmall As X ( t ) isanalyticin byLemma 2 ,wecanrewrite X ( t )= X 0 ( t )+ X 1 ( t )+ 2 X 2 ( t )+... asaTaylorserieswithrespectto about0.Wealsohavethatanysimple eigenvalue & i ( ) isanalyticin byLemma 2 ,sowecanwrite & i ( )= & 0 + "& 1 + 2 & 2 +... Recallthat X (0, )= I forall .Inparticular,taking =0 ,wehavethat X 0 (0,0)= I ,and X i (0,0)=0 forall i % 1 .Thisexpansionyieldsthesystem: X 0 = A 0 X 0 ( t ), X 0 (0, )= I X 1 = A 0 X 1 ( t )+ A 1 ( t ) X 0 ( t ), X 1 (0, )=0 X 2 = A 0 X 2 ( t )+ A 1 ( t ) X 1 ( t ), X 2 (0, )=0 . Attime T thissystemhasthesolution: 55

PAGE 56

X 0 ( T )= e A 0 T X 1 ( T )= e A 0 T 7 T 0 e \$ A 0 t A 1 ( t ) e A 0 t dt . X k ( T )= e A 0 T 7 T 0 e \$ A 0 t A 1 ( t ) e A 0 t X k \$ 1 ( t ) dt k =2,3,... As & ( ) representsthedominantFloquetmultiplierofthesystem,weareinterested inhowthisquantitychangeswhentheperturbationisintroduced.Thatis,wewantto nd & % (0) .Notethatat =0 ,oursystemreducesto X ( t )= A 0 X ( t ) ,whichjust hasthesolution X ( T )= X 0 ( T )= e A 0 T .ByLemma 1 e A 0 T ispositive,soby Perron-Frobenius(Theorem 1.4 ),theprincipaleigenvalueissimple.Weareableto showthatforsmallperturbationswithmean0,theprincipaleigenvalueofthesystem hasderivativezero.Thefollowingresultholdsforallsimpleeigenvalues,notjustthe dominantone. Theorem5.2. Let & 0 beasimpleeigenvalueof X ( T ,0) thatsatisesthesystem X ( t )=( A 0 + A 1 ( t )) X ( t ) ,where A 0 isquasi-positiveand A 1 ( t ) hasmean0over period T .Thenthereexistsananalyticfunction & ( ) denedon ( # 0 0 ) suchthat & ( ) isaneigenvalueof X ( T ) forall & (0)= & 0 ,and & % (0)=0 Proof. Considerthe nxn matrix X ( T ) witheigenvalues & i ( ) i =1,2,..., n .The characteristicpolynomial p ( & ) ofthismatrixcanbewrittenas p ( & )= n 8 k =0 ( # 1) k a k & n \$ k wherethecoefcients a k aretheelementarysymmetricpolynomials.Newton'sidentities canbestatedintermsoftheseelementarysymmetricpolynomials[ 24 ]: ka k ( & 1 & 2 ,..., & n )= k 8 i =1 ( # 1) i \$ 1 a k \$ i ( & 1 & 2 ,..., & n ) p i ( & 1 & 2 ,..., & n ), 56

PAGE 57

where p k ( & 1 & 2 ,..., & n ) isthe k -thpowersumdenedby p k ( & 1 & 2 ,..., & n )= n 8 i =1 & k i k % 1. Notethat p k ( & 1 & 2 ,..., & n )=tr( X ( T ) k ) .BysolvingNewton'sidentitiesrecursively, eachcoefcient a k canbewrittenexplicitlyintermsofthetraceofpowersofthematrix X ( T ) .Thusthecharacteristicpolynomialforan n x n matrixcanbewrittensothateach term(otherthantheconstant)isoftheform C tr( X ( T ) i )(tr( X ( T )) j & k ,where C is arealnumberand 0 i j k n areintegers.When k = n ,theconstanttermisjust a k = a n = & 1 & 2 & n =det( X ( T )) Byimplicitdifferentiation, 9 & % ( ) ( p (& + ( p (" : / / / / =0 =0 foranyeigenvalue & .Letting & representanysimpleeigenvalue, p "( ( & 0 ,0) + =0 ,sowe mustshowthat p "' ( & 0 ,0)=0 .Wewillexamineeachtermofthecharacteristicequation individually. Firstnotethat det X ( T )=det X (0, ) e T 0 tr J ( x 0 s ) ds byLiouville'sFormula (Proposition 1.1 ).Butthen det X ( T )=det \$ Ie tr( T 0 A 0 + A 1 ( s ) ds ) % sincethetraceisalinearfunction = e tr T 0 A 0 ds +tr T 0 A 1 ( s ) ds = e T tr A 0 since A 1 ( t ) hasmean0overperiod T and A 0 isconstant. Thustheconstantterm det X ( T ) doesnotdependon ,sothederivativeofthisterm is0. Nowconsideranytermoftheform tr( X ( T ) i ) foranyinteger i % 1 .Forease,let X 0 = X 0 ( T ,0) and X 1 = X 1 ( T ,0) .Thederivativeofthistermat =0 is tr( X i \$ 1 0 X 1 )+tr( X i \$ 2 0 X 1 X 0 )+...+tr( X 1 X i \$ 1 0 ) sincethetraceisalinearfunction = i tr( X i \$ 1 0 X 1 ) because tr( AB )=tr( BA ) 57

PAGE 58

= i tr \$ e T ( i \$ 1) A 0 T 0 e ( T \$ s ) A 0 A 1 ( s ) e sA 0 ds % = i T 0 tr \$ e ( Ti \$ s ) A 0 A 1 ( s ) e sA 0 % ds becausethetraceisalinearfunction = i T 0 tr \$ A 1 ( s ) e sA 0 e ( Ti \$ s ) A 0 % ds because tr( AB )=tr( BA ) = i tr \$ T 0 A 1 ( s ) e TiA 0 ds % = i tr \$ T 0 A 1 ( s ) dse TiA 0 % =0 since A 1 ( s ) hasmean0overperiod T Thusanytermoftheform tr( X ( T ) i ) & k or (tr X ( T )) i & k willhavezeroderivative. Finally,thederivativeofanyterm tr( X ( T ) i )(tr X ( T )) j & k is0bytheproductrule. Thus & % (0)=0 foranysimpleeigenvalue & Nowconsiderthesystem X ( t )=( A 0 + A 1 + A 2 ( t )) X ( t ) ,where A 0 is quasi-positiveasinthepreviouscase, A 1 > 0 ,and A 2 ( t ) hasmean0overperiod T Thisintroducesaperturbationthathasanon-zeroaverageoveritsperiod T .Weare againinterestedinhowthedominantFloquetmultiplier & ( ) (thedominanteigenvalueof thematrixsolution X ( T ) )changeswithrespectto at =0 .Inthiscaseofintroducing smallperturbationswithpositiveaverage,weshowthatthederivativeoftheprincipal eigenvalueofthesystemispositive. Foruseinouranalysis,wewillwritethissystemas X ( t 1 2 )=( A 0 + 1 A 1 + 2 A 2 ( t )) X ( t 1 2 ) andthenexaminethecasewhere 1 = 2 Theorem5.3. Let & betheprincipaleigenvalueof X ( T ,0) thatsatisesthesystem X ( t )=( A 0 + A 1 + A 2 ( t )) X ( t ) ,where A 0 isquasi-positive, A 1 > 0 ,and A 2 ( t ) has mean0overperiod T .Thenthereexistsananalyticfunction & ( ) denedon ( # 0 0 ) suchthat & ( ) isaneigenvalueof X ( T ) forall & (0)= & ,and & % (0) > 0 Proof. Wewanttond d d & ( " ) / / =0 .Webeginbyexaminingthepartialderivativesof & withrespecttoboth 1 2 individually. 58

PAGE 59

Firstconsider "( "' 1 (0,0) .Since 2 iskeptconstantatzero,thesystemreducesto X ( t 1 )=( A 0 + 1 A 1 ) X ( t 1 ) .Similartotheoriginalperiodiccase,both X ( T 1 ) & ( 1 ) areanalyticin 1 byLemma 2 .Wecanthenusetheexpansion & ( 1 )= & 0 + 1 & 1 + 2 1 & 2 + ... ,so "( "' 1 = & 1 .Since & ( 1 ) isaneigenvalueof X ( T 1 ) ,thereexistsaneigenvector y ( 1 )= y 0 + 1 y 1 + 2 1 y 2 +... suchthat ( X 0 ( T 1 )+ 1 X 1 ( T 1 )+...)( y 0 + 1 y 1 +...)=( & 0 + 1 & 1 +...)( y 0 + 1 y 1 +...). Notethat & 0 isaneigenvalueof X 0 ( T 1 ) ,sothereexistsavector w T 0 suchthat w T 0 ( X 0 ( T 1 ) # & 0 I )=0 .Usingthisandsolvingfor & 1 ,wehave & 1 = w T 0 X 1 ( T 1 ) y 0 w T 0 y 0 Recallthat X 1 ( T 1 )= e A 0 T 7 T 0 e \$ A 0 t A 1 e A 0 t dt .Since A 1 > 0 inthiscase,wehave X 1 ( T 1 ) > 0 ,andthus & 1 > 0 .So "( "' 1 (0,0) > 0 Nowconsider "( "' 2 (0,0) .Since 1 iskeptconstantatzero,thesystemreducesto X ( t 2 )=( A 0 + 2 A 2 ( t )) X ( t ) .Wehavethat A 2 ( t ) is T -periodicwithmean0,andthus byTheorem 5.2 wehavethat "( "' 2 (0,0)=0 .NotethatbyLemma 2 & ( 2 ) isanalyticin 2 Bycontinuityofeigenvalues, & ( 1 2 ) iscontinuousinaneighborhoodabout (0,0) Sinceboth & ( 1 ), & ( 2 ) areanalyticintheirrespectivevariables,byTheorem 1.7 both (& / (" 1 and (& / (" 2 arecontinuousinthesameneighborhoodabout (0,0) .Thus & ( 1 2 ) istotallydifferentiableat (0,0) .Bythechainrule,wehave d d & ( " )= (& (" 1 (" 1 (" + (& (" 2 (" 2 (" Since 1 = 2 = ,wehavethat "' 1 "' = "' 2 "' =1 .Thus d d & ( " )= (& (" 1 (" 1 (" + (& (" 2 (" 2 (" = & 1 > 0. 59

PAGE 60

5.4StabilityoftheDisease-FreeSolution Weagainconsiderthesystem( 51 )-( 56 ),whichhastheconstantperiodic solution x 0 =(0,0,0,0, N 0 V # ) asitsdisease-freesolution.Now,however,instead of ( t ) representingasmallperiodicperturbation,wejustrequire ( t ) > 0 tobe continuouswithperiod T .Thesystemhasthesame4x4JacobianmatrixattheDFE x 0 =(0,0,0,0, N 0 V # ) : J ( x 0 t )= ) * * * + # )' (1 # f ) (1 # f ) & ( t ) V ) # ( \$ + )00 0 \$ # 0 & ( t ) N 0 & ( t ) N 0 0 # . Nowwerewritethismatrixas J ( x 0 t )= A + B ( t ) ,where A = ) * * * + # )' (1 # f ) (1 # f )0 ) # ( \$ + )00 0 \$ # 0 00 0 # , B ( t )= ) * * * + 000 & ( t ) V 0000 0000 & ( t ) N 0 & ( t ) N 0 00 . Notethatthematrix B ( t ) isnonnegativeandeachtermhasperiod T Because x 0 isaperiodicsolution,itisaxedpointofthePoincar ` emap P ( x ) .Thus thestabilityoftheDFE x 0 canbeanalyzedbydeterminingthespectralradiusof DP ( x 0 ) thederivativeofthePoincar ` emap. DP ( x 0 ) istheprincipalfundamentalsolutionofthe T -periodiclinearvariationalsystem X = F x ( x 0 t ) X (58) 60

PAGE 61

where F x ( x 0 t )= J ( x 0 t )= A + B ( t ) .If X ( t ) denotestheprincipalfundamentalsolution ofsystem 58 ,then ( DP ( x 0 ))= ( X ( T )) .Sowewillexaminetheeigenvaluesof X ( T ) whicharetheFloquetmultipliersofthesystem X =( A + B ( t )) X (equivalenttosystem 58 ).Asprovedbelow,thesystemisstablewhentheaverageofthenormof B ( t ) is sufcientlysmallandunstablewhentheaverageof ( t ) issufcientlylarge. Theorem5.4. Let 0 0 denoteamatrixnorm.Thenthereexists > 0 suchthatthe disease-freesolutionofsystem( 51 )-( 56 )islocallyasymptoticallystablewhen T 0 0 B ( s ) 0 ds < Proof. ItfollowsfromTheorem 2.1 thatthematrix A isHurwitzsinceallofitseigenvalues havenegativerealparts.Because A isHurwitz,forthegivenmatrixnorm 0 0 thereexist a k > 0 suchthat 0 e At 0" ke \$ at forall t > 0 byTheorem 1.6 .Forthesystem X =( A + B ( t )) X ,theprincipalfundamentalmatrixsolutionsatises X ( t )= e At + t 0 e A ( t \$ s ) B ( s ) X ( s ) ds Then 0 X ( t ) 0 = / / / / / / / / e At + t 0 e A ( t \$ s ) B ( s ) X ( s ) ds / / / / / / / / "0 e At 0 + t 0 / / / / e A ( t \$ s ) / / / / 0 B ( s ) 00 X ( s ) 0 ds k + k T 0 0 B ( s ) 00 X ( s ) 0 ds since 0 e A ( t \$ s ) 0" ke \$ a ( t \$ s ) k ke 7 t 0 k 0 B ( s ) 0 ds byGronwall'sinequality(Theorem 1.3 ) Rearrangingourprincipalfundamentalsolution,wehave 61

PAGE 62

0 X ( T ) # e AT 0" T 0 0 e A ( T \$ s ) 00 B ( s ) 00 X ( s ) 0 ds k T 0 0 B ( s ) 0 ke k s 0 & B ( z ) & dz ds byinequalityfor 0 X ( t ) 0 k 2 T 0 0 B ( s ) 0 e k T 0 & B ( z ) & dz ds since s T Nowas T 0 0 B ( z ) 0 dz & 0 ,wehave 0 X ( T ) # e AT 0& 0 aswell.Thusastheaverage ofthenormof B ( s ) approacheszero,sodoes 0 X ( T ) # e AT 0 .Since A isHurwitz,all oftheeigenvaluesof e AT areinsidetheunitcircle.Bycontinuityofeigenvalues,the eigenvaluesof X ( T ) areinsidetheunitcircleaswell.Soas T 0 0 B ( s ) 0 ds & 0 ,the disease-freesolutionisstablebyTheorem 1.10 Toanalyzethecasewhentheaverageof ( t ) islarge,wewillneedthefollowing lemma. Lemma3. If C and D arequasi-positivematricessuchthat C D ,then e C e D Proof. Since C and D arequasi-positive,thereexists r > 0 suchthat 0 rI + C rI + D Then e rI + C e rI + D .Thisgives e r e C e r e D ,so e C e D Theorem5.5. Thedisease-freesolutionofsystem( 51 )-( 56 )isunstablewhen T 0 ( s ) ds > ; N 0 V # ( e \$ T # 1), where 0 =min 1 i 4 { a ii } with a ii representingthediagonalentriesofthematrix A Proof. Forthesystem X =( A + B ( t )) X ,theprincipalfundamentalsolutionsatises X ( t )= e At + t 0 e A ( t \$ s ) B ( s ) X ( s ) ds Let r besufcientlylargesothat rI + A + B ( t ) % 0 forall t ,andlet Y ( t )= e rt X ( t ) Then Y ( t )= e rt X ( t )+ rY ( t )=( A + B ( t )+ rI ) Y ( t ) .For Y ( t ) % 0 Y ( t ) % 0 .Sofor Y (0) % 0 Y ( t ) % 0 forall t .Thus X ( t )= e \$ rt Y ( t ) % 0 forall t 62

PAGE 63

Since X ( t ) isnonnegative,wehave X ( t ) % e At forall t > 0 .Sotheprincipal fundamentalmatrixsolutionsatises X ( T ) % e AT + T 0 e A ( T \$ s ) B ( s ) e As ds Itisgiventhat 0 =min 1 i 4 { a ii } ,where a ii representthediagonalentriesofthematrix A Note 0 0 sincethediagonalentriesof A arenonpositive,and A % 0 I .Then e At % e t I forall t > 0 byLemma 3 .So X ( T ) % e T + T 0 e ( T \$ s ) B ( s ) e s ds = e T I + T 0 B ( s ) ds ( Then ( X ( T )) % e T I + T 0 B ( s ) ds (# ,where representsthespectralradiusof thematrix,byTheorem 1.5 .Forthegivensystem,letting b = T 5 6 = T 0 ( s ) ds ,we have T 0 B ( s ) ds = ) * * * + 000 b V 0000 0000 b N 0 b N 0 00 . Theeigenvaluesofthismatrixare 0,0, b ( N 0 V ,so < e T 0 I + T 0 B ( s ) ds 1= = e T 1+ T 5 6 7 N 0 V # # Then X ( T ) willhaveatleastoneeigenvaluethatislargerthanoneaslongas T 0 ( s ) ds > ; N 0 V # ( e \$ T # 1). (59) SobyTheorem 1.10 ,thedisease-freesolutionisunstablewhenCondition 59 is satised. 63

PAGE 64

InFigure 5-2 ,wepresentanumericalsimulationofsystem( 51 )-( 56 )with f =0.5 r =1.5 b = N 0 = V # = T =1 \$ = = ) = =0.5 ( t )= +sin( 2 + t T ) with =2 (so 5 6 =2 )andinitialcondition ( I 1 (0), I 2 (0), R (0), V + (0), S (0), V \$ (0))= (0.05,0,0,0,0.95,1) toshowtheinstabilityofthedisease-freesolution.Withthese hypothesizedparametervaluesCondition 59 issatised,andweseefromthe simulationthat,evenwhenstartingclosetotheDFE,thesolutiondoesnotconverge toit. Figure5-2. Simulationofsystem(5 # 1)-(5 # 6)usingMATLABode45solverwith f =0.5 r =1.5 b = N 0 = V # = T =1 \$ = = ) = =0.5 ( t )= +sin( 2 + t T ) with =2 ,andinitialcondition ( I 1 (0), I 2 (0), R (0), V + (0), S (0), V \$ (0))=(0.05,0,0,0,0.95,1) 5.5StabilityoftheEndemicSolutioninaSpecialCase Weareunabletoanalyticallydeterminethestabilityofanendemicsolutionofthe periodicsystem,butweagainturntothecasewhere f =0 togainsomeinsight.Inthe limitingcase N = N 0 and V = V # ,theequationsofthemodelaregivenby: I 1 = ( t ) S N 0 V + V # # ) I 1 + ( I 2 + R ), (510) I 2 = ) I 1 # ( \$ + ) I 2 (511) 64

PAGE 65

R = \$ I 2 # R (512) V + = ( t ) V \$ V # I 1 + I 2 N 0 # V + (513) S = # ( t ) S N 0 V + V # (514) V \$ = V # # ( t ) V \$ V # I 1 + I 2 N 0 # V \$ (515) Thissystemhastheuniqueendemicperiodicsolution ( I # 1 I # 2 R # V # + S # V # \$ )= N 0 ) + )' N 0 ( \$ + )( ) + ) \$) N 0 ( \$ + )( ) + ) V # + ,0, V # \$ # where V # + ( t ) istheperiodicfunction V # + ( t )= e \$ t \$ \$ ( # + \$ + ) V ( # + \$ )( + \$ ) t 0 & ( s ) ds t 0 ( s ) ( \$ + + ) ) ( \$ + )( ) + ) e s + \$ ( # + \$ + ) V ( # + \$ )( + \$ ) s 0 & ( ) ) d ) ds + V + (0) ( withinitialconditionon V + ( t ) givenby V + (0)= % ( \$ + % + # ) ( \$ + % )( # + % ) 7 T 0 ( s ) e ( s \$ T ) e \$ ( # + \$ + ) V ( # + \$ )( + \$ ) [ s 0 & ( ) ) d ) \$ < & > T ] ds 1 # e \$ ( + \$ ( # + \$ + ) V ( # + \$ )( + \$ ) < & > ) T (516) and V # \$ ( t )= V # # V # + ( t ) > 0 .Noticethat I # 1 I # 2 R # ,and S # havethesameendemic equilibriavaluesasinthecasewhere ( t )= isconstant.Weareabletoprove thefollowingtheorem,similartoTheorem 2.4 ,concerningstabilityoftheendemic equilibrium.Notethatthebeginningoftheproof,whichinvolvesexaminingthelimits oftheendemicequilibriavaluesofthetrees,remainsthesameasintheproofforthe constantcase. WerststatealemmawhichwillbeusedintheproofofTheorem 5.6 Lemma4. Let x ( t )= f 1 ( t ) # f 2 ( t ) x ( t ) and y ( t )= h 1 ( t ) # h 2 ( t ) y ( t ) with m f i h i M for i =1,2 ,all t % 0 ,and m M > 0 .Assumethat f i ( t ) # h i ( t ) & 0 as t &1 for i =1,2 Then x ( t ) # y ( t ) & 0 as t &1 Proof. Dene z ( t )= f 1 ( t ) # h 2 ( t ) z ( t ) forall t .Wewillshowthatboth z ( t ) # x ( t ) & 0 z ( t ) # y ( t ) & 0 as t &1 .Firstconsider d dt ( z # y )( t )= f 1 ( t ) # h 1 ( t ) # h 2 ( t )( z # y )( t ) Since f 1 ( t ) # h 1 ( t ) & 0 as t &1 and h 2 ( t ) % m > 0 ,wehavethat ( z # y )( t ) & 0 as 65

PAGE 66

t &1 .Nowconsider d dt ( z # x )( t )= # f 2 ( t )( z # x )( t )+( f 2 ( t ) # h 2 ( t )) z ( t ) .Notethat z ( t ) isboundedaboveandbelowduetotheequationof z ( t ) ,so ( h 2 ( t ) # f 2 ( t )) z ( t ) & 0 as t &1 .Usingthisandthefactthat f 2 ( t ) % m > 0 ,wehave ( z # x )( t ) & 0 as t &1 Thus ( z # y )( t ) # ( z # x )( t )= x ( t ) # y ( t ) & 0 as t &1 ,asdesired. Theorem5.6. If 5 6 > 0 ,thenallpositivesolutionsofthesystem( 510 )-( 515 )with S (0) < N 0 convergetotheendemicperiodicsolution ( I # 1 I # 2 R # V # + S # V # \$ ) Proof. Firstweobservethat S ( t ) isnonincreasingandboundedbelowby0.So S ( t ) & S as t &1 forsome 0 S < N 0 .Nowlet M ( t )= I 1 ( t )+ I 2 ( t )+ R ( t ) .Sincethetotal populationis N 0 wehavethat M ( t ) & M as t &1 for M = N 0 # S > 0 .Hence I 1 ( t )= M + h ( t ) # I 2 # R where h ( t ) & 0 as t &1 .Thenweconsiderthesubsystem: I 2 = ) ( M + h ( t ) # I 2 # R ) # ( \$ + ) I 2 (517) R = \$ I 2 # R (518) Letting I 2 = #% ( \$ + % )( # + % ) M and R = \$# ( \$ + % )( # + % ) M ,weobservethat 0= ) ( M # I 2 # R ) # ( \$ + ) I 2 0= \$ I 2 # R Wethenperformashiftbydening i 2 = I 2 # I 2 and r = R # R .Then( 517 )-( 518 )can bewrittenas x = Ax + f ( t ) where x = ) + i 2 r , A = ) + # ( ) + \$ + ) # ) \$ # , f ( t )= ) + ) h ( t ) 0 . Usingvariationofparameters,wendthat x ( t )= t 0 e ( t \$ s ) A f ( s ) ds + e tA x 0 66

PAGE 67

Notethatthematrix A isHurwitzsincebotheigenvalueshavenegativerealparts, so e tA & 0 as t &1 byTheorem 1.6 .Thisalsomeansthatthereexist C \$ > 0 suchthat | e ( t \$ s ) A f ( s ) | Ce \$ \$ ( t \$ s ) | f ( s ) | forall t % s % 0 .Let > 0 begiven. Since f ( t ) & 0 ,thereexists + > 0 suchthat | f ( t ) | < '\$ 2 C forall t % + .Forthis xed + ) 0 | e ( t \$ s ) A f ( s ) | ds e \$ t ) 0 Ce \$ s | f ( s ) | ds .Sothereexists + 1 % + suchthat e \$ \$ t ) 0 Ce \$ s | f ( s ) | ds < 2 forall t % + 1 .So / / / / t 0 e ( t \$ s ) A f ( s ) ds / / / / t 0 Ce \$ \$ ( t \$ s ) | f ( s ) | ds = ) 0 Ce \$ \$ ( t \$ s ) | f ( s ) | ds + t ) Ce \$ \$ ( t \$ s ) | f ( s ) | ds e \$ \$ t ) 0 Ce \$ s | f ( s ) | ds + "\$ 2 t ) e \$ \$ ( t \$ s ) ds < 2 + 2 (1 # e \$ \$ ( t \$ ) ) ) forall t % + 1 < Thus x ( t ) & 0 as t &1 .Hence I 2 & I 2 and R & R as t &1 .Therefore I 1 & % # + % M whichwedeneas I 1 .Now I 1 + I 2 > 0 andwehave I 1 ( t )+ I 2 ( t )= I 1 + I 2 + g ( t ) forsome g ( t ) & 0 as t &1 .Wecanthenexpress( 513 )as V + = ( t ) N 0 ( I 1 + I 2 + g ( t )) # ( t ) N 0 V # ( I 1 + I 2 + g ( t ))+ # V + = g 1 ( t ) # g 2 ( t ) V + where g 1 ( t ) and g 2 ( t ) convergetothenonnegativeperiodicfunctions & ( t ) N 0 ( I 1 + I 2 ) and & ( t ) N 0 V ( I 1 + I 2 )+ ,respectively,as t &1 .UsingvariationofparametersandLemma 4 V + ( t ) approachestheuniqueperiodicfunction V + ( t )= e \$ t \$ I 1 + I 2 V N 0 t 0 & ( s ) ds 0 t 0 ( s ) I 1 + I 2 N 0 e s + I 1 + I 2 V N 0 s 0 & ( ) ) d ) ds + V + (0) 1 67

PAGE 68

where V + (0) isgivenbyEquation( 516 ).Hence V + ( t ) > 0 andsothereexists + > 0 and = 1 2 min V + ( t ) > 0 suchthatforall t > + S "# ( t ) N 0 V # S . Thus S ( t ) S ( + ) e & N 0 V ! t & ( s ) ds S ( + ) e & N 0 V K + T & ( s ) ds where K = > t # + T ? = S ( + ) e & N 0 V KT < & > wheretherightsideconvergesto0as t &1 .Therefore S =0 ,whichimpliesthat M = N 0 .Thus I 1 = I # 1 I 2 = I # 2 R = R # ,and V + = V # + InFigure 5-3 ,wepresentanumericalsimulationofsystem( 510 )-( 515 )with r =1.5 b = N 0 = \$ = = ) = = V # = T =1 ( t )= +sin( 2 + t T ) with =1 (so 5 6 =1 )andinitialcondition ( I 1 (0), I 2 (0), R (0), V + (0), S (0), V \$ (0))= (0.05,0,0,0,0.95,1) .Withthesehypothesizedparametervaluestheendemicsolution is ( I # 1 I # 2 R # V # + S # V # \$ )=(0.5,0.25,0.25, V # + ,0, V # \$ ) ,where V # + and V # \$ arethedened periodicfunctions.Weseefromthesimulationthatthesolutiondoesindeedconvergeto theendemicsolution. 68

PAGE 69

Figure5-3. Simulationofsystem(5 # 10)-(5 # 15)usingMATLABode45solverwith r =1.5 b = N 0 = \$ = = ) = = V # = T =1 ( t )= +sin( 2 + t T ) with =1 ,andinitialcondition ( I 1 (0), I 2 (0), R (0), V + (0), S (0), V \$ (0))=(0.05,0,0,0,0.95,1) 69

PAGE 70

CHAPTER6 NUMERICALEXPLORATIONS Themotivationforthefollowingnumericalexplorationsistoattempttoincorporate theeconomicaspectofcitrusgreeningdiseaseintoourmodel.AsinBriggsand Sculpher[ 7 ],weconsiderastochasticmodel.Mostoftheliteratureanalyzesthe benetsagainstthecostsoftreatmentforsickhumans,butthesameprinciplescanbe appliedtocitrusgreeningandtheroguingcontrolstrategy.Ideally,thiscostanalysis mightbeusedtoconvincecommercialgrowersthatroguingmorepartially-infected treesnow,whilereducingthecurrentfruityield,couldactuallysavemoremoneyin severalyearswhenfewerdiseasedtreesarepresent.Unfortunately,atthispointthere isnotmuchdataavailablefromgrowersandtoomanybiologicalunknownstoprovide accurateeconomicinsight. Evenwithouttheabilitytodepicttheeconomicimpact,itisstillworthwhileto considerastochasticmodelandcompareitwiththedeterministicmodelwehave usedsofar.Thestochasticmodelisbasedontheoriginalowdiagram,withthesame statesforboththetreesandvectorsaswellasthesamepossibletransitionsbetween states.WeusethisowdiagramtosetuptheGillespiealgorithm.Eachsteponthe graphinFigure 6-1 representsatransition;weonlyallowonetransitiontooccurata time.Thetimebetweentransitionsisdecidedbyanexponentialmeanwaittime.One differencebetweenthissimulationandouroriginalmodelisthatthevectorpopulation isnotconstant(althoughthetreepopulationisconstant).Forthisreason,weplotthe treepopulationsandthevectorpopulationsondifferentaxes,astheirscalesarevastly different. Werstsimulateourstochasticmodelforasinglegroveoftreeswiththeinitial condition ( I 1 (0), I 2 (0), R (0), V + (0), S (0), V \$ (0))=(5,0,0,50,15,50) andparameter values =100, ) =.5, \$ =.25, =1, f =.9, =.1, r =.5 ,asshowninFigure 70

PAGE 71

6-1 .Usingthesamemodelandvalues,wealsoconsiderthetimeaverages,asshownin Figure 6-2 Figure6-1. Simulationofstochasticsystemwith =100, ) =.5, \$ =.25, =1, f =.9, =.1, r =.5 ,andinitialcondition ( I 1 (0), I 2 (0), R (0), V + (0), S (0), V \$ (0))=(5,0,0,50,15,50) Wenowexploretheincorporationofasecondgroveintoourmodel,allowingfor psyllidmigration.Eachgroveseparatelyfollowstheoriginalowdiagram,butthereis nowanadditionaltransitionthatrepresentsthepossiblemovementofapsyllidfromone grovetoanother.Weconsiderthesameparametervaluesineachgrove: =100, ) = .5, \$ =.1, =1, f =.9, =.1, r =.25, m =.05 .However,therstgrovehasinitial condition ( I 1 (0), I 2 (0), R (0), V + (0), S (0), V \$ (0))=(5,0,0,50,15,50) whilethesecond grovehasinitialcondition ( I 1 (0), I 2 (0), R (0), V + (0), S (0), V \$ (0))=(0,0,0,0,20,100) ThetimeaveragesofthisstochasticmodelaredepictedinFigure 6-3 71

PAGE 72

Figure6-2. Timeaveragesofsimulationofstochasticsystemwith =100, ) =.5, \$ =.25, =1, f =.9, =.1, r =.5 ,andinitialcondition ( I 1 (0), I 2 (0), R (0), V + (0), S (0), V \$ (0))=(5,0,0,50,15,50) Nowwecomparethisstochasticmodeltothedeterministicmodelwithtwogroves. ForthesystempicturedinFigure 6-3 ,wesimulatethecorrespondingdeterministic modelwiththesameparametersandinitialconditions.ThisisdepictedinFigure 6-4 Notethatthelimitingbehaviorofeachstatevariableineachgroveisthesameforboth typesofmodels.However,theactuallimitingvaluesvaryslightlybetweenthestochastic anddeterministicsimulations.Thisisduetothenonlinearityofthemodel. 72

PAGE 73

Figure6-3. Timeaveragesofsimulationoftwo-grovestochasticsystemwith =100, ) =.5, \$ =.1, =1, f =.9, =.1, r =.25, m =.05 ,grove1(top) initialcondition ( I 1 (0), I 2 (0), R (0), V + (0), S (0), V \$ (0))=(5,0,0,50,15,50) andgrove2(bottom)initialcondition ( I 1 (0), I 2 (0), R (0), V + (0), S (0), V \$ (0))=(0,0,0,0,20,100) 73

PAGE 74

Figure6-4. Simulationoftwo-grovedeterministicsystemwith =100, ) =.5, \$ =.1, =1, f =.9, =.1, r =.25, m =.05 ,grove1(top) initialcondition ( I 1 (0), I 2 (0), R (0), V + (0), S (0), V \$ (0))=(5,0,0,50,15,50) andgrove2(bottom)initialcondition ( I 1 (0), I 2 (0), R (0), V + (0), S (0), V \$ (0))=(0,0,0,0,20,100) 74

PAGE 75

REFERENCES [1] R.N.A LLEN Epidemiologicalfactorsinuencingthesuccessofroguingforthe controlofbunchytopdiseaseofbananasinNewSouthWales ,AustralianJournal ofAgriculturalResearch,29(1978),pp.535544. [2] Furtherstudiesonepidemiologicalfactorsinuencingcontrolofbanana bunchytopdisease,andevaluationofcontrolmeasuresbycomputersimulation AustralianJournalofAgriculturalResearch,38(1987),pp.373382. [3] USDAA NIMALAND P LANT H EALTH I NSPECTION S ERVICE Aphisnewsrelease: Usdaannouncesquarantinetopreventthespreadofcitrusdisease http://www. freshfromflorida.com/pi/chrp/greening/citrus_disease-june-2010.pdf 2010. [4] R.B.B ASSANEZI ,L.H.M ONTESINO ,M.C.G.G ASPAROTO ,A.B ERGAMIN F ILHO AND L.A MORIM Yieldlosscausedbyhuanglongbingindifferentsweetorange cultivarsinS aoPaulo,Brazil ,EuropeanJournalofPlantPathology,130(2011), pp.577586. [5] A.B ERMANAND R.J.P LEMMONS NonnegativeMatrices ,SIAM,1979. [6] M.B OCHER Onsemi-analyticfunctionsoftwovariables ,TheAnnalsof Mathematics,12(1910),pp.1826. [7] A.B RIGGSAND M.S CULPHER AnintroductiontoMarkovmodellingforeconomic evaluation ,Pharmacoeconomics,13(1998),pp.397409. [8] M.S.C HANAND M.J.J EGER Ananalyticalmodelofplantvirusdiseasedynamics withroguingandreplanting ,JournalofAppliedEcology,31(1994),pp.413427. [9] C.C HICONE OrdinaryDifferentialEquationswithApplications ,vol.34,Springer, 2006. [10] E.J.D ASBACH ,E.H.E LBASHA AND R.P.I NSINGA Mathematicalmodelsfor predictingtheepidemiologicandeconomicimpactofvaccinationagainsthumanpapillomavirusinfectionanddisease ,EpidemiologicReviews,28(2006), pp.88100. [11] E.B.D YNKINAND A.A.Y USHKEVICH MarkovProcesses:TheoremsandProblems PlenumPress,1969. [12] S.F ISHMANAND R.M ARCUS Amodelforspreadofplantdiseasewithperiodic removals ,JournalofMathematicalBiology,21(1984),pp.149158. [13] S.F ISHMAN ,R.M ARCUS ,H.T ALPAZ ,M.B AR -J OSEPH ,Y.O REN ,R.S A LOMON AND M.Z OHAR Epidemiologicalandeconomicmodelsforspreadand controlofcitrustristezavirusdisease ,Phytoparasitica,11(1983),pp.3949. 10.1007/BF02980710. 75

PAGE 76

[14] A.F ONDA Uniformlypersistentsemidynamicalsystems ,Proceedingsofthe AmericanMathematicalSociety,104(1988),pp.111116. [15] H.F REEDMAN ,S.R UAN AND M.T ANG Uniformpersistenceandowsneara closedpositivelyinvariantset ,JournalofDynamicsandDifferentialEquations,6 (1994),pp.583600. [16] K.G OPALSAMY StabilityandOscillationsinDelayDifferentialEquationsofPopulationDynamics ,KluwerAcademicPublishers,Dordrecht,1992. [17] T.G OTTWALD CurrentEpidemiologicalUnderstandingofCitrusHuanglongbing AnnualReviewofPhytopathology,48(2010),pp.119139. [18] S.E.H ALBERT .Personalcommunication,2010. [19] .Personalcommunication,2012. [20] S.E.H ALBERTAND K.L.M ANJUNATH Asiancitruspsyllids(Sternorrhyncha: Psyllidae)andgreeningdiseaseofcitrus:aliteraturereviewandassessmentofrisk inFlorida ,FloridaEntomologist,87(2004),pp.330353. [21] A.H ODGESAND T.S PREEN EconomicImpactsofCitrusGreening(HLB)in Florida,2006/07-2010/11 ,ProceedingsoftheAmericanMathematicalSociety, (2012). [22] K.J ACOBSEN ,J.S TUPIANSKY AND S.S.P ILYUGIN Mathematicalmodelingofcitrusgrovesinfectedbyhuanglongbing ,MathematicalBiosciencesandEngineering, 10(2013),pp.705728. [23] J.E.K EESLING .Personalcommunication,October2012. [24] D.G.M EAD Newton'sidentities ,AmericanMathematicalMonthly,(1992), pp.749751. [25] R.M.N ISBETAND W.S.C.G URNEY ModellingFluctuatingPopulations ,JohnWiley &Sons,Inc.,605ThirdAve.,NewYork,NY10158,416,1982. [26] T EXAS D EPARTMENTOF A GRICULTURE Texasdepartmentofagricultureandusdaconrmdetectionofplantdiseasethatdamagescitrus trees http://www.texasagriculture.gov/tabid/76/Article/1802/ texas-department-of-agriculture-and-usda-confirm-detection-of-plant\ -disease-tha.aspx ,2012. [27] U NIVERSITYOF C ALIFORNIA I NTEGRATED P EST M ANAGEMENT P ROGRAM A newpestincalifornia,diaphorinacitri(asiancitruspsyllid):Provisionaltreatment guidelinesforcitrusinquarantineareas http://www.ipm.ucdavis.edu/EXOTIC/ diaphorinacitri.html ,2011. 76

PAGE 77

[28] S.L.O TT ,S.J.W ELLS AND B.A.W AGNER Herd-leveleconomiclossesassociated withJohne'sdiseaseonUSdairyoperations ,PreventiveVeterinaryMedicine,40 (1999),pp.179192. [29] R.W.H.P LUKE ,J.A.Q URESHI AND P.A.S TANSLY Citrusushingpatterns, Diaphorinacitri(Hemiptera:Psyllidae)populationsandparasitismbyTamarixia radiata(Hymenoptera:Eulophidae)inPuertoRico ,TheFloridaEntomologist,91 (2008),pp.3642. [30] K.M.R ICH Discosem:Anintegratedepidemiological-economicanalysisoffoot andmouthdiseaseinsoutherncone ,tech.report,Workingpaper,TheRegional EconomicsApplicationsLaboratory,UniversityofIllinois,2004. [31] H.L.S MITHAND H.R.T HIEME Dynamicalsystemsandpopulationpersistence vol.118,AmericanMathematicalSociety,2011. [32] H.L.S MITHAND P.W ALTMAN TheTheoryoftheChemostat ,CambridgeUniversity Press,Cambridge,1995. [33] H.M.T AYLORAND S.K ARLIN AnIntroductiontoStochasticModeling ,vol.3, AcademicPressNewYork,1984. [34] H.R.T HIEME MathematicsinPopulationBiology ,PrincetonUniversityPress, 2003. [35] A.V AN B RUGGEN .Personalcommunication,2010. [36] P.V ANDEN D RIESSCHEAND J.W ATMOUGH Reproductionnumbersandsubthresholdendemicequilibriaforcompartmentalmodelsofdiseasetransmission MathematicalBiosciences,180(2002),pp.2948. [37] C.F.X U ,Y.H.X IA ,K.B.L I AND C.K E Furtherstudyofthetransmissionofcitrus huanglongbingbyapsyllid,DiaphorinacitriKuwayama ,inProceedingsofthe10th Conference,InternationalOrganizationofCitrusVirologists.Riverside,California, 1988,pp.243248. 77

PAGE 78