A SIMULATION MODEL OF THE CONSEQUENCESOF FEED ALLOCATION DECISIONS IN MIXED-SPECIES LIVESTOCK HOLDINGS (FRAME)

FINAL TECImICAL REPORT ON PROJECT R5l.83

Peter Thorne

Livestock SectionNatural Resources Institute

OBJECTIVES, SUMMARY OF ACHIEVEMENTS AND RECOMMENTATIONS

The original objectives of project R5183 were:

To develop a methodology for investigating therelationships between seasonal variation in feedavailability and animal responses in multi-specieslivestock holdings

To use this methodology to investigate the effects ofdifferent feed allocation strategies on productivityand efficiency of feed utilisation within livestockproduction systems.

.

The limited supply of feed that is available in varyingquantities at different times of the year has beenidentified as a major constraint to livestock productionon the mixed farms of smallholders. R5183 was initiatedto develop a tool for evaluating the consequences ofdifferent strategies for allocating limited, andseasonally changing, feed resources across mixed-species,livestock holdings. An interactive, simulation model(FRAME) was constructed from standard, quantitativetreatments of energy and protein transactions in ruminantlivestock. The model uses a simplified input data set -which describes feed quality and availability -topredict the effects of different allocation strategies oncumulative production from individual animals and fromthe herd as a whole. Specifically, the trade-offs betweendifferent productive outputs (meat of different types andmilk) and essential functions (maintenance, provision ofmanure and draught power) that result from the changingpriorities attached to different animals within alivestock holding at different ti.mes of year may beevaluated.

Full testing of FRAME has not been possible at this stageas this will require data from a related project (R5690)which will describe patterns of feed availability andanimal performances in livestock holdings on 32 smallfarms in the Eastern Hills of Nepal. R5690 will becompleted later in 1995. Initial indications from alimited test are that, for periods of up to 5 months,reasonable predictions of patterns of liveweight changemay be made by the model. However, its lack of ametabolism component means that predictions for lactatinganimals in particular are likely to be less reliable.Problems were also experienced in simulating observationsof liveweight changes in growing animals but unreliable

i

test data may have been responsible for this. A morethorough test of the model will be made when all datafrom R5690 become available.

I

Initial conclusions are that the project has highlightedthe shortcomings of feed evaluation and rationing systemsderived for use in temperate production systems whenapplied to the analysis and specification of feedingsystems for the tropics. Whilst the current FRAME modelmight be used for making predictions in relativelyintensive systems, its unrestricted use for derivingpractical feeding strategies for the smallholder farmeris not recommended. However, the approach taken allowsthe consequences of feeding decisions to be simulated ina dynamic way that is consistent with farmers' feedavailability situations, production objectives andfeeding practices. Thus, we believe that the model willprovide a useful framework for incorporating futureimprovements in the evaluation of tropical feeds intopractical analyses of optimum feeding strategies for awide range of situations. It is propsed that a follow-upproject should be developed (see para. 47) to allow theinitial testing of the model in practical situationswhere its limitations apply to a lesser extent. Thisproject would also allow the incorporation of theimproved rationing system currently under development(R6282) and improved predictions of manure and compost

production from project R6238.

ii

TABLE OF CONTENTS

INTRODUCTION . .1.

THE CONSTRUCTION OF FRAME .2

General Structure and Operation. 2Variables. 4Units 4Variable. 5Units or Method of Selection 5

Prediction of Animal Performance from Feed Allocations. 6Energy and Protein Systems 6Object Orientation 6Operation of Simulations 9

Programming Language 11

TESTING FRAME 11

REFERENCES IN. THE TEXT """""""."'..~'.' .28

DISSEMINATION 29

Reports, Publications and Seminars 29Publications in Preparation 29Arrangements for Further Dissemination 29

.iii

I

APPENDICES

APPENDIX 1: INHERITANCE OF ATTRIBUTES AND METHODS BYFRAME OBJECTS .30

APPENDIX 2: PROGRAMME LISTING 47

APPENDIX 3: INPUT DATA FOR TEST SIMULATION ...J'9

iv

LIST OF FIGURES

Figure 1: Structure of the feed resources allocationmodel 3

FRAME animal object hierarchyFigure 2: 7

Figure 3:Availability and sources of feed dry matter(C.B. Bista, 6 Marcr. -25 july, 1994) 14

Figure 4: Comparison of observed and predicted bodyweightchanges for Ox 1 and Ox 2 16

.l8

Figure Sa: Comparison of observed and predictedbodyweight changes for Cow 1

Figure 5b: Comparison of observed and predictedbodyweight changes for Cow 5 19

Figure 6: Comparison of observed and predicted bodyweightchanges for Cow 2 21

F~gure 7: Comparison of observed and predicted values forN ~xcretion. by indigenous Malawi. goats co.nsumingdifferent levels of nitrogen. ' 24

vi

INTRODUCTION

1. Limited availability of livestock feeds compels farmers tomake decisions regarding priorities for the allocation ofthose feeds to the different classes of livestock in theirholdings. Their objectives in making these decisions will be,firstly, to ensure that all the essential functions of theirlivestock may be fulfilled (survival and draught power, forexample), and secondly to optimise total outputs from theanimals kept. Under these conditions, farmers need to balancethe short- and long-term needs of animals with demands foroutputs, and the future consequences of feeding decisions.However, their ability to make feed management decisions whichwill allow them to achieve this may be compromised as theeffects of these decisions may not always be immediatelyapparent. For example, a cow fed at a low level thr9ughoutpregnancy may exhibit poor body condition at calving which isassociated with low milk production, reduced calf viabilityand an extended post-partum anoestrus period. The situationmay be further complicated by the changing extent of nutrientshortages over the year which results from seasonalfluctuations in total plant biomass production, labouravailability and the quality of individual feeds.

..o' °2.. Clearly, animal performance. in smallholders' mixe.d-species

livestock h.oldingso' is detoermined' by quite complex, dynamicinteractions of farmer, animal and feed management factors.However, current advice to these farmers on feed resourcemanagement strategies is usually based on evaluations ofshort-term responses to nutritional and other interventionsand may therefore be less than appropriate for their needs. Atpresent there appears to be little information on, and noappropriate tool for, evaluating the consequences of seasonalchanges in feed supply and animal "requirements" for livestockproductivity and the effects of the complex interactions whichrelate these and other variables to herd performance. This hasmeant that integrating considerations of the dynamic nature offeed availability and allocation into the planning andimplementation of research on feeding systems has beendifficult and expensive -or, more frequently, avoidedaltogether.

Perhaps it is not surprising that the uptake ofresearcher-driven technologies based on static feeding trialshas been poor in the past.

3.

This project was initiated to develop a computer modelcapable of simulating seasonality in the availability of feedresources and the consequences for productive outputs ofdifferent strategies for allocating these limited resourcesacross a mixed-species holding. It was intended that the model

1

would,

eventually, provide a tool for researchers and otherlivestock production specialists to assist them in planningoptimum feeding strategies.

THE CONSTRUCTION OF FRAME

General Structure and Operation

4. FRAME is a dynamic model that simulates the consequences offeed allocations over a specified period for the cumulativeperformance of individual animals within a mixed-specieslivestock holding.

5.

The structure of the model is outlined in Figure 1. Theinput data required to initiate a FRAME simulation (Table 1)comprise:

.

A basic description of the type and chemical compositionof the feeds available over the simulation period (feedsdatabase)

.

.

A description of the types and numbers of animals in thelivestock holding at the start of the simulation (initial,livestock hoJ-ding).

A ~eries of dates on which observatiqnsand'pre~ictionswill be made .(at regular or irregular intervals) and thequantities of individual feeds available on each (feedcalendar) .

6. The progress of a simulation .is determined interactivelyafter the initial data have been entered. On each observationdate further inputs are required from the user (Table 2).These include:

..An~mals removed from the herd (Deaths, sales, loans etc.

Animals entering the herd (Purchases, loans etc.)

.

A work profile for draught animals

Allocations of the available feeds across the livestockholding

7. Animals that enter the herd during the course of asimulation are described in the same way as those that arepresent at its start. Newborns enter the herd automatically atparturition from the mother, the user being prompted todescribe the animals sex. For intervals between observationsof greater than one day, it is assumed that the sameavailability and allocations of feed apply on each day until a

2

Table 1: Input data r~quired to initiate a FRAME simulation

Variables ani ts

Feeds Database

Feed name Text

'rype of feed

Grass,

Grass hay, Cereal crop residue,Legume crop residue, Tree fodder,Grain,

Animal protein supplement,Vegetable protein supplement

Dry matter g / kg

Crude protein 9

/

kg dry matter

Crude fibre g / kg dry matter

Ether Extract g / kg dry matter

MJ / kg dry matterMetabolisable energy

Feed Calendar

Date DD / MM /. yy

Feed name Text (sele"cted. from database)

Quantity kg as fed

Initial Livestock Holding

Species Bos taurus, Bos indicus, Buffalo,Sheep, Goat

Sex Male, Female, Castrate

Class Growing, Mature

if mature female

Day of lactation Positive integer

Day of pregnancy 0 ..285

Working

Liveweight

Yes, no

kg

4

Table 2: Input data supplied interactively by the user at eachobservation date

Units or Method of SelectionVariable

Select from animals currently in herdRe..movals

Animal description as for initial herdstructure

Additions

Work Profile

if Working = Yes

Days worked Integer

hours per dayTime worked

mouldboard plough,Traditional plough,spike-toothed harrow

Type of implement

Feed Allocations

kg as.'fedQuantity of eachfeed allocated toeach animal

or

hours spent grazing per day. Grazingintake (GI) is derived as:

Time grazed by eachanimal

GI = 0.02 * Bodyweight * Time grazed / 24.

5

revision is specified by the user on the subsequentobservation date. Predictions of animal performance are madeiteratively on this basis.

Prediction of Animal Performance from Feed Allocations

Energy and Protein Systems

8. The simulation of animal performance within FRAME is basedon the treatments of nutrient and energy utilisationformulated for metabolisable energy (ME) and metabolisableprotein (MP) (AFRC-TCORN, 1993). There are a number ofphilosophical shortcomings in this approach. These arise fromthe requirement-based nature of the ME and MP systems which isnot well-suited to the predictive nature of FRAME~. However,there is no predictive system available currently that couldbe applied readily to a wide range of tropical feeds so the MEand MP systems were adopted for the model out of necessityrather than aptitude. The consequences of this, and thepotential for improving the treatments of energy and proteinnutrition used by the model, will be discussed later in thisreport.

9. °The. basic input data set 'required by FRAME has beenminimised too allow the model .to be operated in situations wereavail~le data are limited. Therefore, some of the feedparameters required by the model must be derived indirectly.These include the feed metabolisability (q = ME / GrossEnergy), the acid detergent insoluble nitrogen (ADIN) contentof the feed and three parameters which describe the dynamicsof protein degradation in the rumen, a, band c (0rskovandMcDonald,

1979). ADIN, a, band c are required by the MPsystem. The values used are averages for each feed type.

.

Object Orientation10.

FRAME simulates the performance of a number of species andclasses of domestic animals. Some of the equations whichdescribe energy and protein transactions are common acrossspecies and classes whereas others must be defined

-Requiremenc-based systems specify Che level of oucpuc and decermine che nutriencinpucs required co achieve chis. Prediccive syscems accempc Co predicc oucpuc froma specified range of nucrienc inpucs.

6

specifically for each type of animal. The model employs anobject-oriented construction2. This allows the definition ofbase animal types that pass attributes (e.g. liveweight) andmethods (e.g. for the determination of a maintenance energyrequirement from liveweight) to inheritor types in ahierarchical manner. The object hierarchy used by FRAME forsimulating its different animal types is shown in Figure 2 andthe attributes and methods e~ployed by each of the model'scomponent objects are surnrnarised at Appendix 1.

11. The following brief example should serve to illustrate theway in which the object hierarchy has been implemented inFRAME and some of the key features of the object orientedapproach.

12. The ancestor object TRuminant defines a method MP_Requiredwhich returns the amount of dietary metabolisable proteinrequired to support a level of production predicted from theanimal's current energy intake. The implementation of thismethod is:

13. The object TFemaleBovine (describing a mature femalebovine) obviously shares many attributes of the objectTruminant. It is, therefore, defined as a descendant allowingthese attributes and the methods that may be needed todetermine or manipulate them to be inherited. However, anassessment of the metabolisable protein required by aninstance of TFemaleBovine needs to ,include the demands ofpregnancy and lactation when these arise. Thus the Truminantmethod MP_Required must be refined for use by TFemaleBovine:

14.

The binding of data and processes that is the essence ofthe object-oriented approach promotes ,simulations that are

-A detailed description of the object-oriented paradigm is beyond the scope ofthis report. A number of standard computing textbooks are available which describeit in detail. For an example of its application in ecological modelling seeSesqueira et al (1991).

8

both more logical and more economical in programming effort.As modifications to method MP_Required in object TRuminantwill also be expressed in instances of the objectTFemaleBovine, it can be seen that the approach also helps toensure consistency between types when modifications toancestor methods are made.

15.

In general terms, the object-oriented approach is moresuited than traditional, procedural methods for producingsimulations that are representative of real-world situationssuch as agricultural systems. Agricultural systems are alsocomposed of objects -in the case of FRAME, animals withattributes and methods (the physiological processes whichdetermine their function) -and it is a positive developmentto be able to represent them, conceptually, as such insimulations.

Operation of Simulations16. The simulation of animal performance by all FRAME objectsis derived initially from the difference between the MEsupplied in the feed consumed and the maintenance MErequirement appropriate for each type of animal. This iscaicu1:ate.d using the equations sununarised. in AFRC-TCORN, 1993.

A mean, .'da.ily -rate. of production (milk and/or livewei.ghtchange) and total production (or liveweight los.s) during asimulation period is calculated from the amount of ME inexcess or deficit of the maintenance ME required. In thelatter case, weight ~oss is calculated from the amount of bodyreserves mobilised to meet a shortfall in ME for maintenance

(ARC, 1980).

17. The ability of an animal to achieve the level ofperformance predicted from ME intaKe depends on the adequatesupply of protein for turnover and production. The proteincomponent of the model, based on the UK, MP system, uses thesame general approach described by Dewhurst and Thomas (l992)who used the system to evaluate the effects of dietary changeson urine production. Dietary MP supply is checked against theMP required to support the level of production by the energycomponent. If the former is found to be inadequate, acorrection is made to the predicted production level on thebasis of the rate of MP utilisation that the current MP intakewill support. Xf MP supply is inadequate for protein turnover,a weight loss is calculated as specified by AFRC-TCORN (1993).

18. The MP system also allows the prediction of thepartitioning of nitrogen between faeces and urine (Table 3).It is assumed that all excreted nitrogen is incorporated into

9

,~:3:QH-Q

)=..-1~~'0=co

[I]Q)

UQ)

~\I..j

=-.-I[I]JJ~c.J;.J~0=Q

)010~JJ

..-1~01~-.-IJJU-.-I'0Q

)~c.~0\I..j

'0Q)

[I]~[I]=0..-1JJCO

~0'r:1~Q

)1'"'4.QC

O

e-.

-1"1mm.-I

UJ

OJ

UOJ

113r:. aJCI~:J[J]

QJ

.,..jUQJ

0.C

/)

~-foi ~

a>

0.I.J

-foi0

.I.J~

ItS

Q..U

J.foi

e .-!

0 -foi

~

.I.J~

=

'

QJ

r-i.a.~J.)

Ct/J

QJ

QJ

01010.~

~

'OJ.)

C

.~I-!

C

Ul

QJ

Ul

Ul

0~Ul

~0~QJ

010'0~~ tIJQ

JtIJtIJ0~tIJ:30CQ

J010'0C~

->.

~'001- ->

,~'0O

l-

-.LJ.c0'1-.-IaJ~->

<

M"""

0

->,C

O

z"0

~,

><

0'1 04

-~

-.LJ.c0'1-.-IaJ~->

<

L!)M0

.LJ~

aJ

aJ -.-I0'1"0Q

.LJ~aJ

~.LJ.c

Q)

-.-I .LJ

O'I~

~

II,.j-aJQ

)O>

, .LJ~

CO

aJ..Q

.u"0

"0 ~

~

r-! aJ

, "O

Q.u

-.-I t/J ~

0'1 (J~

O-~

-.-I

(J

aJ~.u.uC

O

C)

~0~~

'0 ~

0

~

~

U"I

~

o~ooo

~

QJ

\.C>

U"I tJ

' '-

oo~>

'O~

.~

tJ 0

0 ~

~

,.f"!

QJ~

..c:JJIIJJ~

JJ~~

O'I~

JJZ

.f"! Q

J 0

tJ~

~~

~~

-So.O

'Ir-i

Q)

:>0.Q~rI)~+0.00Q

) .uQ

) ~.c:

0C

/) m

aJ~0.artS

UJ

~ 1.0N0IIZ~+Q

)r-I>

0

0 0

.0 :'3:

~~

(Q

0.0:1:11-1

+J.J.r:.m.,..jO

J~>

<

a)~MN

N

~O

Ln

00

QJ

>-.-IJ.JU=

''00~04

J.Jt:Q

J~Q

J~~-.-I'0~0~t:Q

J010~J.J

-.-It:'0QJ

.0~0U)

.0n3

~0t:0-.-IU)

~QJ

>t:0u~

QJ

0 ut:

>,n3

U

t:t:

QJ

QJ

~-.-I

t:U

-.-I

-.-I n3

~

e~

~

QJ 0

-Q

J~

0.c

Q)

J.J'O

~t=

r-I

U) n3 -

-.-IU

) U

~Q

J ~

~

U)

~004~=

'04

compost together with any feeds that are rejected by animalsthat are not allocated and not conserved at the end of thesimulation period. The current implementation of the modelassumes that there is no diurnal variation in faeces or urineoutput.

Programming Language

19. FRAME is an object-oriented simulation, implemented usingBorland Pascal (BP) with Objects 7.0N. The source code for themodel is shown at Appendix 2. The interface for the currentimplementation of the model has been constructed from theTurbo Vision components supplied with BP with Objects 7.0N toprovide a user-friendly operating environment allowing accessto the model"s functions through a system of pull-down menus.The model should run as a stand-alone programme on any IBM-PCcompatible computer with a minimum of 640K RAM althoughwidespread testing to verify this has not been possible. Thecode may also be compiled into a protected mode applicationallowing access to any available extended memory for thoseusers with hardware running on the Intel 80386 CPU (or above).This would greatly enhance the simulation periods and / orfrequency of observatiops,possible.

TESTING FRAME

20. FRAME is a mechanistic model, albeit at a 'fairly low levelof disaggregation. The parameters used in its c,onstruction arerelated by discrete, identifiable processes that governnitrogen and energy transactions in the animal. The scope forapplying the model depends, therefore, on how generally theseprocesses apply. The key question here is: are the standarddescriptions of nitrogen and energy. transactions (AFRC-TCORN,1993) for animals under European conditions sufficientlyadaptable for animals of different genotypes kept under moreextensive feeding and management systems under tropicalconditions?

21. This question will be addressed in more detail when themodel is tested against individual animal data recorded underfield conditions. Arrangements for this testing have been madeunder a related project entitled "Strategies for theAllocation of Seasonally Varying Feed Resources to OptimiseLivestock Productivity" (R5690). R5690 has been examining feedallocation options and strategies in a hillside, crop-livestock system in the eastern hills of Nepal over a 15month period. One of its objectives was to provide the data

11

set needed to test FRAME. This data set will be available whenR5690 is completed towards the end of 1995.

Seasonal Feed Allocation and Livestock Perfo~ce in theEastern Hills of Nepal

22. Data for one of the Nepalese farms monitored were analysedin more detail and used for an initial test c.f FRAME t;o bereported here. The model's ability to simulate theconsequences of a real feed allocation strategy for some ofthe different animal types in a mixed species holding wasevaluated. A relatively limited simulation period of fivemonths (6 March to 25 July, 1994) on just one, typical farmwas selected. This is normally the critical period forlivestock keepers as it spans the late dry season when theavailability of feed resources is said by farmers to be mostrestricted (Thorne, 1993).

Feed use over the simulation period23.

The farm of Chitra B. Bista was visited on ten occasionsover the period used for the test simulation at,approximately, for.tnightly.intervals as part of the on-farmstudy conducted "under .R56"90. The .structure of the livestockholding on Mr Bista's.farm and changes to it during the periodare presented in Table 4.

24. On each visit, feeds offered to and refused by each animalor group of animals, their liveweights (estimated from bodygirth in the case of large ruminants) and productive outputs(including work by draught oxen) were recorded. Figure 3 showsthe pattern of dry matter usage on the farm by class of feed.In setting the scene for the simul~tion, the main points tonote are:

.the switch from diets based predominantly on tree fodder andcrop residues during the dry period to fresh, cut andcarried grasses. This change coincided with the forage flushbrought about by of the onset of the annual monsoon.

.the relative shortage of dry matter available between mid-April and mid June

25.

It is in this context that Mr Bista's decisions regardingthe allocation of feeds to different animals in his holdinghad to be made during the simulation period. In addition hewould have had to consider the specific needs of particularanimals, in particular the parturition of Cow 1 and a periodof work for his two oxen between mid May and early June.

J.2

'00-.-I~0)0-~0-.-I.I.J~.-I§-.-ItI)0).c.I.J

0'1~-.-I~~'0'00)>~0)tI).00tI)0)0'1~~.cu'0~

-~

~0'10)0'1~

~

~.I.J ~

U

>,

~.-I

~

~.I.J I-JtI)

l/')

O'IN

~-.-I I

'0.-I.c0

U.c

~~.-I~~-.-I

I.D.I.J-M

~

~

~1-1

.I.JtI)I

~

~0)

-

.-I~.0

-

~

UE

-.-

C1

Q)t))

dlIS~U -'0Q

J'0\.I0t)Q

J\.IQ

J..I.JId'U-.-4ItSE

i-ri

~11-10Q

)

~.Q.Hr-IId~-ri

~

'51-1

~\,Q

.u113

b1~"P

"i'0ro-i0~~I-i

lJ)~0(J0J.J

.r:.J.J~-.-I.0Q

)::-~0.u~ItS~0)Q

)--1-10101~~

~0'\.-1,---.-11.00)0)0)0)N

N~

~~

~--N

r--I.O~

tJ'1.-I .-I

1.0 0\

~.

'tj'tjM.-I.-IN

OQ

)Q)

M.I.J .u

.-IItS

ItS

Q

) 1-1

1-1 Q

)1-I1-I,-iQ

)Q)M

II.u.uItS

~~

1tSfJ1 fJ1 e

OM

OM

e .u

ItS

ItS

Q)

Q)

Q)

Q)

.cU

U

~

.c

:I: ~

'0)oM

"""Q)

fJ1 fJ1

tJ) tJ)

tJ) tJ)

:3:J:J~

~~

~Q

)U

UU

UU

U:>

""1 °'"1

' '

" '

OM

'O'O

'O'O

'O'O

M~

~~

~~

~""1

"'"1 '

" '

,-iItStJ)

tJ) tJ)

tJ) tJ)

tJ) .u

0000000~

~~

~~

~E

-4

~~0\0\1-1

.-I-.-I~01~LnN-

rn.I.Jr-iN

Mqoln

.-IN

03333~

:0<:0<

0000O

OU

UU

Ur"-

,~O

M~~.:x:

':!'-

-~

':!'~

0'1

0 0'1

.aM.cm

'

-M

~Q

) -M

c= ~~

0 .:x:

.u':!'

'0-Q

)c=

'O~

~0

0~

C/)

-~0\0\.-I

'00"""11..4Q

)a.§°""1~~1..4Q

)tf1.Q0t))~"""11..4:J'0'0Q

)1..4Q

)~~

~

~eM~CI

-.::C

-.q'~

~l!')~

~N

m.-l

-.-I

.-I ,O

JO

J ~

~

~

~0

~

I-:)U

I-j.-I

Ol!')N

J..)--t:

~

~~

~

~

0 0

0~

~~

-~0\0\r-4

CI)

J.J J.J

U'\

n3 n3

0 0

3 C

-' C

-'0U

""'~

-0'}~

0'}

O'

~o;t'

N-.

\C~Q

J 0'}

II~

~..-i

M

~Q

J .~

~M

O'}

..-i,

II O

J(/j

~:J

~

OJ

U ~

>

""1 O

'}..-i"tJ

..-i .-t

~

OJ

""1 ~

.-tO

J rt3

(/j >

~0

..-i 0

~~

E--

-~(j)(j)~>

."SJL{)N

I

..c:<

.)~('J

:2:to('J-cninm()-~Q

)-ro:2:C

:-o"'CQ

)Q

)U

--0cnQ

)<

.)~:J0U)

"'CC('J>

.:t::

::0('J

"'ffi><"

"

(t)Q)

~:JC)

i:i::::

~~

~~(~a

5)j)~

e:In

0

~ ~~~ QI

E~C)

QI

--' ~cOJ

EOJ

"Q.

Co

:.C/) !.!0

Simulation26. The data described in outline above were used as inputdata to establish a FRAME simulation. They included records ofthe quantities of feed available and consumed by each animal,the chemical composition of those feeds and the initial stateof the herd. For reference, the data are shown in detail atAppendix 3. At the time of writing, chemical analyses of feedsare incomplete so it was necessary to include a number ofestimates based on existing information in setting up thesimulation. Simulations were not conducted for Cow 3 and Cow 4as these left the herd during the simulation period with only2 observations made.

Ox 1 and Ox 227. Figure 4 shows the liveweight changes observed in MrBista's oxen during the simulation period in comparison withthe predictions of the FRAME simulation. The feeding strategypursued by Mr' Bista over the period is represented by theobserved dry matter intakes of the two animals (protein andestimated metabolisable energy intakes follow broadly the samepattern according to the analyses currently completed).Separate feeding of the animals was observed during one visitonly (2.2 May., 1994) when Ox 1 (the'lighter animal). was offered(and consurne'd~ app~oximately do~le the amount of dry matter

offered. to Ox 2. This strategy was probably adopted in orderto bring the lighter animal into condition for the period ofwork which took place between 22 May and 5 June. Its successis inqicated by the increase in the observed liveweight .of Ox1 between 4 April and 7 May.

28. The predicted liveweight changes (although quantitativelyadrift at the end of the 5 month simulation period by 13kg(6.7%)and 18kg (7.9%) for Ox 1 and Ox 2 respectively) appearto reflect the pattern in observed changes reasonably closely.In particular, the simulation was able to replicate the lossin liveweight occurring during the working period and therecovery of liveweight in response to the monsoon seasonforage flush. The growth spurt shown by Ox 1 between 4 Apriland 7 May was not clearly indicated in the simulated data. Itis possible that this was due to inadequate input data. Thesustained liveweight gain observed would suggest thatpreferential feeding of Ox 1 continued for most of the periodbetween 4 April and 7 May. However, the snapshot observationon 25 April limited preferential feeding to the period between25 April and 7 May in the simulation.

15

Cow 1 and Cow 529. Figure 5a shows the observed and predicted pre- and post-partum liveweight changes) for Cow 1 in relation to dry matterintakes observed at each visit. The observed and predictedearly growth curves for Cow 5, which was born on 25 April, toCow 1 are shown in Figure 5b.

30. The predicted values for Cow 1 demonstrate acceptablesimulation of the pattern of liveweight change in the peri-partum period and the initiation of the growth response to theonset of the monsoon season feed flush. The observed andpredicted liveweights of Cow 1 differed by 15kg (5.4%) at theend of the simulation period.

31. When compared with the pattern of dry matter availability(see Figure 3) the pattern of dry matter intakes recorded forCow lover the simulation period suggest that Mr Bistaattached considerable priority to the adequate feeding of thisanimal after parturition. Dry matter intakes during earlylactation were consistently higher than those observed pre-partum despite the relative scarcity of fodder between mid-April and early June. This strategy resulted in a post-partummaternal weight loss that was restricted to 25kg which was notsimulated by the model. The most likely s?urce of.thedifferences between.observedand predic~ted liveweights duringthis .period'.is th.e inability of the ME and MP systems used. inFRAME's construction to predict lactation performance. Theproblems of using the requirement-based ME and MP systems werealluded to in the section of this report which described themodel's construction. There is no direct treatment of themetabolism of absorbed nutrients in the requirement-basedsystems so effective prediction of the partitioning ofnutrients between lactation and other metabolic demands suchas maintenance and bodyweight change is not feasible. This isprobably the main obstacle to the widespread application ofthe model in its current form.

32.

The simulation of bodyweight changes in the newborn calf(Cow 5) was unacceptable. This was expected as the currentimplementation of the model has no treatment for milk consumedby the suckling animal which would be its principal source ofnutrients at this stage. Further data are required to allowthe factors which determine the consumption of milk by thecalf to be incorporated into FRAME. However, any treatmentbased on an understanding of these would be superfluous beforethe model's ability to predict total daily milk production has

been improved.

milk yield data are not available at ~he time of writing

J.7

..-~0u...0-C/)

Q)

C)

cCIS

..ctJ

:EC)

.~~0.Q-cQ)

U'5Q)

...a.

~cCIS

-cQ)

~Q)

C/)

.Q0-0c0C/)

"CCIS

a.E0u"cu"It)Q

)...~C

)

ii:

~0~

§~

~

(6)1) 1~6IaM

a"!1

0~~

:?i

~~0u'-0"-InCJ

0>c:cu.c:(J

:E0>"a;~"00..c"0C

JU'6~Q

."0c:cu"0C

J

~CJ

In..c0""-0c:0In"t=C

OQ

.

E0u.:c~~~0>i.i:

0 8

I/)co ~

g g

<M

8~ 6)1/6)

~

~~(5>

1) -

Ja~w

l.;a

.9,

~

on'Q"

~

~ ~

Cow 2

33. Observed and predicted values for the liveweight of Cow 2differed by 5kg (3.4%) at the end of the five monthsimulation. However, this outcome does not reflect theapparent total failure of FRAME to predict the pattern ofliveweight change in this growing heifer (Figure 6). For mostof the simulation period observed growth rates areconsiderably greater and liveweight losses considerably lessthan those predicted by the model on the basis of nutrientintakes.

34. Failings in the model which might account for thesediscrepancies could be:

.a need for more reliable compositional data for the feedsused.

.the operation of other nutritional factors not accounted forby the ME and MP systems. These would, in any event, have toexhibit quite miraculous growth promoting properties toaccount for the observed effects.

However, the feeds consumed by Cow 2 were substantially thesame as those fed to the other animal types so "these factorsshould have produced similar' results in the simu:1ations .conduc,ted with other types of animals.

34. Clearly the ability of the model to predict theperformance of growing animals needs to be more widelyevaluated against the Nepalese data before firm conclusionscan be drawn regarding this disparity between observed andpredicted values. If the problem is more widespread, the mostlikely explanation would seem to be that the observedliveweights are inaccurate. At relatively mild levels ofunder-nutrition, growing animals may continue to increasetheir frame size at the expense of body reserves. This wouldresult in an increase in body girth (upon which the estimatesof observed liveweight were based) but a decrease in actualweight.

Manure-Compost Component

35. The prototype for the manure-compost component of themodel was developed as part of a review of the role oflivestock in nutrient cycling in sub-Saharan Africa (Romneyetal, 1992). The objective of the original work was to assessthe feasibility of a modeling approach for investigating therole of livestock management in determining soil fertility incrop-livestock systems. This model was a static simulation of

20

.1;

N~0()'-0-(/)Q

)C

)CC

O.c(,)

~C)

°'Q5

~"C0.0"CQ)

U'5Q)

'-a.

"CCCO

"CQ)

~Q)

(/).00-0C0(/)

°1::C

Oa.E0()cO

"Q

)'-:JC

)

i:i:

the effects of a limited range of livestock managementdecisions on the quantity and quality of organic soiladjuvants in crop-livestock systems. For integration intoFRAME, the original model's description of the fate ofnitrogen was improved (Thorne, 1995) and a number ofadjustments made to allow the dynamic prediction required. Acomplete validation of the manure-compost component has notyet been undertaken because of the exploratory natur."e of thework.

36. For the purposes of this report, a simple comparison ofFRAME predictions with experimental observations has beenpresented to illustrate current limitations on the predictivecapacity of the model. The input data used for this simulation(Table 5) are derived from a nitrogen balance experimentconducted with indigenous Malawi goats offered a range of hay-supplement combinations (Reynolds, 1981) which was selectedat random from a number of suitable, published studies. Faecaland urinary nitrogen production observed in the experiment andpredicted by FRAME using the input data derived from thisstudy are shown in Figure 7. The diets used in the studyexhibited a range of nitrogen contents. Accurate quantitativeprediction of the individual experimental observations was notachieved by the model as, in general, pred'icted 'values" felloutside the ranges ot the. errors a,ssociated with the observedvalues.

However, the,direction and rates of the predictedresponses to increasing dieta"ry N were significantly"correlated with those observed in vivo (faecal N, r2 = 0.676;urine N, r2 = 0.981). Similar inaccuracies were observed inthe quantitative prediction of nitrogen balance (Table 6) butagain, correlation over the range of treatments wassignificant (r2 = 0.959). Growth performance data were notavailable for the experiment. However, weight changespredicted by the FRAME model, also shown in Table 6, wereconsistent with the observed nitrogen balance data -animalsthat were in negative N balance in the experimental study lostweight in the simulations.

37. Inaccuracies in the predictions, for urine N production atleast, would appear to be errors of quantification rather thandue to the general formulation of the relationships whichdefine the model. This implies that there are inaccuracies inthe data used within the simulations. It is suggested that theparameters most likely to be responsible are those whichdescribe protein degradation in the rumen (a, b and c) andADIN which contributes directly to faecal N levels. Valuesused in the model are derived means for each different classof feed but, for a, band c, are likely to be subject to quitelarge inaccuracies as these parameters are highly variable

22

,rJ)

'0~0c=:>.

OJ

~-ro:I.Jro'0'0O

J..c:

rJ)"M~.0~0..

11-10c=0-MJ.I~~§"MrJ)

~011-1

roJ.Iro'0J.I~0..c=H -.-jCX

)0'\.-j..LnQ

).-!..0C

t3

E-o

QJ

0->.

E-I

'0QJ

QJ

~

-~

~Q

) O

M

Q

'0 Q

)=

,JJC1

~

0 ~

U~

"AI

C1

- -~O

J O

J Q

'0 ~

~..oO

)~

..-i

~t)1i-I

"0)

- -.LJ~

~

U 0

OJ

In.c

~

0'1.LJ.LJ~r%

1~

~

0'1

-~

~Q

) C

>'.1..10)

~

.1..1 .!(

o~ e

0)

-

QJO

1;or-i~

QJ

,a'J-Jro

I-:J roC

J) ~ E

t t

r-i>,J-J

0 01

CJ)

~

~

QJ

.LJ QJ

-Q

J ~

~~

~Q

'0G)

G)

~

MN

M~

>,>

,>,>

,~

~

~

~

..c..c.c..c---

~~

~~

MN

Mt:t:t:t:---~

~

~

~

~~

~~

Q)Q

)Q)

~

~

~

~

.I.J .I.J

.I.Jt1'It1'It1'It1'I~

~~~

~

~

{J) {J)

{J) {J)

.I.J .I.J

.I.J""1 ""1 ""1 ""1

~

~

~~

~~

~Q

)Q)Q

)O

OO

OU

UU

r-ir-ir-i~

~

~

~.c:r'::.t:A

0

0 0

tJvtJvuuu

>.ItS

.cmmItS

-~

-=

=-

0 =

=

-

Ll100Nr'-C

X>

Mm

NO

MN

Or-i

cx>m

mm

mm

m

O\M

O\O

OO

Or-.

Ll)Ll)Ll)Ll)OM

qt,-iN

M

[,,-U'lN

U

'I 0°0\

Nr-1""'~

"",OO

\~

~r-1~

r-i

0000\00011)0\~

0\1.01.0

OO

Ll'lOLl'l~

~~

~r-i~

C"-~

N

C"-C

"-OO

r.:r-ir-ir-ir-ir-ir-i

J.JItS001

'0Q)

J.JItS\..IJ.JrnItSC

J

Q)

\..I~J.JItSe01~0MQ

)C0t/)~.,..1ItSJ.JQ

)'0r-iItSe.,..1

c~

Figure 7: Comparison of observed and predicted valuesfor N excretion by indigenous Malawi goatsconsuming different levels of nitrogen(observed values from Reynolds, 1981)

8

7

->.cU~

tJ)-'0Q).!oJQ)~tJXQ)

=Q)tJ)0~.!oJ

-.-fZ

6

5

4

3

7.

1.

Nitrogen intake (g / day)

I

I

Table 6: Predicted and observed nitrogen balances andpredicted growth rates based on data from Reynolds (1981)

Nitrogen balanceTreatment Predict~d

weightchange

(g / day)Observed Predicted

-1.81:tO.394-1.20:tO.357

1.11:tO.3127.45:tO.378

HayHay + Concentrate 1Hay + Concentrate 2Hay + Concentrate 3

25

-1.85

0.003.989.61

-94

-31

3517

(Webster, 1993) even when determined for the same samples atdifferent laboratories. The use of ADIN to representundegradable, undigested nitrogen has been adopted relativelyrecently, therefore, only limited data are available on whichto base model predictions. Thus, increasing the accuracy withwhich a, b, c and ADIN are quantified is likely to be the mostsignificant barrier to improving the quantitative predictivecapacity of the model in f~ture. Direct input of theseparameters for feeds used in simulations could be expected toincrease accuracy of prediction. However, these data are notwidely available for practical situations and insistence ontheir use would probably compromise the utility of the model.

Arrangements for Further Testing

38. Analysis of the data set from Nepal will be completedunder an extension to project R5690 during the latter part of1995. This work will include a detailed evaluation of thepredictive capacity of FRAME when applied to the animals onthe 32 livestock holdings monitored by the study.

39. A key requirement for potential FRAME users is likely tobe the model's ~ility 40 test the foll9wing:

.alternative allocations of feeds acrOES species;

.optimum timing of the use of feed supplements;

.alternative interventions under particular conditions offeed availability;

.interactions of feed availability and allocation with otheraspects of herd dynamics (by manipulation of calving datesand selective culling for example).

40. As the testing of the model described here was relativelylimited in its scope, it was not considered appropriate toillustrate the use of the model in this way. However, a fullevaluation of FRAME as a tool for effecting this kind ofanalysis will be conducted as part of the full scale test ofthe model

41. Further testing of the manure-compost component will notbe undertaken on the current implementation of the model. Workunder a new project entitled "Implications of LivestockFeeding Management for Long-term Soil Fertility in SmallholderMixed Farming Systems" (R6238) will include refinement of theapproach adopted in FRAME. This project will also examine thescope for integrating the model's manure-compost outputs withsoil nutrient models for predicting the consequences of

I

I

I 26

I

feeding management for the maintenance of soil fertilityThe modeling phase of this study will be undertaken incollaboration with soil scientists from Wye College,University of London and field studies in Nepal willgenerate data for testing the model.

CONCLUSIONS

42. The work conducted under project R5183 hasdemonstrated the general suitability of the simulationmodeling approach adopted for analysing the feedingsystems operated by smallholder farmers with mixed-species livestock holdings. However, a number ofshortcomings in the current version of the model havebeen identified as the work has progressed. These relatebroadly to the different nature of smallholder feedingsystems (when compared with more intensive systems) and alack of specific information on the productivity oftropical breeds of livestock from tropical feeds. Themost serious aspect is the indication that the ME and MPsystems used to construct the model are not ideallysuited to the predictive nature of FRAME. In particular,the lack of a treatment for the metabolism of absorbednutrients restricts the model's applicability. It isacknowledged, therefore, that FRAME's use at presentwould be limited to more intensive production systemswhere outputs may be "pre-specified" to allow a morerequirement-based approach.

43. Currently, there are a number of opportunities whichcould support the further development of FRAME to addressthese deficiencies. This further work will not requirespecific funding as it can be effected as part of two newprojects.

.

44. Project R6282 entitled "Development of a PracticalDairy Feed Rationing System Appropriate for Use inDeveloping Countries" will provide a predictive systemfor rationing dairy cattle in the tropics. In principle,it will be possible to extend the system for use withother classes of livestock and use it as a completereplacement for the ME and MP systems currently used byFRAME. This should address the model's difficulties inpartitioning of absorbed nutrients that compromise itsuse with lactating animals.

46.

Project R6283 will reassess the requirements for ananimal model which can be used to analyse the role oflivestock in nutrient cycling in mixed farming systems.

27

The manure-compost component of FRAME will be used as astarting point for this work and the outputs of R6283will ultimately enhance the model's capacities in thisarea.

47. A follow-up project will be required to allow theinitial testing of the model in practical situationswhere the limitations discussed in para. 42 apply to alesser extent. This project would also support theincorporation of the improved rationing system currentlyunder development (R6282) and improved predictions ofmanure and compost production from project R6238. Thefuture version of FRAME incorporating these innovationswould then be tested for its suitability for thefollowing applications:

.pre-testing improved treatments for researchersconducting experiment on improved feeding strategies;

.analysing potential nutritional interventions byplanners and designers of development projects;

.support to extension staff planning recommendations onpractical feeding strategies;

.training of livestock officers and extension staff.

A concept note describing the proposed follow-up projectwill be submitted for livestock production programmefunding in due course.

REFERENCES IN THE TEXT

AFRC-TCORN (1993) Energy and Protein Requirements ofRuminants. Wallingford, UK, CAB International. 159pp..

Dewhurst R J and Thomas, C. 1992. Modelling of nitrogentransactions in the dairy cow and their environmentalconsequences.

Livestock Production Science, 31: 1-16.

0rskov, B.R. and McDonald I. (1979) The estimation ofprotein degradability in the rumen from incubationmeasurements weighted according to rate of passage.Journal of Agricultural Science (Cambridge), 92: 499-503.

Reynolds L. 1981. Nitrogen metabolism in indigenousMalawi goats. Journal of Agricultural Science (Cambridge)96:

347-351. I

Romney,

D.L., Thorne, P.J. and Thomas, D. (1992) A Reviewof the Role of Livestock in Nitrogen Cycling in Mixed

I28

Farming Systems in Sub-Saharan Africa. UnpublishedReport, Chatham, UK, Natural Resources Institute.

Sesqueira, R.A., Sharpe, P.J.H., Stone, N.D., El-Zik,K.M. and Makela, M.E. (1991) Object-oriented simulation:

plant growth and discrete organ to organ interactions.Ecological Modelling, 58: 55 -89.

Thorne, P.J. (1993) Report on a Visit to Nepal toInitiate a COllaborative Study of Seasonal Feed ResourcesAvailability with the Pakhribas Agricultural Centre.Unpublished Report: R2009(S), Chatham, UK, NaturalResources Institute.

Thorne, P.J. (1995) Modeling the effects of livestock onnutrient flows in mixed crop-livestock systems. In:Proceedings of an International Conference on Livestockand Sustainable Nutrient Cycling in Mixed Farming Systemsof Sub-Saharan Africa held at ILCA, Addis Ababa,

Ethiopia, 22 -26 November, 1993. Eds. J.M. Powell, T.O.Williams and S. Fernandez Rivera. Addis Ababa, Ethiopia,International Livestock Centre for Africa. (In press).

Webster A J F. 1993. The metabolizable protein system forruminants.

In: Haresign Wand Garnsworthy P C, (eds),Recent Advance in Animal Nutrition -1992, Butterworths,

London, UK pp. 93 -110.

DISSEMINATION

Reports, Publications and Seminars

Thorne,

P.J. (1993) Strategies for the Allocation ofSeasonally Varying Feed Resources to Optimise LivestockProductivity.

Seminar presented at Pakhribas AgriculturalCentre, Nepal. 30 May, 1993. .

Thorne, P.J. (1995) Modeling the effects of livestock onnutrient flows in mixed crop-livestock systems. In:Proceedings of an International Conference on Livestockand Sustainable Nutrient Cycling in Mixed Farming Systemsof Sub-Saharan Africa held at ILCA, Addis Ababa,Ethiopia, 22 -26 November, 1993. Eds. J.M. Powell, T.O.Williams and S. Fernandez Rivera. Addis Ababa, Ethiopia,International Livestock Centre for Africa. (In press).

Publications in Preparation

Thorne, P.J. Modeling the consequences of seasonalchanges in feed availability and farmers' feed allocation

29

I

decisions on production from mixed-species livestockholdings. Poster to be submitted for the Fourth

International Symposium on the Nutrition of Herbivores,Clermont Ferrand, France.

Arrangements for Further Dissemination

Two further publications describing the model'sconstruction and its testing against the Nepalese datawill be prepared. These will be submitted to a refereedjournal such as Agricultural Systems.

Further dissemination funding will eventually be soughtfor the preparation of a user manual to accompany a finalversion of the model to be distributed on disc.

I

30

m'0Q)

Q)

11-1

Q)

r-I.Q~r-I0.-/

~~~11-10m~00.-/.j.J0.-/m0~0U.-I~U0.-/

=Q)

.r:tJ..r-!IC

t:

Q)

r-I.Q~E

-I

or::u~ro~1.0ro.j.JmOM

~~U-

->.

.-I~If)N -qt~~.-i

Q)

Q.

>.

I- c:(1)-0...c..

(1)"C~...(.)...(1)=toE~c"'CQ

)Q

)u. Q

)~

..c\t::Q)

"'C~~u t)cu...XQ

)...Q

).c:-W "'C

>

-Q

)-a>_..os-m

m

Q

)E

.~

c::

~oQ

)C

I)..oQ)

wm-Q

)

E

-C)

~-C)- ~cC

)~"0- ~cC

>~--,~

Q)

:J~(/) Q

)Q)Q

)Q)

"C"C

"""C"C

"CQ

)"C0

0E

o---Q

)Q):JQ

)Q

)Q)C

>Q

)Q)

I-I--JI-

~~

~~

N~

~O

~~

NO

NN

N~

~~

~o~

m~

~o~

~~

~m

mm

m~

~~

~~

N

~~

Nm

~~

NN

N~

N

v~W

NN

vMm

~W

O~

NN

N~

OM

ro~voN

~m

~~

m~

omm

rov~

~~

~~

~~

~

~

~

M~

N~

WO

WO

N~

roM~

NN

NN

~O

~N

~~

ro~~

W~

M~

~O

OO

Mro

N~

NN

N

~~

N

MM

MM

~

w~

~~

ro~~

ro~w

omw

w~

~m

M~

N~

~N

~M

~M

~~

~~

O~

~

W~

W~

O~

~~

~W

NN

W~

N~

~w

Nm

m~

ro~~

~~

~m

roomw

w~

OO

~O

OM

OO

mm

OO

~~

mO

Om

OO

Mm

~

~~

~

---~

NM

---Q)

'-C/)C

/)C/)U

)~~

O

C~

C/)C

/)U)U

)CU

-'- "--

It, C

U0

CU

C

U

CU

C

U

U0

", It,

'" '-

'- '-

'- '-

0'-

", C

U

:. C

0

.c ~

C

) C

) C

) C

)"'C

'-

OC

U.c-O

~-C

'-C

U.cE

cE>

--~~

oQ)Q

)Q)"'C

"'C"'C

cu~E

~

_,,--NN

NQ

)Q)Q

)---C

C

U

CU

C

U .-"'C

'-".~

'- X

X

X

C

/) .cC

U

CU

Q

) ..c

..c..c ~

0

CU

C

U

CU

C

U

.-~

Q)

~~

~U

UU

CG

~~

~~

~~

~~

Z

c::Q

)

0'-'-'-'-

0-'-Q

)Q)Q

)Q

)Q

)"'0"'0"'0

-"'0"'0"'0"'0

..0"'0000

cuo---C

f)c::c::Cf)C

f)Cf)C

f)""'-Q

) Q

) Q

) C

f) C

f) C

f) C

f) C

f) Q

) Q

)

Q)

Q)

Q)

CU

C

U

CU

C

U

CU

C

U

CU

Q)

C>

Q

),-,-,-'-'-'-'-'-'-'-

'-G

GG

GG

GG

>..-

'0Q)

aOM4.)~0tJM<Q

)r-i.Q~E

-t

Q)

=mE~c c::"0;-0I-0..Q

)"'C~l-

t) Q)

.c~Q)

"C::J

U"CQ)

Q)

u. G)

c..>

-I- (..)~0-

XQ)

0-

Q)

..c:W "'0

>-

Q)-C

>-..Q

I-C

UC

UQ

)E

.~

c:.--Q

)-0U

)..QQ

)W

cu-Q)

E

-C)

~-C)- ,~0C

>~"C

>'-" ~c0)~-~~-

Lt)""""C\I~

OO

,,,,""Lt)O~

OQ

)L()<X

:)O<

X:)~

~N

M

<x:)""""~

N

<x:)

I- I-

Q)

Q)

"C"C

"C"C

0 0

--Q)

Q)

Q)

Q)

I- I-

1-1-

Ll)<D

~~

Ll)omm

~N

~O

~N

~

",""N~

~

N~

N<

OC

X)<

OLr>

OO

~Lr>

Q)~

cx)Lr>~

Lr>

cx)~~

N

N~

~

NLr>

~Lr>

mrom

o~~

Lt)~

~

~

~~

~

Lt)CO

OLt)",,"O

~~

N~

NLt)Lt)C

O

ON

NO

OO

>M

N&

r>"

NO

O&

r>"

~~

~

C1)

Q)

~:-QC1)

Q)

'-.S

D

-ca

0'-

'-"u

Q)

[:O>

~

0 ~

2 'i

m-[:~

_0

[: m

-

.-m

Q..-

C/)

~

~.c0

-'i'i>

-Q)Q

. Q

)Q.

[:-E

Q)Q

)-Q)

Q)

Q)

.--u u

-:>~

m

m

:J m

~

~

c..c..c..O::O

::U)U

)1-

(/)Q

)~"C"0-Q

)(/) '-(/)Q

.(U

0

'- '-

GU

.c:u~In~1.0

In+JUl

OM

~~tJ-'00OM~aJ01

I:::0OM+JInr-i§OMUl

aJ.c:+

JOl

I:::O

M~;j'0Ul

'0aJaJ~Ij..j0>

.+

JO

Mr-iO

M,0Inr-i°r-iIn~N~aJ-r-iqt,00'1In

0'1E

-tri

>.

r-I~I-JLt)N

-"'C

~U

)cuC

)~-Q

)

::c~Om>cuU)

Q)

:;:::;::ccu:J0-cQ

)Q

)La.

~0-<D0 M0-~N ~a-~a ~0-I.()

N IJ')0-r 0 I.()0-NN C

D0-LC

)0 C

D0-~N f"'-a-I.()a r 0-Lr>N

0:(t)

~0~

0.Lt)

0.M Q)

Lt)

("II("II ~~

m.

~ !'-;m

~M

m~

~ciN C

D;

('II

0.<D

~

"'":0~.ro

0 0.~~0

~0 ~O

0"'""

Ll)Ll)o~

~OL()

C")C

")L()

(X);

"f:t~

M.

com

(");a ~0

o~MLn

N

, CX)

om"","

~~

M CX

)

CD

0.(\I

M;

N Lt';~~

0:~ 0:~

---~

N~

---Q)

-f/)

f/) f/)

f/) ~

"5

~o

C~

f/)f/)f/)f/)~

~...~

~~

~u-

0~~

>-Q

)Q)

0 0

C

o~~

~o§~

c ~

~~

~~

~~

~~

~

~E

cE>

-~~

~oQ

)Q)Q

)~~

~S

~>

-Q)~

E

c ~

~

~

=

~

~

.~.~

.~

"~

~

~

~

f/) ~

.S

';

E~

~

Q

) ~

~

~

~

0

~

~

~

~

~

Q)

~

~

~

~~

~U

UU

C~

~~

~~

~~

~~

z~~

~

~OJ

~~".-I.&

J~0UN1::1:

OJ

r-i.a~E

-t

"CQ)

Q)

LI- -"0Q)-

(/)cuC

)~"--"Q

)

::c~om>cu(/)Q

):;:;c:cu~0

M0-CD

0 M0-T-

C\/

~0-.~0 ~a-I.()N a.n0-, 0 Lt)0--NN c.o0-Lt')0 C

D0-~N t"-o--L!)0 f'-0-Ll)N

Q..qoM

~~

ciN

N..-N

" 0

..-

Ln.0 0)Lt>

Lt>000

T"'

0

OM

, W

ON

N CO

C\lO

~O

a)C

\I (;')

I'-..NN

aio~

(D~

MO

~

~0 C\!

0 CD

:0

Q)

c:os.s

s:co

co-:5: ~

co

0 -

co..- c..

.-U)

.,cU)

'm'coco

Q)Q

)-Q)g.Q

)~

~~

~~

~

Lt')M~

~

Ll")

CX

)

CD

:(J)..- ~T

-

,~+JCf)

.~~.~.U-00.~~Q)

a.~0.~+J~~;js.~Cf)

Q)

~+J01~.~~;j-0C

f)~0.~+J~U0~~~-0Q)

Q)

rz...MQ

)~,Q~E

-I ,:>

,.-t~'":>

L!")NI

.c:U~~~1.0

-":3'0'\0'\.-I

.-~

N~

o.00u C

\IC

)~~

o .

~

.~00u C

O)

~~

"ft"00

U "to~0u

-.~

~

~

~""o

cCta

...' .0. .-ta

-~~

""-taQ

)ta>.Q

..o1/)~

.~.o~

cE

Q) Q

)-'ta.-

uu-fD

~

Q

) ta

~

tao~

ZD

..D..O

:O:(/)

-~ ~d

0)~M ~~

cx)~N

~0

..0..O

M

.0

0

- .

<0,

"'(')~

oooO

ON

O

"'co~co

N~

.0

..~

.~

O

0

Mo)(O

N

~.'0

OM

.

0

ro N

NIl)

..0O

M

.0

C\lM

M~

oM

~M

.

0 0

(..)

C¥)'O

t

~N

0u-MM

~ :~

0 0

U ~~0(,) ~0 C'f

><

0

CD

CD

~~

CD

'O

~.N

..

000

""'Il)~~

co .--."of"

..-0- lot');

M ~c;...,.

&I)

0);"ff"

Eco.C

o~

-.-0~ -

~

~

-=3:

.sC

) '-

cco 0

...' °.c .-'-

C-

O

~~

-co.cQ

»-c-(/) '"Q

)~

E.~

cE

Q)-Q

),

co'- (,)-~

-co~

CO

~.-

CO

>N

ID~

a..a..D:u)u)

~0u ('1M

~0<.>

(W)

~0u "to~0(.,)

~0 ~)(0

C')-

CD

.N .-C

D

CX

);C

")

"'-('\10)Q

) .~

.0 .

N "'-CD

M

~C")

0~ CO

;~~~.c~°

~

~

~

:s~

0Q).s

~

8-"'0-

~-

c..c°Q»-C

/) ~

<E

~.~

c:

Q) -Q

)I

'v ._(.)->

~~

.c~m

.-~>

omu~

c..~cncn

~"'":co

..-~0(.)

C\I

~0(..)

(W)

S:

0u 'ot-

~0(,)

-0 ('!

0

N CO

oq-C")

N

...00

oq-

..."':0 0>C

'!('II

-ro..;N

. .

r-.CO

It)N.'-

.M

~

.-0

CO

);C

O)

«)

ci

==.=~Q

) ~

0

~

2 -

.'5~

-3: oS

Q»=

0

c=

0I""

~

~

0....=

- ~

~

~-

o..E-Q

)=-

0..1/) -~

N-"'Q

) Q

)-cC

).-I/).=;E

Q)~

Q)

.A=

o=~

==

~.2=

3:N

(DG

~~

c..c..c..~U

)U)

~0u C'-j

~0u CO

)

~0u ~S:

0(.) ,...

c C'I

)(0

an;N o~-0 0)

~~ ~0.~ NC

'\!~

~

~~

co

>

co0

~

'-

~Q

) co-

>-c:o~

'-

II)-I:;

--C

O

~

.-CO

II)

CO

~co.c_Q

)Q)

Q)

1="C

~=

O~

.c,

c~~

co.-co>Q

UO

~c..~

(/»

M.

N (#)~('II

(#) .,...

.0 .

~

('II

,...""':0 ~~N 0>~C

O)- ~~0 -~N 0>

C'!

N

..-~0u C'I/

3:0(.)

(¥)

~0(.,)

~ca~

c:~

>-~

caoE.=

caEE

:E,cC

/)~

c:-~

Q)Q

)-I

ow

Co)

Co) -

Nca.c~

ca

Nm

UO

~~

(/)

~S:

0l> -0 C'I

C

CO

N.'"

..'..-N

..-M.

0

N r-.:0 It)C

f'J;N It')

CO

;N

C"I

~0(,)

-~0u (W)

~0<..>

~~0(..)

-~

'- I/)

~~

~

I/) co

Q)

0 ~

.~

~c

~C

) '-

~oQ

)Q)Q

).a ~

7>

-.~

.~

-Q)

-Q)

.-coco-U-oC

It) co~

~

co.- co>

o~~

~a..[k:(/»

C'I

0

..,-CO

)"C

O)

O\i)~ r-co~~

-It) ~

N

..0O

O~

CO

.

C'I!

CO

)

a)a>co.~ 0)r-.;M -~...

C'I

S:

0u (0)

~0u~~0u ~~0<..>

-~ C'I

6<

CO

)

0...-In;- ,...'"":-

C')~co.

~("II

"-:Q

)

0)0""':It) -q-C

)C"), q-O

IX) ...O

NO

co-,...~."1,...0

0) ...

0"10

«)~I/)

..qo ...o~

M ('II

a);0-

-cQ.)

~OJ-

~

~

3:Q

)CO

g,~

c3: ~

C'ta~

C>

~

~

'U)

~E

EQ

.)~.o'U

) "C

6"7

c co

.~

)( Q

.) Q

.) -Q.)

-co .r.

co .-u u

ro .r.

Nm

U~

~~

~U

)~

C\I

~0<..>

(W)

S:

0u ~S:

0<.>

~~~0(,)

C'I

C

N"'!0 a)

~0

"'co.'"~

.

~O

N

-cQ)

e!S

...11)~

II)

ootcoo~~

~

-!.co =

EQ

)Q)~

~

NN

(C'-'-X

~at)

co co

co .-co

Qm

~~

~(/) ."'":

It) ~r-..'0..-0'""':co... N

~'0~

.

~O