See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/222204767

Agroecosystem Analysis

Article in Agricultural Administration · December 1985

DOI: 10.1016/0309-586X(85)90064-0

CITATIONS READS

422 17,676

1 author:

Gordon Conway
Imperial College London
65 PUBLICATIONS 11,166 CITATIONS

SEE PROFILE

All content following this page was uploaded by Gordon Conway on 15 December 2017.

The user has requested enhancement of the downloaded file.

 AGROECOSYSTEM ANALYSIS

GORDON R CONWAY

ICCET SERIES E No 1 1983

Centre for Environmental Technology and

Department of Pure and Applied Biology,
Imperial College of Science and Technology,
London, SW7 ILU, United Kingdom

Imperial College Centre for Environmental Technology

Series E "The Dynamics of Environmental Systems"

Rep irts in this serie are working papers and are intended for discussion purposes
only. This report does not contribute a formal publication and should not be
quoted without permission of the author.

This report has been submitted for publication in

Agricultural Ecosystems Environment

Imperial College Centre for Environmental Technology

Imperial College of Science and Technology
48 Prince's Gardens
London SW7 ILU

83-E-1

-,I­
 CONTENTS

PAGE

ABSTRACT 3

INTRODUCTION 4

THE CASE FOR A HOLISTIC APPROACH 4

THE CONCEPTUAL BASIS 10

THE PROCEDURE OF ANALYSIS 18

OBJECTIVES AND DEFINITIONS

OBJECTIVES 23

SYSTEM DEFINITION 23

PATTERN ANALYSIS 26

SPACE 26

TIME 27

FLOW 32

DECISIONS 36

SYSTEM PROPERTIES 36

KEY QUESTIONS 39

RESEARCH DESIGN AND IMPLEMENTATION 42

DISCUSSION 43

ACKNOWLEDGEMENTS 46

REFERENCES 47

2
 In recent years there has been a growing demand for a more

multidisciplinary and holisric content to agricultural research

and development and for the formulation of methods by which this

can be acheived. Farming systems research and integrated rural

development are two responses to this demand but, in common with

other multidisciplinary approaches, they face the problem of

trying to encompass a breadth of expertise while at the same time

generating a common agreement on worthwhile practical ,%ction.

Resort to bureaucratic methods or to formal systems analysis is

unsatisfactory for a variety of reasons. The procedure of agro­

ecosystem analysis which is described and illustrated here steers

a middle course, combining a rigorous framework with sufficient

flexibility to encourage genuine interdisciplinary interaction.

This procedure has been designed and tested in several workshops

held in Thailand over the past five years.

At the heart of the procedure are the concepts of the

system, system hierarchies, system properties and the agro­

ecosystem. The participarts begin by defining the objectives of

the analysis and the relevant systems, their boundaries and

hierarchic arrangement. This is followed by pattern analysis, the

systems being analysed by all the participating disciplines in

terms of space, time, flows and decisions. These patterns are

important in determining the important system properties of agro­

ecosystems, namely productivity, stability, nustainability and

equitability. The outcome of the analyses are a set of agireed ke"

questions for future research or alternatively a set of tentative

guidelines for development.

Experience suggests that the procedure can be applied at any

3
time in a project's life, but is particularly useful at the

beginning when data are scarce. Ideally it should be repeated and

updaced at regular intervals.

INTRODUCTION

Farmers, of necessity, adopt a multidisciplinary, holistic

approach to their work and it would seem logical that this should

also apply to the design and implementation of ag icultural

development programmes in the less developed countries (LDC's).

Indeed m.ny programmes have approached their goals in this way,

but for the most part they have tended to focus on a limited

number of factors - hiqh yielding varieties of key food grains,

irrigation water, fertilisers, pesticides, the provision of

credit - which promise a quick and high return. It is only in

the last decade that there has been a significant demand for a

more multidisciplinary and holistic content to agricultural

research and development and for the formulation of methods and

procedures by which this can be achieved. In this pap r I report

one such procedure which has been developed and practised in

Thailand over the past five years.

THE CASE FOR A HOLISTIC APPROACH

Part of the justification for a more holistic approach to

agricultural development lies in the recent performance of the

agricultural revolution that has been taking place in the LDC's

since World War II. While real and gratifying increases in per

4
capita food production have been achieved (up by about 8% since

the early 1960's for the LDC's as a whole, FAO, 1977), the incre­

mental returns to the varieties and inputs on which the revo­

lution depends have begun to diminish. There are several reasons

for this. Since the beginning the adoption of the new !ochno­

logies has been highly uneven; the technologies were primarily

designed for the better favoured classes of farmer and for the

best endowed agroclimatic regions; they have been slow to spread

to the poorer farmers and the more marginal areas (IRRI, 1976).

Even when conditions are considered favourable, the gap between

performance on the agricultural research station and in the

farmer's field has proved highly persistent (IRR1,1978,1979a).

The new technologies have also been accompanied by a numbei of

serious short and medium term problems. These include increasing

incidence of pest, disease and weed problems (Nickel,1973;

IRRI,1979b), sometimes aggravated by pesticide use (IRRI,1980,

1981), deterioration in soil structure and fertility (Hauri,1974;

McNeil,1972), increased indebtedness and inequity (Collier,1977;

Collier et al.,1974; Murdoch,1980; Palmer,1976; Pearse,1980).

Finally the oil crisis of the mid 1970's generated soaring costs

of precisely those inputs on which incteased agricultural prod­

uction was becoming dependent. The potential returns relative to

the costs of inputs have become less dramatic and in many cases

inputs, even if still profitable, are no longer affordable by

farners with poor access to credit.

One answer aas been to tackle these various issues and

problems individually as they arise. However there has been a

5
growing realisation that many, if not all, of the problems are

essentially systemic in nature; they are linked to each other and

to the performance of the system as a whole. Moreover they arise

as a consequenze of fundamental incompatabilities between tle

existing agricultural systems and the newly introduced techno­

logical elements. It is, of course, almost axiomatic that revo­

lutions, agricultural or otherwise, involve such incompatabil­

ities. The size and magnitude of change is of the essence of

sticcess and undue preoccupation with problems and side-effects

may impede the realisation of the main objectives. Nevertheless,

recent experience suggests that the problems are often so great

that they directly threaten the main objectives themselves

(Conway and McCauley,1983). Even where there is some success in

attaining increased yields, the success may be short lived if

attention is not quickly diverted to side effects which threaten

other equally important development goals.

The last decade has also been characterised by the return of

large numbers of LDC agricultural scientists from postgraduate

studies in the induscrialised countries with, far too often, a

training which reflects the increasing degree of specialisation

that characterises much of modern agricultural education. As a

consequence they are often overwhelmed by the comple.ity of

agricultural 6 velopment issues and find it easier to treat

problers in purely disciplinary terms. in practice there is

often little or no interaction betweei. the agricultural discip­

lines, notwithstanding physical proximity within a university

faculty or agricultural research station. It is true that many

recent technological advances have been made by multidisciplinary

6
teams, but for the most part these were small teams, with narrow

goals and composed of a limited range of traditional agricultural

disciplines. Contemporary problems require teams which have a

broader perspective and draw on a wider range of disciplines, in

particular spanning the natural and social sciences.

The argument, it should be stressed, is not that we should

dispense wizh or by-pass specialisation. Given the complexity of

agricultural systems, anything more than a superficial understan­

ding requires the insight,; of highly trained specialists of all

kinds. Compared to the farmer the agricultural specialist has a

narrower range of knoledge, which is deeper and, within its

range, more complete. There is much that the specialist knows

that the farmer is ignorant of and could use. But where the

farmer tends to be superior, not least at the practical level, is

in those areas of knowledge where the specialist disciplines

overlap (Figure 1) and it is precisely in such areas that the

current problems of agricultural development lie.

One response has been the development of tarming ..

y,;tems

research ('SR) (Gilhert et al.,1990; Harwood,1979; Norman,1980;

Shaner et al.,1981). This focusses on the farm as the Ibasic

system for research and development and seeks, through the

sequential stages of analysis, design, testing and evaluation, to

discover and implement technological improvements that are appro­

priate to the farm's capacities and capable of overcoming its

constraints. This approach aro!;s primarily as a specific

response to the problems of small farms in the LDC's and is also

characterised by a strong involvement of the farmers themselves

at all stages in the process, capitalising on their knowledge of

problems and goals and their capacities as experimenters and

innovators.

A second approach has been that of integrated rural develop­

ment (IRD) (Conde et al.,1979; FAO,1975; Gomez and Juliano,1978).

This is even more holistic in scope but directed primarily toward

the development end of the research and development spectrum.

Its focus is the organisation of development projects which

deliberately go beyond a consideration of Zhe needs of improved

agricultural production to encompass such aspects as increased

fish, forest and handcraft production, more opportunities for

off-farm employment, and hetter provision of health, education

and other communal services. In practice IRD projects are

commonly seen as a means of improving coordination and better

working relations between different government agencies.

Both these approaches have proved of practical value,

although they are not without their critics. The major problem

they face, which is common to all holistic research and develop­

ment enterprises, is that of trying to encompass a breadth of

view and range of disciplines and talents while at the same time

generating a common agreemen on sorthwhile practical acion for

research and development. One solution has been to rely on

bureaucratic procedures and a hierarchic leadership of the

research or development team. This may be effici-nt but it is

likely to become rigid and non-innovative with time, losing the

valuable insights anc cross-fertilisation of ideas that is

possible with a multiiisciplinary group. An alternative is to

rely on formal techniques of systems analysis, using mat iematical

8
 FARMER'S KNOWLEDGE

DISCIPLINE A DISCIPLINE B

FIGURE Ov-rlapping )nowledge bhrtwoen farmers and research

qpecialists ( eq soil s:ientists and agricultural
ecoiomists). I1tched .nd striped areas indicate
extent o ke ' 'd r ,

9
or computer models. But this has the drawback of requiring

relatively specialised skills, so excluding the breadth of exper­

ience of the research and development workers whom the approach

is meant to involve and help.

The procedure I report here is intended to steer a middle

course. It is developed from tha basic concepts of systems

analysis and is thus rigorous and well focussed. Yet it is also

flexible in design and encourages wide and easy participation and

the flow of new ideas and insights. The procedure is not

intended an an alternative to FSR or 1RD but is offered as a

technique that can be used within the framework of these

approaches and, indeed, in any multidirciplinary research and

development programme whether the focus be the crop, field plot,

farm, village, watershed, or region.

In this paper I describe the underlying philosophy of the

approach and the details of the procedure and then qive some

examples of its application, drawn from several workshops held

recently in Thailand. Detailed reports un thoce workshop- are

published in Gypiantasiri et al. (1980) and KKU--Ford Cropping

Systems Project (1982a, 1982b). Summaries and a discussion of

the findings will be presentcd in two further papers in this

series.

THE CONCEPTUAL BASIS

The goal of multidisciplinary analysis is to achieve an

interaction between the disciplines that produces insights which

significantly transcend those of the individual disciplines.

10
Arranging the physical or social environment so as to promote

ease of communication among the disciplines is an essential

prerequisite, but the process of interaction may remain casual,

too often producing results that are superficial and mundane.

Experience suggests that the generation of good interdisciplinary

insights also requires organising concepts and frameworks and a

relatively formal working procedure which encourages and engi­

neers cross-disciplinary exchange.

The concepts have to be simple and basic and involve a

minimal set of assumptions that are acceptable to all the discip­

lines participating in the exercise. At the core of the pro­

cedure reported here is the concept of the system and the related

concepts of system hierarzhy, system properties and the agro­

ecosystem.

A system is here defined as an assemblage of elements coi'­

tained within a boundary such that the elements within the boun­

dary have strong functional relationships with each other, but

limited, weak or non-existent relationships with elements in

other assemblages; the combined outcome of the strong functional

relationships within the boundlry is to produce a distinctive

behaviour of the assemblage such that it responds to many stimuli

as a whole, even if the* stimulus is only applied to one part.

We can conceive of the natural living world as a nested

hierarchy of such systems (gene-cell-tissue-organ-organism­

population-community-ecosystem-biome-biosphere) each with a more

or less well defined boundary and a distinctive system behaviour.

It is assumed that systems higher in the hierarchy tend to con­

trol those b2neath them and, most important for the task of

11
analysis, that the behaviour of higher systems is not readily

discerned simply from a study of the behaviour of lower systems

(Checkland,1981; Milsum,1972; Simon,1972; Whyte et al.,1969).

For the three ecological systems (populations, communities and

ecosystems) the system behaviour can be disassembled into three

system properties: productivity, stability and sustainability.

These properties are relatively easy to define, although not

equally easy to measure. Productivity is the net increment in

numbers or biomass per unit of time. Stability is the degree to

which productivity is free from variability caused by small

disturbances inherent in the normal fluctuations of climate and

other environmental variables; it is most conveniently measured

by the reciprocal of the coefficient of variation in numbers or

biomass. Sustainability can be defined as the ability of a

system to maintain productivity in spite of a major disturbance

such as is caused by intensive stress or a large perturbation. A

stress is here defined as a regular, sometimes continuous, rela­

tively small and predictable disturbance, for example the effect

of soil salinity. A perturbation is an irregular, infrequent,

relatively large and unpredictable disturbance, such as is caused

by a rare drought or flood. Satisfactory methods of measuring

sustainability have still to be found, however. Lack of sustai­

nability may be indicated by declining productivity but, equally,

collapse may come suddenly and without warning.

In aqricultural development, ecosystems are transformed into

hybrid agroecosystems for the purpose of food or fibre produc­

tion. These too can be arranged in a hierarchic scheme (e.g.

12
 IRRIGATION vDYKE

Competi ;on ~DK

MICROORGANISMS

SOIL& WATER NUTRIENTS

Competition competition Competition

(% RIC ISH.
4-
WEDSL *C+ANTSS
u u

Herbivory

PESTS

Predation

NATURAL ENEMIES
V
FIGURE 2 Representation of a ricefield as a systerm showing key
elments and functional relationsh.ips

13
 HIGii LOW

PRODUCTIVITY

Time Time

STABILITY

Time Time

Perturbation Perturbation

Time Time
SUSTAINABILITY

Stress Stress

Time Time

EQUITABILITY o.

M.

Income Income

FIGURE 3 The system propeities of agroecosystems

14
field-farm-village-watershed-regi on). The transformation

involves several significant changes. The systems become more

clearly defined, at least in terms of their biological and

physico-chomical boundaries. These become sharper and less per­

meable; the linkages with other systems are limited and chan­

nelled. The s jtems are also simplified by the elimination of

many species and various physico-chemical elements. A good

example is the ricefield (Figure 2): the water-retaining dyke or

bund forms a strong, easily recognisable boundary, while the

irrigation inlets and outlets represent some of the limited

outside linkages.

The important system properties remain essentially the same,

except that productivity is now measured in terms of yield or net

income (Figure 3). In the ricefield productivity is in terms of

rice, fish and crabs, but the important functional relationships

which determine this property remain essentially ecological in

character, involving competition, herbivory and predation.

However, at the higher levels of the agroecosystem hierarchy

the inclusion of human beings, their social, cultural and econ­

omic activities, reintroduces considerable complexity, but of a

different nature. Figure 4 orders the components of a typical

LDC farm. Some of the important functional relationships remain

ecological or, at least in dynamic terms, are analoqous to ecolo­

gical processes. For example, forms of competition, mutualism

and predation can be discerned not only in the natural inter­

actions but also in the socio-cultural and economic interactions.

An important new system property consequent on the inclusion of

human beings is equitability. This expresses how evenly the

15
 products of an agroecosystem are distributed among its human

beneficiaries; the more equitable the system the more evenly are

the agricultural products shared among the members of, say, a

farm household or a village. Equitability can he readily des­

cribed using statistical distribution parameters (Figure 3).

In natural ecological systems the operation of natural sele­

ction on the reproductive success of individual organisms favours

numbers, and hence the productivity of ecological systems, and

survival, and hence sustainability. It may or may not select for

stability, as here defined. In agroe-osystems human manipul­

ation, operating on both individuals and whole systems, partly

replaces natural selection. Different system properties are

favoured according to predominant human goals and values. It is

thus possible to view agricultural development as a progression

of changes in the relative values of the important system

propezties.

In general, traditional agricultural systems such as swidden

cultivation (shifting cultiviaiton) have low productivity and

stability, b,,t high equitability and sustainability (pattern A in

Table I). Traditional sedentary cropping systems tend to be more

productive and stable, yet retain a high degree of sustainability

and ,:ome of the equitability (B). However the introdution of new

technology, while greatly increasing the produczivity, is likely

to also lead to lower values of the other properties (C). This

was particularly true, for example, of the introdution of the new

high yielding rice varieties, such as IR8 and its relatives, in

the 1960's; yields fluctuateO widely, but have tended to decline,

16
 U PRODUCE
Innovations/Skill SOCIAL TRAFFIC

F mily Crafts/
P roduc e mark ets \ Cap ital I K in - . . .. .
Off-farm employent \ Labour 'eighbour
Input markets \ Knowledge a
ehb
- -- u _ ' 1
Consumer goods FAR ---- co peatv
Temple
Soils Crp Government e -activity/
Topography Livestoc
/ Water Fish \/
/Timber Welfare /
~~Fertilizers Crops _
~Pesticides Livestock /
~Seed,Watr Fish /
kk Machinery Timber /

?IGURE 4 Comonents of a typic&l less developed country farm

±ndicotinq linkages with other s'stens

17
 in part due to growing pest and disease attack. More recent

varieties combine high productivity with high stability, but

still have poor sustainabi]ity (D). The ideal goal is probably

pattern E or on marginal lands, where there is a conflict between

productivity and sustainability, pattern F may be more

appropriate.

THE PROCEDURE OF ANALYSIS

The procedure has evolved over the past five years from one

originally designed for ecosystem analysis (Walker et al. 1978).

It rests on the concepts described above and on five further

assumptions:

1. It is not necessary to know everything abcut an agro­

ecosystem in order to produce a realistic and useful analysis.

2. Understanding the behaviour and important properties of an

agroecosystem requires knowledge of only a few key functional

relationshios.

3. Producing significant improvements in the performance of an

agroecosystem requires changes in only a few key management

decisions.

4. Identification and understanding of these key relationships

and decisions requires that a limited number of appropriate key

questions are defined and answered.

5. Since there is, as yet, no easy guide to defining key

questions or identifying key relationships and decisions, the

most productive approach is for a multidisciplinary team to

18
 OBJECTIVES

1p
SYSTEM DEFINITION

PATTERN ANALYSIS

AND

SYSTEM PROPERTIES

IMPLEMENTATION

FIGURE 5 Basic steps of the procedure for aqroeco,ystem analysis

19
Table I

Agricultural development as a function of agroecosystem

properties

Productivity Stability Sustainability Equitability

A Swidden LOW LOW HIGH HIGH

rultivation

B Traditional
cropping MEDIUM MEDIUM HIGH MEDIUM
system

C Improved HIGH LOW LOW LOW

D £mproved HIGH HIGH LOW MEDIUM

E ?Ideal HIGH MEDIUM HIGH HIGH

(best land)

F ?Ideal MEDIUM HIGH HIGH HIGH

(marginal
land)

20
 attempt to collectively deFine questions through a process of

system definition, analysis of system patterns and the invest­

igation of system properties.

The basic steps of the procedure are described in Figure 5.

Experience has shown that the procedure is best followed in a

seminar or workshop environment, in which meetings of the whole

team are interspersed with intensive work sessions involving

small groups of individuals. Although the first workshop in

Thailand (Gypmantasiri et al., 1980) ran intermittently for a

period of a year, more recently they have been confingd to one

week, but with a month-long preparatory period for data

acquisition. Table II describes an appropriate timetable.

The key to success lies in clear communication between the

different disciplines present. In the Pattern Analysis phase, in

particular, it is importanz for the Participants to strive to

present their disciplinary and specialist knowledge in such a

fashion that all other members of the workshop can easily grasp

its significance. This process is greatly helped by the use of

diagrams and in the workshops extensive use has been made of

maps, transects, graphs, histograms, flow diagrams, decision

trees, venn diagrams and any other pictorial device that appears

to aid communication. One practical, but essential, requirement

is for the workshop room to be well ecaipped with overhead pro­

jectors, transparencies, pin boards, graph paper etc.

21
 Table II

Timetable for a week-long workshop of agroecosystem analysis

Day 1 Participant introductions.

Conceptual basis and details of procedure.
Examples from previous workshop results.

Day 2 System Definition by whole workshop team.

Break into sub-groups, each assicned a level in
the system hierarchy (e.g. field plot-farm-village
-region or one of a series of agroecosystems
(e.g. different farms or villages). Each group
carries out Pattern Analysis and Key Question
Identification.

Day 3 Continuation of Day 2. (Brief whole team

meetings if necessary).

Day 4 Field visits to agroecouystems.

Day 5 Presentation by subgroups of findings.

Whole team discussion of Key Questions and
Research Design and Implementation.

Day 6 Continuation of Day 5 as necessary.

Writing of draft report by editcrial team.

Day 7 Completion of draft report.

22
OBJECTIVES AND DEFINITIONS

Objectives

As in all exercises in systems analysis the quality of the

final resutls depends crucially on a having a definition of

objectives at the outset which is couched in simple, precise and

unambiguous language and is acceptable to all members of the

team. Recent workshops have had objectives of the form:

1. To identify research priorities that will lead to

improvements in the level and stability of net income of farm

households in the x region.

2. To identify tentative guidelines for improving agricultural

productivity of the poor farmers in y village.

Precise definition of targets is crucial. For example, is the

objective to improve mean agricultural productivity of an area,

_r the productivity of the poor farmers in the area (the former

may not imply the latter)? Also is the aim to increase producti­

vity only, or is improved stability, sustainability or equitabi­

lity to be explicitly included?

System Definition

This phase involves identification of systems, system boun­

daries and system hierarchies.

At the outset the identificalion of systems and their boun­

daries is subjective and tentative. The biological and chemico­

physical boundaries are often fairly clear: the ricefield is

23
 bounded by a dyke; the valley by the cxtent of the watershed.

nut the cultural and socio-economic boundaries ar, r,ore

elue;ive.
For example, defininc a farm household solely in terms of the
farm itself - th.- land that is cultivated or sth,=rwise rxooited

- is frequently inadequate. A mc,mber of tr F:rm household may

be deriving income from- f-r away; the sale of rrodrce ma, depend
on distant markets and the farmer's goals and vale may he
influenced by politii l or reliqious movements of a complex:

origin. In Northaast Thailand .-embers of the- Familv way ho

working ter-.,orari]v in S-,di Arabia; she price of a major crop,

cassava, is influenced by quota,,; estahli -hd 'v tho European

Economic Community and the velu-s of Buddhist farmers m,-y be
influenced by religious ve'sa.ita in Sri L.inka. The ',rlswr i3

to translate these, as far as )o-sible, into physical or qeoqr­

aphic terms and to elaborate system hierarchies that link or

combine systems whose boundaries are defined in different terms

(Figure 1;.

The systems and boundaries can be revised as the workshop

proceeds, particularly in the light of a grow,;ing understanding
of
the system properties since the extent and quantificatios
of the
important functional relationships contributing to these proper­
ties will provide nore objective criteria upon .:hich to draw
the
boundaries. The :,rcdIura of analysis will als;o indicate which
siystems are strong in terms of their relevance to the nbjectives

of the workshop -nd increasingly only these will b analysed in

any detail.

24
 WORLD INTERNATIONAL MARKFT

NATION

PROVINCE LOCAL MARKET

ECOSYSTEM WATERSHED I
I TRIBE DISTRICT MIDDLE MAN

IRRSTIOH I I
KIN GROUP VILLAGE COOPERATIVE

CO.DINITY FARM FAhILY FAMILY FARM

POPULATICN FIELD

ORGANISM CROP PLANT

ORGAN

CELL

GENE

NATURAL AGRICULTURAL COMIUSAL POLITICAL ECONOMIC

FIGURE 6 System hierarchies in a less developed country (Conv;.y 1983)

25
 PATTERN ANALYSIS

Four patterns are chosen as likely to reveal the key func­

tional relationships that determine system properties.
Three of

these - space, time and flow - are knoan to be important in

determining the properties of ecological systems (May,19Hl). All

three are significant factors in productivity. Variability or
heterogeneity in space is also an important promoter of
stability
and sustainability (e.g. predator-prey system, , Hass;o 11,197F).

Variability in time, however, can be destabi lisino; eystems with

long time lags are often very unstable (Mtay, 1975). Flows are
either stabilising or destabilising depending on .h. ther the
feedback loops are negative or positive (Levins,197.).

All three patterns also have the vi rtue of being neutral

with respect to scientific disciplines. Space, time and fiow a-o

equally important patterns for both natural and social science

analysis and hence provide a basis for the gmeration of cross­

disciplinary insights. The fourth pattern - decisions - arises

directly from the need to understand the consequences of human

manipulation and to identify those decisions which bear most

significantly on agroecosystem sustainability and equitability.

Although this pattern is primarily the ooject of socioesonomic

analysis, experience shows that it generates lively discussion

among both social and natural scientists.

Space

Spatial patterns are most readily revealed by simple maps

and transects. Overlay maps are useful in uncovering potentially

important functional relationships. Thus in the Chiang Mai

26
 Valley of Northern Thailand they indicated that cropping inten­

•sity was more a function of the farm irrigation system rather

than soil type (Figure 7). Subsequent analysis of the pattern of

decision making in irrigation in the Valley suggested that triple

cropping is likely to be more reliable in traditiona'. and tube or

shallow dug well systems because farmers exercise greater control

and hence the water supply is more reliable.

Transects are particularly useful in identifying problem

areas and the impcrtant spatial relationships both between and

within farms. In the analysis of Northeast Thailand agroeco­

systems the recognition of the mini-watershed agroecosystem and

its subdivisions pin-pointed the role of the upper paddy fields

as the generator of instability in rice production (Figure 8).

Time

Patterns in time are best expressed by simple graphs. Threi

patterns appear to be important for agroecosystems. The first is

that of seasonal change and can be analysed by the superimpos­

ition or simultaneous graphing of cropping sequences, labour,

credit peahs, prices etc. on various agrometeorological para­

meters. This helps, in particular, to idenfity those periods in

the year where the timing of operations and the availability of

resources is critical for productivity and stability (Figure 9).

Longer term changes, in prices, production, climate, demog­

raphic parameters etc., can be graphed in a conventional manner

(10 years of data is minimum requirement). These reveal trends

in productivity and a measure of stability (Figure 10), possible

27
 Single cropping
I Rainfed
RID systems
Double cropping
L systems
Traditional
Triple cropping
Tube wells
i Dug wells

a. b.

FIGURE 7 Spatial patterns in the Chi3nq Mij Valley, 'h - (, roppiv- -nten-ity, (b)
,RID) and non-government irr~rnwion -ys~teis governr.ient
al. 19R0)
 Hamlet
\X\. -Field

shelter

Upland Upper Paddy Lower Paddy Upper Paddy Upland

Soils Paleustult Paleaquult/ Paleaquult

Paleustulc

Crops Cassava Cassava Rice

Kenaf Rice followed by
Sugarcane vegetables
Water melon

Problems Drought Insufficient Occasional

Erosion water flooding

FIGURE 8 Transect of a min'watershed in Northeast Thailand

(KKU-Ford Cropping Systems Project 1983a)

29
 300- rainfall
-- *v P- apot.silalion

- 200
,E,
"' "
too
" -- - -.
.----.
. . . .

10

21100n KENAF NAF

Glutinous rice

2­
o nn n n
Tobacco

Sso80
40

20

aKenaf

4
2

MA M JJ A SO0N D J FM A MJ J A S

FIGURE 9 Seasonal pa'.tterns of rainfall, crop-inq and

labour demand for an area of Northeast T;ahiland
(KKU-Ford Crop[-inq, Systems Project 1982a)

30
 10-
10

8­

0
t--6
o 6-6

00
.o
•o 7
oZ ­
0 'Regional rice requirement

1960 1462 19'64 1966 19'68 19'70 I92 19'74 19'76 1978
Year

FIGURE 10 Rice production in Northeast Thailand (KKU-F-d

Foundation Cropping Systems Project 1982a)

31
time lags in the system and other causes of instability (Figure

11) and any signs of lack of sustainability (Figure 12).

The final pattern in time is of the response of important

variables to stress and perturbation. Stresses, ar defined ear­

lier, include soil deficiencies and toxicitie2s, posts, diseases

and weeds etc. Perturbations include major floods or droughts or

a sudden outbreak of a pest or disease. The distinction between

the two forms of disturbance rests on the degree of predict­

ability. In some regions of the world, for example in Northeast

Thailand, floods and droughts are so common as to constitute

stresses; in Northern Thailand where wet season rice pests are

relatively unimportant an outbreak of a new pest, such as the

brown planthopper (Nilaparvata lugens) would constitute a per­

turbation. In the analysis actual and possible stresses and

perturbations need to be identified and the known or likely

responses of the variables graphed (Figure 13).

Flow

Included under this heading are the patterns of flows and

transformations of enerqy, materials, money, information etc. in

the agroecosystems. While these may be described by conventional

flow diagrams the aim should not be to trace out all the detailed

relationships. Flows should be principally analysed for the

major causes an effects and for the presence of stabilising or

destabilising feedback loops. The flow diagrams should thus be

kept as simple as possible (Figure 14). Tables, matrices, bar

histograms (Figure 15) and regression graphs may also be useful

in indicating important relatiornships.

32
 o-ci Cassava
7- o---0 Kenaf

- Maize

!-
'5­

CL 4"

0­

3­
,0

, 2­

0-1 3 175 19 77 1979

Year

FIGURE 11 Annual fluctuations in price and planLed area for

major crop- in Northeast Thailand (22 baht = US$1
ap'rc-"; 1 rai. = .16 ha) (! KU-Ford Cropping
Systrms Projocc 1982a)

33
7.0

6"0

40­
" - I II I I I I I

1969 71 73 75 77 79
Years

rIGURE 12 Declining rice yields under intensiv-i cropping cn

iiresearch staLion in Northern Thailand
(Gypmantasiri el al. 1980

34
 6"5 CROP I II CROP 2 I I CROP3

rice- soybean- mungbean

SOIL rice-soybean- fallow

FLOODED

pH ,, rie-a
,c alo

5.5

•
* I
* I

4.5 ',

4.0

AS O N D J F A M J J A SON D

FIGURE 13 Fluctuations in soil acidity under three cropping

systems in Northe':n Thailand (plH measured 4n 0.01M
CaCd ) (Gypmantas~ri et al. 1980)

t I+
- -.-.-.Urban Poorjt - -
- I I

+ - I
+ I

Early + Planted + Labour - I

rain area Ric-/
capita

+ Production
per farm

Late + Yield FIGURE 14 Flow diagram of rico

rain production, economics
and labour relations for
Northeast Thaiilnd

35
 Decisions

Decisions, ranging from those of national agricultural po­

licy to the individual farmer's day-to-day choices,

occur at aIll
levels in the hierarchy of agroecosystems. Two patterns are
important. The first is of the choices made in a given agro­
ecosystem under differing conditions and is best drscribed by
means of a decision tree. Construction of the tr. helps to
reveal both the goals of the decision maker and the constraints

on choice that are present in the agroecosystem.

Decision trees

produced for Northeast Thailand agroecosystems suggested the

importance of labouL irAd land type constraints cn farm and vil­

lage production (Figure 16).

The second pattern is of the spheres of influence of

deci­
sion makers. Here analysis is primarily required in order to
identify the critical decision makers in the system
hierarchy and
simple diagrams are useful in distinguishing the points
of con­
tact and overlap in decision making. Analysis of irrigation
water control in the Chiang Mai Valley, for example, reveals the

extent of farmer participation in decision making

under different

systems (Figure 17).

SYSTEM PROPERTIES

Discussion of system properties should guide the form of

pattern analysis, helping to indicate the likely key relation­
ships and decisions. However at the end of the pattern analysis

phase it may be useful to summarise what has been learnt of

system properties and to tabulate the most important
contributing

36
 l Net farm income: R= rice, C= cash crop
[II Trade, home industry, etc.
D] Off- farm income

26,000 Income from livestock

22,000­

18,000-

Ec 14, 000­

0
U
S 10,000­
. c

6,000 R R R R R R R R R R R

2,000

Villages

FIGURE 15 Components of far-i income for 16 adjoininq

villic,?s in Northeast Thailand (22 baht
US$1 zpi)-ox.) (KKU Ford CropPinq Sstomn
Project 1982h)

STRATGIES

Io, hi11hquality No

TOBACCO OR OFF-FAR

NoYe

RIFETobacco
Ooes rice crop grou
.ndr
teet
cntric,
subsistence?

Yea TOBACCO OR WATEKMLON

oes soil require
improvement?

YesRice-w~te
e IAuon-bend
Ys LATheNLOs H - sroNntontinusly
Does dIsLa,,.
bud p?
Ye. Rotatiion )f rice- rt,-n
j
I'll lo-d by too or,,
11. ly

FIGURE 16 Decision tree For farviinr str,.eqirs irl on area

of Northeast Thai]and

37
 a.
Catchment

RID [lead office

Dam 6 laterals"'
Project Engineer

Water MIasters

- - --- -Zonemnen
-

C'ate Tenders

-. Waters users assocn.

Sublateral Common irrigator

Farm ditch Field ditch

Farmer

Catchment Annual meeting of

muang leaders

Dam i t " "

laterals . ....
&

Village headman

Farm ditch Fa me r
Field ditch
FIGURE 17 Diarriim so:'iC points of rontaclt ,idovorl;ip in
.rrigtioi Ceuision ,ckrtj North,ri 'Thaiilond:
(a) 4ov,.njent (R.D) ,.ystoms; (b) tririt.Jonal
nystems (.n each diagram the physical systems are
.;nthe left and the deciior. making systems on the
riqht).

36
 relationships and variables (Table III).

KEY QUESTIONS

Key question3 arise throughout the procedure, during system

definition, pattern analysis and the discussion of system pro­

perties. They should be noted down as they emerge and then

collectively revised by the members of the subgroup in the light

of all the information available. Experience suqgests that a

field trip to the agroecosystem sites is useful at this stage:

some questions may be quickly answered; others may be revealed as

poorly based or inappropriate. The full list of key questions

should then be extensively discussed by the workshop team as a

whole.

Experience shows that the questions take a variety of forms.

In many, if not most, situations the analysis is likely to reveal

broad gaps in knowledge which require further investigation to

uncover suspected key functional relationships. A typical gap

question is:

"What are the most appropriate meteorological parameters for

characterising the agricultural seasons in the Valley?"

(In the Chiang Mai Valley farmers appear to recognise at least

four distinct cropping seasons but their meteorological defini­

tion is not clear.)

However most questions will be more narrowly defined, focus­

sing on suspected key processes or decisions. For example:

"Can new rice varieties be bred to produce more stable yields on

the upper poorly watered paddy fields?"

39
 Table III

Examples of key relationships and variables

determining the system properties of agroecosystems of
Northeast

Thailand.

PRODUCTIVITY
Demand by world markets (especially EEC)
Government rice and fertilizer policies
Water resource development

STABILITY
Rainfall, especially floods and droughts
Rice production in upper paddy fields
Human migration
Diversification of production

SUSTAINABILITY
Increasing salinity
Increasing indebtedness
Deterioration of communal mutual help arrangements

EQUITABILITY
Subsistence rice crop
Diversification of production
Government rural works programme
Availability of credit

40
 (The high instability of rice yields in Northeast Thailand is
largely a function of poor performance in the upper elevation

fields and of the mini watersheds.); or

"What is the optimal application of fertilizers to traditional

rice varieties under highly variable rain-fed conditions?"

(In Northeast Thailand rainfall is highly variable and

it is not
clear that encouragement of higher fertilizer use would
produce a

reasonable return to the farmer or to the region.)

Many important questions span different levels in the

system

hierarchy. For example:

"How is the form and productivity of crcpping systems in the

Chiang Mai Valley affected by government policy on the price
of

rice?"

(Various government price policies essentially mean

that farmers
get a relatively poor retu.n for rice, the basic crop of most

existing cropping systems.)

However in our experience the most powerful questions are

those that directly address system properties and in particular

the actual or potential trade-offs between them:

"To what extent are the gains in productivity and stability

from
land consolidation in the Chiang Mai Valley likely to be
offset
by a decline in sustainability and equitability?"

(Land fragmentation in the Valley, although promoting ineffic­

iency, seems also to encourage crop diversity and hence
sustaina­
bility. It is also probably more equitable.) Questions of this

kind also often implicitly raise doubts about the conventional

wisdom.

41
 Where the object of analysis is to identify possible
ways of
developing an agroecosystem the key questions may be framed in

the form of tentative guidelines:

"It is likely that crop production in village x will

be signif­
icantly improved by the provision of bctter quality second
-rop

seed."

(Under intensive rotational cropping good establishment

of the
second crop following rice is critical to success.) Although
written in this form, t.ie implicit question aod hypothesis are
apparent. If better quality seed is provided it should be seen
strictly as an experiment and the results used to modify the

overall analysis.

As far as possible the key questions should be turned

into
carefully phrased, test hypotheses so that by the end of the
workshop there is a list of questions each accompanied by a
hypothesis, a discussion of the issues involved
and some indicat­

ion of the investigations now required.

RESEARCH DESIGN AND IMPLEMENTATION

The remaining phase of the procedure is one of

conventional
research. The hypotheses are tested as appropriate:
by laboratory

or field experiments, field surveys or extension trials, or by

development trials in which guidelines are enacted
and assessed.
The multidisciplinary _ctivity of the workshop may or may not
extend into the research phase; many of the key
questions will be

phrased in terms of single disciplines and are best answered

by
the appropriate specialists. T-, this extent the outcome of the
workshop may appear superficially similar ts research programmes

42
 arising from a collection of individual initiatives, but will

crucially differ in that the individual research projects are the

direct consequence of a multidisciplinary systems analysis and

the results feed back to and modify that analysis. The research

has thus a better contextual basis and is likely to be more

appropriate and relevant, while the results have a greater chance

of being acted upon.

It is, of course, not necessary that all the key questions

be tackled by the workshop team. Some of the questions may raise

issues or require methods of approach that lie outside the compe­

tence of the group But if the questions are well phrased and

their importance clearly justified they should interest nd

excite other research workers to find answers.

DISCUSSION

Figure 18 shows the final detailed form of the procedure.

The arrowed lines connecting the various stages indicate that the

procedure is intended to be itera-ive. New knowledge and perspec­

tives at each stage are likely to require revision of earlier

stages; in particular answers to the key questions when they are

found will modify earlier assumptions, updated at regular inter­

vals. Experience suggests that the procedure can be applied at

any time in a project's life but it is particularly useful at the

beginning of a project when data are scarce. Ideally it should be

repeated and updated at regular intervals. Phe first run­

through of the procedure is likely to produce more "gap" than

43
 SYSTEH OUNAI
DEFINITION CIIY

SPACE4- ---------- - I E

PATTERN
ANlALYS
IS

FLOW<
4- - ECI SI 01S

IZY
QUESTIONS

GUIDELINES

HYPOTHCSES

DESIGN
AND LABORATORy i:XI-L:Rl . ILTS
IMPLEMENTATION FIFLD EXPERIMENTS
F I ILLDSUKVEYS
EXTENSION?TRIALS
OLVELOIPHLNT[:XI'ERI.,fE [S

FTGURE 18 Th' fu.ll prou. dure of : Crccosystem analysis

4
.'.
 true key questions; but with time and answers the original ques­

tions will be replaced by better, more precisely focussed ques­

tions.

It should be stressed that the procedure is intended to be

flexible. Our experience is that it changes in quite important

respects from workshop to workshop, depending on the objectives

of the workshop and the background, experience and interest of

the participants. New ways of presenting, analysing and interp­

reting information should be actively encouraged.

Despite its foundation on the concepts of systems analysis

the procedure makes no explicit mention of the role of mathem­

atical models. We have deliberately avoided tne conventional

approach of using a large scale simulation model as the focus of

analysis. This is partly because many individuals may be exc­

luded from the analysis through a lack of skill or inclination

to interact with the model, and partly because in such large

scale modelling exercises the key issues and questions tend to he

obscured by a preoccupation with the details of construction.

Nevertheless it is clear that the potential for use of a wide

variety of models (matrix models, regression, linear programming

models, simulation models etc.) exists throughout the procedure,

for example in the analysir of the patterns of time, space, flow

and decisions and of the dynamics of system properties, in formu­

lating key questions and hypotheses and, indeed, in some cases in

answering these questions. The extent to which models will be

used in practice depends on the interests and skills of the

workshop participants.

45
ACKNOWLEDGEMENTS

the workshops have involved a total of well over a hundred

people. Those most involved in the development and refining of

the procedure are Dr. Terd Charoenwatana, Mr. lain Craig, Ms.

Laxmi Ganjapan, Dr. Terry Grandstaff, Mr. Phrek Gypmantasiri,

Dr.

Rapeepan Jaisaatd, Dr. Viriya Limpinuntana, Dr. Aran Patanothai,

Dr. Benjavan Rerkasem, Dr. Kanok Rerkasem, Dr. Manu Seetisarn,

Ms. Nongluk Suphanchaimat and Ms. Aree Wiboonbongse. Financial

support for the workshops was provided by the Ford Foundation

c.nd

in one case by the US Agency for International Development.

46
 REFERENCES'

Checkland, P.,198l. Systems Thinking, Systems Practice. John

Wiley & Sons, Chichester.

Collier, W.L.,1977. Technology and peasant production: a discu­

ssion. Development and Change 8: 351-362.

Collier, W.L., Soentoro, Wiradi, G. aid ilakali,1974. Agricul­

tural technology and institutional change
in Java. Food Research

Institute Studies 13: 169-194.

Conde, J., Paraiso, M.J. and Ayassou, V.K.,1979. The Integrated

Approach to Rural Development Health and Population.

Development
Centre Studies, Development Centre of the
O.E.C.D., Paris.

Conway, G.R.,1983. Applying Ecology. Centre for Environmc.*tal

Technology, Imperial College of Science and

Technology, London.

Conway, G.R. and McCauley, D.S.,1983. Intensifying tropical

agriculture: the Indonesian experience. Nature 302: 288-289.

F.A.0.,1975. Integrated Rural Development. Food & Agriculture

Organisation, Rome.

F.A.0.,1977. The Fourth World Food Survey. Food & Agriculture

Organisation, Rome. Statistic Series No.11, Food & Nutrition

Series No.10, 128 pp.

Gilbert, E.H.,Norman,D.W. and Winch, F.E.,1980. Farming systems

research: a critical appraisal. MSU Rural Development

Paper no 6.
Department of Agricultural Economics, Michigan State University,

47
 East Lansing, Michigan.

Gomez, A.A. ani Juliano, P.A. (Editors),1978.

Rural Development:
The Philippine Experience. The Philippine Training Centers for
Rural Development, University of the Philippines at Los Banos,

College, Laguna.

Gypmantasiri,P.,Wiboonpongse,A.,Rerkasem,B.,Craig,I.,Rerkasem,K.,

Ganjapan,L., Titayawan,M., Seetisarn,M., Thani,P., Jaisaard,R.,

Ongprasert,S.,Radanachaless,T., and Conway,G.R,1980. An Inter­

disciplinary Perspective of Cropping Systems in the Chiang
Mai
Valley: Key Questions for Research. Faculty ot Agriculture,
Uni­
versity of Chiang Mai, Chiang Mai, Thailand.

Harwood, 'R. R.,1979. Small Farm Development: Understanding and

Improving Farming Systems in the Humid Trcpics.

Westview Press,

Boulder, Colorado.

Hassell, M.P.,1978. The Dynamics of Arthropod Predator-Prey

Systems. Princeton University Press, Princeton.

Hauri, I.,1974. Le projet cerealier en Tunisie: Etudes aux

niveaux national et local. UNRISD Report No.74.4, Geneva.

International Rice Research Institute,1976. Research highlights.

In: Annual Report for 1975. IfRI, Los Banos, Philippines,

pp 1­

3.

International Rice Research Institute,1978. Research highlights.

In: Annual Report for 1977. IRRI, Los Banos, Philippines, pp

xiii-xv.

48
 International Rice Research Institute,1979a. Farm-Level Con­

straints to High Rice Yields in Asia: 1974-1977. IRRI, Los Banos,

Philippines.

International Rice Research Institute,1979b. Brown Planthopper:

Threat to Rice Production in Asia. IRRI, Los Banos, Philippines.

International Rice Research Institute,1980. Control and mana­

gement of rice pests. In: Annual Report for 1979. IRRI, Los

Banos, Philippines, pp 202-207.

International Rice Research ]iiintitute,1981. Control and mana­

gement of rice pests. In: Annual Report for 1980. IRRI, Los

Banos, Philippines, pp 202-203.

KKU-Ford Croppig Systems Project,1982a. An Agroecosystem Ana­

lysis of Northeast Thailand. Faculty of Agriculture, University

of Khon Kaen, Khon Kaen, Thailand.

KKU-Ford Cropping Systems Project,1982b. Tambon and Village Agr­

cultural Systrm in Northeast Thailand. Faculty of Agriculture,

Universisty of Kho:a Kaen, Khon Kaen, Thailand.

Levins, R.,1974. Qualitative analysis of partially specified

systems. Annals N.Y. Acad. Sci. 231: 123-138.

McNeil, M.,1972. Lateri ic soils in distinct tropical environ­

ments: Southern Sudan and .razil. In: Farvar, M.T. and Milton,

J.P. (Editors), The Careless Technology: Ecology and Inter­

national Development.The Natural History Press, Doubleday and

Co.,New York, pp 591-608.

49
 May, R.M.,1975. Stability and Complexity
in Model Ecosystems (2nd

Edn.). Princeton University Press, Princeton.

May, R.M. (Editor',1981. Theoretical Ecology: Principles and

Applications. Blackwell Sci. Publns.,
Oxford.

Milsum, J.H.,1972. The hierarchical basis for general living

systems. In: Klir, G.J. (Editor). Trends in General Systems

Theory. Wiley, New York.

Murdoch, W.W.,1980. The Poverty of Nations: The Political Eco­

nomies of Hunger and Population. John Hopkins Press, Baltimore.

Nickel, J.L.,1973. Pest situations in changing agricultural

systems - a review. Bull. Ent. Soc. Am. 19: 136-142.

Norman, D.W.,1980. The farming systems approach: relevance for

the small farmer. MSU Rural Development Paper No.5, Department
of Agricultural Economics, Michigan State University, East

lansing, Michigan 48824.

Palmer, I.,1976. The New Rice in Asia: Conclusions from Four

Country Studies. UNRISD Report No.76.6, Geneva.

Pearse, A.,1977. Technology and peasant production:

reflections
on a global study. Development and Change, 8: 125-159.

Pearse, A.,1980. Seeds of Poverty, Seeds of Want: Social and

Economic Implications of the Green
Revolution. Clarendon Press,

Oxford.

5O
 Shaner, W.W., Wilson, A.G. and Wilson, D.(Editors),1969. Hierar­

chical Structures. Elsevier, New York.

Simon, H.A.,(1962. The architecture of complexity. Proc. Amer.

Phil. Soc. 106: 4C7-482.

Walker, B.H., Norton, G.A., Barlow, N.D., Conway, G.R., Bailey,

M. and Comins, H.N.,1978. A procedure for multidisciplinary

ecosystem research with reference to the South African Savanna

Ecosystem Project. J. Appl. Ecol., 15: 481-502.

Whyte, L.L., Philipp, P.F. and Schmehl, W.R.,1982. Farming Sys­

tems q:search and Development: Guidelines for Developing Coun­

tries. Westview Press, Boulder, Colorado.

51

View publication stats