Sunteți pe pagina 1din 24

Computers, Environment and Urban Systems

26 (2002) 89–112
www.elsevier.com/locate/compenvurbsys

Expert systems and GIS: an application of land


suitability evaluation
S. Kalogiroua,*
a
Department of Geography, University of Newcastle, Daysh Building, Newcastle Upon Tyne, NE1 7RU, UK

Accepted 13 September 2001

Abstract
In this paper expert systems and geographical information systems technologies are com-
bined to help with an implementation of a land suitability evaluation model. The result is the
LEIGIS software, which allows empirical work within the framework of this paper. The
model used is based on the FAO land classification for crops, and data which describe an
agricultural area in terms of soil mechanics and environment. The land evaluation has two
parts; the physical evaluation and the economic evaluation. For the physical evaluation of the
land, data for 17 land characteristics have been used and a Boolean classification method has
been applied. The implementation includes models for general cultivation and five (wheat,
barley, maize, seed cotton, sugar beet) specific crops. A new interpolation function is intro-
duced to map values to scores in terms of land characteristics. The economic evaluation
includes income-maximization taking into account market restrictions. The expert system has
been designed to help with the evaluation of land and to allow alteration in its rules based on
different performance observed in local areas. The GIS functions help in managing the spatial
data and visualizing the results. The software developed allows the evaluation and presenta-
tion of any equivalent spatial dataset and does not require special computer skills. # 2002
Elsevier Science Ltd. All rights reserved.
Keywords: Physical and Economic land evaluation; Expert Systems in Agriculture; Custom GIS

1. Introduction

The problem of selecting the correct land for the cultivation of a certain agri-
culture product is a long-standing and mainly empirical issue. Although many
researchers, organizations, institutes and governments have tried to provide a fra-
mework for optimal agricultural land use, it is suspected that much agricultural
land is used at below its optimal capability. The increased need for food production

* Tel.: +44-191-222-6359; fax: +44-191-222-5421.


E-mail address: stamatis.kalogirou@ncl.ac.uk (S. Kalogirou).

0198-9715/02/$ - see front matter # 2002 Elsevier Science Ltd. All rights reserved.
PII: S0198-9715(01)00031-X
90 S. Kalogirou / Comput., Environ. and Urban Systems 26 (2002) 89–112

and the shortage of resources stimulate a need for sophisticated methods of land
evaluation to aid decision makers in their role to both preserve highly suitable lands
and satisfy producers’ demand for increased profit.
This paper presents a methodology for the physical evaluation of the land based
on the FAO’s (1976) framework. It introduces a rather simplistic economic evalua-
tion using the results of a physical evaluation and information on crop yields. The
methodology is applied using a piece of specially written software that incorporates
expert system and GIS components. The area of interest is agricultural land in
Central Macedonia, Greece. The information concerning yields and prices has been
available for 1995 from FAO Statistics for Agriculture (FAO, 2001).
The main aim of this paper is to illustrate how important it is for rural planners to
provide the methodology for physical land evaluation along with economic evalua-
tion, and to motivate further research on the latter. The initial aim of this research is
to provide the geographic society with a stand-alone, powerful and user-friendly
application for land use assessment that has customized capabilities and function-
alities. However, the focus of the paper is on discussing a method for economic as
well as physical evaluation of the land. It aims also to link the theory with the real
world. Thus, it provides empirical work based on the results of the land evaluation
and the real economic conditions of the area of study. This paper mentions potential
research questions that have not been investigated due to time constrains and a lack
of data for the area of interest.
It is important to clarify how the integration of some advanced information tech-
nologies could be used in research and how this research could be applied in the real
world. The system described here could be applied to other geographical or spatial
problems (such as urban planning, human resource management); indeed, anywhere
there is a need for a rule-based classification and evaluation of information. An
argument of this paper, concerning the described software, is that expert systems are
suitable for solving classification problems and allowing on-the-fly customisation. The
GIS component was used in managing spatial data and visually presenting the results.
In this stage of research, no new survey methods or classification models are sug-
gested. Instead, the FAO classification model for crops (FAO, 1976, 1984, 1985) is
adopted, which allows land evaluation based on soil and environmental character-
istics into five classes of suitability (three suitable and two not suitable) for certain
crops. To date, evaluation for general cultivation as well as for the cultivation of wheat,
barley, maize, seed cotton, and sugar beet have been implemented. Current research
examines the participation of social and economic characteristics, such as local culture
and labour force, product prices and market conditions, for the model to be more realistic
and useful in rural planning. In this paper a simple approach is used: the allocation of
specific crop cultivation in each parcel of land is based on the maximum estimated
income out of the five crops possible (wheat, barley, maize, seed cotton, sugar beet).
In the literature it has been suggested that fuzzy classification (Ahamed, Rao, &
Murthy, 2000; Burrough, 1989; Davidson, Theocharopoloulos, & Blocksma, 1994;
Hall, Wang, & Subaryonon, 1992) and multi-criteria decision-making (McClean,
Cherrill, & Fuller, 1995; Pereira & Duckstein, 1993) give more accurate results for
land suitability than Boolean classification. However, some authors (Burrough, 1989)
S. Kalogirou / Comput., Environ. and Urban Systems 26 (2002) 89–112 91

used only three classes for their Boolean classification compared to seven or more
classes for their Fuzzy classification. Other problems with earlier studies include the
scale of maps used, the form of data (nominal, ordinal, ratio), and the land character-
istics measurement methods (survey, calculation from satellite images, estimation).
The main problem with map scale is that soil maps have been produced for
scales 1:250,000 (soil map of the UK) to 1:1,000,000 (FAO’s soil map of the world).
Many of those maps have been produced with coarse soil sampling and soil classi-
fication of large areas (hundreds or thousands of hectares) using satellite images.
For the latter method large area satellite images with pixels of 1 km have been used
along with land-use indexes (such as NDVI for vegetation). The creators of those
soil maps have tried to estimate the underlying soil based on the grounds land-use
and vegetation.
Fuzzy methods are more suitable for data extracted from maps such as soils where
there is a high possibility of inaccuracies. For example one polygon at a scale of
1:250,000 can overlap as many as a few thousand land parcels and tens of thousands
of land properties. One should note that in the area of interest a land property used
in agriculture could be as small as 1000 square feet. During empirical work using the
method presented in this paper, it was found that it is important for a soil survey to
be conducted for more accurate results to be obtained. When this is not possible, a
low-cost soil sampling could be conducted instead. Obviously, this method can only
be applied for small areas (district level) where data might have been already col-
lected by local government.
The database used here has been produced using existing soil, geology and land-
use maps, along with a soil survey that gave a more accurate image of the area. For
the analysis, maps in vector format were used. Because of the intersection of maps in
different scales (a geology map is usually in a different scale than a soil map), the
spatial dataset consists of approximately 4000 polygons covering an area of 32,000
hectares. Urban areas, water streams and water reservoirs can be clearly identified in
the maps (Figs. 1–4). Polygon sizes vary from less than a hectare to a 1000 ha for
some homogeneous areas. A polygon is equivalent to a parcel of land.
This paper argues that the existing methods for producing soil maps and the scale
of the latter are not sufficient for effective land use capability evaluation. The exist-
ing soil maps in developed countries are often at a very coarse scale and thus not
appropriate for the state of agricultural practice in Greece which is characterized by
low availability of land for agriculture and generally small land properties. Accord-
ing to the National Statistics for the number of agricultural holdings by order of
magnitude of agricultural land in Greece, 12.1% of the holdings are less than 1.25
acres in area, 25.4% of the holdings are less than a hectare and only 2.6% are over
20 ha (National Statistical Service of Greece, 2001). Thus, methodologies in the lit-
erature that use soil maps of scales from 1:250,000 up to 1:1,000,000 that assume a
certain level of soil homogeneity are not applicable in this case.
Another peculiarity of the Greek landscape is the strong variation of some of the
land characteristics such as gravel type and magnitude, slope and flood danger. This
is based on a different land development culture among individual farmers. The low
availability of resources for agriculture and the high cost of land development
92 S. Kalogirou / Comput., Environ. and Urban Systems 26 (2002) 89–112

Fig. 1. Land suitability evaluation for general cultivation.

Fig. 2. Land suitability evaluation for wheat cultivation. Map Detail.


S. Kalogirou / Comput., Environ. and Urban Systems 26 (2002) 89–112 93

Fig. 3. Economic land evaluation of five crops (maximum expected income). Suggested cultivation.

Fig. 4. Economic land evaluation of three grains (maximum expected income). Suggested cultivation.
94 S. Kalogirou / Comput., Environ. and Urban Systems 26 (2002) 89–112

methods deterred farmers to develop large land parcels. Therefore, it is important


for the use of this methodology that fine soil surveys take place and data are collected
for the area of study. This becomes more important under the framework of the
contemporary agriculture culture. The latter strongly focuses on the production of
organic food, thus it is necessary that accurate information about the land is col-
lected frequently. Consequently, the cost of such a method is high.
However, it is the responsibility of the local government and individual farmers to
adopt and use scientific approaches for the long-term profit maximization and nat-
ural resources preservation. In line to this the National Agricultural Research
Foundation (N.AG.RE.F.) has a research team and funds projects for the conduc-
tion of soil surveys, soil mechanics analysis aiming to produce accurate soil maps for
all the Hellenic territory. These maps are at a scale of 1:20,000. For the latter,
satellite imagery and aerophotography methods have been incorporated along with
soil sampling (National Agricultural Research Foundation, 2001). More informa-
tion about the local status of agriculture can be also found in the official website of
the Hellenic Ministry of Agriculture (Hellenic Ministry of Agriculture, 2001).

2. Background

The expert system used in this work has been implemented using the expert system
shell ‘‘CLIPS’’. Jackson (1999) discusses expert systems in general, several expert
system shells and programming languages, a few very well known problems and the
applications of expert systems. In the last chapter he introduces programming in
‘‘CLIPS’’. Most of the information concerning ‘‘CLIPS’’ used in this work can be
also found on the Internet (see Web references for CLIPS) and in Giarratano (1993)
as well as in Giarratano and Riley (1998). ‘‘CLIPS’’ has been developed by a team of
researchers at the Lyndon B. Johnson Space Centre, NASA and is a freeware pro-
duct (Software Technology Branch, 1993a; 1993b; 1993c). Recent versions 6 and
6.01 can be found in the official site as well as in links to it (see Web references for
CLIPS).
A fundamental text discussing land evaluation is the ‘‘A framework for land
evaluation’’ by FAO (1976). It provides standards, definitions and a description of
land qualities that can be formed from land characteristics. It also sets the guidelines
for physical and economical land evaluation. Guidelines for land evaluation for
rainfed and irrigated agriculture were provided by FAO in Soils Bulletin 52 (FAO,
1984) and Soils Bulletin 55 (FAO, 1985), respectively. However, FAO allows local
variations and many different implementation methods. In fact, not all researchers
follow this framework, but policy makers should do since this has been so far the
only standard approach to land classification. Other texts include a book on soil
survey and land evaluation by Dent and Young (1981); a very detailed text useful
for rural development in topics and subtropics by EUROCONSULT (1989); and the
Handbook of soil science (Sumner & Malcolm, 1999).
Two main categories of papers are relevant to this research. The first cate-
gory includes papers that discuss the use of a combined GIS and expert systems
S. Kalogirou / Comput., Environ. and Urban Systems 26 (2002) 89–112 95

technology and the second includes other approaches to implementing a land suit-
ability evaluation model. The former provide more evidence supporting the useful-
ness of expert systems but the applications discussed concern mainly animal habitats
and forestry. For the latter, expert systems have not been always used. Papers for
forestry planning provide a great range in different way of physical land evaluation.
Almost all authors do not include economic and social factors in their research.
Keechoo and Tim (1996) discuss a hybrid travel demand problem with GIS and
expert systems (ES). By integrating the same technologies, Kirkby (1996) tries to
identify and manage dryland salinization. The classification of urban land cover
based on expert systems and GIS are discussed both by Moller-Jensen (1997) and
Luckman, Jessen, and Gibb (1990). In a recent paper, Plant and Vayssieres (2000)
discuss the combination of ES and GIS technologies to implement a state transition
model of oak woodlands.
Other approaches of implementing land suitability evaluation models include a
multiple criteria decision-making methods (Pereira & Duckstein, 1993), and fuzzy
classification methods (Ahamed et al., 2000; Burrough, 1989; Davidson et al., 1994;
Hall et al., 1992). Bojorquez-Tapia, Diaz-Mondragon, and Ezcurra (2001) present a
GIS-based multivariate application for land suitability assessment with a public
participation base. Their work is one of the few including spatial structure (distance
from roads, coast) and competing land uses (agriculture, aquaculture, fisheries) but
the use of land characteristics is quite poor. Joerin, Theriault, and Musy (2001)
introduce a method called ELECTRI-TRI and account sustainable development
and economic competitiveness. They provide an example of suitability of land for
housing, but it would be interesting to examine whether their approach could be
applied in rural planning since it is more sophisticated than the others. The latter two
papers along with a linear programming and GIS for land use modelling (Chuvieco,
1993) focus on land use allocation, but partly discuss land suitability and provide
findings that can stimulate further research in the inclusion of social and economic
factors in land evaluation modelling.
It is worth mentioning the mixed qualitative/quantitative approach presented by
Van Lanen, Hack-ten Broeke, Bourma, and De Groot (1992), the Cropping Systems
Model PERFECT as a quantitative Tool in Land Evaluation (Thomas, Gardener,
Littleboy, & Shields, 1995) and the integration of three land classifications within
the Decision Support System for land use planning by McClean et al. (1995).
Hall et al. (1992) discuss a comparison between the Boolean and fuzzy classifica-
tion methods, which is a good example of the Boolean method itself. They con-
cluded that although the Boolean method is capable of aiding decision makers, the
fuzzy approach gives more realistic results due to overcoming the two level classifi-
cation of the former method. However, they use data which are already classified
such as Cation Exchange Capacity (CEC) and include only 4 classes for the Boolean
approach. They include only 13 land characteristics, while the FAO framework
suggests more than 50 land characteristics to measure 25 land qualities (FAO, 1985)
with more than half of them considered significant. This makes their evidence
weaker, and thus further research is required to support the argument that the fuzzy
approach is more realistic than the Boolean.
96 S. Kalogirou / Comput., Environ. and Urban Systems 26 (2002) 89–112

Burrough (1989) concluded that fuzzy set theory offers a way around incon-
sistencies occurred using the Boolean approach by allowing users to define flexible
class membership functions that match practical experience. He recognized that the
main sources of bias are the classes and weights chosen for land characteristics
rather than the methodology and encouraged research in addressing those. However
he rejected the Boolean method as inadequate to help in an exploration of data and
model the real world. Again in that paper less than 10 land characteristics were used,
and only three classes used in the Boolean approached whereas the fuzzy approach
was modelled using seven classes.
The immediate question is why those researchers have tried to increase the com-
plexity of the methodology by involving fuzzy theory and have not put effort
towards the incorporation of more land characteristics in their empirical work?
Fuzzy theory works effectively with numerical data. However, it does not serve its
aim to improve the quality of the classification since many of the variables used in the
model have already had their values classified (such as many of the soil datasets are).
It is argued here that research should focus on developing a more realistic model of land
evaluation by incorporating more sources of data, such as geographical, climatological,
socio-economic and cultural along with environmental and agro-ecological.
Expert systems work without problems with both Boolean and Fuzzy approaches.
The Boolean approach, which has been implemented in this paper, has been tested
and gave correct results in a few milliseconds per polygon.
Rossiter (1995) recognizes the value of an economic land evaluation for rural
planning and resource management. He identified a gap in the literature concerning
economic evaluation along with the plethora of works in physical evaluation. The
two key points of his paper are that the physical attributes of the land affect its
economic value, and that these effects can be quantified in economic terms by the
land evaluator.
Rossiter (1990) has also discussed the Automated Land Evaluation System
(ALES). ALES is a computer program that allows evaluators to build their own
knowledge-based systems with which they can compute the physical and economic
suitability of land map units. Using ALES, decision trees can be build. Customiza-
tion is fully supported in ALES. However, the implementation of the software does
not seem to be very user-friendly and it is rather difficult for an non-IT-expert to
make use of it. ALES is one of the few implementations with a knowledge-base
based on FAO’s Framework for Land Evaluation (FAO, 1976), therefore it should
be considered as an alternative solution.

3. Methodology

This research is multidimensional. The aim of the work is to provide a clear phy-
sical land evaluation methodology; to introduce a rather simple economic land
evaluation methodology; and to integrate computer technologies.
The physical evaluation consists of a model that assigns a score to every land
parcel based on its value on 17 characteristics (Section 4). To do this, initially scores
S. Kalogirou / Comput., Environ. and Urban Systems 26 (2002) 89–112 97

are assigned to individual land characteristics. The latter are then combined to form
three major groups of characteristics and then the total score is calculated. This
procedure is necessary in order for every characteristic to contribute with different
weight to the final score. This is important because the methodology is based in a
hierarchical importance of land qualities and a land quality is described with one or
more than land characteristics.
The data provided from a soil survey are often continuous data and therefore it is
necessary to apply a classification scheme that assigns scores to individual land
characteristics. This scheme is based on linear interpolation functions that map
value intervals to score intervals. If the observed value is x and it falls into the
interval [a,b] it needs to get a score y that falls into the interval [c,d]. The formula to
calculate y is:

dc dc cd ad  bc


y¼ xþc a or y¼ xþ
ba ba ab ab

The code in CLIPS is:


(deffunction interpolation (?observed ?a ?b ?c ?d)
(+(* ? observed (/ ( ?c ?d) ( ?a ?b))) (/((* ?a ?d) (* ?b ?c)) ( ?a ?b))))
For land characteristics with their values already classified, such as flood hazard,
the middle score of the class was given. For example, a land parcel with its flood
hazard equal to F2 will get a score for flood hazard equal to 75, the middle score of
FAO class 2 (65–85). After the physical evaluation finishes and scores have been
assigned to each land parcel, the parcel is classified into one of the FAO classes (N2–
S1) based on that score (Table 1). The score intervals for each class are the same for
the individual land characteristics and the total score a land parcel gets.
For the economic evaluation, the expected yield is calculated based on the score of
the land parcel for cultivation, and the corresponding maximum yield. Then the
expected income for all possible types of cultivation is calculated and the cultivation
that gives the highest expected income is selected.
In order to obtain a better understanding of the state of practice in the agriculture
industry, some fieldwork, mainly interviews of farmers took place. The result was
the creation of statistics concerning the average producer value for crops in Mace-
donia, and some indication of the expenses attached to growing these crops. The
latter have been included in this paper because it was not possible for their accuracy
to be verified from official sources. The local experience concerning producers price
of these products as well as the observed yields is similar with the corresponding
averages recorded by both word statistics (FAO, World Bank) and national statis-
tics (Hellenic Ministry of Agriculture, Hellenic Statistics Office). However, the yield
and product quality in Macedonia is not the same as in that of Thessaly, the region
that contributes the most in the national yield. Therefore the average values are
overestimates of the observed, thus the latter is used here.
Based on the individual farmers’ experiences it can be argued that the cost of the
cultivation varies across land properties. Major factors include the rent of the land if
98 S. Kalogirou / Comput., Environ. and Urban Systems 26 (2002) 89–112

Table 1
Summary table of land characteristics and score assignment. Classes suggested for General Cultivation

Land characteristics/ Class 0 1 2 3 4


classes and scores
Score 100–98 98–85 85–65 60–40 <40

Factor A
Soil toxitities var1 % Organic >1.5 1.5–1 1–0.6 <0.6 –
var2 % Base Saturation (BS) >60 60–50 50–40 <40 –
var3 Cation Exchange >18 18–12 12–6 <6 –
Capacity (CEC)
var4 % Carbonate (CaCO3) 0.3–10 10–30 30–50 50–80 >80
var5 % Sulfate (CaSO4) 0–2 2–4 4–10 10–15 >15
var6 Reaction (pH) 6–7.5 5.5–6 4.5–5.5 4.0–4.5 <4
8.5–7.5 9.0–8.5 9.0–9.5 >9.5
Rooting var7 Depth >90 90–60 60–40 40–20 <20
conditions var8a % Fine Gravel Volume 0–15 15–40 40–75 >75 –
% Coarse Gravel 0–3 3–15 15–40 40–75 >75
Volume
var9b % Stones Volume 0–3 3 3–15 15–40 40–75

var10 % Slope 0–3 (A) 3–12 (B) 12–18 (C) 18–36 (D) >36 (E)
var11 Erosion Hazard E0 E1 E2 E3 E4
var12 Soil Mechanics SiCL, SCf, Si, SCL, SL, Cm, LS, S,
SiL, L, C–60f SiCm* SC+60m C+60m
C–50f
Factor B
Excess of salts var13 Salinity (EC) 4–0 8–4 10–8 14–10 >14
var14 Sodicidy (ESP) 8–0 12–8 20–12 30–20 >30
Factor C
var15 Water level >120 60–120 40–60 20–40 <20
var16 Flood Hazard F0 F1 F2 F3 F4
var17 Drainage A B C D or E F or G
a
Gravel type {Fine, coarse, Stones}.
b
% Gravel in soil (0,100).

not owned by the farmer, the cost of the tractor’s oil and maintenance for tillage and
the cost of the seed. The most profitable way is for the farmer to use part of the
previous year’s yield as the current year’s seed, own the land and own the machinery
necessary to carry out the cultivation. Another finding of the qualitative research
was that new seeds have been developed in order to give higher yields based on the
local climate and soil mechanics. Therefore, it is necessary for the evaluator to be
able to review the classification scheme periodically and alter the value intervals of
the land characteristics that correspond to the appropriate suitability score.
The tasks of the land evaluation are achieved through the creation of software
termed Land Evaluation using an Intelligent Geographical Information System. The
acronym of this long term is LEIGIS. The software supports three main functions:
the application of the model to a set of data for an individual land parcel; the
application of the model in a rural area that consists of many land parcels, which are
S. Kalogirou / Comput., Environ. and Urban Systems 26 (2002) 89–112 99

represented as polygons on a map; and the presentation of the results on a map. The
expert system was implemented separately for land suitability evaluation for general
and specific cultivation (six files).
The programming environment for the software is the graphic language MS Visual
Basic (VB). VB is well documented and supported by both its vendor (Microsoft
Corporation) and many web sites of individuals and groups of developers. Techni-
cally speaking, the code written in VB manages the input to and the output from
the expert system component and the input to the map controls (GIS component).
The expert system has the form of a text file for every crop which can easily be altered.
At run time, the expert system shell ‘‘CLIPS’’, in the form of a MS Windows Dynamic
Link Library (dll), allocates memory for the fact base, loads and compiles the indivi-
dual expert systems (text files), and waits for data to satisfy the conditions of rule(s)
and start running the classification model. The idea is that an expert system is a set
of rules with conditions and a set of facts that satisfy these conditions. The facts
activate rules that run and produce other facts and so on and so forth, until there are
not enough facts within the fact base to satisfy any rule and the process ends. During
that process, classification values are being produced and error messages are being
reported. Those results then are saved in the corresponding databases.
GIS functionality was implemented using the MapObjects ActiveX Controls
developed by ESRI (ESRI Press, 1996). MapObjects technology is based on the
architecture used in ESRI ArcView. MapObjects support the visualization of the
results (GIS component). The reason they have been selected in the place of other
software (e.g. ArcView) is for the user to save time and money, as well as for LEIGIS
to be a standalone tool. The GIS component allows vector maps (ESRI shape file
format) to be plotted, data for every polygon to be explored and results to be
presented. The system reads the land attributes from the loaded map and after the
classification it stores the results in the map’s database. However it is possible to
store those results in an external database along with the id of each polygon to allow
a link between the spatial and classification data using that common id. The user can
define a colour for each class of suitability, and the system then can create a choro-
pleth map, ready for printing.
Concerning the classification model itself; it is executed by the rules of the expert
system. Firstly, it checks the validity of the values entered for a field (e.g. pH is
always between 0 and 14); secondly it maps continuous values to distinct classes (1–
5); and finally it calculates the suitability for a crop by combining the single vari-
ables linearly. When data for a land characteristic are not nominal, an interpolation
method is used to assign a score of 1–100 rather than a class (S1–N1) to that char-
acteristic. For nominal or already classified values the middle score of a class was
assigned to allow numerical computing (e.g. score 99 for class S1).

4. Data, analysis and results

This research started during author’s undergraduate dissertation, which was


conducted with a joint supervision of the Department of Informatics and the
100 S. Kalogirou / Comput., Environ. and Urban Systems 26 (2002) 89–112

Department of Agriculture, University of Thessaloniki, Greece. The first part of this


research concentrated in the software development and not in the land evaluation
modelling. It focused in selecting and integrating computer technologies to address
classification models. Data and some experience were provided from colleagues in
Agriculture Department in that stage. This paper reviews existing work in this field
and introduces findings of further research.
The database used here was produced for another project and was provided under
the condition that it will be used for academic purposes only. Data have been used
under the condition that no identification of individual properties and land perfor-
mance could be possible from published information. Thus, results are presented
using choropleth maps only. However, new data will be available as soon as the
Greek Cadastre is completed which will provide boundary data for individual fields
(land properties) with up to date information for non-agriculture land, forest and
land use in digital form (vector GIS). Contemporary methods in soil surveys, land
development and EU agricultural policy, have introduced new ways of land suit-
ability evaluation and land use in general.
For the physical evaluation 17 land characteristics were combined. Those char-
acteristics form three Factors that then are combined to result in a class of land
capability.
The three Factors include the following land characteristics:
Factor A: Soil Mechanics and Toxicities, Slope, Erosion hazard, Rooting Condi-
tions.
Factor B: Excess of Salts.
Factor C: Water level, Flood hazard and Drainage.
The classes of land suitability (capability) suggested by FAO are five:
Class S1: Highly Suitable.
Class S2: Moderately Suitable.
Class S3: Marginally Suitable.
Class N1: Currently not Suitable.
Class N2: Permanently not Suitable.
However, the equal interval classification (five intervals of 20% in the range 0–
100) was not used in this work. Instead, for the highly suitable class, the score
interval 98–100 was adopted which makes this score almost impossible for any land.
The detailed classes of suitability and scoring for land qualities for general cultiva-
tion are given in Table 1. For an explanation of those see Sys (1985), Silleos (1990)
and FAO (1984, 1985). It is beyond of the aim of this paper to analyse in detail the
reasons why this classification scheme was used. However, evidence for the impor-
tance of this scheme can be seen in Table 2, in which the expected land performance
and profit for each class is given. Note that for a land to be highly suitable, mini-
mum input (irrigation, fertilizers, land development) is required, thus the profit is
maximized.
The Boolean classification was implemented in a way that for a value that is
already classified (e.g. flood hazard=F2) the average score of the class was given
S. Kalogirou / Comput., Environ. and Urban Systems 26 (2002) 89–112 101

Table 2
Expected land performance and profitability

Class FAO class Score Expected performance: Expected profit:


percentage of the perfect percentage of profit
performance in perfect conditions

0 S1 98–100 >90 >75


1 S2 85–98 60–90 25–75
2 S3 65–85 35–60 <25
3 N1 40–65 <35 –
4 N2 <40 – –

(e.g. 75) whereas for continues values an interpolation function used to assign a
score (e.g. for water level 44 cm the score is 69). Scores are available for each poly-
gon along with the final classification (S1–N2). An indication of the weights used to
calculate the final score can been obtained from the calculation formulas follow:

Final Score ¼ ðFactor AÞ ðFactor BÞ ðFactor CÞ=1000:

Factor A ¼ ðsoil tox þ var7 þ var8 þ var9 þ var10 þ var11 þ var12Þ=6

soil-tox ¼ ðvar1 þ var2 þ var3 þ var4 þ var5 þ var6Þ=6

Factor B ¼ ðvar13 þ var14Þ=2

Factor C ¼ ðvar15 þ var16 þ var17Þ=3

It is very common for the attributes to have some degree of measurement error.
Further research is required to address this issue. If extreme values or no values
presented, the model returns a zero score prompting the user to examine the data in
the corresponding polygon. The analysis of the given dataset resulted in an accurate
suitability map for general cultivation, and different suitability maps for individual
crops. It was interesting that every polygon was assigned a class for general culti-
vation, but this was not the case for the five specific crops. The land requirements for
barley, wheat, maize, seed cotton and sugar beets resulted in a zero score for many
polygons. That was because the single characteristics’ classes are formed with dif-
ferent value intervals compared with the general cultivation. The detailed classes of
suitability and scoring for land qualities for specific cultivation are given in Appen-
dix A at the end of the paper. The land some of the big polygons represent is not
necessarily agricultural, however their score can motivate land development.
Some of the results are presented in the map series section; Figs. 1 and 2. In those
maps, urban areas represent any residential, commercial or industrial area. Water
represents rivers, streams, water reservoirs, moist soils and seasonally dry streams.
The ‘‘No data’’ class represents lands that had zero score after the evaluation. That
zero score was obtained because of wrong data, missing or out of classes values. The
102 S. Kalogirou / Comput., Environ. and Urban Systems 26 (2002) 89–112

reason those lands were not assigned the class N1 was because a more detailed
investigation should be made for them.
Fig. 1 shows the class assigned to each polygon in terms of its suitability for gen-
eral cultivation. The histogram of the map illustrated that there are no fields with
suitability of classes S1 and S2. This suggested that the land requires development
and fertilizers. However, the evaluation for specific crops showed that some fields
where moderately suitable (S2) and more fields where suitable marginally suitable
(S3) than unsuitable comparing with the suitability for general cultivation. The
suitability scoring for maize cultivation showed that the area of interest was unsui-
table for this cultivation. Thus, for an average yield to be produced, high input will
be required, which increases the cost for the producer. However, for seed cotton and
sugar beets cultivation the land is capable after an average input to produce the
average yield. Most of the land parcels were classified in N1, S3 and S2 classes in
term of their suitability for the latter types of cultivation. A detail of the evaluation
map for wheat if given in Fig. 2. Most of the land parcels were either marginally
suitable or currently not suitable for cultivating barley or wheat. However, the cul-
tivation of such grains does not require much input, therefore is considered as more
profitable than the other. A certain amount of fertilizers could significantly increase
the land capability and allow more profitable cultivation.
Going one step further, the maximum income per polygon using average produc-
tion per hectare and producer price was estimated. To accomplish that, information
about the market of agricultural products (producer and retail prices) in Greece in
1995 published by FAO was extracted from the on-line Statistics for Agriculture
(FAO, 2001) and international market prices were used to confirm the former (The
World Bank, 2001). The analysis of those statistics resulted in the calculation of the
average yield per cultivation per hectare. Furthermore, the expected income in
Euros per hectare using the average producer price of the product in the same year
(Table 3) was calculated. However, prices vary based on the quality of the product,
whereas the yield is correlated with the land capability. The performance was esti-
mated in correlation with the land evaluation suitability score using the information
provided in Table 2. Table 4 shows equations for calculating the expected perfor-
mance as a percentage of the perfect performance (interpolation).

Table 3
Average price per kilogram, yield per hectare (in kg), and expected income per hectare. Empirical values
for yield reflect the experience in Greece

Greece 1995 Producer price Average yield Empirical yield Average income
(Lc/kg) (kg/ha) (kg/ha) (Euros/ha)

Barley 45.470 2632.8 1000–5000 351


Wheat 63.290 2634.1 1000–5000 489
Seed Cotton 280.510 3157.1 500–4000 2599
Sugar Beets 17.734 63,290.3 60,000–90,000 3294
Maize (Corn) 44.640 10,076.2 10,000–15,000 1320

Source: FAO—Price for Maize estimated using grain prices in 1995 by The World Bank. Lc, Local cur-
rency (Drachma); kg, kilogram; ha, hectare; Euros, 1 Euro=340.75 Drachmas (fixed currency rate).
S. Kalogirou / Comput., Environ. and Urban Systems 26 (2002) 89–112 103

Table 4
Equations to calculate expected land performance from scores

Class FAO Class Score Expected performance (EP): percentage of the perfect performance

0 S1 98–100 EP=0.2 * Score+80 90–100


1 S2 85–98 EP=2.3 * Score 135.5 60–90
2 S3 65–85 EP=1.25 * Score  46.25 35–60
3 N1 40–65 EP=1.4 * Score 56 0–35
4 N2 <40 0 0

Table 5
Average price per kilogram, yield per hectare (in kg), and expected income per hectare. Empirical values
for yield reflect the experience in Greece

Greece 1995 Expected maximum Producer price Maximum income


yield (kg/ha) (Euros/t) (Euros/ha)

Barley 5305.1 133.441 707.92


Wheat 5307.7 185.737 985.84
Seed Cotton 6361.6 280.510 1784.49
Sugar Beets 127,530.0 52.044 6637.17
Corn 20,303.5 131.005 2659.86

Assuming that an average yield should be the result of cultivating a land with an
average capability, that land should have a suitability score equal to 75 (middle
value of the middle FAO class). According to information provided in Table 4
(equation for S3) the expected performance (EP) is the 47.5% of the perfect perfor-
mance. [EP=1.25*Score46.25)(Score=75) EP=47.5]. Thus, the perfect per-
formance should be the 1/0.475=210.5% of the expected performance which is the
average in this case. Comparing this to the empirical findings for maximum yields
(some extremes and not the average of the maximum reported) based the farmers’
reports it has been decided to adjust this factor to 201.5%. After multiplying aver-
age performance given in Table 3 with 2.015, values for the perfect performance are
given in Table 5. An immediate observation is that the expected maximum perfor-
mance is probably overestimated compared with the empirical range of expected
yield. However, the linear multiplication of the values for the five crops does not
change their ranking order based on the expected income. Since the score for most
of the lands is N1 or S2, those results can be accepted.
The objective of this step was to provide a methodology for the modelling of the
maximization of the expected income in order to suggest a land use breakdown for
the area. To do this it was necessary to calculate the expected performance (% of
maximum) based on the score of the land for each of the five possible cultivations;
multiply that with the appropriate maximum expected yield (to obtain the expected
yield); multiply the result of the latter with the average producer price of the crop (to
obtain the expected income); and finally select the maximum value out of the five
calculated. An example of the methodology described above is provided in Table 6.
104 S. Kalogirou / Comput., Environ. and Urban Systems 26 (2002) 89–112

Table 6
Example of a land evaluation in terms of maximum expected income

Product Score FAO class Expected performance Expected yield Expected income
(% of maximum) (kg/ha) (Euros/ha)

Barley 56 N1 22.4 1188.3 158.6


Wheat 52 N1 16.8 891.7 165.6
Seed Cotton 64 N1 33.6 2137.5 599.6
Sugar Beets 59 N1 26.6 33,923.0 1765.5
Corn 63 N1 32.2 6537.7 856.5

Max (158.6, 165.6, 599.6, 1,765.5, 856.5)=1765.5 thus, suggested cultivation is ‘‘Sugar Beets’’

A random polygon with scores for the five different crops was selected and the
results of the three phases of calculation (Expected performance; Expected Yield;
Expected Income) are shown in the corresponding columns in Table 6. The sug-
gested cultivation was that of Sugar Beets. It seems that this was the dominant cul-
tivation since the production per hectare is very high even though the price of the
product is the lowest. This was observed in the whole area of interest (Fig. 3) and
motivated further investigation.
The economic evaluation should always take into account local condition of the
market and possible production restrictions. A study of the local news and reports
by the Hellenic Ministry of Agriculture (2001), and newspapers made clear that the
production of sugar beets is limited (European Agreements) and most of it has long
term contracts with the National Sugar Industry. That makes it impossible for the
area to be used only for sugar beets cultivation. Thus the suggested method needs to
be enhanced with more economical attributes in order to help in decision making for
rural planning. In fact the production is limited under the European Agriculture
Agreement for all products. However, for crops such as grains, there is more flexi-
bility since the product can be used as livestock food and partly not be registered as
a profitable investment. Because of that, a second evaluation took place, this time
only including the cultivation of barley, wheat and maize. The results were accep-
table and are presented in Fig. 4.
Another issue that has to be examined is the affect of the contiguity of the fields.
Zones with homogeneous cultivation are more likely to give higher profit than
heterogeneous cultivation. Evidence for this argument is both the cost of cultivation
and the environmental effect of contiguity. For example, the cultivation of maize
requires more water and takes place in different time periods than that of barley and
that of wheat. Dry fields next to a field of maize may reduce the yield of the latter, or
may generate the need for more irrigation to keep the soil’s moistness at the required
level. Homogeneous areas allow the use of more cost-effective machinery and
require less transportation. Similar studies have been conducted in order to design
homogeneous zones using socio-economic data (Openshaw & Alvanides, 1999).
However, the concept of designing zones with specific properties for the need of
the analysis can be taken further as demonstrated by Alvanides, Openshaw, and
MacGill (2001). For example, it is possible to group many fields with specific soil
S. Kalogirou / Comput., Environ. and Urban Systems 26 (2002) 89–112 105

and size characteristics into larger crop zones in order to minimise the cultivation
cost and help in maximising the land’s profitability. Further research is required to
address this issue since it is beyond the scope of this work.

5. Conclusions

The research has resulted in stand-alone software called LEIGIS for supporting
rural planners with a first view of the land suitability for cultivation of certain
crops. This work examined soil and environmental characteristics to produce a
physical evaluation of land capabilities and used those results to provide an eco-
nomical evaluation of the land for different types of agriculture. The latter was
based in the selection of the crop that generated maximum income.
Scientists have been heavily criticized for the use of expert systems. This work
supports the argument that the use of expert systems is a faster and more flexible
solution in addressing classification problems compared with procedural sequential
algorithms and the corresponding software. Another advantage of expert systems is
that they produce classifications based on given knowledge rather than by making
assumptions or by using heuristic algorithms.
A review of methods of physical evaluation concludes that the fuzzy approach
compared with the Boolean increases the complexity of the modelling without
making the results more efficient. It is necessary to test the accuracy of soil data used
by conducting soil surveys or sampling. The estimation of land capability based in
remotely sensed images is more likely to give inaccurate results. Finally, the scale of
the area should be the level of individual fields. The technology today can handle
large datasets but specific soil data such as those used in this research are not avail-
able for that scale and for all areas of a country.
The physical and economical evaluation led to the conclusion that it is necessary to
conduct both to aid rural planning. That was because both phases are necessary
to minimize the risk of decisions. The physical evaluation suggested here used a
classification method that set very specific requirements for land to be classified as
highly suitable. However, those parcels of land which are highly suitable require
minimum input during the cultivation process. The economic evaluation was rather
basic, but is still important. It tried to provide an indication of the role economic
factors may play in altering land-use planning in order to maximize profitability.
The latter is the first aim of the producers who do not necessarily care about the
protection of the land.
This work showed that further research should be conducted to address issues
such as the contiguity of the lands, the enhancement of the physical evaluation by
involving more land characteristics (such as climate) and the development of a
complete economic evaluation. The latter should include socio-economic variables
such as available agriculture labour force, and spatial variables such as accessibility,
distance from urban areas (travel cost of workers) and industrial areas (transporta-
tion cost of the products). It should look at the maximization of profit for the pro-
ducer, along with the preservation of lands at risk of permanent unsuitability.
106 S. Kalogirou / Comput., Environ. and Urban Systems 26 (2002) 89–112

Acknowledgements

To all my colleagues in the who helped me with their comments during this work.
To my father who helped with fieldwork collecting information about the state of
the market in agriculture in Greece as well as the empirical yield of crops presented
in Table 3 within Section 4. To Professor Peter Fisher for his encouragement, valu-
able comments and the update with some previous publications relative to this
paper. To Professor Stewart Fotheringham with his help with the language of the
paper. The dataset used in this work is the copyright of the Laboratory of Remote
Sensing and Geographical Information Systems, Department of Agriculture, Uni-
versity of Thessaloniki, Greece.

Appendix A. Summary tables of land characteristics and score assignment for


barley, sugar beets, seed cotton, maize and wheat.

Table A1 Summary table of land characteristics and score assignment. Classes


suggested for Barley Cultivation

Land characteristics/ Class 0 1 2 3 4


classes and scores
Score 100–98 98–85 85–65 60–40 <40

Factor A
Soil var1 % Organic >0.5 0.4–0.5 <0.4 – –
toxicities var2 % Base Saturation >80 80–50 50–35 <35 –
(BS)
var3 Cation Exchange >16 16–8 <8 – –
Capacity
(CEC)
var4 % Carbonate (CaCO3) 3–20 0–3 or 30–40 40–60 >60
20–30
var5 % Sulfate (CaSO4) 0–3 3–6 6–12 12–20 >20
var6 Reaction (pH) 5.3–6.7 5.1–5.3 4.9–5.1 4.7–4.9 <4.7
Rooting var7 Depth >90 90–50 50–20 20–10 <10
conditions var8a % Fine Gravel Volume 0–15 15–40 40–75 >75 –
% Coarse Gravel 0–3 3–15 15–40 40–75 >75
Volume
var9b % Stones Volume 0–3 3 3–15 15–40 >40

var10 % Slope 0–3 (A) or 12–18 (C) 18–36 (D) >36 (E) –
3–12 (B)
var11 Erosion Hazard E0, E1 E2 E3 E4 –
var12 Soil Mechanics SiL, L, SC-60, SL, C+60, LS, Cm+60, S
SCL, CL, C-60, SiC, SC+60 SCm+60
SiCL, Si SiC, C

Factor B var13 Salinity (EC) 0–4 4–8 8–12 12–16 >16


var14 Sodicidy (ESP) 0–15 15–25 25–35 35–45 >45
S. Kalogirou / Comput., Environ. and Urban Systems 26 (2002) 89–112 107

Land characteristics/ Class 0 1 2 3 4


classes and scores
Score 100–98 98–85 85–65 60–40 <40

Factor C Var15 Water level >90 50–90 20–50 10–20 <10


Var16 Flood Hazard F0, F1 F2 F3 F4 –
Var17 Drainage A or B C D E F or G
a
Gravel type {Fine, Coarse, Stones}.
b
% Gravel in soil (0,100).

Table A2 Summary table of land characteristics and score assignment. Classes


suggested for Sugar Beets Cultivation

Land characteristics/ Class 0 1 2 3 4


classes and scores
Score 100–98 98–85 85–65 60–40 <40

Factor A
Soil toxicities var1 % Organic >2.5 1.2–2.5 0.6–1.2 <0.4 –
var2 % Base Saturation (BS) >80 80–50 50–15 < 15 –
var3 Cation Exchange >24 24–16 16–8 <8 –
Capacity (CEC)
var4 % Carbonate (CaCO3) 3–20 0–3 or 30–40 40–60 >60
20–30
var5 % Sulfate (CaSO4) 0–2 2–4 4–10 10–15 >15
var6 Reaction (pH) 5.8–7.5 5.2–5.8 4.8–5.2 4.2–4.8 <4.2
Rooting var7 Depth >90 90–50 50–30 30–15 <15
conditions var8a % Fine Gravel Volume 0–15 15–40 40–75 > 75 –
% Coarse Gravel 0–3 3–15 15–40 40–75 >75
Volume
var9b % Stones Volume 0–3 3 3–15 15–40 >40

var10 % Slope 0–3 (A) or 12–18 (C) 18–36 (D) >36 (E) –
3–12 (B)
var11 Erosion Hazard E0, E1 E2 E3 E4 –
var12 Soil Mechanics L, SCL, Si, SL, C+60, LS, Cm+60, S
SiC, CL, SiL, C–60 SC+60 SCm+60
SiCL, SC

Factor B
var13 Salinity (EC) 0–8 8–10 10–14 14–18 >18
var14 Sodicidy (ESP) 0–15 15–25 25–35 35–45 >45

(Table continued on next page)


108 S. Kalogirou / Comput., Environ. and Urban Systems 26 (2002) 89–112

Land characteristics/ Class 0 1 2 3 4


classes and scores
Score 100–98 98–85 85–65 60–40 <40

Factor C
var15 Water level >120 90–120 60–90 30–60 <30
var16 Flood Hazard F0, F1 F2 F3 F4 –
var17 Drainage A or B C D or E F or G
a
Gravel type {Fine, Coarse, Stones}.
b
% Gravel in soil (0,100).

Table A3 Summary table of land characteristics and score assignment. Classes


suggested for Seed Cotton Cultivation

Land characteristics/ Class 0 1 2 3 4


classes and scores
Score 100–98 98–85 85–65 60–40 <40

Factor A
Soil toxicities var1 % Organic >1.5 0.8–1.5 0.4–0.8 <0.4 –
var2 % Base Saturation (BS) >80 80–50 50–35 <35 –
var3 Cation Exchange >25 25–16 16–10 10–6 <6
Capacity (CEC)
var4 % Carbonate (CaCO3) 1–10 0–1 or 10–20 20–30 30–40 >40
var5 % Sulfate (CaSO4) 0–2 2–4 4–10 10–15 >15
var6 Reaction (pH) 5.0–7.5 4.5–5.0 4.0–4.5 3.5–4.0 <3.5
Rooting var7 Depth >90 90–50 50–20 20–10 <10
conditions var8a % Fine Gravel Volume 0–15 15–40 40–75 >75 –
var9b % Coarse Gravel 0–3 3–15 15–40 40–75 >75
Volume
% Stones Volume 0–3 3 3–15 15–40 >40

var10 % Slope 0–3 (A) 12–18 (C) 18–36 (D) >36 (E) –
or 3–12 (B)
var11 Erosion Hazard E0, E1 E2 E3 E4 –
var12 Soil Mechanics SiL, L, SC, CL, Cm+60,
SCL, CL, SiCf, Si SiCm, LS
SiCL

Factor B
var13 Salinity (EC) 0–8 8–10 10–12 12–16 >16
var14 Sodicidy (ESP) 0–15 15–20 20–30 30–40 >40

Factor C
var15 Water level >90 50–90 20–50 10–20 <10
var16 Flood Hazard F0, F1 F2 F3 F4 –
var17 Drainage A or B C D or E F G
a
Gravel type {Fine, Coarse, Stones}.
b
% Gravel in soil (0,100]).
S. Kalogirou / Comput., Environ. and Urban Systems 26 (2002) 89–112 109

Table A4 Summary table of land characteristics and score assignment. Classes


suggested for Maize Cultivation

Land characteristics/ Class 0 1 2 3 4


classes and scores
Score 100–98 98–85 85–65 60–40 <40

Factor A
Soil toxicities var1 % Organic >1.5 0.8–1.5 <0.8 – –
var2 % Base Saturation (BS) >50 50–35 <35 – –
var3 Cation Exchange >24 24–16 16–8 <8 –
Capacity (CEC)
var4 % Carbonate (CaCO3) 0–6 6–15 15–25 25–40 >40
var5 % Sulfate (CaSO4) 0–2 2–4 4–10 10–15 >15
var6 Reaction (pH) 5.4–6.8 5.2–5.4 5.0–5.2 <5.0 –
Rooting var7 Depth >90 90–60 60–40 40–20 <20
conditions var8a % Fine Gravel Volume 0–15 15–40 40–75 >75 –
% Coarse Gravel 0–3 3–15 15–40 40–75 >75
Volume
var9b % Stones Volume 0–3 3 3–15 15–40 >40

var10 % Slope 0–3 (A) 12–18 (C) 18–36 (D) >36 (E) –
or 3–12 (B)
var11 Erosion Hazard E0, E1 E2 E3 E4 –
var12 Soil Mechanics SiC, SL, L, Cm+60, – –
SCL, CL, C-60, SiL SCm+60, LS
SiCL, SCf, Si

Factor B var13 Salinity (EC) 0–2 2–4 4–6 6–8 >8


var14 Sodicidy (ESP) 0–8 8–15 15–20 20–25 >25

Factor C var15 Water level >90 60–90 40–60 20–40 <20


var16 Flood Hazard F0, F1 F2 F3 F4 –
var17 Drainage A or B or C D E F G
a
Gravel type {Fine, Coarse, Stones}.
b
% Gravel in soil (0,100).

Table A5 Summary table of land characteristics and score assignment. Classes sug-
gested for Wheat Cultivation

Land characteristics/ Class 0 1 2 3 4


classes and scores
Score 100–98 98–85 85–65 60–40 <40

Factor A
Soil toxicities var1 % Organic >1.5 0.8–1.5 0.4–0.8 <0.4 –
var2 % Base Saturation (BS) >80 80–50 50–35 <35 –
var3 Cation Exchange >16 8–16 <8 – –
Capacity (CEC)
(Table continued on next page)
110 S. Kalogirou / Comput., Environ. and Urban Systems 26 (2002) 89–112

Land characteristics/ Class 0 1 2 3 4


classes and scores
Score 100–98 98–85 85–65 60–40 <40

var4 % Carbonate (CaCO3) 3–20 0–3 or 30–50 50–70 >70


20–30
var5 % Sulfate (CaSO4) 0–3 3–6 6–12 12–20 >20
var6 Reaction (pH) 5.4–7.0 5.2–5.4 5.0–5.2 4.8–5.0 <4.8
Rooting var7 Depth >90 90–50 50–30 30–20 <20
conditions var8a % Fine Gravel Volume 0–15 15–40 40–75 >75 –
% Coarse Gravel 0–3 3–15 15–40 40–75 >75
Volume
var9b % Stones Volume 0–3 3 3–15 15–40 >40

var10 % Slope 0–3 (A) or 12–18 (C) 18–36 (D) >36 (E) –
3–12 (B)
var11 Erosion Hazard E0, E1 E2 E3 E4 –
var12 Soil Mechanics C, SiC, SiCL, CL, L, SiL SL, C, S, LS
SC–60 SCL SCm+60

Factor B
var13 Salinity (EC) 0–4 4–8 8–12 12–16 >16
var14 Sodicidy (ESP) 0–15 15–25 25–35 35–45 >45

Factor C
var15 Water level >90 60–90 40–60 20–40 <20
var16 Flood Hazard F0, F1 F2 F3 F4 –
var17 Drainage A or B C D E or F G
a
Gravel type {Fine, Coarse, Stones}.
b
% Gravel in soil (0,100).

References

Ahamed, T. R. N., Rao, K. J., & Murthy, J. S. R. (2000). GIS-based fuzzy membership model for crop-
land suitability analysis. Agricultural Systems, 63, 75–95.
Alvanides, S., Openshaw, S., & Macgill, J. (2001). Zone design as a spatial analysis tool. In N. J. Tate, &
P. M. Atkinson (Eds.), Modelling Scale in Geographical Information Service. New York: John Wiley & Sons.
Bojorquez-Tapia, L. A., Diaz-Mondragon, S., & Ezcurra, E. (2001). GIS-based approach for participa-
tory decision making and land suitability assessment. International Journal of Geographical Information
Science, 15, 129–151.
Burrough, P. A. (1989). Fuzzy mathematical methods for soil survey and land evaluation. Journal of Soil
Science, 40, 477–492.
Chuvieco, E. (1993). Integration of linear programming and GIS for land-use modelling. International
Journal of Geographical Information Systems, 7, 71–83.
Davidson, D. A., Theocharopoulos, S. P., & Blocksma, R. J. (1994). A land evaluation project in Greece
using GIS and based on Boolean and fuzzy set methodologies. International Journal of Geographical
Information Systems, 8, 369–384.
Dent, D., & Young, A. (1981). Soil survey and land evaluation. London: George Allen & Unwin.
ESRI Press (1996). Building applications with MapObjects. Redlands, Calif.: Environmental Systems
Research Institute Press.
S. Kalogirou / Comput., Environ. and Urban Systems 26 (2002) 89–112 111

EUROCONSULT. (1989). Agricultural Compendium for rural development in the tropics and subtropics.
Amsterdam: Elsevier.
Food and Agriculture Organization of the United Nations (FAO) (1976). A framework for land evalua-
tion. Soils bulletin, No.. 32. Rome: FAO.
Food and Agriculture Organization of the United Nations (FAO) (1984). Guidelines for land evaluation for
rainfed agriculture. Soils bulletin, No. 52. Rome: FAO.
Food and Agriculture Organization of the United Nations (FAO) (1985). Guidelines: land evaluation for
irrigated agriculture. Soils Bulletin, No. 55. Rome: FAO.
Food and Agriculture Organization of the United Nations (FAO) (2001), FAOSTAT agriculture Data.
Available: http://apps.fao.org/page/collections?subset=agriculture, last accessed 10 April 2001.
Giarratano, C. J. (1993). CLIPS: C language integrated production system, CLIPS user’s guide-version 6.0.
Athens, Ga.: COSMIC, University of Georgia.
Giarratano, C. J., & Riley, G. (1998). Expert systems: principles and programming (3rd ed.). Boston: PWS
Publishing Company.
Hall, G. B., Wang, F., & Subaryono (1992). Comparison of Boolean and fuzzy classification methods in
land suitability analysis by using Geographical Information Systems. Environment and Planning A, 24,
497–516.
Hellenic Ministry of Agriculture (2001). Available: http://www.minagric.gr, last accessed 30 July 2001.
Jackson, P. (1999). Introduction to expert systems. New York: Addison-Wesley.
Joerin, F., Theriault, M., & Musy, A. (2001). Using GIS and outranking multicriteria analysis for land-
use suitability assessment. International Journal of Geographical Information Science, 15, 151–174.
Keechoo, C., & Tim, T. J. (1996). A hybrid travel demand model with GIS and expert systems. Compu-
ters, Environment and Urban Systems, 20, 247–259.
Kirkby, S. D. (1996). Integrating a GIS with an expert system to identify and manage dryland saliniza-
tion. Applied Geography, 16, 289–303.
Luckman, P. G., Jessen, M. R., & Gibb, R. G. (1990). Use of expert systems and GIS in land evaluation.
New Zealand Geographer, 46, 15–20.
McClean, C. J., Cherrill, A. J., & Fuller, R. M. (1995). The integration of three land classifications within
a decision support system for land use planning. In P. Fisher (Ed.), Innovations in GIS 2 (pp. 137–149).
London: Taylor & Francis.
Moller-Jensen, L. (1997). Classification of urban land cover based on expert systems, object models and
texture. Computers, Environment and Urban Systems, 21, 291–302.
National Agricultural Research Foundation (2001). Available: http://www.ethiage.gr, last accessed 30
July 2001.
National Statistical Service of Greece (2001). Available: http://www.statistics.gr/en/data/tables/
table71.htm, last accessed 30 July 2001.
Openshaw, S., & Alvanides, S. (1999). Applying geocomputation to the analysis of spatial distributions. In
P. A. Longley, M. F. Goodchild, D. J. Maguire, & D. W. Rhind (Eds.), Geographical information systems
(Vol. 1), principles and technical issues (2nd ed.) (pp. 267–282). New York: John Wiley & Sons, Inc.
Pereira, J. M. C., & Duckstein, L. (1993). A multiple criteria decision-making approach to GIS-based
land suitability evaluation. International Journal of Geographical Information Systems, 7, 407–424.
Plant, R. E., & Vayssieres, M. P. (2000). Combining expert system and GIS technology to implement a
state transition model of oak woodlands. Computers and Electronics in Agriculture, 27, 71–93.
Rossiter, R. G. (1990). ALES: a framework for land evaluation using a microcomputer. Soil Use and
Management, 6, 7–20.
Rossiter, R. G. (1995). Economic land evaluation: why and how. Soil Use and Management, 11, 132–140.
Software Technology Branch, Lyndon B. Johnson Space Center (1993a). Clips reference manual: basic
programming guide (Vol. I). CLIPS Version 6.0. Houston, Tex.: The Artificial Intelligence Section.
Software Technology Branch, Lyndon B. Johnson Space Center (1993b). CLIPS reference manual:
advanced programming guide (Vol. II). CLIPS Version 6.0. Houston, Tex.: The Artificial Intelligence
Section.
Software Technology Branch, Lyndon B. Johnson Space Center (1993c). CLIPS reference manual: CLIPS
interfaces guide (Vol. III). CLIPS Version 6.0. Houston, Tex.: The Artificial Intelligence Section.
112 S. Kalogirou / Comput., Environ. and Urban Systems 26 (2002) 89–112

Sumner, M. E., & Malcolm, E. (1999). Handbook of soil science. Boca Raton: CRC Press.
Silleos, N. (1990). Mapping and evaluation of agricultural soils and lands. Thessaloniki: Giachoudi & Gia-
pouli.
Sys, C. (1985). Land evaluation. Ghent, Belgium: State University of Ghent.
The World Bank (2001). Pink sheet—January 1998: commodity price data. Available: http://www.world-
bank.org/html/ieccp/pkjan98.html, last accessed 10 April 2001. Readers should contact The World
Bank Pink Sheets Editor at http://www.worldbank.org/prospects/pinksheets/
Thomas, E. C., Gardner, E. A., Littleboy, M., & Shields, P. (1995). The cropping systems model PER-
FECT as a quantitative tool in land evaluation: an example for wheat cropping in the Maranoa area of
Queensland. Australian Journal of Soil Research, 33, 535–554.
Van Lanen, H. A. J., Hack-ten Broeke, M. J. D., Bouma, J., & De Groot, W. J. M. (1992). A mixed
qualitative/quantitative physical land evaluation methodology. Geoderma, 55, 37–54.

Web references for CLIPS

Athena Information Services (2000). Available: http://ourworld.compuserve.com/homepages/olsont/


HOMEPAGE.htm, last accessed 31 July 2001.
FuzzyCLIPS (2001). Available: http://ai.iit.nrc.ca/IR_public/fuzzy/fuzzyClips/fuzzyCLIPSIndex.html,
last accessed 31 July 2001.
Intelligent Software Professionals (2001). Available: http://www.isphouston.com, last accessed 31 July
2001.
Mark Tomlinson—CLIPS Dynamic Link Library / ActiveX (2001). Available: http://ourworld.compu-
serve.com/homepages/marktoml/clipstuf.htm, last accessed 31 July 2001.
NASA CLIPS Rule Based Language (2001). Available: http://www.siliconvalleyone.com/clips.htm, last
accessed 31 July 2001.
CLIPS—A Tool for Building Expert Systems (1999). Available: http://www.ghg.net/clips/CLIPS.html,
last accessed 31 July 2001.
CLIPS Download files (2000). Available: http://www.ghgcorp.com/clips/download, last accessed 31 July
2001.

S-ar putea să vă placă și