Previous PageTable Of ContentsNext Page

A conceptual model for assessing agricultural land suitability at a catchment level using a continuous approach in GIS

Sumbangan Baja1, David M. Chapman, and Deirdre Dragovich

Division of Geography, School of Geosciences, The University of Sydney
NSW 2006, Australia; ph +61-2-93514589, fax +61-2-93513644
1
Corresponding author, email: sbaja@mail.usyd.edu.au

Abstract

For the purposes of maintaining and developing agricultural land use at a catchment level on a spatial and timely basis, it is vital to identify the levels and geographical patterns of biophysical constraints and hence, land suitability for a given purpose. This is the central task of land evaluation. The paper presents a spatial modelling procedure for the assessment of land suitability using available biophysical information. The soil landscape map (scale 1:100,000), which is widely used for planning purposes in New South Wales, Australia was used in conjunction with a Digital Elevation Model (DEM) to derive land suitability indices (LSIs) for cropping. The LSIs were modelled based on fuzzy set methodology in geographical information systems (GIS). The available land suitability classification map (at scale 1:50,000) produced by NSW Agriculture (1995) was employed for cross-checking with the model outputs. It reveals that the output is consistent with such a land suitability classification scheme, and the model developed produces fine discrimination of land units. The procedures developed show the significance of this model for micro-mapping purposes at a comparatively low cost.

Introduction

From a land use planning perspective, the systems of land use should be well matched with the inherent characteristics of the land to ensure long-term productivity and sustained use of the land. Land suitability assessment (LSA) plays an important role in this regard. Imhof et al. (2000) stated the following:

“Enhancement of existing mapping is required to allow for improved matching of land use with land suitability and capability, and the identification of areas where higher value sustainable land uses can be promoted. This information will also be needed to provide the basis for disseminating information on best management practices that allow increased production while enhancing the quality of land resources and preventing off-site degradation.”

Land suitability assessment is concerned with ‘the process of estimating the potential of land for alternative kinds of land use’ (Dent and Young, 1981). Land use could be in the contexts of agriculture, engineering, forestry, or recreation; but agriculture may be the most popular area where LSA is applied. In the agricultural context, the evaluation is directed to a specific kind of land utilisation (FAO, 1976), at specified units of homogeneous delineation. The size and homogeneity of the units depend on the level of detail of information required for a given purpose. Traditionally, LSA requires the comparison of the land qualities with the requirements or limitations of a specific land use to determine the overall potential of the land. The results are expressed as a spatial distribution of suitability classes or indices for a given purpose.

In New South Wales (NSW), use of soil landscape maps has been very popular for planning purposes. For use at a large catchment, the data resolution is considered sufficient, but at more detailed scale, for instance, a small scale catchment (e.g., 1000 sq. km or less) it might be too restrictive. The main shortcoming is that a considerable variation of slope commonly occurs within each soil landscape mapping unit. Therefore, use of ancillary data such as a digital elevation model (DEM) is useful in enabling such variation to be examined, and offers an opportunity to observe spatially-based approaches for land suitability assessment in a GIS.

This paper presents a methodological approach and application of a continuous method of LSA which is designed for use at a small catchment level, using the commonly-available spatial data in NSW: soil landscapes series (mapped at 1 : 100,000 scale), and a three-second DEM. A theoretical background on LSA is presented first, followed by methodology and application in GIS environment, before the results of model testing including the analyses of sensitivity and spatial comparison are presented.

Theoretical background

Key issues in land use and land suitability

A land suitability assessment should be viewed from three different perspectives: (i) productivity, (ii) workability, and (iii) sustained use of the land. A land area where agricultural overproduction occurs cannot necessarily be associated with that of land highly suitable for that particular type of land use. Increasing yields could possibly result from an intensive use of fertilizers and irrigation. With current technology, land with low workability like steep slopes, shallow profiles, and high content of cobbles, boulders, and stones, can still be managed to attain a certain level of productivity, but at a relatively high cost. The issue is whether or not that area will be capable of sustaining such a use for a prolonged period. In a region where land degradation becomes one of the main issues of environmental management, land evaluation systems should place more emphasis on the sustainable use of land.

From the perspective of catchment management, LSA plays an important role in the identification of areas or zones where higher value sustainable land uses can be highlighted, and where the use of a particular land utilisation type may be restricted. Because a catchment is a discrete biophysical system, mismatch of the land with the land use on a particular area will result in not only on-site effects but also off-site degradation for a prolonged period.

Components of land suitability assessment

An LSA procedure, in essence, involves eight main components (Figure 1). The assessment of land quality for a specific type of land use should be based on land use requirements and constraints. Such requirements and constraints are then used as the basis for establishing what are termed ‘evaluation criteria’ or ‘decision criteria’. With reference to these decision criteria, a set of algorithms is then employed to match the existing quality of land and the requirements of that particular type of land use. The matching procedure (see FAO, 1976) then gives rise to a ranking of the potential of land for a given purpose, whether categorical or continuous grades. Regardless of the approaches employed, the final result of land evaluation is a map that portrays the divisions of the area of interest into suitability classes or land unit indices for a nominated land use. The first is produced from an approach called a categorical system, and the second from a continuous method of land suitability evaluation. This latter method is the one discussed in this paper. Furthermore, as seen in Figure 1, model validation is also considered as one of the main components of land evaluation.

Figure 1 Main components of land suitability assessment

2.3 Continuous methods of land suitability assessment

Use of a continuous method for land resource assessment has recently received considerable attention in soil and environmental studies. Such a method is used to refine a Boolean-based technique, and to improve analytical capability of a categorical-based land evaluation system. For such purposes, a fuzzy set classification technique is commonly used (e.g., Burrough, 1989; Wang et al., 1990; Burrough et al., 1992; Tang and van Ranst, 1992; Oberthur et al., 2000), but there are also other techniques for undertaking these tasks. For example, use of the Wymore’s standard scoring function (SSF) and its derivatives has also been very popular (e.g., Harris et al., 1994; Karlen and Stott, 1994; Glover et al., 2000). Other similar types of continuous methods have also been used by Hess et al. (2000) and Hellkamp et al. (2000) in assessing ecological condition of agricultural lands.

A continuous method of land suitability evaluation can be viewed in two different perspectives. In the first, criterion ratings are made based on the continuous values (e.g., from 0 to 1) for individual mapping units which have ‘hard’ boundaries. Here, an overall suitability (or occasionally called Joint Membership Function) of a given hardened mapping unit is presented as a value ranging from 0 to 1 (see for example, Wang et al., 1990; Davidson et al., 1994). Use of other similar methods as demonstrated by Karlen and Stott (1994), Glover et al. (2000) also fall in this category, where suitability is given in continuous grades, while the outputs use hard boundaries. In the second, both suitability ratings and output presentations are given in continuous grades. Examples for such a method can be seen in Triantafilis and McBratney (1993), Burrough et al. (1992), and Baja et al. (2001). In a GIS environment, these suitability levels are represented on a cell-by-cell basis.

This paper presents suitability rating strategies based on the variant of fuzzy set methodology, called a ‘semantic import’ (SI) model (see also Burrough et al., 1992; McBratney and Odeh, 1997; Burrough and McDonnell, 1998).

Model structure

For a catchment system, there are at least two important groups of land attributes for land suitability analysis: inherent qualities of soils and external characteristics. The former are soil attributes which have the function for accommodating plant growth, while the latter are those determining the level of ‘workability’ (see FAO, 1976), runoff, sedimentation, and erosion in the catchment. In this study, soil attributes and topographic elements were used to represent both groups. The analytical procedures developed entail the following steps (see also Figure 2):

Step 1: Determination of individual ratings of land characteristics

Each of the land attributes within the two groups of variables (i.e., soils and topography) are first independently examined and rated using an appropriate SI-based fuzzy set model. The rating of soil attributes is done on the basis of the available soil mapping units (SMUs) which have ‘crisp’ boundaries, while that of topography is based on the DEM which has continuous values. Therefore, at this stage the unit for individual soil attributes is SMUs, while for topography (slope) it is data cells. Each land attribute is rated with values ranging from 0 (minimum) to 1.0 (maximum) according to its suitability for a nominated land use.

Step 2: Derivation of group ratings of land characteristics

Land attributes within each group are then combined using a ‘convex combination’ function, which utilises criterion weights. This operation permits tradeoffs between land attributes of the same group. For instance, low soil organic matter content can be compensated for by an excellent soil texture; degree of compensation depends on the quality level of land attributes under consideration and criterion weights. The rating for this group of land attributes is represented as joint membership function (JMF), also ranging from 0 to 1.0.

Figure 2 Basic structure of the model developed

Step 3: Calculation of overall land suitability indices

At this stage, two sets of JMFs would have been determined: one for soil attributes, and the other for topography. A multiplicative function, which is based on the cell-by-cell operation in GIS database, is then employed between the two groups of land variables to produce the overall indices of land suitability for cropping. This reveals that no compensation is allowed between these two groups of land properties, meaning that a land area with very steep slope cannot be compensated for by, for instance, excellent quality of soil profile, or vice versa (see Storie, 1978; Pierce et al., 1983; and Sys, 1985). Such a multiplicative operation then gives rise to land suitability indices (LSIs) which are expressed in continuous values, ranging from 0 (not suitable) to 1.0 (highly suitable).

Methodology

Study area

The study area selected covers some portions (approximately 37,000 ha) of the Lower Hawkesbury-Nepean catchment, in the west of Sydney Region (Figure 3). Various types of land use exist in the area. Agriculture occupies around 40 %, mainly pasture, and various types of orchards and market gardening. Forest had long been the dominant land use in the area, particularly in the western sections. Scattered blocks of residential sites exist, mainly along the main roads, except in Penrith, Richmond, and Windsor where considerable urban/rural residential development has taken place.

The mean annual rainfall recorded in Richmond is 860 mm (Monthly Weather Review, Bureau of Meteorology, NSW), with the wettest months occurring from January to March and the driest from July to September. Various geological structures exist in the study area forming two distinct physiograpic regions: the Cumberland Lowlands and the Blue Mountains Plateau (Jones and Clark, 1991). The complex nature of geology in the study region gives rise to various geomorphic classes and soil types. Soil types, known as ‘soil landscapes,’ were originally developed from five different landscape groups: residual, colluvial, erosional, alluvial, and aeolian deposit (Bannerman and Hazelton, 1990).

Figure 3 Site location of study area

Fuzzy set methodology: semantic import (SI) model

Model functions used to calculate membership function (MF) of land attributes are depicted in Figure 4 (adapted from Burrough et al., 1992; Burrough and McDonnell, 1998). Model 1 (Figure 4) is used to determine the membership grades of land qualities with symmetric functions (Model 1 Figure 4), where only one ideal point or central concept exists. Another type of symmetric function is shown in Model 2 Figure 4, where the central concept consists of a range of values (from b1 to b2). Further, there are also situations where only the lower and upper boundary of a class have practical importance (Burrough and McDonnell, 1998). In such circumstances, an asymmetric function needs to be applied (Models 3 and 4). An asymmetric left function is used for the lower boundary of a class, while an asymmetric right is employed for an upper boundary. If MF(xi) represents individual membership value for ith land property x, then in the computation process these model functions (Models 1 to 4) take the following form:

MF(xi) = [1/(1 + {(xi – b)/d}2)] …………………………………….....….…………………….. (1)

In addition, the following forms also apply to Models 2, 3 and 4 in calculating membership functions of land attributes:

For optimum range (Model 2):

MF(xi) = 1 if (b1 + d1) < xi < (b2d2) …..…………………..…..……...……….…...……... (2)

For asymmetric left (Model 3):

MF(xi) = [1/(1 + {(xi – bi – d1)/d1}2)] if xi < (b1 + d1) .………...……..….……..….…….. (3)

Figure 4. Fuzzy set models used for rating land attributes (adapted from Burrough et al., 1992; Burrough and McDonnell, 1998). MF represents membership function of land properties; d, d1, and d2 are width of transition zone, i.e., x at MF = 0.5 or crossover point; LCP and UCP are lower and upper crossover points; and b, b1 and b2 are values of land attributes at the ideal point.

For asymmetric right (Model 4):

MF(xi) = [1/(1 + {(xi – b2 + d2)/d2}2)] if xi > (b2 – d2) .…….…….…...………..……….. (4)

The membership functions of individual land characteristics under consideration are then combined using a convex combination function to produce a joint membership function, JMF(X) as follows:

n

JMF(X) = Σλi MF(xi) ……………………………………………..…….....…………..…… (5)

i =1

where 0 < JMF(X) < 1, and 0 < MFxi < 1

n

Σλi = 1, and 0 < λi < 1

i =1

JMF(X) is joint membership function for soils and topography

λi is a weighting factor for the ith land property x

MF(xi) denotes a membership grade for the ith land property x

n indicates number of land attributes under consideration.

Evaluation criteria, fuzzy set model parameters, and weightings

As many as nine soil characteristics (internal land variable) and slope gradient (external variable) were selected as evaluation criteria for ‘cropping’ land utilisation type. Climate in the area is not limiting for most types of cropping (Zhang, 1989). Such criteria were adapted from Sys (1985) and Zhang (1989) (Table 1). Soil attributes defined by ‘hard’ boundaries were derived from the soil landscape map series for Penrith (Bannerman and Hazelton, 1990) and SALIS (Soil and Land Information System) data base, held in the Department of Land and Water Conservation, while slope was derived from a three-second digital elevation model (DEM) which is represented on a continuous scale.

Table 1 Evaluation criteria for cropping and selected types of fuzzy set models and MF parameters

Land variables

Type of
Data#

Model function

MF parameters for cropping

Weight

(2FD)

LCP

b

UCP

D$

Internal land variables

             

Available water capacity, AWC (%)

C

3

12.5

20

-

7.5

0.095

Site drainage

[O, 5]

4

-

1

3

2

0.190

Texture and structure*

[O, 5]

4

-

0

2

2

0.190

Cobbles, boulders, and stones (CBS)

[O, 5]

4

-

1

3

2

0.095

Solum depth

[O, 5]

3

3

5

-

2

0.190

Cation exchange capacity, CEC (topsoil)

[O, 5]

3

2

4

-

2

0.048

Organic carbon, OC (topsoil)

[O, 5]

3

3

5

-

2

0.048

Soil pH

C

2

4.75

5.5-8.0

8.8

0.85

0.048

Salinity, Electrical Conductivity (EC)

[O, 5]

4

-

1

2

1

0.190

External land variables

             

Slope gradient (%)

C

4

-

2

12.5

10.5

-

Note:

#C = cardinal; O = ordinal. [O, 5] means ordinal data with 5 categories: 1, 2, 3, 4, and 5 (Table 2), which were used for deriving MF

*MF for texture and structure were determined based on limitation degree from 0 (none) to 5 (very severe) (Zhang, 1989)

$ The value of d for soil pH is based on the LCP, as there is no pH value of soil units greater than 8.0 (or upper boundary of b value).

One of the crucial stages in implementing the SI-based fuzzy set methodology is to determine the values of such model parameters as b (ideal point), LCP (lower crossover point), UPC (upper crossover point), and d (width of transition zone) (see Burrough et al., 1992). These parameters were determined based on the optimal level of land property for b, and marginal level for d (LCP or UCP) for selected land use types (see guidelines in Zhang 1989; Burrough et al., 1992; and the available land suitability classification systems as proposed by Sys, 1985 and van Gool and Moore, 1998). According to the data sets in the Soil Landscape Series (Bannerman and Hazelton, 1990), two types of formats were used: cardinal (C) and ordinal (O) (Table 1). Table 2 presents categorical rankings of ordinal data, which were used for calculating membership functions (see Burrough et al., 1992).

Table 2 Categorical rankings of land attributes measured on an ordinal scale

 

Ordinal/categorical rankings

Land attributes

1

2

3

4

5

 

Very low

Low

Moderate

High

Very high

CEC (me%)

< 6

6 - 12

12 - 25

25 - 40

> 40

OM (%)

0 - 0.5

0.5 - 1.5

1.5 - 2.5

2.5 -5.0

> 5.0

Salinity, EC (dS/m)

0 - 2.0

2.0 - 4.0

4.0 - 6.0

6.0 - 8.0

> 8.0

Drainage class

Well

Moderately well

Imperfect

Poor

Very poor

Solum depth (m)

Very shallow

Shallow

Moderate

Slightly deep

Deep

< 0.15

0.15 – 0.4

0.4 – 0. 7

0.7 – 1.0

> 1.0

CBS*

Absent

Few

Frequent

Common

Abundant

*Cobbles, boulders, and stones

In the weighting procedure, these soil attributes were grouped into three categories, ranked according to their importance in descending order. Hence, attributes within each group were weighted equally. Such ranked groups, in descending order, are as follows: group I (soil texture and structure, effective depth, and salinity (EC)); group II (site drainage, AWC, and CBS); and group III (soil pH, OM, and CEC). The weights as seen in Table 1 are generated based on a two-fold difference in terms of degree of importance for the above ranked groups. Alterations of weightings were also applied to examine the sensitivity of the results, as will be described later.

GIS-based analyses and results

Calculating individual membership functions

As soil data sets were stored in the Arc/Info vector format, calculation of membership function of each soil characteristic was done in the attribute file (PAT). Therefore, the soil landscape unit is used as a delineation unit of individual properties in the spatial representation of membership function (MF) and joint membership function (JMF). Slope layer was in raster format, so that computation was performed on a cell-by-cell basis. The model function used for each criterion depends on the ‘trend of performance’ of the respective land attribute in giving a favourable condition for cropping (Baja et al., 2001). For instance, very low and very high values of soil pH are limiting for most agricultural applications, and the optimum values (i.e, central value, b with MF value = 1.0) might range from slightly acid to neutral or slightly alkaline (see Table 1). Therefore, a symmetric function (Model 2) would be the most appropriate one used for deriving MF for soil pH. With a similar rule, an asymmetric left model or ‘more is better’ is used for calculating membership values for soil depth, cation exchange capacity, and organic carbon content, and so forth. Such land attributes as slope, salinity, and the content of cobbles, boulders, and stones perform inversely.

Slope layer was represented in a grid format, as it was derived from raster-based DEM. The grid-based IDRISI 32 program was employed for computing membership function of slope cells.

Deriving JMF and LSI

Once the membership values of soil attributes have been determined, the JMF is then calculated using equation (5). Here a relative weight of each land attribute, as described earlier, is employed. Four alterations of weightings were made to allow for the analysis of sensitivity. As the computation of LSI will be done in a raster structure, the data conversion process was undertaken to transform the vector-based soil layer (JMF) to a raster format. At this stage, membership function for slope was already in a raster structure. To calculate the overall land suitability indices (LSIs) for cropping, a pixel-by-pixel based multiplicative function was employed between JMFs of soils and topography. Figure 5 shows the spatial distribution of LSIs derived from using a ‘two-fold difference’ weighting (2FD).

Sensitivity analysis and validation

Sensitivity analysis is intended to examine the extent of variation of outputs or model response resulting from the uncertainty in the input factors (Crosetto et al., 2000), either individually or in combination, while validation refers to the level of accuracy in the predictions of the model (O’Keefe et al., 1991). In the present study, both types of analyses were undertaken simultaneously. Four sets of weightings were chosen to produce LSI layers: equal weight (EQU), two-fold difference (2FD), three-fold difference (3FD), and 0.5, 0.3, and 0.2, respectively, for group I, II, and III (FTT). A three-fold difference of importance is considered maximum, because the difference between the lowest weight in group III and the highest in group I is nine-fold (i.e., 0.231 and 0.026, respectively). Davidson et al. (1994) used a slightly lower gap of criterion weights between those with the lowest and the highest importance.

Commonly-used approaches employed in validation include testing for predictive ability and comparison against performance standards (Harrison, 1991; O’Keefe et al., 1991). In this study, a cross-comparison technique was employed to evaluate model performance: testing for close agreement between outputs of the model developed and that deemed as a performance standard. Here, a 1 : 50 000 scale land suitability map (NSW Department of Agriculture, 1995) was employed as a performance standard. This scheme employs five classes (class 1 to 5). As the limitations increase with increasing class number, the number of suitable crops decrease. For instance, class 1 is suitable for most types of agricultural land use. Regular cultivation can only be considered on land with classes 1 and 2, and to some extent class 3, while classes 4 to 5 are for grazing.

For this model testing, an archetype area of 9,775 ha (coordinates 279050 to 290550 mE and 6276850 to 6285750 mN) was selected. The area covers a great deal of variation in land characteristics, and hence LSIs, and all land suitability categories exist and are well represented. A spatial comparison was then undertaken in GIS between the units of suitability class (NSW Department of Agriculture, 1995) and the corresponding pixels of land suitability indices. This procedure was similar to that employed by Tang and van Ranst (1992) and Davidson et al. (1994) in examining the relationships between the values of fuzzy membership values of land attributes and the individual suitability categories of land characteristics for a nominated land use. Figure 6 shows the relationships between the LSIs and suitability categories, and also reveals the sensitivity of the weights given and the variation of LSIs within each suitability class.

Figure 5 Land suitability indices (LSIs) for cropping; the index values range between and including 0 (very poor) to 1.0 (highly suitable)

Figure 6 Graph depicting relationship between LSIs and suitability class

Discussion and conclusion

It has been demonstrated that the model developed integrates two important land variables: soils and topography. The analytical procedures used infer that the danger of full tradeoffs between these two groups of variables can be avoided, because they were rated and evaluated separately, before being combined in one function (i.e., multiplicative operation). In other words, one variable with a poor index cannot be compensated for by another one having a good index. Moreover, use of a continuous function can resolve shortcomings found in a Boolean technique, in which key and positive land properties may be masked by less important ones (Davidson et al., 1994).

The analysis of sensitivity has shown that there is no important change of LSIs from the application of different sets of weightings. The overall variation of LSIs from different weights given seems to be sufficient for such a large area. Furthermore, based on spatial comparison analysis, it was found that the variations of LSIs are larger in the land units with low (poor) suitability classes. This accords to a land suitability classification study conducted by Burrough et al. (1992), who found that use of a Boolean-based categorical system of land suitability analysis had resulted in the rejection of considerable suitable areas. As a result, the poorer the suitability class, the higher the LSI variations, as indicated by the standard deviation in Figure 6. The model developed thus demonstrates its significance for fine discriminations of land quality in the area of interest.

It is recognised that more parameters are required in the assessment of external land characteristics on a catchment system, particularly those related to the properties that govern water runoff and erosion. Work is in progress in modelling potential soil loss, a parameter that may be later incorporated to the model described in this paper.

References

Baja, S., Chapman, D. M., Dragovich, D. (2001). Using GIS-based continuous methods for assessing agricultural land use potential in sloping areas. Environment and Planning B: Planning and Design (accepted, in press).

Bannerman, S.M., Hazelton, P.A. (1990). Soil Landscapes of Penrith 1 : 100,000 Sheet. Soil Conservation Service of NSW, Sydney.

Burrough, P.A. (1989). Fuzzy mathematical methods for soil survey and land evaluation. Journal of Soil Science, 40, 477-492.

Burrough, P. A., McDonnell, R. A. (1998). Principles of Geographical Information Systems, Spatial Information System and Geostatistics, Oxford University Press, New York.

Burrough, P.A., MacMillan, R. A. van Deursen, W. (1992). Fuzzy classification methods for determining land suitability from soil profile observations and topography. Journal of Soil Science, 43, 193-210.

Davidson, D.A., Theocharopoulos, S.P., Bloksma, R.J. (1994). A land evaluation project in Greece using GIS and based on Boolean and fuzzy set methodologies. International Journal of Geographic Information Systems, 8, 369-384.

Dent, D. and Young, A. (1981). Soil Survey and Land Evaluation. George Allen & Unwin, Boston.

FAO. (1976). A Framework for land evaluation. FAO Soils Bulletin No. 32, Rome.

Glover, J.D., Reganold., J.P., Andrews, P.K. (2000). Systematic method for rating soil quality of conventional, organic, and integrated apple orchards in Washington State. Agriculture, Ecosystem, and Environment, 80, 29-45.

Harrison, S.R. (1991). Validation of agricultural expert system. Agricultural Systems, 35, 265-285.

Hellkamp, A.S., Bay, J.M., Campbell, C.L., Easterling, K.N., Fiscus, D.A., Hess, G.R., McQuaid, B.F., Munster, M.J., Olson, G.L., Peck, S.L., Shaver, S.R., Sidik, K., Tooley, M.B. (2000). Assessment of the condition of agricultural lands in six Mid-Atlantic States. Journal of Environmental Quality, 29, 795-804.

Hess, G.R., Campbell, C.L., Fiscus, D.A., Hellcamp, A.S., McQuaid, B.F., Munster, M.J., Peck, S.L., Shaver, S.R. (2000). A conceptual model and indicators for assessing the ecological condition of agricultural lands. Journal of Environmental Quality, 29, 728-737.

Imhof, M., Rampant, P., Bluml, M. (2000). The future of land resource assessment in Victoria. ACLEP Newsletter, 9, 6-16.

Jones, D.C., and Clark, N.R. (Eds). (1991). Geology of the Penrith 1 : 100,000 Sheet 9030. NSW Geological Survey, Sydney.

Karlen, D.L., Stott, D.E. (1994). A framework for evaluating physical and chemical indicators of soil quality. In: Doran, J.W., Coleman, D.C., Bezdicek, D.F. and Stewart, B.A. Defining Soil Quality for Sustainable Environment. SSSA Special Publication No 35. Soil Science Society of America, Inc., Madison, 53-72.

McBratney, A.B., Odeh, I.O.A. (1997). Application of fuzzy sets in soil science: fuzzy logic, fuzzy measurements and fuzzy decisions. Geoderma, 77, 85-113.

Nogues, J., Herrero, J. Rodriguez-Ochoa, R., Boixadera, J. (2000). Land evaluation in a salt-affected irrigation district using an index of productive potential. Environmental Management, 25, 143-152.

NSW Department of Agriculture. (1995). Agricultural Land Classification Atlas: Sydney Basin, including the Lower Nepean-Hawkesbury Catchment 1: 50 000 Scale, NSW Government, Orange.

Oberthur, T., Dobermann, A., Aylward, M. (2000). Using auxiliary information to adjust fuzzy membership functions for improved mapping of soil qualities. International Journal of Geographical Information Science, 14, 451-454.

O’Keefe, R.M., Osman, B., Smith, E.P. (1991). Validating expert system performance. In: Gupta, U.G. (Ed.). Validating and Verifying Knowledge-Based System. IEEE Computer Society Press, Washington, 2 –11.

Pierce, F.J., Larson, W.E., Dowdy, R.H., Graham, W.A.P. (1983). Productivity of soils: assessing long-term changes due to erosion. Journal of Soil and Water Conservation, 38, 39-44.

Storie, R.E. (1978). Storie Index Soil Rating (Revised). University of California, Division of Agricultural Science special publication 3203. University of California.

Sys, C. (1985). Land Evaluation: Part I. University of Ghent, Ghent.

Tang, H., van Ranst, E. (1992). Testing fuzzy set theory in land suitability assessment for rainfed grain maize production. Pedologie, 41, 129-147.

Triantafilis, J., McBratney, A. B. (1993). Application of Continuous Methods of Soil Classification and Land Suitability Assessment in the Lower Namoi Valley, Divisional Rep. No. 121, Division of Soils, CSIRO Australia, Melbourne.

van Gool, D., Moore, G. (1998). Land Evaluation Standards for Land Resource Mapping: Guidelines for Assessing Land Qualities and Determining Land Capability in South-West Western Australia (Draft), Resource Management Technical Report No. 181, Natural Resource Assessment Group, Agriculture Western Australia, Perth.

Wang, F., Hall, G.B., Subaryono. (1990). Fuzzy information representation and processing in conventional GIS software: database design and application. International Journal of Geographical Information Systems, 4, 261-283.

Zhang, H. (1989). A Land Suitability Evaluation System for Specific Rural Purposes in New South Wales. A PhD Thesis (unpublished). The University

Previous PageTop Of PageNext Page