Using GIS-based continuous methods for assessing agricultural land-use potential in sloping areas

The dynamic nature of land qualities both in space and in time suggests that, whatever the stages of the development of an area, it is always crucial to assess land qualities in terms of their potential and suitability for specific kinds of land use on a sustainable basis. In this paper we demonstrate an integrated method of biophysical land-suitability assessment, which combines two basic land-evaluation principles: allowing trade-offs among evaluation criteria, and a limiting condition approach. Decision criteria consist mainly of biophysical parameters, based on internal and external groups of land properties. Internal variables include various physical and chemical characteristics of soils, whereas external variables are those of topographic attributes. A fuzzy set methodology was employed to calculate values of a membership function, MF (0 4 MF 4 1:0), of each land variable. These fuzzy membership values were then combined to produce land-suitability indices, LSIs (0 4 LSIs 4 1:0), which can be used to assess the potential of land for a nominated land-use type. The analysis procedure implements a convex combination, which permits trade-offs between criteria of the same group, and a multiplicative function, which takes no account of compensation between groups (internal and external variables). A limiting condition approach is also incorporated in this system to resolve shortcomings found when using a convex combination technique alone. The results show that LSIs are consistent with categorical classes generated from applying the well-established method based on the Food and Agriculture Organisation framework for land evaluation. However, the LSIs, as derived from a cell-by-cell operation, produce more detailed subdivisions of land in terms of their potential for a given purpose. The main advantages of using this model include: (1) it minimises subjectivity in the evaluation procedure; (2) it takes into account the combined effects of the factors considered, while preserving information on biophysical constraints for a given purpose; and (3) it is amenable to further computer-based applications, especially for use in geographical information systems. DOI:10.1068/b2706 } Permanent address: Department of Soil Science, Hasanuddin University, Makassar 91425, Indonesia. assignments: true or false (McBratney and Odeh, 1997). Davidson et al (1994) identified the main disadvantages of land evaluation based purely on Boolean logic as follows: (1) the masking of key and positive land properties by less important ones which may depress the overall suitability class; (2) the inability of the system to take into account the effect of properties which happen to have values near to the class boundary. Such phenomena may have a significant effect on results. Burrough et al (1992) have shown that, with this technique, a large number of suitable areas were rejected. This also occurred in the study of Heuvelink and Burrough (1993). The use of a CF model (with fuzzy set functions) in a geographic information system (GIS) offers some flexibility in spatial-based analytic procedures. According to McBratney and Odeh (1997), fuzzy set theory has been used in land evaluation to solve problems related to ambiguity and vagueness, and to handle inexactness. When such a method is used within GIS, information sets from different formats can easily be integrated, spatial georeference of information is preserved, and the results of analyses may be presented in continuous grades and in more realistic forms. In New South Wales (NSW), Australia, basic soil surveys on reconnaissance soil landscape mapping are being carried out by the Department of Land and Water Conservation (DLWC). The project is currently very active in the regions where development is likely to occur. The data inventory provides an aid to the formulation of land-use planning based on sustainable perspectives, and has become one of the major sources of soil information for use in a wide range of environmental studies in the state. The NSW soil data system, SALIS (Soil and Land Information System) has also been developed by DLWC staff to store soil survey information, at over 40 000 sites across NSW, on a spatial basis. However, it is generally acknowledged that use of soil landscape units as the basis for biophysical land-resource assessment is too restrictive, because a lot of variation occurs, particularly in relation to the slope within each mapping unit. In land-evaluation perspectives, such variation can be dealt with by using other available information such as a DEM within GIS. The aims of this research project are three-fold. The first is to develop analytical procedures of land-suitability evaluation in sloping areas, by using the existing concept of fuzzy set methodology, with a particular emphasis on the Semantic Import model. The second is to show how the rigid soil landscape units could be dealt with by using developed techniques to establish detailed subdivision of land based on its qualities. The final aim is to demonstrate how this model can be implemented within GIS. The procedures described in the sections that follow are intended first to give some general overview about the existing land-evaluation system using fuzzy set methodology, followed by the method used in this project. The final sections then deal with application and model testing. 2 Fuzzy set methodology in land evaluation Fuzzy set theory (Zadeh, 1965) has been widely adapted for use with environmental applications. It has been used for soil classification (McBratney and de Gruijter, 1992; Triantafilis and McBratney, 1993), land-evaluation studies (Burrough, 1989; Burrough et al, 1992; Tang and van Ranst, 1992), and has also been applied with GIS technology (for instance, Banai, 1993; Beinroth et al, 1998; Davidson et al, 1994; Wang et al, 1990). A recent comprehensive review of the application of fuzzy sets in soil science is given in McBratney and Odeh (1997). The central aim of using the fuzzy set classification technique is to give the solution to the problem that may occur when, as in using the Boolean set (crisp set), a membership value is expressed solely as being true or false (Burrough et al, 1992); in other words, being full (MF value 1) or none (MF value 0) 4 S Baja, D M Chapman, D Dragovich