Degree of site suitability measurement in a GIS: the effect of standardisation method

Degree of site suitability measurement in a GIS: the effect of standardisation method

Site suitability analysis is performed to identify suitable land units (i.e. grid cells) for a specific purpose so that management decisions can be made in a site-specific manner. However, these grid cells are rarely equally suitable in the real world. They may vary substantially in their degree (or level) of suitability. Yet, the discrimination between suitable cells is often beyond the scope of conventional site suitability analysis. Widening the scope of conventional site suitability analysis to include a degree of site suitability (DoSS) measurement is therefore crucial for managing sites in a truly site-specific manner. Conventionally, site suitability analysis involves weighted linear combination (WLC) of standardised input factors (e.g. land use, slope, distance from stream, etc.) within a Geographic Information Systems (GIS) framework. In a conventional site suitability analysis, factor attributes are standardised using discrete classification method. Yet, the effect of this standardisation method on the DoSS measurement is unknown. Therefore, the objective of this study was to quantify the effect of the discrete classification methods of input factor attribute standardisation on the DoSS measurement.

In this study, seven input factors affecting the suitability of an agricultural land for site-specific application of animal waste as fertiliser were selected, pre-processed and standardised. Discrete classification method of standardisation, which replaced continuous or discrete factor attributes with a fixed number of differentially weighted classes, was employed. Three different classification and weighting schemes were adopted. Firstly, the attributes of each input factor were classified in up to five equal-sized classes to examine the effect of class number on the DoSS measurement. These classes were weighted with equally incremented weights that added up to 100. Secondly, they were classified into three sets of three classes each using equal area, equal interval and defined interval methods of classification to examine the effect of the class size on the DoSS measurement. These classes were also weighted with equally incremented weights that added up to 100. Thirdly, the attributes of each input factor were classified into two sets of three classes each, using equal area method of classification to examine the effect of differential weighting on the DoSS measurement. These sets were respectively weighted with equally and unequally incremented weights that added up to 100. Finally, the standardised input factors were correspondingly combined within GIS framework to produce 10 different composite maps (i.e. five for varying class number, three for varying class size and two for varying class weight). The DoSS measurements of each of the composite maps was quantified using the descriptive statistical parameters such as weighted average (WA), coefficient of variation (CV), value range (VR), and coefficient of skewness (CS) to make them comparable.
The conventional discrete classification method of standardisation resulted in a series of suitability maps that varied widely depending on the class number, the class size, and the method of weighting the classes. The WA varied between 700 (CV=0 & VR=0) and 221.9 (CV=6.31 & VR=100) for class number ranging between one and five. The WA for various class sizes and weight distribution between classes were less dramatic. However, they have resulted in DoSS measurements that were clustered and skewed.

The comparisons of results from these tests have highlighted the inconsistencies in the DoSS measurement when using various discrete classification methods of input factor attribute standardisation. It was found that the variations in terms of the class number, the class size, and the weight distribution between classes were the major contributing elements towards measurement inconsistencies. Therefore, it was concluded that the usefulness of this method of standardisation is limited for obtaining a comparable and repeatable DoSS measurement unless a more robust technique could be developed through further research.