A spatial-statistical approach to the visualisation of uncertainty in land suitability analysis

Abstract

A formal GIS-based procedure known as land suitability analysis (LSA) is used to determine the most appropriate crops for cultivation in different geographical locations. The approach is based on multi-criteria decision analysis utilising biophysical measurements (including rainfall, temperature, pH) and expert opinion captured from regional workshops. The issue of uncertainty in model predictions and its importance is discussed, and a method is described for its analysis and visualisation in LSA maps. Experimental results using Monte Carlo simulation are presented for ryegrass/sub-clover and winter wheat crops grown in south-western Victoria. It was found that uncertainty in the prediction of land suitability, as described by the coefficient of variation (CV), ranged from 0.13 to 0.18 for ryegrass, and much higher at 0.28 to 0.30 for the winter wheat crop. Results showed that, for close matches between crop type and production location, over 90 percent of the standard deviation in the prediction was accounted for by uncertainty in expert opinion rather than uncertainty in biophysical data.