Uncertainty in Multi-Criteria Evaluation Techniques When Used for Land Suitability Analysis

Abstract

Uncertainty analysis is rarely considered in the application of predictive models in agriculture, resource planning and land suitability analysis. Uncertainty in modeling land suitability for agricultural production arises from a variety of sources. An important source of error is due to uncertainty in model inputs and parameters, especially in the case of multi-criteria analysis requiring data from physical measurements or expert opinion from regional workshops. The concept, scope and taxonomy of uncertainty analysis are discussed in the context of resource planning and land suitability analysis. The model used for land suitability was derived using the Analytic Hierarchy Process (AHP) originally introduced by Saaty in the mid 1970s. The general approach is also appropriate to modeling suitability to pasture and forestry as well as agricultural crops. The deterministic AHP approach produces point estimates only, with no indication of error or confidence in the output. We have integrated the AHP approach with a stochastic simulation model for uncertainty assessment. Since the AHP approach is deterministic, procedural adjustments are required to estimate uncertainty in predictions. The approach taken was to represent expert judgements and ratings by probability distributions and to implement a graded series of stochastic simulations. Variable weight values were subject to constraints of range and unit-sum for each level of the hierarchy in the AHP model. Results for uncertainty analysis are presented for land-use suitability in south-west Victoria in Australia for the crops ryegrass/sub-clover and winter wheat. The work was carried out in the context of a program supporting climate change adaptation funded by the Victorian Government. Estimates of uncertainty for the AHP approach were conservative in nature and a primary objective was to explore and develop further a generalised approach to uncertainty assessment for the AHP model and similar multi-criteria evaluation techniques.