Comparison of a deductive and an inductive approach to specify land suitability in a spatially explicit land use model

In this paper, two research approaches to specify the relation between land use types and their explanatory factors are applied to the same modelling framework. The two approaches are used to construct land suitability maps, which are used as inputs in two model applications. The first is an inductive approach that uses regression analysis. The second applies a theoretical, actor decision framework to derive relations deductively using detailed field data. Broadly speaking, this classification coincides with the distinction between empirical and theoretical models and the distinction between deriving process from pattern and pattern from process. The two modelling approaches are illustrated by a scenario analysis for a case study in a municipality in the Philippines. Goodness-of-fit of the deductive approach in predicting current land use is slightly lower compared to the inductive approach. Resulting land use projections from the modelling exercise for the two approaches differ in 15 percent of the cells, which is caused by differences in the specification of the suitability maps. The paper discusses the assumptions underlying the two approaches as well as the implications for the applicability of the models in policy-oriented research. The deductive approach describes processes explicitly and can therefore better handle discontinuities in land use processes. This approach allows the user to evaluate a wide range of scenarios, which can also include new land use types. The inductive approach is easily reproducible by others but cannot guarantee causality. Therefore, the inductive approach is less suitable to handle discontinuities or additional land use types, but is well able to rapidly identify hotspots of land use change. It is concluded that both approaches have their advantages and drawbacks for different purposes. Generally speaking, the inductive approach is applicable in situations with relatively small land use changes, without introduction of new land use types, whereas the deductive approach is more flexible. The choice of modelling approach should therefore be based on the research and policy questions for which it is used.