Modelling smallholder land use decision-making with ABM; Assessing the robustness and system assumptions of the LUSES Model

Abstract

The LUSES-model is an Agent-Based model of decision-making processes in a smallholder farming community, based on the Consumat-framework. A further (quantitative) evaluation of the LUSES-model was required before further development and exploration of the model. In addition to this, further model evaluation could add to the knowledge on the use of the socio-psychological Consumat approach in land use ABMs. This thesis resulted in a quantitative evaluation of the LUSES model, providing a validation of the model output and internal validation to the model by assessing the parameter robustness (including interaction effects) and exploring the model assumptions. The research is built on a literature study of the most relevant concepts for model evaluation. After preparatory analyses, the robustness of the parameters of the LUSES model was assessed by a One-factor-at-a-time sensitivity analysis. A Sobol’ sensitivity analysis was performed to analyse the interaction effects. The system assumptions were explored by analysing alternative scenarios. The research showed a sensitivity of the model to certain price changes. A large drop of the prices of a crop that used to be steady had extreme effects on almost all output variables: fluctuations and extreme values, especially for more extreme parameter values. The analysis also showed an extreme lack of response to parameter changes in the social need satisfaction variable, though this variable was shown to contribute though interaction effects. The intricacy of this ABM was shown in the Sobol’ analysis, which indicated that all main parameters (directly or indirectly) effect almost all output variables. The Existence need satisfaction variable is most influential taking into account interaction effects, the Aspiration variable is most influential without. No abnormalities were found that could not be explained by taking a closer look at the model logic, though the analysis of the system assumptions showed the model should not be used with settings that deviate from the baseline settings or parameter bounds that were tested in the OFAT analysis. From this can be concluded that the LUSES-model is very sensitive to price changes. This is influenced by the Cognitive effort variable. The Aspiration and Existence need variables have the strongest influence on the output variance. The model is sound, but the use of parameter settings that deviate from the baseline settings should be attempted with caution.