Interactive uncertainty and sensitivity analysis: building trust on soil erosion models and making better informed decisions

Soil erosion models are complex and inherently uncertain. This limits their perceived trustworthiness and hinder their use in practical decision making, in particular by non-technical users. Paradoxically, understanding, reducing (when possible) and explicitly incorporating this uncertainty is crucial for building trust and improving decision-making. For this purpose, we are developing iMPACT-erosion, an open-source soil erosion modelling toolbox based on Jupyter Notebooks. Integrating interactive elements and visualization, iMPACT-erosion fosters a more fluent user-model conversation, targeting students, professors, researchers, and decision-makers. The toolbox comprises three components: iMPACT-start (basic concepts and initial steps), iMPACT-test (model evaluation), and iMPACT-explore (scenario assessment). This work focuses on iMPACT-test and iMPACT-explore, emphasizing the combination of Monte Carlo simulations and interactivity. This approach addresses not only uncertainty quantification and attribution but also tests model behaviour plausibility, identifies controlling factors and conditions leading to unsustainable soil loss rates, and reduces uncertainty by optimizing model evaluation against field measurements. This toolbox empowers users to gain a comprehensive understanding of model behaviour, assess model suitability for specific applications, and ultimately make more informed and robust decisions regarding soil conservation and land management.