Model Decisions and Model Cultures

Abstract:

Give different people the same recipe and ingredient, the final dish will still taste different. The same is true when talking about modelling; give different modellers the same model, the results might still look different. I investigated the scientific practice around modelling, rooted in Science and Technology Studies. This is particularly relevant for the use of hydrological models, now that more and more scientific studies are conducted, and conclusions are drawn, solely based upon models.

It is widely acknowledged that environmental models are prone to uncertainty, still scientists and decision makers often perceive models as being something ``objective''. However, in the process from problem to simulation, there are several steps that require the modeler’s input and decisions. These can influence or even steer the model’s result. Addor & Melsen (WRR, 2019) demonstrated that model choice is closely related to the institute of the first author. This leads to a geographical bias in model use across the world. In this follow-up study, fourteen modellers from three different institutes were interviewed (interviews took place in Feb - May 2020), to track their motivations for certain model decisions, for instance related to resolution and data use.

Based on the interviews, three pillars that influence modelling decisions were identified. Situational factors related to infrastructure often determine which model is used (for instance a high performance cluster with a compiled model available), social situational factors influence several modelling decisions (for instance, colleagues that recommend an interpolation technique), and finally personal idiosyncrasies; the modeler simply prefers a certain procedure or tinkered with the procedure in interaction with the model. The results show that model use is a social activity, and as such, models should be perceived as social constructs.