A framework to assess model structure stationarity

Abstract

In geosimulation models it is often, implicitly or explicitly, assumed that the relationship between the system state change and its explanatory processes, i.e. model structure, is stationary. Potential systemic changes influencing the stationarity of this relationship are thereby disregarded. We present a framework to assess model structure stationarity and apply it on a case study in which we model land use change in São Paulo state, Brazil, from 2004 to 2012. We find that systemic changes do occur during this period, which is a warning that the assumption of stationarity is likely to cause an underestimation of simulated future uncertainty.