Does optimal complexity that is required to model a basin depend on its hydrologic characteristics?

Abstract:

It is necessary to understand system complexity in order to better understand and model basin scale hydrology. We introduce an index of basin complexity based on its optimal model complexity. We use a simple hydrological model based on k-nearest neighbors and quantify its complexity for different parameterizations. Further, by a model of optimal complexity we mean a robust representation of the underlying processes (here to predict streamflow), i.e. it remains a good conceptualisation of underlying processes as new information in the form of new datasets are introduced. Relationships between hydrologic characteristics of around 430 MOPEX basins such as location, hydrologic ratios, saturated hydraulic conductivity and NDVI and respective basin complexities are investigated. The results suggest that optimal complexity and aforementioned hydrologic characteristics are linked, thereby suggesting that model complexity may have a physical interpretation. It also suggests that optimal model complexity can be interpreted as an index of physical similarity, and that minimum complexity that is required to model a basin depends on its hydrologic characteristics.