Advertisement

Learning from model improvement: On the contribution of complementary data to process understanding

Authors

  • Fabrizio Fenicia,

    1. Public Research Center – Gabriel Lippmann, Belvaux, Luxembourg
    2. Water Resources Section, Faculty of Civil Engineering and Geosciences, Delft University of Technology, Delft, Netherlands
    Search for more papers by this author
  • Jeffrey J. McDonnell,

    1. Water Resources Section, Faculty of Civil Engineering and Geosciences, Delft University of Technology, Delft, Netherlands
    2. On leave from Department of Forest Engineering, Oregon State University, Oregon, USA.
    Search for more papers by this author
  • Hubert H. G. Savenije

    1. Water Resources Section, Faculty of Civil Engineering and Geosciences, Delft University of Technology, Delft, Netherlands
    Search for more papers by this author

Abstract

[1] A priori determined model structures are common in catchment rainfall-runoff modeling. While this has resulted in many ready-to-use modeling tools, there are several shortcomings of a one-size-fits-all model structure. The uniqueness of catchments with respect to their hydrological behavior and the need to adapt model complexity to data availability challenge this status quo. We present a flexible approach to model development where the model structure is adapted progressively based on catchment characteristics and the data described by the experimentalist. We demonstrate this approach with the Maimai catchment in New Zealand, a location with a large availability of data, including stream discharge, groundwater levels, and stream isotope measurements. Different types of data are introduced progressively, and the architecture of the model is adjusted in a stepwise fashion to better describe the processes suggested by the new data sources. The revised models are developed in a way to strike a balance between model complexity and data availability, by keeping models as simple as possible, but complex enough to explain the dynamics of the data. Our work suggests that (1) discharge data provides information on the dynamics of storage (represented by the “free” water in the reservoirs) subject to pressure wave propagation generated by rainfall into the catchment, (2) groundwater data provides information on thresholds and on the contribution of different portions of the catchment to stream discharge, and (3) isotope data provides information on particle transport and mixing of the rainfall with the storage present in the catchment. Moreover, while groundwater data appear to be correlated with discharge data, and only a marginal improvement could be obtained adding this information to the model development process, isotope data appear to provide an orthogonal view on catchment behavior. This result contributes to understanding the value of data for modeling, which may serve as a guidance in the process of gauging ungauged catchments.

Ancillary