Entropy-based grouping of river flow regimes

Abstract

In the context of changing environment, and specifically a possible climate change, a river flow regime type turns from being a merely general, average, descriptive characteristic to a tool for monitoring changes in flow seasonality both in time and space. Utilization of river flow regimes as a diagnostic tool for the output of climate models and in flow sensitivity studies demands an objective grouping of flow series into regime types. A hierarchical aggregation of monthly flow series into flow regime types, satisfying chosen discriminating criteria, is effectively performed by means of minimization of an entropy-based objective function. This function is based on the concept of an `information loss' resulting from such an aggregation and describes the difference between the series aggregated into one group, i.e. inaccuracy of aggregation. The main advantage of the approach, operating on river flow series for individual years (and not only on long-term means), is its ability to consider also the temporal regularity of the seasonal flow patterns, neglected, as a rule, in other approaches. Meanwhile, both flow volumes and seasonal patterns of flow are sensitive to climate fluctuations. A strict formulation of the `stopping rules' in the hierarchical flow regime grouping, directly related to the aggregation principles, is suggested. The approach allows different formulations of criteria for discriminating flow regime types. It is illustrated on a regional river flow sample for Scandinavia for two different formulations of discriminating criteria.