Keywords
som, self-organizing map, model evaluation, optimization
Start Date
1-7-2008 12:00 AM
Abstract
Basically, any statement on hydrological model behaviour depends on our possibilities to differentiate between model time series. Applied within a model identification context, aggregating statistical performance measures are inadequate to capture details on time series characteristics as essentially different model results can be produced with close to identical performance measure values. It has been readily shown that the loss of information on the residuals imposes important limitations on model identification and -diagnostics and thus constitutes an element of the overall model uncertainty. In this contribution we present an approach using a Self-Organizing Map (SOM) to circumvent the identifiability problem induced by the low discriminatory power of aggregating performance measures. Instead, a Self-Organizing Map is used to differentiate the spectrum of model realizations, obtained from Monte-Carlo simulations with a distributed conceptual watershed model, based on the recognition of different patterns in time series. Further, the SOM is tentatively used as an alternative to a classical optimization algorithm to identify the model realizations among the Monte-Carlo simulations that most closely approximate the pattern of the measured discharge time series. The results are analyzed and compared with the manually calibrated model as well as with the results of the Shuffled Complex Evolution algorithm (SCE-UA).
Towards model evaluation using Self-Organizing Maps
Basically, any statement on hydrological model behaviour depends on our possibilities to differentiate between model time series. Applied within a model identification context, aggregating statistical performance measures are inadequate to capture details on time series characteristics as essentially different model results can be produced with close to identical performance measure values. It has been readily shown that the loss of information on the residuals imposes important limitations on model identification and -diagnostics and thus constitutes an element of the overall model uncertainty. In this contribution we present an approach using a Self-Organizing Map (SOM) to circumvent the identifiability problem induced by the low discriminatory power of aggregating performance measures. Instead, a Self-Organizing Map is used to differentiate the spectrum of model realizations, obtained from Monte-Carlo simulations with a distributed conceptual watershed model, based on the recognition of different patterns in time series. Further, the SOM is tentatively used as an alternative to a classical optimization algorithm to identify the model realizations among the Monte-Carlo simulations that most closely approximate the pattern of the measured discharge time series. The results are analyzed and compared with the manually calibrated model as well as with the results of the Shuffled Complex Evolution algorithm (SCE-UA).
