Sensitivity of catchment model parameters to raingage network characteristics
Résumé
This paper investigates the sensitivity of the parameters of three catchment models to the size and spatial characteristics of raingage networks. Using randomly generated subsamples of a large existing raingage network on the Yonne river (France), we define four classes of parameter behavior: a) parameters able to make an efficient use of increasing information (converging towards a catchment-specific value), b) parameters overly sensitive to the calibration period (converging towards a period-specific value), c) stable parameters (easily identifiable ones) and d) poorly identifiable parameters. The approach presented here will be useful to water resources managers, to assess the size and characteristics of raingage networks needed for rainfall-runoff analysis, as well as to modellers seeking a better understanding of model parameter behavior.