SCALE-SELECTIVE VALIDATION OF MORPHODYNAMIC MODELS

Abstract

Although it is generally acknowledged that the practical predictability at smaller scales may be limited, output of high-resolution morphodynamic area models is mostly presented at the resolution of the computational grid. The so-presented fields typically are realistic looking, but not necessarily of similar quality at all spatial scales. Unfortunately, commonly used single-number validation measures do not provide the necessary guidance as to which scales in the output can be considered skilful. Also, differences in skill throughout the model domain cannot be discerned. Here, we present a new, scale-selective validation method for 2D morphological predictions that provides information on the variation of model skill with spatial scale and within the model domain. The employed skill score weighs how well the morphological structure and variability are simulated, while avoiding the double penalty effect by which point-wise accuracy metrics tend to reward the underestimation of variability. The method enables us to tailor model validation to the study objectives and scales of interest, establish the resolution at which results are ideally presented and target model development specifically at certain morphological scales.