Mieke van Arkel

Model outcomes inherently involve a degree of uncertainty. This is due to the high degree of natural variability and our limited understanding of underlying processes. In data poor environments not much data on the coastal area under study is available. The limited data availability complicates model calibration and validation. Consequently the degree of uncertainty of model outcomes in data poor environments is higher compared to data rich environments.

Conducting a sensitivity analysis is a way to deal with modeling under uncertainty. Sensitivity analysis is the study of how the uncertainty in the output of a model can be apportioned to different sources of uncertainty in the model input. A sensitivity analysis can be very time-consuming due to the amount of computation time a numerical process based model can require and (ii) the huge number of model executions required for particular sensitivity methods.

Engineers use approaches to speed up the sensitivity analysis through: the use of simpler models to avoid excessive computational effort, (ii) reduce the size of the input data set, and (iii) the use of sensitivity methods requiring a small number of model executions. All three approaches speed-up the sensitivity analysis, but involve drawbacks in accuracy as well. Therefore the speed-accuracy tradeoff plays an important role in all the approaches. To conduct an efficient sensitivity analysis it is necessary to make an informed decision on this tradeoff. For this it is fundamental that hydraulic engineers know how the use of different models, input reduction methods and sensitivity methods affect the results of a sensitivity analysis. This is currently unclear.

The aim of this research is to develop a quick-start guide that can assist in the decision-making on the speed-accuracy tradeoff related to the three approaches. This quick-start guide is useful for engineers who aim to obtain quick insights in the coastal area under study, especially when this area is a data poor environment.