Data poor environments

Abstract

The objective of this study is to propose a method that is able to deal with data poor environments within coastal studies. Numerical models are useful tools to get insight in the coastal processes of the system being modelled. Inaccurate or poor input data leads to inaccurate or even incorrect model results. By stochastic modelling insight is obtained in the uncertainty propagation through the model and will, in contrast to deterministic modelling, provide output distributions. The coastline of interest suffers from structural erosion and consists of a rocky reef lying oblique in front of the coastline, which causes complex rip current patterns. Accurate information about the reef geometry and sufficient calibration and validation data is missing, which means we are dealing with a data poor environment. It became clear that the Monte Carlo method is the most appropriate method for this study. The Monte Carlo method is very robust, is relatively easy to implement and is based on random sampling. To generate a random sample the uncertainties of the used input and model parameters need to be quantified in probability distributions. The uncertainties are quantified by describing the total wave and wind climate in six separate probability distributions while taking correlation between the wave and wind variables into account. The uncertainty related to the reef geometry is taken into account by using the unknown reef elevation as stochastic variable. Roughness height is the only model parameter that is used stochastically. This is a commonly used calibration parameter and therefore partly obviates the missing calibration and validation data. As for the reef elevation appropriate and substantiated parameters are chosen. The obtained output distributions show that it is possible to provide an indication of the expected results. The reliability is, as expected, lower for the sediment fluxes than for the hydrodynamics. By examining scatter plots and correlation coefficients insight in the relationships and relative variable importance is obtained. The reef causes complexity and leads to different variable importance for (spatially) different output variables. Surprisingly, the results show that for the total sediment balance of the system the wind speed and wave period are the most important variables of the system. A regression analysis is used to quantify the relative importance of the used variables. It gives insight in the direct effect of the variable on the model output. Correlation between input variables makes it more complex but regression analysis is able to determine the direct and indirect effect of correlated variables on the model output. The proposed method gives rise to many applications and opportunities. It has shown an indication of the expected results can be given including an estimation of the reliability of the results. The corresponding probabilities can be used for risk analyses. The system is better understood and insight in complex coastal processes is obtained. Uncertainty analyses are able to provide the variable importance of the system being modelled, which for this study show unexpected results. Knowledge about the variable importance is valuable for coastal engineers and decision makers as it enables to focus on the most important model aspects and variables. By imposing stochastic model parameters the need for calibration can be determined. Furthermore a conceptual method has shown that it is possible to determine the (financial) value of obtaining more (accurate) data. This can lead to economic efficiency in coastal engineering practice, as it determines the necessity of data collection and calibration. Given the applicability of the proposed method it is recommended that uncertainty analyses like this become a more common approach within coastal studies, regardless of dealing with a data poor environment.