Quantification of Predictive Uncertainty for Reversible Degradation of Diclofenac under Biotic, Denitrifying Redox Conditions

Drinking water resources and the associated delicate aquatic ecosystem are threatened by several contaminants. Diclofenac poses major concerns due to its persistent nature and frequent detection in groundwater. Despite some evidences of its biodegradability under reducing conditions, Diclofenac attenuation is often interpreted through geochemical models which are too simplified, thus potentially biasing the extent of its degradation. In this context, we suggest a modeling framework based on the conceptualization of the molecular mechanisms of Diclofenac biodegradation which we then embed in a stochastic context. The latter enables one to quantify predictive uncertainty. Biotic and denitrifying reference conditions are taken from a set of available batch experiments that evidence the occurrence of a reversible degradation pathway. Our model is subject to stochastic calibration through Acceptance-Rejection Sampling. The associated results fully embed uncertainty quantification and support the recalcitrance of Diclofenac in groundwater. Our results show that data scarcity and/or redundant model parametrization seem to deteriorate the quality of some parameter estimates, a feature that appears to be associated with the degree of information contained in the available dataset, which is addressed towards specific model processes. We then address the issue by reducing the complexity of the model and embed the resulting formulations within a multi-model context. The resulting models are calibrated through a Maximum Likelihood approach assisted by modern sensitivity analyses techniques, the performance of each candidate model being then assessed (in a relative sense) through classic model identification criteria. Our results suggest that an optimal trade-off in terms of model complexity (i.e., level of parametrization) given data availability can be assessed to satisfactorily interpret the system dynamics.