Bayesian calibration of mechanistic aquatic biogeochemical models and benefits for environmental management

Aquatic biogeochemical models have been an indispensable tool for addressing pressing environmental issues, e.g., understanding oceanic response to climate change, elucidation of the interplay between plankton dynamics and atmospheric CO₂ levels, and examination of alternative management schemes for eutrophication control. Their ability to form the scientific basis for environmental management decisions can be undermined by the underlying structural and parametric uncertainty. In this study, we outline how we can attain realistic predictive links between management actions and ecosystem response through a probabilistic framework that accommodates rigorous uncertainty analysis of a variety of error sources, i.e., measurement error, parameter uncertainty, discrepancy between model and natural system. Because model uncertainty analysis essentially aims to quantify the joint probability distribution of model parameters and to make inference about this distribution, we believe that the iterative nature of Bayes' Theorem is a logical means to incorporate existing knowledge and update the joint distribution as new information becomes available. The statistical methodology begins with the characterization of parameter uncertainty in the form of probability distributions, then water quality data are used to update the distributions, and yield posterior parameter estimates along with predictive uncertainty bounds. Our illustration is based on a six state variable (nitrate, ammonium, dissolved organic nitrogen, phytoplankton, zooplankton, and bacteria) ecological model developed for gaining insight into the mechanisms that drive plankton dynamics in a coastal embayment; the Gulf of Gera, Island of Lesvos, Greece. The lack of analytical expressions for the posterior parameter distributions was overcome using Markov chain Monte Carlo simulations; a convenient way to obtain representative samples of parameter values. The Bayesian calibration resulted in realistic reproduction of the key temporal patterns of the system, offered insights into the degree of information the data contain about model inputs, and also allowed the quantification of the dependence structure among the parameter estimates. Finally, our study uses two synthetic datasets to examine the ability of the updated model to provide estimates of predictive uncertainty for water quality variables of environmental management interest.     

Resilience of Touristic Island Beaches Under Sea Level Rise: A Methodological Framework

This contribution addresses the need for a simple model for managers to employ when planning strategies for the management of touristic beaches under sea level rise. A methodological framework was developed and tested in two Aegean archipelago islands (Lesvos and Rhodes, Greece). The scheme can represent the status of touristic island beaches, based on easily obtained variables/indicators and projections of beach erosion/retreat under different scenarios of mean sea level rise (MSLR) and extreme events. Information on beach geomorphological characteristics, environmental setting, water quality, management, and services (such as those used in the “Blue Flag” classification) was collated/collected and beach erosion/retreat due to CV & C was estimated through suitable ensembles of cross-shore (1-D) morphodynamic models. A Strength-Weaknesses-Opportunities-Threats (SWOT) framework was employed to assist in the selection of indicators and multicriteria analysis used to optimize indicator weights and rank beaches according to their sustainability under sea level rise. Implementation of the framework at the two islands has shown that: the majority of Lesvos and Rhodes beaches (82% of a total of 217 beaches and 58% of a total of 97 beaches, respectively) can be classified as beaches with no, or minimal, human interference, suggesting that under environmentally sound coastal management further touristic development might be afforded; there could be very significant effects of the sea level rise on the carrying and buffering capacities of the most developed (“Blue Flag”) beaches, with some expected even under conservative projections to be completely eroded by 2100, unless technical adaptation measures are taken; and using the proposed framework, touristic beaches can be rapidly ranked in terms of their resilience to sea level rise and their development potential, allowing prioritization of effective management responses.