Identifiability and uncertainty analysis of bio-irrigation rates

Bio-irrigation is often quantified through incubations where an inert tracer is added to the overlying water of a core or a benthic chamber, and subsequently following the tracer distribution in either the overlying water or the porewater. The interpretation is based on fitting data with a model containing several unknown parameters such as the enhancement over molecular diffusion or non-local exchange. In this paper, we test under what conditions the results obtained through this fitting are robust. We first use identifiability analysis to investigate the minimum data requirements for two types of sediments, representative for deep-sea and shallow-water settings. We then use two different representative data-sets to estimate uncertainties of the fitted parameters, based on a Bayesian technique, the Markov Chain Monte Carlo. Using only the concentration change in the overlying water, it is not possible to constrain both the rate and the mechanism of bio-irrigation, thus, sampling the porewaters at the end of the incubation is a necessity