The effects of ionic strength and organic matter on virus inactivation at low temperatures: general likelihood uncertainty estimation (GLUE) as an alternative to least-squares parameter optimization for the fitting of virus inactivation models

Abstract

This study examined how the inactivation of bacteriophage MS2 in water was affected by ionic strength (IS) and dissolved organic carbon (DOC) using static batch inactivation experiments at 4°C conducted over a period of 2 months. Experimental conditions were characteristic of an operational managed aquifer recharge (MAR) scheme in Uppsala, Sweden. Experimental data were fit with constant and time-dependent inactivation models using two methods: (1) traditional linear and nonlinear least-squares techniques; and (2) a Monte-Carlo based parameter estimation technique called generalized likelihood uncertainty estimation (GLUE). The least-squares and GLUE methodologies gave very similar estimates of the model parameters and their uncertainty. This demonstrates that GLUE can be used as a viable alternative to traditional least-squares parameter estimation techniques for fitting of virus inactivation models. Results showed a slight increase in constant inactivation rates following an increase in the DOC concentrations, suggesting that the presence of organic carbon enhanced the inactivation of MS2. The experiment with a high IS and a low DOC was the only experiment which showed that MS2 inactivation may have been timedependent. However, results from the GLUE methodology indicated that models of constant inactivation were able to describe all of the experiments. This suggested that inactivation time-series longer than 2 months were needed in order to provide concrete conclusions regarding the time-dependency of MS2 inactivation at 4°C under these experimental conditions.

Human intervention in rivers

“…The model input distributions obtained through inverse modelling can be used as input in forward analysis in the other quadrants, as is often done (e.g. Fonseca et al, 2014;Mayotte et al, 2017;Inam et al, 2017). However, this requires some additional assumptions regarding extrapolation and stationarity (figure 7.1).…”

“…Formally, this is achieved through Bayesian inference, which produces probability distributions of parameter values given the (subjective) likelihood that the model, given these values, confirms experimental observation. A well known Bayesian inference methodology is GLUE (generalised likelihood uncertainty estimation), originally developed for non-uniqueness problems in hydrology (Beven and Binley, 1992;Fonseca et al, 2014;Mayotte et al, 2017). However, previous work has shown that estimating more than one friction source based on water levels can result in unidentifiable parameter distributions, which are characterised by wide and unconstrained shapes (Werner et al, 2005b).…”