Global sensitivity analysis of the harmonized Lemna model

Abstract

Mechanistic effect modelling is becoming increasingly important for environmental risk assessment in the framework of pesticide authorization. For instance, the European Food Safety Authority (EFSA) has judged the model for the aquatic macrophyte test organism Lemna as “ready for use.” Nevertheless, national regulatory authorities are still hesitant to accept mechanistic effect modelling studies.

In order to increase the confidence in the Lemna model, in this study we performed a two-step global sensitivity analysis (GSA) of the harmonized model. GSA notably allows identifying and ranking the importance of i) toxicokinetic (TK) and toxicodynamic (TD) parameters, ii) physiological and ecological parameters, iii) environmental driving variables, and iv) initial conditions.

In a first step a Morris sensitivity screening was conducted to filter out non-influential input factors. In a second step, a true variance-based GSA was carried out with the Sobol method. The GSA was conducted for four different concentration levels and three different exposure regimes: constant, pulsed and realistic. Moreover, two different sets of input distributions of TKTD parameters were examined. The target variables were the effects of the pesticide on Lemna biomass and average growth rate.

Both Morris and Sobol GSA showed that for a specific substance three physiological parameters (optimum and minimum growth temperature, maximum photosynthesis rate) and the initial biomass BM0 were more important than the five TKTD parameters. Hence, for predictive applications of the model outside a laboratory context, BM0 must be chosen carefully, and uncertainty in the main physiological parameters must be reduced to a minimum.