Bayesian calibration of the Pasture Simulation model (PaSim) to simulate emissions from long-term European grassland sites: a case study at Laqueuille (France)

Abstract

Mechanistic vegetation models with large parameter sets and high temporal resolution are currently used in grassland studies. They need a parsimonious screening method to identify the most influential parameters for the grassland system in specific contexts (weather, soil, management). This is basic to better understand and make use of the outputs from these models. In this study, Morris’ method was applied to test the sensitivity of a variety of outputs of the Pasture Simulation model (PaSim) to its parameters in six European multi-year grassland sites (one of them run under both extensive and intensive management regimes). Twenty-eight parameters related to plant physiology and animal digestion were screened and ranked for their sensitivity (under two distributional assumptions of parameter values), with the objective of determining their stability across sites and the minimum requirements for parameter calibration. The sensitivity analysis results proved that PaSim response is fairly stable across European sites, with only a few differences. Key results are that (1) seven influential parameters of vegetation development, aboveground growth and carbon/nitrogen partitioning were globally identified with both uniform and Gaussian distributions of parameter values, (2) two additional parameters (associated with leaf and stem fibre content) were also recognized as relevant for animal CH4 emissions target output, (3) listing of key parameters differed, but not widely, across sites and targeted outputs, and between distributions (ranking was more plastic), (4) first-order sensitivity rank (strength, mu) was generally similar to (or higher than) higher-order sensitivity (spread, a), indicating that parameters showing high interaction with other parameters or non-linearities are those with also a high direct effect on output. Overall, Morris’ method proved to be an effective and reliable tool to identify key vegetation parameters for the use of PaSim in the European conditions.

Publication
In Proceedings of Greenhouse Gases and Animal Agriculture, Dublin, June
Date