Stochastic sensitivity analysis of glyphosate biochemical degradation

A comprehensive numerical modelling of soil biogeochemical dynamics allowed us to explore uncertainties in the fate of the herbicide glyphosate (GLP) and its crucial byproduct aminomethylphosphonic acid (AMPA) in soil. GLP and AMPA are both toxic and have the potential to disrupt complex ecological and biogeochemical processes. This study aims at identifying the more influential sources of uncertainty in GLP biochemical degradation. Microorganisms may evolve different strategies for scavenging nutrients and energy from anthropogenic molecules depending on the surrounding environmental conditions. GLP can be catabolized by soil bacteria along two pathways. One biotic pathway produces AMPA, which can be degraded biologically to non-toxic end products, while a second biotic pathway produces non-toxic byproducts. Recent studies have shown that GLP and AMPA can also undergo fast chemical degradation to non-toxic byproducts in the presence of birnessite mineral, in which Mn3+ and Mn4+ ions act as catalysts. Therefore, a comprehensive GLP degradation reaction network was tested numerically by means of the BRTSim solver to assess GLP and AMPA degradation potential within a network that integrates several biochemical processes. Chemical and biological processes were described by Michaelis-Menten (MM) kinetics and the Monod growth model, respectively. The biochemical reactions describing the reaction network and the corresponding kinetic parameters were retrieved from the literature. In this numerical study, GLP was applied at typical rates in a soil control volume representing the top soil of an agricultural plot. GLP and AMPA concentrations were modelled over time as a function of both biological and chemical processes. A suite of sensitivity analyses on input Michaelis-Menten-Monod (MMM) parameters were run to assess the effect of biological parametric uncertainties and to quantify the influence of specific biological processes or specific group of MMM kinetic parameters to the overall model output. Parameter values were randomly chosen from a Gaussian distribution with mean equal to the corresponding experimentally-retrieved parameter value and standard deviation equal to 5, 10, 15, 20, 25, and 30% of that value. We found that, in the lack of birnessite mineral, variability in the reaction rate constant increased GLP equilibrium concentration, while variability in the half-saturation concentration constant and the biomass yield decreased it. The action of birnessite mineral shrank output variability and decreased GLP concentrations by 5 times. Overall, the more GLP was biodegraded the more AMPA was produced, which accumulated due to its slow biodegradation.