Mr Gavin Styles1, Ms Miguela Martin1, Professor Antonio Patti1, Dr Mick Rose2, Associate Professor Lukas Van Zwieten2
1Monash University, Melbourne, Australia, 2New South Wales Department of Primary Industries, Wollongbar, Australia
Sorption is considered one of the most influential processes on the mobility, bioavailability and rate of degradation of herbicides in soil. Site specific estimates of soil sorption coefficients are therefore a prerequisite for the accurate prediction of herbicide persistence in different soils. The adsorption affinity of a soil for a given herbicide is controlled by a wide variety of factors such as mineralogy, organic matter content, and organic matter speciation. However, direct measurement of sorption requires expensive and laborious construction of adsorption isotherms. The development of an effective predictive model could circumvent this testing, and advance our capabilities to predict herbicide persistence and ecological and agronomic risks.
Mid-infrared spectrometry (MIR) is a cheap, rapid technique that, given adequate calibration, can quantify a wide variety of factors including organic matter content and speciation, particle size distribution and mineralogy. This makes it especially promising for predictive modelling of small organic molecule sorption. There have been previous attempts to use MIR based modelling to predict herbicide adsorption affinity of soils, however, due to a limited dataset; these models have lacked the generality required for wider application.
The aim of this work is build a soils dataset with widely varying properties, and then to develop a compressive model capable of predicting herbicide sorption, based on MIR. By using a larger data set, comprised of 40+ sites from across Australia, a relatively universal model may be produced, capable of predicting adsorption affinity in soils outside of the initial data set. In this work, we assessed the adsorption of glyphosate, clopyralid and imazamox using MIR combined with principle component analysis, partial least squares and principle component regression techniques. Results will be presented in the context of herbicide sorption relationships to soil organic matter and mineralogy.
Gavin Styles is a PhD researcher at Monash University working with the New South Wales Depmartment of Primary Industry. His work focuses on herbicide dynamics and bio availability in the soil, with a particular focus on sorption, preictive modelling, and the development of new testing methods for herbicide presence and bioavailability in soil.