Applying a GLM based approach to model the influence of soil properties on the toxicity of phenmedipham to Folsomia candida

Purpose: Soil properties are the main explanation to the different toxicities obtained in different soils, due to their influence on chemical bioavailability and the test species performance itself. However, most prediction studies are centred on a few soil properties influencing bioavailability, while their direct effects on test species performance are usually neglected. In our study we develop prediction models for the toxicity values obtained in a set of soils taking into account both the chemical concentration and their soil properties. Materials and methods: The effects on the avoidance behaviour and on reproduction of the herbicide phenmedipham to the collembolan Folsomia candida is assessed in twelve natural soils and the OECD artificial soil. The toxicity outcomes in the different soils are compared and explanatory models are constructed by Generalized Linear Models (GLMs) using phenm edipham concentrations and soil properties. Results and discussion: At identical phenmedip ham concentrations, the effects on reproduction and the avoidance response observed in the OECD soil were similar to those observed in natural soils, while effects on survival were clearly lower in this soil. The organic matter and silt content explained differences in the avoidance behavior in different soils; for reproduction, there was a more complex. Conclusions: Our results highlight the need for approaches taking into account all the soil properties as a whole, as a necessary step to improve the prediction of the toxicity of particular chemicals to any particular soil pattern involving several soil properties