Predicting pesticide leaching under climate change: Importance of model structure and parameter uncertainty

The assessment of climate change impacts on pesticide leaching requires a careful consideration of both parameter uncertainty and model structural error. Our aim was to assess (i) model structure uncertainty, with specific focus on the effects of temperature on pesticide sorption and diffusion, and its relation to parameter uncertainty and (ii) the importance of these sources of uncertainty in long-term predictions of pesticide leaching in the perspective of climate change. We introduced optional functions describing the effects of temperature on sorption and diffusion into the pesticide fate and transport model MACRO5.2, which resulted in four structurally different versions. The generalised likelihood uncertainty estimation (GLUE) method was applied to these four structurally different model versions in order to generate an ensemble of acceptable parameter combinations for long-term predictions of pesticide leaching under climate change. The analysis was performed in a two-step procedure. First, for each model version, 80,000 parameter combinations were tested against a comprehensive field data set of bromide and bentazone concentrations in drainflow and in the soil profile to identify acceptable parameter combinations. In a second step, we performed 30-year predictions based on these acceptable parameter sets for three hypothetical compounds that differed in their sorption strength. The period 1970–1999 was used to represent the current climate for a region in south-west Sweden. The future climate time series was derived for the period 2070–2099 by perturbing the observed time series based on monthly change factors calculated from climate projections of the regional climate model RCA3 with boundary conditions of the global circulation model ECHAM5 forced by the A2-emission scenario. For all three pesticide compounds, total losses were predicted to increase for autumn applications due to projected increases in winter rainfall, but decrease for spring applications. Differences between model versions were found, despite the large uncertainty in parameter estimates. Thus, temperature dependent diffusion was shown to be important for weakly sorbed compounds, while inclusion of temperature dependent sorption significantly affected the predictions for moderately and strongly sorbed compounds. The response to climate change was consistent for the different model versions, but the models that included temperature dependent sorption showed relatively more leaching in a future climate. Our findings stressed the importance of considering temperature dependent processes when modelling future pesticide losses.

• Climate change studies rarely investigate effects of impact model uncertainty.
• Acceptable parameter sets for MACRO were identified within a GLUE-framework.
• Model predictions of pesticide losses were made for present and future climates.
• A consistent response to climate change was found despite parameter uncertainty.
• Temperature effects on sorption and diffusion should be considered in impact studies.