Mapping atrazine persistence in soils of central Argentina using INLA

Abstract

The length of time an herbicide remains in soils can be used as an indicator for diffuse pollution assessment. Atrazine, a residual herbicide with a wide range of reported environmental impacts, has been extensively applied in agricultural soils cultivated with gramineous crops. While atrazine persistence in soils has been widely studied by quantifying half-life from experimental dissipation curves, the spatial variability of half-life at the landscape scale has not been sufficiently explored. The objectives of this work were to estimate atrazine dissipation in a variety of soils and to map its spatial variability by modelling half-life as a function of environmental variables. Based on laboratory determination and environmental data, atrazine dissipation in soils was mapped across the territory of Córdoba province, central Argentina. First, we determined the dissipation parameters of atrazine under a variety of environmental conditions in 60 sites. Second, we fitted a Bayesian hierarchical model for the estimated atrazine half-lives, considering a soil site-specific property (soil organic carbon, SOC) and land use (gramineous crop, non-gramineous crop, pasture, and forest) as covariates. We used Integrated Nested Laplace Approximation and stochastic partial differential equations to model the spatial dependence on random effects. The results suggest that atrazine half-life can be spatially predicted by SOC and land use. Atrazine dissipation was faster in sites with previous gramineous crops and high soil organic matter. The digital map allowed us to quantify the environmental variability of atrazine dissipation in Córdoba (3.2–99.5 days). The modelling strategy might be applied to other herbicide compounds by selecting proper environmental covariates.