Modeling the risk of water pollution by pesticides from imbalanced data.

The pollution of ground and surface waters with pesticides is a serious ecological issue that requires adequate treatment. Most of the existing water pollution models are mechanistic mathematical models. While they have made a significant contribution to understanding the transfer processes, they face the problem of validation because of their complexity, the user subjectivity in their parameterization, and the lack of empirical data for validation. In addition, the data describing water pollution with pesticides are, in most cases, very imbalanced. This is due to strict regulations for pesticide applications, which lead to only a few pollution events. In this study, we propose the use of data mining to build models for assessing the risk of water pollution by pesticides in field-drained outflow water. Unlike the mechanistic models, the models generated by data mining are based on easily obtainable empirical data, while the parameterization of the models is not influenced by the subjectivity of ecological modelers. We used empirical data from field trials at the La Jaillière experimental site in France and applied the random forests algorithm to build predictive models that predict "risky" and "not-risky" pesticide application events. To address the problems of the imbalanced classes in the data, cost-sensitive learning and different measures of predictive performance were used. Despite the high imbalance between risky and not-risky application events, we managed to build predictive models that make reliable predictions. The proposed modeling approach can be easily applied to other ecological modeling problems where we encounter empirical data with highly imbalanced classes.

Predictive quality of 26 pesticide risk indicators and one flow model: A multisite assessment for water contamination.

Stakeholders need operational tools to assess crop protection strategies in regard to environmental impact. The need to assess and report on the impacts of pesticide use on the environment has led to the development of numerous indicators. However, only a few studies have addressed the predictive quality of these indicators. This is mainly due to the limited number of datasets adapted to the comparison of indicator outputs with pesticide measurement. To our knowledge, evaluation of the predictive quality of pesticide indicators in comparison to the quality of water as presented in this article is unprecedented in terms of the number of tested indicators (26 indicators and the MACRO model) and in terms of the size of datasets used (data collected for 4 transfer pathways, 20 active ingredients (a.i.) for a total of 1040 comparison points). Results obtained on a.i. measurements were compared to the indicator outputs, measured by: (i) correlation tests to identify linear relationship, (ii) probability tests comparing measurements with indicator outputs, both classified in 5 classes, and assessing the probability i.e. the percentage of correct estimation and overestimation (iii) by ROC tests estimating the predictive ability against a given threshold. Results showed that the correlation between indicator outputs and the observed transfers are low (r<0.58). Overall, more complex indicators taking into account the soil, the climatic and the environmental aspects yielded comparatively better results. The numerical simulation model MACRO showed much better results than those for indicators. These results will be used to help stakeholders to appropriately select their indicators, and will provide them with advice for possible use and limits in the interpretation of indicator outputs.