Pesticides are now occurring worldwide in almost all water resources, thus threatening the health of humans and other life. As a consequence, there is a strong social demand for designing safe cropping systems with less or no hazardous pesticides. Safe cropping systems can be designed now using pesticide transfer models. These models are mathematical tools that allow to predict the flow and concentration of pesticides in a field or a watershed. Here, we review the effects of agricultural practices on runoff, leaching, erosion, and drift from eight watershed models and nine field models. Our main findings are the following: (1) though models claim they account for practices, their effects cannot be represented. We present a method and four practice levels to assess the effects of practices in models, using tillage as an example. (2) The conceptual structure of the model highly influences the predicted distribution and transfer of pesticides. For instance, the pesticide levels remaining on the soil surface after plowing ranges from 0 % of the dose applied for the MIKE SHE-DAISY model to 100 % for GLEAMS, annAGNPS, SoilFug, and PestLCI. Only the Root Zone Water Quality Model (RZWQM) simulates pesticide interception by mulch during pesticide application. (3) Models should better take into account mulching, e.g., plastic, crop residues and associated crops, and other innovative practices. (4) A change in scale is needed for drift in watershed models. Here, topological watershed representations are the most promising way for upscaling the effects of practices. (5) Nonconservative calculations of pesticide interception by watershed mitigation structures (SWAT) should be carefully checked because these calculations underestimate the risk of pollution at the outlet. How models simulate practices will no longer be a secret for model users who apply our methodology and recommendations when selecting a model. We provide recommendations for improving tools to assess practices.

No abstract

The understanding of factors affecting pesticide transfers to catchment outlet is still at a very early stage in tropical context, and especially on tropical volcanic context. We performed on-farm pesticide use surveys during 87 weeks and monitored pesticides in water weekly during 67 weeks at the outlet of a small catchment in Martinique. We identified three types of pollution. First, we showed long-term chronic pollution by chlordecone, diuron and metolachlor resulting from horticultural practices applied 5-20 years ago (quantification frequency higher than 80%). Second, we showed peak pollution. High amounts of propiconazole and fosthiazate applied at low frequencies caused river pollution peaks for weeks following a single application. Low amounts of diquat and diazinon applied at low frequencies also caused pollution peaks. The high amounts of glyphosate applied at high frequency resulted into pollution peaks by glyphosate and aminomethylphosphonic acid (AMPA) in 6 and 20% of the weeks. Any intensification of their uses will result in higher pollution levels. Third, relatively low amounts of glufosinate-ammonium, difenoconazol, spinosad and metaldehyde were applied at high frequencies. Unexpectedly, such pesticides remained barely detected (<1.5%) or undetected in water samples. We showed that AMPA, fosthiazate and propiconazole have serious leaching potential. They might result in future chronic pollution of shallow aquifers alimenting surface water. We prove that to avoid the past errors and decrease the risk of long-term pollution of water resources, it is urgent to reduce or stop the use of pesticides with leaching potential by changing agricultural practices.