Our view: the 2016 and 2018 Atrazine Risk Assessments

In 2016 and 2018, farmer and agriculture groups responded to our calls to action by submitting comments to EPA on two key risk assessments: the  Draft Ecological Risk Assessment on Atrazine in 2016 and the Draft Human Health Risk Assessment in 2018. Both are assessments remain under review at EPA. Below is a summary of our concerns on each assessment.

2016 Draft Ecological Risk Assessment on Atrazine
Specific concerns about the draft ecological assessment and proposed aquatic community level of concern (LOC). EPA  recommended an aquatic life level of concern (LOC) of 3.4 parts per billion over 60 days. This LOC is currently 10 ppb and all credible science says it should be 25 ppb or higher. This ultra-low LOC would basically ban the use of atrazine in much of farm country. As detailed in our public comments, the preliminary ecological assessment contains many examples of mathematical and procedural errors and departures, including:

  • Aquatic LOC: Despite the advice and recommendations of multiple SAPs (2007, 2009, 2012) and other independent data reviews, EPA used a set of low quality studies that were not scientifically defensible to calculate a preliminary LOC for aquatic plants of 3.4 μg/L. Quality data show no significant effects at concentrations less than 30 μg/L when acceptable studies only are considered.
  • Water Monitoring: Quality control errors in the water database and methodological errors led to extraordinary overestimates of aquatic and terrestrial exposure. Errors were made by EPA in reporting, duplicate entries, infilling errors for periods when samples were not available, and use of non-detect samples with levels of detection much higher than their proposed LOC. EPA also used highly conservative modeling to estimate aquatic exposure concentrations. Modeled exposure concentrations were as much as 260-fold higher than the corresponding monitoring data, which was conducted around the country under the direction of EPA as a condition of registration resulting in a very large dataset that targeted the most vulnerable watersheds and high use areas.
  • Birds: The assessment erroneously lowered the chronic no-effect level for birds by a factor of three, without new information or data to support such a change, and used highly over conservative assumptions in areas such as dermal absorption. EPA’s modeling was so outrageously conservative that it incorrectly claimed that atrazine use would result in more than 35% of birds dying, which we know to be absolutely wrong.
  • Fish: The assessment relies on a study that was not conducted in accordance with required guidelines to lower the fish endpoint 12-fold, while overlooking the results of a more recent, guideline-compliant study, as well as other fish studies conducted by EPA itself.
  • Frogs: EPA reversed their previous conclusions on frogs and instead revived discredited, low quality frog studies.

2018 Draft Human Health Risk Assessment
Unlike the ecological risk assessment, this assessment by-and-large took into consideration the extensive review by the SAP dating back to 2009. However, EPA’s assessment of the large and broad-ranging drinking water database was very concerning

  • Concerns over drinking water data sets used (and not used) by EPA
    • We have a water monitoring database for atrazine. For drinking water derived from surface water sources (reservoirs and streams), monitoring data contains robust and comprehensive sampling from multiple targeted surface drinking water programs. This includes the Atrazine and Simazine Monitoring Programs conducted by the registrant (under the direction of EPA as a condition of registration), USDA Pesticide Data Program, and the USGS-EPA Pilot Monitoring Program, extensive monitoring by states under the Safe Drinking Water Act, and numerous other monitoring conducted by universities and other sources, accounting for nearly a million water samples.
    • We were surprised as to why EPA did not use this large collection of data sets as a basis of exposure, instead choosing to use overly predictive modeling approaches that highly overestimated the presence of triazines in drinking water. As we stated in our public comments, the modeled estimated concentrations significantly exceeded comparable monitoring data from surface drinking water supplies and are the result of multiple conservative model input assumptions that are not supported based on the available environmental and agronomic data. Given the robust surface drinking water monitoring dataset available, monitoring data should have been the basis for the derivation of surface drinking water concentrations rather than model predictions.