While modeled results are often a driving factor behind decisions in pesticide risk assessment, monitoring data is an important set of supplemental data.  To get a full picture of monitoring data for a pesticide, it is beneficial to gather as much sample data as possible, often from a variety of publicly available data sources.  Gathering these data and performing quality control can be a time-consuming hurdle in preparation for risk assessment.  By using automation to gather and process data, the time required to produce a dataset can be greatly reduced.  Several accessible data sources (e.g. United States Geological Survey’s NWIS, Unites States Environmental Protection Agency’s STORET, California Environmental Data Exchange Network, and some state Agencies) often contain overlapping data records with varying formats and levels of quality.  Using machine learning clustering algorithms to identify common monitoring locations and samples across data sources provides a more efficient method of creating a base dataset than manual comparison without sacrificing quality. Further automation techniques can then be utilized to standardize units, merge metadata, and correct other data errors.  The resultant dataset is comparable to one produced through manual processing in a fraction of time spent.  To make the data easily accessible, a spatial dataset is automatically created and published to a user-friendly webtool.  The webtool allows users to examine data by compound, region, concentration, source type and other filters.  The web map enables users to easily identify high-risk areas for mitigation efforts and provides summary stats for the selected pesticide.

Using automation techniques and machine learning algorithms to create a multi-source pesticide monitoring database

Fall 2022 abstracts, posters, and presentations will launch in mid-September.

ACS Fall 2022

The official archive and repository for all ACS Meetings & Expositions abstracts and select posters or presentations.