Grantee Research Project Results
Final Report: Topological Simplification and Machine Learning for Real-time Prediction of Off-target Pesticide
EPA Contract Number: 68HERC22C0030Title: Topological Simplification and Machine Learning for Real-time Prediction of Off-target Pesticide
Investigators: Ball, Kenneth
Small Business: Geometric Data Analytics, Inc.
EPA Contact: Richards, April
Phase: I
Project Period: December 1, 2021 through May 31, 2022
Project Amount: $99,993
RFA: Small Business Innovation Research (SBIR) Phase I (2022) RFA Text | Recipients Lists
Research Category: Small Business Innovation Research (SBIR) , SBIR - Toxic Chemicals
Description:
Off-target pesticide spray drift is an economically and environmentally significant problem. Off-target drift can damage susceptible crops, orchards, and apiaries, induce herbicide resistant weed development, damage native pollinator habitats, create human-health related pesticide exposure hazards, and create social and legal strife as a result of adjudicating exposure incidents.
The SprayCast project uses mathematical and statistical techniques to provide more accurate and easily accessible forecasts of drift related weather conditions and actual drift potential in real time to pesticide applicators. SprayCast is meant to facilitate communication among stakeholders, improve visibility into drift potential, and ultimately provide a more reliable tool to quantify and mitigate risk of post application off-target drift.
The SprayCast product is a web-based application that provides a drift prediction forecast in real-time via an easily accessible user interface. A prototype is remotely deployed on the Google Cloud Platform and can be accessed at https://spraycast-app-cr-v1-zyht7nnmqq-ue.a.run.app/.
Summary/Accomplishments (Outputs/Outcomes):
We have built a prototype web-based application called SprayCast that incorporates a user's location information, retrieves an up to date forecast of drift-related weather conditions, and resamples that forecast to the user's location and time. SprayCast then models drift potential given environmental conditions and user supplied application height and droplet size parameters and overlays contours of that drift potential on an interactive map. SprayCast also provides time series forecast information relative to longer term hazards associated with vapor drift from post-application pesticide volatilization, and provides a recommendation on spray drift risk based on interpolated 10 foot wind speeds.
We were able to build out and remotely deploy the SprayCast application, which delivers in real-time a forecast, rendered drift model, and spray recommendation in a matter of a few seconds to a remote user. All algorithmic and computational tasks in the SprayCast application pipeline are scalable and feasible for delivery in a commercial application. We found that the forecast data underlying SprayCast’s analyses was reliable and enables us to provide hyper-local drift forecasting at a competitive resolution. The forecasts that SprayCast processes were found to be more reliable on average than nearest neighbor official weather station data in a localized data validation experiment.
Conclusions:
Our results and assessment of our prototype support the potential for SprayCast to provide a remotely accessible, competitively accurate forecast of weather conditions with minimal computational overhead and an easy to use interface. SprayCast provides significant value to stakeholders at various levels of sophistication in the agricultural pesticide application market, and helps to reduce the impact of one of the largest and most expensive problems in modern agriculture. Direct feedback from stakeholders, experts, and potential customers has been both positive about the SprayCast product and affirming of the need for better predictive and planning solutions for post application pesticide drift.
Phase I development tasks related to commercialization have been primarily related to developing and validating a commercialization pathway through stakeholder engagement and feedback. Pesticide drift is an enormous and expensive problem for applicators, equipment manufacturers, regulators, and agrochemical manufacturers. SprayCast is an easy to use product with a user interface and API that can support commercial stakeholder needs at a variety of levels.
SprayCast clearly improves upon standard practice for pesticide drift risk assessment and can provide value by reducing actual incidents of contamination and driving commercial use towards pesticides whose standard formulations and label requirements are directly integrated into the SprayCast product. The SprayCast API can provide real-time modeling and simulation of drift potential, and we have demonstrated integration with a third-party maintainer of a pesticide susceptible site database. SprayCast offers a tangible value proposition to our diverse set of stakeholders.
In 2020 Monsanto (a subsidiary of Bayer) agreed to a $400 million settlement with farmers who experienced damage from off-target drift of dicamba. Because wind speed and direction have a prominent impact on off-target drift potential, reliable, hyper-local, and easily accessible drift forecasts like that delivered by SprayCast will increase the effectiveness and applicability of pesticide products while reducing risk of liability to manufacturers. Mitigating off-target drift can also increase the effective lifespan of novel pesticides; for instance drift exposure can result in weeds that are resistant or tolerant to the applied herbicide.
Individual farmers and commercial pesticide applicators have a direct economic interest in reducing the risk of off-target spray drift, as pesticide drift beyond buffer zones can damage and reduce yields of sensitive crops. Furthermore, reliable forecasting helps farmers and applicators to mitigate legal liability associated with off-target spray drift resulting from off-label application of pesticides either caused by or concurrent with wind and weather conditions that would prohibit application. Applicators can be held liable for damage from off-target drift, and a forecast tool like SprayCast — especially one that incorporates local wind speed and direction and local topography — will reduce legal and regulatory ambiguity for individual applicators.
SBIR Phase II:
SprayCast: Mitigating Pesticide Drift through High Resolution Forecasting, Modeling, and OptimizationThe perspectives, information and conclusions conveyed in research project abstracts, progress reports, final reports, journal abstracts and journal publications convey the viewpoints of the principal investigator and may not represent the views and policies of ORD and EPA. Conclusions drawn by the principal investigators have not been reviewed by the Agency.