« Back to Resource Library

Stone is excited to again be sponsoring and participating in the American Chemical Society's (ACS) Fall 2026 Meeting, from Sunday, August 23, to Thursday, August 27, 2026, in Chicago.

We have long supported the conference, and our employee-owners are eager to share another year's worth of experience in multiple presentations covering a range of topics related to pesticide application, pollinator risk assessment and data analysis, and the use of machine learning in safety assessments. We'll also be a sponsor for the conference's AGRO program.

See below for a list of presentations with detailed descriptions. Feel free to connect with our Ben Brayden; Jonnie Dunne; Tim Dupuis; John Hanzas; Jacob Mitchell; Dwayne Moore; Hendrik Rathjens; Jody Stryker; or Michael Winchell.

Presentations and sessions

Monday, August 24, 2026

Session: Pollinator Risk Assessment Methods and Data Analysis Approaches in a Tiered Risk Assessment System (Room: S502b McCormick Place Convention Center)

Stone Senior Ecotoxicologist Dwayne Moore is co-chair of this session, which begins at 9 a.m. EDT.

10:20 - 10:45 a.m., EDT

Presentation: Using Pollinator Mobility Data in Precision Application Risk Assessments: A Case Study with Lepidopterans (4541328)

Presenter: Dwayne Moore, Stone Environmental. Co-Authors: R. Scott Teed, Stone Environmental; Amund Løvik; Charlotte Elston.

Abstract: Most screening-level pesticide risk assessments assume that pollinators occur only in recently treated areas and thus receive maximum possible exposures. However, many pollinators are mobile and thus move between treated areas, areas receiving runoff and drift, and unexposed areas. Further, with the advent of precision agriculture, only portions of crop fields are treated rather than entire fields. Thus, a better understanding of how individuals move and forage in agroecosystems is required to develop refined exposure models for pollinator species. Technological advancements in recent years have led to an exponential increase in the availability of pollinator movement data at various spatial scales. Such data can be used to develop spatially-explicit exposure models by combining pesticide application and off-target transport maps with pollinator movement trajectories to estimate exposure at each location visited. We are compiling a movements database for pollinators and other non-target terrestrial invertebrates. The database includes relevant characteristics of test organisms (e.g., species; life stage; size; type of movement such as foraging, dispersal, diurnal movements, and mating), foraging range, preferred habitats, environment (e.g., habitat type; crop stage; date; spatial and temporal scales for movement recordings), experimental setup, methods, and other metadata. These data will be used to establish which movement models best describe observed movement patterns (e.g., random walk; correlated random walk; Lévy flights) and the key factors affecting movements (e.g., crop; season; environmental conditions). In this presentation, we will illustrate how movement data can be used in refined assessments for lepidopteran species.

Session: Precision Application of Agricultural Pesticides in the Digital Age (Room: S503a McCormick Place Convention Center)

4:20 - 4:45 p.m. EDT

Presentation: Tiered Framework for Environmental Exposure Assessment of Precision Pesticide Applications (4541048)

Presenter: Hendrik Rathjens, Stone Environmental. Co-Authors: Olivier de Cirugeda Helle; Neil Mackay; Sebastian Multsch; Dirk Nickisch; David Patterson; Jörn Strassemeyer; Robin Sur; Bernhard Gottesburen.

Precision application (PA) technologies fundamentally alter the spatial and temporal distribution of plant protection products across agricultural fields, creating both opportunities and challenges for environmental risk assessment (ERA). Unlike conventional broadcast applications, PA introduces spatial heterogeneity, e.g., through band, patch, spot, or variable-rate strategies, that existing regulatory exposure frameworks were not designed to address. While adapting the European FOCUS ERA framework for PA can be approached within a conceptually familiar tiered structure, scenario development at each tier requires new concepts regarding two aspects, (1) the spatio-temporal resolution at which exposure information is required for effect assessments, and (2) the translation of real-world precision spray patterns into technology-specific regulatory scenarios. To address the first, the framework introduces the Representative Elementary Area (REA) concept to define the minimum spatial scale at which exposure is relevant for effect assessment, accounting for organism mobility. To address the second, four components (field coverage ratio, application zone characterization, spatial distribution analysis, and edge analysis) are proposed as a statistical basis for characterizing spray patterns and mapping them to idealized regulatory scenarios. These two concepts are integrated into the tiered framework, which proceeds from Tier 1, where standard FOCUS models remain applicable with simple conversion or reduction factors, through Tier 2, where harmonized GIS datasets and technology-specific refinements (e.g., drift reduction data from ISO 22866-compliant trials) enable regionally representative scenarios, to Tier 3, where field-specific data (e.g., drone imagery, spray maps, LiDAR-derived vegetation structure, and local meteorological conditions) support field-level, situation-specific exposure estimates. This contribution presents the tiered assessment framework alongside examples of how each tier could be realized, including an initial application of the statistical framework to classify real-world spray patterns and link them to idealized scenarios. Rather than presenting finalized scenarios, this work illustrates the building blocks of a coherent, scalable assessment pathway that can accommodate the diversity of PA technologies while maintaining regulatory consistency and progressively capturing the exposure-reducing potential of precision application.

Session: Applications of Machine Learning and Large Language Models in Pesticide Safety Assessments (Room: S502b McCormick Place Convention Center)

Stone Senior Environmental Modeler Hendrik Rathjens is co-chair of this session, which begins at 3 p.m. EDT.

5:40 - 6:05 p.m., EDT

Presentation: Random Forest Modeling of Drift Interception by Plant Species (4533374)

Presenter: Jonnie Dunne, Stone Environmental. Co-Authors: Jens Kiesel, Stone Environmental; Michael Winchell, Stone Environmental; Milos Zaric; Jeff Golus; Richard Brain.

Abstract: Risk assessments for pesticide spray drift exposure in the United States do not account for variability in drift capture efficiency among off-field non-target plant species, which vary markedly in life history, structure, surface texture, and physiology. However, conducting drift exposure experiments to measure capture efficiency for all potential combinations of operational parameters and plants is not feasible. While the effects of many different aspects of pesticide applications and plant canopies on drift deposition have been studied individually, they have not been comprehensively incorporated into regulatory exposure models. The objective of this research is to model drift capture efficiency across a diverse range of plant species, nozzle types, and distance from application. Random forest regression modeling is a machine learning approach that is well-suited to assess the relative influence of these factors, interpolate differences in drift capture efficiency, and guide the collection of additional data for model refinement. Here we develop a random forest model that is predictive of these variables (r2 = 0.83 validation performance), trained on capture efficiency data from wind tunnel trials with three different spray nozzles and 12 herbaceous plant species native to the midwestern US, arranged in three staggered rows. The drift capture efficiency predictions are based on descriptors of canopy shape derived from photogrammetric scans, leaf flexibility measured via mechanical testing, droplet size distribution statistics from laser imaging, and distance from spray nozzles. Regulatory model refinement integrating application and non-target plant heterogeneity related to drift capture efficiency values would provide a clearer understanding of potential pesticide exposure risk under a variety of scenarios.

6:05 - 6:30 p.m., EDT

Presentation: Agricultural BMPs, Water Quality, and Pesticide Risk Assessment: A Machine Learning Framework for Watershed-Scale Mitigation Effectiveness Evaluation (4532577)

Presenters: Hendrik Rathjens, Stone Environmental. Co-Authors: Jens Kiesel, Stone Environmental. Richard Brain; Wenlin Chen.

Abstract: Agricultural best management practices (BMPs) represent a primary management approach for reducing nutrient, sediment, and pesticide losses from farm fields to surface water. Existing evidence of BMP effectiveness is largely derived from field-scale studies or hypothetical scenarios using watershed-scale modeling approaches. Direct quantification of BMP effectiveness based on information sources including surface water monitoring data, catchment characteristics, and level of BMP adoption within watersheds therefore represents an important opportunity to better inform land management decisions and regulatory frameworks such as the EPA’s Mitigation Menu. This ongoing study is structured in three phases. Phase 1 establishes the data foundation and machine learning architecture using a harmonized dataset of over 257 million water quality records from the USGS National Water Information System and the EPA’s Water Quality Portal, spanning watersheds across the United States varying in agricultural intensity. Phase 2 builds a machine learning framework of tree- and neural network-based algorithms using a wide range of catchment characteristics as predictors to simulate metrics derived from the water quality database. We assess BMP effectiveness by evaluating whether incorporating BMP implementation data into the machine learning framework improves model predictive performance for key water quality indicators such as total suspended solids, total nitrogen, nitrate, and total phosphorus. Phase 3 extends the analysis to pesticides, exploring whether the developed machine learning framework can be applied to predict catchment scale pesticide vulnerability, thus potentially complementing pesticide monitoring efforts. This presentation will focus on the machine learning architecture and Phase 1 results, using all available water quality data to capture a gradient in runoff and erosion vulnerability. Using random forest and gradient boosting algorithms within a site-year framework, we identify key environmental and landscape predictors of water quality and establish the analytical foundation for subsequent BMP effectiveness assessment.

Session: Environmental Fate, Transport, and Modeling of Agriculturally Related Chemicals (Room: A2 – South Hall ChemPod 1 McCormick Place Convention Center)

12:05  to 2:30 p.m., EDT

Presentation: Dicamba and 2,4-D Levels in Midwestern United States Rainwater in 2022 and 2023 (4532262)

Presenter: Sara Whiting. Co-Authors: Lance Shuler; Naresh Pai; Allen Olmstead; Brent Toth; Jacob Mitchell, Stone Environmental; Karla Gage; Scott Nolte; Reid Smeda.

Abstract: Previous studies have detected agricultural herbicides in rainfall. This study measured ambient concentrations of dicamba and 2,4-D in rainwater collected at 23 locations across six Midwestern states in 2022 and eight in 2023. Samples were collected for 12 to 17 weeks from March or April through August coinciding with typical dicamba and 2,4-D application periods for early post-emergent herbicide use in row crops. Rainwater sampling locations, except for the reference locations, were placed in high agricultural use areas. Based on the USDA Cropland Data Layer assessments, 43 to 92% of land surrounding non-reference sites consisted of corn, cotton, soybean, pasture, sorghum, or other crops with potential dicamba and/or 2,4-D use. Sub-daily or weekly rainwater sampling was conducted at a height of 2 m using an automated atmospheric wet deposition sampler. The 2022 data showed that 90^th percentile and maximum values for dicamba were 1.8 and 2.9 µg/L, respectively and for 2,4-D were 4.2 and 8.2 µg/L, respectively. The data in 2023 indicated that the 90^th percentile and maximum values for dicamba were 0.20 and 2.0 µg/L, respectively and for 2,4-D were 0.52 and 3.6 µg/L, respectively. No discernable temporal trends were observed during the sampling periods. Rainfall residues were converted to an equivalent pesticide application rate in lb/A for comparison with a regulatory no adverse effect rate (NOAER). The USEPA NOAER’s for the most sensitive species were 0.000261 lb/A for soybean (dicamba) and 0.0075 lb/A for tomato (2,4-D). Among 1028 samples collected over two years, one exceeded the dicamba NOAER at an equivalent rate of 0.00034 lb/A (representing <0.1% of samples collected), while none exceeded the 2,4-D NOAER. Because the analyses quantified only herbicide acids, detection cannot be attributed to any specific products, tank mixes, or application methods. 

5:40 - 6:05 p.m., EDT

Presentation: Method Development and Application of a Geospatial Refinement Approach to Pesticide Aquatic Exposure Modeling and Risk Assessment for Endangered Species (4544136)

Presenter: Michael Winchell, Stone Environmental. Co-Authors: Hendrik Rathjens, Stone Environmental; Scott Teed, Stone Environmental; Christopher Holmes; Ralph Warren; Huajin Chen; Patrick Havens; Tilghman Hall; Paul Whatling; Rebecca Haynie.

Abstract: Strategies for efficiently conducting endangered species pesticide risk assessments and determining necessary mitigations for reducing pesticide exposure have been developed by the US EPA, including distinct strategies for herbicides and insecticides for assessing agricultural pesticide uses. These strategies have purposefully relied upon conservative aquatic exposure modeling methods and assumptions, for efficiency and to help ensure species protection. Screening-level exposure estimates are then compared against population-level effects thresholds to quantify the magnitude of difference (MoD) between exposure and effects, subsequently determining the level of mitigation needed to reduce exposure and provide further protection of endangered species. Given the conservatism in aquatic exposure estimates, these MoDs can often be large (100 or greater), resulting in implied mitigation levels that may not be necessary for the protection of many species. An alternative approach for aquatic species is to incorporate exposure modeling methods that account for species-specific cropping patterns and regional information concerning pesticide use practices to provide more precise information from which to estimate pesticide exposure. These species-specific exposure estimates then lead to more realistic risk characterization and mitigation strategies. We have developed a multi-step aquatic exposure modeling methodology designed for national endangered species assessments that incorporates three geospatial refinement elements to derive species-specific exposure: crop-weighted catchment exposure scenarios, Percent Cropped Area, and Percent Crop Treated. The methodology, designed to be both flexible and efficient, has been tested for two pesticides, dimethoate (an insecticide) and pendimethalin (an herbicide), and compared with the exposure, risk assessment, and mitigation requirement results derived from the US EPA’s final Insecticide Strategy and Herbicide Strategy. The geospatial modeling refinements produced a range of species-specific exposure estimates based on local landscape variability resulting in EECs substantially lower than the generalized screening-level exposure estimates, leading to more representative risk characterization and a more effective and targeted mitigation approach. We provide a framework for how this refined aquatic exposure modeling approach could be integrated with the current regulatory endangered species risk assessment strategies.

6:30 - 6:55 p.m., EDT

Presentation: Geospatial Refinement of Terrestrial Pesticide Plant Exposure for Endangered Species Risk Assessments (4542596)

Presenter: Michael Winchell, Stone Environmental. Co-Authors: Hendrik Rathjens, Stone Environmental; Scott Teed, Stone Environmental; Ralph Warren; Eric Henry; Christopher Hassinger.

Abstract: The US EPA’s Herbicide Strategy (HS) considers terrestrial non-target plant exposure to herbicides due to runoff from adjacent agricultural fields. The conceptual model for the exposure assessment assumes that a field treated with an herbicide is immediately adjacent to a non-target plant community that occupies a 30-meter terrestrial plant exposure zone (TPEZ). Terrestrial plant communities beyond this 30-meter zone are not impacted by direct runoff from a treated field, though could still be impacted by spray drift. One limitation of the current HS approach is that it does not explicitly account for pesticide runoff scenarios that are directly relevant to the geographic location of the species range. It also does not account for other important geospatial information, including the portion of the species range that falls within the TPEZ adjacent to potentially treated fields, the hydrologic connectivity between the potentially treated fields and the TPEZ, and the fraction of potentially treated agricultural fields that are likely to be treated with the pesticide. We have developed a flexible, multi-step approach to refine TPEZ pesticide exposure that accounts for runoff exposure conditions relevant to each endangered plant species range, using either regional (HUC2-scale) high vulnerability runoff exposure scenarios, or spatially explicit runoff scenarios that are directly related to soil, slope, and weather conditions occurring with the potentially treated crops in the species range. Using either of these exposure runoff scenario approaches, subsequent refinement steps then account for (1) the fraction of the unexposed species range (outside the 30-m TPEZ), (2) the fraction of the TPEZ that is unexposed due to upslope topographic position (hydrologically disconnected), and (3) the fraction of potentially treated fields receiving pesticide applications (Percent Crop Treated). We applied this step-wise refinement approach to 169 terrestrial plant species identified by EPA in the HS and compared the exposure estimates and resulting mitigation requirements to those derived following the current HS approach. The analysis showed that each step in the geospatial refinement approach provided additional species-specific potential exposure information that helps in better understanding the population-level risks to endangered plants and in determining the appropriate level of mitigation at the species range scale.

Thursday, August 27, 2026

Session: Unmanned Aerial Systems (aka Drones): Pesticide Spraying and Other Agricultural Applications (Room: S503a McCormick Place Convention Center)

9:05 - 9:30 a.m., EDT

Presentation: Investigation of Batch-Specific Recovery Variability in Commonly Used Sample Media for UAV Spray Drift Studies (4535841)

Presenter: Tim Dupuis, Stone Environmental. Co-Authors: Brent Toth; Frank Donaldson; Rajeev Sinha; Andrew Hewitt.

Abstract: Accurate quantification of spray drift using fluorescent tracer dyes relies on the consistent recovery of analytes from sampling surfaces. During a recent unmanned aerial vehicle (UAV) spray drift field study and separate storage stability analysis, unexpectedly low recovery values were observed from a specific batch of biaxially-oriented polyethylene terephthalate (commonly called BoPET or Mylar) sample cards. This phenomenon was anomalous, as previous field trials utilizing different batches of polyester film sample media product yielded acceptable, consistent recovery rates. To identify a root cause, a follow-up laboratory study was conducted to evaluate the influence of material batch variability on dye recovery.

This study compared the recovery efficiencies of spiked fluorescent dye onto four distinct batches of polyester film sample media surfaces, including the batch of BoPET associated with low recoveries. To assess potential concentration or volume-dependent effects, samples were treated with multiple spike volumes and dye concentrations, spanning the range of dye masses observed during previous spray drift field study samples. All samples were processed using standardized fluorometric analysis protocols under GLP (Good Laboratory Practice) standards.

Findings confirm significant recovery deficits unique to the suspected batch of sample media compared to the three alternate batches, at a range of recovery levels comparable to historical field data. The data suggest a substantial lack of uniformity in similar or identical batches of spray drift monitoring sample media and highlight the critical value of quality control samples, like matrix and stability spikes, in the design of spray drift field studies.

9:30 - 9:55 a.m., EDT

Presentation: Key Factors Affecting Drift Deposition from Unmanned Aerial Vehicle Applications (4541708)

Presenter: Dwayne Moore, Stone Environmental. Co-Author: Sebastian Castro-Tanzi.

Abstract: Unmanned Aerial Vehicles (UAVs) have become increasingly important in pesticide applications due to their precision, efficiency, and environmental benefits. However, downwind drift deposition from drones varies significantly based on nozzle type and environmental conditions. Using the data collected by the Unmanned Aerial Pesticide Application Systems Task Force (UAPASTF) from studies conducted in 2023 and 2024 at 8 study locations worldwide, we undertook multivariate statistical analyses to answer the questions: (1) Does study location influence ground sprayer and UAV drift deposition, and (2) Which application practices and environmental conditions best explain the variation observed in spray drift deposition from ground sprayer and UAV applications? Our analyses included a Principal Components Analysis (PCA) to determine how study locations differed regarding meteorological conditions during UAV applications, and a Canonical Correlation Analysis (CCA) to determine the relative importance of pesticide application method (ground, UAV), spray quality (fine, medium, coarse), and meteorological conditions on downwind spray drift deposition. The PCA indicated high overlap between locations regarding meteorology, i.e., within-location variability in meteorology was as or more significant than between location variability. The general lack of separation between study locations was likely due to the tight constraints on allowable windspeeds for applications to proceed (i.e., 2.0 to 5.0 m/s or 4.5 to 11.2 mph). The separation between locations that did occur in the PCA tended to be between locations that had low temperatures and high relative humidities and those that had the opposite conditions. The CCA confirmed that meteorological conditions differed only slightly between locations and, as a result, played only a small role in predicting downwind drift. Application method and, to a lesser extent, spray quality had a more important influence on drift deposition. Drift deposition was higher following UAV applications compared to ground applications, and increased as droplet size decreased (fine > medium > coarse).

9:55 - 10:20 a.m., EDT

Presentation: Comparison of Spray Deposition from Ground and Unmanned Aerial Vehicle Pesticide Applications in a Multi-Location Field Experiment (4541629)

Presenter: Dwayne Moore, Stone Environmental. Co-Authors: Michael Winchell, Stone Environmental; Sebastian Castro-Tanzi; Bradley Fritz.

Abstract: Unmanned aerial vehicles (UAVs) provide a versatile alternative to traditional aerial and ground applications of pesticides. However, UAV applications lack the required regulatory framework to maximize their utility. One key question is how off-target spray deposition from UAVs applications compares to other pesticide application platforms. The prediction of off-target spray drift from UAVs is challenging, in part due to the complexities of their multi-rotor flow fields interacting with a crosswind. Thus, it is difficult to predict downwind displacement of the spray swath relative to the flight line. Accounting for this spray swath displacement is important to ensure agronomic effectiveness and to reduce off-target drift. UAV and ground spray deposition data collected from the Unmanned Aerial Pesticide Application Systems Task Force (UAPASTF) field study at eight locations spread over five continents were compared. In-swath and off target spray deposition were analyzed through ordinary least squares and generalized linear mixed models, respectively. A post-hoc correction using a center of deposition (CoD) approach to address spray deposition displacement from the UAV applications was also explored. Segmented regression analysis was used to quantify the relative inflection distance where spray deposition from UAV applications transitioned from a slow to a fast-decreasing pattern. These breakpoints served as additional benchmarks to compare the original and displacement-corrected UAV deposition datasets. The statistical analysis indicated that UAV in-swath deposition was significantly lower compared to ground applications, whereas off-target downwind drift was higher. The CoD displacement approach was not effective in determining appropriate in-field offsets to improve off-target spray deposition predictions. This result was likely due to the approach being designed for single-swath applications whereas the UAPASTF studies involved multi-swath applications. Overall, we found that the segmented regression modeling approach was more successful for estimating downwind spray drift offsets for UAV applications.

Session: Protection of Agricultural Productivity, Public Health, and the Environment (Room: S502a McCormick Place Convention Center)

9:55 - 10:20 a.m., EDT

Presentation: Simulating Integrated Environmental and Agronomic Outcomes in Vermont Agricultural Systems, with an APEX-Based Decision Support Tool (4541673)

Presenter: Jody Stryker, Stone Environmental. Co-Authors: Hannah Rubin, Stone Environmental; Jens Kiesel, Stone Environmental; Michael Winchell, Stone Environmental.

Abstract: Growers are under immense pressure from rising costs, unstable market dynamics, inadequate policy and financial support, unpredictable weather, and the increased demand for agricultural commodities from a shrinking resource. Supporting sustainable and regenerative agriculture is critical for ensuring we can meet future human needs and protect environmental resources. We present the enhancement of an existing decision support tool originally focused on meeting water quality goals in Vermont agriculture, to additionally predict agronomic indicators and environmental outcomes related to soil health, greenhouse gas emissions, and carbon sequestration. This involved a multi-objective calibration of the Agricultural Policy eXtender model (APEX), validation of predicted responses in interrelated metrics to long-term agronomic management scenarios, and build-out of a web-based tool. The comprehensive model evaluation included an innovative calibration approach incorporating different types of observed data from multiple agricultural sites and time periods to establish a global parameter set that resulted in robust performance across many model outputs. Model validation included model batch runs designed to statistically replicate soil health datasets in the region across a large range of field conditions representative of agriculture in Vermont. Our modeling approach shows the APEX model reasonably captures observed patterns in key indicators (e.g., carbon sequestration, soil organic matter, soil respiration, phosphorus and nitrogen losses, crop yields, and others) and appropriately represents the environmental impacts of long-term agronomic management on these indicators. The Farm-PREP tool now allows users to evaluate field and farm specific outcomes of agronomic management scenarios on a suite of sustainability indicators, enabling more informed navigation of the potential benefits as well as trade-offs that arise from management decisions. The goal of this enhanced tool framework is to increase adoption of agricultural management practices that enhance and/or maintain healthy soils, reduce greenhouse gas emissions and sequester carbon, reduce nutrient and pesticide losses, and meet the agronomic needs of farmers through customized scenario analysis.