Introduction
Crop producers in the United States are under pressure to reduce the amount of pesticides they use to mitigate environmental loading and reduce off-target movement (US EPA 2024). Additionally, rising input costs are always a challenge for producers if they wish to maintain profitability (USDA-NASS 2025). Weeds are often not uniformly distributed across a field; instead, they typically emerge in clumps or patches (Cardina et al. Reference Cardina, Johnson and Sparrow1997; El Jgham et al. Reference El Jgham, Abdoun and El Khatir2023; Franz et al. Reference Franz, Gebhardt and Unklesbay1991; Metcalfe et al. Reference Metcalfe, Milne, Coleman, Murdoch and Storkey2019; Rew and Cousens Reference Rew and Cousens2001; Sapkota et al. Reference Sapkota, Singh, Neely, Rajan and Bagavathiannan2020; Stafford and Miller Reference Stafford and Miller1993; Wiles et al. Reference Wiles, Oliver, York, Gold and Wilkerson1992). Site-specific management of weeds offers an opportunity to reduce herbicide usage. Real-time targeted herbicide applications provide a way to reduce environmental loading, potentially improve profitability, and efficiently control weeds without drastic changes to production practices (Avent et al. Reference Avent, Norsworthy, Patzoldt, Schwartz-Lazaro, Houston, Butts and Vazquez2024; El Jgham et al. Reference El Jgham, Abdoun and El Khatir2023; Gianessi and Reigner Reference Gianessi and Reigner2007; Stafford and Miller Reference Stafford and Miller1993; Wiles et al. Reference Wiles, Oliver, York, Gold and Wilkerson1992).
In 2022, the John Deere company (Moline, IL) released the See & Spray technology that features a dual-boom sprayer capable of simultaneously applying broadcast and targeted herbicides (Deere & Company 2025; Gizotti de Moraes Reference Gizotti de Moraes2024; Lazaro et al. Reference Lazaro, Houston and Patzoldt2024). In 2023, the company introduced a single-tank platform, which can be purchased new or retrofitted to some existing sprayers. Both systems use machine vision and deep learning to detect weeds in fallow systems or to distinguish weeds from crops, allowing real-time applications of herbicides specifically to the weeds (Fu et al. Reference Fu, Padwick, Ostrowski and Anderson2022; Padwick et al. Reference Padwick, Patzoldt, Cline, Denas and Tanna2023; Redden Reference Redden2023). Previous research with soybean and corn (Zea mays L.) comparing targeted applications and broadcast applications demonstrated comparable weed control with a program approach (Avent et al. Reference Avent, Norsworthy, Patzoldt, Schwartz-Lazaro, Houston, Butts and Vazquez2024; Leise et al. Reference Leise, Singh, Cafaro La Menza, Knezevic and Jhala2025). In Nebraska, Leise et al. (Reference Leise, Singh, Cafaro La Menza, Knezevic and Jhala2025) reported that targeted applications of herbicides to weeds growing among corn crops provided ≥90% Palmer amaranth control when the Greeneye spraying system was used (Greeneye Technology, Tel Aviv-Yafo, Israel). In the soybean experiment, targeted applications with See & Spray provided slightly greater control of waterhemp [Amaranthus tuberculatus (Moq.) J.D. Sauer] than broadcast applications (92% vs. 91%; Leise et al. Reference Leise, Singh, Cafaro La Menza, Knezevic and Jhala2025). Avent et al. (Reference Avent, Norsworthy, Patzoldt, Schwartz-Lazaro, Houston, Butts and Vazquez2024) reported no biological control differences in Palmer amaranth between targeted and broadcast applications; both provided ≥94% control after early postemergence applications. Both publications reported using specific machine settings and emphasized how altering the settings could affect the results. However, Leise et al. (Reference Leise, Singh, Cafaro La Menza, Knezevic and Jhala2025) used settings at the highest weed-detection level, and Avent et al. (Reference Avent, Norsworthy, Patzoldt, Schwartz-Lazaro, Houston, Butts and Vazquez2024) used a medium detection setting.
In 11 independent experiments with corn, cotton (Gossypium hirsutum L.), and soybean in 2022, Avent et al. (Reference Avent, Norsworthy, Zimmer, Young, Contreras, Everman, Hager, Patzoldt, Schwartz-Lazaro, Houston, Butts and Vazquez2025) evaluated how changes in the spray sensitivity setting influenced the likelihood of treating weeds before they emerge. Changing the spray setting from broadcast to the lowest spray sensitivity reduced the likelihood of treating Amaranthaceae weeds between soybean rows from >99% to approximately 55% when the weeds were 1.9 cm tall and 2.5 cm wide. The substantial reduction in the ability to treat these weeds is concerning due to the widespread herbicide resistance by Palmer amaranth and waterhemp (Carvalho-Moore et al. Reference Carvalho-Moore, Norsworthy, Porri, dos Santos, Barber, Sudhakar, Meiners and Lerchl2025; Evans et al. Reference Evans, Strom, Riechers, Davis, Tranel and Hager2019; Foster and Steckel Reference Foster and Steckel2022; Heap Reference Heap2025; Randell-Singleton et al. Reference Randell-Singleton, Hand, Vance, Wright-Smith and Culpepper2024). Missing weeds at susceptible growth stages is unacceptable from a herbicide resistance management perspective (Bagavathiannan and Norsworthy Reference Bagavathiannan and Norsworthy2012; Heneghan and Johnson Reference Heneghan and Johnson2017; Norsworthy et al. Reference Norsworthy, Ward, Shaw, Llewellyn, Nichols, Webster, Bradley, Frisvold, Powles, Burgos, Witt and Barrett2012; Spaunhorst et al. Reference Spaunhorst, Devkota, Johnson, Smeda, Meyer and Norsworthy2018).
Another limitation of targeted applications is the potential for reduced residual herbicide use. Since operators usually prioritize greater efficiency by reducing the amount of herbicides they use and by spraying at high speeds (Butts et al. Reference Butts, Barber, Norsworthy and Davis2021; MSU 2025), applicators may be inclined to target-apply residuals using single-tank systems. To mitigate herbicide resistance evolution, weed scientists across the United States advocate for overlapping the application of residual herbicides (Barker et al. Reference Barker, Pawlak, Duke, Beffa, Tranel, Wuerffel, Young, Porri, Liebl, Aponte, Findley, Betz, Lerchl, Culpepper, Bradley and Dayan2023; Boyd et al. Reference Boyd, Moretti, Sosnoskie, Singh, Kanissery, Sharpe, Besançon, Culpepper, Nurse, Hatterman-Valenti, Mosqueda, Robinson, Cutulle and Sandhu2022; Chahal et al. Reference Chahal, Ganie and Jhala2018; de Sanctis et al. Reference de Sanctis, Barnes, Knezevic, Kumar and Jhala2021). Additionally, research on new commercial spray technologies that perform simultaneous detection and application is limited, particularly when considering the ecological effects on the weed seedbank from continuous use of these technologies or their use at different detection settings. Therefore, this research aimed to examine how targeting herbicides using the highest and lowest weed detection spray settings compares to traditional broadcast applications in soybean fields.
Materials and Methods
Site, Design, and Field Data Collection
A 3-yr experiment was initiated in fields where soybean is grown at the Northeast Research and Extension Center in Keiser, Arkansas (35.674281N, 90.077891°W), on May 3, 2022. Plots were monitored yearly to record the weed seedbank changes by marking the four corners of the trial area immediately after harvest. Plots consisted of eight rows, 7.7 m wide by 67 m long, with 96.5 cm between each row. In 2022, AG45XF0 soybean was planted at 350,000 seed ha−1, and due to lodging, B4885XF soybean was planted in 2023 and 2024 at the same seeding rate. Lodging across the field was <15% and not dependent upon treatment. The experimental site had a Steele loamy sand soil (sandy over clayey, mixed, superactive, nonacid, thermic Aquic Udifluvent) consisting of 64% sand, 16% silt, 20% clay, 6.8 pH, and 2.1% organic matter (Soil Testing and Research Laboratory, Marianna, AR). The production system consisted of conventional tillage with raised beds hipped in the fall before the beds were shaped in the spring. Other practices included no preplant soil amendments due to historical soil test values, and furrow irrigation was provided as needed to ensure the crop received 2.5 cm of moisture throughout the week. Soybean planting occurred on May 4, 2022, May 18, 2023, and April 23, 2024.
In-season production practices are listed in Table 1. Early-postemergence applications of herbicides occurred 19 to 39 d after planting, with mid-postemergence applications occurring 11 to 21 d after the early-postemergence application. The reason for different application intervals from year to year was to simulate a program approach in which applications were triggered when newly emerged Palmer amaranth plants of approximately 7.5 cm tall were found. Additionally, rain and irrigation partly affected when herbicides were applied each year. If no new weeds were observed at the mid-postemergence application, the application occurred as soon as the soybean reached the R1 growth stage.
Table 1. Planting, herbicide application, and harvest dates.
The experiment was conducted using a randomized complete block design with three treatments (factors) and six replications. Each treatment consisted of a broadcast application at planting, with the same herbicide combination (Table 2), but with different postemergence application methods. The application methods included a broadcast standard, targeted applications at the lowest See & Spray sensitivity, and targeted applications at the highest sensitivity. All herbicide were applied with a scaled, dual-boom See & Spray small-plot machine previously described by Avent et al. (Reference Avent, Norsworthy, Patzoldt, Schwartz-Lazaro, Houston, Butts and Vazquez2024). The sprayer was mounted to the front-end loader of a tractor and equipped with 10 nozzles spaced 38.1 cm apart. The sprayer was calibrated to deliver 140 L ha−1 at 283 kPa while traveling 19.3 kph with rearward-inclined PS3DQ0005 nozzles (medium droplet classification; Deere & Company, Moline, IL). The small spray buffer setting was used for postemergence applications. The detection models used throughout the years were subject to change due to system updates, and each year, the detection algorithms matched what was available to commercial sprayers.
Table 2. Herbicides applied during the 3-yr soybean experiment.
a Preemergence herbicides were broadcast at planting. Early postemergence and mid-postemergence herbicides were either broadcast or targeted.
b Herbicide products: flumioxazin + pyroxasulfone, Fierce 76 WDG (Valent U.S.A., Walnut Creek, CA); paraquat, Gramoxone SL 2.0 or 3.0 (Syngenta Crop Protection, Greensboro, NC); nonionic surfactant, Preference (Winfield Solutions, St. Paul, MN); glufosinate, Liberty 280SL (BASF, Research Triangle Park, NC); glyphosate, Roundup PowerMAX 3 (Bayer CropScience, St. Louis, MO); S-metolachlor, Dual Magnum 7.62 EC (Syngenta); acetochlor, Warrant (Bayer).
c Costs are based on the calculated 10-yr average reported using the Mississippi State Archived Budget Publications from 2016 to 2025 (https://www.agecon.msstate.edu/whatwedo/budgets/archive.php). Prices for flumioxazin + pyroxasulfone are based on a 9-yr average and acetochlor is based on a 7-yr average. The nonionic surfactant cost is based on the general term surfactant reported in the budgets.
In-season data included density assessments of all weed species when postemergence herbicides were applied. Postemergence applications included blue dye (Super Signal Blue; Precision Labs, Kenosha, WI) at 0.25% (v/v) to indicate whether weeds were missed after application. When herbicides were applied, weed counts and misses were determined for the length of each plot between soybean Rows 2 and 3 and Rows 6 and 7, equating to 129.3 m2 plot−1. Missed weeds were divided by the weeds present at application, and total weed counts per 129.3 m2 were extrapolated to hectares. Palmer amaranth was the most abundant weed, but other species included prickly sida (Sida spinosa L.), morningglory (Ipomoea spp.), horse purslane (Tranthema portulacastrum L.), and broadleaf signalgrass [Urochloa platyphylla (Munro ex C. Wright) R.D. Webster].
At harvest, Palmer amaranth escapes capable of reproducing were counted in the entire plot. All plots were harvested with a John Deere S690 harvester (Deere & Company), and grain samples were collected from each plot to standardize yields back to 13% grain moisture. Plot-to-plot contamination from wind-blown chaff exiting the combine was not prevented, but to prevent additional contamination of settled chaff, beds were reshaped within 3 d after harvest and marked to keep plots consistent from one year to the next. Additional data included area sprayed metrics recorded by the small-plot sprayer.
Exhaustive Germination Data Collection
At planting in 2022 and after harvest in 2024, 20 soil samples were collected from each plot to estimate the weed seedbank when the trial began and when it ended. Each sample was cut to a depth of 7.5 cm with an 11.4-cm-diam cup cutter. After soil samples were collected they were homogenized, the total mass recorded, and a 2-kg subsample was placed in a plastic greenhouse tray (52 by 40 by 5 cm) filled with a 1:1 ratio of potting soil (PRO-MIX BX; Greenhouse Megastore, Danville, IL) and field soil. Trays were brought to a greenhouse in Fayetteville, Arkansas, where nighttime and daytime temperatures were maintained at approximately 21.1 and 30 C, respectively, with a 16-h photoperiod. Each tray was watered over the top twice daily, and no fertilizer amendments were added. Emerged weeds by tray were counted and pulled for 4 wk. Thereafter, trays were placed in a subzero freezer for 2 wk. This cycle was repeated four times. The number of weeds per tray was extrapolated to square meters by accounting for the cup-cutter cross-sectional area, soil mass collected per plot, and the 2-kg subsample. Palmer amaranth ending counts were subtracted from the beginning counts to calculate the change in Palmer amaranth per square meter.
Economics Data Collection
To determine the economics of the various treatments, 10-yr average costs were determined for all herbicides, interest rates, and labor from the Mississippi State archived budgets and adjusted for inflation over the previous 5 yr (MSU 2025). The interest and labor rates were then input to the University of Arkansas System Division of Agriculture Enterprise Budgets, which we then used to calculate fixed and variable costs, assuming US$600,000 for a broadcast sprayer and approximately $120,000 for the two See & Spray Premium upgrades (Watkins Reference Watkins2024; https://configure.deere.com/). The two See & Spray Premium sprayer costs were added to the budget at $640,000 and $680,000, respectively. The $40,000 difference between a broadcast and See & Spray Premium assumes two comparable sprayers were already equipped with the required GPS and nozzle bodies and required only the addition of a camera and vision processing units, and installation of a wiring harness. The $80,000 upgrade assumes a base-model sprayer that needs all upgrades to facilitate targeted applications. The economic analysis compared differences within each application timing and total costs.
Each scenario incorporated the application, subscription, and herbicide costs on a per hectare basis. Subscription costs are reported at $12.36 ha−1 on the area not sprayed. For instance, if 40% of a hectare was sprayed, the cost would be $7.41. On average, broadcast sprayers in Arkansas travel approximately 24.2 kph (Butts et al. Reference Butts, Barber, Norsworthy and Davis2021), whereas the See & Spray Premium is limited to 19.3 kph. The different speeds affected fuel and labor costs, while the fixed and repair costs differed due to machine valuation. The fixed, repair, fuel, and labor costs for a broadcast sprayer were $14.97, $1.19, $0.82, and $0.32 ha−1, respectively. Fixed, repair, fuel, and labor costs for targeted applications were $19.49, $1.55, $1.01, and $0.40 ha−1, respectively, for the $40,000 upgrade; and $21.21, $1.68, $1.01, and $0.40 ha−1, respectively, for the $80,000 upgrade.
Data Analysis
All data were analyzed using JMP Pro software (v.18; SAS Institute, Cary, NC). Yearly data were analyzed as a repeated measure with an unstructured covariance considering application method, application timing (only for counts and misses), and year as fixed effects. Changes in Palmer amaranth density from exhaustive germination were analyzed as a one-way analysis. Blocks were considered random effects. The collective density of all weeds at both postemergence applications and the density of Palmer amaranth escapes at the end of the season were analyzed using generalized linear mixed models with a Poisson distribution. Weed misses were analyzed with a normal distribution due to skewed residuals with Poisson, gamma, lognormal, negative binomial, and exponential distributions. Yield, area sprayed, and application costs passed equal variance and residual assumptions and were analyzed as standard least squares. Yield and total postemergence application costs were analyzed using only the year as a repeated measure.
All data subjected to ANOVA were considered significant at P ≤ 0.05, and the means were separated using Tukey’s HSD (α = 0.05). All data figures are displayed with box and whisker plots of the observed data, and compact letters displayed on graphs are based on the least square means. Additionally, a planned orthogonal contrast was used to address whether targeted residual applications at early postemergence increased the number of weeds present at mid-postemergence relative to the broadcast treatments. Another important consideration was the presence of potential outliers in the data set. Since all collected data were continuous, outliers were not excluded from the analysis.
Results and Discussion
A significant interaction between application method by year and application timing by year occurred for weed density (Table 3). In 2022, weed densities were similar in each application timing, which averaged 1,500 and 270 plants ha−1 at early and mid-postemergence, respectively (Figure 1). The similarities at trial initiation are important to consider since differences indicate the need to either consider the initial population as a covariate in the analysis or to calculate the change in population density. The weed density following broadcast and targeted applications at the highest sensitivity did not increase over the years. However, targeted applications at the lowest spray sensitivity increased yearly, with the greatest overall density in 2024. Similarly, the soil seedbank estimates indicated that after three soybean production seasons, Palmer amaranth germination increased by 783, 1,077, and 2,246 plants m−2 after broadcast, targeted at the highest sensitivity, and targeted at the lowest sensitivity, respectively (Table 4). The increase can be explained by the main effect of application method on Palmer amaranth escapes that were present at harvest (Table 5). Averaged over the years, targeted applications at the lowest spray sensitivity resulted in the greatest number of Palmer amaranth escapes at 1,380 ha−1 compared with 268 and 343 escapes ha−1 after broadcast and highest spray sensitivity applications, respectively (Table 5; Figure 2).
Table 3. Effect summary for in-season weed counts and area sprayed with targeted applications. a
a P-values were calculated with JMP Pro software (v.18.0; SAS Institute, Cary, NC) using generalized linear mixed models for counts and standard least squares for area sprayed and misses.
b Broadcast treatments were excluded from area sprayed due to no variation in the response.
Figure 1. Effect of application method by year interaction (left) and timing by year interaction (right) on weed density averaged over application timing. Box and whiskers are based on observed data. Levels not containing similar letters represent different least square means according to Tukey’s HSD at α = 0.05. Figures 1–3 and 3–7 were generated using the graph builder function in JMP Pro software (v.18; SAS Institute, Cary, NC).
Table 4. Effect of application method on Palmer amaranth counted from exhaustive germination evaluations.a,b
a Differing letters in a column indicate significantly different means based on Tukey’s HSD at α = 0.05.
b P-values were calculated with JMP Pro software (v.18.0; SAS Institute, Cary, NC) using standard least squares.
Table 5. Effect summary of Palmer amaranth escapes at harvest and soybean yield. a
a P-values were calculated with JMP Pro software (v.18.0; SAS Institute, Cary, NC) using generalized linear mixed models for escapes and standard least squares for soybean yield.
Figure 2. Effect of application method on weeds missed at application per hectare averaged over application timing and years. Box and whiskers are based on observed data. Levels not containing similar letters represent different least square means according to Tukey’s HSD at α = 0.05.
The weed density increased by 67% after broadcast and targeted applications at the highest spray sensitivity each year, whereas weed density increased by 262% each year after herbicides were applied at the lowest spray sensitivity (Figure 1). The Palmer amaranth seedbank combined with the weed density increase in plots that received broadcast and targeted applications at the highest spray sensitivity means that some plot-to-plot contamination cannot be ruled out. Palmer amaranth has demonstrated the ability to laterally grow more than 5 m from one year to the next (Norsworthy et al. Reference Norsworthy, Griffith, Griffin, Bagavathiannan and Gbur2014), meaning that plants at the edge of the plots could have easily dispersed seeds to the experiment plots. The increasing weed density is concerning from a herbicide resistance management perspective (Norsworthy et al. Reference Norsworthy, Ward, Shaw, Llewellyn, Nichols, Webster, Bradley, Frisvold, Powles, Burgos, Witt and Barrett2012), especially with the increasing number of reports of herbicide-resistant Palmer amaranth in the United States (Brabham et al. Reference Brabham, Norsworthy, Houston, Varanasi and Barber2019; Carvalho-Moore et al. Reference Carvalho-Moore, Norsworthy, González-Torralva, Hwang, Patel, Barber, Butts and McElroy2022; Foster and Steckel Reference Foster and Steckel2022; Heap Reference Heap2025; Hwang et al. Reference Hwang, Norsworthy, Piveta, Souza, Barber and Butts2023; Randell-Singleton et al. Reference Randell-Singleton, Hand, Vance, Wright-Smith and Culpepper2024). In Arkansas, for example, a Palmer amaranth population was recently reported to be resistant to the sites of action of seven postemergence herbicides (Carvalho-Moore et al. Reference Carvalho-Moore, Norsworthy, Porri, dos Santos, Barber, Sudhakar, Meiners and Lerchl2025).
Based on the weed density results, targeted applications of herbicides at the lowest spray sensitivity is resulting in an increasing selection for herbicide resistance genes through the yearly increase in weed density (Jasieniuk et al. Reference Jasieniuk, Brûlé-Babel and Morrison1996), especially when weeds are missed at the early postemergence stage and when herbicides are applied at the lowest spray sensitivity, which also led to greater weed densities when herbicides were applied at mid-postemergence (Figures 3 and 4). Additionally, compared with the other application methods, targeted applications of S-metolachlor at early postemergence with the lowest spray sensitivity led to a greater weed density at mid-postemergence (Table 6). Averaged over the years and after subtracting the weeds that were missed at early postemergence, the weed density at mid-postemergence averaged 840, 989, and 4,310 plants ha−1 after broadcast applications, targeted applications at the highest sensitivity, and targeted applications at the lowest sensitivity, respectively.
Figure 3. Effect of application method on the proportion of Palmer amaranth escapes, averaged over years. Box and whiskers are based on observed data. Levels not containing similar letters represent different least square means according to Tukey’s HSD at α = 0.05.
Figure 4. Photographs of the same plot treated with targeted applications at the lowest sensitivity in 2024. Many of the weeds that were observable at early postemergence were missed, they became too large to control at mid-postemergence, and they became reproductive by harvest.
Table 6. Orthogonal contrast of weed counts at the mid-postemergence timing to determine whether targeted residual herbicides at early postemergence increase the subsequent weed population. a
a Orthogonal contrast was calculated in the fit model platform.
b Weed density was averaged over years and represents the weeds counted at mid-postemergence application minus the weeds missed at early postemergence.
The overall weed density increase at the lowest sensitivity from one year to the next appears to follow the exponential growth phase commonly observed in ecological infestations, while the densities for other application methods appear to be in the lag phase (Mack et al. Reference Mack, Simberloff, Lonsdale, Evans, Clout and Bazzaz2000). The increasing weed density could also reduce the subsequent area sprayed at the lowest sensitivity in the following years by increasing the proportion of the field that becomes infested. These effects could not be captured in a 3-yr experiment and should be considered in future experiments.
For the targeted applications, both the highest and lowest sensitivities provided similar reductions in area sprayed ranging from 20% to 90% relative to broadcast applications (Figure 5). The only advantage in the area sprayed for the lowest sensitivity occurred at the early-postemergence application timing, which averaged 41.3% each year compared with 57.9% with the highest sensitivity. In the United States, newly registered and re-registered herbicides considered under the Federal Insecticide, Fungicide, and Rodenticide Act will incorporate the potential for population-level impacts to endangered species (US EPA 2024). During consideration, herbicides may be required to incorporate mitigation strategies to protect endangered species. Targeting herbicides is one of the proposed strategies. Both the highest and lowest spray sensitivities provide savings that fall within the medium efficacy classification for Environmental Protection Agency mitigation measures, providing two mitigation points; however, this research featured a specific herbicide program, and using a different preemergence residual herbicide could alter the savings by increasing the weed density when postemergence herbicides are applied (Woolard et al. Reference Woolard, Norsworthy, Roberts, Thrash, Barber, Sprague and Avent2025).
Figure 5. Effect of application method by timing interaction on area sprayed, averaged over years. Box and whiskers are based on observed data. Levels not containing similar letters represent different least square means according to Tukey’s HSD at α = 0.05.
Soybean yield was greater after targeted applications at the lowest sensitivity compared with yield after broadcast applications (Figure 6). Averaged over the years, soybean yield was 4,220 kg ha−1, 4,350 kg ha−1, and 4,480 kg ha−1 after broadcast, highest sensitivity, and lowest sensitivity treatments, respectively. The high yield from the lowest sensitivity treatment is difficult to explain since labeled rates of postemergence herbicides were used in all 3 yr. One explanation could be that the higher yield was due to a reduced area sprayed with labeled rates of glufosinate, glyphosate, S-metolachlor, and acetochlor. All these herbicides are detoxified through metabolic pathways that require energy expenditures (Breaux Reference Breaux1986; Moldes et al. Reference Moldes, Medici, Abrahão, Tsai and Azevedo2008; Pline et al. Reference Pline, Wu and Hatzios1999), and by targeting herbicides, some soybean plants could have devoted more resources toward development. However, the only difference in the area sprayed between targeted applications was observed at early postemergence, suggesting the early-season applications may be a causal effect (Figure 5).
Figure 6. Effect of application method on soybean yield, averaged over years. Box and whiskers are based on observed data. Levels not containing similar letters represent different least square means according to Tukey’s HSD at α = 0.05.
The 260 kg ha−1 increase in yield after the lowest sensitivity treatment compared to the broadcast may not justify the 198% increase in the Palmer amaranth seedbank over 3 yr that occurred with the lowest sensitivity (Figure 5; Table 4), especially considering the limited postemergence herbicide options available for northeast Arkansas soybean producers (Carvalho et al. Reference Carvalho-Moore, Norsworthy, Porri, dos Santos, Barber, Sudhakar, Meiners and Lerchl2025). If an effective postemergence herbicide is lost due to resistance, the newly introduced herbicide tends to be more expensive, thereby affecting profitability in the future (Kniss et al. Reference Kniss, Mosqueda, Lawrence and Adjesiwor2022; Livingston et al. Reference Livingston, Fernandez-Cornejo and Frisvold2016; Norsworthy et al. Reference Norsworthy, Ward, Shaw, Llewellyn, Nichols, Webster, Bradley, Frisvold, Powles, Burgos, Witt and Barrett2012). Ultimately, a producer must decide whether the short-term benefits (improved herbicide savings and increased yield) outweigh the risk of herbicide resistance evolution. Future research is needed to determine the cause of yield improvement and whether it persists in subsequent years, and to quantify whether a shift in Palmer amaranth resistance is due to targeted herbicide applications.
Based on the total application costs, both targeted applications provided a return relative to broadcast programs at both application timings (Figures 5 and 7). Averaged over the years, targeted applications of all herbicides cost $140.89 ha−1 and $118.57 ha−1 for the highest and lowest spray sensitivities, respectively, whereas broadcast herbicides cost $227.22 ha−1. The lowest sensitivity provided the highest return on investment for total postemergence costs, which is attributed to greater savings at early postemergence (Figure 5). Using the range of savings depicted in Figure 7, producers who adopt this technology at the $40,000 upgrade cost could expect to pay off the investment by treating 266 ha−1 to 1,229 ha−1 if using the highest sensitivity, or 218 ha−1 to 909 ha−1 with the lowest sensitivity. At the $80,000 upgrade cost, the extremes shift to 220 ha−1 and 1273 ha−1, respectively. At no point did the savings overlap with the broadcast program costs, indicating that targeting herbicides with both residual- and postemergence-active chemistries can provide an economic return in soybean production following a broadcast-applied preemergence herbicide at planting.
Figure 7. Effect of application method and year on application costs (US$). Application costs account for the 10-yr average of herbicide, interest, and labor costs; subscription fees; and equipment cost accounting for efficiency. Broadcast applications are excluded from the figure but are displayed with the dashed line. Assuming a $600,000 broadcast sprayer valuation, the dashed line represents the cost associated with the total broadcast postemergence herbicide program at $227.22 ha−1. Box and whiskers are based on observed data. Levels not containing similar letters represent different least square means according to Tukey’s HSD at α = 0.05. Blue boxes are for a See & Spray Premium machine valued at $640,000 and are separated using uppercase letters. Purple boxes are for a See & Spray Premium machine valued at $680,000 and are separated using lowercase letters.
Practical Implications
The weed scientists with the University of Arkansas System Division of Agriculture will continue to recommend broadcast-applying residual herbicides when using targeted applications (J. Norsworthy, personal communication). Due to time constraints and the potential to increase the return on investment, producers may still be inclined to target-apply residual herbicides. The research demonstrated here illustrates the ability of postemergence targeted applications to provide similar weed control and improve producer profitability when using the highest spray sensitivity. Using the lowest spray sensitivity may reduce herbicide costs in the short-term, but the greater risk for escapes each year will likely have negative long-term consequences (e.g., greater weed density, potential herbicide resistance, and lower savings), which will be determined through continuation of this experiment. Both targeted application methods demonstrate the potential for this technology to provide mitigation points for endangered species and reduce the environmental loading of pesticides through spraying smaller areas. Additionally, producers may notice a subtle yield benefit by targeting herbicides rather than broadcasting them across the field, but this result may not be reciprocated to different sites of action or in areas with higher weed densities than evaluated here. The latter would likely increase the proportion of the field treated, and additional research is needed to validate these results.
It is important to note that these results may not translate to all soybean production areas in the United States. This research was conducted in a conventionally tilled, bedded soybean production system, which is typical of the midsouthern United States. Furthermore, the preemergence herbicides included a premixture of pyroxasulfone and flumioxazin, which have demonstrated excellent Palmer amaranth control out to 28 d after emergence (Houston et al. Reference Houston, Barber, Norsworthy and Roberts2021). Results could be vastly different if a less or more effective preemergence residual herbicide is used than was evaluated here. Machine settings and nozzle selection may also provide varying results. The sprayer was set to use a small-low buffer, which dictates how many and how long nozzles are activated to treat a weed. PS3DQ0005 nozzles also provide a 100-degree fan angle with a medium droplet classification at 283 kPa (unpublished data; John Deere US 2018). Using different sprayer setups could affect performance, and more research is needed on buffer and nozzle selections.
Acknowledgments
We thank the personnel at the Northeast Research & Extension Center in Keiser, Arkansas, and the technical and resource support personnel from Blue River Technology and John Deere.
Funding statement
Funding for this research was provided by Blue River Technology (Deere & Company) and the United Soybean Board.
Competing interests
The authors declare they have no competing interests.
Open access
