Impact of soil residual auxin herbicide on seedling emergence and biomass differs between soil types, water pulse event, and seedling species
Author contributions: AMF, EAL, DH conceived and designed the research; DH performed the experiments; CVC, MMR, RWRS analyzed the data; RWRS, MMR, EAL, DH, CVC, JW, AMF wrote and edited the manuscript.
Abstract
Synthetic-auxin herbicides are often used to control woody plants and aid in grassland restoration. Seed-based restoration is common alongside herbicide applications and there may be unintended effects of these herbicides on dryland plants at the seed and seedling stages. Additionally, abiotic conditions at the time of herbicide application may influence herbicide–soil–plant interactions. We conducted a greenhouse study to examine the effects of a common shrub-control herbicide mix and its interaction with soil type and a post-herbicide water pulse on common desert plant seeds and seedlings. In this greenhouse study, we found that a subset of species responded negatively to soil residual herbicide activity of a mixture of aminopyralid, clopyralid, and triclopyr at the seed and seedling stages. Species sensitive to soil herbicide residues were primarily shrub and forb species that are often the target species of herbicide applications for woody plant control, such as honey mesquite (Prosopis glandulosa) and creosote bush (Larrea tridentata). However, two shrub species (four-wing saltbush [Atriplex canescens], soaptree yucca [Yucca elata]) and one perennial grass species (Arizona cottontop [Digitaria californica]), which are used in dryland restoration projects, were found to be particularly sensitive to soil residual herbicide activity. Thus, if using these herbicides to control woody plants and restore herbaceous vegetation via active seeding or relying on the in situ seed bank, considerations should be given to what species are used in the seed mix, what species are already present in the soil seed bank, and other details of the circumstances of herbicide application.
Implications for Practice
- Soil residual activity of a commonly used mixture of aminopyralid, clopyralid, and triclopyr herbicides differentially impact seedling emergence and growth of monocot and dicot restoration species, which could alter plant community trajectory.
- Depending on the species, short-term soil residual herbicide activity may interact with soil type and a post-herbicide application pulse of water, such as a large rain event.
Introduction
Drylands face numerous ecosystem level threats, including climate change, invasive species, and land use changes (Hoover et al. 2020). These threats can lead to the ultimate risks of ecological vulnerability and shifting ecological states (Bestelmeyer et al. 2018; Hoover et al. 2020). One common ecological state shift is the global increase of woody vegetation in drylands historically classified as grasslands and savannas (Ravi et al. 2010; Archer & Predick 2014; Bestelmeyer et al. 2018). Woody plant encroachment has resulted in ecological state transitions from grass-dominated ecosystems to shrublands, with varying consequences to biodiversity and ecosystem services (Barger et al. 2011; Eldridge et al. 2011; Bestelmeyer et al. 2018).
Common management approaches to reduce woody plant cover include mechanical removal (Mndela et al. 2023), prescribed fire (Limb et al. 2016; Ansley et al. 2022), and chemical shrub treatment (Tunnell et al. 2006; Eddy et al. 2020). Target specificity, cost-effectiveness, and restoration efficacy vary by method (Hamilton 2004). The application of selective herbicides to target woody species is particularly effective at reducing woody plant cover in shrub-encroached grassland ecosystems at the landscape scale, especially in sites that do not have the fine fuel base to carry fire (Tunnell et al. 2006; Pierce et al. 2019; Eddy et al. 2020). Three synthetic-auxin herbicides demonstrated to be effective at killing established woody species, especially those in the Fabaceae family, are a mix of aminopyralid and clopyralid (trade name: Sendero, Corteva Agriscience, Indianapolis, IN, U.S.A.) and triclopyr (trade name: Remedy Ultra, Corteva Agriscience, Indianapolis, IN, U.S.A.) (Hamilton 2004; Eddy et al. 2020). When applied in optimal conditions, these synthetic-auxin herbicides can act synergistically to produce higher rates of whole plant mortality than when applied alone (Bovey & Whisenant 1992; Ansley & Castellano 2006). This in turn can reduce the density and cover of established woody plant species and release desirable plants from competition, allowing them to recover (Radosevich & Bayer 1979; Hamilton 2004).
In addition to high treatment efficacy on established woody plants, these synthetic-auxin herbicides have relatively short half-lives (<70 days) in soil (Bukun et al. 2010; Sakaliene et al. 2011; Li et al. 2018). These short half-lives result in reduced soil residual activity over-time compared to herbicides with different modes of action, such as acetolactate-synthase inhibitors, and have been shown to have negligible adverse impacts on established graminoids (Ansley & Castellano 2006; Douglass et al. 2016; Eddy et al. 2020).
Despite the shorter half-lives and reduced soil residual activity of synthetic-auxin products, their soil residual activity can have unintended, adverse impacts on seedlings of some species of forbs and grasses if seeds germinate prior to herbicide breakdown (Wagner & Nelson 2014; Douglass et al. 2016; Anésio et al. 2018). This is particularly important as the application timing of these synthetic-auxin herbicides often coincides with the beginning of the growing season when target plants are actively growing and are the most receptive to herbicide (Hamilton 2004). This, however, is when seeds may be germinating and young plants are also most susceptible (Hamilton 2004; Wagner & Nelson 2014; Eddy et al. 2020). Thus, herbicide soil residual activity may pose an unintended barrier to desired plant emergence and establishment (Wagner & Nelson 2014; Douglass et al. 2016; Peterson et al. 2020).
Plant–herbicide interactions may also be influenced by soil properties such as texture, cation exchange capacity (CEC), and organic matter, as well as environmental conditions such as pulsed rain events (Hamilton 2004; Sakaliene et al. 2011; Hirsch et al. 2012). Aminopyralid, clopyralid, and triclopyr are moderately mobile in soil and can be degraded by soil microbes (Hamilton 2004; Bukun et al. 2010; Li et al. 2018). These herbicides leach more readily in coarse-textured soils than in fine-textured soils (Bukun et al. 2010; Sakaliene et al. 2011; Li et al. 2018). However, in soils with both fine texture and high organic matter, the herbicides may be quickly adsorbed to soil particles or degraded by microbes before they can be leached from the soil column (Smith & Aubin 1989; Bukun et al. 2010; Douglass et al. 2016).
Examining the adverse impacts of herbicides and their interactions with soil on emerging seedlings is crucial to understanding restoration success because active seeding is often coupled with herbicide application to meet ecological restoration goals in drylands (Barr et al. 2017; McManamen et al. 2018; Shackelford et al. 2021). Seeding efforts which have been found to be the most successful in arid and semiarid environments use seed mixes with a diverse assemblage of functional groups and species at high seeding rates (Barr et al. 2017; Shackelford et al. 2021), and are planned around near-term weather events and regional climate trends (Butterfield et al. 2017; Hardegree et al. 2018). When coupling herbicide application with active seeding, or relying on the in situ seed bank community, it is important to understand the potential risks herbicides may pose to seedling recruitment (Wagner & Nelson 2014; Douglass et al. 2016; McManamen et al. 2018).
In this study, we conducted a greenhouse experiment to elucidate the impacts of soil herbicide residues of a commonly used combination of aminopyralid, clopyralid, and triclopyr herbicide on dryland plant species' seed emergence and subsequent early plant growth. Within the framework of identifying herbicide influence on dryland seedling emergence and growth, we asked four primary questions: (1) How does soil residual herbicide activity affect the emergence and growth of dryland plant species? (2) Does herbicide residue influence differ between soil types? (3) Does a precipitation pulse immediately after herbicide application influence emergence and growth? And (4) does this vary by species? To assess this, we tested the application of this herbicide mixture on two different soil types to determine if the seedling response is affected by the abiotic conditions of soil properties and a post-herbicide-application water pulse (simulating a large rain event). We hypothesized that soil residual activity of the combination of aminopyralid, clopyralid, and triclopyr impacts forb and shrub species by reducing seedling emergence and decreasing seedling biomass, and that grasses will demonstrate a neutral response. In addition, we hypothesized that soil type interactions with soil residual herbicide will cause species-specific responses to herbicide impacts, and that flushing soil with water immediately after application will ameliorate the effect of herbicide on emergence and biomass.
Methods
Soil Collection and Preparation
In February 2021, we collected soils from two different locations in south central New Mexico, U.S.A. at the New Mexico State University Chihuahuan Desert Rangeland Research Center. Collection sites were chosen to represent two distinct ecological sites—gravelly, desert shrub (hereafter “Gravelly”) and sandy, desert shrub (hereafter “Sandy”)—with similar climatic conditions (Soil Survey Staff 2023). The Gravelly site (lat. 32.50917°N, long. −106.78873°W) is dominated by creosote bush (Larrea tridentata) and the dominant soil unit is classified as a loamy-skeletal, mixed, thermic Typic Calciorthid. The Sandy site (lat. 32.54436°N, long. −106.79518°W) is dominated by honey mesquite (Prosopis glandulosa) and the dominant soil unit is classified as a coarse-loamy, mixed, thermic Typic Haplargid. Bulk soil at both sites was collected from the upper 10 cm of the soil profile at random points along a transect at each site and homogenized. After field collection, soil was sieved through a 2 mm mesh to reduce soil aggregation and to remove gravel and debris. Soil property analyses (Table 1) were conducted by Ward Laboratories Inc. (Kearney, NE, U.S.A.) on three samples of each soil type.
| Soil property | Gravelly | Sandy |
|---|---|---|
| pH | 8.3 ± 0.0 | 8.2 ± 0.0 |
| Organic matter L.O.I (%) | 0.7 ± 0.0 | 0.5 ± 0.0 |
| Cation exchange capacity (CEC, me 100 g−1) | 19.1 ± 0.3 | 7.7 ± 0.3 |
| Nitrate-N (ppm N) | 16.5 ± 1.7 | 10.9 ± 0.7 |
| Olsen-P (ppm P) | 10.9 ± 0.5 | 11.7 ± 0.3 |
| Calcium (ppm Ca) | 3,553.0 ± 50.9 | 1,193.0 ± 50.3 |
| % Sand | 76.7 ± 0.7 | 82 ± 1.2 |
| % Silt | 6 ± 0.0 | 0.7 ± 0.7 |
| % Clay | 17.3 ± 0.7 | 17.3 ± 0.7 |
| USDA soil texture class | Sandy loam | Sandy loam |
| % Coarse fragments (>2 mm, by volume) | 22.9 ± 4.5 | 0 ± 0.0 |
Study Species
To account for different plant traits and functional groups, we used 10 species found in the northern Chihuahuan Desert, Omerick Ecoregion III (Table 2). The suite of species consists of two shrub species commonly targeted during woody plant control (P. glandulosa and L. tridentata) and eight species commonly used in dryland restoration and revegetation. The two shrub species were field collected by the authors in December 2020 from the areas where the soils were collected, P. glandulosa at the Sandy site and L. tridentata from the Gravelly site. Black grama (Bouteloua eriopoda), a native perennial grass species of particular focus for restoration in southwestern U.S. desert grasslands; and Lehmann lovegrass (Eragrostis lehmanniana), a non-native perennial grass planted for range improvement and reclamation but has become invasive in the desert southwest, seeds were purchased from Native American Seed and Granite Seed Company, respectively (Holechek & Herbel 1982; Buerdsell & Lehnhoff 2023). The remaining six species were field-collected at the time of maturity in the summer and fall of 2019 by the Bureau of Land Management Seeds of Success program (Haidet & Olwell 2015). These six field-collected species consisted of native perennial forb (desert marigold [Baileya multiradiata] and twinleaf senna [Senna bauhinioides]), native perennial grass (cane bluestem [Bothriochloa barbinodis] and Arizona cottontop [Digitaria californica]), and shrub (four-wing saltbush [Atriplex canescens] and soaptree yucca [Yucca elata]) species that are commonly used in ecological restoration seeding applications. All seeds were stored in a cool dry location prior to the start of the study. Plant taxonomy and functional groups throughout are based on the United States Department of Agriculture (USDA) PLANTS Database (USDA, NRCS 2023).
| Species scientific name | Species common name | Family | Growth habit |
|---|---|---|---|
| Atriplex canescens (Pursh) Nutt. | Four-wing saltbush | Chenopodiaceae | Shrub |
| Baileya multiradiata Harv. & Gray ex A. Gray | Desert marigold | Asteraceae | Perennial forb |
| Bothriochloa barbinodis (Lag.) Herter | Cane bluestem | Poaceae | Perennial graminoid |
| Bouteloua eriopoda (Torr.) Torr. | Black grama | Poaceae | Perennial graminoid |
| Digitaria californica (Benth.) Henrard | Arizona cottontop | Poaceae | Perennial graminoid |
| Eragrostis lehmanniana Nees | Lehmann lovegrass | Poaceae | Perennial graminoid |
| Larrea tridentata (DC.) Coville | Creosote bush | Zygophyllaceae | Shrub |
| Prosopis glandulosa Torr. | Honey mesquite | Fabaceae | Shrub |
| Senna bauhinioides (A. Gray) H.S. Irwin & Barneby | Twinleaf senna | Fabaceae | Perennial forb |
| Yucca elata Engelm. | Soaptree yucca | Agavaceae | Shrub |
Experimental Design
To explore how soil residual herbicide activity affects the emergence and growth of selected species, we conducted a greenhouse experiment at New Mexico State University in Las Cruces, New Mexico, U.S.A. We used a completely randomized design to compare two levels of herbicide (applied or not applied) in two field collected soil types (Gravelly or Sandy; Table 1), two levels of soil water inundation (soil not inundated with a pulse of water or soil inundated with a pulse of water after herbicide application), using 10 plant species (Table 2) and 12 replicates resulting in 960 total pots (2 soil × 2 herbicide × 2 water × 10 species × 12 replicates = 960 pots). Greenhouse conditions included temperatures ranging from 4 to 32°C and ambient light conditions. Cone-tainers (3.81 cm diameter, 13.97 cm depth; total volume of 107 mL [Greenhouse Megastore, West Sacramento, CA, U.S.A.]) were lined at the base with polyester quilt padding and filled with 100 mL of either Gravelly or Sandy soils.
The herbicide used was a mixture of Sendero (Corteva Agroscience, Indianapolis, IN, U.S.A.) and Remedy Ultra (Corteva Agroscience, Indianapolis, IN, U.S.A.), plus modified vegetable oil surfactant (Dyne-Amic, Helena Chemical Company, Collierville, TN, U.S.A.). Sendero, applied at 2.05 L/ha, consists of aminopyralid (122.95 g ae/ha) and clopyralid (565.55 g ae/ha). Remedy Ultra, applied at 0.59 L/ha, consists of triclopyr (283.1 g ae/ha). Dyne-Amic was applied at 0.29 L/ha. The application rates used were within the label rates and were based on recommended guidance from New Mexico State University for effective control of P. glandulosa (Duncan & McDaniel 2015; Young & Spackman 2021). The mixture was applied to the soil surface of half of the cone-tainers using a CO2-powered backpack sprayer at 25 psi. The other half of the cone-tainers were sprayed with the same volume of water. The total volume of the spray solution was equivalent to 37.4 L/ha.
The application of the herbicide or water (control) took place at the Leyendecker Plant Science Research Center, south of Las Cruces, New Mexico on a calm morning. The cone-tainers were placed outside on the ground in holding racks with their soil surface approximately 20 cm off the ground. Herbicide (or water) was then applied over the cone-tainers as the applicator with the backpack sprayer walked past the cone-tainers at 4.6 km/hour with the spray boom approximately 80 cm above the cone-tainers.
Within 24 hours after herbicide application, half of all soil-filled cone-tainers received an inundation treatment by adding 40 mL of water per cone-tainer, simulating a 3.5 cm pulsed rain event that saturated the entire soil column but did not result in gravitational flow. These cone-tainers were subsequently left to dry for a period of 1 week prior to seeding.
To help guide planting rates and quantify germination potential within the greenhouse experiment, germination testing of six of the ten seed lots was conducted prior to the start of the greenhouse experiment (Fig. S1). In addition to germination potential, to account for pot size, larger seeded-species such as soaptree yucca (Yucca elata) and P. glandulosa were planted at a rate of three seeds per individual cone-tainer, while all other species were planted at five seeds per cone-tainer. P. glandulosa required a pre-treatment to improve germination, thus all seeds were scarified by breaking the seed coat with a small incision. All seeds were planted at a shallow depth (~2 mm) that maintained seed–soil contact and likely would not insert the seed below the herbicide layer.
For the duration of the study, we watered plants with 8.33 mL cone-tainer-1 day-1 (0.73 cm cone-tainer-1 day-1) using tap water. Seedling emergence was monitored every other day and new emergence was recorded distinguishing between consecutive emergence in the same cone-tainer. To avoid competition, we thinned additional seedlings approximately one week after recording emergence of the first seedling in a given cone-tainer. Seeds were planted on 1 May, 2021, and plants were harvested on 29 and 30 June, 2021 (60 growing days). Total seedling biomass (root plus shoot) was accounted for by removing soil media from each harvested seedling's roots. Harvested biomass was dried at 60°C for 72 hours and then weighed using a four-digit scale. These metrics allowed for response variables of emergence (% of seedlings emerged per cone-tainer) and total (root plus shoot) seedling biomass (grams).
Statistical Analysis
Despite having good (42.5 ± 2.5% germination) preliminary germination potential (Fig. S1), desert marigold (B. multiradiata) had poor emergence in the greenhouse cone-tainers across all treatments (only 8 of 480 planted seeds emerged). Thus, B. multiradiata was removed from the analysis. The analyzed dataset contained information for nine species and 864 replicates (12 replicates per treatment condition). Even after removing B. multiradiata, emergence and total biomass data were highly zero-inflated, non-normal, and over-dispersed. We therefore adapted the zero inflated model for the analyses of these data.
To reduce model complexity quantifying the impact of the herbicide's soil residue, we measured the effect size of the soil residual herbicide activity by computing the difference (Δ: herbicide minus control) of both the emergence (%) and total biomass (g), of the herbicide and control replicates for each species, soil type, and soil inundation treatment (Rinella & James 2010; James et al. 2011). Whereby, effect sizes with a negative value indicate that the soil residual herbicide activity negatively affected the emergence or growth of seeds and seedlings, respectively, positive values indicate a beneficial effect of the soil residual herbicide, and values near zero indicate a neutral effect of the soil residual herbicide activity.
Analyses were conducted using R v4.2.2 (R Core Team 2022), with a user defined function for model fit coded in C++17. To analyze the effect sizes, it was necessary to adapt the zero-inflated model with the T-student distribution within the Bayesian statistics framework. To account for the experimental nature of the data, species, soil type, and the post-herbicide water pulse were included as treatment factors in the analysis, following the Bayesian analysis of variance technique (BANOVA; Gelman 2005). The resulting model was a robust alternative to explaining atypical values, which might be poorly explained by a normal distribution, and to model the observed excess of zeros.
To completely specify the model, prior distributions for the unknown parameters were defined as follows. The prior distributions for and are Inverse Gamma(0.01, 0.01) and Gamma(2, 0.1), respectively. The prior distribution for the degrees of freedom, , is the one proposed by Juarez and Steel (2010). For the other parameters in the model, we used the default prior distributions defined for the BANOVA approach in Gelman (2005). Graphical displays showing mean effect size estimates and 95% Bayesian credible intervals (CRIs) of the a posteriori distributions, and the BANOVA tables were computed using the last 25,000 iterations, taken every 10th, of a total of 100,000 iterations (Figs. S2 & S3).
Bayesian credible intervals have relatively simple interpretations, making them well suited for quantifying the impact of the soil residual herbicide activity on seedlings (Rinella & James 2010). If the mean of the effect size estimate and credible interval (CRI) does not overlap zero, then there is greater than a 95% chance that the herbicide had a non-neutral effect on either emergence or growth. We interpreted these effect sizes for each species' soil type × water treatment interactions and attribute effect sizes with CRI's that do not overlap zero to be significantly different from zero due to the impact of the soil residual herbicide activity. To interpret an individual species' soil type and water treatment combination interactions, when two credible intervals of the same species do not overlap, the probability is greater than 95% that the treatment with the most negative mean value interval experienced the largest negative effect of the herbicide residues.
Results
Emergence Rate Differences
The effect of soil herbicide residues on seedling emergence, measured as the effect size (the mean difference in emergence of herbicide minus control for a given species × soil type × water combination), was found to have a three-way interaction between species, soil type, and water treatments (Fig. S2). Soil residual herbicide activity had a significant negative effect (95% CRI < 0 in at least one soil type and water treatment group) on the emergence of three shrubs (Atriplex canescens, Larrea tridentata, and Yucca elata), the perennial forb (Senna bauhinioides), and the perennial grass (Digitaria californica; Figs. 1 & S4). Two species used in restoration seeding activities in the southwestern United States, the perennial grass D. californica and the shrub Y. elata, experienced significant reductions in seedling emergence in all four treatment combinations for the respective species. Reductions in mean emergence for D. californica ranged from 15 to 51% (95% CRI does not overlap zero). Reductions in mean emergence for Y. elata ranged from 26 to 59% (95% CRI does not overlap zero). The only species × soil type × water treatment combination where the water pulse significantly lessened the impact of the herbicide residual activity on emergence was for D. californica in the Gravelly soil. The perennial grass cane bluestem (Bothriochloa barbinodis) had significantly increased emergence in soil that received an herbicide application, but only in the Gravelly soil type that received a pulse of water following herbicide application (mean of 18% increased emergence, 95% CRI does not overlap zero).
Total Biomass Differences
The effect of soil herbicide residue on seedling biomass, measured as the effect size (the mean difference in total seedling biomass of herbicide minus control for a given species × soil type × water combination), was found to have a three-way interaction between species, soil type, and water treatments (Fig. S3). Soil residual herbicide activity had a significant negative effect (95% CRI < 0 in at least one soil type and water treatment group) on the early growth of the four shrubs (A. canescens, L. tridentata, Prospips glandulosa, and Y. elata), the perennial forb (S. bauhinioides), and the perennial grass (D. californica; Figs. 2 & S5). Significant reductions in mean total biomass across all four treatment combinations occurred only for D. californica and ranged from 0.02 to 0.05 g (95% CRI does not overlap zero), which equated to approximately 50% reductions in seedling biomass in each soil type and water treatment combinations (Fig. S5). The only species × soil type × water treatment combination where the water pulse significantly lessened the impact of the herbicide residual activity on total biomass was for Y. elata in the Gravelly soil.
Discussion
The application of synthetic-auxin herbicides to control woody plant encroachment is a commonly used and effective ecological restoration tool (Hamilton 2004; Tunnell et al. 2006; Eddy et al. 2020). A range of nontarget impacts on seed emergence and young seedling growth due to soil residual synthetic-auxin herbicide activity has been shown, from negative impacts to no observed impacts (Wagner & Nelson 2014; Douglass et al. 2016; Tucker et al. 2018). However, abiotic conditions such as soil type or post-herbicide-application water pulses (Hirsch et al. 2012) can interact with herbicide application and species to influence these outcomes. Studies such as this can identify potential unintended barriers to ecological restoration. In this greenhouse study, we observed a range of impacts—positive, neutral, and negative—from the soil residual herbicide activity that depended on species, soil type, and post-herbicide water pulse. However, the most prevalent impacts observed on seedling vitality were negative, potentially resulting in an unintended barrier to seed-based restoration in drylands.
In considering herbicide impacts on seed emergence and early seedling growth in the greenhouse, both commonly targeted shrub species (Prosopis glandulosa and Larrea tridentata) and those common in dryland restoration (Atriplex canescens, Digitaria californica, Senna bauhinioides, and Yucca elata) were found to be generally negatively affected by soil residual herbicide activity of the standard mixture of aminopyralid, clopyralid, and triclopyr. These findings agree with other studies who have found that soil residual activity of synthetic-auxin herbicides can alter nontarget graminoid, forb, and shrub species vitality at the seed and seedling stages (Wagner & Nelson 2014; Douglass et al. 2016; McManamen et al. 2018).
Emergence
While interactions were present, soil residual herbicide activity had an overall negative effect on the emergence of a subset of our seeded species in this greenhouse study. Our hypothesis that synthetic-auxin soil residual herbicide activity would have a negative impact on emergence of shrubs and perennial forbs was supported, as these taxa (A. canescens, L. tridentata, S. bauhinioides, and Y. elata) displayed reduced emergence in at least one of the soil × water treatment combinations that received an herbicide application. Y. elata was also particularly sensitive to soil herbicide residues and was negatively affected across all treatment combinations. Three of the four grass species (Bothriochloa barbinodis, Bouteloua eriopoda, and Eragrostis lehmanniana) were not negatively impacted by the soil residual herbicide; however, one perennial grass (D. californica) displayed significantly reduced emergence in all soil type and water treatment combinations, indicating that D. californica may be particularly sensitive to soil residual herbicide activity.
When looking at the ambient water treatment only, our soil type treatment appears to have caused species-specific responses to the herbicide's effect on emergence. For example, A. canescens and L. tridentata emergence was not negatively affected by herbicide in the Gravelly soil type but was negatively affected in the Sandy soil type. While we did not test this, potential mechanisms for the lack of negative soil residual herbicide impacts in the Gravelly soil may be due to increased microbial degradation (Smith & Aubin 1989), sorption to organic matter or silt (Bukun et al. 2010), or photodegradation on the soil surface following herbicide application (Radosevich & Bayer 1979; Newton et al. 1990; Eyheraguibel et al. 2009) prior to seeding and seedling emergence (Douglass et al. 2016). Further studies on the microbial relationship across these different variables could further elucidate the mechanisms behind seedling emergence outcomes.
Contrary to our expectations, a pulse of 40 mL of water (simulating a 3.5 cm rain event) immediately after herbicide application did not consistently ameliorate negative herbicide impacts on emergence. Rather, the pulse of water may have more evenly distributed the herbicide throughout the soil column to saturate the soil but not initiate gravitation flow and leaching of the herbicide. Whereas in the ambient treatment without an additional flush of water, the herbicide likely remained more concentrated on the soil surface where seeds were planted. The inconsistent interactions with soil types and water pulse may be due to the Gravelly soil's greater capacity to sorb the herbicides (higher CEC, organic matter, and clay) throughout the upper portion of soil column following the pulse of water and remain at toxic levels. Conversely, the herbicide may have been dissipated enough in the upper portion of the Sandy soil column following the pulse of water to not hinder emergence of species moderately sensitive to the herbicide (Bukun et al. 2010; Sakaliene et al. 2011; Douglass et al. 2016). Additional research of the finer-scale movements of herbicides at the sub-centimeter scale in the soil column, microbial activity between soil types, as well as a more comprehensive herbicide sensitivity assay for these species would be needed to understand these interactions more thoroughly.
Total Biomass
The herbicide mix of aminopyralid, clopyralid, and triclopyr and its soil residual herbicide activity had an overall negative impact on the total biomass of plant seedlings grown for 60 days in the greenhouse. Our hypothesis that synthetic-auxin soil residual herbicide activity would have a negative impact on total biomass of shrubs and perennial forbs was supported, as all the taxa in these groups (A. canescens, L. tridentata, P. glandulosa, S. bauhinioides, and Y. elata) displayed reduced total biomass in at least one of the soil × water treatment combinations that received an herbicide application. Of the four grass species in the study, three were not negatively affected (B. barbinoides, B. eriopoda, and E. lehmanniana) by the soil residual herbicide; however, D. californica displayed reduced total biomass, equating to 50% reductions in mean total biomass, in each of the soil type and water treatment combinations.
When looking at the ambient water treatment only, herbicide and soil type interacted with seedling biomass at the species level, causing species-specific responses on total biomass. For example, the total seedling biomass of A. canescens and L. tridentata were not negatively affected by herbicide in the Gravelly soil type but were negatively affected in the Sandy soil type. The lack of effect in the Gravelly soil as was considered for emergence rates may also be due to increased microbial degradation (Smith & Aubin 1989), sorption to organic matter or silt (Bukun et al. 2010), or photodegradation on the soil surface following herbicide application (Radosevich & Bayer 1979; Newton et al. 1990; Eyheraguibel et al. 2009) prior to emergence and initial growth. Alternatively, those two species may exhibit moderate sensitivity to the soil residual herbicide activity. By contrast, Y. elata biomass was not negatively affected in the Sandy soil type but was significantly reduced in the Gravelly soil type. While there is much to be understood about this relationship, the inverse interaction displayed across the soil types by Y. elata may be related to it contractile root structure. Contractile roots have relatively high root permeability near the base of the shoot and pull the plants down further into the soil to protect them from unfavorable aboveground abiotic conditions (North et al. 2008; Prado-Tarango et al. 2019). Thus, the herbicide may have degraded for reasons explained above and may have been at low enough concentration in the Sandy soil to not negatively impact biomass. Whereas in the Gravelly soil, there may have been higher concentrations of herbicides across depths where both the roots and shoots encountered toxic levels of herbicide (Bovey & Richardson 1991). We, however, did not test this directly and suggest that further research addressing herbicide interactions between plant functional traits, such as rooting depths, and abiotic drivers, such as soil type, would be illustrative.
As with our emergence results, contrary to our expectations, a pulse of 40 mL of water immediately after herbicide application did not consistently ameliorate negative herbicide impacts on total biomass. Rather, the pulse of water tended to exacerbate the negative impact of the herbicide, depending on the species and soil type. This was particularly true for P. glandulosa and L. tridentata, species which produce a long vertical taproot, as well as Y. elata which produces both a vertical taproot and a horizontal adventitious root. The pulse of water likely distributed the herbicide throughout the soil column of both the Gravelly and Sandy soil types, as the pulse was enough water to, in theory, fully saturate the soil but not reach gravitations flow and leach the herbicide from the soil column. This is supported by Sakaliene et al. (2011) and Bovey and Richardson (1991) who found clopyralid concentrations remained at toxic concentrations in the upper 5 and 15 cm of soils, respectively, across a range of moisture conditions more than 15 days post-herbicide application. Thus, it is likely that the seeds were able to germinate, emerge, and grow, but once the roots intercepted the layer of soil residual herbicide, subsequent growth was negatively affected.
Although herbicide use is commonly implemented and an important restoration tool in many ecosystems (Hamilton 2004), our greenhouse study results demonstrate the potential for nontarget species level impacts. These nontarget impacts may guide land managers to consider how the use of this common mixture of synthetic auxin herbicides may affect seeds and seedling recruitment of shrub species that are targeted by herbicide application and desired species either actively planted through propagule additions, or those already present in the soil seed bank.
It should be noted that herbicide application in this experiment was similar to aerial application from a fixed-wing aircraft, which is a nontargeted broadcast strategy. The nontarget impacts demonstrated in this experiment could be ameliorated with more precise herbicide applications such as spot spraying or cut stump methods. This study also emphasizes the importance of the abiotic conditions surrounding herbicide application—soil type and rainfall pulses following application—as they were found to interact with the effects of the soil residual herbicide activity, again depending on the species.
There are many possible strategies to mitigate potential nontarget impacts for practitioners using this mixture of synthetic auxin herbicides for restoration and land management. These strategies could consist of changing the herbicide application timing (dormant season vs. growing season) or the method of the herbicide application (individual plant application compared to broadcast application) if practicable; adapting seeds of desired species to resist effects of soil residual herbicide (e.g. potentially using an activated charcoal seed coat to neutralize the herbicide) (Terry et al. 2021); or altering seed mix composition to favor species less sensitive to soil residues of these herbicides. For cases in which these strategies are not practical, or the effects of soil residual herbicide are not known for key species or environmental conditions, delaying planting of seed mixes until after the herbicide concentrations have likely dissipated below phytotoxic levels would be the most robust approach across a variety of restoration contexts. Conducting soil tests to measure the concentration of soil herbicide residues could supplement this approach for time-sensitive projects with the appropriate resources to do so. These considerations are inherently context dependent on the ecosystem being operated in, and strategies used to reduce the likelihood of posing unintended barriers to ecological restoration should also reflect this.
Acknowledgments
We would like to thank E. Morris for his assistance with applying the herbicide to the cone-tainers used in this study. We also would like to thank undergraduates and technicians S. Lasché, M. Barraza, G. Morrill, M. Wright, D. Burton, W. Biakeddy, and M. Mason, for their help in the greenhouse. We thank the comments, suggestions, and edits of four anonymous reviewers whose critiques have improved the manuscript. Funding for this research was provided by the USDA NRCS, grant number NR188C30XXXXC003, and the USDOI BLM, funding agreement L19AC00300.
