Management Implications
Dittrichia graveolens (stinkwort) is an emerging invader in California, characterized as a species of concern because of its high invasive potential and high potential impact. Currently, D. graveolens is most often found along roadsides and other disturbed areas in western North America, but the nature of this association is unresolved. Our research shows that this species is strongly limited by competition from other plants, especially by belowground competition. While D. graveolens can establish when plant competition is limited by disturbance, it struggles to persist in dense, competitive vegetation and therefore finds refuge along roadsides and in disturbed sites where competition is low.
This means managers can best reduce the risk of D. graveolens invasion by promoting robust plant cover in vulnerable areas, especially trails, construction zones, or (where feasible) roads. Practices that reduce competition, especially belowground disturbances that expose soil, are likely to increase vulnerability to invasion and should be timed or managed carefully.
Given that the wind-dispersed seeds of D. graveolens can travel far into established vegetation, early detection and spot treatments of outbreaks in disturbed patches are particularly important. The control of roadside populations as seed sources, combined with strengthening plant communities to enhance competition, offers a proactive strategy that could reduce the need for repeated control efforts over time.
Introduction
Disturbance is a critical process in many ecosystems, providing heterogeneity and influencing diversity at both the patch level and landscape scale (Greipsson Reference Greipsson2011; Hobbs and Huenneke Reference Hobbs and Huenneke1992). Disturbance reduces competition and frees up space and resources that colonizing species can exploit (Catford et al. Reference Catford, Daehler, Murphy, Sheppard, Hardesty, Westcott, Rejmánek, Bellingham, Pergl, Horvitz and Hulme2012; Hobbs and Huenneke Reference Hobbs and Huenneke1992). Thus, disturbance also plays an important role in the invasion process for many non-native species (Catford et al. Reference Catford, Daehler, Murphy, Sheppard, Hardesty, Westcott, Rejmánek, Bellingham, Pergl, Horvitz and Hulme2012; Hobbs and Huenneke Reference Hobbs and Huenneke1992; Minchinton and Bertness Reference Minchinton and Bertness2003; Pimm Reference Pimm, Drake, Mooney, DiCastri, Groves, Kruger and Williamson1989). Roadsides, transportation corridors, and other disturbed environments facilitate primary spread of non-native species (Baker Reference Baker1974; McDougall et al. Reference McDougall, Lembrechts, Rew, Haider, Cavieres, Kueffer, Milbau, Naylor, Nuñez, Pauchard, Seipel, Speziale, Wright and Alexander2018; Tyser and Worley Reference Tyser and Worley1992). The movement of species away from roadside edges and into established vegetation (secondary spread) is a critical phase in the invasion process (Ward et al. Reference Ward, Taylor, Dixon Hamil, Riitters and Fei2020).
Roadside soil is often disturbed from initial construction and ongoing road maintenance, characterized by distinct texture and mineral composition, increased compaction, and erosion (Lázaro-Lobo and Ervin Reference Lázaro-Lobo and Ervin2019; Mills et al. Reference Mills, Mamo, Schacht, Abagandura and Blanco-Canqui2020). Road runoff increases soil salinity and can contain contaminants from vehicles (Lázaro-Lobo and Ervin Reference Lázaro-Lobo and Ervin2019; Trombulak and Frissell Reference Trombulak and Frissell2000), and roadsides often have lower nutrient availability (Liu et al. Reference Liu, Yang, Ji, Dong, Li, Wang, Han and Chen2021). Due to maintenance regimes and physical impacts from vehicular traffic, roadsides have high disturbance rates and lower plant cover (Christen and Matlack Reference Christen and Matlack2006).
A classic question in ecology is whether species distributions reflect an affinity for certain environmental conditions or competitive exclusion from higher-quality habitats. Some plant species are more successful in specific abiotic conditions (Gioria et al. Reference Gioria, Hulme, Richardson and Pyšek2023; Wamelink et al. Reference Wamelink, van Dobben, Goedhart and Jones-Walters2018), leading to strong patterns of association with these soils or environmental conditions. In contrast, species may be widely distributed, but biotic interactions with more competitive plant species, herbivores, or pathogens may limit their ranges (Hutchinson Reference Hutchinson1959; Gioria et al. Reference Gioria, Hulme, Richardson and Pyšek2023). Understanding the nature of an association with roadsides is crucial for effective management. If a species has an affinity for disturbed soils, then land managers should focus early detection on those areas. If biotic interactions, such as escape from competitive pressure, limit the distribution of a species, then land managers can focus on reducing disturbance and promoting competition.
We investigated abiotic soil conditions and reduced plant competition (resulting from disturbance) as two potential factors causing stinkwort [Dittrichia graveolens (L.) Greuter] (Figure 1), to grow along roadsides. This annual herb was introduced to California in the early 1980s and was originally found in disturbed areas along railroad tracks and roads in the County of Santa Clara (Brownsey et al. Reference Brownsey, Kyser and DiTomaso2013; Preston Reference Preston1997). Native to the Mediterranean basin in Europe, D. graveolens is often found in bare, disturbed habitats, including roadsides, agricultural lands, gravel riparian areas, and ruderal zones associated with annual or biennial weeds (Brullo and de Marco Reference Brullo and de Marco2000; Rameau et al. Reference Rameau, Mansion, Dumé and Gauberville2008; Šajna et al., Reference Šajna, Adamlje and Kaligarič2017). Since its initial detection in 1984, D. graveolens has spread across California, now occupying more than 83% of counties, with a range extending >400 km north, >200 km east, and >690 km south of its original point of introduction (Calflora 2024; Lustenhouwer and Parker Reference Lustenhouwer and Parker2022).
Figure 1. Growth stages of Dittrichia graveolens showing (A) seeds, (B) newly germinated seedling, (C) seedling at the time of field transplanting, (D) juvenile starting to bolt in the wild, (E) non-reproductive adult growing in mowed grassland, and (F) flowering adult.
In California, D. graveolens has been observed spreading into wildlands and rangelands away from roads (Brownsey et al. Reference Brownsey, Kyser and DiTomaso2013; Melen et al. Reference Melen, Snyder, Fernandez, Lopez, Lustenhouwer and Parker2024). These observations highlight the spread potential of D. graveolens and the invasion risk this species poses. However, little is known about this invasion process and what conditions enable D. graveolens to spread away from roadsides into more intact plant communities. Here we studied whether D. graveolens populations preferentially grow along roadsides because of beneficial soil conditions or if they are limited to growing along roadsides due to competition. We characterized germination and growth responses to engineered fill and field topsoil in lab and greenhouse experiments. Then, in four separate field experiments over 2 yr, we manipulated competition, disturbance, and shading in an established grassland to determine the effects of above- and belowground competition on growth and survival to reproduction.
Materials and Methods
Study Sites
The County of Santa Clara (37.36°N, 121.97°W) is located at the southern end of the San Francisco Bay in California, USA. The county encompasses the Santa Clara Valley, bordered by the Diablo Mountain Range to the east, the Santa Cruz Mountains to the southwest, and San Francisco Bay to the northwest. The valley experiences a mild Mediterranean climate with warm, dry weather much of the year (Grossinger et al. Reference Grossinger, Striplen, Askevold, Brewster and Beller2007) due to its proximity to the moderating effects of the San Francisco Bay. The rainy season is predominantly from November to April with 375 ± 125 mm SD of annual precipitation (McKee et al. Reference McKee, Leatherbarrow, Pearce and Davis2003), and the average daily mean temperature in the San Jose region ranges from 27.9 C to below freezing (Hanson et al. Reference Hanson, Li and Faunt2004).
Soil Sources
We compared the response of both seeds and plants to field topsoil and to engineered fill as a proxy for the physical microenvironment of roadside soils. Field topsoil was collected from a woodland site, and engineered fill was taken from a nearby construction project at University of California Santa Cruz (hereafter UC Santa Cruz); several cubic tons of these soils were in storage at a central campus location, and we subsampled several buckets of each soil type for our experiments (Figure 2).
Figure 2. Soils used in germination and greenhouse experiments were collected from a central soil storage location at UC Santa Cruz. Field topsoil was collected from a woodland site on campus, and engineered fill was taken from a campus construction project. The soils were exposed to outdoor conditions, which allowed microbial communities to persist. We subsampled each soil type and homogenized the collected soil for our experiments. Samples were sent to UC Davis Analytical Laboratory, Davis, CA, for analysis.
Seed Germination in Roadside and Field Soil
In September and October 2020, we collected D. graveolens seeds from 16 populations in the County of Santa Clara. These sites represent a mix of roadside and vegetated habitat (Supplementary Table S1 in Melen et al. Reference Melen, Snyder, Fernandez, Lopez, Lustenhouwer and Parker2024). For each population, we collected from ⪆10 individuals at 3-m intervals along a randomly placed transect and combined seeds into a single sample.
In June and July 2021, we compared germination behavior of seeds on the two substrates, placing 10 seeds in each of 80 petri dishes for each substrate (engineered fill and field topsoil). We visually inspected each seed beforehand to ensure we used only fully developed seeds. Petri dishes were sealed with Parafilm M™ (Amcor, Thurgauerstrasse 34CH-8050, Zürich, Switzerland) and placed in a randomized block design in an incubation chamber with a 16-h day and a 23/19 C day/night temperature cycle. We scored germination each day until no further germination was observed, then for 7 more days (engineered fill = 12 d, field topsoil = 11 d). When scoring for germination, we looked for the first emergence of the root radicle or the cotyledon and removed any germinated seeds. We kept soil moist by misting with deionized water.
We used R v. 4.3.1 (2023-06-16; R Core Team 2023) for all statistical analyses. Our general approach for each response variable was to run mixed-effects models with, at a minimum, fixed effects for soil and block as a random effect.
We compared the germination rate on two substrates (engineered fill and field topsoil) using a mixed-effects Cox proportional hazards model (coxme and survival packages; Therneau Reference Therneau2022, Reference Therneau2023), with soil type as a fixed effect, and dish (N = 5 replicates) nested within population (N = 16) as random effects. We evaluated the main effect of soil type using a type II partial likelihood ratio test (car package; Fox and Weisberg Reference Fox and Weisberg2019).
Plant Growth Response to Disturbed Soil
We assessed the response of D. graveolens to competition and abiotic soil conditions in the same two substrates with three competition treatments: D. graveolens grown alone, with soft brome (Bromus hordeaceus L.), or with Italian ryegrass (Lolium perenne L.). These non-native European annual grasses were selected because they are commonly found in California’s annual grasslands (Dawson et al. Reference Dawson, Veblen and Young2007; HilleRisLambers et al. Reference HilleRisLambers, Yelenik, Colman and Levine2010; Seabloom et al. Reference Seabloom, Harpole, Reichman and Tilman2003), including at our field site described later (Melen et al. Reference Melen, Snyder, Fernandez, Lopez, Lustenhouwer and Parker2024). We collected B. hordeaceus seeds from Blue Oak Ranch Reserve (37.38°N, 121.74°W) and L. perenne seeds from Younger Lagoon Reserve (36.96°N, 122.07°W) on the UC Santa Cruz Coastal Science Campus.
We germinated D. graveolens seeds in the conditions described earlier (see “Seed Germination in Roadside and Field Soil”). We germinated grasses in potting mix trays under fluorescent light banks with 16-h days. We filled D16 Deepots (Stuewe & Sons, Inc., Tangent, OR, 97389; 5-cm diameter, 18-cm height) with engineered fill or field topsoil and then transplanted seedlings in sets of three (one for each treatment) after radicles and cotyledons emerged. We randomized pots into a block design with each block consisting of 2 D. graveolens seedlings from each of the 16 seed-source sites for each of the 3 competition treatments, N = 96 per block by 8 blocks (768 total). We grew plants in a greenhouse for 4 mo before harvesting. We clipped D. graveolens aboveground biomass at the crown and dried it in a 60 C oven for 3 d before weighing.
We fitted a generalized linear model (GLM) to analyze the effect of competition on D. graveolens biomass (car package; Fox and Weisberg Reference Fox and Weisberg2019). The model used a gamma distribution with a log link function; fixed effects were competition and soil. We initially used population and block as random effects (lme4 package; Bates et al. Reference Bates, Maechler, Bolker and Walker2015), but both prevented model convergence. Because Akaike information criterion scores showed that the random effects did not contribute importantly to the model, they were not included. We evaluated the main and interaction effects using type II likelihood ratio tests (car package; Fox and Weisberg Reference Fox and Weisberg2019). We conducted post hoc pairwise comparisons using estimated marginal means, comparing them using Welch t statistics with Satterthwaite degrees of freedom and a Bonferroni adjustment for multiple comparisons (emmeans package; Lenth Reference Lenth2024).
Field Experiments Year 1: Response to Competition and Disturbance
To assess the response of D. graveolens to competition in a field setting, we conducted an experiment at Blue Oak Ranch Reserve, part of the University of California Natural Reserve System. Blue Oak Ranch Reserve is located on the western slopes of Mount Hamilton in the Diablo Range, just east of San Jose, CA, USA (37.381914°N, 121.736264°W). The reserve is a former rangeland, representing a key habitat type threatened by the invasion of D. graveolens. We established a 10 by 26 m fenced field site in a non-native grassland with a mixture of annual grasses and forbs. A subset of treatments from this experiment, combined with a separate study focused on local adaptation among seed sources, was the focus of an earlier paper (Melen et al. Reference Melen, Snyder, Fernandez, Lopez, Lustenhouwer and Parker2024). Here, we focus on the response to competition and disturbance.
We quantified D. graveolens’ response to four competition treatments: (1) control, which was the grassland including the year’s plant growth as well as the previous year’s thatch; (2) thatch removal, which involved raking and removing the previous year’s thatch; (3) aboveground removal, where we used a string trimmer to trim grassland vegetation to 8 to 13 cm above the ground; and (4) and above+below removal, where we tilled the soil to completely remove above- and belowground biomass.
In January 2021, we germinated D. graveolens seeds in the conditions described earlier and transplanted them into D16 Deepots (5-cm diameter, 18-cm height) with field topsoil collected from Blue Oak Ranch Reserve in December 2020. We grew the seedlings in the greenhouse for about 8 wk until the first true leaves had emerged and lengthened for all plants. We did not directly seed D. graveolens into the field site due to biosafety concerns about this noxious weed.
We used a randomized block design with 10 blocks of 1.5-m2 plots (Supplementary Figure S1). From February 27 to March 24, 2021, we planted 16 D. graveolens seedlings into each plot using dibblers (640 seedlings; 40 plots total). Seedlings were planted in a 4 by 4 grid centered on the plot. The distance between plants within plots was 33 cm, and plots were separated from each other by a 25-cm buffer. We surveyed plants weekly throughout the experiment to assess D. graveolens’ survival, and during the first month we also replaced any dead plants. We assessed bud development as a key phenology stage for terminating plants to ensure no seeds were released into the site. Weekly plant surveys continued until all plants had either produced buds or perished.
We measured height, and dried and weighed aboveground biomass after harvest. Height and biomass were correlated (r = 0.58, N = 213), and results for the response variables were similar. Therefore, we present only the results for final aboveground biomass as our measure of performance and proxy for reproductive output.
For all statistical models, competition treatment was the fixed effect and population and block were included as random effects. Here we describe the structures of the models.
We analyzed survival in two ways. First, the probability of surviving to reproduction (budding) was compared across competition treatments with a GLMM using a binomial family with a logit link function (lme4 package; Bates et al. Reference Bates, Maechler, Bolker and Walker2015). We evaluated the main effect of competition treatment using a type II Wald chi-square test (car package; Fox and Weisberg Reference Fox and Weisberg2019). Second, time to death was analyzed using a mixed-effects Cox proportional hazards model (coxme and survival packages; Therneau Reference Therneau2022, Reference Therneau2023). We evaluated the main effect of competition treatment using likelihood ratio tests (car package; Fox and Weisberg Reference Fox and Weisberg2019).
We analyzed biomass at reproduction using a linear mixed-effects model (lme4 package; Bates et al. Reference Bates, Maechler, Bolker and Walker2015). The significance of the competition treatment was assessed using a type III Wald F-test with Kenward-Roger degrees of freedom (car package; Fox and Weisberg Reference Fox and Weisberg2019). We did post hoc comparisons using the differences among estimated marginal means (emmeans package; Lenth Reference Lenth2024), using the Bonferroni method.
Field Experiments Year 2: Response to Disturbance—Aboveground versus Belowground Competition
In January 2022, we germinated D. graveolens seeds in the conditions described earlier and transplanted them into 10.16-cm height by 8.89-cm width injection-molded pots with potting media (ProMix® HP® BioFungicide™ + Mycorrhizae™; Premier Tech, 1, avenue Premier, Rivière-du-Loup, QC, G5R 6C1 Canada), where they grew for about 8 wk until the first true leaves had emerged and lengthened for all plants. We conducted three experiments related to above- and belowground competition using a subset of the same 1.5-m2 plots as the previous year (Supplementary Figure S2).
Aboveground and Belowground Competition
In Year 2 we replicated our test of the relative effect of above- and belowground competition on D. graveolens, with 54 planting locations randomly assigned to three treatments: (1) control, which was the grassland including the year’s plant growth as well as the previous year’s thatch; (2) aboveground removal, where we clipped grassland vegetation to 1- to 3-cm high; and (3) above+below removal, where we dug holes to a depth of 45 cm to completely remove above- and belowground biomass. We did not include a thatch removal experiment in Year 2. We planted two D. graveolens seedlings in each planting location (N = 108) and maintained clipping treatments weekly, recording survival for 9 wk. Because no plants survived to flower (see “Results and Discussion”), we did not measure biomass.
We analyzed seedling survival using a mixed-effects Cox proportional hazards model (coxme and survival packages; Therneau Reference Therneau2022, Reference Therneau2023), with treatment (control, clipping, belowground competitor removal) as a fixed effect, and a random effect of plot. We assessed the significance of the treatment using type II likelihood ratio tests (car package; Fox and Weisberg Reference Fox and Weisberg2019).
Aboveground Shading
To separate the effects of shading from other aboveground interactions, we used an artificial shading experiment with 11 plots, which were each hoed to remove above- and belowground competition. The plots were divided into four quadrants; four bamboo stakes were placed in the corners of each quadrant and were randomly assigned a treatment of control (no shade cloth) or shade cloth (GCI Landscaper’s Choice Premium 5-ounce Woven Landscape Fabric 500 Series; Ground Cover Industries, Inc., Santa Rosa Beach, FL 32459). Shade cloth was attached to the bamboo stakes and maintained for the duration of the experiment. On March 17, 2022, we planted two D. graveolens seedlings per quadrant (N = 88 plants total, 4 seedlings per treatment per plot). We weeded the plots weekly and recorded D. graveolens’ survival for 9 wk.
We analyzed seedling survival using a mixed-effects Cox proportional hazards model (coxme and survival packages; Therneau Reference Therneau2022, Reference Therneau2023), with treatment (no shade and shade) as a fixed effect, and a random effect of plot. We assessed the significance of the shading treatment using type II likelihood ratio tests (car package; Fox and Weisberg Reference Fox and Weisberg2019).
Mechanisms of Belowground Competition
To investigate the mechanisms involved in belowground interactions, we used a two-factor factorial design with trenching and resource addition treatments in 28 grassland plots. In each plot, we dug six holes, each 15 cm in diameter and 45-cm deep; three were lined with weed cloth fabric (GCI Landscaper’s Choice Premium 5oz Woven Landscape Fabric 500 Series), and the other three were left unlined. We placed PVC collars, each 15 cm in diameter and 12-cm deep, into all holes, positioning them with a 2-cm lip above the soil surface to prevent runoff and inserting them to a depth of 10 cm. The original soil was then replaced. We planted two D. graveolens seedlings in each treatment site. To release plants from the effects of belowground competition for water and nutrients, we used three resource addition treatments: control, the addition of 283.49 g of water, and the addition of 283.49 g of water plus 5 g of fertilizer (Osmocote 14-14-14; ICL, Tel Aviv, 6107025, Israel). The soil surface was scratched using a fork in all treatments to encourage infiltration. We maintained treatments weekly and recorded D. graveolens’ survival for 9 wk.
We analyzed seedling survival using a mixed-effects Cox proportional hazards model (Coxme and Survival packages; Therneau Reference Therneau2022, Reference Therneau2023), with treatment (control, water, water + nutrients) and competition (weed cloth, no weed cloth) as fixed effects, and a random effect of plot. We assessed the significance of the interaction and main effects by comparing models using likelihood ratio tests.
Results and Discussion
Seed Germination in Roadside and Field Soil
The proportion of germinated seeds was high in both soil types (engineered fill = 81%, field topsoil = 84%), with seeds showing a slightly higher chance of germinating in field topsoil than in engineered fill (5% higher; relative risk of 1.05 ± 0.06 SE;
$\chi $
2
1 = 105.57, P < 0.001; Figure 3).
Figure 3. Cumulative proportion germinating per day of Dittrichia graveolens seeds germinated on engineered fill (gray circles) and field topsoil (green triangles). Values shown are means ± 1 SE, showing variance across 16 seed sources (sites).
Plant Growth Response to Disturbed Soil
Dittrichia graveolens grown in competition with non-native annual grasses were significantly smaller than D. graveolens growing alone (
$\chi$
2
2 = 48.03, P < 0.001), and D. graveolens had greatly reduced growth in engineered fill (
$\chi$
2
1 = 431.87, P < 0.001; Figure 4). In addition, we saw a significant interaction between competition and soil (
$\chi$
2
2 = 470.32, P < 0.001). When grown with B. hordeaceus, competition reduced D. graveolens biomass by 14-fold in field topsoil (
$\bar X$
alone/
$\bar X$
competitor = 14.3, t = 25.47, P < 0.001) compared with only a 77% reduction in biomass in engineered fill (
$\bar X$
alone/
$\bar X$
competitor = 1.77, t = 5.45, P < 0.001). Similarly, competition with L. perenne caused a 55-fold reduction in D. graveolens biomass in field topsoil (
$\bar X$
alone/
$\bar X$
competitor = 54.6, t = 38.16, P <0.001) and only a 91% reduction in engineered fill (
$\bar X$
alone/
$\bar X$
competitor = 1.91, t = 6.18, P < 0.001).
Figure 4. Biomass (g) of Dittrichia graveolens grown in a greenhouse experiment alone or with each of two grass competitors (Bromus hordeaceus and Lolium perenne), planted into field topsoil (green) or engineered fill (gray). Boxes correspond to the median, first, and third quartiles, and whiskers extend to the furthest value within 1.5× the inter-quartile range. Note the log scale.
Field Experiments Year 1: Response to Competition and Disturbance
Overall survival to reproduction was strongly affected by treatment (
$X$
2
3 = 136.01, P < 0.001), with about a 50% absolute increase in overall survival in the above+below removal treatment compared with the control. Surprisingly, survival in the thatch removal treatment was 15 percentage points lower than in the control (Figure 5A). Likewise, survival analysis showed variation in the timing of mortality, with an 84% reduction in mortality risk in the above+below removal treatment (
$X$
2
3 = 200.1, P < 0.001; Figure 5B).
Figure 5. (A) Proportion of Dittrichia graveolens individuals that survived to reproduction by treatment in Year 1 of the field experiment. Plants were transplanted as seedlings into control plots with undisturbed grassland, plots from which dry thatch was removed, plots where aboveground biomass was clipped, and plots where both above- and belowground biomass of all plant neighbors was removed. (B) Survival probability for D. graveolens over time; plants were censored from the analysis if they began to flower (and were harvested). Treatments indicated by color as in A. Kaplan-Meier survival curves with 95% confidence intervals reflect fixed effects (Treatment) only.
Aboveground biomass was also significantly different across the treatments (F(3, 203.6) = 154.57, P < 0.001; Figure 6). The post hoc pairwise comparison revealed that plants in the control treatment were significantly (almost 19-fold) smaller than in the above+below removal treatment (
$\bar X$
1−
$\bar X$
2 = 0.18, P < 0.0001), but did not differ in size from the aboveground removal (
$\bar X$
1−
$\bar X$
2 = 0.14, P = 1.00) or thatch removal (
$\bar X$
1−
$\bar X$
2 = 2.9, P = 1.00) treatments.
Figure 6. Differences in Dittrichia graveolens aboveground biomass for plants transplanted as seedlings into control plots with undisturbed grassland, plots from which dry thatch was removed, plots where aboveground biomass was clipped, and plots where both above- and belowground biomass of all plant neighbors was removed. Boxes correspond to the median, first, and third quartiles, and whiskers extend to the furthest value within 1.5× the interquartile range.
Field Experiments Year 2: Response to Disturbance–Aboveground versus Belowground Competition
In the second year of experiments, the above- and belowground treatment (control, clipping, belowground competitor removal) had a significant effect on the survival of D. graveolens (
$X$
2
2 = 15.18, P < 0.001; Figure 7A). Clipping marginally significantly increased survival by 32% over the control (Z = 1.64, P = 0.100), and belowground competitor removal increased survival by 61% (Z = 3.85, P < 0.001). Shading treatment reduced survival by almost 4-fold (
$X$
2
1 = 46.161, P < 0.001; Figure 7B). In the trenching experiment, controlling belowground competition increased survival by 33% (
$X$
2
1 = 10.37, P = 0.0013; Figure 8). However, there was no significant effect of the watering and nutrient treatments (
$X$
2
2 = 2.94, P = 0.23) or the interaction between trenching and water + nutrients (
$X$
2
2 = 1.11, P = 0.58). We did not harvest biomass, because all the plants died before reaching reproduction. Plants in Year 2 died after 56 d on average (±0.67 SE), representing around one-third of their life span (until bolting) compared with plants in Year 1 (mean age 157 ± 2.60 SE). We are unable to explain the cause of this sudden mortality in Year 2, which might have reflected a combination of herbivory, temperature, and precipitation patterns, and the fact that seedlings were established slightly later than in Year 1.
Figure 7. Survival of Dittrichia graveolens plants transplanted as seedlings into field plots in 2nd-year experiment. (A) Control plots with undisturbed grassland (solid lines, red), plots where aboveground biomass was clipped (dotted lines, blue), and plots where above- and belowground biomass of all plant neighbors was removed (dashed lines, green). (B) Plots with no shade cloth (dotted lines, yellow) and with shade cloth (solid lines, gray).
Figure 8. Survival of Dittrichia graveolens in 2nd-year experiment exploring mechanisms of belowground competition. Plants were either protected from competition with weed cloth (solid lines) or open to competition (dotted lines), and received one of three resource addition treatments (control, in red; water, in blue; or water + fertilizer, in green). Kaplan-Meier survival curves with 95% confidence intervals reflect fixed effects (Treatment) only.
Taken together, our results suggest that D. graveolens’ spread is limited by competition and that its association with roadsides reflects reduced competition, rather than an affinity for the roadside soil conditions. Competition strongly reduced D. graveolens’ performance in two greenhouse and multiple field experiments conducted over 2 yr with contrasting weather patterns, lending generality to the findings. Restoration practitioners can use this information to prioritize managing soil disturbances and promoting competition close to transportation corridors and aggressively controlling roadside populations before they have the chance to spread.
Other studies concur with our conclusion that D. graveolens is a poor competitor. Brownsey et al. (Reference Brownsey, Kyser and DiTomaso2014) found that D. graveolens develops shallow roots early, with significant growth starting in May, after most resident species are established. In contrast, Bromus hordeaceus initiates root growth by March and shoot growth by April, enabling it to outcompete D. graveolens by exploiting winter rains before senescing in summer. This delayed root development limits the ability of D. graveolens to compete with graminoids that capitalize on California’s winter rains and senesce during the dry summer months. Similarly, Brinkmann (Reference Brinkmann2020) observed that D. graveolens struggled to establish in straw-mulching experiments when germinating alongside forbs. In our greenhouse experiment, aboveground biomass of D. graveolens was strongly suppressed by competition with two different European annual grasses. The rapid growth of annual grasses intensifies competition aboveground for light and belowground for nutrients and water (Coleman and Levine Reference Coleman and Levine2007), and their removal could facilitate D. graveolens’ spread into rangelands.
Other invasive species that spread along roadsides also show limited competitive ability. In its invaded range in Europe, annual ragweed (Ambrosia artemisiifolia L.) recruitment increased when competitors were removed and soils were disturbed (Fumanal et al. Reference Fumanal, Gaudot and Bretagnolle2008). Similarly, establishment of common reed [Phragmites australis (Cav.) Trin. ex Steud.] is constrained by competition, with germination and seedling establishment occurring most successfully in bare roadside soils (Brisson et al. Reference Brisson, De Blois and Lavoie2010). Horseweed [Erigeron canadensis L.; syn. Conyza canadensis (L.) Cronquist] also tends to dominate disturbed soils and is suppressed in sites with dense perennial cover (Weaver Reference Weaver2001). Together, these examples illustrate that many roadside invaders succeed not through strong competitive ability, but by exploiting disturbance and the absence of established vegetation.
Primary Drivers of Competition in Dittrichia graveolens
The primary mechanism limiting D. graveolens in field experiments appeared to be belowground competition, although aboveground competition also played a role. In Year 1, removal of aboveground biomass or thatch had no effect, while removal of above+belowground competition resulted in both higher survival and higher biomass. In Year 2, removing belowground competition increased survival by 61%, while removing only aboveground competition increased survival by only 32%. Our trenching and weed cloth treatments showed increased survival with reduced belowground competition, while adding water or nutrients had no measurable effect, leaving open the question of the mechanism behind root competition. Finally, shading likely also negatively affected survival, consistent with findings by Brownsey et al. (Reference Brownsey, Kyser and DiTomaso2014) and Brinkmann (Reference Brinkmann2020); however, disproportionate herbivory in shaded treatments may have contributed to this effect in our study. High mortality in Year 2, caused by unseasonal heat and intense drought, complicated efforts to fully disentangle competition mechanisms.
Soil played an important role in D. graveolens’ performance. In our greenhouse experiment, D. graveolens performed worse in engineered fill (a roadside soil proxy) compared with nutrient-rich field topsoil. The reduced growth in engineered fill was likely due to its significantly lower nutrient content, with approximately 11 times less nitrogen, 13 times less carbon, and 22 times less phosphorus than field topsoil. These fertility differences likely drove the observed growth patterns. Seed germination was slightly lower in engineered fill than in field topsoil, contrasting with the species’ frequent association with disturbed roadside soils. Organic matter in field topsoil may have influenced soil moisture and microbial communities, subtly affecting germination outcomes. However, the effects of soil type on germination were quite modest overall.
Managing Invasion away from Roadsides
Although D. graveolens is commonly found along transportation corridors in both its native range and in California (Lustenhouwer and Parker Reference Lustenhouwer and Parker2022; Melen et al. Reference Melen, Snyder, Fernandez, Lopez, Lustenhouwer and Parker2024), its ability to thrive in nutrient-rich soils suggests the potential to spread beyond these disturbed areas. This pattern mirrors other invasive species spreading from roadsides into adjacent habitats (e.g., Gelbard and Belnap Reference Gelbard and Belnap2003; McDougall et al. Reference McDougall, Lembrechts, Rew, Haider, Cavieres, Kueffer, Milbau, Naylor, Nuñez, Pauchard, Seipel, Speziale, Wright and Alexander2018; Sărățeanu et al. Reference Sărățeanu, Moisuc and Cotuna2010). Across multiple field experiments over 2 yr, we consistently found evidence that D. graveolens is a poor competitor, primarily limited by belowground root competition. We also found that the effect of competition on D. graveolens in the greenhouse was much stronger in the nutrient-rich field topsoil than in the engineered fill. Taken together, our results suggest that the distribution of D. graveolens is constrained to roadsides by competition. Gioria et al. (Reference Gioria, Hulme, Richardson and Pyšek2023) highlight that invaded ecosystems often include frequent disturbance, and both natural and anthropogenic disturbance create space for colonization.
Our results have several implications for management. Spread is opportunistic for D. graveolens, which exploits disturbed or patchy areas, while competition from resident plant communities restricts its establishment. Where feasible, land managers should support dense vegetation cover near spread vectors (e.g., transportation corridors, footpaths, riparian zones, construction areas) to protect sensitive habitats. Roadside verges present a special challenge: vegetation is often deliberately kept sparse for safety and visibility, making it unrealistic to rely on plant competition in these areas. In such contexts, early detection and rapid response will be essential, with spot-checking disturbed roadside sites being a critical step to prevent new establishment from wind-dispersed seeds. In adjacent habitats, perennials and early-germinating annuals with dense canopies and root systems may provide strong early-season competition, limiting D. graveolens’ survival and plant size at reproduction. Thus, management practices that reduce competition in the spring may facilitate D. graveolens’ establishment, and disturbances from cattle, mowing, and agricultural equipment that create bare soil are likely to increase invasion risk. Overall, this work underscores a broader principle in restoration ecology, that fostering strong, competitive native plant communities remains our best defense against opportunistic invaders like D. graveolens, even as roadside verges continue to pose unique management constraints.
Supplementary material
To view supplementary material for this article, please visit https://doi.org/10.1017/inp.2025.10029
Acknowledgments
We thank the UC Santa Cruz Greenhouse staff for facilities and plant care, especially Sylvie Childress and Laura Palmer, and undergraduate assistants Josie Borden, Tania Cooley, Becky Tapia, and Toni Jaroszewska. Many thanks to Marc Douvia, who helped us locate the soils used in the germination and greenhouse experiments and provided additional support in researching historical construction records. We are grateful to Zac Harlow and Zac Tuthill from Blue Oak Ranch Reserve for their knowledge, assistance, and use of facilities and research space. We also thank Cole Jower, Eric Lynch, Matt Hinshaw, Kaili Hovind, Steven Rosenau, Marina Rosenau, Shirley Gruber, and Randy Melen for all their help in the field. We are thankful for the editors and the two anonymous reviewers for their thoughtful feedback. We respectfully acknowledge the Ramaytush and Tamien Ohlone peoples, the original inhabitants of the land where we collected seeds and conducted field experiments, and the Awaswas-speaking Uypi Tribe, whose unceded territory was used to conduct our greenhouse experiments. We thank Kathleen Kay, Karen Holl, and Virginia Matzek for edits on early versions of this article. We thank the permitting agencies that allowed us to collect seed for this study: Santa Clara Valley Water District and the Don Edwards San Francisco Bay National Wildlife Refuge.
Funding statement
The study was funded by the United States Department of Agriculture, National Institute of Food and Agriculture, Agriculture and Food Research Initiative Grant 2020-67013-31856 to IMP. NL acknowledges support from the Swiss National Science Foundation (Early Postdoc. Mobility fellowship P2EZP3_178481), UKRI Natural Environment Research Council (Standard Grant NE/W006553/1), and the UKRI Horizon Europe Guarantee Research Scheme (Marie-Sklodowska-Curie European Fellowship EP/X023362/1). Financial support was also provided by the UC Santa Cruz Department of Ecology and Evolutionary Biology, the Center to Advance Mentored Inquiry-based Opportunities (CAMINO), the UC Santa Cruz Doris Duke Conservation Scholars Program, the UC Santa Cruz Plant Sciences Fund, the California Native Plant Society Marin County Chapter, and the Jean H. Langenheim Graduate Fellowship.
Competing interests
The authors declare no conflicts of interest.
Open access
