Skip to main content
Dryad logo

Native plant species show evolutionary responses to invasion by Parthenium hysterophorus in an African savanna


Oduor, A. M.O. (2021), Native plant species show evolutionary responses to invasion by Parthenium hysterophorus in an African savanna , Dryad, Dataset,


Invasive plant species often competitively displace native plant species but some populations of native plant species can evolve adaptation to competition from invaders and persist in invaded habitats. However, studies are lacking that examine how variation in abiotic conditions in invaded landscapes may affect fitness of native plants that have adapted to compete with invasive plants. I tested whether invasion by Parthenium hysterophorus in an African savanna may have selected for native plant individuals with greater competitive ability than conspecific naïve natives in nutrient-rich and mesic soil conditions. I compared vegetative growth and seed yields of invader-experienced and conspecific naïve native individuals. Invader-experienced natives grew shorter than naïve natives regardless of growth conditions. Nevertheless, the two groups of native plants also exhibited treatment-specific differences in competitive ability. Invader-experienced natives displayed plasticity in seed yield under drought treatment, while naïve natives did not. Moreover, drought treatment enhanced competitive effects of invader-experienced natives on P. hysterophorus, while nutrient enrichment relaxed competitive effects of experienced natives on the invader. The results suggest that P. hysterophorus may have selected for shorter native plant genotypes that also exhibit plasticity in competitive ability under drought conditions.


Seed sampling in the field

In July 2018, I collected seeds of P. hysterophorus and those of seven native species (Table 1) with which it co-occurred within Nairobi National Park. The park covers an area of 117 km2 and hosts more than 30 large mammalian herbivore species (Ogutu et al., 2013). The park forms a part of the larger Athi-Kapiti savanna ecosystem that exhibits spatial variation in soil moisture due to pulsed rainfall pattern (Imbahale et al., 2008). A shrub-grassland formation that occurs within the central and southern zones comprises a substantial part of the park where large herbivores spend most of their time feeding (Ogutu et al., 2013). The other vegetation types in the park include a forest to the western zones and a riverine woodlot to the south (Ogutu et al., 2013). The shrub-grassland formation harbours 219 native plant species (54 grass species and 165 forb species) from 50 families (Oduor et al., 2018). Seed sampling for the present study was conducted within the shrub-grassland formation. I collected seeds of each of the seven native plant species from five populations (i.e., five separate locations) that were separated from each other by at least four kilometers. Within each population, I collected seeds of native plant individuals that grew inside thick stands of P. hysterophorus (i.e., invader-experienced natives) and seeds of conspecific native individuals that occurred in adjacent open spaces without P. hysterophorus (i.e., naïve natives). In each population, I collected seeds from five mother plants each for invader-experienced and naïve natives (a total of 10 mother plants per population). To minimize a potential bias due to environmental heterogeneity (e.g., soil moisture and nutrients) within a sampling population, I collected seeds from pairs of invader-experienced and naïve individuals that were separated by 10 - 30 meters. Density of invader-experienced natives varied among the sampling sites from 4 to13 individuals per m2.  For each native species, seeds from the different mother plants were then bulked per native plant status (i.e., invader-experienced vs. naïve natives). Seeds of P. hysterophorus were also obtained from the same populations above and bulked. All seeds were then sun-dried and stored at room temperature until use in the experiment described below. Although bulking of seeds from different mother plants per species may have reduced inferential power for each species under the different treatment combinations, the use of seven native species in three families (Table 1) enables extrapolation of the results to other systems invaded by P. hysterophorus (sensu van Kleunen et al., 2014).

Experimental set-up

To test the two predictions above, I performed a fully factorial experiment in a glasshouse between November 2018 and May 2019. Competitive ability can be broken into two components — competitive effect (suppression) and competitive response (tolerance) (Goldberg & Landa, 1991). Therefore, I assessed competitive interactions between native plants and P. hysterophorus by measuring the degree to which P. hysterophorus suppressed growth and reproductive performance of the native plants (i.e., competitive effects of P. hysterophorus) and the degree to which P. hysterophorus tolerated the impact of native plants (i.e., competitive response of P. hysterophorus) (sensu Gibson et al., 2018). In the experiment, invader-experienced and conspecific naïve individuals of the seven native species were grown in 2-L pots individually and in pairwise competition with P. hysterophorus, and under different scenarios of soil nutrient enrichment and moisture availability. In the first week of November 2018, I sowed seeds of P. hysterophorus and the seven native species in plastic trays that had been filled with sterilized sand in a glasshouse with natural photoperiod of 12 h: 12 h, light: dark, temperature range of 19 - 28oC, and 35 - 56% relative humidity. All seeds were sterilized for two minutes in a 1% sodium hypochlorite solution and rinsed well with distilled water. On 28th and 29th November 2018, I transplanted three-week old seedlings to 2‐L plastic pots with four holes at the bottom that had been filled with a forest soil with no history of invasion by P. hysterophorus. The soil was obtained from a stand of Cupressus lusitanica at the Kenya Forestry Research Institute, Muguga center (1° 17.524'S, 36° 49.3168'E). The soil was excavated from the top 20 cm and had the following mean (± 1 standard deviation) chemical properties based on an analysis of six randomly selected samples: pH = 6.98 ± 0.058; electrical conductivity (mS/cm) = 0.063 ± 0.018; Nitrogen = 0.61 ± 0.24%; Phosphorus (ppm) = 3.0 ± 1.26; and Potassium (ppm) = 433.4 ± 20.63. As the objective of the study was to test for direct competition for soil moisture and nutrients rather than indirect competition mediated by soil microbes, I used soil from a neutral location to avoid variation that would otherwise be introduced by differential soil conditioning caused by invader-experienced and naïve natives.

Three weeks after transplant, I started the nutrient enrichment and drought treatments. The nutrient enrichment treatments had three levels: (i) no nutrient enrichment (None), (ii) low-nutrient enrichment (Low), and (iii) high-nutrient enrichment (High), while drought treatment had two levels: (i) experimentally-induced drought (Dr+), and (ii) no-drought (Dr-). These nutrient and drought treatments were fully crossed with two levels of native plant competition against P. hysterophorus: (i) with competition (Comp+) vs. (ii) no competition (Comp-). Each treatment combination was replicated five times, which resulted in 840 experimental pots: 7 native species x 2 levels of native plant status (invader-experienced natives [Exper] vs. naïve natives [Naïve]) x 2 levels of competition with P. hysterophorus x 3 nutrient levels x 2 levels of drought treatments x 5 replicates. As a control, P. hysterophorus was grown without competition and under the same levels of nutrient enrichment and drought treatments as above and replicated five times, which resulted in additional 30 pots. All the 870 pots were assigned in random positions on two benches within a glasshouse with the same climatic conditions as above. The nutrient enrichment treatments were achieved by using a liquid fertilizer ROSASOL–N® (Twiga Chemicals, Nairobi, Kenya) that had an N:P:K ratio of 30:10:10, and trace elements: Boron (100 ppm), Copper (75 ppm), Iron (260 ppm), Manganese (320 ppm), and Zinc (230 ppm). The fertilizer was mixed with tap water at the rate of 3g/litre as recommended by the manufacturer. In the treatment with low nutrient enrichment, I added 50 ml of the fertilizer solution to the soil in the respective pots only once for the entire duration of the experiment. For the high-nutrient enrichment treatment, 50 ml of the same fertilizer solution was applied to the soil in the respective pots every 14 days for a total duration of eight weeks (i.e., four rounds of nutrient enrichment). These nutrient enrichments were applied to simulate spatial and temporal variability in nutrient enrichments in the savannas that are caused by territorial feeding and resting behaviour of mammalian herbivores (Augustine et al., 2003; Veldhuis et al., 2018). The nutrient enrichment treatments were applied four hours after application of water treatments. Pots in the experimentally-induced drought treatment were watered less frequently than pots in the no- drought treatment. That is, in the first six weeks of the experiment, pots in the experimentally- induced drought and no-drought treatments received, respectively, 200 ml of water every third and second day. Plant water demand grew as the plants grew larger and the growth season became warmer. Therefore, to avoid severe plant mortality in later stages of development, watering was increased to 350 ml every third day for the experimentally-induced drought treatment, and every second day for the no-drought treatment. After five and a half months of growth, I recorded survival rates and seed yields for each individual plant. Immediately thereafter, all experimental plants were harvested individually, and their roots washed free of soil particles using tap water. The plants were then oven-dried to constant biomass at 80°C. Biomass of individual roots and shoots were then recorded to an accuracy of 0.1 g.

Statistical analyses    

To test whether invader-experienced natives had significantly greater competitive ability than naïve natives under nutrient enrichment and high moisture availability but not under drought treatment, I fitted linear mixed-effect models with the lmerTest package (Kuznetsova et al., 2017). In the models, biomass (root, shoot, and total) and root-to-shoot ratios of the individual native plants were treated as dependent variables, while native plant status (invader-experienced natives vs. naïve natives), competition against P. hysterophorus (Competition vs. No competition), nutrient enrichment (None, Low, and High), drought treatment (experimentally-induced drought vs. no-drought) and all possible two-way, three-way, and four-way interactions were specified as fixed-effect independent variables. Species identity of the native plants was specified as a random-effect independent variable. Statistical significance of factors was determined by Satterthwaite’s approximate F-test (Kuznetsova et al., 2017). To test whether seed yields of the native plants was influenced by the treatments above, I fitted linear mixed-effect models with the lme function in the nlme package (Pinheiro et al., 2007) with a Poisson error distribution. Statistical significance of each factor was determined with a likelihood ratio test statistic, with a chi-square distribution.

To test for competitive response of P. hysterophorus against native plants, I conducted separate analyses in two steps. First, to test the general effects of presence of native competitors on P. hysterophorus performance, I constructed linear mixed-effect models to compare biomass and seed yields of P. hysterophorus when grown without competition (n=30) versus when grown in pairwise competition with each of the seven native species (n=420). The model fixed part included competition (presence vs. absence of a native plant in a pot), while the random part of the model included species identity of the competitors. Second, to test whether the different levels of nutrient enrichment and drought treatments and native plant status had main and interactive effects on P. hysterophorus growth and reproductive output, I selected the subset of cases in which P. hysterophorus was grown in pairwise competition with native plants (n = 420) and constructed linear mixed-effects models. The model fixed part included nutrient enrichment treatment (None, Low, and High), drought treatment (experimentally-induced drought vs. no-drought), native plant status (invader-experienced natives vs. naïve natives), and all possible two-and three-way interactions. The random part included species identity of the competitor species.

To eliminate any potential bias in the results that could arise from environmental maternal effects of the field-sampled seeds, I included heights of three-week old individual seedlings (measured at transplant) as co-variates in the models. Less than 1% of all experimental plants suffered mortality; hence, I did not test whether survival rates were significantly different among the treatment combinations. Seed yields and height at maturity were analyzed for six native species while excluding Cynodon dactylon because only 10 individuals of the species produced seeds and it had mostly horizontal growth (long horizontal runners on the surface of the tables). Nevertheless, an analysis of biomass including or excluding C. dactylon produced similar results. In cases where there were significant treatment effects, I used the glht command in the multcomp package (Hothorn et al., 2008) to perform Fisher's LSD post-hoc comparisons between treatment levels. All data were analyzed with R v3. 6.2 (R Development Core Team, 2020)


National Geographic Society, Award: WW-155R-17

British Ecological Society, Award: 5693/6737

Alexander von Humboldt-Stiftung, Award: 3.4‐8151/GA-Nr. 19 016 & V-8151/17017

Deutscher Akademischer Austauschdienst, Award: 91787649