The contribution of phenotypic traits, their plasticity, and rapid evolution to invasion success: insights from an extraordinary natural experiment

of to understand why some species become succesfull and others not. understanding studying closely-related invasive and non-invasive alien taxa sharing the the same We identified this unusual in Kenya where the individuals that founded invasive Prosopis juliflora and non-invasive P. pallida populations are still present in original plantations. We evaluated field-measured traits, conducted glasshouse experiments simulating different nitrogen and water availability treatments, and did reciprocal transplants to compare functional traits and plasticity between the founders of both species (i.e. ‘invasive–non-invasive congeners’ com-parison), and between P. juliflora individuals from plantations and invaded sites (i.e. testing for rapid evolution during invasion). We found that planted individuals of P. julifora and P. pallida differed in a number of key traits related to performance and spread (root:shoot ratio, number of stems and susceptibility to seed damage) as well as in levels of phenotypic plasticity in growth responses to resource availability, which may explain their differential invasiveness. Offspring of invasive P. juliflora individuals had higher seed mass and production, germination, survival, produced more stems, matured earlier and had higher plasticity compared with those of founder individuals, indicative of rapid post-introduction evolution. this exceptional study sys-tem, we show that differences in values of only a few key traits, increased phenotypic plasticity and post-introduction evolution have all contributed to the success of P. juliflora as an invasive species in Kenya.


Introduction
Several studies have addressed whether particular traits can be linked to the likelihood of a species becoming invasive (Rejmanek and Richardson 1996, Goodwin et al. 1999, Pyšek and Richardson 2007. This research effort has resulted in many ecological and evolutionary hypotheses to explain invasion success (Catford et al. 2009, Enders et al. 2018 and to identify generalities that could aid risk assessments and management efforts (Novoa et al. 2020). However, the context-depency of biological invasions makes it difficult to understand why some species become succesfull invaders and others not (Catford et al. 2019). For such understanding it is necessary to perform comparative studies on organisms with different levels of invasiveness (van Kleunen et al. 2010a) sharing the same introduction history, residence times (Wilson et al. 2007) and propagule pressure (Colautti et al. 2006), in the same environment. However, these circumstances are rare anywhere on Earth.
From an ecological viewpoint, many invasion hypotheses are formulated around functional traits, including those related to dispersal, growth responses and reproduction (Catford et al. 2009, Enders et al. 2018. Invasiveness has also been attributed to the release from natural enemies in the invaded areas (i.e. enemy release hypothesis; Keane andCrawley 2002, Heger andJeschke 2018). Differential species invasiveness may also be associated with differences in phenotypic plasticity of traits -broadly defined as the capability of one genotype to display different phenotypes in response to different environmental conditions, and thus increasing survival and reproduction (Richards et al. 2006, Davidson et al. 2011, Liao et al. 2016. The evolutionary school of thought suggests that nonnative species often undergo rapid evolution to become invasive (Bossdorf et al. 2005, Maron et al. 2007, Moran andAlexander 2014, van Kleunen et al. 2018). Such evolution can occur through deterministic processes, e.g. in response to novel selection pressures such as herbivory, mutualistic interactions and altered abiotic conditions (Barrett et al. 2008, Prentis et al. 2008, Moran and Alexander 2014, Zenni et al. 2014, Reznick et al. 2019. In contrast, non-adaptive shifts in genotype and phenotype frequencies in invasive populations may stem from demographic events during range expansion such as strong founder events, genetic drift or spatial sorting (Shine et al. 2011, Perkins et al. 2013, Schrieber et al. 2017, Clarke et al. 2019. Studies searching for evidence of rapid post-introduction evolution often use the analogy of 'ancestor-descendant' comparisons, by studying the genetic basis of phenotypic variation within and among populations from the native and the invaded range (Rogers and Siemann 2004, Keller and Taylor 2008, Stutz et al. 2018. However, such studies rarely involve the actual 'ancestors' of invasive lineages. This is because the original founders of invasive populations are either dead or cannot be located; the native geographic origins of the introduced populations are unknown; invasive populations may comprise genetic admixtures between different native sources; or invasive populations are severely bottlenecked (Keller andTaylor 2008, Colautti andLau 2015).
Some of the world's worst invasive trees belong to the genus Prosopis (Leguminosae). Prosopis species have been introduced to more than 129 countries and have become naturalized or invasive in 122 of these, having negative effects on ecosystems, economy and society (Shackleton et al. 2014). Prosopis introductions in eastern Africa provide a truly unique opportunity to study differences in invasiveness between species and the eco-evolutionary processes that underlie successful invasion. In Baringo County, Kenya, plantations of two species, Prosopis julifora and P. pallida , that were established in the early 1980s (Johansson 1990, Otsamo and Maua 1993, Choge et al. 2002, are still present today at known locations (Choge et al. 2002, Mbaabu et al. 2019). Thus, they share similar residence times under similar ecological conditions. Introductions of Prosopis individuals also involved seeds from a few individuals of each species (Choge et al. 2002). Large-scale invasion of landscapes surrounding these plantations by P. juliflora has been documented multiple times (Choge et al. 2002, Castillo 2019, Mbaabu et al. 2019), while P. pallida individuals have only occasionally been detected outside plantations (Castillo 2019). Here, we aim to investigate why one species (P. juliflora) has become invasive while the other (P. pallida) has not, and whether rapid post-introduction evolution occurred during invasion by P. juliflora. To do this we compared functional traits and their plasticity of a) the offspring of trees of both species that were originally planted and b) the offspring of planted P. juliflora trees, and trees that have invaded areas outside plantations. The first comparison could provide insight into the determinants of invasiveness at the species level, while the second comparison would be, to our knowledge, one of the first to compare the founders of a species with their invasive descendants to assess evidence for rapid evolution during invasion.
Specifically, we tested two hypotheses: 1) that planted (i.e. founder) P. juliflora differs from planted P. pallida in functional traits, their plasticity and allometric growth allocation, supportive of differences in their invasiveness; and 2) that invasive individuals of P. juliflora differ from planted P. julifora individuals in functional traits, their plasticity and allometric growth allocation, supportive of rapid evolution during invasion. To this end, we evaluated field-measured traits related with reproductive output of planted P. juliflora and P. pallida trees and invasive P. juliflora individuals, and conducted glasshouse and reciprocal transplant experiments to compare traits associated with invasiveness, involving offspring from founder trees as well as offspring from invasive P. juliflora trees.

Study area and study species
Our study area was located in Baringo County, Kenya, which is ca 50 km north of the equator. The climate is semi-arid (Owen et al. 2004) with a yearly average temperature of 24.6°C (Kassilly 2002). The area has two wet seasons, and the mean annual rainfall is 635 mm (Kassilly 2002).
Prosopis juliflora and P. pallida share similar environmental tolerances and morphological characteristics, with some authors considering them a species compex (Pasiecznik et al. 2001). In some non-native ranges, both species are seen as valuable sources of fuelwood and livestock fodder, while in other areas they are condisered highly invasive and as ecosystem transformers (Pasiecznik et al. 2001, Shackleton et al. 2014. It is thought that the global invasion success of Prosopis species is linked to their high performance under conditions of low soil nitrogen and water availability (Geesing et al. 2000, Pasiecznik et al. 2001, Shackleton et al. 2014. For example, in Ethiopia P. juliflora invades large areas with nutrient-poor and dry soils (MLC pers. obs.) where it negatively impacts biodiversity and ecosystem services (Linders et al. 2019(Linders et al. , 2021. Further details of our study area and both species are provided in the Supporting information.

Seed collection
For common garden experiments, we collected seeds from 55 healthy and mature and haphazardly selected trees, representing both planted (n = 28) and invasive (n = 27) trees from randomly selected sites. Planted trees were from seven plantations, and with either one or both of the Prosopis species present. Invasive trees were from nine sites, all of them having mature trees and seedlings, indicating ongoing reproduction and spread. Details on morphological classification of study species are provided in the Supporting information.
Hereafter 'origin' refers to either plantation or invaded areas while 'origin site' refers to each one of the seven plantation sites or nine invaded sites. Each mature tree was characterized according to its origin and species type, i.e. as being either invasive P. juliflora, plantation P. juliflora or plantation P. pallida (Supporting information). We did not observe any P. pallida trees in invaded areas and genetic analyses confirmed that only P. juliflora is invasive in the area (Castillo 2019, Castillo et al. 2021. From these trees, between five and 30 seed pods were collected per individual. Differences in the number of seeds collected per tree were due to high variation in seed set and seed damage between individuals within and between sites. From subsets of the trees, we also determined the number of seeds per pod (n = 51 individuals), the percentage of undamaged seeds per pod (n = 51 individuals), and the seed mass per pod (n = 50 individuals). Although 55 trees were sampled, only 54 had enough replicates to be used in the experiments (Table 1 for methodology for data collection of traits).

Comparison of reproductive output between planted Prosopis juliflora and P. pallida
We fitted linear mixed-effect models (LMMs) using the nlme R package (Pinheiro et al. 2017) for seed mass, number of seeds (both log-transformed to satisfy model assumptions) and percentage of undamaged seeds, including the tree species as fixed effects. Origin site was set as random effect to account for spatio-temporal autocorrelation. To account for maternal effects on number of seeds and percentage of undamaged seeds, parent trees was added as random effect, nested in origin site.

Comparison of reproductive output between invasive and planted P. juliflora individuals
We fitted LMMs as above (i.e. same random factor terms for each response variables), for the same response variables, but including the origin of P. juliflora as fixed factor (i.e. invasive or planted). Table 1. Summary of models fitted for different trait data from the reciprocal transplant experiment. The first set of models comparing the two Prosopis species included transplant site and tree species (plantation P. juliflora and plantation P. pallida) as fixed factors, while the second set of models comparing planted and invasive P. juliflora included transplant site (invaded and plantation) and origin of P. juliflora (i.e. invasive versus plantation) as fixed factors. A random factor was included in models with height and stem diameter as response variables. This random factor was not included in models with the number of stems as response variable as it caused model overfitting (see Material and methods for model details and transformation of variables).

Response variable
Fixed effects Random effects Trait N ind

Transplant site Tree species Interaction
Plantation-invaded site pair/ origin site/parent tree

Reciprocal transplant experiment
To evaluate the determinants of invasiveness at species level and to test for rapid evolution during P. juiflora invasion, we conducted a reciprocal transplant experiment involving offspring from planted trees of both Prosopis species and offspring from invasive P. juliflora trees. We then compared traits related to invasivennes between the planted tree species (i.e. 'invasive-non-invasive congener' comparison) and between planted and invasive P. juliflora trees (i.e. 'ancestor-descendant' comparisons). The experiment was conducted using a subset of our sampled sites: three plantation sites where both P. juliflora and P. pallida were originally planted (and are still present), and five invaded sites (Supporting information). We paired 1-2 invaded sites randomly to a plantation site as there is 1) weak genetic structure between founder and invasive P. juliflora populations, 2) no genetic structure between invasive P. juliflora populations (Castillo 2019), 3) no significant difference between invaded and plantation sites in soil chemistry, nor soil textures, 4) no spatial correlation between sites for soil characteristics and the landscape variables that influence the distribution of Prosopis in Baringo (e.g. elevation, precipitation in the wettest month, Eckert et al. 2020; Fig. 1, Supporting information). In July of 2016 seeds from the selected plantation and invaded sites were germinated. Five seeds from the same tree (i.e. same maternal line) were sown in the same pot containing soil from their origin site in a common garden. Prosopis seedlings were randomly weeded after emergence, leaving only one seedling per pot. Seedlings were transplanted in the field between 14 and 21 November 2016 (details of the experimental design are provided in the the Supporting information). For each plantation/invaded site group, transplants only involved seedlings from that specific combination ( Fig. 1, Supporting information). The number of maternal lines replicated in each plantation and invaded site is provided in the Supporting information. Around 150 individuals for each plantation-invaded site pair were included (total of 447 individuals for the three pairs).
At 15 months after transplanting individuals in the field, we recorded stem diameter, height and the number of stems for each individual. Traits related with reproduction such as any sign of flowering, age at maturity and number of inflorescences and pods per individual were also recorded 17 months post-germination.

Comparisons of planted Prosopis juliflora and P. pallida
We fitted models with transplant site (plantation or invaded) and the tree species as fixed effects for height, stem diameter (log-transformed) and number of stems. LMMs were fitted for height and stem diameter with parent tree nested in origin site which was nested in plantation-invaded site pair as random factor. For number of stems, random factors indicated above caused overfitting of the model, therefore, we used generalised linear models (GLMs) with Poisson distributions and logit link functions. We report on model results using parent trees instead of seed mass as a random effect because the effect of seed mass did not change the significance of any of the fixed terms and/ or interaction terms and seedlings from parent trees without this data had to be removed from the analyses. For comparison, results of the effect of both parent tree and seed mass are shown in Table 2 and Supporting information. Since only fixed effects were of biological importance for assessing the factors that contribute to invasiveness, we pay particular attention to these in the 'Results' section. See also the Supporting information for testing of the significance of fixed factors and their interactions.

Comparisons of invasive and planted P. juliflora individuals
We fitted models as above (i.e. using the same random factors), with the same response variables, but including transplant site and the origin of P. juliflora as fixed factors. A significant interaction term between fixed factors would indicate local adaptation (Kawecki and Elbert 2004). In the absence of a significant interaction, a significant origin effect would still provide evidence for rapid evolution between plantation and invasive P. juliflora individuals. For comparison, results of the effect of both parent tree and seed mass are shown in Table 2 and Supporting information. See Supporting information for distribution and link function used in models to test for significance of fixed factors and their interactions.

Glasshouse experiment
We also performed a glasshouse experiment to compare the means of key performance traits, phenotypic plasticity of growth traits and biomass allocation strategies, between planted (founder) Prosopis tree species to test for differences of invasiveness at inter-specific level; and between planted and invasive P. juliflora trees to test for rapid evolution during invasion of this species. The experiment was conducted at Stellenbosch University, South Africa, between June and November 2017, using seeds from parent trees originating from seven plantation and seven invaded sites. The number of parent trees used for seed collections of planted P. juliflora and P. pallida and invasive P. juliflora, is provided in the Supporting information. Five seeds from the same maternal line were sown in 30 cm-deep pots filled with a 2:3 silica sand-vermiculite mixture. A total of 350 pots were included in the experiment. The percentage of germination and mean emergence time (MET) were recorded for each individual pot. Seedlings were randomly weeded out 15-20 days after emergence, leaving only one seedling per pot. At this time seedling survival for each individual pot was recorded. After four weeks of growth, pots were selected at random for each treatment and rotated weekly until harvesting. For each treatment, a mean of 18.5 replicates (i.e. number of individual pots/treatment) for plantation P. pallida, 20 replicates for plantation P. juliflora and 30.5 replicates for invasive P. juliflora were included; leading to a total of 276 individual treatment combinations (total number of replicates was reduced due to germination failure or no survival of seedlings in some pots). The number of maternal lines replicated in each treatment is shown in the Supporting information. The following treatments were applied in a full factorial design: water (low and high availability) and nitrogen (low and high availability), the details of which are provided in the Supporting information.
After 20 weeks of growth, eight traits were measured for each seedling: root and stem length, biomass of roots, leaves and shoots, root to shoot ratio (RSR), number of leaves and total plant biomass (see the Supporting information for methodology).

Comparisons of planted Prosopis juliflora and P. pallida
We fitted models with tree species as fixed factor for percentage of germination, MET and survival. The latter was estimated at the time of weeding the pots and therefore it was not affected by the water and nitrogen treatment. General mixed-effect models (GLMMs) were fitted for percentage of germination and survival, with parent tree as random factor using the lme4 R package (Bates et al. 2017). For MET, we used GLM since the model was overfitted with parent tree as random factor.
LMMs were fitted for root length, stem length, root biomass, stem biomass, RSR, leave biomass and total plant biomass (all log-transformed). For number of leaves we fitted GLMMs. These models were fitted with tree species, nitrogen treatment, water treatment and their interactions as fixed factors. Parent tree nested in origin site was considered as random variable, except for number of leaves, were only parent tree was added as random factor due to overfitting of the model.
MET was not added as covariable in the models because it did not have a significant effect on survival (χ 2 = 0.12; p = 0.73). Similarly, as per reciprocal transplant experiment, Figure 1. Experimental design of the reciprocal transplant experiment in Baringo, Kenya. Plantation sites were randomly grouped with invaded sites, each group consisting of one plantation site where both P. juliflora and P. pallida trees were present, and one or two invaded sites away from the plantation and where only invasive P. juliflora trees were present. Seedlings from these trees were planted in their site of origin or transplanted. For example, in the case of plantation 1 and invaded site 1; invasive P. juliflora individuals were planted in invaded site 1 (same origin site = 'home') as well as in plantation 1 (i.e. different origin site = 'away'). Around 150 individuals for each plantationinvaded site group and a total of 447 individuals for the three groups were included. The number of individuals per planted P. juliflora and P. pallida and invasive P. juliflora, planted/transplanted in each site is provided in the Supporting information.   we tested seed mass as a random term instead of parent tree, but the effect of seed mass did not change the significance any of the fixed and/or interaction terms. Therefore, results are based on models using parent tree instead seed mass as a random effect. See the Supporting information for results of the effect of seed mass as random factor. The distribution and link function used in models and methods for testing of the significance of all fixed factors and their interactions in all models are given in the Supporting information.
We also estimated the phenotypic plasticity of growth traits by determining the coefficient of variation as the standard deviation of trait means/mean of trait means (Schlichting and Levin 1986, Valladares et al. 2002. We then calculated differences between planted tree species in: 1) plasticity for each trait, 2) the mean plasticity for nitrogen and water availability independently and 3) in the mean plasticity across nitrogen and water treatments.
Lastly, we estimated differences in allometric biomass allocation, i.e. the degree of investment in a given trait per unit biomass produced, across treatments for each growth trait as the relationship between Ln(total plant biomass) and each Ln-transformed trait measure, using standardized major axis (SMA) regression (Sokal and Rohlf 1995) that includes the variability of both variables. SMA slopes (i.e. ratio between traits) were tested for significant differences between tree species under different nitrogen and water availability conditions. From an allometric viewpoint, different slopes or different elevations suggest that growth allocation is affected by the treatments, indicating differences in static allometric slope for that particular trait. SMA regressions and tests were implemented using the SMATR package in R (Warton and Ormerod 2007).

Comparisons of invasive and planted P. juliflora individuals
We fitted models as above, for percentage of germination and survival, but including transplant site and the origin of P. juliflora as fixed factors. For the eight growth traits, we fitted the same models as per tree species comparison (all log-transformed except stem length) but with the origin of P. juliflora as fixed factors. MET did not have a significant effect on survival (χ 2 = 3.07; p = 0.07), so it was was not included as covariable in models. For comparison, results of the effect of both parent tree and seed mass are shown in Table 3 and the Supporting information. See also the Supporting information for distributions and link functions used in models for testing significance of all fixed factors and their interactions.
As per planted P. juliflora and P. pallida comparisons, but between P. juliflora origins, we calculated differences in plasticity for: 1) each growth trait, 2) water and nitrogen treatments and 3) across treatments; and compared allometric biomass allocation of growth traits across treatments. We performed all analyses using R statistical language (<www.rproject.org>).

Comparisons of invasive and planted P. juliflora individuals
There was no origin × transplant site interactions for stem diameter, height and number of stems. Prosopis juliflora seedlings growing in plantations were taller (Table 1, Fig. 3a) and had larger stem diameters (Table 1, Fig. 3b) than those growing in invaded sites, and invasive individuals had, on average, 14% more stems than planted individuals (Table 1, Fig. 3c). At the time of the data collection (17 months post-germination), seven invasive P. juliflora seedlings had reached reproductive maturity (first production of flowers and/or presence of pods) while plantation P. juliflora did not show any signs of reproduction. These seven invasive P. juliflora individuals had between two and nine flowers and between one and 17 seed pods.
Survival of plantation P. pallida and plantation P. juliflora seedlings did not differ (χ 2 = 0.67, p = 0.41, Fig. 4c). Prosopis pallida seedlings had 36% longer stems and 21% lower RSR than plantation P. juliflora seedlings, independent of the treatments (Table 2, Fig. 5b and e, respectively). Under high water availability, plantation P. pallida had 15% longer roots (Tukey HSD: p < 0.001). Under low nitrogen availability plantation P. pallida tended to have longer roots than plantation P. juliflora (Tukey HSD: p = 0.05, Table 2, Fig. 5a), and this difference remained significant when including seed mass as random factor instead of parent tree (Tukey HSD: p < 0.05, Supporting information). Overall, plantation P. juliflora showed different plasticity in various traits compared to plantation P. pallida (Table 3, Fig. 6). Across plant sizes, plantation P. pallida had proportionally longer stems and allocated more biomass to them than plantation P. juliflora, while the allocation of biomass to Figure 2. Differences in reproductive output of parent trees of invasive P. juliflora (red), plantation P. juliflora (blue) and plantation P. pallida (green) depicted as (a) average number of seeds per pod, (b) the percentage of undamaged seeds per pod (i.e seeds not aborted or showing signs of herbivore damage or fungal infection), (c) average seed mass per pod. These seeds were used in the common garden experiment. Separate models were used: one between tree species (i.e. plantation P. juliflora and plantation P. pallida) and a second between P. juliflora origins (i.e. invasive P. juliflora and plantation P. juliflora). Bars represent mean ± standard error. Statistical significance: *, p < 0.05; **, p < 0.01; ***, p < 0.001 (p < 0.05; F -tests). roots and leaves was proportionally similar between the two species (Fig. 6b-d and f ). Also, P. pallida seedlings of the same size differed in their root length and RSR (Fig. 6b and e).
Overall, invasive P. juliflora showed higher plasticity compared with plantation P. juliflora (Table 3). Compared with plantation P. juliflora individuals, invasive P. juliflora individuals of bigger size also have proportionally longer stems (this difference was marginally significant when excluding outliers; p = 0.06), and allocated proportionally more biomass to stems ( Fig. 6b and d). Figure 6. Standardized major axis (SMA) regression relationships between total plant biomass and (a) root length, (b) stem length, (c) root biomass, (d) stem biomass, (e) root-shoot ratio (RSR) and (f ) leave biomass in invasive P. juliflora, plantation P. juliflora and plantation P. pallida, in response to water and nitrogen availability treatments: high water-nitrogen (open blue circles), high water-low nitrogen (closed blue circles), lower water-high nitrogen (open orange squares) and low water-nitrogen (closed orange squares). SMA slopes were tested for significant differences between tree species (i.e. plantation P. juliflora and plantation P. pallida) and between P. juliflora origins (i.e. invasive P. juliflora and plantation P. juliflora) along the different water and nitrogen treatments. Differences in SMA elevations (Elev) were evaluated when slopes were similar. Different letters in brackets indicate significant differences in SMAs slopes or elevation (p < 0.05). For (b) and (d), differences in SMA slopes were found between P. juliflora origins (red letters) and differences in SMA elevations were found between tree species (black letters). R 2 = Pearson correlation coefficients for evaluated relationships. Statistical significance: *, p < 0.05; **, p < 0.01; ***, p < 0.001; ns, not significant.

Discussion
The invasiveness of species may stem from a combination of pre-adapted traits of ecological importance, their plasticity and rapid evolutionary changes post-introduction (Colautti andLau 2015, Gallagher et al. 2015). We hypothesized that planted individuals of Prosopis juliflora and P. pallida differ in functional traits, their plasticity and allometric growth allocation, supportive of differences in their invasiveness. We also hypothesized that these differences exist between planted and invasive individuals of P. juliflora, supportive of rapid evolution during invasion. We found support for both hypotheses by showing that the means of, and plasticity in, key performance traits differ between planted Prosopis species that differ in their invasiveness, and that invasive P. juliflora individuals have diverged from those originally planted in the study area.
The traits that differentiate invasive and non-invasive species often depend on conditions in the new range (van Kleunen et al. 2010b, Kueffer et al. 2013) and/or the stage of invasion (Richardson and Pyšek 2012, Catford et al. 2019, Milanović et al. 2020. Despite these context-dependencies, species traits that underlie high performance in the native range can confer pre-adaptations to become invasive (van Kleunen et al. 2010b). For instance, root-to-shoot ratio has been put forward as an important trait for seedling performance and biomass production, particularly under drought conditions (Lloret et al. 1999). Increased allocation to roots may increase uptake, or more efficient use of, water and other soil resources during dry periods (Padilla and Pugnaire 2007). Conversely, less allocation to above-ground biomass translates into less transpiration, and thus less water loss. Rapid development of roots after germination, leading to high RSRs has been previously reported from both Prosopis species we studied (Pasiecznik et al. 2001). In our glasshouse experiment, we found that the offspring of plantation P. juliflora had higher RSRs compared to those of plantation P. pallida. Also, plantation P. pallida seedlings of the same size differed in their root length but in both plantation P. pallida and P. juliflora the allocation of biomass to roots was proportionally the same. This is probably due to differences in root architecture (i.e. root diameter, root proliferation) or physiological traits not evaluated in our study. Our data suggest that differences between planted P. juliflora and P. pallida in strategies and (likely preadapted) traits related to resource use and acquisition may explain their differential invasiveness in Baringo, and could be particularly important in invasive species of arid and semi-arid systems (Funk 2013).
Among the ecological attributes that have been linked to plant invasiveness (van Kleunen et al. 2010b, Gallagher et al. 2015, Gioria and Pyšek 2017, increased height and aboveground biomass production often seem to benefit invasion success Richardson 2007, van Kleunen et al. 2010b). We found the offspring of plantation P. pallida to be, on average, taller than those of plantation P. juliflora in both our transplant and glasshouse experiments. As for many trees, above-ground biomass production is positively correlated with stem diameter in Prosopis (Muturi et al. 2011). Therefore, even though P. pallida grew taller in transplant sites, P. juliflora likely produced more above-ground biomass due to its thicker stems. In our glasshouse experiment, and across plant sizes, plantation P. pallida grew longer stems and allocated more biomass to them (per unit of total plant biomass produced) compared with plantation P. juliflora. However, this was not linked to higher stem biomass in P. pallida. We also found that the offspring of planted P. juliflora individuals produced more stems than those of P. pallida. The development of numerous stems could be related to stress tolerance under low moisture availability (Pasiecznik et al. 2001) and multi-stemmed plants can survive, and continue growing, if one stem dies (Götmark et al. 2016). This trait may therefore further promote the invasiveness of P. juliflora.
Our finding of longer emergence time in P. juliflora compared with plantation P. pallida disagrees with the general consensus that invasive species germinate earlier than noninvasive congeners (Gioria and Pyšek 2017). However, we found that early emergence was not related to survival of seedlings of both species. Early germination would also allow individuals to reach larger sizes when becoming reproductive (Donohue et al. 2010), which could be the case for P. pallida. Future studies should evaluate the long-term benefits of early germination such as higher establishment during late seedling development, size or fecundity in trees (Gioria et al. 2018).
Invasive plants are often liberated from their specialist natural enemies (i.e. the enemy release hypothesis; Keane and Crawley 2002) leading to higher growth and reproduction in their new ranges. The absense of specialist enemies may also create opportunities for the reallocation of (often costly) resources associated with defence strategies towards performance via rapid evolution, in what has been termed the 'evolution of increased competitive ability (EICA)' hypothesis (Blossey and Nötzold 1995, Bossdorf et al. 2005, Rotter and Holeski 2018. In Baringo County, we found Prosopis pods to be attacked by a number of seed-feeding insects. The fact that plantation P. juliflora individuals had significantly less damaged seeds than P. pallida individuals suggests that P. juliflora is experiencing higher levels of enemy release than P. pallida. These differences in enemy loads could result in differences in the survivorship, fecundity, biomass or other demographic parameters, i.e. performance, of these tree species. Phenotypic plasticity is also often linked to plant invasion success (Gallagher et al. 2015, Huang et al. 2015, partly because individuals with higher plasticity levels would be better equipped to survive and reproduce under heterogenous environmental conditions (Richards et al. 2006). However, phenotypic plasticity does not always benefit invasion (Palacio-López and Gianoli 2011) or it may only be important during the initial stages of establishment (Bossdorf et al. 2005, Lande 2015, as plastic responses may not always translate into higher fitness (Davidson et al. 2011). Our results provide evidence to suggest that differences in phenotypic plasticity between P. julifora and P. pallida in growth responses to resource availability, can separate invasive and non-invasive species. We found the offspring of invasive P.