Road traffic and landscape characteristics predict the occurrence of native halophytes on roadside verges

Road management practices, such as winter de-icing create ideal habitats and competitive advantage for salt-tolerant species. We aimed to map the occurrences of halophytes along roads in Hungary. Furthermore, we tested factors that might play a role in the roadside occurrences of five chosen native halophytes from rare to common, we encountered during our field surveys. These were Festuca pseudovina, Limonium gmelinii subsp. hungaricum, Podospermum canum, Puccinellia distans and Spergularia media. We found, that at least one halophyte species was documented in 71% of the total sampling points. Germination experiments indicated that substrate salt concentration significantly decreased germination rates in each of the five species, but in case of L. gmelinii subsp. hungaricum, or P. distans germination occurred on extremely high salt concentrations. Traffic intensity, the presence of other halophytes at the sampling point and the presence of a given species in the surrounding landscape had a significant positive effect on the occurrence of four of the five model species. Our results suggest that the studied species are mostly in the early stage of their roadside spread, colonizing roadsides close to their native distribution ranges. The possibility of a future range expansion along roads cannot be excluded.


Materials and methods
Road surveys. In order to survey roadside occurrences of the five model species we carried out fieldwork between April and September 2018. We surveyed a total of 519 random sampling points along roads, covering an approximately 2600 km long road section across Hungary, with sampling points randomly designated every 4-5 km, at points where the traffic situation and rules allowed stopping. At every sampling point, we surveyed the roadside along a 100-m-long transect and we recorded the presence or absence of the five halophyte model species and we also recorded the presence of other halophytes besides our model species. The number of individuals belonging to these species, their distance from the edge of the road along with geocoordinates and altitude of the sampling locations were also recorded. The distance of individuals from the road edge was examined at a width of 6 m from the road. Given that cars cannot stop at random points on motorways, these surveys were restricted to gas stations and rest areas. We used the online database of Hungarian Public Road Nonprofit Pte Ltd Co. 2 to categorize the surveyed roads as follows: motorways, main roads (1st main roads, 2nd main roads) and lower category roads [i.e. connecting roads (connecting settlements), conjucting roads (connecting settlements to the national road network), roads leading to railway station, unknown numbered roads].
Road and landscape parameters. To characterize roads, we gathered information about the cross-sectional annual average daily traffic data of road sections surveyed during this study. The latter data was obtained for the year 2018 from the online database of the Hungarian Public Road Nonprofit Pte Ltd Co. 2 . In order to assess whether nearby occurrence of a model species affected its roadside occurrences, we gathered information on the number of occurrences of the studied species in flora mapping quadrats (6,25 × 5,55 km) surrounding our surveyed roadside sampling points using the Hungarian Flora Mapping Database 52 . In the latter case we calculated the total number of occurrences of a species in the quadrate in which the sampling point was located as well as in the eight quadrates surrounding this using Quantum GIS version 3.4 53 . We gathered this data for all five model species.  54 . All five species studied have highly characteristic morphology and their identification is straightforward. No voucher specimens were collected, but occurrences were documented using photographic evidence (Fig. 1).
In order to estimate individual seed production by our model species, we collected 30 individuals/species during fieldwork in 2018. We determined their following characteristics to assess seed production: number of flowering shoots of 30 individuals, number of inflorescences per shoot (in this case also the number of flowers per inflorescence in 30 inflorescences) or number of flowers per shoot, and the number of seeds per fruit in 30 fruits. Thousand seed mass of the species was determined following Török et al. 80, by calculated it from the data obtained by measuring the mass of 3 × 100 seeds.
The germinability of the seeds of the five species in different salt concentrations was tested in October 2019 in an in vitro experiment. Seeds did not need stratification. Seeds for the germination test of Limonium were collected on 16 September 2019 near Balmazújváros, seeds of Podospermum were collected on 02 June 2019 near Karcag, seeds of Spergularia and Festuca were collected on 25 June 2019 near Sárszentágota, while Puccinellia seeds were collected from the side of the road 33 between Debrecen and Hortobágy on 18 June 2019. Prior to the start of the experiment all seeds were stored at room temperature (22-23C°) under similar conditions. Seeds of Festuca, Podospermum, Puccinellia and Spergularia were stored for approximately four month and seeds of Limonium were stored for one month. Seeds were placed to germinate in Petri dishes containing 1% agar substrate with one of the following 14 different concentrations of NaCl in mass percent [m/m%] (and molar solutions (mM) and soil salinity of saturated extract (ECe in dS/m) converted following Food and Agriculture Organization (FAO) classification 55  A total of 2100 seeds were placed to germinate per species, 50 seeds per Petri dish, and each salt concentrations was tested in triplicate. Germinability test was started on 13 October 2019 performed in parallel for the five species and lasted for 33 days. Petri dishes were kept at room temperature, close to a window, subject to natural light conditions throughout the experiment. The number of germinated seeds was checked and recorded daily. Criteria for germination was the appearance of the first green parts, the cotyledon (in case of Limonium, Podospermum and Spergularia) or coleoptile (in case of Festuca and Puccinellia).

Data analyses.
Statistical analyses were carried out in R (version 3.4.1) 56 . In order to test whether there is a difference between species regarding their distance from the edge of the road (thus testing the occurrence in different microhabitats) we built linear mixed effect models (LMMs) using package lme4 56,57 . We used the distance from the road of individual plants as a dependent variable, while species identity was included as sole explanatory variable. Given that observations within a sampling point are non-independent, sampling point identifier was included as a random factor in the model. To assess significance, we used the Satterthwaite approximation for degrees of freedom as recommended to minimize type I error (1) and implemented in the lmerTest package (2) and likelihood ratio tests for multiple-df tests.
To analyze germinability of seeds at various salt concentrations we used a binomial GLMM using package lme4 57 , where germination outcome (germinated/non-germinated) of seeds was used as dependent variable and www.nature.com/scientificreports/ NaCl concentration was the sole explanatory variable, included as a second-degree orthogonal polynomial. All models included the identifier of the Petri dish as a random factor. If the non-linear effect was not significant, we re-fitted the model using only the linear NaCl concentration as an explanatory variable. Predicted germination probabilities and associated prediction intervals were obtained using the 'predictInterval' function 58 .
In order to explore whether landscape flora and road characteristics predict roadside occurrences of the five species studied here, we used generalized linear models (GLM) using package lme4 57 , with binomial error distributions (i.e. for presence/absence data) built separately for each of the five model species. Presence or absence of a given model species was used as dependent variable and we used altitude, traffic intensity, number of other halophyte species (besides our five model species) found at the sampling point, and the sum of occurrences of the candidate species in the surrounding landscape (flora mapping quadrats) as explanatory variables in the models. Only ten species were considered as 'other halophytes' in the analysis, namely species whose phenology allowed detection during the period of our field surveys. These were the following: Artemisia santonicum, Atriplex tatarica, Festuca pseudovina, Hordeum hystrix, Limonium gmelinii subsp. hungaricum, Podospermum canum, Puccinellia distans, Spergularia maritima, S. rubra, S. salina. In a GLM where we used a given model species as dependent variable, this species was not included in the sum of other halophytes. We further included the county as random factor.
Nomenclature used in this work follows The Plant List (2013) 59 .
Ethical statement. The presented experimental research and field studies on plants, including the collection of plant material, comply with relevant institutional, national, and international guidelines and legislations. Plant material was collected from public lands and none of the studied species are protected by law in Hungary, therefore collection did not require special permits.

Mapping halophytes along roads in Hungary.
Here we surveyed a total of 517 random sampling points at roadsides in Hungary, covering different road categories with various traffic and disturbance intensities (Fig. 2). The survey resulted in the documentation of 15 halophyte species besides our model species occurring in roadside verges of public roads (Table S1). The large majority (over 71%) of the surveyed sampling localities harboured at least one halophyte species. Only 27 points with salt-tolerant plants harboured none of the five model species studied in detail here. The most commonly documented halophyte along roads in Hungary was Puccinellia distans, whose presence was documented at more than half of our sampling points, with a total of nearly 19,000 individuals. The rarest of the five model species was Spergularia media, which was found in only six sampling locations (Table 1). www.nature.com/scientificreports/ Interestingly, when the relative frequency of the five target species was assessed for different road categories, Festuca, Podospermum and Puccinellia showed similar frequency along motorways. All three of these species were present at more than the 60% of the sampling points taken along motorways. Limonium was less frequent along motorways, while Puccinellia was the most frequent in this category ( Reproductive traits and germination experiment. We have found that the estimated potential seed production per individual was the highest in case of Limonium, but was similarly high in the case of Spergularia ( Table 3).
The germination experiment showed marked germinability differences among the five model species, with a high proportion of seeds germinating in the case of Festuca, but only a few germs being observed in the case of Spergularia (Table 4). Some seeds germinated already by the first day of the experiment in case of each species, but this usually happened only at lower salt concentrations (0.00-0.30%). At the highest salt concentration (3.00%) only Limonium was able to germinate in very small numbers. All species germinated at the four lowest concentrations (0.00%; 0.15%; 0.30%; 0.45%) and all species had the highest germination rate on the control media and the lowest germination rates occurred on the highest NaCl concentrations.
Germinability decreased significantly or marginally significantly (in case of Puccinellia) with increasing substrate NaCl concentration as shown by GLMMs performed separately for each species (Table 5). The effect of NaCl concentration was non-linear in case of Limonium and Podospermum, but was linear in case of Festuca, Puccinellia and Spergularia (Fig. 3).
Effect of road and landscape characteristics on roadside occurrences. Traffic intensity had a significant positive effect on the occurrence of Festuca, Podospermum Puccinellia and Spergularia, but had no effect on Limonium. The probability of occurrence at roadsides increased significantly with increasing number of other halophytes present at the sampling site in all species except Spergularia. Moreover, sum of occurrences in the  www.nature.com/scientificreports/ surrounding flora mapping quadrats significantly positively influenced occurrence probability in Festuca, Limonium, Podospermum and Spergularia, but not in Puccinellia (Table 6). Altitude had significant negative effect on the occurrence of only two halophytes, Festuca and Limonium.

Discussion
Halophytes are increasingly recognized to colonize disturbed roadside verges. In this study we highlight the frequency and species richness of halophytes colonizing roadside verges, by showing that over 70% of random sampling points harbour at least one of the 20 identified species colonizing verges. Moreover, by examining the occurrences of five commonly encountered halophyte species in Hungary we highlight the role of traffic intensity, local conditions (e.g. presence of other halophytes at sampling) as well as the influence of the surrounding landscape's flora in influencing the probability of roadside colonization in several halophytes.  Table 5. Effect of NaCl concentration on germination: (poly (NaCl concentration) 1 shows the linear effect and poly (NaCl concentration) 2 shows the non-linear effect of NaCl on germination (germinated/nongerminated seeds)). www.nature.com/scientificreports/ has been found along North American roads as well 79,80 . Finally, the occurrence of Spergularia media along roads is well documented in Hungary [39,2018] as well as in several other countries, such as Austria 81 , the Czech Republic 82 , Germany 68 , the United Kingdom 73 , Ireland 83 or in the USA 84 . Data generated by our field surveys indicate that roadsides provide potentially suitable habitats for halophytes. It is likely that due to the increased salinity of the soil in the proximity of roads, halophytes find favorable conditions here, where their salt tolerance might provide them with competitive advantage over other species. Limonium did not form large monodominant patches along roadsides, whereas Podospermum, Festuca, and Puccinellia occurred in many cases in dense contiguous patches along roadsides. Several other disturbance-tolerant species occurred at our surveyed points, the most common of which were the Matricaria recutita and Atriplex tatarica.

Estimate SE t-value p-value
The spread of halophytes was hypothesized to be more likely along roads with higher-traffic intensity (e.g. motorways) due to higher air turbulence and due to the passing of more vehicles potentially transporting seeds. In accordance with this hypothesis, we found that three of our five model species occurred most frequently along motorways, however, Limonium and Puccinellia were more common along roads of intermediate traffic intensity. Based on our results and our personal observations in the field we believe that the spread of most halophytes in Hungary began along lower-category roads that pass through the natural habitats of the focal species, where roadside verges form a natural transition to the surrounding landscape. As a second step, propagules from established populations along lower category roads reach motorways where they can start a faster spread due to a more intense traffic. In support of this hypothesis, a series of halophytes documented on road edges during our field survey were found only along main roads, but not along motorways. These inlcuded Hordeum hystrix, Plantago tenuiflora, P. maritima, Lepidium perfoliatum, Myosurus minimus, Artemisia santonicum, A. ponticum and Spergularia salina. It remains to follow the spread of these species along roads, especially from already established roadside populations. Plantago maritima was also found along roads in Scotland 85 . However, during our roadside surveys, Hordeum hystrix was found only at a few points nearby original saline habitats, propagules of this species was found to be carried by cars in large numbers in Australia 86 .
A proposed key component determining the distance of halophyte individuals from the road is the distribution of de-icing salt content within the verge, as well as the diminution of the competitive advantage with increased distance from the road's edge of halophytes over other ruderal, stress-tolerant species. In case of our five model species, each appeared to occupy different microhabitats at the roadsides, which might also reflect their competitive ability. Festuca occurred generally farthest from the road edge mostly in dense closed grassy places and Spergularia individuals were the closest to the road edge, but mostly where free surfaces were available. www.nature.com/scientificreports/ These microhabitats are very similar to the natural microhabitats they colonize on the salt steppes regarding the relative vegetation cover and bare ground 87 .

Reproductive traits and germination experiments.
Previous studies have shown that low seed weight is one of the most important functional traits in plant dispersal, as low seed weight is usually associated with significant seed production, persistent seed bank, and efficient wind propagation 88,89 . Upon comparing to the average thousand-seed weight of Hungarian salt-tolerant species, we found that Puccinellia, Festuca and Spergularia had much lower seed weight, than the overall average. We gathered the seed mass of 60 native salt steppe species based on Török et al. 90,91 to compare them with the seed mass of our model species. The average seed mass of them was 0.7661 g () and comparing to this, Podospermum was the only species with higher seed mass while Limonium was close to the average value. The high seed mass of the latter two species can be compensated for by other reproductive traits in terms of propagation. In case of Limonium the higher seed mass is accompanied by significant individual seed production, while the propagation of large seeds of Podospermum is winged by the pappus.
Previous studies have shown a positive correlation between seed number per fruit and spatial spread 92 . Here we found that the two rarest of our model species (Limonium and Spergularia) have the largest seed productions, while the most widespread model species, Pucinellia produced the second lowest seed counts, indicating that seed production is unlikely to be the key factor explaining variation in the distribution of native saline species along roads. Another important trait could be the life-form of these halophytes, most of them being hemicryptophytes, which are well known to be high resistant to mowing thanks to their overwintering organs. The success of our model species along the roads can also be increased by their high degree of salt tolerance, which was also demonstrated by our germination experiment. In consistent with our experiment, other studies have also been documented the negative effect of NaCl on germination rates in case of Spergularia media 93 and Puccinellia distans 94 . Soils with a soluble salt content of 0.25-0.50% can already be classified as highly saline, where only few halophytes can survive 95 . In all five of our model species, germination was documented at concentrations much higher than this threshold. Although NaCl alone does not represent the composition of the soluble saltcontent of saline soils 95 . According to the FAO classification 55 , four of our model species germinated on NaCl concentrations that corresponded to very strongly saline soils in nature. On such high salt concentrations only a few tolerant crops are able to produce satisfactory yields. Moreover, Podospermum germinated on substrates that corresponded in salt concentration to strongly saline circumstances. Halophytes requires a series of physiological and morphological adaptations in order to make them truly tolerant to high salt concentrations. The regulation of increasing salt levels in plant tissue has led to the induction of specific changes in plant cell tissue and organs  96,98 . Limonium, furhtermore has salt-glands facilitating the recretion of salts 99 . Generally, monocotyledonous halophytes, that show optimal growth under 50 mM NaCl 100 are characterized by lower salt tolerance than dicotyledonous halophytes, that shows optimal growth up to 250 mM NaCl 101 . According to Pătruţ et al. 102 Limonium and Puccinellia belong to species adapted to intensely salinized biotopes, while Festuca and Podospermum adapted to soils with lower, but different degrees of salinization. This also reflects our results, since Limonium and Puccinellia were characterized by the highest salt tolerance according to the germination experiment. Overall we conclude, that the salt tolerance of our model species does not predict their frequency along roads, suggesting that extreme salt tolerance might not be an important factor in the roadside success of halophytes.
Factors influencing roadside occurrences. The roadside occurrences of our studied species seemed to be strongly positively influenced by traffic intensity and by the distribution of the focal species in the surrounding landscape. Furthermore, the number of other halophytes present at the sampling points was also a significant predictor of higher probability of presence. For two species, F. pseudovina and L. gmelinii subsp. hungaricum, a significant negative effect of altitude was observed. Our result indicating the positive effect of traffic intensity is not surprising, since several studies have already indicated the role of traffic in the spread of plant species 37,103 . The correlation between the number of other halophytes and the presence at the roadside of our model species indicates that it is likely that salt-tolerant species colonize these edges from nearby saline habitats, potentially from habitats identical to those where our model species originates. On the other hand, more halophytes at sampling locations might simply be indicative of local conditions that are highly suitable for salt-tolerant taxa, and therefore for our model species. In four of our model species the positive effect on roadside occurrence of the presence in the surrounding flora mapping quadrats clearly indicates that they currently mostly occur at roadsides where they also occur in the surrounding landscape. In the case of our most common roadside model species, Puccinellia, this correlation could not be detected. This probably indicates the fact that this species has already reached the stage of faster roadside dispersal outside its native distribution range. It is also important to note, that Puccinellia is known from 521 quadrats from Hungarian Flora Mapping Database, whereas our rarest roadside model species Spergularia is known only from 34 quadrats, which shows that their frequency along roads is also strongly influenced by frequency in the landscape.
In conclusion, several of our native salt-tolerant species appear to be at a stage of local colonization of roadsides, mostly situated within their native range. However, Austrian, Czech and German roadside occurrences of Limonium 61-63 may be good examples of the potential long-term spread of our native halophytes, which seems to be an occasional phenomenon for now, but with increasing traffic intensity it is likely to occur more frequently. These results highlight how certain species (even potentially invasive ones) reach the roadside and exit their original range. This provides data also for understanding plant invasions.