Endangered plant species under differing anthropogenic interventions: how to preserve Pterygopleurum neurophyllum in Wondong wetland?

Study area

The study area is located in Wondong wetland of Yangsan-si in South Korea (35°22′N, 128°54′E; Fig. 1), which is considered as the last natural habitat of P. neurophyllum in the country (National Institute of Biological Resources, 2021). The study area also provides habitats for another endangered wetland plant, Viola raddeana Regel (Park et al., 2021). This area is subjected to the warm-temperate climate with a hot, humid summer and a cold, dry winter, in addition to the concentrated rainfall event between July and September. Comprising the last 10 years, the average annual air temperature and precipitation are 14.9 °C and 1473, respectively. The Wondong wetland extends for approximately 57.3 ha, alongside the lower Nakdong River basin; thus, the study area has an altitude close to the sea level exhibiting minimum slope. Because of concentrated rainfall event during the summer, flooding seasonally occurs when Nakdong River overflows.

The study area has been affected by two types of anthropogenic wetland interventions for decades: wetland interventions without or with soil disturbance. The former results from continuous vegetation removal for the access to freshwater fishing; therefore, this area comprises the linear transects that contain a reduced vegetation coverage with marginal disturbance to the soil (area under vegetation disturbance only, Fig. 1A). On the other hand, the later originates from land reclamation along Nakdong River, which buries native wetland vegetation and creates bare soil patches. This area is, thus, characterized by reduced vegetation coverage, rocky and compacted soil surface, and terrestrial pioneer plant community such as Setaria viridis (L.) P. Beauv. (area under both soil and vegetation disturbances, Fig. 1A).

Flora taxa and P. neurophyllum incidence investigations

A total of 32 plots were randomly established within the study area for flora monitoring in July and August 2021. This monitoring period was selected given that the flowering of P. neurophyllum occurred in the middle of the summer season. The size of each plot was 2 × 2 m, but it was extended to 5 × 5 m if the vegetation within the plot consisted of any mature tree species (Salix type vegetation, see below) considering the canopy width and height of arborescent communities. All plant taxa were listed (Table S1), and the exotic plant species were specified as well. The incidence and population size (number of individual) of P. neurophyllum were also checked. Species richness was calculated as the number of flora taxa in each plot.

Each plot was assigned to one of four vegetation types, namely Setaria (n = 3), Miscanthus (n = 16), Phragmites (n = 8), and Salix (n = 5). The Setaria type is dominated by S. viridis, and tends to occur around the terrestrial and compacted soil conditions (Fig. S1). The Miscanthus and Phragmites types are characterized as dense Miscanthus sacchariflorus (Maxim.) Hack. and Phragmites australis (Cav.) Trin. ex Steud. communities (Fig. S1). The overstory of the Salix type is covered by Salix chaenomeloides Kimura, Salix pierotii Miq., and Salix koriyanagi Kimura ex Goerz, while the understory is primarily vegetated by M. sacchariflorus and P. australis (Fig. S1).

The Salix type prefers the upper, drier wetland area compared to the Miscanthus and Phragmites types (Son et al., 2002). Both Miscanthus and Phragmites types occupy the lower, wetter conditions, but M. sacchariflorus more favors the micro-topography on sediment mounds than P. australis (Park et al., 2021). Furthermore, the Miscanthus, Phragmites, and Salix types have historically dominated Wondong wetland (Son et al., 2002), while the Setaria type is eventually introduced to the area under both soil and vegetation disturbances after land reclamation (Park et al., 2021). Accordingly, these four vegetation types are assumed to reflect both natural environments and recent impacts of anthropogenic interventions around the study area.

Measurements for vegetation coverage and distance from disturbed area

Vegetation coverage was quantified in each plot using a drone (Phantom 4 pro V2.0, DJI, China) and the freeware software Imagej to estimate the potential competitive plant density (Banu, Borlea & Banu, 2017; Engelbrecht & Herz, 2001; Johnson, Manby & Devine, 2020). A photograph of the canopy was taken with a drone at 12 m height above the center of each plot at the same time of the flora investigation. Each photograph was clipped based on the edge of each plot, and processed following the vegetation coverage analysis protocol of Engelbrecht & Herz (2001); the type of each photograph changed from color to the grey, which led to the images with the black and white pixels representing vegetated and none-vegetated areas, respectively. Vegetation coverage of each plot was estimated by the percentage of the black pixels within the total number of pixels.

Distances from the area under vegetation disturbance only (D_V) and under both soil and vegetation disturbances (D_SV) were analyzed using drone mapping. A total of 547 photographs were taken at 40 m altitude across the study area using a grid mission procedure of Pix4Dcapture software (Pix4D, Prilly, Switzerland) with a drone. The photographs were integrated and mapped into a single image containing GPS data using Agisoft PhotoScan software (Agisoft, Saint Petersburg, Russia). D_V and D_SV were then estimated according to the displacement from the center of each plot to the line transect throughout the middle of the disturbed area on the mapped image data with QGIS software (QGIS Development Team 2021).

Soil sampling and analyses

Coarse rock content, soil bulk density, soil pH, and soil electroconductivity were measured to detect potential variations in soils under the anthropogenic wetland interventions. In each plot, a soil core was collected at 0–10 cm depth using a cylindrical soil sampler (407 cm³). Soil electroconductivity was simultaneously measured in the field using a time-domain reflectometry sensor (TDR 150, Spectrum Technologies Inc., USA). Soil samples were air-dried and processed through a two mm sieve to separate coarse rocks (larger than two mm) from the fine soils (smaller than two mm). Coarse rocks and fine soils were oven-dried at 105 °C and then weighed to determine coarse rock content and soil bulk density. Coarse rock content was calculated by dividing the oven-dry mass of coarse rock to that of the total soil, and soil bulk density was based on the proportion of the oven-dry mass of the total soil relative to the volume of the soil sampler. Soil pH was measured by a 1:5 soil-to-water ratio with air-dried soil samples and a refillable electrode (Orion ROSS Ultra pH/ATC Triode; Thermo Scientific, Waltham, MA, USA).

Statistical analyses

Plots were treated as the unit of replication for statistical analyses in the present study (n = 32, comprising three for Setaria type, 16 for Miscanthus type, eight for Phragmites type, and five for Salix type). Here, general linear model with Tukey’s HSD test was performed to assess the variation related to vegetation types (P < 0.05). Considering that all P. neurophyllum were detected in the Miscanthus type (Table 1), plots with P. neurophyllum (n = 6 out of 16) were analyzed separately from the other plots in the Miscanthus type (n = 10 out of 16) to find any differences related to P. neurophyllum incidence. Log transformation was applied to vegetation coverage, species richness, coarse rock content, soil electroconductivity, D_V, and D_SV for normalization. Levene’s test was used to assess the homogeneity of variance considering the unbalanced experimental design (P > 0.05). Exotic plant incidence was analyzed as a binary variable (presence or absence) on the basis of Bernoulli distribution. For vegetation coverage, species richness, and exotic plant incidence, plot size was used as a random factor because the Salix type had larger plot size than the other types. These analyses were carried out with ggpubr and lme4 packages in R 4.2.1. software (R Core Team, 2022).

Permutational multivariate analysis of variance with post-hoc Bonferroni test was conducted with 9999 permutations to summarize the multivariate difference across D_V, D_SV, vegetation coverage, species richness, exotic plant incidence, coarse rock content, soil bulk density, soil electroconductivity, and soil pH (P < 0.05). This analysis was based on Bray-Curtis dissimilarity, for which input data were square root transformed. Given that our experimental design was unbalanced, permutational analysis of multivariate dispersion with 9999 permutations was accompanied to confirm the homogeneity of multivariate variance (P > 0.05). Permutational multivariate analysis of variance and permutational analysis of multivariate dispersion were performed with vegan and pairwiseAdonis packages of R 4.2.1. software.

Multimodel inference was adopted to elucidate major influential factors for P. neurophyllum population size. A series of models were created using all possible combinations of explanatory variables, and ordered in accordance with corrected Akaike information criterion (AIC_C). Then, candidate models were selected until the cumulative Akaike weight exceeds 0.95, and used to calculate model-averaged coefficients (Symonds & Moussalli, 2011). Relative importance was estimated by summing Akaike weights of candidate models that include each explanatory variable, and significance of the full model was tested prior to multimodel inference (Symonds & Moussalli, 2011). Square root transformation was implemented for input data of multimodel inference. Simple linear and multiple regressions were also used to reveal any notable pairwise relationships between the variables (P < 0.05). Multimodel inference and regression tests were performed with MuMIn and berryFunctions packages of R 4.2.1. software (R Core Team, 2022).

Flora taxa and P. neurophyllum incidence

A total of 61 flora taxa were detected in the study area (19 orders, 23 families, 48 genus, 57 species, one subspecies, and three varieties). Asteraceae (nine taxa), Poaceae (eight taxa), Polygonaceae (seven taxa), and Cyperaceae (six taxa) plants were prevalent in the study area and accounted for the half of the detected flora taxa. Monocotyledons and dicotyledons consisted of 42 and 19 flora taxa, while pteridophytes and gymnosperms were undetected in the study area. When each vegetation type was separated, the Setaria, Miscanthus, Phragmites, and Salix types included 31, 34, 20, and 21 flora taxa, respectively (Table 1). List of the flora taxa is available at Table S1.

The target endangered plant, P. neurophyllum was observed only in the Miscanthus type, but not in the other vegetation types (Table 1). Out of 16 plots belonging to the Miscanthus type, P. neurophyllum was found in 6 plots, which were located near the area under vegetation disturbance only. As a result, the plots with P. neurophyllum generally featured a less dense canopy relative to the plots without P. neurophyllum. Each population of P. neurophyllum included one to 17 individuals, and 36 P. neurophyllum individuals were found in total.

Ten exotic plant species were identified in the study area (Ambrosia artemisiifolia L., Ambrosia trifida L., Conyza canadensis (L.) Cronquist, Conyza sumatrensis (Retz.) E. Walker, Cosmos bipinnatus Cav., Xanthium orientale L., Amaranthus patulus Bertol., Chenopodium album L., Amorpha fruticosa L., and Rumex nipponicus Franch. & Sav.) (Table S1). The Setaria, Miscanthus, Phragmites, and Salix types contained eight, three, two, and two exotic plant species, respectively. All three exotic plant species of the Miscanthus type were found in the plots with P. neurophyllum, whereas there was no exotic plant in the Miscanthus type without P. neurophyllum. The Setaria type had the highest exotic plant incidence ( P < 0.005), and all plots belonging to the Setaria type included exotic plant species (Table 1).

Variations related to vegetation type and P. neurophyllum incidence

Permutational analysis of multivariate dispersion confirmed the homogeneity of multivariate variance among D_V, D_SV, vegetation coverage, species richness, exotic plant incidence, coarse rock content, soil bulk density, soil electroconductivity, and soil pH (P > 0.05, Table 2), despite the unbalanced number of replication. Permutational multivariate analysis of variance and post-hoc test showed that the Miscanthus type with P. neurophyllum significantly differed from the Miscanthus without P. neurophyllum (F = 13.26, P = 0.01) and the Phragmites type (F = 11.44, P = 0.02). It suggests that the environmental conditions of the area with P. neurophyllum were possibly distinguishable from those of the area without P. neurophyllum, even when they shared the same vegetation type.

All variables complied with the assumption of homoscedasticity (Levene’s test, P > 0.05), despite the unbalanced number of replication according to vegetation type and P. neurophyllum incidence. General linear model and pairwise comparisons demonstrated that the Setaria type and Miscanthus type with P. neurophyllum were lower in vegetation coverage (F = 13.10, P < 0.005, Fig. 2A) and higher in species richness (F = 17.74, P < 0.005, Fig. 2B) than the Miscanthus type without P. neurophyllum, Phragmites type, and Salix type. The Setaria type also showed the highest coarse rock content (F = 8.33, P < 0.005, Fig. 2C) and soil bulk density (F = 3.80, P = 0.01, Fig. 2D). Conversely, no significant difference was detected for soil electroconductivity (F = 1.42, P = 0.26, Fig. 2E) and soil pH (F = 1.18, P = 0.34, Fig. 2F). The Miscanthus type with P. neurophyllum had the lowest D_V (F = 4.13, P = 0.01, Fig. 2G), while the Setaria type exhibited the lowest D_SV (F = 4.28, P = 0.01, Fig. 2H).

Influential factors for P. neurophyllum population size

Multimodel inference selected 113 candidate models from 512 sets of created models, among which the first-ranked model included coarse rock content and vegetation coverage only (AIC_C = 74.3, Table 3). These two explanatory variables also had significant model-averaged coefficients and higher relative importance than D_V, D_SV, species richness, exotic plant incidence, soil bulk density, soil electroconductivity, and soil pH (Table 3). Notably, all candidate models contained vegetation coverage, and correspondingly, relative importance of vegetation coverage became 1.00. Multiple regression test found a positive interaction effect between coarse rock content and vegetation coverage, while each of them was negatively related to P. neurophyllum population size (Fig. 3). Incidence of P. neurophyllum was concentrated at the area that had relatively low vegetation coverage and coarse rock content; however, P. neurophyllum did not occur if either vegetation coverage or coarse rock content was high (Fig. 3).

The two most influential factors (vegetation coverage and coarse rock content) for P. neurophyllum population size were significantly related to D_V and D_SV (Fig. 4). Vegetation coverage increased with D_V, and P. neurophyllum occurred in the plots having low vegetation coverage and D_V (Fig. 4A). Such pattern was especially remarkable within the Miscanthus type (Fig. 4B). Meanwhile, coarse rock content decreased with D_SV, and P. neurophyllum was found in the plots containing low coarse rock content and high D_SV (Fig. 4C). These results reflect differing effects between the reduced vegetation coverage without soil disturbance and the elevated coarse rock content under both soil and vegetation disturbances.

Pterygopleurum neurophyllum incidence and influential factors

Our results show several distinctive characteristics of P. neurophyllum microhabitats, although all of them existed in the area dominated by M. sacchariflorus. Especially, the Miscanthus type with P. neurophyllum was clearly distinguishable from the Miscanthus type without P. neurophyllum by lower vegetation coverage, higher species richness and exotic plant incidence, and closer distance to the area under vegetation disturbance only. This pattern is inconsistent with previous reports on the degradation of endangered plant populations because of anthropogenic interventions and invasion of exotic species (Dueñas et al., 2018; Loope, Starr & Starr, 2004). Instead, the decreased M. sacchariflorus coverage following anthropogenic wetland interventions seemed to provide an opportunity for P. neurophyllum (and exotic plants) to survive, even though M. sacchariflorus is native to the study area.

It is notable that vegetation coverage was the most influential factor, negatively affecting P. neurophyllum population size in the Miscanthus type. In fact, Miscanthus plants are known by their invasive characteristics to the moist ecosystems, and could overcrowd both natural wetlands and abandoned agricultural lands (Schnitzler & Essl, 2015). Such introduction of Miscanthus plants might intensify the competition for light availability, and simplify plant community composition by shading out other coexisting plants and seed bank beneath the soil (Hager et al., 2015). Furthermore, Miscanthus plants primarily distribute their root-rhizome systems to the surface and subsurface soils for nutrient and water uptake (Neukirchen et al., 1999), which could exclude P. neurophyllum that has a similarly shallow rhizome depth (Park et al., 2021; Rim & Chung, 1966). These competition effects might hinder the growth and population development of P. neurophyllum beneath the dense Miscanthus canopy. In this context, our findings suggest that dominance by M. sacchariflorus could discourage the abundance of P. neurophyllum unless wetland interventions without soil disturbance decreased vegetation coverage.

If taking all vegetation types into account, vegetation coverage alone was not enough to explain overall variations in the incidence and population size of P. neurophyllum. Specifically, the Setaria type lacked P. neurophyllum incidence despite its low vegetation coverage as the Miscanthus type with P. neurophyllum. Mechanical soil disturbance might contribute to this pattern considering that the level of coarse rock content acted as the second influential factor for P. neurophyllum population size. Actually, the Setaria type was dominant across the area under both soil and vegetation disturbances, and featured higher coarse rock content than the Miscanthus type with P. neurophyllum. Rocky and compacted soil surface are considered to decline plant growth by reducing penetration ability of root systems, forcing the topsoil roots to become redistributed into the subsoil, and restricting nutrient uptake from the soil surface (Taylor & Brar, 1991; Unger & Kaspar, 1994). As P. neurophyllum is known to form a root-rhizome system mainly at the shallow soil depth (Rim & Chung, 1966; Park et al., 2021), the rocky topsoil condition might be crucial for nutrient acquisition of P. neurophyllum and consequently block the survival under anthropogenic soil disturbance.

Differing effects of anthropogenic wetland interventions

Our results demonstrate that wetland interventions with soil disturbance could have a negative effect on P. neurophyllum population size by generating the rockier and compacted soil conditions. This negative response is not unanticipated because the excessive anthropogenic wetland interventions have been frequently pointed out as major threats of endangered plants (Ren et al., 2012). Excessive wetland interventions may distort the native plant community and surrounding environments by accelerating the invasion of competitive species, reorganizing the level of spatial heterogeneity, compacting soil profile, diminishing seed bank, reducing survival rate of remaining population, and altering hydrological regime and nutrient availability (Magnano et al., 2018; Mao et al., 2016; Schweiger et al., 2016; Wisheu & Keddy, 1991). When the magnitude of such interventions exceeds the ecosystem resilience, it may lead to the loss of regional wetland patches and habitat fragmentations (Chi et al., 2018; Deane et al., 2017). Although soil disturbance in the study area was not as severe to cause the entire habitat loss and extinction of P. neurophyllum at the current stage, its potential risk is undeniable for future fitness of the remaining population given the landscape alteration due to the overexploitation around the lower Nakdong River basin (Im et al., 2017; Oh et al., 2016).

The effects of wetland interventions without soil disturbance seemed to contrast with those of the interventions with soil disturbance. Reduced vegetation coverage owing to wetland interventions without soil disturbance had a positive effect on P. neurophyllum population size by lowering the magnitude of inter-species competition. In other words, the canopy without such vegetation disturbance might be too dense for P. neurophyllum to maintain their population in the study area. Obata et al. (2012) similarly predicted the improvement of endangered plant communities resulting from the restoration activity to remove invasive exotic plants in Watarase wetland, a historical P. neurophyllum habitat. These findings also affirm that creating canopy gaps could reorganize the monotonous landscape and enhance plant diversity (Anderson & Leopold, 2002; Scanga & Leopold, 2012; Wisheu & Keddy, 1991).

A similar situation may occur in the habitats of other endangered wetland plants. For instance, V. raddeana, an endangered plant cohabiting with P. neurophyllum is threatened by anthropogenic habitat loss and climate change (Jeong et al., 2018); nevertheless, Sawada et al. (2010) observed that prescribed burning boosted the occurrence of V. raddeana seedlings by decreasing the dominance of competitive plants in a wet secondary grassland. In fact, anthropogenic interventions sometimes mimic the role of historical, natural disturbances to wetland biodiversity and rare plant populations when their intensity does not surpass the level of ecosystems’ resilience (Zhang et al., 2020; Wisheu & Keddy, 1991). These tendencies denote the importance to develop conservation strategies for the endangered plants that are experiencing strong inter-species competitions and anthropogenic wetland interventions.

Conservation implications

Several implications can be cautiously suggested for P. neurophyllum conservation in accordance with the differing effects of wetland interventions. First and foremost, any extreme land reclamations should be excluded from the P. neurophyllum habitat. This is essential to prevent further progression of heavy soil disturbance and habitat loss for remaining P. neurophyllum populations. Second, partial reduction of competitive plants can be implemented to provide extra spaces for P. neurophyllum so that they can avoid intense resource competitions beneath the dense vegetation coverage. This treatment should involve neither the soil compaction by heavy machineries nor the incorporation of rocky materials to minimize soil disturbance. Since establishing canopy gaps could accelerate the unintended invasion of exotic plants as well (Brewer, 2010), such treatment on vegetation coverage should be followed by careful monitoring regarding the shift in plant community composition.

Direct reinforcement of P. neurophyllum population should also be addressed because small population size is generally unfavorable in terms of maintaining the safe sites for fundamental niches and minimum viable population (Zedler, 2000). Therefore, if necessary, conservation programs may consider the inclusion of artificial propagation and reintroduction into the natural habitat to help enlarge P. neurophyllum populations. Although several institutes have established ex-situ collections of P. neurophyllum and studied artificial seed germination (Kwon, Kim & Kim, 2020), sufficient number of individuals for large-scale restoration may not be available because of the extremely small wild population. In this regard, a small-scale experimental reintroduction would be an option to initiate the successive population reinforcement and clarify factors impacting the ecophysiology of planted P. neurophyllum more in detail.

Another point deserving consideration is the conflict around the study area between wetland protection and modification for agricultural, industrial, and recreational purposes. Similar conflicts also prevail across other wetlands alongside the lower Nakdong River basin (Im et al., 2017), which can be alternative sites for the experimental reintroduction of P. neurophyllum. This situation is hindering the feasibility of conservation activities, although South Korean government declares P. neurophyllum as one of the species requiring urgent restoration. Thus, policy makers and scientists should cooperate to pursue socioeconomic agreements regarding how to manage the remaining habitat for P. neurophyllum.