Vegetation and soil seedbank dynamics in Parthenium hysterophorus L. invaded subtropical grassland in Nepal

Parthenium hysterophorus L. is a noxious invasive weed and is ever expanding in its introduced range including Nepal. Understanding vegetation dynamics including soil seedbank in Parthenium invaded communities and the growth pattern of the weed itself is essential for effective management of Parthenium. We monitored growth of Parthenium (height, density, cover and soil seedbank) and species composition of associated plant species for 5-year period from 2009 in a grassland invaded by Parthenium in south-central Nepal. We found that Parthenium cover and height decreased from 2009 to 2010 and then slightly increased in 2013. Parthenium density decreased from 2009 to 2010 and then was variable until 2013. Year × grazing interactions had significant effect on Parthenium cover and density. Parthenium soil seedbank was eight times higher near the soil surface (0–5 cm) than in deep soil (5–10 cm). It increased from 2009 to 2012 but decreased in 2013. Seedbank was also affected by interactions of year × depth, depth × grazing, and year × depth × grazing. Altogether, 87 plant species were recorded in Parthenium invaded sites and their species richness decreased until 2012 but slightly increased in 2013. The composition of associated plant species was affected by animal grazing intensity, Parthenium density, cover, and their interactions. Parthenium invasion has been ever increasing in our study site and many palatable plant species are under potential threat. Thus, there is an urgent need to carry out awareness campaign, formulate proper management plans, and implement such plans properly to manage Parthenium weed in Nepal.


Introduction
In terrestrial ecosystems, impacts of biological invasions on associated plant community structure is observed high when the invader is producer and the recipient ecosystem is grassland (Mollot et al. 2017). This condition leads to the loss of native species and economic damages (Theoharides and Dukes 2007). Invasiveness of an invader depends on persistence of seeds in the soil (Richardson and Kluge 2008). Once the status of seedbank is known it would be easy to formulate the control and management measures for invasive species. Though seedbank is difficult to destroy but reducing seed output could be achieved by using biological control agents (Kluge and Neser 1991) or destroying it mechanically, or by using chemicals (Dhileepan et al. 2018). In Nepal, seed bank of Parthenium hysterophorus L. (Asteraceae) (hereafter named as Parthenium) was investigated (Shrestha et al. 2015) but the long-term patterns are not well documented. Thus, we have investigated how the seedbank of Parthenium changes over a 5-year period in a same locality in Nepal.

International Society for Tropical Ecology
Electronic supplementary material The online version of this article (https ://doi.org/10.1007/s4296 5-020-00085 -7) contains supplementary material, which is available to authorized users.
Invasive species are known to have a wide range of ecological impacts to recipient ecosystems; these includes changes in nutrient cycling and soil properties, productivity, fire regime, habitat structure, and species composition and their relative abundance (Simberloff 2011;Vilà et al. 2011). Among them change in nutrient cycling is the most commonly reported example of ecosystem impacts of invasive species (Simberloff 2011). However, community structure (such as species richness and composition) changes long before the impact of invasive species on nutrient cycling is detected (Vilà et al. 2011). Decline in species richness due to invasion has been reported in several studies (Mollot et al. 2017). Similarly, change in species composition without significant change in species richness has also been reported (Hejda et al. 2017). The impact of invasion on community structure depends on trophic position of invader, taxonomic position of invader as well as the organism in the recipient community being considered, the ecosystem/habitat type and disturbance regime (Bulleri et al. 2016;Mollot et al. 2017). It is important to understand changes in species composition of the invaded community and growth dynamics of the invasive species itself over the time so that it would be possible to devise proper management strategies for the control of invasive plant species.
In Nepal, out of 179 naturalized flowering plant species (Shrestha 2019a), Parthenium is arguably one of the most damaging invasive weeds in grassland (Shrestha et al. 2015;Shrestha 2019b). It is also problematic in agroecosystems of Asia, Africa and Oceania (Adkins and Shabbir 2014;Bajwa et al. 2016). In Nepal, Parthenium was first reported in 1967 (Tiwari et al. 2005). Since then, the weed has been spreading continuosly and become widespread throughout Nepal upto 2000 m elevation due to absence of a rigourous control program (Shrestha et al. 2015. Although several environmental impacts of Parthenium are widely studied, the temporal growth pattern of this plant species has been merely studied (Adkins and Shabbir 2014;Bajwa et al. 2016;Navie et al. 1996). Soil nutrients, habitat quality, seed dispersal and disturbances are some of the important factors that contribute to the growth of invasive species during one of the four stages of invasion process: introduction, existence as casual aliens, naturalization and spread (Milbau and Stout 2008). Out of many factors, grazing is considered as an important factor that contributes in plant invasions (Keeley et al. 2003) and is important to see how it affects invasive plant growth during certain time period.
Shortly after establishment Parthenium weed tends to dominate vegetation in disturbed habitats of urban, periurban and natural areas (Gupta and Narayan 2006;Karki 2009;. The weed has been reported to reduce biodiversity in natural ecosystems including rangelands and protected areas (Adkins and Shabbir 2014). However, only few studies examined plant species diversity in Parthenium invaded sites using appropriate experimental designs and there was lack of consistent pattern in the impacts of Parthenium on plant species diversity of the recipient community Nigatu et al. 2010;Timsina et al. 2011). Nigatu et al. (2010 reported a decline in species diversity (Shannon's index) with increasing cover of Parthenium in grazing lands of Ethiopia. A similar decline in species diversity (Shannon's index; aboveground as well as seedbank) in quadrats with high Parthenium density was reported in a grazing grassland of Queensland, Australia . But, Timsina et al. (2011) showed higher species richness (number of species per 1 m 2 quadrat) in intermediately invaded quadrats than in the adjacent non-invaded quadrats of the grazing grasslands in Nepal. In addition to the lack of consistent patterns of the impact, the previous studies did not examine how species diversity in Parthenium invaded sites varies over time. It is highly likely that abundance of Parthenium weed as well as the species diversity of the invaded community change over time and understanding of this pattern is important for planning management of Parthenium weed. There are cases when abundance of invasive species declines ('bust' phase), even in absence of management interventions, after a period of high abundance ('boom' phase) due to processes mainly related to invasion and environmental changes (Strayer et al. 2017). To understand dynamics of Parthenium weed abundance and plant community structure in the invaded grazing grassland, we specifically addressed the following research questions: (1) How does growth of Parthenium (height, density, cover and soil seedbank density) vary between years in central Nepal? (2) How does soil seedbank of Parthenium change over a time period and vary according to soil depth? (3) How species richness and composition of different plant species change over certain time period in a grassland invaded by Parthenium? To answer above-mentioned questions, we collected the data on Parthenium density, cover, and height in different quadrats, associated plant species, grazing intensity and soil seedbank of Parthenium from 2009 to 2013 in Hetauda, south-central Nepal.

Study area
The study was undertaken at Lamsure Danda of Hetauda Municipality in Makawanpur district, south-central Nepal. The study site (27° 25′ N latitude, 85° 03′ E longitude, 500 m a.s.l.) lies in a dun valley of Siwalik region. The region has four distinct seasons: cold and dry winter (December-February), hot and dry summer (March-May), hot and humid monsoon (June-August) and warm and moist fall (September-November). Analysis of weather data between 1976 and 1 3 2005 revealed that annual precipitation was 2430 mm with monsoon rain contributing to about 82% (Practical Action Nepal 2009). Mean minimum and maximum temperatures were 16.3 °C and 29.2 °C, respectively. Shorea robusta Gaertn is the dominant tree species in forests of the region. The area lies in Tarai-Duar Savanna and Grassland ecoregion of the world (MFSC 2014).
The study grassland is a property of government owned Hetauda Cement Industries Limited but local communities have been using the grassland for grazing livestock and harvesting forage. The area was used for farming by local communities until 1977. In 1978, the area was purchased by Hetauda Cement Industries Limited for mining soil, and since then the site remained abandoned. This 'secondary' grassland has the highest species richness of the invasive alien plants among the natural vegetation in the region (Dhakal 2017). Parthenium, Lantana camara L., Senna tora (L.) Roxb., Xanthium strumarium L., Spermacoce alata Aubl. are the common invasive alien plants in this grassland while Clerodendrum viscosum Vent., Imperata cylindrica (L.) Beauv., Chrysopogon aciculatus (Retz.) Trin. are the common native species. The area was not subjected to any weed management. A few instances of slashing, uprooting and firing of Parthenium weed were observed but these activities were not common and less likely to have significant impacts on abundance of Pathenium weed in the study site. Zygogramma bicolorata Pallister (Coleoptera: Chrysomelidae), a leaf feeding beetle used as a biological control agent of Parthenium, was reported from Hetauda including the present study site in 2009 (Shrestha et al. 2010).

Parthenium survey and associated plant sampling
Sampling was carried out during August-September from 2009 to 2013. In the study area, three sites were selected that were about 150 m apart with relatively high cover (> 50%) of Parthenium. At each site, 10 square quadrats each of 1 m 2 with > 50% cover of Parthenium were defined within big ca. 20 m × 20 m plot. Altogether, there were 30 quadrats sampled each year. Vascular plants growing in each quadrat, Parthenium growth (cover, density, and height of the tallest plant) were noted down. Parthenium cover was estimated visually in percentages, Parthenium density was counted in number (plant individuals within 1 m 2 quadrat) and height was measured in centimeter. Grazing intensity in each quadrat was also estimated in the scale of 0 (no sign of livestock grazing) to 3 (vegetation removal by livestock in > 50% area of the quadrat). Each quadrat was marked by fixing a green colored polyvinyl chloride (PVC) pipe (length, 40 cm).
Associated plant species were recorded in each quadrat and voucher specimens were collected. Identification was carried out by comparing specimens deposited at National Herbarium and Plant Laboratories (KATH, Godawari, Lalitpur, Nepal) and referring 'Plant Diversity of Eastern Nepal: Flora of Plains of Eastern Nepal' (Siwakoti and Verma 1999). Nomenclature was followed from Press et al. (2000) and Catalogue of Life (Roskov et al. 2019).

Soil sampling and seedbank
Each quadrat was relocated in November with the help of fixed PVC pipe for soil sampling when most of the plants including Parthenium dispersed their seeds. Soil samples were collected with the help of soil core sampler of ten centimeters diameter from the center of each quadrat after removal of ground vegetation and plant litter (but not seeds). Soil was collected in two steps: (i) from surface up to 5 cm depth, and (ii) from 5 cm up to 10 cm depths. Altogether there were 60 soil samples collected from 30 quadrats each year. Soil samples were kept in plastic bag and brought to laboratory at Central Department of Botany, Tribhuvan University, Kathmandu for further analysis.
In each collected soil sample, seedling emergence method was used to assess the germinable soil seedbank (hereafter Parthenium seedbank) as it was difficult to extract seeds from the soil and count them (Leck et al. 1989). Collected soil samples were kept into germination for two months in the greenhouse without any controlled environment to determine Parthenium seedbank ( Supplementary Fig. 1). For this, earthen pots each of 15 cm in diameter were taken and in each pot 2/3rd parts were filled with heat sterilized sand and then soil sample was spread over it, uniformly such that depth of the soil sample was less than 2 cm. Sand and soil sample was separated by a layer of newspaper to prevent their mixing. All the pots were watered regularly with 100 ml of water for each pot. They were observed regularly for newly emerging seedlings. Number of seedlings of Parthenium that emerged in each pot were recorded weekly until the end of the experiment (i.e. two months). In each observation, seedlings were removed to avoid crowding.

Data analysis
The analyses were carried out in three steps. First, we explored the determinants of Parthenium growth (cover, density and height). Here, we used Parthenium height, cover and density as response variables, and year, grazing and their interactions were used as explanatory variables. Second, to find out determinants of Parthenium seedbank, we used number of emerged seedlings as response variable, and year, depth of soil, grazing, and their interactions as explanatory variables. Third, we assessed determinants of species richness of associated species. Here, number of different plant species growing in each quadrat (1 m 2 ) was termed as species richness. In this analysis, species richness was used as response variable and explanatory variables used were year, grazing, Parthenium cover/density/height and their interactions.
All analyses were carried out by using the generalized linear mixed model (GLMM) with Poisson distribution using lme4 function in lmerTest package in R 3.3.2 (R Development Core Team 2018). We used Poisson distribution because data were normally distributed. In all tests, the effect of each quadrat was also taken into account by using them as random factor in above mentioned models. To derive the p-value, we first used drop1 function implemented in R and then again used Chi-square test (R Development Core Team 2018). All the figures were drawn by using STATISTICA 12 (StatSoft Inc 2015). To find out variation in significant values from one another, we used Tukey's post-hoc Test in Statistica 12.
Multivariate test was used to see the effect of Parthenium cover, density and height and also grazing on composition of associated plant species recorded in different quadrats. We used Redundancy Analysis (RDA) because of a short gradient length of 2.01 (Lepš and Šmilauer 2014) by using Canoco 5.12 (ter Braak and Smilauer 2012). In RDA, model was built in the same way as it was done in GLMM. First, we tested significance of year and grazing intensity. If they were significant then the significant factor was taken as covariate. In the final model, we tested the effect of year, grazing intensity (only significant variables), Parthenium cover, height, density and their interactions on species composition of associated plant species. The significance was tested by performing Monte Carlo permutation test (n = 4999). To make the test possible, we added fictive species which was present in all the sampling quadrats. In RDA, we down weighted the rare species to reduce their effect on the results.

Results
The correlations test showed that Parthenium cover was positively correlated with Parthenium density (p < 0.001, r = 0.414) and plant height (p < 0.001, r = 0.598). From the year 2009, Parthenium cover and height significantly decreased in 2010 but increased afterward (Fig. 1a, b; Table 1). Parthenium density decreased from 2009 to 2010 and then the pattern was variable in remaining years ( Fig. 1c; Table 1). Parthenium density ranged from 15 to 703 stem m −2 whereas height ranged from 26 to 188 cm. All these growth parameters of Parthenium decreased along with increasing grazing intensity. There was significant effect of interactions of year × grazing on Parthenium cover and density in our study area (Table 1).
Seedling emergence showed that Parthenium seedbank was eight times higher in shallow soil depth than the deep soil (Fig. 2a). There were 127 to 42,675 seeds m −2 in shallow soil (0-5 cm) and no seeds to 16,815 seeds m −2 in deep soil (5-10 cm). There was significant increase in Parthenium seedbank from 2009 to 2012 but a sharp decline in 2013 (Fig. 2b). Though the seedbank was affected by year, soil depth, grazing intensity, and their interactions, the soil depth was the most important factor in maintaining seedbank ( Table 2). Interactions of year × depth is the second most important factor in maintain soil seedbank whereas other significant factors have less contributions.
Out of all 87 plant species recorded in Parthenium invaded sites, 81 species (12 species identified up to genera level) belonging to 62 genera and 21 family were identified (Supplementary Table 1); 6 species could not be identified due to absence of reproductive parts in the specimens collected during field sampling. It was found that species richness of plant species significantly decreased until 2012 but slightly increased in 2013 (Table 3; Fig. 3). Species richness of associated plant species decreased with increasing Parthenium height (Table 3; Fig. 4). There was a significant effect of year × Parthenium density on the plant species richness ( Table 3).
The multivariate analysis of species composition of associated species was significantly affected by grazing intensity, Parthenium density and cover (p = 0.039). Similarly, all two ways interactions of different factors affected species composition (Table 3).

Discussion
Long-term planning for effective management of invasive alien plants require an understanding of how species composition of the invaded community and abundance of the invasive species itself change over time. Using data collected for five years in a Parthenium invaded grazing grassland of south-central Nepal, we showed that despite annual variation the abundance of Parthenium has in general increased over time with a significant shift in species composition. In a context of widespread occurrence of this weed in Nepal , the present results indicate that the persistence of many plant species will be threatened seriously with direct negative impacts on  Relationship between Parthenium associated plant species with grazing and Parthenium plant density. The 1st canonical axis explained 6.12% and the 2nd 2.56% of the total variation in the data set biodiversity, ecosystem processes and livestock farming if management of Parthenium is not initiated immediately. The invasion of Parthenium has increased over the time in our study site and a similar situation could be expected in other regions invaded by Parthenium in Nepal ). This fact is illustrated by our observations showing increase of Parthenium cover and height of the plant individuals from 2010 to 2013. Rate of increase of invasive plant with time is also found in different studies in Nepal and around the world (Adkins and Shabbir 2014;Huang et al. 2011;Russell et al. 2017;Shrestha et al. 2015). The decrease in Parthenium cover, density and height from the year 2009 to 2010 might be due to the herbivore damage caused by a biological control agent Zygogramma bicolorata that was first reported from Hetauda in Nepal in 2009 (Shrestha et al. 2010). This beetle is considered as an effective biological control agent that feed on leaves of Parthenium and significantly reduce growth, vigor and seed production (Dhileepan et al. 2018). However, there may be year-to-year variation in the damage caused by Z. bicolorata on Parthenium weed due to variation in weather conditions (Dhileepan 2003).
The positive correlation between cover with density and height of Parthenium in our study shows that either cover or density could be used to represent one another. Parthenium density in our study (15-703 stem m −2 ) is very high compared to data available from Pakistan (Javaid and Riaz 2012). However, previous study by Karki (2009) in the same region showed a mean density of 118 stem m −2 showing a sharp increase in abundance of Parthenium over a time period. The plant height in our study (26-188 cm) was less than in highly invaded areas in Australia where they were up to 2.5 m in height although most plants do not exceed 1.5 m (Haseler 1976). Mean height estimated from different parts of Nepal is 128 cm, with maximum value reaching up to 240 cm ). It appears that the maximum height reported from Nepal and Australia are similar but it was lower in the present study site. Either edaphic or climatic factors might have prevented the individuals of the present study site to grow tall. In addition, negative impacts of defoliation by the biological control agent could not be ruled out but this was not evaluated in the present study.
Parthenium is normally avoided by cattle (Shrestha et al. 2015) but grazing intensity varied with the abundance of Parthenium. Low grazing intensity in highly invaded quadrats (i.e. with high abundance of Parthenium) suggests low visitation by cattle, probably due to low availability of forage in such quadrats. Reduction of native plant biomass nearly to 1/3 rd has been reported due to dense growth of Parthenium in three dun valleys of Nepal including Hetauda (Shrestha 2019b).
Prevalence of more seeds in shallow top soil compared to deep soil is an well-established phenomenon (Leck et al. 1989;Fenner 2000). These seeds remain viable for a very long time period and eradication of Parthenium is difficult once it starts growing in some places. Adkins and Shabbir (2014) have found that more than 70% of seeds below 5 cm remain viable for 2 year, with a half-life of 7 years. Similarly, White (1994) found seeds to be viable for 4 to 6 years and other studies Navie et al. 1998) also reported that buried seeds remain viable for much longer period than seeds on the soil surface. The impact of interactions of year × depth means that there is variability of seed bank according to the soil depth and is variable in different years. This is probably because of the plant growth and seed production that vary from one year to next.
There was high number of seed production (up to 42,675 seeds m −2 accumulated in shallow soil and 16,815 seeds m −2 in deep soil) by Parthenium in our study site. This is similar to other studies (Navie et al. 1998;Nguyen et al. 2010) but far less than the report by Joshi (1991) in India, who reported about 200,000 seeds m −2 . The increase in seedbank size could be attributed to the addition of seed to the persistent soil seed bank of Parthenium Adkins and Shabbir 2014;Navie et al. 1998;White 1994). Increasing soil seedbank from 2009-2012 was alarming, which suggests the need of immediate management to prevent further degradation of the grassland due to Parthenium invasion. Sharp drop in seed bank from 2012 to 2013 might be due to low production of seeds as the year 2013 was exceptionally dry due to low precipitation (DoHM 2015).
In Hetauda, we found high number (n = 87) of plant species in Parthenium invaded sites and many of them are palatable meaning that the site is rich in plant community. This number in our study is higher compared to other studies in Nepal (Khatri-Chettri 2012; Timsina et al. 2011), 72 species in Ethiopia (Nigatu et al. 2010), 30 associated species in Pakistan (Shabbir and Bajwa 2006) and in different part of the world (Adkins and Shabbir 2014). The sharp decrease in plant species richness from 2009 to 2012 have shown that invasion of Parthenium is ever increasing with serious adverse effect on many native species. This adversity will in near future may lead to severe degradation of habitat with loss of biodiversity (Timsina et al. 2011) and brings about changes in existing ecosystem in larger scale (Adkins and Shabbir 2014;Shrestha et al. 2015). A significant effect of year × Parthenium density on the plant species richness means that the density of Parthenium is variable in different years and its' growth probably depends on favorable temperature or precipitation and disturbance in a year.
Effects of different environmental factors on species composition of the Parthenium invaded community is in line with other studies (Adkins and Shabbir 2014;Nigatu et al. 2010;Shrestha et al. 2015;Timsina et al. 2011). Each environmental factor acts differently and thus result in differences in associated plant species composition in a given space. Taxonomic identity of plant species growing in Parthenium invaded sites differ from place to place as shown in Nepal (Timsina et al. 2011) and elsewhere in the world (Nigatu et al. 2010;Adkins and Shabbir 2014). Our multivariate analysis shows that high number of associated plant species do not prefer Parthenium invaded quadrats resulting in variation in species composition of associated plant species. However, there was no effect of time period on associated plant composition meaning that despite Parthenium invasion most of the plant species still can persist at quadrats where the abundance of Parthenium is low. This is also in agreement with our previous report that plant species richness can be higher in plots that are intermediately invaded by Parthenium than in the heavily invaded, or noninvaded quadrats (Timsina et al. 2011).
The results show that Parthenium is ever increasing in our study site at Hetauda. In addition to habitat degradation and changes in community structure, the increase in invasion severity may in future cause serious problems in health for both livestock and human due to its toxic nature Nigatu et al. 2010). A wide range of management plans have been reported recently by Adkins and Shabbir (2014). However, depending on human and economic resources available, the locally most appropriate and cost-effective methods should be used for Parthenium management. Further, as there are many important species in Parthenium invaded sites a proper competition experiment should be conducted to identify the vulnerable and competitive species in the communities. The competitive species can potentially be used as suppressive plants as a part of integrated management of Parthenium weed (Khan et al. 2019).

Conclusion
Present multi-year study in a grazing grassland at Hetauda in south-central Nepal has provided in-depth knowledge on dynamics of Parthenium weed invasion. It has been found that the weed is a serious threat to the associated plant species that are palatable and important to maintain existing ecosystem. Variation of species richness of plant species over 5-year time period and also increasing size of soil seedbank shows that Parthenium is currently a potential threat to biodiversity that is also increasing with time due to accumulation of seeds in seed bank. Thus, awareness campaign, formulation of proper management plans and their implementation is needed to avoid serious threats posed by Parthenium weed invasion in different parts of Nepal.