Endemic shrubs in temperate arid and semiarid regions of northern China and their potentials for rangeland restoration

The shrubs Amygdalus pedunculata, Amygdalus mongolica and Ammopiptanthus mongolicus are endemic species in temperate northern China. They have developed adaptations to characteristic arid and semiarid ecosystems. A. pedunculata prefers low hills and sandy land in semiarid regions. A. mongolica prefers gravel deserts in semiarid regions. A. mongolicus prefers sandy land in arid regions. They play critical roles in maintaining or restoring these ecosystems. However, they are in danger of extinction. As part of a general conservation effort, they can be used as nurse plants to facilitate vegetation establishment in the engineering of rangeland restoration.


Introduction
Rangeland desertification in semiarid and arid regions constitutes a global challenge for human welfare and social development (Dong et al. 2012;D'Odorico et al. 2013). With desertification, vegetation productivity and diversity are typically reduced, which further leads to serious disasters such as sand storms, sand drifts and sand burial (Kurosaki et al. 2011). Temperate northern China, with vast arid and semiarid rangeland, has experienced intensive desertification and associated vegetation and biodiversity loss for many years (Wang et al. 2013). Although major efforts have been launched to restore desertified rangeland, there is still disagreement about whether woody species or grasses are most appropriate for these restoration activities (Wang et al. , 2010. Some scientists propose restoring the desertified land using grasses, rather than water-consuming woody species, while others worry that it will be difficult for grasses to become established on bare dunes without initial sand-stabilization provided by woody species, such as shrubs (Yang et al. 2006;Cao 2008;Padilla et al. 2009). The original vegetation of these regions should provide insights into this controversy, given that the resident plant species have so far survived the harsh environments of these regions and have presumably adapted to them (Wilson and Roberts 2011;Szitá r et al. 2014).
The shrubs Ammopiptanthus mongolicus (Maxim. ex Kom.) Cheng f., Amygdalus pedunculata Pall. and Amygdalus mongolica (Maxim.) Ricker are endemic species, mainly distributed in arid and semiarid regions of northern China, and they have been officially designated as rare species deserving conservation efforts (Browicz and Zohary 1996;Liu 1998). These species now exist in remote locations that were never subject to severe overgrazing and desertification. They and associated plants constitute distinct communities of local original vegetation. As resident species, they may be promising candidates for rangeland restoration (Eldridge et al. 2011;Caperta et al. 2014;Kleinhesselink et al. 2014). We hypothesize that they have developed special adaptations to the conditions found in temperate arid and semiarid regions and can serve as nurse species to facilitate rangeland restoration. We therefore designed a study to investigate and analyse the diversity and structure of the original plant communities associated with these species, from which we wanted to collect factual evidence for the hypothesis.

Data collection
The study area was located in arid and semiarid regions of northern China, where annual precipitation is mostly lower than 400 mm. Population patches of the three endemic shrub species were scattered in remote flatland and hills. During July and August of 2013, we selected 25 sites for field surveys, where nature reserves or protected areas have been established ( Fig. 1). At each site, we randomly sampled three 10 × 10 m plots at intervals of 1 km. We measured higher shrubs (ca. .40 cm) individually, recorded their long and short crown diameters and calculated their crown area using the oval area formula. We classified these shrubs by species and calculated percent cover of each species, which was the total crown area of all shrubs of the same species divided by plot area. For other shrubs and grasses, we estimated percent cover of each species by counting their individual or patch numbers and measuring diameters of sample individuals or patches. All observed species were identified in the field by consulting handbooks of species identification (Lu et al. 2012). Because some traits were invisible or missing, we had to identify some plants at the genera level, belonging mainly to the families of Gramineae and Compositae. We also collected plot information such as longitude, latitude and elevation (using GPS, Garmin eTrex 30, Olathe, USA) and habitat type (i.e. sandy land, gravel land and rocky hills). Afterwards, we added the data of average annual precipitation (AnPre) and annual average temperature (AnTem) for every plot, by consulting published data for world climates (http://www.worldclim.org), covering the period from 1950 to 2000.

Statistical analysis
We analysed niches of the three endemic shrub species in relation to AnTem and AnPre, using the HyperNiche 2 software (MjM Software, Gleneden Beach, OR, USA) (Carrick 2011). The data used for the analysis were percent cover of these species and values of AnTem and AnPre. The software explores the functional niche of a species through a series of parameter-flexible models. We chose the model type of quantitative local mean Gaussian and simulated with AnTem and AnPre concurrently. The software automatically performed iterative simulations and evaluated them by adjusted R 2 . We chose the simulations with the highest R 2 as the best, where detailed significances ( p) were not reported. We also conducted an Indicator Species Analysis using PCord 6.0 (MjM Software) to compare requirements of these shrub species for habitats (Peck 2010). A higher indicator value for a habitat type represents higher values of cover and frequency in habitats of this type. We did not simulate effects of the geographical factors of longitude, latitude and elevation, because they affect vegetation diversity and distribution indirectly and are too complicated to be well simulated.
The plots were sampled from an area spanning more than 1000 km, and their species composition was very heterogeneous. We found 184 species and 110 genera in these plots, many of which appeared only in one or two plots. We classified all observed species by families and calculated sum cover of every family in every plot, which was the sum of percent cover of all species of the same family. We sorted all the plots using PCord 6.0 with the non-metric multidimensional scaling (NMS) method, where vegetation data were the sum cover of every family and environmental data included AnPre, AnTem, longitude, latitude, elevation and habitat type (Peck 2010). The plots were categorized subjectively by the three endemic shrub-species.
All species were classified into three functional types according to life forms, i.e. short-lived grasses with longevity of 1 or 2 years, long-lived grasses with longevity .2 years and shrubs (including semi-shrubs) normally with longevity .2 years. We counted species number (richness) of every functional type in the plots. Based on species numbers, we calculated species ratios of each functional type in the plots; a species ratio of a functional type in a plot was species number of the functional type in the plot divided by total species number of all functional types in the plot. Then, we performed one-way ANOVA via Origin 8.0 (OriginLab, Northampton, MA, USA) to check whether species ratios of every functional type differed among the three plot-groups, which corresponded to the three endemic shrub-species. In addition, we evaluated how many species are inclined to co-occur with the three endemic shrub species by comparing their co-occurrence rates with these endemic shrub-species. The rates were times that a species co-occurred with an endemic shrub species divided by number of the plots where the pertaining endemic shrub-species occurred.

Niches and habitats of the endemic shrub-species
The best simulations with the highest R 2 demonstrate niches of the three endemic shrub species for climates (Fig. 2). Optimal conditions for the species are discriminative. Amygdalus pedunculata grows best where annual average temperature is ,4 8C and annual precipitation is more than 350 mm, with relatively higher cover than on other conditions. Amygdalus mongolica grows best where annual average temperature is 5.5 8C and annual precipitation is 280 mm. Ammopiptanthus mongolicus grows best where annual average temperature is more than 6 8C and annual precipitation is ,200 mm. These shrub species are also discriminative in terms of habitat requirements. Amygdalus pedunculata prefers low hills and sandy land, A. mongolica prefers gravel deserts and A. mongolicus prefers sandy deserts (Table 1).

Main families and their distributions
The NMS ordination plot shows that AnTem and AnPre are the key factors determining vegetation structures and distributions, as indicated by their longer arrows (Fig. 3). In contrast, longitude and elevation play subordinate roles, perhaps because they are indirect factors to reduce or distort influences for intervening factors or processes. The ordination plot also reflects that: communities of A. pedunculata (CBBT, Chang-Bing-Bian-Tao in Chinese) tend to appear in relatively colder and wetter conditions, with higher plant diversity; those of A. mongolicus (SDQ, Sha-Dong-Qing in Chinese) tend to appear in warmer and drier conditions, with lower plant diversity and those of A. mongolica (MGBT, Meng-Gu-Bian-Tao in Chinese) often appear in transitional conditions between the communities of CBBT and SDQ, yet exhibiting higher similarity to CBBT communities and lower similarity to SDQ communities.

Functional structure of plant communities
The life form structures vary with community or plot types (Fig. 4). Species ratios of long-lived grasses are the highest in CBBT communities and the lowest in SDQ communities (F 2 , 72 ¼ 28.02, P , 0.001). Those of shrubs (including semi-shrubs) are the highest in SDQ communities and the lowest in CBBT communities (F 2 , 72 ¼ 20.20, P , 0.001). The features of MGBT communities are not as noticeable. This finding demonstrates that shrub species possess better potential to survive arid conditions than long-lived grasses. Especially, species ratios of short-lived grasses behave insensitively to arid and semiarid environments, i.e. show no significant differences among these community types (F 2 , 72 ¼ 0.14, P ¼ 0.87).

Responses of shrubs and grasses to environments
The species ratios of shrubs (including semi-shrubs) increased from semiarid (CBBT communities) to arid conditions (SDQ communities), entirely opposite those of long-lived grasses and also unlike those of short-lived grasses. This reflects that native shrubs exceed grasses to endure arid conditions; that is, they have developed better adaptations to arid environments. For example, A. pedunculata grows well where annual precipitation is more than 350 mm, A. mongolica grows well where annual precipitation is 280 mm and A. mongolicus grows well where annual precipitation is ,200 mm. Moreover, there are many other drought-enduring shrubs and semi-shrubs, such as Nitraria spp., Zygophyllum xanthoxylum, Artemisia frigida, Artemisia ordosica and Reaumuria songarica. In contrast, long-lived grasses are inclined to occur in semiarid conditions, presenting higher species ratios in semiarid CBBT communities than in arid SDQ communities. Plant diversity is found to be higher in semiarid communities, but is mainly attributed to long-lived grasses, rather than shrubs. Short-lived grasses do not show dependence on either arid or semiarid conditions, with approximate species ratios in SDQ, MGBT and CBBT communities. The distinctive responses of plants in our system are likely explained with their life forms (Schulze et al. 1996;Throop et al. 2012;Venn et al. 2014). Grasses normally have most of their roots shallower than 20 cm, whereas shrubs have only part of their roots at such a shallow depth and allow many more roots to grow in soil deeper than 20 or 30 cm (Schulze et al. 1996;Yang et al. 2014). This difference determines their survival potentials. In semiarid regions, annual precipitation ranges from 200 to 400 mm, and in arid regions, it can be lower than 200 or 100 mm. With decreasing precipitation, sunny days increase and ground evaporation intensifies. Such evaporation takes place primarily in and near ground surface, and scarcely at depths .20 cm (Yuan et al. 2008). As a result, the surface soil from semiarid to arid regions turns increasingly drier, and accordingly becomes more intolerable for long-lived grasses. The situation is not so serious for many native shrubs because their roots can easily reach deeper than 20 or 30 cm, where snow melt and heavy showers are adequate to penetrate for storing and yet avoid ground evaporation (Berndtsson et al. 1996). Relying on such moisture storage, native shrubs can survive better than long-lived grasses, that is, stand vigorously to maintain an ecosystem (Schulze et al. 1996). As for short-lived grasses, because their roots are also shallow and their lives are transient, they usually act as wanderers among long-lived grasses and shrubs, seeking new spots for transient establishment (Bowers et al. 1995). Their occurrences depend mostly on uncertain chances, unlike long-lived grasses and shrubs, which compete for soil moisture at different depths through special roots and persistent endeavours. For this reason, species ratios of short-lived grasses remain approximate in different communities, seemingly independent of environmental aridity.

Integration of rangeland restoration and species conservation
Now that native shrubs can exceed grasses to survive arid conditions, some of them may be qualified as nurse species to facilitate recovery of arid and semiarid ecosystems (Yang et al. 2006;Zhao et al. 2007;Kleinhesselink et al. 2014). After these ecosystems suffer serious desertification, their vegetation-free surfaces usually become sensitive to wind erosion and sand drifts (Li et al. 2004;Xu 2006). Plants have difficulty colonizing such bare ground, except for a few annuals that grow only on favourable days (Yang et al. 2006). Every spring, winds are particularly strong and ground surface is very dry, thus causing serious ground erosion, sand drifts and sand burial. In this case, grasses may be dug out and killed unless their cover is high enough to tightly stabilize ground sand (Yang et al. 2006). Relative to grasses, shrubs embed their roots much deeper, achieving more chances to survive. They therefore survive better than grasses by protecting most roots from exposure to the sun (Wu and Yang 2013). Sand burial is also common in arid areas, but is not as harmful as sand erosion, and may be favourable for seed germination and plant establishment (Shi et al. 2004). Branches and stems of pioneer shrubs can slow winds and stabilize sand, thereby facilitating sand and seed deposition around them (Yang et al. 2006; Zhang et al. 2013). This condition is also beneficial for seed germination and seedling growth (Shi et al. 2004;Zhao et al. 2007;Zhang et al. 2013). As a result, grasses and other shrubs can follow such pioneer shrubs to become increasingly established.
We suggest that the three endemic shrub-species, as native rarities already adapted to temperate arid and semiarid regions, are potential candidates for restoration projects in degraded rangeland systems (Zhao et al. 2007;Kleinhesselink et al. 2014). Amygdalus pedunculata is suitable for semiarid regions. Amygdalus mongolica is suitable for semiarid gravel regions. Ammopiptanthus mongolicus is suitable for arid sand regions. The seedling stage is a critical period for these species to become established because seedlings are highly sensitive to climatic events and damage by livestock (Padilla et al. 2011;Li et al. 2013). Once their seedlings get through this stage, they can grow to stabilize sand ground and facilitate subsequent establishment of more plants (Chillo et al. 2011;Le Stradic et al. 2014). In particular, as rare species deserving conservation, if they are chosen for rangeland restoration, conservation efforts for them are greatly advanced. On one hand, their growth area can be enlarged, so that they are no longer restricted to nature reserves and protected areas. On the other hand, people can benefit from these efforts, at least ensuring the process of rangeland recovery.

Conclusions
The endemic shrubs of A. pedunculata, A. mongolica and A. mongolicus have developed adaptations to arid or semiarid environments. They are promising nurse species to initiate rangeland restoration, by facilitating sand stabilization and plant establishment. Because these endemics are also unique and precious, people may become enthusiastic to grow them in their degraded rangelands, as long as official encouragement is available.