Wind-dispersed Seed Deposition Patterns and Seedling Recruitment of Artemisia halodendron in a Moving Sandy Land

Abstract

• Background and Aims Artemisia halodendron is a native sub-shrub that occurs mainly in moving and semi-fixed sandy lands in Inner Mongolia, China. Information on the spatial patterns of wind-dispersed seed deposition and seedling recruitment of A. halodendron inhabiting moving sandy lands is very limited. The aim of this study was to examine wind-dispersed seed deposition patterns and post-dispersal recruitment of A. halodendron seedlings.

•Methods The spatial patterns of wind-dispersed seed deposition and seedling recruitment of A. halodendron were examined by investigating the numbers of deposited seeds, emerged and surviving seedlings using sampling points at a range of distances from the parent plant in eight compass directions for two consecutive growing seasons.

•Key Results Wind-dispersed seed deposition showed considerable variation between directions and years. Wind transported A. halodendron seeds only a few meters away from the parent plant in all eight directions. Seedling emergence and establishment also showed between-direction and between-year variability, but the spatial pattern of seedling distribution differed from that of seed deposition. Only a very small fraction (<1 %) of the deposited seeds emerged in the field and survived for long enough to be included in our seedling censuses at the end of the growing season.

•Conclusions The spatial variation in wind speed and frequency strongly affects the pattern of seed deposition, although the variation in seed deposition does not determine the spatial pattern of seedling recruitment. Seeds of A. halodendron are not dispersed very well by wind. The low probability of recruitment success for A. halodendron seedlings suggests that this species does not rely on seedling recruitment for its persistence and maintenance of population.

INTRODUCTION

The Horqin Sandy Land of eastern Inner Mongolia is one of the four well-known sandy areas in northern China (Li et al., 2003a). Over the last few decades, the areas being affected by desertification in this region have been spreading at an annual average rate of about 2000 km² per year, primarily due to poor land management, such as over-cultivation (conversion of grassland to farmland), intensive firewood collection and overgrazing (Wang et al., 2003). The consequences of land desertification are a substantial loss of the region's biodiversity and productivity, as well as a range of negative environmental and socio-economic impacts (Jiang et al., 2003). For example, once the vegetation in sandy grasslands has been destroyed by overgrazing or other human activities, the soils become less stable and highly erodable (Li et al., 2003a, 2005a), which, in the long-term, will result in a new source of material that could contribute to the dust-storms that occur frequently in the arid and semi-arid lands of northern China (Xuan et al., 2000). Efforts have been made since the late-1970s (particularly in recent years) to combat desertification in the Horqin region. One of the most effective practices has been to plant native species that are adapted to a sandy land environment on severely degraded land (T. H. Zhang et al., 2004). However, the success of this technique is strongly dependent on the selection of the right species for re-vegetation.

Artemisia halodendron is a native sub-shrub that occurs primarily on moderately-to-severely degraded semi-fixed and moving sandy lands in the Horqin Sandy Land and is often the dominant species in these grassland types (Li, 1991). Due to its low nutrient requirements and high capacity to produce offspring through vegetative propagation under conditions of sand burial, A. halodendron easily colonizes bare patches to establish populations in unstable, nutrient-poor moving sandy lands (Li and Zhang, 1991; Chang et al., 1994; Li et al., 2002). Thus A. halodendron, as an ideal sand-fixing plant, has been widely used for re-vegetation on severely degraded sandy land. The value of this species is that once a population has been established, it will act both as a seed accumulator by physically trapping dispersing seeds (Aguiar and Sala, 1997; Pugnaire and Lázaro, 2000) and as a sink for resources, either actively through root uptake of soil water and nutrients (Hook et al., 1991; Gutiérrez et al., 1993; Burke et al., 1995) or passively by accumulating wind-blown dust and litter (Barth and Klemmedson, 1982; Garner and Steinberger, 1989) in a feedback mechanism that facilitates invasion and colonization by other plant species under or near its canopy. This results in an increase in biodiversity of the moving sandy land ecosystem (Pugnaire et al., 1996; Li et al., 2003b).

A number of recent studies of A. halodendron have examined its ecological and morphological attributes (Li and Zhang, 1991), its variation patterns of reproductive allocation under different habitats (Li et al., 2005b), its distribution patterns of populations in differently degraded sandy lands (Chao et al., 1999), and its mechanisms of physiological adaptation to the harsh sandy land environments, which are characterized by frequent drought, high temperature and sand burial (Wang and Zhou, 1999; Zhou, 1999; Zhou et al., 1999). However, few studies have examined the combination of the spatial patterns of seed deposition by wind dispersal of A. halodendron and its post-dispersal germination and seedling establishment. The aim of this study was to address the following: (1) the spatial pattern of wind-dispersed seed deposition; (2) the consequences of post-dispersal seedling recruitment, including seedling emergence and establishment; and (3) to determine if there is spatial concordance between seed deposition and seedling distribution (i.e. does the spatial variation in seed deposition determine recruitment pattern?).

MATERIALS AND METHODS

Study site and experimental species

The study was conducted from 2002 to 2003 on a natural moving sandy land at Shelihu in Naiman County, Inner Mongolia Autonomous Region, China (42°55′N, 120°44′E; altitude approx. 360 m a.s.l.), 520 km north-east of Beijing (Li et al., 2003a). Naiman is situated in the southern part of the Horqin Sandy Land, which is roughly 400 × 400 km in size and represents one of the worst examples of land desertification in northern China (Andrén et al., 1994). The landscape in this area is characterized by sand dunes alternating with gently undulating lowland areas. The soils are sandy, loose in texture, and particularly susceptible to wind erosion (Li et al., 2003a). The climate is temperate, semi-arid and continental, receiving 360 mm annual mean rainfall, with 75 % of this falling in the June–September period. Highly variable rainfall in the growing season (April–September: coefficients of variation 75 %, 54 %, 51 %, 60 %, 51 % and 124 %) results in unreliable plant growth and development in most years. Climatic conditions in both years of the study were dry, with 2002 receiving only 58 % (207 mm) of the long-term average and 2003 receiving 69 % (249 mm). There is little rain or snow in the late autumn–early spring period and this is coupled with frequent and high winds. Hence the period from late October to May has been shown to be the major wind-erosion season (Li et al., 2003b), as well as the major dispersal season for plant seeds (diasporas). Based on wind data (1997–2002) from the weather station of the Naiman Desertification Experiment Station, the annual mean wind speed ranges between 3·4–4·1 m s⁻¹ at 2 m height, and the prevailing wind directions over the erosive season are S, SSW, SW, NNW, WNW, NW, N and NNE. These winds occur with high speeds in excess of 4 m s⁻¹ at 2 m height (a threshold wind speed to initiate sand movement; H. Zhang et al., 2004) in most days of the erosive season (Fig. 1).

Fig. 1.

Wind speed and frequency (the number of hourly measurement points at which wind was blowing from that direction) for 16 compass directions over the potential dispersal period from late-October to May between 1997 and 2002 (mean ± s.e.). (A) Based on the data set excluding winds that occurred with a speed less than 4 m s⁻¹ at 2 m height. (B) Based on the complete data set. N = 350–11°, NNE = 12–34°, NE = 35–56°, ENE = 57–79°, E = 80–101°,ESE = 102–124°, SE = 125–146°, SSE = 147–169°, S = 170–191°, SSW = 192–214°, SW = 215–236°, WSW = 237–259°, W = 260–281°, WNW = 282–304°, NW = 305–326°, NNW = 327–349°.

The study area was originally a grass-dominated steppe community with sparsely distributed woody species (mainly elm, Ulmus spp.). When the study was initiated, the original vegetation had been substantially degraded, primarily due to overgrazing by livestock (Li et al., 2000). Degraded grassland is generally classified into three main forms: fixed or stabilized (light degradation), semi-fixed or semi-stabilized (moderate degradation) and moving or unstabilized (severe degradation), and these three forms represent a realistic range of historical grazing activities and impacts (Zhao et al., 2003). The severely degraded moving sandy land is characterized by a very low vegetation cover dominated by annual plant species. The most abundant species are the annual chenopods Corispermum macrocarpum Bge., Salsola collina pall. and Agriophyllum squarrosum (L.) Moq., and the annual grass Setaria viridis (L.) Beauv., which accounted for over 90 % of the total cover.

At the study site, Artemisia halodendron Turcz. ex Besser (Asteraceae) is distributed primarily in moving and semi-fixed sandy lands, but rarely in fixed sandy lands. It is a deciduous sub-shrub, with well-developed rhizome system. In general, A. halodendron flowers in early July and sets seed in early August, and seed matures in early October (Li and Zhang, 1991). After ripening, A. halodendron seeds (achenes) naturally fall on the ground around the parent shrubs. According to our estimation from sampling 60 individual adult plants, mean seed production was 29 g per plant and mean thousand-seed weight was 6·1 g ± 0·21 (±s.e.).

Experimental design and data collection

The experiment was conducted on a natural moving sandy land that had been protected from livestock grazing since 2000. The experimental site is open and level, covering an area of about 30 ha. In early spring 2002, six individual adult A. halodendron plants of similar sizes (plant height/canopy diameter: 70/278 cm, 84/310 cm, 85/290 cm, 63/241 cm, 64/269 cm, 67/281 cm) were chosen to be ‘target plants’. Each of the six target plants was growing at least 30 m away from any other A. halodendron shrubs, to ensure seed deposition patterns around each target plant were not being affected by seed rain from other plants. To determine the spatial patterns of wind-dispersed seed deposition, sampling lines were set up along eight compass directions (at 45° intervals, i.e. north, north-east, east, south-east, south, south-west, west, north-west) centred on each target plant. In each of the eight directions, six sampling points were placed at distances of 0·5, 1, 2, 3, 4·5 and 6 m away from the target plant.

Measurements of densities of deposited wind-dispersed seeds were made during the first week of April 2002 and 2003, before the seed rain and germination period. Thus, all viable seeds found in the soil can reliably be assumed to have originated from previous years, and all had passed through at least one winter season with the cold stratification needed to break dormancy (Baskin and Baskin, 1998). At each of the six sampling points for each direction, a soil seed bank sample of 20 × 20 cm and 5 cm deep was collected. This sampling depth was chosen for two reasons. One is that most A. halodendron seeds are retained at this depth, based on a previous field study (Li et al., 2003c). The other is that almost no seedlings emerged when A. halodendron seeds were buried at a depth of >5 cm according to a laboratory germination experiment (L. Y. Zhao, unpubl. res.). The collected soil samples were germinated in plastic trays. The trays were first filled with seed-free fine loam about 70 mm deep, and then the soil samples were spread to form a uniform, thin layer (4–6 mm) and covered with 1–2 mm of seed-free fine sand. All trays were placed in an unheated greenhouse and the viable seed density estimated from counts of seedlings that emerged over the following 12-week period. The seedlings were counted at 3–4 d intervals (Zhao et al., 2003). Interpretation of seed deposition patterns requires caution as variation in deposited seed density may include both variations in pre- and post-dispersal seed losses from the soil.

To examine the consequences of post-dispersal seedling recruitment, we investigated the number of A. halodendron seedlings that emerged at each of the six sampling points of each direction in both years. For each year, seedling censuses were made at two dates: one at the end of June (representing mid-stages of the growing season) and again at the end of August (representing late stages of the growing season), using 1 m² quadrats that were placed in close vicinity to the sampling points of the seed bank. To assess the success of recruitment, the recruitment process was divided into two distinct phases: seed to emergent seedlings, and emergent seedlings to established seedlings. The first phase includes the process of emergence, and the probability of emerging success (the mean number of seedlings per viable seed) is described as the emergence, which was calculated as a percentage of the cumulative number of emerged seedlings over a period from late-April to the end of June across all six sampling points of each direction relative to the cumulative number of deposited seeds in the 5 cm soil layer. The second phase, including the process of post-emergence establishment and the probability of establishing success (the mean number of established seedlings per emerged seedling), is described as the seedling survival, which was calculated as a percentage of the cumulative number of established seedlings at the end of the growing season relative to the cumulative number of emerged seedlings. The overall recruitment success relies not only on emergence success but also establishment success, so the probability of recruitment success can be described by the recruitment index (RI), which is decomposed into the product of emergence and survival.

Data analysis

The repeated measures analysis of variance (ANOVA) of the general linear model was performed to test for the effects (using a random model) of year, direction and their interactions on the number of deposited seeds at each of the six sampling points. The repeated measures ANOVA was also used to test for the effects of year, direction and their interactions on the cumulative numbers of deposited seeds, emerged seedlings and established seedlings, as well as on emergence, survival and RI. Differences between directions were compared using Tukey's tests and the significance of differences between the two years of the study was determined using paired t-tests. When a significant interaction between year and direction was detected for a response variable, we analysed the effect of direction within years and the effect of year within directions. Regression analysis was used to determine the relationship between deposited seed density and dispersal distance. To explore the relationship between seed deposition pattern and seedling distribution pattern, linear regression analysis was also performed to test whether deposited seed density affected emergence. The data were tested for normality using the criteria of skewness (Webster, 2001). A log(n + 1) transformation was performed for the numbers of deposited seeds, and emerged and established seedlings, and an arcsine square-root transformation was used for emergence, survival and RI prior to analysis. However, untransformed values are presented in the text.

RESULTS

Spatial patterns of wind-dispersed seed deposition

Although the cumulative number of deposited seeds over all sampling points and directions was similar between years (mean ± s.e.: 1262 ± 154 vs. 972 ± 88 seeds m⁻² in 2002 and 2003, respectively; F_1,95 = 0·05, P = 0·3094), the cumulative number of deposited seeds over sampling points and years varied between directions (F_7,95 = 6·16, P < 0·0001). The cumulative number of deposited seeds was significantly higher in the south-west than in other directions, except for the south-east and the west (Fig. 2). There was a significant year–direction interaction (F_7,95 = 8·73, P < 0·0001) on the cumulative number of deposited seeds. In 2002, the maximum seed deposition density was noted in the south-west and the minimum deposition density in the north, whereas the maximum seed deposition density was found in the north and the minimum deposition density in the south in 2003 (Fig. 2).

Fig. 2.

Cumulative number of deposited seeds across all six sampling points for each direction in each of the two years, and the combined data (mean ± s.e.). Different letters indicate a significant difference between directions (P < 0·05 from Tukey's tests). ^*Indicates a significant difference between means of the two years (P < 0·05 from paired t-tests).

Over 90 % of the dispersed seeds were deposited within 2 m of the parent plant but deposition densities were very low beyond 2 m when averaged across years (Fig. 3). Similar results were observed when the data were analysed for each year, despite some differences in the spatial pattern of seed deposition between the two years (data not shown). Regression analyses showed a linear decrease of seed deposition density with increasing distance in all eight directions, although there were some differences in the relative rate of decrease between directions (Fig. 4).

Fig. 3.

Relative density (% of cumulative number of deposited seeds) and absolute number of deposited seeds at each of the six sampling points at varying distances from the parent plant in eight directions (mean ± s.e.). Data from the two years are combined. N = 350–11°, NE = 35–56°, E = 80–101°, SE =125–146°, S = 170–191°, SW = 215–236°, W = 260–281°, NW = 305–326°.

Fig. 4.

Relationship between deposited seed density and dispersal distance from the parent plant for eight different directions. Data from the two years are combined. N = 350–11°, NE = 35–56°, E = 80–101°, SE = 125–146°, S = 170–191°, SW = 215–236°, W = 260–281°, NW = 305–326°.

Spatial patterns of seedling emergence and establishment

Although seedling emergence in A. halodendron commences in late April, a major emergence period occurs between May and June (according to our observations). The cumulative emergence of seedlings over a period from late-April to the end of June varied significantly between years (F_1,95 = 139·34, P < 0·0001), with a 10-fold higher cumulative emergence count in 2003 (58 ± 6 seedlings m⁻²) than in 2002 (6 ± 0·7 seedlings m⁻²). A significant difference in cumulative emergence of seedlings was also observed between directions (F_7,95 = 4·38, P = 0·035), with a significantly higher cumulative number of emerged seedlings in the north-east than in other directions except for the east, south-east and north-west when averaged across years (Fig. 5).

Fig. 5.

Cumulative numbers of emerged seedlings over a period from late-April to the end of June, and surviving seedlings at the end of the growing season across all six sampling points for each direction (mean ± s.e.). Data is presented for each of the two years, and for both years combined. Different letters indicate a significant difference between directions (P < 0·05 from Tukey's tests). *Indicates a significant difference between means of the two years (P < 0·05 from paired t-tests). N = 350–11°, NE = 35–56°, E = 80–101°, SE = 125–146°, S = 170–191°, SW = 215–236°, W = 260–281°, NW = 305–326°.

The spatial pattern in seedling emergence varied between years (significant year–direction interaction: F_7,95 = 2·44, P = 0·0254). In 2002, the cumulative emergence counts were significantly higher in the north-east than in other directions except for the east and south-east, whereas in 2003 the corresponding value was significantly higher in the north-east, east and south-east than in other directions except for the north-west and north (Fig. 5). The spatial pattern of seedling distribution differed remarkably from that of seed deposition. Seed deposition density was lowest in the north-east while the cumulative emergence of seedlings was highest in this direction. Likewise, seed deposition density was highest in the south-west, but the cumulative emergence of seedlings was lowest in this direction (Figs 2, 5).

Post-emergence establishment of seedlings also varied between years (F_1,95 = 14·94, P = 0·0062), with a significantly higher cumulative number of surviving seedlings at the end of the growing season in 2003 (9·2 ± 1·4 seedlings m⁻²) than in 2002 (1·3 ± 0·2 seedlings m⁻²). Although no statistical difference was found in cumulative survival counts between directions (F_7,95 = 0·81, P = 0·6145), the cumulative number of surviving seedlings was highest in the north-east and lowest in the south-east when averaged across years (Fig. 5). There was also a significant interaction between year and direction on seedling survival (F_7,95 = 9·31, P < 0·0001), indicating that the effect of year on this variable varied between directions (Fig. 5).

Recruitment success and the effect of seed deposition density on recruitment

Although emergence was significantly higher (F_1,95 = 166·22, P < 0·0001) in 2003 (8·1 ± 1·1 %) than in 2002 (0·8 ± 0·1 %), seedling survival was similar in both years (31·5 ± 5·6 % vs. 26·3 ± 4·1 % in 2002 and 2003, respectively). There were striking differences in emergence and survival between directions (F_7,95 = 6·23, P < 0·0001 for emergence and F_7,95 = 6·99, P < 0·0001 for survival; Fig. 6A, B). Both emergence and survival showed a significant interaction between year and direction (F_7,95 = 2·20, P = 0·0426 for emergence and F_7,95 = 5·16, P = 0·0023 for survival), indicating that the variation observed in emergence and survival between directions varied between years (Fig. 6A, B).

Fig. 6.

(A) Emergence, (B) survival and (C) recruitment index (RI) of Artemisia halodendron seedlings across directions and years (mean ± s.e.). See Methods for definition of RI. Different letters indicate a significant difference between directions (P < 0·05 from Tukey's tests). *Indicates a significant difference between means of the two years (P < 0·05 from paired t-tests).

Recruitment index (RI) varied significantly between years (F_1,95 = 61·42, P < 0·0001). In 2002, only 0·2 % of the deposited seeds emerged in the field and survived for long enough to be included in our seedling census at the end of the growing season, compared with 1·6 % in 2003 when averaged across directions. There was also a considerable difference in RI between directions (F_7,95 = 5·76, P < 0·0001), varying from 0·1 % in the south-east to 2·6 % in the north-east when the data for the two years were analysed together (Fig. 6C). A significant interaction between year and direction was found for RI (F_7,95 = 3·38, P = 0·0033). In 2002, no statistical difference was observed in RI between directions, but RI was significantly higher in the north-east than in other directions except for the east in 2003 (Fig. 6C).

Regression analyses indicated a negative relationship between seed deposition density and emergence in both years tested. A negative relationship between seed deposition density and emergence was also detected in six of the eight directions, and in particular the relationship was much stronger in the south-east and south-west, which had the highest seed deposition densities (Fig. 7).

Fig. 7.

Effects of deposited seed density on seedling emergence, which was calculated as a percentage of the cumulative number of emerged seedlings relative to the cumulative number of deposited seeds. The data were analysed for each direction combined over the two years (n = 12) and for each year combined over the eight directions (n = 48).

DISCUSSION

Wind-dispersed seed deposition patterns

This study indicated that the local deposition pattern of wind-dispersed A. halodendron seeds was strongly affected by the spatial variation in wind speed and frequency over the dispersal season. In the study area the potential dispersal season occurs during the period from late-October (after seeds mature) to the end of May in the following year (Zhao, 2004), which coincides with the erosive season (Li et al., 2003a). A significantly higher density of deposited seeds was found in the south-western and south-eastern aspects than in the northern, north-eastern and southern aspects when averaged across the two years (Fig. 2). The spatial variation observed in seed deposition can largely be explained by the spatial variation in the frequency and speed of the prevailing winds over the dispersal season (Fig. 1). Our study also showed a significant difference in the spatial pattern of seed deposition between the two years tested by finding that far more seeds were deposited in the south-western aspects in 2002 and in the northern aspects in 2003 (Fig. 2). Although the present year study cannot explain the cause of the year-to-year variation, we believe that this between-year variability in seed deposition is most likely to be a result of the combined effects of a range of factors including annual wind conditions, seed production, seed rain input, seed germination, secondary tumble dispersal, and pre- and post-dispersal seed losses. Further research is needed to test the effects of individual factors and the interactive effects of different factors on the spatial and temporal pattern of seed deposition in A. halodendron.

The results have also shown large differences in dispersal distances between directions. These differences might reflect variation in wind speed with direction. Many studies have suggested that wind dispersal distance is highly dependent on wind speed (McEvoy and Cox, 1987; Greene and Johnson, 1989; Yang and Zhu, 1995; Bullock and Clarke, 2000). The fact that wind transported A. halodendron seeds only a few meters away from the parent plant suggests that this species is not dispersed very well by wind because of its relatively large seed mass. ‘Wind dispersal’ in this study refers to a mixed effect of primary wind dispersal (i.e. the ‘flight’ itself) and secondary tumble dispersal, because we were unable to distinguish between the two. In a recent study, Tackenberg et al. (2003) reported that some Artemisia species are not dispersed very well by wind, consistent with our results. Other studies have demonstrated differential wind dispersal potentials for different plant species because of different terminal velocities, resulting primarily from the great differences in seed size, weight, shape and surface roughness (e.g. Bond, 1988; Matlack, 1992; Greene and Johnson, 1993; Benkman, 1995; Lisci and Pacini, 1997; Fort and Richards, 1998; Tackenberg et al., 2003). A recent study by Vander Wall and Longland (2004) has suggested that for many plant species the dispersal of their seeds (propagules) is a complex, multi-step process involving a range of alternative dispersers, with wind dispersal as only the first step of the whole dispersal process. In this study, we only examined the spatial pattern of seed movement away from the parent plant that is the result of initial dispersed by wind; the secondary dispersal pattern of A. halodendron seeds by seed-caching rodents and birds, or other dispersal agents remain unexplored.

Post-dispersal seedling recruitment patterns

There was a significant difference in either seedling emergence or post-emergence establishment between the two years tested, providing evidence that the process of recruitment was strongly affected by annual climatic conditions. In this study, the significantly greater cumulative emergence and survival in 2003 than in 2002 could largely be attributed to the higher rainfall in 2003 (249 mm) than in 2002 (207 mm), which might result in a different soil moisture environment for emergence and establishment of A. halodendron seedlings in the two years. The importance of the abiotic soil environment (e.g. water and nutrient levels) in determining the performance of individual plants and recruitment success of plant populations has been well documented by many investigators (e.g. Winn, 1985; Aguilera and Lauenroth, 1995; Bisigato and Bertiller, 1999; Wijesinghe et al., 2005).

The results also demonstrated the between-direction variability in cumulative emergence of seedlings. The cumulative seedling counts were more than four times higher in the north-eastern aspects than in the western aspects (which had the lowest cumulative emergence) when the data for the two years were analysed together. However, the spatial pattern of seedling distribution was found to differ from that of seed deposition, suggesting that there was no spatial correspondence between seed deposition, seedling emergence and establishment; or in other words the spatial variation in seed deposition did not determine recruitment pattern. An explanation for this result may be that A. halodendron seeds were deposited much more in the south-western aspects because of the wind direction, whereas seedlings emerged more in the north-eastern aspects, most likely because of the higher soil moisture. Therefore, the observed between-direction variability in seedling distribution could partly be related to soil resource heterogeneity between different directions. However, our analysis of the relationship between seed deposition density and emergence provided a further explanation for this result. In these analyses, a significant negative relationship was found between seed deposition density and emergence in both years and in six of the eight directions (Fig. 7), suggesting a negative effect of deposited seed density on emergence of seedlings. In addition, the negative relationship between seed deposition density and emergence was far stronger in high seed deposition directions (e.g. south-western and south-eastern aspects) than in low seed deposition directions (e.g. northern, north-eastern and north-western aspects, Fig. 7). A possible explanation may be that relative to the directions with lower seed deposition densities, those with higher seed deposition densities had much lower rates of emergence, most likely because of the stronger below-ground (seed-to-seed) competition for resources such as soil moisture and nutrients. All these results emphasize the importance of negative density-dependent regulation in determining seedling distribution pattern. Negative density dependence for emergence of seedlings has been documented by many studies in both natural plant communities and artificially constructed assemblages of species (e.g. Linhart, 1976; Fowler, 1986; Murray, 1998; Goldberg et al., 2001; Lortie and Turkington, 2002).

Another important result emerging from the present study was the low probability of recruitment success of A. halodendron seedlings. Overall, only a very small fraction (<1 %) of the deposited seeds in the top 5 cm of soil germinated or emerged in the field and survived for long enough to be included in our seedling census at the end of the growing season. The low recruitment rate, resulting primarily from a low emergence rate and a high post-emergence mortality rate, might suggest that A. halodendron persists and maintains its population by not relying on seedling recruitment in moving sandy lands. A recent study by Li et al. (2005b) has suggested that sexual reproduction is relatively favoured in A. halodendron plants inhabiting less-eroded semi-fixed sandy land, whereas vegetative propagation is the most important recruitment mechanism for A. halodendron plants inhabiting highly erodable mobile sandy land. Although the mechanism behind the lower recruitment success of A. halodendron seedlings remains unclear, the following factors could be important. Firstly, the moving sandy land in which A. halodendron is growing is characterized by a highly erodable environment, which may lead to seeds landing on ground where they are more likely to be buried by moving sand. Successful germination of the seeds would be more difficult following burial at a significant depth. In a laboratory germination test using a range of burial depths under controlled water conditions, Zhao (2004) found that the germination rate of A. halodendron seeds declined with increasing depth and virtually no seedlings were able to emerge from a depth of >5 cm. Secondly, low and highly variable rainfall in the May–June period (a major seedling emergence period) may result in low emergence success. Thirdly, a high post-emergence mortality rate of juvenile seedlings could primarily be accredited to desiccation during dry spells over the growing season. The study area has a strong continental climate with highly variable rainfall over the growing season (Li et al., 2003b); hence short periods of drought occur frequently. Furthermore, because of the low vegetation cover on moving sandy lands, the rain-wetted surface layer would soon dry out when exposed to sunshine and high winds (Li et al., 2002). Consistent with this view, we found that after rainfall, the wetted top 7–9 cm soil layer in the bare sandy land dried out rapidly when exposed to direct solar radiation in the summer months (pers. obs.). Fourthly, evidence from other studies has suggested that A. halodendron plants have a relatively weak capacity to compete for resources, such as water, from the surface soil layer compared with other herbaceous plants (Wang and Zhou, 1999; Zhou, 1999). In a pot experiment using a range of water supply treatments, Zhou et al. (2004) assessed the drought-resistance of two shrub species (A. halodendron and A. frigida) that are distributed widely in the Horqin sandy steppes by measuring changes in their gas exchange, shoot water potential and leaf chemical characteristics. This study indicated that A. frigida seedlings had superior capacity to compete for water compared with A. halodendron seedlings under severe water stress conditions.

In conclusion, this study improves our understanding of the population dynamics of A. halodendron inhabiting moving sandy lands and is helpful for developing appropriate management practices for conservation of A. halodendron-dominant grasslands.