Contrasting responses of two typical plant communities to precipitation variability in horqin sandy land, northeast China

The ongoing modification of precipitation regime highlights the necessity to further investigate underlying vegetation processes. To figure out the key precipitation characteristics that largely affected the biomass and species richness of different plant functional groups in the semiarid sandy land, we compared the 15-year (from 2005 to 2019) variation in the aboveground biomass (AGB), precipitation use efficiency (PUE) and species richness of two typical biotopes (fixed dune and sandy grassland) in Horqin Sandy Land, a semiarid sandy land in northeast China. Results showed that the sandy grassland had higher AGB and PUE, lower species richness than the fixed dune; the annuals contributed 60.16% of the AGB and 60.32% of the species richness in sandy grassland, while contributed 53.40% and 75.77% in fixed dune, respectively; the AGB and species richness in the fixed dune had medium positive correlation with annual precipitation, while that in the sandy grassland had medium positive correlation with the total amount of specific rainfall events in the growing season (e.g. cumulative precipitation of 10–40 mm rain events and number of ≥ 10 mm rain event). There was no significant correlations were detected between PUE and annual precipitation, and weakly positive correlations were detected between PUE and amount and number of medium rainfall events in growing season. Our results suggested that the biomes dominated by annual plants were more responsive to the variability of growing season rainfall pattern rather than annual precipitation. In semiarid sandy grassland communities, the frequent medium rainfall events (5–50 mm) have significant role on the biomass, and the high species richness is likely to depend on the number of larger ones (≥ 10 mm). The AGB and species richness of grassland communities in semi-arid sandy land was the result of the combined effect of amount size, number and interval of rain event.


Introduction
Currently, the effects of precipitation variability on ecosystems are receiving growing attention from ecologists (Fernandez-Going et al 2012, Cleland et al 2013, Kulmatiski and Beard 2013, Zhang et al 2019, Paschalis et al 2020, Verbruggen et al 2021, Yue et al 2023), because of the incontrovertible evidence for changes in precipitation regimes IPCC Climate Change (2022).Most general circulation models project that the precipitation regime is changing and will intensify in the coming decades as a consequence of global warming, including the change in total amount, increasing variability in event size and event intervals, and increases in the frequency and severity of storms and droughts (IPCC 2012).Similar to the expectation that temperature-limited ecosystems should respond strongly to warming, the dryland ecosystems were expected to be affected profoundly by precipitation regime changes (Weltzin et al 2003, Knapp et al 2008), and some effects are likely to be irreversible IPCC Climate Change (2022).
However, communities with different species composition may respond to climate variation in different direction and rate (O'Connor et al 2001, Grime et al 2008).Soil type and plant life-form function groups modify the response to precipitation change (Grime et al 2008, Fernandez-Going et al 2012, Harrison et al 2015, Case and Carla 2018).There are a few experimental studies on effects of altered precipitation regimes on communities with contrasting soil and vegetation characteristics (Grime et al 2008, Heisler-White et al 2009, Fernandez-Going et al 2012), which suggested that vegetation and soil properties can influence a community's response to shifts of precipitation regime.For example, the aboveground biomass and precipitation use efficiency in some grassland communities composed of stress-tolerant species showed less responsive to precipitation variability (O'Connor et al 2001, Grime et al 2008, Fernandez-Going et al 2012); annual plants are more sensitive to rainfall changes than perennials because of specific functional traits (Kandlikar et al 2022); herbaceous vegetation responded immediately to the different rainfall scenarios, while the woody vegetation had a weaker and slower response (Verbruggen et al 2021).
The Horqin Sandy Land is located in the northern China, which is the largest sandy land in China and one of the main sources of wind-blown sand in the Beijing-Tianjin-Hebei region.The Horqin Sandy Land covers an area of about 66,300 km 2 , of which 35,100 km 2 is sandy land.The original natural vegetation of sandy land was a grass-dominated steppe with scattered trees (mainly elms, Ulmus spp.), which has been a desert-like landform characterized by a mosaic of mobile dune, semi-fixed dune, fixed dune, and lowland among dunes due to longterm overgrazing and fuel gathering (Zuo et al 2014).The desertified land accounts for 26.83% of the total sandy land.A classical vegetation succession occurs from the sand pioneer plant in the mobile dune to the low shrub communities in the semi-fixed or fixed dune then toward the annual herb dominated communities in the sandy grassland (Zuo et al 2009).Annual plants, the main species in sandy land, contributed a substantial proportion to biomass and species abundance, which showed more sensitive to climatic variations (Yue et al 2016(Yue et al , 2023)).The fixed dune and lowland grassland among dunes act as converge of material and species flow, and they are also the source of species diffusing.The semi-arid grassland was supposed to be more sensitive to inter-and intraseasonal rainfall variability (Knapp et al 2008, Hsu et al 2012).However, the responses of semiarid grassland to precipitation regime change were linked to plant traits soil fertility, and community functional composition (Luo et al 2021(Luo et al , 2023)).To better understanding the relative sensibility of typical plant communities to precipitation variability, we compared the 15-year (from 2005 to 2019) variations in above-ground biomass, precipitation use efficiency and species richness of two typical biotopes (Fixed dune and sandy grassland) in Horqin Sandy Land, and analyzed the relationship between their dynamics with precipitation variability.We hypothesized that the above-ground biomass and species richness of the two plant community will have contrasting correlations with precipitation variability, and the plant community in infertile fixed dunes may be less responsive to the precipitation variability than that in fertile sandy grassland.Our results will contribute to predict the effects of potential precipitation regimes, and allow land managers to better allocate limited resources, manage and utilize land properly in the future.

Study area
The study was conducted at comprehensive test plot of the Naiman Desertification Research Station (NDRS, 42°5 5'N, 120°42'E; 360 m asl), Chinese academy of Science.The comprehensive test plot is located in the southwest of Horqin Sandy Land, northeastern China (figures 1(a) and (b), and have been fenced to exclude grazing since 1992.. The study area is a farming-pasture zone with a semi-arid continental temperate monsoonal, and the local landscape is characterized by a mosaic of sandy grassland and sand dune (figures 1(b) and (c).The mean annual temperature is 6.4 °C, ranging from −13.1°C in January and to 23.7 °C in July, and the mean annual precipitation and potential evaporation are 342 mm and 1935 mm, respectively.About 60%-80% of the total precipitation occurs in growing-season (from May to August), while there is little precipitation in winter (from December to February) (Yue et al 2016).The study area belongs to the West Liaohe river basin, and the groundwater table decreased from 3-4 m to 5-6 m over 10 years (from 2005 to 2015) (Zhong et al 2018).The soil is classified as a Cambic Arenosol in the United Nations Educational, Scientific and Cultural Organization (UNESCO), and Food and Agriculture Organization (FAO) soil classification system (FAO and ISRIC 1988), and the underlying strata are loose alluvial and aeolian sediments of quaternary, mainly medium and fine sand (diameter of 0.10 to 0.50 mm, table 1).The soil in the sandy grassland biotope was finer and more fertile than that in the fixed dune biotope (tables 1 and 2).There were differences in soil and vegetation characteristics between two biotopes.To monitor the plant community succession without human activities interference, some sandy lands have been fenced to exclude grazing since 1992.

Data
Vegetation data: Vegetation survey was conducted monthly on the two landforms in a fenced sandy land (7.9 ha) from May to September since 2005.The vegetation data in late August was selected as the annual AGB due to the biomass reaches its maximum at the end of August.Ten quadrats were selected for each landscape type, and the sampling points were not the same with the ones of previous year.AGB was harvested by clipping plants at the ground level, and the AGB was classified by life-form (annual grass, perennial grass and perennial shrub).Then the clippings were dried at 65 °C for 72 h and weighed.Species richness was represented by the number of plant species per quadrat.
Precipitation data: the daily precipitation data corresponding to vegetation data was obtained from the meteorological station of Naiman Desertification Research Station, Chinese Academy of Sciences, and the data Calibration was carried out according to data from meteorological observation stations of China Meteorological  Administration.In this study, we defined a precipitation event below 2 mm as an invalid precipitation event, because that the precipitation event larger than 2 mm could induce a significant increase in surface (0-10 cm soil layer).The annual precipitation is defined as the total amount from September of the previous year to August of the current year.Precipitation-use efficiency (PUE) was the ratio of AGB and annual precipitation (O'connor et al 2001), which was used to be a critical indicator for quantifying the responses of grassland ecosystems to climate variability (Bai et al 2008, Yang et al 2010).
Data analysis: the AGB, PUE and species richness between two biotopes were compared using a t-test.Pearson product-moment correlations were used to test the significant relationship between precipitation variables (annual precipitation, the growing season rainfall amount, mean event size, mean interval) and other variables (AGB, PUE and species richness).Linear regressions were used to fit the relationship between AGB, PUE, species richness, and precipitation variables for two biotopes.All the statistical analysis was performed using the SPSS software (v.22.0,Chicago, IL, USA).

Vegetation characteristics
The number of species found in sandy grassland and fixed dune from 2005 to 2019 were 34 and 25 respectively, and 18 species have been found in both plant communities, e.g.Artemisia scoparia, Salsola collina, Chenopodium acuminatum, Setaria viridis, Lespedeza davurica.The species of two biotopes were primarily consisted by Asteraceae, Gramineae, Chenopodiaceae and Leguminosae, which accounted 92.77% of total AGB in sandy grassland and 93.71% in fixed dune (table 3).However, the percentage contribution of different species to total AGB was contrasting in two communities.The percentage contribution to AGB of Asteraceae, Chenopodiaceae, Gramineae, and Leguminosae was 34.55%, 12.68%, 40.61% and 4.93% in sandy grassland, and 33.26%, 27.25%, 17.58% and 15.62% in fixed dune community, respectively.
The species that contributed the most AGB in sandy grassland were Artemisia scoparia, Setaria viridis, Phragmites australis, Chenopodium acuminatum, and Pennisetum centrasiaticum, which accounted 34.55%, 20.46%, 8.62%, 8.93% and 5.92% of the total biomass, respectively.In fixed dune, the species that contributed the most AGB were Artemisia halodendron, Melissilus ruthenicus, Bassia dasyphylla, Corispermum macrocarpum, and Salsola collina, which accounted 24.73%, 11.97%, 8.57%, 7.21% and 7.60% of the total biomass, respectively (table 3).Annual grass species are the primary component of two communities.24 species are annual or biennial species in sandy grassland, while 15 species are annual or biennial species in fixed dune (table 3).The contribution of annual or biennial species to total AGB was 76.01% in sandy grassland and 53.11% in fixed dune respectively.
The AGB in the sandy grassland biotope was higher than that in the fixed dune biotope.The mean AGB in the sandy grassland biotope was 159.44 g g m −2 , ranging from 72.91 g m −2 (2011) to 344.26 g m −2 (2018), while that in the fixed dune biotope was 90.12 g m −2 , ranging from 52.62 g m −2 (2014) to 164.95 g m −2 (2018) (figure 2(a)) .The AGB in the fixed dune biotope was lower than that in the sandy grassland biotope, significantly (figure 2(a)).The mean PUE in the sandy grassland biotope was 0.58 g m −2 mm −1 , ranging from 0.28 g m −2 mm −1 (2008) to 1.00 g m −2 mm −1 (2018), which was higher than that in the fixed dune biotope (0.31 g m −2 mm −1 ), ranging from 0.18 g m −2 mm −1 (2017) to 0.49 g m −2 mm −1 (2018) (figure 2(b)).The species richness ranged from 4.4 to 9.3 in the sandy grassland biotope, while it ranged from 6.4 to 12.0 in the fixed dune biotope (figure 2(c)).The species richness in the sandy grassland biotope was 6.25, which was lower than that in the fixed dune biotope (8.42, figure 2(c)).The AGB and PUE showed lower coefficient of variation in fixed dune than that in the sandy grassland biotope (table 4).
The sandy grassland biotope was consisted primarily by annual species (e.g.Artemisia scoparia, Setaria viridis, Chenopodium acuminatum) and a few perennial grasses (e.g.Phragmites australis, Pennisetum centrasiaticum).The mean of annual species AGB in the sandy grassland was 128.67 g m −2 , which was significantly higher than that of perennials.The inter-annual variability in AGB of annual species was higher than that of perennial species (figure 3(a)).The mean species number of annuals in the sandy grassland was 3.77, which was slightly higher than that of perennial species (2.48, figure 3(b)).The fixed dune biotope was characterized by some annual species (e.g.Artemisia scoparia, Euphorbia humifusa, Corispermum macrocarpum, Setaria viridis, Chenopodium acuminatum) and some native perennial species (e.g.Artemisia halodendron, Melissilus ruthenicus, Bassia dasyphylla, Corispermum macrocarpum) (table 3).The mean of annual species AGB in the fixed dune was 49.65 g m −2 , accounted for 53.40% of the total AGB, which was closed to that of perennial species (46.60%, figure 3(c)).The AGB percentage of perennial species in the fixed dune was higher than that of annual species before 2009 and fell to lower since 2015 (figure 3(c)).The mean species number of annuals in the fixed dune was 6.38, accounted for 75.77% of total species richness, which was significantly higher than that of perennial species (2.04, figure 3(d)).Annuals in the fixed dune showed a larger inter-annual variability of species   The species number of the annuals was both strongly correlated with the species richness in both biotopes (P < 0.05).However, the strong correlation between the species richness of the perennials and the total species richness was only detected in the sandy grassland (R = 0.78, P < 0.05).

Correlation between AGB, PUE, species richness and precipitation variability
There was positively strong correlation between PUE and AGB in the two biotopes (figures 4(a) and (d), P < 0.05), and negatively weak correlations between species richness and PUE in sandy grassland biotope (figure 4(b)).No correlation was detected between AGB and species richness in the two biotopes.
During the 15 years, the mean annual precipitation was 311.0 mm, ranging from 239.6 mm (2015) to 401.2 mm (2019) (table 5).The relatively dry years appeared in 2008-2010 and 2014-2015, and the relative wet years appeared in 2016-2019.About 50%-80% of the total precipitation fell in the growing season (219.6 ± 48.3 mm, May to August, table 5).During the growing season, the rain event number ( 2 mm) ranged from 14 to 26, the mean event size ( 2 mm) ranged from 6.5 mm (2010) to 17.7 mm (2007), and the mean dry interval length ranged from 4.7 d (2005) to 10.8 d (2007).No significant temporal trends in annual precipitation, seasonal rainfall amount, mean rain event size, and mean dry-interval length were observed.The mean number of rainfall events was 14.8, which varied from a low of 9 (2007) to a high of 21 (2005).The maximal size of rainfall event was 147.7 mm and the maximal dry interval length was 30 d, which all occurred in 2007.Most of the long dry intervals appeared in May and June (except in 2009 and 2011) and most of the heavy rainfall events occurred in June and July (except in 2006).
AGB in the sandy grassland had medium positive correlation with precipitation characteristics in growing season, such as precipitation from 2 mm to 60 mm rain event in growing season (figure 5  AGB of perennials in the sandy grassland had a medium positive correlation with growing season precipitation (figure 9(a)) and precipitation from 2 mm to 60 mm rain event in growing season (figure 9(b)), while a strong positive correlation with number of 2 mm event (figure 9(c)).The species richness of perennials in the sandy grassland had significantly positive correlation with number of 2 mm event (figure 9(f)), while that in the fixed dune had medium negative correlation with precipitation from 2 mm to 60 mm rain event in growing season (figure 9(e)).

Discussion
Our results show that the annual species play a substantial role on the variation in AGB and species richness of both biotopes in Horqin sandy land.However, the contribution of perennial species to total AGB in fixed dune was significant higher than that in sandy grassland, and the dominated annuals in fixed dunes were Corispermum macrocarpum, Salsola collina and Bassia dasyphylla, which have stronger drought tolerance than annual plants in sandy grassland (e.g.Artemisia scoparia, Setaria viridis and Chenopodium acuminatum).This may be attributed to that the coarse and infertile soil in fixed dune because that the infertile soil always favors the stress-tolerant species (Fernandez-Going et al 2012, Harrison et al 2015).In this experiment, the two types of biotope were all dominated by annual plant species.However, soil of the fixed dune biotope was relatively coarser and more infertile than sandy grassland.The lower AGB and relatively smaller inter-annual variation in AGB in fixed dune  Value is mean ± standard deviation.a No. of events represents the number of 2 mm rain event; b intervals represents the duration between two events of 2 mm. .By contrast, due to the high capacity of resource capturing and photosynthesis, annual species showed higher growth rate in wet periods, but more vulnerable than perennials during dry periods.Precipitation variability was considered the key factor in variation in the AGB and species richness in semiarid grassland ecosystems (Chesson et al 2004, Zhang et al 2019).We found that there was no significant correlation between the AGB, species richness and annual precipitation in the sandy grassland, which was inconsistent with some similar studies).Previous reports found there were positive correlations between annual precipitation and AGB (Yang et al 2008, Munson et al 2016).Different from the negative correlations between annual precipitation and PUE in many grasslands of inner Mongolian Plateau (Bai et al 2008, Sun andDu 2017), the PUE of two grassland communities in this study had no significant correlation with annual precipitation in the Inner.These differences may be attributed to the species life-history traits of two biotopes, which were dominated by annual species.Generally, the AGB of deep-rooted perennials have a positive relationship with annual precipitation (e.g., perennial species), while shallow-rooted and short-lived plants have a weak or nonlinear relationship (e.g., annual grass) (Knapp et al 2006).Annual precipitation is the dominant factors driving the floristic composition and diversity in temperate grasslands (Jiao et al 2017, Bai et al 2021).The inter-annual variation of precipitation has little effect on species richness.The weakly positive correlation between species richness and annual precipitation in both biotopes was similar with Grime et al (2008) and Harpole et al (2007), who found change of annual precipitation had no significant effect on grassland plant diversity.The correlation between aboveground biomass, precipitation use efficiency and species richness was also not detected in this study.
Rather than annual precipitation, the amount and size distribution of the growing-season rainfall play more important role in the productivity and species richness of grassland (Swemmer et al 2007, Miranda 2009, Zhang 2018).In this study, we found that there is a significantly positive correlation between AGB, species richness and the growing-season rainfall amount in the sandy grassland.We found rainfall amount contributed by relatively intermediate rain events (e.g. from 2 mm to 60 mm, from 5 mm to 50 mm, from 10 mm to 40 mm), and the number of the relative large rain event (e.g. 5 mm, 10 mm) of the growing season had medium positive correlation with the AGB and species richness in both biotopes.This may be attributed to the considerable proportion of annual plants.Plants with different life forms have different responses to the same rainfall variability.For example, the AGB of perennials was more responsive to the precipitation from 2 to 60 mm rain events and the number of 2 mm (or 5 mm) rain event, while the species richness of annuals was more responsive to the precipitation amount from 15 to 30 mm rainfall events.This result implied that the frequent medium rainfall events (5-50 mm) have a significant role on the productivity of the semiarid sandy annual grassland, but the higher species richness is likely to be attributed to the larger ones.Similarly, the weak correlations between PUE and larger rainfall events (e.g.precipitation from 10 to 40 mm rain events, precipitation from 15 to 30 mm rain events, and number of 10 mm rain event) in sandy grassland further illustrated the large rainfall had higher precipitation use efficiency than the small rain events.Because a considerable proportion of the sand grain size of 0.25-1.0mm (accounted for 20%-60%, table 2) results a poor water holding capacity, small rain events only moisten the soil surface and evaporate quickly, while the heavy events can infiltrate deep soils inaccessible to shallow-rooted annual plants.Although we were not able to identify a significant relationship between the mean event size and AGB, we found that the correlation between the precipitation of 10-40 mm rainfall events, the number of 10 mm rain events and the AGB of the sandy grassland was higher than other rain-size, which demonstrated that the rain event size had important effects on the productivity of semiarid grasslands (Heisler-White 2008, Thomey et al 2011).That the higher correlation between AGB, species richness of perennials and number of 2 mm rain events indicated that the small rainfall events are beneficial to perennial plant vitality under drought stress.However, the more extreme precipitation characterized by heavy rain events and longer dry event-intervals may reduce the precipitation use efficiency in semi-arid grassland.In this study, we found the largest event size and the maximum AGB of the sandy grassland did not appear simultaneously in the same year, which may be attributed to the extreme dry intervals resulted by the stochasticity of rainfall events.Yue et al (2023) discovered that dry intervals significantly reduced the annual grassland productivity, even if the rainfall amount remains stable.For example in 2010 and 2015, the amount of the growing season rainfall were approximate, the AGB of sandy grassland in 2010 was higher than that in 2015, which may be attributed to the relatively even distribution of rainfall events in 2010.The higher precipitation but lower AGB of sandy grassland in 2007 and 2017 than that in 2005, 2012, and 2018, may be attributed to the longer dry spells ( 27 days).Droughts resulted by reducing precipitation also impacted the productivity and species richness of semiarid grassland ecosystems (Cherwin andKnapp 2012, Craine et al 2012).The decreasing dominance of the perennial species and the increasing dominance of annual plants in the fixed dune may be attributed to the droughts from 2008 to 2012 (figure 2), because that less precipitation and longer dry seasons can reduce the coverage of perennials (Zeiter et al 2016).The impact of droughts occurring at the mid-growing season on grassland productivity will be more significant than one occurring at the end of the growing-season (Craine et al 2012).In sandy grassland, the higher AGB in 2016 than that in 2019 may be attributed that the occurrence of maximum dry interval, which in 2016 occurred in August and that in 2019 occurred in June.Furthermore, in this region, the long-term mean annual precipitation was 342 mm, closer to that in semi-humid regions, which may be the reason that the plants in this region may have weaker drought resistance capacity than that in arid regions.Knapp et al (2008) predicted that the more extreme rainfall regimes with intensified rainfall events and prolonged event intervals will increase the primary productivity in xeric ecosystems because of the higher sensitivity to rainfall than droughts.The results in this manuscript supported that the intensified rainfall events will increased productivity of semiarid sandy grassland dominated by annual plants.However, the annual species also showed more vulnerable to droughts, which may affect the species composition of annual grassland by modify the reproductive growth of plants (Monroe et al 2019), long-term changes in water availability would probably reduce productivity and diversity in these grassland communities (Miranda et al 2009, Gherardi andSala (2015), even trigger plant death, along with contribution to desertification of semiarid sandy grassland in conjunction with grazing activities (Reichstein et al 2013).

Conclusion
The fixed dune and sandy grassland have contrasting responses to precipitation variability in Horqin Sandy Land.The infertile fixed dune was dominated with drought tolerant species and more sensitive to variation in annual precipitation, while the fertile sandy grassland was characterized by annual grasses and more sensitive to the growing-seasonal rainfall variability.The increasing annual precipitation would increase the biomass, species richness and dominance of annual species in fixed dune.The increasing growing season precipitation would be likely to increase the species richness in sandy grassland.The more intensified rainfall events during the growing season would be likely to increase the biomass of sandy land, especially biomass of annuals in sandy grassland.However, the negative effect of more extreme droughts may offset, even surpass the positive effect of increasing precipitation or intensifying rainfall events.
number than the perennials (figure 3(d)).The species number of annuals in the fixed dune decreased to a minimum in 2008 and increased to a maximum in 2013 (figure 3(d)), while the number of annual species was maximal in 2007 and minimal in 2009 in the sandy grassland biotope (figure 3(b)).The total AGB had strong correlation with the AGB of annual species in the sandy grassland (R = 0.96, P < 0.05) and moderate correlation in the fixed dune (R = 0.64, P < 0.05).
(c)), precipitation of from 5 mm to 50 mm rain event in growing season (figure 5(d)), precipitation from 10 mm to 40 mm rain event in growing season (figure 5(e)), number of 10 mm event in growing season (figure 5(i)).The AGB in the fixed dune had medium positive correlation with annual precipitation (figure 5(a)), growing season precipitation (figure 5(b)), precipitation from 2 mm to 60 mm rain event (figure 5(c)) and precipitation of from 5 mm to 50 mm rain event in growing season (figure 5(d)).

Figure 3 .
Figure 3. Variations in AGB and species richness of annual species and perennial species for two biotopes from 2005 to 2019.Value is mean ± standard error.Different lowercase letters indicate significant differences between two biotopes (P < 0.05).

Figure 4 .
Figure 4. Correlations between (a, d) AGB and PUE, (b, e) PUE and species richness, (c, f) species richness and AGB for two biotopes.R represents the Pearson correlation coefficient.* represents P < 0.05.

8
Environ.Res.Commun.5 (2023) 115017 X Yue et al than the sandy grassland biotope, may be also attributed to that the stress-tolerant species is less sensitive to climatic variation (Fernandez-Going et al 2012, Harrison et al 2015).The perennials have a long-lived life form, low specific leaf area, and deep and dispersed root systems (Ning et al 2022), which help them buffer the environmental fluctuations (Grime et al 2008, Evans et al 2011)

Figure 5 .
Figure 5. Correlations between AGB and (a) annual precipitation, (b) growing season precipitation, (c) precipitation from 2 mm to 60 mm rain event in growing season, (d) precipitation from 5 mm to 50 mm rain event in growing season, (e) precipitation from 10 mm to 40 mm rain event in growing season, (f) precipitation from 15 mm to 30 mm rain event in growing season, (g) number of > 2 mm rain event in growing season, (h) number of > 5 mm rain event in growing season, (i) number of > 10 mm rain event in growing season .R represents the Pearson correlation coefficient; * represents P < 0.05.

Figure 6 .
Figure 6.Correlations between PUE and (a) annual precipitation, (b) growing season precipitation, (c) precipitation from 2 mm to 60 mm rain event in growing season, (d) precipitation from 5 mm to 50 mm rain event in growing season, (e) precipitation from 10 mm to 40 mm rain event in growing season, (f) precipitation from 15 mm to 30 mm rain event in growing season, (g) number of > 2 mm rain event in growing season, (h) number of > 5 mm rain event in growing season, (i) number of > 10 mm rain event in growing season.R represents the Pearson correlation coefficient; * represents P < 0.05.

Figure 7 .
Figure 7. Correlations between species richness and (a) annual precipitation, (b) growing season precipitation, (c) precipitation from 2 mm to 60 mm rain event in growing season, (d) precipitation from 5 mm to 50 mm rain event in growing season, (e) precipitation from 10 mm to 40 mm rain event in growing season, (f) precipitation from 15 mm to 30 mm rain event in growing season, (g) number of > 2 mm rain event in growing season, (h) number of > 5 mm rain event in growing season, (i) number of > 10 mm rain event in growing season .R represents the Pearson correlation coefficient; * represents P < 0.05.

Figure 8 .
Figure 8. Correlations between AGB of annual species and (a) precipitation from 2 mm to 60 mm rain event in growing season, (b) precipitation from 10 mm to 40 mm rain event in growing season, (c) number of > 10 mm rain event in growing season; Correlations between species richness of annual species and (d) annual precipitation, (e) growing season precipitation, (f) precipitation from 15 mm to 30 mm rain event in growing season.R represents the Pearson correlation coefficient; * represents P < 0.05.

Figure 9 .
Figure 9. Correlations between AGB of perennial species and (a) growing season precipitation, (b) precipitation from 2 mm to 60 mm rain event in growing season, (c) number of > 2 mm rain event in growing season; Correlations between species richness of perennial species and (d) growing season precipitation, (e) precipitation from 15 mm to 30 mm rain event in growing season and (f) number of > 2 mm rain event in growing season.R represents the Pearson correlation coefficient; * represents P < 0.05.

Table 1 .
Soil mechanical composition of two biotopes.

Table 2 .
Soil bulk, soil organic carbon, soil nitrogen, and soil phosphorus in 0-20 cm soil layer of two biotopes.

Table 3 .
Mean percentage contribution of species to AGB for two biotopes.

Table 4 .
The Coefficient of variation (CV) of AGB, PUE and species richness.(Standard deviation / Mean) × 100%, represents the dispersion of the value.The CV implies the interannual change in AGB, PUE and species richness in this study.

Table 5 .
Rainfall characteristics of the growing seasons (May to August) over 15 years.