The study was carried out at Ordos Sandy Land Ecological Station (39° 29′ 37″N, 110° 11′ 29″E; Fig. 1) of the Institute of Botany of the Chinese Academy of Sciences. The study site is situated in the northeastern part of the Mu Us Sandy Land. The annual mean temperature is approximately 6.0° C-8.5° C, with mean monthly temperatures ranging from -9.5° C to -12.0° C in January and from 22.0° C to 24.0° C in July. The mean annual precipitation is approximately 350.0 mm with 60%-70% occurring in the period from July to September.

Fig. 1

	Figure Option ViewDownloadNew Window
	Fig. 1 Location of the Mu Us Sandy Land

The typical plant is Artemisia ordosica Krasch., a deciduous subshrub, and it commonly occurs on sand dunes (fixed and shifting) in this area (Zhang, 1994). An autogenic successional pathway in vegetation patterns is associated with the development from shrub-dominated community to herb-dominated community (Wang et al., 1997; Guo, 2000). This succession began with bare-sand dunes, succeeded with semi-fixed sand dunes dominated by A. Ordosica and finished with fixed sand dunes also dominated by A. ordosica. It then shifted to an A. ordosica+moss community on fixed-sand dunes and reached the climax stage characterized by A. ordosica+Stipa bungeana Trin. community (Zhang, 1994). Accordingly, three typical communities in the Mu Us Sandy Land were selected for the field measurements (Fig. 2): shifting dune (no vegetation); shrub-dominated A. ordosica community on semi-fixed-sand dune; and herb-dominated A. ordosica+S. bungeanaon the fixed-sand dune.

Fig. 2

	Figure Option ViewDownloadNew Window
	Fig. 2 Landscape of shifting dune (a), shrub-dominated community (b) and herb-dominated community (c)

Community characteristics and plant aboveground biomass of the three sites were investigated in August 2012. Three 5 m× 5 m plots were used to study the shrub community characteristics, and three 1 m× 1 m quadrats were arranged in the center of each shrub plot for plant surveys. Belowground biomass was investigated in 2013 at three trenches (1 m× 1 m) of 80 cm deep at 10-cm intervals. Biomasses of grass roots, < 2 mm (fine root) shrub roots, and > 2 mm (coarse root) shrub roots were separately determined and living roots were separated from dead roots via visual inspection. Soil samples were collected from the three trenches at 10-cm intervals. The collected soil samples were first air-dried and then sieved through a 2-mm screen for physical and chemical analysis. Particle size composition was determined via the hydrometer method (Liu, 1996) and soil organic matter was determined via the K₂Cr₂O₇-H₂SO₄ oxidation method of Walkley and Black (1934).

At each site, ECH₂O EC-5 (Decagon Devices, Pullman, Washington, USA) probes were installed at 5, 15, 30, 50, and 70 cm below the ground to automatically measure the volumetric soil water content (%) and these probes corresponded to the 0-10, 10-20, 20-40, 40-60, and 60-80 cm layers. Soil water content data were calibrated according to the empirical calibration equation to attain approximately ± (1%-2%) accuracy (Cobos and Chambers, 2010). The rainfall amount in each rainfall event was measured at 10-min intervals with a tipping bucket rain gauge (Model TE 525) (note: rainfall data from 20 July to 26 July were missed due to sensor malfunctioning). Soil water data were used to estimate the average cumulative infiltration. The cumulative infiltration was calculated using the following equation:

$W=\sum{_{i=1}}({{\theta }_{ei}}-{{\theta }_{i}})\times k.$ (1)

Where W is cumulative infiltration (mm); $\theta _{ei}$ is volumetric soil water content (%) of layer i (2 hours after each rainfall event); $\theta _{i}$ is initial soil water content in layer i (%); k is thickness of layer i (mm). Infiltration coefficient ($\alpha$) was calculated using the following equation:

$\alpha=W/R$. (2)

Where R is rainfall amount (mm). Soil water storage was calculated using the following equation:

$D=\sum{_{i=1}}({{\theta }_{ri}}\times {{k}_{i}}).$ (3)

Where D is soil water storage (mm); θ _ri is volumetric soil water content in layer i(%); k_i is thickness of layer i (mm). Instant infiltration rate was calculated using the following equation:

$IR=\text{ }\!\!\Delta\!\!\text{ }{{D}_{t}}/t.$ (4)

Where IR is instant infiltration rate (mm/min) and $\Delta D_t$ is increment of soil water storage (mm) within the time of $t$ (min) during a rainfall event. Based on the increments of volumetric soil water content in soil profile (0-80 cm) for the three sites, isograms were made using Kriging interpolation. Instant infiltration rate was expressed by slope of soil water storage versus time when the soil water storage was increased with time.

Differences in measured parameters between different samples were analyzed using one-way analysis of variance (ANOVA) and Fisher’ s protected least significant difference (LSD). ANCOVA was used to test whether the relationship between cumulative infiltration (dependent variable) and rainfall (covariate variable) differed significantly between sites. All statistical analyses were conducted using version 19.0 of the SPSS software (SPSS Inc. Chicago, IL, USA) at the P=0.05 level of confidence. Graphic plotting, slope, and derivative of curves were generated with Origin 9.2 (Origin Lab, Northampton, MA, USA) and Surfer 10.0 (Golden Software, Golden, Colorado, USA) software.

Again, the measurements were conducted at three sites: shifting dune (no vegetation), shrub-dominated community, and herb-dominated community (see Fig. 2). Plants species and biomass were significantly different between the shrub-dominated community and herb-dominated community (Table 1) and the plant cover of the shrub-dominated community (~40%) is lower than that of the herb-dominated community (~60%). The herb layer of the shrub-dominated community was mainly composed of such annual plants as Euphorbia humifusa, Corispermum candelabrum, Setaria viridis, and Hedysarum leave. The herb-dominated community contained such annual plants asSetaria viridis, Corispermum candelabrum, Bassia dasyphylla, as well as such perennial plants as S. bungeana, Poa sphondylodes, Astragalus melilotoides, and Stellaria filicaulis. The herb species diversity was higher in the herb-dominated community than that in the shrub-dominated community. Obviously, the herb biomass in the herb-dominated community was also significantly higher than that in the shrub-dominated community (P=0.009). However, the leaf biomass of A. ordosica was larger (P=0.000) in the shrub-dominated community than that in the herb-dominated community. Soil texture and organic matter content were also significantly different among the stages of succession (i.e., among the three sites). There were no obvious differences between the shifting dune and shrub-dominated community, but clay and silt contents and organic matter content in the herb-dominated community were significantly higher than those in the shrub-dominated community.

Table 1

Table 1

Table 1 Characteristics of communities under different cover types Site Shifting dune Shrub-dominated community Herb-dominated community Artemisia ordosicacoverage (%) 32.25± 1.61a 3.62± 0.18b Community coverage (%) 42.79± 2.14b 64.44± 3.22a A. ordosicaleaf biomass (g/m2) 17.58± 0.88a 3.21± 0.16b Grass biomass (g/m2) 14.23± 0.71b 61.26± 3.06a Soil organic matter content (%) 1.80± 0.09b 1.81± 0.09b 6.83± 0.34a Sand (%) 95.62± 4.78 95.62± 4.78 85.62± 4.28 Silt (%) 2.00± 0.10 2.00± 0.10 8.50± 0.43 Clay (%) 2.38± 0.12 2.38± 0.12 3.88± 0.19 Dominant species A. ordosica; Hedysarum leave; Euphorbia humifusa; Corispermum candelabrum; Setaria viridis A. ordosica; H. leave; Stipa bungeana; Incarvillea sinensis; Heteropappus altaicus Note: Values followed by different lowercase letters in the same row indicate significant differences (P< 0.05). Soil organic matter content of each site is the average value for 0-20 cm soil layer. Soil texture (i.e., sand, silt, and clay) in the shifting dune is the same with that in the shrub-dominated community. Mean± SE.

Table 1 Characteristics of communities under different cover types

The vertical distribution patterns of fine root biomass in the shrub-dominated community were also significantly different from those in the herb-dominated community (Fig. 3). Herb roots mainly distributed in the range of 0-40 cm depth in the shrub-dominated community and the range spanned from 0 to 80 cm depth in the herb-dominated community. In shallower layers (0-20 cm), herb root biomass of the herb-dominated community was 7 times higher than that of the shrub-dominated community. In addition, ~75% of shrub fine roots distributed in the 0-40 cm depth in the shrub-dominated community, but only ~40% of shrub fine roots distributed in the 0-40 cm depth in the herb-dominated community.

Fig. 3

	Figure Option ViewDownloadNew Window
	Fig. 3 Vertical distribution of fine root biomass in the shrub-dominated community (a) and the herb-dominated community (b). Bars indicate standard errors.

In general, the diurnal soil water content closely reflected the rainfall patterns (Li et al., 2013). To avoid measurement errors resulted from ice and snow, we collected daily soil water content data only in the growing season (from 1 April to 31 October) in 2012 to evaluate soil water dynamics. Total 68 rainfall events were recorded during the observation period. Approximately, 52% of rainfall events had the amount of less than 5.0 mm, 34% between 5.0 and 20.0 mm, and 14% more than 20.0 mm. The most intensive rainfall event (146.0 mm) occurred on 21 July (2012).

As shown in Figure 4, the diurnal soil water content at the 0-60 cm depth was higher in the herb-dominated community than in the shrub-dominated community during the observation period; while the diurnal soil water content at 60-80 cm depth was higher in the shrub-dominated community than in the herb-dominated community during the same period. The main fluctuations in the diurnal soil water content were tightly related with the intense rainfall events (> 10.0 mm). The fluctuations were expectedly the largest during and after those extremely intense rainfall events, such as ones on following days: 23 June (21.4 mm), 27 June (26.4 mm), 21 July (146.0 mm), 26 August (32.6 mm), 1 September (34.8 mm), and 24 September (29.2 mm). In addition, the soil water content was more sensitive to rainfall in the upper soil layer (0-20 cm) than in the deeper soil layer (20-80 cm) under “ normal” rainfall events (i.e., less intense or/and less prolonged ones). However, the soil water contents both in the shallower soil layer (0-20 cm) and in the deeper soil layer (20-80 cm) were sensitive to rainfall under “ abnormal” rainfall events (i.e., extremely intense or/and prolonged ones).

Fig. 4

	Figure Option ViewDownloadNew Window
	Fig. 4 Daily time series of rainfall (a) and volumetric soil water content in the shifting dune site, shrub-dominated community, and herb-dominated community at soil depths of 0-10 (b), 10-20 (c), 20-40 (d), 40-60 (e), and 60-80 (f) cm. Rainfall data from 20 July to 26 July were missed for the sensor damage.

Figure 5 shows that there are positive linear correlations between the rainfall amounts during the 17 individual rainfall events (each with 5.0-35.0 mm rainfall) and the corresponding cumulative infiltrations. As shown in Figure 5, the dependency of the cumulative infiltrations on the rainfall amounts was significantly stronger in the shifting dune (no vegetation) than in other two sites (shrub-dominated community and herb-dominated community) and the infiltration initiated more readily under “ less-intense” rainfall conditions in the shifting dune. However, the differences between the shrub-dominated community and the herb-dominated community were not significant at all. It should be further noted that during the 17 individual rainfall events (each with 5.0-35.0 mm rainfall), the total rainfall amount was 253.2 mm. The corresponding cumulative infiltration amount was 220.1 mm in the shifting dune, 104.0 mm in the shrub-dominated community, and 115.8 mm in the herb-dominated community. Infiltration coefficient was the ratio of the cumulative infiltration to the rainfall. The ratio normally increases with the increasing rainfall amount and eventually reaches a stable value. When the rainfall amount of an individual rainfall event was less than 35.0 mm, the largest infiltration coefficient was 0.91, 0.56, and 0.63 in the shifting dune, in the shrub-dominated community, and in the herb-dominated community, respectively. Correspondingly, the stable infiltration coefficient was 0.91, 0.64, and 0.74.

Fig. 5

	Figure Option ViewDownloadNew Window
	Fig. 5 Response of cumulative infiltration to rainfall based on 17 rainfall events (> 5.0 mm) collected from April to October 2012. Cumulative infiltration was calculated by changes of volumetric soil water content before and after (2 hours) rainfall.

To unveil the relationships between the soil water content and the rainfall amounts under different vegetation types (or sites), we interpolated incremental soil water content in the soil profile (0-80 cm) based on the data of the 17 selected rainfall events (5.0-35.0 mm). Three observations can be made from Figure 6. First, soil water content in the shifting dune was rather sensitive to rainfall amount. That is, relatively small rainfall events (e.g., < 10.0 mm) can readily affect the soil water content in shallower layer (i.e., 0-20 cm) and relatively large rainfall events (e.g., > 25.0 mm) can affect the soil water content in deeper layer (e.g., up to 80 cm). Second, relatively small rainfall events (e.g., < 10.0 mm) can only affect the soil water content in very shallower layer (e.g., < 10 cm) under both the shrub-dominated community and the herb-dominated community. Third, even the relatively large rainfall events (e.g., > 30.0 mm) cannot penetrate to deeper layer (e.g., > 50 cm) under the herb-dominated community.

Fig. 6

	Figure Option ViewDownloadNew Window
	Fig. 6 Isograms based on Kriging interpolation for increments of volumetric soil water content in soil profile (0-80 cm) for shifting dune (a), shrub-dominated community (b) and herb-dominated community (c)

To evaluate the responses of soil water storage to the selected rainfall events under different plant communities (or sites), we plotted the total soil water storages in the entire soil layer (0-80 cm) versus times in Figure 7. It can be seen that infiltrations both in the shrub-dominated and the herb-dominated communities lag behind that in the shifting dune, further demonstrating the ecohydrological buffering effect of vegetation surfaces (compared with bare surface). It is also apparent that the herb-dominated community had a considerably higher soil water storage than those of other two cases (i.e., shifting sand dune and shrub-dominated community) under any circumstances. To further explore the relationships between rainfall amounts and infiltration rates, we examined five different rainfall events with varying amounts of rainfall: 146.0 mm (21 July), 68.0 mm (25 July), 15.4 mm (26 July), 30.8 mm (30 July) and 22.6 mm (1 August). As shown in Figure 8, the instant infiltration rate was higher in the shifting dune than in other two sites. The highest instant infiltration rate appeared when the rainfall amount was the highest (146.0 mm). The instant infiltration rate was 1.10 mm/min in the shrub-dominated community (lower) and 1.43 mm/min in the herb-dominated community (higher) when the rainfall amount was 146.0 mm. However, the instant infiltration rate was higher in the shrub-dominated community than in the herb-dominated community when the rainfall amount was less than 146.0 mm (e.g., 68.0, 30.8, 22.6, and 15.4 mm). It should be noted that under any circumstances, the instant infiltration was higher in the shifting dune than in other two sites.

Fig. 7

	Figure Option ViewDownloadNew Window
	Fig. 7 Ten-minute soil water storage during several rainfall events from 20 July to 1 August at different sites. Rainfall amount was 146.0 mm on 21 July, 68.0 mm on 25 July, 15.4 mm on 26 July, 30.8 mm on 30 July and 22.6 mm on 1 August. No data were available for ten-minute rainfall dynamics of 20 July to 26 July due to sensor malfunctioning.

Fig. 8

	Figure Option ViewDownloadNew Window
	Fig. 8 Instant infiltration rate of rainfall events from 20 July to 1 August

Infiltration is a physical process that transfers rainfall into soil water in the rhizosphere soil layer (Wang et al., 2007) and vegetation cover and soil texture are the two factors known to affect water infiltration in significant ways (Castellano and Valone, 2007). The rainfall intercepted by plant canopies is primarily evaporated into the atmosphere and thus the soil water recharge was effectively reduced. It was reported that the vegetation canopy interception loss can account for up to 40% of total rainfall (Liu, 1997; Carlyle-Moses, 2004) and that the average canopy interception loss of A. ordosica can account for 26.8% of the total rainfall (Wang et al., 2004). Our results show that due to no rainfall intercepted by canopies ofA. ordosica and grasses, the infiltration rate of an individual rainfall in the shifting dune (no vegetation) was significantly higher than those in other two cases (i.e., shrub-dominated community and herb-dominated community). For the same reason, the cumulative infiltration and infiltration coefficient in the shrub-dominated community and herb-dominated community were significantly lower than those in the shifting dune (Fig. 5). Infiltration rate depends on various factors such as the initial soil moisture and soil texture (Wang et al., 2003; Angelaki et al., 2013). Our results show that infiltration rate in the shrub-dominated community was marginally higher than that in the herb-dominated community when the rainfall amount was moderately high (e.g., 30.8, 22.6, and 15.4 mm; Fig. 7). This may be attributable to higher silt and clay contents in the herb-dominated community and to lower silt and clay contents in the shrub-dominated community. Based on a linear regression model, soil infiltration rate (IN) could be predicted by silt content (SL) and clay content (CL): IN=37.87-0.315SL-0.527CL (Hajiaghaei et al., 2014). It should be noted that instant infiltration rate in the three sites would be underestimated when the rainfall amount was extremely high (e.g., 68.0 and 146.0 mm).

Small rainfall events can only cause minor fluctuations in surface soil moisture and major portion of the rainfall can be directly lost as soil evaporation (Heisler-White et al., 2008). It was reported that only rainfall events that exceed 2.0 mm can cause physiological responses in plants (Sala and Lauenroth, 1982; He et al., 2012) and that a single large rainfall event or two consecutively-occurred moderate rainfall events could bring about major fluctuations of soil water in deeper layer (i.e., 60-80 cm) in arid and semi-arid regions (He et al., 2012). Our study shows that the replenishment of soil water in deeper layers was mainly resulted from the accumulation of several rainfall events or one large event. Our study also shows that the rainfall events larger than 30.0 mm still failed to replenish the soil water in deeper soil layer (60-80 cm) under the herb-dominated community. The failure to replenish deeper-layer soil water under the herb-dominated community may be resulted from the changes in hydraulic conductivity. The changes in hydraulic conductivity was associated with the changes in soil texture and the latter (soil-texture changes) was brought about by the vegetation succession from the shifting dune (no vegetation), through the shrub-dominated community, to the herb-dominated community (Wilcox et al., 2003; Li et al., 2008).

Previous studies concluded that hydraulic conductivity in soil was controlled by the porosity and continuity of macro pores (Lebron et al., 2007; Zhang et al., 2013). With the A. ordosicacommunity (i.e., the shrub-dominated community) established on the shifting dune, there were no significant changes in soil texture (i.e., no increases in silt and clay contents). But, at the higher stage of succession in the A. ordosica community (i.e., the herb-dominated community), silt and clay contents increased in the shallower layer (i.e., 0-20 cm), effectively impeding penetration of infiltrated water into deeper soil layers. Thus, rainfall penetration depth in the herb-dominated community was much shallower than those of the shifting dune and shrub-dominated community. Due to fact that most rainfall events in arid and semi-arid regions were less than 5.0 mm (Sala et al., 1982; Li et al., 2013), the soil water content in the deeper layer (i.e., 60-80 cm) is thus destined to being lower under the herb-dominated community.

Soil organic matter can be another factor contributing to the failure for relatively large rainfall events to replenish deeper-layer soil water under the herb-dominated community. Soil water holding capacity has been found to exhibit a strong positive correlation with organic matter content (Hudson, 1994; Konwar, 2016). Compared with the shrub-dominated community, soil in the herb-dominated community was able to store more water in shallower layers simply due to a higher water-holding capacity. The higher water-holding capacity in the herb-dominated community was simply resulted from a higher content of organic matter. That is, the organic matter-enriched shallower layer in the herb-dominated community allowed the water stay in the shallower layer and prevented the water from penetrating to the deeper layer.