The Mu Us Sandy Land covers an area of about 40× 10³ km²and its majorportion is situated within the Inner Mongolia Autonomous Region, northern China. The annualmean temperature ranges from 6.0° C to 8.5° C and the mean annualprecipitation from 250 to 440 mm from the northwest to the southeast. Precipitation is mainly concentrated in summer months from July to September (Cheng et al., 2011; Yan et al., 2013). In recent years, with rapid development of petrochemical industries in the area, the demand for water greatly increased and the groundwater is perceived to be a reliable water source for the industries. This surely poses an enormous threat to the ecosystems that heavily rely on the groundwater and the threat thus warns us pay particular attentions to the ecological need of the groundwater. However, we have little knowledge about the relationship between vegetation and groundwater(Cheng et al., 2011).

The study site is located in the center of the Mu Us Sandy Land (Fig.1). The site is featured by stable sand dunes and dune slacks. Four vegetation typescan be readily differentiated along a transect from the leeward side (i.e., site 1 in Fig. 2), through the upper windward slope (site 2) and the lower windward slope (site 3), to the far front of the dunes (site 4). That is, Poplar alba distributes at the leeward side of sandunes; Artemisia ordosicaat the upper windward slope and Salix psammophilaat the lower windward slope; andCarexenervisat thefar front of sandunes or dune slacks (hereinafter the four sites mentioned above will be referred as PA, SP, AO, CE sites, respectively). Generally, the heights of a single S.psammophila, A.Ordosica, P. albaand C. enervisare around 2.2, 1.5, 8.5 and 0.2m, respectively; and the crown diametersare around 1.8, 1.5, 5.0 and 0.1 m, respectively. The percentages of plant coverageare around 75%, 40%, 50% and 45% of the land surface atthe CE, AO, PA, and SP sites, respectively. Groundwater exists in the eolian sand aquifer with a greater than 50-m saturation thickness and the aquifer has good transmissivity and storativity. There are only two local shepherd householdsresiding within the 2-km radius of the study site. Thus, groundwater used by other processes, such as groundwater extraction and irrigation, can be neglected.

Fig. 1

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

Fig. 2

	Figure Option ViewDownloadNew Window
	Fig. 2 Schematicmap showing four vegetation types at the study sites. The numbers of 1, 2, 3 and 4 denote the four monitoring sites covered by Carexenervis, Salix psammophila, Artemisia ordosica andPoplar alba, respectively.

Monitoring wells were constructed in each one of the four vegetation zones in May 2013 and a 4.2-cm diameter PVC screen was installed in each well. The water table and the barometric pressure were automatically recorded at 5-min intervals during the plant growing seasons from 25 May to 11 November 2013. A tipping bucket typological rainfall gauge with an accuracy of 0.1 mm was installed to record the on-site precipitation.

Estimation of ET_g rates by the White method was based on the daily groundwater table fluctuations observed at the monitoring wells. The rates were calculated according to Equation 1 (White, 1932).

Where S_y (dimensionless)is the specific yield of the aquifer sediments; r isthe rate of the groundwater table change between 00:00 and 04:00 (24-h clock); sis the net change of the groundwater table during a 24-h period. The value of r(dimensionless) was quantified daily at each site as the slope of the water-tablegraph between 00:00 and 04:00(Fig. 3). During this period (i.e., between 00:00 and 04:00), the ET_g is rather limited due to a rather weak photosynthesis and thereforer represents the groundwater influx rate. This method assumes that the daily groundwater influx rate is constant and that the water-table change in response to ET_gcan be quantified by taking the difference between the actual groundwater table depth at the end of the day and the groundwatertable depth expected to be if the groundwater table at the site was controlled only by the groundwater flow.

Fig. 3

	Figure Option ViewDownloadNew Window
	Fig. 3 Dailyhydrograph for estimating groundwater evapotranspiration (ETg)

Based on the soil-water retention model (Brooks and Corey, 1964), Duke (1972) proposed an expression, i.e., Equation 2, to capture the S_ythat varies with the depth of groundwater table.

Where λ (dimensionless)and h_a are the pore size distribution index and soil air-entry pressure head of the Brooks and Corey water retention model, respectively; dis the depth of groundwater table, and θ _s and θ _rare the saturated water content and the residual volumetric water content, respectively. Cheng et al. (2014) proposed an analytical expression to describe the relationship between the S_y and the depth of groundwater tables using the van Genuchten soil-water retention model (Eq. 3).

Where α is a parameter related to the inverse of the air-entry value; m and n are dimensionless parameters; The n value is generally restricted to values greater than one so that the slope of the soil water retention curve is zero as the water content approaches saturation; m and nare related through the equation m=1-1/n.

Nachabe (2002) proposed a closed-form expression for determining S_y given the depth of groundwater table and the duration of drainage after a step change in the depth of the groundwater table. This method is based on the Brooks and Corey water retention model. This expression can provide insight into the dependency of S_yon both the duration of drainageand the depth of groundwater table. The expression is as Equation 4 when an initial groundwater table d₁ drops to d₂,

Where K_sis saturated hydraulic conductivity; t is time; φ is matric potential; the dimensionless exponent i can be set equal to (2+3λ )/λ , the water table fluctuation Δ =d₂-d₁, Θ _b, Θ _sur and Θ _suriare the dimensionless normalized water content, surface soil water content and initial surface soil water content, respectively.

and Θ _sur is determined from the following equation:

The S_yvalues calculated by the above three equations (Eqs. 2-4) were compared and discussed in the following section.

During growing seasons, the hydrographs at the four wells all showed typical patterns of diurnal fluctuations, being rather consistent with observations by other researches. The groundwater table tends to decline after sunrise due to strong photosynthesis, and then to riseafter sunset because ofthe lateral flow of groundwater recharge andthe weakened photosynthesis(White, 1932; Loheide et al., 2005; Butler et al., 2007). Generally speaking, groundwater table declined to a trough at around 18:00 to 20:00 and rose to a crest at around 07:00 to 09:00. In addition, the diurnal fluctuations gradually disappeared when vegetation began to enter into defoliation and dormancy. In June and July, the daily amplitude of groundwater table fluctuation had a maximum value of around 60, 30, 15 and 40 mm at the CE, SP, AO, and PA sites, respectively; and after July it reduced gradually until the regular waving fluctuations disappeared (Fig. 4). It should be particularly noted that the unexpected smoothhydrographs in June at the four wells should be taken as “ invalid” because the observed excessive decline of the groundwater table exposedthe pressure transducersabove the groundwater table, resulting in “ incorrect” measurements.

The White method is not applicable during rainfall events because the assumption of the constant groundwater influx rate throughout the day is not satisfied. Therefore, it is necessary to identify the groundwater table fluctuations resulted from the rainfall events. A total rainfall amounts of 379.7 mm was recorded during the monitoring period and the most of the amountsoccurred between June and September (Fig. 4). The bottom panel of Figure 4 (i.e., Fig. 4e) shows the rainfall events during the monitoring period.

The S_y values were estimated using Equations 2, 3 and 4, respectively. The groundwater tables in Equations 2 and 3 were obtained from the daily average value of diurnal variation. The initial groundwater table (d₁) in Equation 3 was obtained from the groundwater table at 00:00 each day and the Δ in Equation 4 was the declined values of groundwater table throughout the day. The relevant parameter values in the three equations were listed in Table 1. The calculated S_y values at the four sites were shown in Figure 5, respectively.

Fig. 4

	Figure Option ViewDownloadNew Window
	Fig. 4 Variations of the depths of groundwater tablesat Carexenervis (CE site, a), Poplar alba (PA site, b), Artemisia ordosica (AO site, c)and Salix psammophila (SP site, d)

The S_y values obtained from Equation 2 were greater than those from Equation3 when the groundwater table was less than ~2 m while the S_y values from Equation 2were approximately equal to those from Equation3when the groundwater table was greater than ~2 m (Fig. 5). The two water retention curves, one from the Brooks and Corey model and the other from the van Genuchten model, had differentshapes at smaller depth of the groundwater table, and this difference was resulted from the inconformity of the S_y from Equations 2 and 3. The maximumdifference ofS_yvalues betweenthese two curves occurred at the groundwater table of0.5 m. Equation 4was also derived based on the Brooks and Corey model, and the S_y values from this equation (Eq. 4) were also greater than those from Equation 3 when the groundwater table was less than ~0.9 m. In addition, when both the duration of drainage and the depth of groundwater table were considered in Equation4, the S_y values from this equation were less than those from Equation2 in which S_y was only dependent on the depth of groundwater table.

Table 1

Table 1

Table 1 Adopted parameters of the Brooks-Corey model for the eolian sand in the study site θ s (cm3/cm3) θ r (cm3/cm3) ha (cm) λ Ks(m/d) 0.365 0.078 21 2.6 2.47

Table 1 Adopted parameters of the Brooks-Corey model for the eolian sand in the study site

The water retention curve for the eolian sand has an S-shape, i.e., it fits the van Genuchten model, and theS_y is the function of both the depth of groundwater table and the duration of drainage. Theoretically speaking, anS_y expression, which was derived from the van Genuchten model and took into account both the depth of groundwater table and the duration of drainage, is more accurate for estimating the ET_g in the eolian sand aquifer using the White method. From this point of view, the three S_y expressions listed in Equations 2-4 did not adequately meet the expectation for accurately estimating the ET_g rates. However, Figure 5 illustrated the transient S_y expression (Eq. 4), which could estimate the value than the ultimate S_y (Eqs. 2 and 3) more accurately when the groundwater table was greater than ~0.9 m, whereas the ultimate S_y expressions may overestimate the S_y values. For example, at the groundwater table of 2.5 m, the S_y value from Equation4 was around 0.26, while the S_y value from Equations2 and 3 was ~0.29. Equations 2 and 3 may overestimate the S_y value by as much as 11.5% at this groundwater table. Therefore, when the groundwater table was greater than ~0.9 m, the effect of the drainage time on the S_y value was greater than that of the model-characterized water retention curve (e.g., the van Genuchten model or the Brooks and Corey model). However, when the groundwater table was less than 0.9 m, the opposite is the case.

Fig. 5

	Figure Option ViewDownloadNew Window
	Fig. 5 CalculatedSpecific yield (Sy) values using Equations 2, 3 and 4, respectively

In this study, Equation4 was applied for estimating the S_y value when the groundwater table was greater than 0.9 m and Equation3 was applied when the groundwater table was less than 0.9 m.

The ET_g rates were estimated using the White method at the four sites that were covered with four different types of vegetation.The dailyET_g rates were shown in Figure 6 and the average daily ET_g rates of each month were shown in Figure 7. The ET_g rates were largely controlled by the meteorological variables such as net solar radiation and temperature (Butler et al., 2007). Consequently, the ET_g rates in July and August were higher than thosein other months. However, the maximum monthly ET_g of the four sites did not occur simultaneously. Specifically, the maximum appeared in July at CE and AO sites and the maximum occurred in August at PA and SP sites, suggesting that C.enervis and A.ordosica reached their maximal growth earlier than P.alba andS.psammophila. In addition, because P.alba and S. psammophilaare deciduous species, the ET_g at PA (P.alba) and SP (S. psammophila) sites decreased sharply within a short period when leaves fellin late September. In contrast, the ET_g for the non-deciduous C. enervis andA. ordosica gradually became weak at the beginning of October when they entered into the winter dormancy.

Fig. 6

	Figure Option ViewDownloadNew Window
	Fig. 6 Estimated daily ETg rates at CE site (a), PA site (b), AO site (c) and SP site (d), respectively

The total values of ET_g at SP, AO, PA, and CE sites during May-September (2013) were 361.87, 372.53, 597.86 and 700.76 mm, respectively. The ET_gvalues at SP and AO sites were approximately equal to the rainfall amount of 379.7 mm for the observation period while the ET_gvalues at PA and CE sites were far greater than the rainfall amount. However, the groundwater table as a whole rose rather than declined from the late May to the late September, indicating that the study sites belong to a groundwater discharge zone and that the recharge rate both from infiltrating rainfall and from the net lateral inflow of groundwater were greater than the ET_g rates.

The estimated ET_g rate wasalso species-dependent.For example, C. enervishadthe highest ET_g rate among the four vegetation types and the descending order of the ET_grate for the four vegetation was: C. enervis, P. alba, A. ordosica, and S. psammophila. In addition, the depth of groundwater table has an obvious effect on the ET_g rate and the effect varied with the vegetation types. The results also obtained in the Mu Us Sandy Land by Cheng et al. (2013) showed that the average monthly ET_g ratesat SP site (S. psammophila) in July, August and September with a depth of groundwater table of ~2.35 mwere 0.70, 0.45 and 0.32 mm/day, respectively while the average monthly ET_g rates at AO site (A. ordosica) with a depth of groundwater table of 5.35 m did not exhibit any monthly difference. However, the same study (Cheng et al., 2013) showed that the average monthly ET_g ratesat SP site in July, August and Septemberwith a depth of groundwater table of 1.0-1.5 mwere up to 3.36, 3.55 and 1.06 mm/day, respectively, while the the average monthly ET_g rates at AO site in those three months with a depth of groundwater table of ~3.0 mwere 3.99, 3.03 and 2.56 mm/d, respectively. These results demonstrate that the ET_g varies with the depth of groundwater table. This finding is consistent with those by others. For example, Cooper et al. (2006) found that a 1.6-m drop in the depth of groundwater table has resulted in a 62% reduction in ET_g in the SanLuisValley of southern Colorado, USA. Although the reported 62% reduction in the SanLuisValley may be an exaggerated expression due to large-scale groundwater pumping, it is still valid in confirming that the depth of groundwater tablehas significanteffects on the ET_g.

Fig. 7

	Figure Option ViewDownloadNew Window
	Fig. 7 Average daily ETg rates in each month at CE site, PA site, AO site and SP site, respectively

A number of researches have been conducted to obtain the evapotranspiration rates or the transpiration rates with respect to the aforementioned four vegetation types in the Mu Us Sandy Land using various methods and the used methods include steady porometer, Quick-weighing technique, photosynthetic measurements, and water balance method. It should be noted that most of the measurements were performed under the conditionswhere the groundwater table wasdeeper than the reach of vegetation roots. Therefore, the water supply for evapotranspiration (ET_s) or soil transpiration was originated solely from unsaturated layer above the groundwater table. The reported results (Table 2) showed thatthe measured ET_s or transpiration rates for each of fourvegetation typeswere not consistent among different researches. However, under all of the four vegetation types, the transpiration rates measured using steady porometer and photosynthetic system were expectedly less than the ET_g rates measured using the White method and the larger ET_g was most likely resulted from the fact that the estimated ET_g was composed of both evaporation and transpiration. It appears that those results obtained using different methods are not comparablebecause different assumptions were made and different measurement conditions were encountered (i.e., plant covers and meteorological conditions). However, these results are still somewhat validin distinguishing evapotranspiration-dominated conditions from transpiration-dominatedconditions within the unsaturated layers.This point can be further illustrated by this study. For example, the ET_g at AO site (A.ordosica) can be regarded to have soly resulted from groundwater transpiration because the depth of groundwater table (~3.0 m) was beyond the reach of evaporation. At AO site the estimated monthly ET_srate was 0.08 mm/d for September and the monthly rate was1.29 mm/d for the rest of monitoring period (June, July, and August), being much less than the ET_gestimated in this study (Table 2). Similar patterns were also observed at other sites (CE site, PA site, and SP site).

Table 2

Table 2

Table 2 Comparison of the measured or estimated ET rates or transpiration rates of S. psammophila, A. ordosica, P. alba and C.enervisusing different methods Species ETg, ETsor transpiration rates(mm/d) Measurement methods Data sources Jun Jul Aug Sep C.enervis - 0.20 0.22 - Steady porometer, transpiration Xu et al. (2008) - 0.17 - - Steady porometer, transpiration Yang et al. (2005) 3.08 - - - Quick weighing, transpiration Zhang et al. (2004) 4.70 6.62 5.77 4.27 White method, ETg This study P. alba - - 2.18 - Steady porometer, transpiration Fang et al. (2006) 1.95 5.65 6.59 4.10 White method, ETg This study A. ordosica - - - 0.08 Portable photosynthesis system, transpiration Zhang et al. (2007) 0.89 1.50 1.49 - Water balance, ETs Zhang et al. (2005) 1.78 3.99 3.03 2.56 White method, ETg This study S. psammophila - - - 0.01 Portable photosynthesis system, transpiration Zhang et al. (2007) 0.85 1.67 1.63 - Water balance, ETs Zhang et al. (2005) 2.24 3.36 3.55 1.06 White method, ETg This study

Table 2 Comparison of the measured or estimated ET rates or transpiration rates of S. psammophila, A. ordosica, P. alba and C.enervisusing different methods