Introduction
A comprehensive and accurate understanding of aquifer hydraulic properties is the basis for efficient management of groundwater resources, which can be developed by pumping test determining aquifer hydrological parameters for groundwater modeling, aquifer vulnerabil ity assessment, among others (Wen, Wu, Yeh, & Tseng, 2010; Li, Qian, & Wu, 2014; Criollo, Velasco, Vázquez-Suñé, & García-Gil, 2016). The accuracy of data derived from pumping test is of concern to parameters calculation, which fluctuating along with the variety of barometric pressure, sea tides and so on (Guo & Jiao, 2008; Guo, Jiao, & Weeks, 2008). In the process of studying, hydrogeologists devoted themselves to theoretical and practical work on pumping tests, it was found the influence of barometric pressure on water level cannot be ignored, that pressure efficiency is expressed by a simple linear equation at first (Jacob, 1940). After long time duration of studying, for the pumping test process, it was found that the groundwater level (especially the deep well water level) response is not linear, but rather a complex nonlinear response, which makes it inappropriate to use a linear regression method for the pressure correction (Quilty & Roeloffs, 1991; Seo, 2001). Among lots of researchers, Rasmussen and Crawford (1997) used a regression deconvolution method to determine pressure correction and achieved good results.
The hinterland of the Taklimakan Desert, which belongs to a sedimentary area in the southern section of the Tarim Basin, is an ancient alluvial flood plain that formed from many rivers in the northern slope of the Kunlun Mountains. Precipitation and condensed water supply no recharge to groundwater basically (Yu, 1992), which is mainly affected by the Kunlun Mountain piedmont plain groundwater lateral runoff recharge. There are rich oils and natural gas resources, but the climate is extremely dry, and the precipitation is below 50 mm with an annual evaporation of 3 000 mm, no surface water exists generally (Xu, Li, Zhou, & Zhou, 2001). Therefore, the water used for production and living comes entirely from the exploitation of groundwater, unconfined water, particularly, which is formed by river infiltration and lateral groundwater infiltration belonging to pore water stored in Quaternary deposits, whose main hydrochemical types are Na-Mg-Cl-SO4 and Na-Cl-SO4 (Wei, Fan, Xu, Jin, & Zhou, 2016), and salinity is 3-10 g/l. The depth of groundwater between longitudinal dune usually is 1-5 m, and over 20 m at top of ones. It was found that when the groundwater depth is less than 1.5 m, the groundwater is affected by strong evaporation (Fan et al., 2010), coincide with the terrain trend, the water table is relatively flat, with a flow velocity measured between 21 m/year and 205 m/year (Lei, Li, Jin, & Xu, 2008).
In this paper, a single-well pumping test established at the core area of pumping in the hinterland of the Taklimakan Desert was used to calculate the hydrogeological parameters, in order to provide a reference of the future groundwater research, especially the construction of hydrological modeling.
Methods and experiments
We observed the water level of the whole year and conducted single-well pumping test from 2009 Jul. 27th to Aug. 12th day in the core area of water supplying in the oil field with observation time interval was every 30 min. The pumping well is located in the hinterland of the Taklimakan Desert (38º 58’ 20.77”, 83º 39’ 39.60”) (figure 1), with the observation well (38° 58’ 21.78”, 83° 39’ 37.35”) logging in the northwest at a distance of r = 58.6 m (figure 2). The elevation of pumping well is 1 104.9 m. The thickness and effective thickness of unconfined aquifer are 66.00 and 40 m, respectively in the study area where the depth of groundwater is 6m, the formation is single fine sand of Quaternary. The casing pipe of the partial penetrating observation well is hard PVC material and depth was measured by Solinst Levelogger Model 3001. In the experiment, discharge rate was set as 32 m3/h and it was regarded as having achieved stable state when the fluctuation of the drawdown was not more than 1% (Wang et al., 2013). The regression deconvolution was performed in order to adjust the groundwater level using BETCO from Sandia Labs (USA). The hydrogeological parameters were calculated by Aquifer Test 2011.1 designed by Waterloo Hydrogeologic Inc. (incorporated into Schlumberger Water Services since 2005) to provide a reference for the construction of hydrological modeling.
Barometric pressure correction
In measuring groundwater level, many natural forces have important theoretical and practical significance to the borehole water level (Guo & Jiao, 2008). In particular, when the impact of barometric pressure on the water level cannot be ignored compared to the larger circumstances (such as the late period pumping test), it is important to perform barometric pressure calculations in order to obtain accurate aquifer parameters (Wang et al., 2012). During an extended period of pumping test processes, the delayed response, due to the transmission of the barometric pressure perturbation through the unsaturated zone and borehole storage or skin effects, causes a water level difference between the aquifer and the pumping well. Rasmussen and Crawford proposed the regression deconvolution method could be used to correct the groundwater level in order to eliminate the influence of barometric pressure (Rasmussen & Crawford, 1997). After the response function δ(i) to a barometric pressure pulse is obtained by using ordinary least square method linear regression, the step response α(j) can be found by summing the impulse responses:
Then, it was following correction variables of the water levels that being found (Toll & Rasmussen, 2005):
where W* is the corrected variable for each observation within time t from m to n; ΔBt is the change in the barometric pressure at time t; m is the maximum lag selected by the user, and n is the total number of observations in the dataset.
To gain more accurate data of the water level, we corrected the observed pumping test data by BETCO, which is the application of the regression deconvolution (Toll et al., 2005).
Calculation of hydrogeological parameters
In Aquifer Test, methods for unconfined aquifer analysis usually are either Neuman or Boulton. Theis with Jacob correction can also be used for late-time of the pumping test. Due to unsteady flow within the pumping tests by the artificial control of water stability, we need to determine the water level changes with time in order to obtain hydrogeological parameters in an unconfined aquifer of unstable movement. In this paper, the Neuman model, the Boulton model, and the Theis model with the Jacob correction (Theis with Jacob) for unconfined aquifer are used to fit the water level variation curve of the pumping process.
Jacob (1940) proposed the following correction:
Where scor is the corrected drawdown, s is the measured drawdown, and D is the original saturated aquifer thickness. There are no additional type curve parameters for this solution method.
The equation developed by Neuman representing drawdown in an unconfined aquifer is given by (Neuman, 1975):
Where W(μA . μB . β) is known as the unconfined well function; μA = r2s/4Tt is the Type A curve for early time steps; μB = r2Sy/4Tt is the Type B curve for later time steps; β = r2Kv /KH, KV, KH: vertical and/or horizontal permeability; r is the distance to the observation well; S is storativity; Sy is specific yield, usable probe volume; T is transmissivity.
The method developed by Boulton (1963) can be performed as follows:
Where H is defined as the average head along the saturated thickness, and other parameters defined as mentioned above.
Results and discussions
Barometric pressure correction results
From the figure 3, we can conclude that, at the period of Jan. 1st to Feb. 21st and Sep. 23rd to Dec. 31st, the water level was roughly stable, which is consistent with the main water use period of the project from March to November. Though the pumping well was performed in a production well, the test was started after the water level reaching stable judged by long time observation of recovery period.
The previous researchers have done so many about barometric pressure correction (Spane, 2002; Dong, Shimada, Kagabu, & Yang, 2015) and hydrogeological parameters’ calculation (Lubczynski & Roy, 2005; Samuel & Jha, 2014), But for the combination of the two methods, relevant research is relatively small (Allen, 2008; Vilhelmsen, Christensen, Behroozmand, & Søltoft, 2014), and most of them used barometric efficiency (Acworth & Brain, 2008; Delcourt-Honorez & Scholz, 2014) to correct the influence of barometric pressure, which didn’t consider the delay effect to vadose zone soil layer. The delay effect is considered in BETCO software, and the result of pressure correction is more accurate.
The regression deconvolution method is used to correct the deviation that barometric pressure variation causes the measured groundwater level, and the results seem to be good. And it is easy concluded that barometric pressure correction is crucial to the accuracy of groundwater level in pumping tests, which is of great concern to delay factor. In the BETCO, though we conduct that maximum response time as 12, it can be noticed that at the preliminary stage of observation, there are still exits the phenomenon of delaying. After the correction for barometric pressure, the annual water depth has been changed; the maximum of the initial water depth was 8.518, though, with the correction, it was 8.491.
It is important to note that other natural conditions, such as tidal, meteorological, and hydrological factors will also affect the water level fluctuation. If these effects are taken into consideration for correction, we can determine a more accurate water level correction.
The upper graph is the observation water depth of whole year, and the inferior graph is the depth during the pumping test.
Hydrogeological parameters
The Boulton model does not achieve fitting results. The remaining two models are more successful and achieve the results shown in figure 4.
Through curve fitting, the hydrogeological parameters of the Theis model with the Jacob correction are as follows:
And the hydrogeological parameters of the Neuman model are as follows:
The permeability coefficient of the two models calculated is 13.6 m/d and 13.3 m/d, which is good agreement with previous K = 12.85 m/d (Fan et al., 2010) and K = 13.32 m/d (Fan, Jin, Lei, Xu, & Zhou, 2013). These hydrogeological parameters are calculated using the water level process curve model fitting based on the pumping test, and the results are useful for the construction of the hydrological model.
Conclusions
In this paper, for the unsteady flow pumping test in an unconfined aquifer, three kinds of unconfined models were used to calculate hydrogeological parameters. Boulton cannot successfully gain the results, the parameters that Theis with Jacob calculated are T = 8.96 E2 m2/d, S = 6.11 E-2 and those calculated by Neuman are T = 8.76 E2 m2/d, Sy = 5.00 E-1, Kv /KH = 1.12 E-3, Sy / S = 1 .00 E1, the permeability coefficient calculated by two models is 13.6m/d and 13.3 m/d. From the calibration results in this paper, considering the unsaturated zone delay effect, the results are closer to the true value. But in the pumping test, the study area is used as a homogeneous isotropic field. The calculated permeability coefficient is one of the point values of the region, which can’t reflect the heterogeneity of the scale. However, these parameters results provide a certain reference value to further study on hydrogeological conditions in the study area and the methods used in this paper can provide solutions to the problems that may be encountered in similar projects.