Application of numerical modeling to reservoir immersion assessment and control in dual-formation hydrogeological unit

Reservoir immersion is a serious environmental geological issue in a dual-formation structural reservoir bank (DFB) induced by dynamic surface water impoundment (SWI) that has implications for low-lying farmland and buried infrastructure. It is a major challenge to identify the dynamic immersion process and make economic and scientific joint mitigation measures for controlling groundwater immersion. Here, we develop a three-dimensional groundwater flow model and apply it to evaluate and control reservoir immersion in the typical low-lying DFB of Xingan Navigation and Power Junction Project (XGNPJ) across Ganjiang River in Jiangxi Province, China. The field-scale model is well calibrated to predict where the groundwater immersion could potentially occur. Furthermore, the effectiveness of the countermeasures adopted for the reduction of reservoir immersion areas were analysed based on the simulation model by considering the projected future combination scenarios of engineering measures. Results indicate that without engineering mitigation measures, SWI generates groundwater inundation across 23% of the total study area. Comprehensive comparative analysis on different seepage control schemes reveals that the joint engineering measures can effectively control the immersion range to 5% of the total area. The findings can provide scientific basis for groundwater immersion assessment and guide immersion control of XGNPJ project.


INTRODUCTION
Over the past three decades, massive water conservancy projects in the form of navigation and hydropower junction have been constructed in China to bring tremendous benefits in flood control, water scarcity alleviation, and clean energy generation (Liu et al. ; Jia ). However, water conservancy development has caused serious groundwater immersion due to surface water impoundment (SWI) in the dual-formulation structural reservoir bank (DFB) (Azamathulla et  The analytical method with numerical simulation approach is often applied to evaluate reservoir immersion in a plain reservoir with a dual-formation structure (Mahdavi ; Hu et al. ). During early engineering planning, the evaluation of the immersion area is usually through a prediction method recommended by the Kaminski formula (КАМЕНСКИЙ ). However, many scholars have found that the immersion range obtained by the Kaminski method using a simplified calculation model is much larger when applied to a DFB where the permeability coefficient of the lower main aquifer is 2-3 orders of magnitude larger than that of the upper weak aquifer (Hu et al. ). As a consequence, the Kaminski the bank with a dual-formation reservoir structure was evaluated, and the range of immersion and degree of immersion hazard in the study area were predicted, thereby providing a scientific basis for the layout of the immersion control scheme in the early stage of the project.
Water conservancy designers and local state government recognize that ongoing SWI will necessitate the development of adaptive design standards with regard to The main function of XGNPJ is navigation and power generation with a normal storage level of 32.5 m, about 2.0 m higher than the average annual water level of Ganjiang River.
Reservoir immersion is a serious geo-environmental issue for agriculture and building structures in plains areas along Ganjiang River. Taking Shihutang Navigation and Power Junction Project as an example, an adjacent reservoir project in the upstream of XGNPJ in Ganjiang River, the immersion happened after one year of reservoir impoundment (Yin & Lu ). The groundwater level rose and exceeded the critical immersed groundwater depth, even creating water logging in the surrounding residential and agricultural cultivated land, especially in areas where the ground elevation is lower than the impounded water level of 56.5 m. In the recent three years, the process of groundwater immersion has been effectively controlled as the result of engineering management measures such as establishment of relief wells, a ditch drainage network system, lifting low-lying fields and so on, spending tremendous financial and material resources.
Impounding and operation of XGNPJ will inevitably produce immersion issues in the reservoir area where the topographic and hydrogeological conditions are extremely similar to Shihutang reservoir region. To avoid socio-economic losses caused by immersion disaster, it is essential to identify the dynamic immersion process, evaluate the immersion range, and then optimize the combination of mitigation measures for controlling immersion.

Site description
The left reservoir bank is selected as the research area for several reasons: (1) the left bank region is a closed and independent hydrogeologic unit between Ganjiang River and Yuan River, and is predicted to be threatened by a relatively high degree of reservoir immersion (18 km 2 ) by steady state analytical equation; (2) the left bank is rich in agricultural resources and intensive socioeconomic activities and it is thus urgent to conduct scientific assessment and control for reservoir immersion; (3) an enormous amount of different types of hydrological and hydrogeological data (from the Geological Survey; Geological Data Information Network of Jiangxi Province, http://www.jxgtt.gov.cn/ dzzlg/Index.shtml; National Meteorological Information Center, http://www.nmic.cn/web/index.htm) has allowed us to establish a three-dimensional groundwater model to evaluate the processes in the alluvial groundwater system.

Meteorological and hydrological conditions
The study area receives annual mean precipitation of

Hydrogeological conditions
The left bank of XGNPJ can be considered to be a relatively independent hydrogeological unit covering 83.9 km 2 . The  In this region, the Ganjiang River is deeply downcut with an average river-bed elevation of 23-27 m above sea level based on the SRTM 30 m DEM Digital Elevation Database (SRTM Data https://srtm.csi.cgiar.org/srtmdata/). In addition, high linear correlation coefficients (R SG ) were determined between groundwater level and nearby surface water level based on the synchronously observed data from groundwater and surface water monitoring (1986)(1987)(1988)) (Guo & Chen ). The average R SG was 0.81 at Xingan station (Ganjiang River) and 0.78 at Yichun station (Yuan River). Based on the above DEM data and the dynamic relationship between surface water level and groundwater level (Figure 3), we conclude that the channels of the Ganjiang River and Yuan River cut through the alluvial aquifer. The area enclosed by these water bodies is an independent hydrogeological unit surrounded with specific head boundaries. Rainfall is the main source of recharge of groundwater in the study area and it is then discharged to surface water for most of a hydrological year. However, surface water recharges groundwater between June and August when the surface water level rapidly increases with the increasing of inflow from upstream areas. Therefore, groundwater and river water exchange seasonally, groundwater flowing into rivers along the exposing aquifers in the dry and flat seasons and river water also recharging aquifers, replenishing the aquifer in the wet season.

METHODOLOGY
In order to identify the dynamic immersion process and make economic and scientific joint mitigation measures for controlling groundwater immersion, a groundwater numerical modeling based methodological framework was proposed, as shown in Figure 5. Firstly, we developed a three-dimensional groundwater flow model in the typical low-lying DFB of XGNPJ reservoir region. With regard to model reliability, the procedure of model calibration and sensitivity analysis were employed resulting in the  can be described as follows: where K xx , K yy , K zz are hydraulic conductivity along the x, y, Model setup

Model calibration
We assume homogeneous and isotropic material property in each sub-domain as shown in Figure 6. Three fitting criteria including the root mean squared error (RMSE), the mean absolute error (MAE), and Nash-Sutcliffe coefficient of efficiency (NSE) were used to calibrate the numerical model. The observed data and calculated results highlight that both the surface water level and groundwater level showed obvious annual variation and the trends in their change were largely consistent, indicating that a close hydraulic connection exists between the main watercourses and the alluvial aquifer.
Based on the relative elevations of stream stage and the groundwater head adjacent to the stream, groundwater mainly discharged to surface water bodies at the sites close to the rivers, while surface water recharged groundwater between June and July when the surface water level rapidly increases with the increasing of inflow from upstream areas.

Sensitive analysis
Sensitivity analysis (SA) was applied to vary model input parameters and evaluate how model results of groundwater flow dynamics change with these variations. In this study, the sensitivity of hydrogeological parameters including net infiltration rate, the specific yields of the L1 and L2 unconfined aquifers, the hydraulic conductivity and specific storage of L3 and L4 confined aquifers are analyzed to provide valuable understanding of both model implementation and the underlying physical processes, thus providing insight into model behavior. The relative sensitivity index (S r ) of the any parameter is estimated by using the following equation:

Critical immersed groundwater depth
Future influences of reservoir immersion need to be evaluated with respect to the positioning of the groundwater depth with critical immersed groundwater depth (GD C ), which is the evaluation criteria for reservoir immersion.
When the groundwater depth is less than GD C after the normal reservoir impoundment, it is deemed as the immersion area. The GD C can be calculated according to the following formulation (Li et al. ): where, H K is the height of capillary rise [L], and ΔH is the According to the results of field experiment test, the height of capillary rise is 0.6 m in the study area. Therefore, the GD C corresponding to the slight and high immersion is 1.1-2.1 m and <1.1 m, respectively.

Engineering measures for controlling reservoir immersion
The objective of this modeling study was to investigate how the combination of the engineering measures affects the groundwater system and the process of reservoir on the left bank of XGNPJ from 1/1/2018 to 1/1/2021. Therefore, to reasonably forecast the effects, we considered 8 schemes with different combinations of engineering countermeasures, including cut-off wall, relief wells, and land lift reclamation projects, as shown in Figure 9.
The cut-off wall constructed along the embankment are embedded below the bedrock surface, starting from the Dam Site A and ending at Node C, was divided into two sections.
One is the section of high pressure jet grouting impervious wall from Dam Site A to Node B, and the other section is the water jetting anti-seepage wall from Node B to Node C.
The thickness of high pressure jet grouting and water jetting   Table 2, including Scheme S0 without any anti-immersion engineering measures, and Schemes S1-S7, the combination schemes with different K values of cut-off wall, and drainage level of relief wells and their vertical distance to the cut-off wall.

Evaluation of the dynamic immersion process
Based on the 3D numerical model, the inside boundary conditions, such as the specified head boundary of Ganjiang River, were changed according to the process of SWI to predict the groundwater immersion process without any anti-immersion engineering measures. The surface water level after the normal operation of XGNPJ will be main-    Table 3, the dynamic immersion process calculated for Profile 2 was similar to that for Profile 1.
The simulation results show that during the early stage of SWI, the main driver of dynamic changes in groundwater in the upper unconfined aquifer was high head pressure in the lower confined aquifer, whereas the recharge from precipitation infiltration was the second-most influential factor. However, between the second year and the end of the third year, as the groundwater level increased in the unconfined aquifer, there is a decrease in the difference of groundwater level between lower and upper aquifers, resulting in a weakening hydraulic gradient from the lower confined aquifer. On the other hand, with the rise of groundwater level in the unconfined aquifer, the distance from surface infiltration recharge to the unconfined water table shortened, with some areas even becoming waterlogged.
In addition, the results of the sensitivity analysis showed that the model was most sensitive to the net recharge coefficient of precipitation infiltration; therefore, the effect of rainfall infiltration recharge will gradually become the dominant factor driving groundwater immersion during the middle and later period. According to these results, the   Under Scheme S0, serious seepage immersion is predicted to occur along the left bank of the river and along low-lying areas within three years, with Table 4 showing the immersion area. The regional numerical groundwater flow model, as described in Section 3, was used in a comparative analysis of immersion control in the low-lying reservoir area under 7 combinations of engineering control measures, S1-S7 (Table 2), with the simulation results summarized in Table 4. Furthermore, Profile 1 and Profile 2 were selected to compare the groundwater level at selected points on the two profile lines under the different engineering control schemes, as illustrated in Figure 12.

Analysis of engineering measures on groundwater immersion control
Under the control schemes S1-S3, while the surface water in the reservoir river slowly advanced into the embankment after SWI, and the groundwater level inside the embankment could be controlled to remain below 30 m ( Figure 13), with the immersion area in the study area reduced by about 60% compared to that under Scheme S0.
Under the constant drainage level (29 m), well spacing (30 m), and distance to the cut-off wall (30 m) of the relief wells, with a decrease in the hydraulic conductivity of the cut-off wall, the total immersion area almost remains To sum up, in order to reduce the unfavorable influence of SWI on the agriculture and residential life on the bank of the reservoir, appropriate measures should be taken to reduce the groundwater level inside the cut-off wall and to reduce the thickness of the capillary saturation zone using relief wells in order to achieve effective protection. An anti-seepage wall can easily block the rapid seepage of groundwater; however, it breaks the original nature exchange between surface water and groundwater in the riparian zone.
Especially in the wettest season, the rising groundwater inside the embankment recharged from precipitation is difficult to discharge to the river, which leads to the groundwater inundation. For the dualistic structure stratum, relief wells can be used not far away from the dike and can greatly reduce the water table and reduce the pressure head at the bottom of the impermeable layer for the purpose of immersion control. The drainage level of the relied well is the most important parameter for immersion control, which directly affects the immersion area and drainage rate. Therefore, aiming at the phenomenon that the water storage of XGNPJ leads to a high groundwater level on the left bank of the reservoir, it is suggested to arrange the joint countermeasures of Scenario S2 with a cut-off wall and relief wells along the axis of the auxiliary dam, in which the hydraulic conductivity of the cut-off wall is less than 17.28 × 10 À3 m/d; the drainage water level of relief wells is 29 m, to alleviate the immersion of the XGNPJ reservoir bank.

CONCLUSIONS
The surface water impoundment (SWI)-driven groundwater immersion in the reservoir region is expected to be a serious environmental geological issue in a dual-formation structural reservoir bank. Aiming to deal with this issue, a three-dimensional (3D) groundwater model was established for reservoir immersion evaluation and control in the downstream area of the left bank of XGNPJ.
(1) The results of model calibration and sensitivity analysis indicate that the net recharge coefficient is most sensitive to groundwater flow dynamics, and a better fit of simulated values to targets is obtained, which can be used as a numerical tool for prediction of groundwater immersion in the plain reservoir area.
(2) Based on the 3D groundwater flow model, groundwater immersion under the effect of SWI was evaluated without any anti-seepage measures (S0). By the end of the third year of SWI, the 32 m groundwater level contour advanced 0.09-1.04 km into the embankment region, and the immersion area gradually increased to 10.91 km 2 . It can be expected that the actual extent of reservoir immersion in the future would be more severe than the model prediction. Better strategies for controlling groundwater inundation and scheme comparison will be necessary to choose optimal countermeasures to protect the adjunct aquifers from immersion by SWI.
(3) The joint engineering countermeasures, including the cut-off wall, relief wells, and land lift reclamation projects can effectively control reservoir immersion. With the construction of the cut-off wall, the drainage water level of the relief well is the most sensitive parameter for controlling reservoir immersion, while the hydraulic conductivity of the cut-off wall and the distance between the relief well and cut-off wall are the secondary sensitive parameters. According to the comprehensive comparative analysis, it is suggested that the joint immersion control engineering measures should be set as follows: the hydraulic conductivity of the cut-off wall is less than 17.28 × 10 À3 m/d; the drainage water level of relief wells is 29 m; the relief well spacing is 30 m; the vertical distance between the relief wells to the cut-off wall is 30 m; and the elevation of land lifting reclamation outside the embankment is more than 33 m.