Journal of Environmental Science and Engineering A9 (2020) 77-89 doi:10.17265/2162-5298/2020.03.001 D DAVID PUBLISHING

The Groundwater Model for Part of the Water Supply Source Aquifer for the City of Using FEFLOW Simulation

Narantsogt Nasanbayar Department of Environmental Engineering, School of Civil Engineering and Architecture, Mongolian University of Science and Technology,Ulaanbaatar14191,

Abstract: The cold, semi-arid environment shows a high variability in precipitation and river discharge, with a general tendency towards decreasing water availability due to increasing air temperatures and, thus, rising potential evaporation levels. The main watercourse near the city is the , fed by precipitation in the nearby Khentii Mountains. However, due to the absence of precipitation during winter and spring, the riverbed usually runs dry during these seasons, and observations show that the dry period has been extending within the last years. For many decades, the water supply of Ulaanbaatar has been exclusively based on the use of groundwater in the Tuul valley. However, in parallel with the city’s development, the extended groundwater aquifer shows a clear decline, and the groundwater levels drop significantly. Therefore, a groundwater management system based on groundwater model and MAR (Managed Aquifer Recharge) is proposed and a strategy to implement these measures in the Tuul valley is presented. The groundwater model research purposes of artificially recharging the Tuul River aquifer are to provide information for future improvement in solving shortage water supply related issues and to find simple, low cost, cheap, and reliable flow control methods to eliminate the Tuul River drying out in low flow season.

Key words: Finite difference, hydraulic head, conductivity, recharge.

1. Introduction The length of the river from its origin to the is 1,341km[2]. The origins of the river are Tuul Northern Mongolia is the part of a semi-arid, located on the southern slopes of the Khentii highly continental region where the Tuul River flows mountains. The upper part of the river basin is through the boundary between the last of the Siberian mountainous and almost entirely covered by forests. Taiga forest and Mongolian steppe lands bordering The study area lies within the south western spurs of with the [1].The region’s main waterway Khentii-the main ridge of the mountain region withthe is the Tuul River, which is a right-bank tributary of same name. The district is characterized by erosionand the Orkhon Riverflowing to the Selenge River in denudation relief due to the dismemberment spurs of Mongolia (See Fig. 1).The Tuul River basin is Khentii,the Tuul River valley and its tributaries. The bordered on the north by the river basinsof Kharaa landscape of the area from the source of the river Tuul and Eruu, in the east by the Kherlen River basin, to Ulaanbaatar city is determined by medium in the south by the Central Asian internal mountains, usually withsoft smooth contours, and drainagebasin and the west of the Orkhon River basin. below is mostly hilly ridges. The height of the The total length of the Tuul River is 742 km to the mountain peaks is in the range of 1,300-2,800m Orkhon river confluence, catchment area is 50,400km2. altitude, and marks the bottom of the Tuul River

 Valley in the district range of 1150 to 1450 M[3]. The Corresponding author: NasanbayarNarantsogt, MEng, main research field:Hydraulics. water level of the Tuul River fluctuates according to

78 The Groundwater Model for Part of the Water Supply Source Aquifer for the City of Ulaanbaatar Using FEFLOW Simulation

Fig. 1 Main water flow basins Mongolia—drainage basins.(a) hydrological drainage basin map of Mongolia; (b) map of river basins. annual high to low-flow cycles, with its average water warm period beetween May and September [7]. The flow being 26.6 m3/s [2]. daily maximum precipitation reaches 75 mmm . The climate of the region is extreme, with large During the year, in the Ulaanbaatar area is fluctuations in daily and annual amplitudes of air dominated by north west, north and south easterly temperature and low precipitation. winds. The highest wind speed is observed in spring The average annual temperature is minus 3.1°С and autumn [8]. with absolute maximum 39°С in July and an absolute Quaternary alluvial sediments are widespread along minimum of minus 49°С in December. Average the Tuul River valley and occur on the south facing absolute minimum air temperature is minus 41°C. slopes of high mountains, and in aquifers. The Low winter temperatures in the low cover snow cause following sediments are distinguished by their permafrostin soil 10-15m thick [4]. Below showed lithologic structure: recorded extreme climate parameters in the Alluvial sediimentaQII-IV: Alluvial sediments are Ulaanbaatar city wetter station. The Ulaanbaatar widespread in the Tuul River Basin. The sediments terminal site is due to evaporation and infiltration; the consist of mostly sandy loam and pebbles with clayey soils are slight to medium heaving soils in diameters from 2.3-3.2cm to 17.0-22.0 cm, seasonal freezing depth [5].The duration of the cold respectively. Boulders and gravel have diameters from period is 187 days with average daily air temperature 23cm to 24cm. Alluvial sediments are fromo 15.0 m to under 0°C. The average beginning and lasting dates of 29.0m in thickness in the vallleys, with most sediments the frosty period are October 9th and 14th of April, ranging from 12.0 to 15.0 m [9]. respectively. Annual rainfall is 261mm[6]. dpQI (Diluvial pluvial sediments) are widespread in Distribution of rainfall during the year is uneven, the 1st and 2nd flood terraces of river valleys, on the about 90% of annual precipitation falling during the south facing slopes of high mountains, in

The Groundwater Model for Part of the Water Supply Source Aquifer for the City of 79 Ulaanbaatar Using FEFLOW Simulation water-collecting depressions, and at the feet of boulder fractiondecreases, and the amount of silty clay mountains. particles increases.Detrital, fragment material detritus The sediments consist of pebbles, gravel with clay consists of well rolled up, petroal presented metamorphic loam, sandy loam, land waste, rock waste, boulders schist, sandstone, quartzite, granite and other rocks of and occasional rare boulders. The pluvial deposits the mountain valley of the Tuul River framing. have depth of 13.0 to 18.0m depending on the The thicknessof the upper stratum ranges from 2 to hillslope, while the depth of diluvial sediments ranges 37 m [9]. As a rule, there is an increase in thickness from 4.2 to 7.8 m [9]. from the sides of the valley to the center and down the Tuul River alluvial deposits identify two stratigraphic valley. In some places there is an increase of 2 stratums: upper layerαQ III-IV—alluvial deposits of thicknessof the upper layeron the main channel and quaternary period, gravel and pebble, boulder filled up the flow of the river Tuul (See Fig.2). with sand, and lower layer dominatesαQII-III1—clay In the eastern part of aquifer where simulation area with cement, conglomerates, argillite of Mesozoic locate, cross sections II-IItoIV-IV, have sustained ganozoic era [9]. capacity of 10-12 m. Near the east side of the valley Upper stratumalluvial formation is ubiquitous and there is marked reduction in sedimentary thickness up represented by gravel pebble deposits with boulders to 5m5 . Illustrations of this water table contours were and sand filling. Sand is usually assorted, with small surveyed in September 1978 and March 1979 drawn. areas of medium-coarse grained. According to Thickness of sedimentary increases from the eastern mechanical analysis of aquiferdeposits contents of part to the center. gravel and boulders are in range from 29% to 66% of In most of the considered site aquifer has a the total weight of the rocks, while the amount of gravel two-layer structure. The upper aquifer horizon is varies from 18%-39% sand content to 12%-36%[9]. confined to deposits of the surface, to the top of the Silt and clay fractions in aquifer either contained lower layers having a sufficiently high water respectively in amount of 0.4%-2.9% and permeability. The lower aquifer is enclosed in the 0.2%-1.1%[9].Asdepth of the gravelcontent increases, bottom sediments of the rock formation.

Fig. 2 Longitudinal hydrogeological section along the Tuul River aquifer. Based on data from RIBES (1979) [9].

80 The Groundwater Model for Part of the Water Supply Source Aquifer for the City of Ulaanbaatar Using FEFLOW Simulation

Fig. 3 Tuul River valley monitoring cross section II-II[9](see Fig. 2).

Lower alluvial stratum has developed almost amount of sand, silt and clay fractions increases. The everywhere in the study area except for some areas base alluvial formations of Tuul River valley close to the sides of the valley. It is represented by deposited carbon or Neogene sediments, sometimes pebble gravel sedimentation inclusion of boulders and aeolian, and deluvial alluvial formation. Sandstones filler in the form of silty clayey sand. In the content of carbon wasdiscovered in open wells under the the stratum, lenses or bands of sandy loam and loam alluvium in the eastern part of the central sources on can be traced almost everywhere, sometimes cross sectionsII-II, V-V. containing more gravel and pebbles and less sand. Along the Tuul River valley occur quaternary Thickness of individual layers varies from 1m to alluvial sand and gravel wiith sandy loam, clay and 5-8m[9]. pluvial sand, sandy loam deposits occur in The thickness increasesfrom the cross section aquifers.Within the study area, there are groundwater III-IIIto the aquifer center, western direction. of alluvial and diluvial deposits in river valleys, The minimum thickness is not more than 5-15m in sedimentation waters of Neogene and Cretaceous cross sections II-II and III-III, typical for the eastern sediments and fracture waters of Paleozoic rocks and part of the aquifer. For the rest of the large part of the granites. aquifer, the thickness of the lower stratum varies from Alluvial aquifer valley Tuul river deposits having 4 to 35 m or more, usually increasing from the sides to ubiquitous within the Central source area, the center of the valley. characterized by report material RIBES(Research Typically, the deepest interlayers situate wells in Institute on Building Engineering Studies(1979)and the bottom layers. Lower ground layers, for example other organizations). on the domestic drinking water intake site, have Both layers are hydraulically interconnected but particle size distribution within the following ranges: differ in the composition of water-bearing rocks, and pebbles and boulders 18%-35%, gravel 25%-37%, watery.Whole aquifer and the upper aquifer stratum sand 25%-44%, silt particles 3.1%-3.8%, clay are unconfined and contain free groundwater. Waters particles 1.0%-1.9% [9]. With depth, the content of enclosed in the lower aquifer, sometimes becoming gravel, pebbles and boulders decreases, and the head due to the presence in its roof loamy lenses and

The Groundwater Model for Part of the Water Supply Source Aquifer for the City of 81 Ulaanbaatar Using FEFLOW Simulation

Fig. 4 Tuul River valley monitoring cross section III-IIIbased on data from RIBES (1979) [9]. (see Fig. 2).

Fig. 5 Tuul River valley monitoring cross sections IV-IVbased on data from RIBES (1979) [9]. (see Fig. 2).

Interlayers. When drilling wells the overall level of its lowest position [10]. both aquifer stratumswas set. Pluvial alluvial formation, occurs on the left Under the conditions of undisturbed mode, the bank of the valley, from tributaries of the Tuul depth of the water table in the lowest position River, composed of alternating layers of loam, sandy (winter-spring period) varies from surface 4-8m, and loam and sand, containing crushed stone, gravel, at the highest within 0.5-3m (summer-autumn period). pebbles, boulders. Right bank of the northern side On the site of intake, depth of the decline in of the valley is dominated by pluvial alluvial groundwater levels for individual wells reaches formation valleys from the tributaries of the Tuul 10-13m at high standing groundwater table, 15-19m at River: Selbe, Uliastai rivers and other, as well as

82 The Groundwater Model for Part of the Water Supply Source Aquifer for the City of Ulaanbaatar Using FEFLOW Simulation diluvial-pluvial formation loops, possibly with high significant drawdown and shortages caused by the terrace outliers. operation of group water intakes of Ulaanbaatar city Deposits attributable to terraces present alternation water supply and its enterprises. of loam, sandy loam, sand with inclusions of gravel, The depth of the groundwater table from surface pebbles and boulders. Under the above-described low increases in the direction from the Tuul river to the diluvium formations or on the right side of the valley sides of the valley its tributaries and in water intake lie the Neogene sediments (interbedded clays, areas and varies by the season. siltstones, sandstones), carbon (sandstones, The bottom of the upper layer, usually passing conglomerates, shales) [9]. along the border of the upper and lower stratums, is Groundwater table is currently characterized by sometimes held below, in the lower layer.

Fig. 6 Monitoring and water supply wells in aquifer for water supply city of Ulaanbaatar.

Fig. 7 Groundwater table degradation by abstraction of water supply wells [6,10].

The Groundwater Model for Part of the Water Supply Source Aquifer for the City of 83 Ulaanbaatar Using FEFLOW Simulation

Fig. 8 UB aquifer zones classified by disturbance of GW gradient.

The general slope of the groundwater table, down to evaporation rate 11mm/year as inflow on top [1]. the valley, is marked byy local increase and decrease 3. FEFLOW Simulation of Model Area comparing tonormal undisturbed groundwater gradient, depending on location of group of intake wells. Nowadays, groundwater simulation and modelling are one of the main tools for groundwater aquifers, 2. Concept Model which visualise situation and condition water in Concept model is the first important step of underground poorous media for the protection of modelling attempt and requires data of geology, groundwater, as well as restoration and development hydrogeology, hydrology, and groundwater flow of aquifers. Groundwater level loggers collect more regime, water balance in the area of simulation. data, but since they cannot be controlled manually, The city has been supplied by deep water only software modelling and processing provide an exploitation wells that draw on groundwater sources option that is both fast and accurate [11]. from an unconfined aquifer that runs along the They are two modes of numerical model software: riverbed, exploiting additional alluvial deposits from  MODFLOW—finite differences method the Tuul River and tributaries. Groundwater is the  FEFLOW—finite elements method main source of drinking water supply for Ulaanbaatar, The water intake area with group of wells named A the capital city of Mongolia. zone is in the eastern part of the Central source area For the groundwater model of aquifer estimation and has 22 wells, located in three rows. In this zone also used RIBES monitoring borehole data starting two rivers Uliastai from north side and Khul river from upper cross section from I to down IV cross from south side are distributedand the area of A zone section also dam axis cross section data of flow restricts west side with Uliastai river and south-eastern control reservoir survey data. side with Tuul river. The wet months with the highest precipitation are The data of groundwater drawdown in monitoring June, July, August and September. A FEFLOW wells 51, 68are shown in Fig. 10 like red banner used simulation neglected winter snow precipitation as Eastern and Western boundary conditions and wells andonly used difference between rainfall and N8, 58, as calibration for groundwater fluctuation.

84 The Groundwater Model for Part of the Water Supply Source Aquifer for the City of Ulaanbaatar Using FEFLOW Simulation

Fig. 9 Conceptual model box of inflow and outflow water rate of aquifer.

Fig. 10 Simulated elevation of aquifer zone A-A, yellow points are water supply wells.

The main source of groundwater operational Fig. 11 graph showed that hydraulic conductivity reserves in the water intake area in the dry season variation is more sensible in simulation results. The between December to April, wherethereis lack of river calibration of simulated and measured values ratio flow, annuallyfills river runoff infiltration loss of the was chosen hydraulic conductivity. Tuul river surface water. 5. Calibration and Validation 4. Sensitivity Analyses The calculation of the operational groundwater The sensitivity analyses for FEFLOWsimulationare reserves in accordance with the above conditions for to identify which parameter variation ismore sensible their formation, and also taking into account the need to optimize simulation results [12, 13] and calibration. to solve the loop and inverse problems for evaluating In the FEFLOW model some parameterswere varied the functional reliability of the assumed design by simulation run and it noted which simulated parameters and obtaining the initial conditions, can parameter ismore sensible to the simulation results be realized only in the result of realizations of and measured groundwater level. mathematical models [14] describing both steady and

The Groundwater Model for Part of the Water Supply Source Aquifer for the City of 85 Ulaanbaatar Using FEFLOW Simulation

Fig. 11 Sensitivity analyses of FEFLOW simulation for groundwater model. transient filtering groundwater in a three-dimensional validation, reverse validation methodwas used, which mode in terms of unconfined flow. The calibration presents as changing small-simulated area to large ofinitial data ofhydraulic conductivity and area and ex boundary condition well monitored as for determination of calculated hydrogeological a validation. parameters was usedfor the calculation of reserves. In 6. FEFLOW SimulationResult the calibration Microsoft Excel functions—CORREL, DEV SQ average and mean squared deviation As mentioned before, the hydraulic conductivity of function, were used. the upper layer is 330m/day, and the hydraulic Validation is to create the correct model. It wasused conductivity of the lower layer 30m/day [15], giving to determine that the model is an accurate more suitable results (see Table 1) in visual representation of the real system. Validation usually fluctuation similarity of groundwater level in middle achieved by calibrating the model, the iterative part of area in wells in simulation. The red line process of comparing the model with the actual presents groundwater level in manually measured behavior of the system and using inconsistencies values, andthe silver blue lines present simulation between them and the data obtained to improve the results (see Fig. 13).When measured and simulated model. This process repeats until the model accuracy data are similar and demonnstrate calibrated curves is to be acceptable. Validation process builds more such as Fig. 13, then we can manage some scenarios extended model area of central source of A-A zone, in simulation that shows results in computer which are compared to the small FEFLOW model including suitable in practice use. In this FEFLOW simulation intake wells, input hydraulic head boundary weused groundwater level logger data in years from transformed to the real system (see Fig. 12). For a 2009 to 2011 same as measured data.

86 The Groundwater Model for Part of the Water Supply Source Aquifer for the City of Ulaanbaatar Using FEFLOW Simulation

Table 1 Calibration and validation in FEFLOW. Well Correlation Average square dev. Mean square dev. Eastern BC 68-1327.5 Sum of Average Mean Number of Western BC 51-1310.34 Correlation Sum of Sum of Sum of monitoring square square simulation coefficient Correlation ASD MSD Monitoring N8-1315.64 wells deviation deviation wells 58-1321.76 No. 68 0.952 12.28 24.10 No. 51 0.890 7.74 42.60 5 293-12 N8 0.917 1.79 29.76 No. 58 0.923 3.682 1.84 3.16 24.97 37.74 134.20 No. 68 0.979605 10.78 35.44 No. 51 0.976686 15.83 36.94 160 330-40 N8 0.936719 7.03 31.39

Calibration Calibration No. 58 0.919286 3.812 1.856 8.03 41.67 50.95 154.71 No. 68 0.972978 10.22 34.65 No. 51 0.97307 14.67 35.62 162 292-25 N8 0.840523 5.8 31.41 No. 58 0.835059 3.622 1.676 6.72 37.41 49.75 151.43 No. 68 0.979467 10.63 35.11 330-30 No. 51 0.971521 48.05 78.16 170 /1305.5- 132326\ N8 0.900138 6.69 29.59 No. 58 0.901302 3.653 1.856 7.15 722.51 48.78 191.64 No. 68 0.987448 11.28 35.7

validation validation 500-40 No. 51 0.972535 50.75 83.71 176 /1305.5- 132326\ N8 0.952723 9.19 32.04 No. 58 0.921597 3.653 1.856 8.88 80.1 47.74 199.15

Fig. 12 Extended area part of aquifer A-A zone for reverse validation.

The Groundwater Model for Part of the Water Supply Source Aquifer for the City of 87 Ulaanbaatar Using FEFLOW Simulation

Fig. 13 Groundwater level measured and simulated values in wells.

7. Discussion weather condition. To create ice storage from artificial sources in the In latest decades as the city develops and expands, winter cold season to increase groundwater levels by the consumption of domestic and industrial water melting during the dry season of March and April. supply of the city is increasing intensively but To understand the balance between consumption availability of water supply both now and in the future, and recharge in order to develop strategies for solving has become a pressing issue. Therefore, in this paper it water shortages using MAR (Managed Aquifer considered groundwater model built in FEFLOW Recharge). simulation and compared with measured results for To determine the most suitable and efficient model in upper part of Ulaanbaatar aquifer. conventional and non-conventional MAR The simulation should be as simple as possible, but methods,and combination of them, in extreme cold not as simpler and results should be close to the true weather conditions. measurement. Therefore, some aquifer data should be FEFLOW simulated results comparing scenarios as simple as possible. For example, hydraulic including changing recharging boundary conditions, conductivitywasnumbered equally in x, y, z directions. diversion drainage canals by old river channels, When weget good simulation results similar to building underground ice dams, and storing ice in measured data, then weshould possibly do following groundwater source areas. activities in FEFLOW program software. To recharge aquifer storage via infiltration of 8. Conclusions surface water during high flow season for use during The main parameter to determine simulation was low flow period such as winter for water supply. hydraulic conductivity. The hydraulic conductivity Aquifer infiltration rate from artificially recharging determined by RIBES [9]in 1979 shows in each drainage canal to study percolation of monitoring monitoring well different values from 4m/day to underground water flux during severe Mongolian 293m/day in two layers (see Table 1, Figs. 15 and 16).

88 The Groundwater Model for Part of the Water Supply Source Aquifer for the City of Ulaanbaatar Using FEFLOW Simulation

Fig. 14 Hydraulic conductivity of the first upper layer.

Fig. 15 Hydraulic conductivity of the second lower layer.

Fig. 16 Hydraulic conductivity in x direction.

The Groundwater Model for Part of the Water Supply Source Aquifer for the City of 89 Ulaanbaatar Using FEFLOW Simulation

To run FEFLOW simulation, some simplification in Mongolia, Tsukuba, Japan, 16-9. [4] Dashjamts, D.2015.“Permafrost and Geotechnical was applied in that the layers have the same hydraulic Investigations in Nalaikh Depression of Mongolia.” conductivity. Sciences in Cold and Arid Regions7(4): 438-55. The calibration is compared with the measured and [5] Dashjamts, D. 2013.“Geotechnical Problems of simulated results of hydraulic conductivity by Construction on Permafrost in Mongolia.”Sciences in Cold and Arid Regions5(5): 667-76. correlation coefficient and checked again by average [6] GIM (Geo-ecology Institute Mongolia). 1997.The square deviation and average mean deviation. All Research Works Report of Tuul River Water Reserves these results are then compared, calibrated and validated. Decreases Reason, Protection provision UB. Ulaanbaatar, Calibration and validation demonstrate that Mongolia: GIM. hydraulic conductivity of the groundwater aquifer on [7] GIM (Geoecology Institute of Mongolia). 1999.The Ecological Assessment of Tuul River. Ulaanbaatar, the upper side of the central source of Ulaanbaatar in Mongolia: GIM. upper stratum yields 330m/day; lower stratum yield of [8] GIM(Geoecology Institute of Mongolia).2010.The Report 30m/day is suitable for FEFLOW simulation [15]. of Water Quality, Aquatic Environmental Ecology Study. The simulation provides information about future Ulaanbaatar, Mongolia: GIM. [9] RIBES(Research Institute on Building Engineering improvements in solving Ulaanbaatar’s water supply Studies). 1979.Technical Report of EngineeringSurvey issues [17,18] in a simple, low cost and reliable manner. TOM-2. Moscow: RIBES. For this purpose, create useful FEFLOW simulation to [10] Unurjargal. D. 2009.Report of Research Works on understand natural hydrological conditions of the Tuul Groundwater Monitoring Measurements Data. Ulaanbaatar. [11] Kresic. N. 2007.Hydrogoelogy and Groundwater River, that close to the natural hydrological regime Modelling(2nd ed.).Boca Raton, FL: CRC Press. and groundwater flow. All simulation scenarios result, [12] Sltelli, A. 2002.“Sensitivity Analysis for Importance and evaluation of MAR methods have beenwritten in Assessment.”Risk Analysis22(3): 1-12. following article in Journal Water [13] McElwee, C. D., and Yukler, M. A.1978.“Sensitivity of Groundwater Models with Respect to Variations in (https://www.mdpi.com/2073-4441/11/12/2548) [1]. Transmissivity and Storage.”Water Resources Research References 14(3): 451-9. [14] Anderson, M. P.,and Woessner, W. W. 1992.Applied [1] Nasanbayar,N., and Morhlok, U. 2019. “Evaluation of Groundwater Modeling: Simulation of Flow and MAR Methods for Semi-arid, Cold Advective Transport.London: Academic Press, 281. Region.”Water11(12):2548.https://doi.org/10.3390/w111 [15] Nasanbayar,N. 2019.“Icing Phenomena for Managed 22548. Aquifer Recharge (MAR) and Its FEFLOW Simulation [2] Davaa, G. 2008.Surface Water Resources in Mongolia. Result.”PMAS59 (1): 229. Ulaanbaatar, Mongolia. [16] SIWRMM. “Strengthening Integrated Water Resources [3] Gombo, D.,andErdenetuya, M. 2004.“Hydrological Management in Mongolia.” Project.SIWRMM. Tuul river Changes in the Upper Tuul River Basin.”In Proceedings basin integrated water resources management assessment of the 3rd International Workshop on Terrestrial Change report. 2012.