Exploitation of alluvial aquifers having an overlying zone of low permeability: examples from Bangladesh

Transcription

1 Hydrological Sciences-Journal-dés Sciences Hydrologiques, 42(1) February Exploitation of alluvial aquifers having an overlying zone of low permeability: examples from Bangladesh MOHAMMAD MIRJAHAN MIAH Department of Water Resources Engineering, Bangladesh University of Technology, Dhaka 1000, Bangladesh K. R. RUSHTON School of Civil Engineering, University of Birmingham, Birmingham B15 2TT, UK Abstract Alluvial aquifers can often supply large quantities of water but if they are overlain by a low permeability zone, the recharge may be restricted with the result that the long term safe yield of the aquifer is greatly reduced. This paper describes a study of an alluvial aquifer in Bangladesh where there is a low permeability layer overlying the main aquifer. Pumping tests carried out in this aquifer were analysed using a numerical model which represents both the main aquifer and the overlying low permeability zone. Using the aquifer parameters deduced from the pumping test analysis, the numerical model was then used to represent five years of pumping. This long term simulation indicated that there would be a serious decline in the pumped levels and that a water table would develop in the main aquifer and would fall at a rate of almost five metres per year. Exploitation d'aquifères alluviaux recouverts d'une couche peu perméable: exemples du Bengladesh Résumé Les nappes alluviales peuvent souvent fournir de grandes quantités d'eau mais, si elles sont recouvertes d'une couche de terrains moins perméables, l'infiltration peut être limitée et la production de la nappe garantie à long terme peut s'en trouver considérablement réduite. Le présent article décrit l'étude d'une nappe alluviale du Bangladesh où une couche moins perméable recouvre l'aquifère principal. Les essais de débit effectués sur cette nappe ont été analysés à l'aide d'un modèle numérique représentant à la fois la nappe principale et la couche moins perméable. En utilisant les paramètres obtenus par les essais de débit, le modèle est ensuite utilisé pour simuler cinq années de pompage. Cette simulation à long terme indique que le niveau de l'eau dans le puits subirait un important rabattement tandis que le niveau de la nappe s'abaisserait d'environ cinq mètres par an. INTRODUCTION Alluvial aquifers are used extensively in supplying water for irrigated agriculture; a major concern is that the exploitation of the aquifer often leads to serious reductions in the water table elevation and pumped levels. However, there are situations where the main alluvial aquifer, which consists primarily of sand with layers of clay and silt, is covered by an upper silty-clay layer. This is a common situation in Bangladesh in the alluvial sediments of the three major rivers, viz. the Ganges, the Bramaputra and the Meghna. Although significant potential recharge occurs during the monsoon season, the upper silty clay layer regulates the flow into the main aquifer. This paper explores the important flow mechanisms when pumping occurs from an alluvial aquifer overlain by a low conductivity layer. Open for discussion until 1 June 1997

2 68 M. M. Miah & K. R. Rushton Studies of field conditions in areas with overlying clay layers have raised the possibility that water can pass through the clay to recharge the main aquifer system. Ahmad (1974) described early groundwater exploration studies in Bangladesh and showed that in many areas there is in excess of 30 m of clay between the ground surface and the main aquifers. He concluded that this lowers the water yielding potential of the aquifer systems and suggested that these areas are unsuitable for high capacity tubewells. In the Calcutta region of the Bengal Basin in India, much of the aquifer consists of coarse to medium grained sand overlain by a thick clay zone (Banerji, 1983). There is uncertainty about the possibility of recharge through the clay layers; even in the early 1980s, there were indications that the groundwater resources were not being fully recharged. In the Madras aquifer (Krishnasamy & Sakthivadivel, 1986) pumping from the deeper aquifer zones has exceeded the rate at which water can pass through the overlying clay layers with the result that new water table conditions have formed beneath the clay layers resulting in a perched water table above the clay layers. As well as limiting the recharge, the presence of the clay layers can lead to subsidence. Ramnarong (1983) described the uppermost aquifer in Bangkok which is overlain by a blue to grey to yellowish marine clay about 25 m thick. The initial concept was that this clay was impermeable but water balance studies indicated that vertical recharge could occur. The consequent settlement in Bangkok confirmed that some water has drained out of the overlying clay. The particular aquifer selected for this study is the Madhupur aquifer in Kapasia, Bangladesh. Groundwater development in this area has been limited because the hydrogeological conditions in the area are believed to be poor. However, a recent study (Mott MacDonald International, 1990) revealed that the thickness of the screenable formation in the top 120 m varies from 40 to 90 m. An examination of the available borehole logs indicated that the aquifer is a complex mixture of sand, silt and clay. A clay layer of thickness ranging from 5 to 30 m overlies the aquifer. The average annual rainfall in the study area is 2370 mm. About 80% of this rainfall occurs during the monsoon months of July-October. The first stage of the analysis was to identify the important flow processes. Next, a numerical model was developed to represent pumping from this aquifer. The twozone radial flow model of Rathod & Rushton (1991) was used but with developments to represent flow mechanisms in the overlying low conductivity layer. Aquifer parameters were chosen so that the model represented pumping tests and the subsequent recovery. The same model was then used to examine the long term response of the aquifer system and determine whether a deterioration of the aquifer conditions might be likely to occur. FORMULATION OF PROBLEM First stage: description of the system The first stage in the formulation of the problem was to develop a diagram which represents the important flow mechanisms. This diagram did not need to represent all the detailed complexities of the specific problem but it had to represent the major

3 Exploitation of alluvial aquifers 69 flow mechanisms. The formulation was considered in two parts: the flow through the main aquifer and the response of the overlying low permeability layer. Figure 1 shows the aquifer system divided into two; the main aquifer is represented in Fig. 1(b) and the overlying zone in Fig. 1(a). (a) RECH overlying zone * * t M M INF T FWT (b) low permeability zone upper permeable zone lower permeable zone Fig. 1 Idealization of the two parts of the aquifer system (a) overlying zone, (b) main aquifer (a) wis. -4 f- RECH -W-4 M * RECH FWT GW BASE -4 f- ^H XvGW INF = VPERM WT-GW WT - BASE FWT = INF-RECH xdelt SYO INF = VPERM m - BASE WT - BASE FWT = INF - RECH x DELT SYO RECH if RECH < VPERM INF = RECH FWT= 0.0 L.i INF rwt i-v-gw if RECH > VPERM INF = VPERM FWT= INF-RECH ynf T SYO Fig. 2 Method of calculating infiltration (INF) from the overlying zone (OZ) to the main aquifer (MA).

4 70 M. M. Miah & K. R, Rushton Considering first the main aquifer, detailed studies have shown that an alluvial aquifer can be represented as a layered system (Kavalanekar et ah, 1992). Figure 1(b) shows the major flows through an alluvial aquifer towards a pumped borehole. This situation can be idealised as two permeable zones with an intermediate low permeability zone. The borehole draws water from both of the permeable zones. Inflow (INF) to the upper surface of the aquifer system can occur due to recharge or flow from an overlying low permeability zone. Next, inflows to the low permeability overlying zone (Fig. 1(a)) occur due to recharge (RECH) from the soil zone. Another aspect of the water balance is the fall of the water table (FWT) in the overlying zone. The quantity of water moving from the overlying zone into the main aquifer system also depends on the vertical hydraulic gradient in the overlying zone. Figure 2 illustrates three possible conditions: (a) the water table (WT) and the groundwater head (piezometric head) of the main aquifer (GW) both lie within the overlying zone, Fig. 2(a); this means that the main aquifer is under confined conditions; (b) the groundwater head in the main aquifer (GW) is below the overlying zone, and consequently unconfined conditions apply whilst perched water table conditions apply in the overlying zone; and (c) most of the water has drained out of the overlying aquifer with the result that the water table (WT) is close to, or at the base of, the overlying zone. Second stage: quantify vertical flow through overlying zone The second stage of the formulation was to develop a method of analysing the problem as described above. Numerical methods are suitable for this type of problem. The two-zone radial flow model described by Rathod & Rushton (1991) can represent the flow in the main aquifer as shown in Fig. 1(b); full derivations of the finite difference equations were given in that paper. However it was necessary to adapt the overlying layer simulation to represent the conditions in the low permeability overlying zone. The relevant equations for the overlying zone are shown in Fig. 2. Two quantities need to be calculated: (i) the inflow, INF, from the overlying zone to the main aquifer; and (ii) the fall in the water table in the overlying zone, FWT. Four further parameters are required: VPERM, the vertical permeability of the overlying zone; BASE, the elevation of the base of the overlying zone; SYO, the specific yield of the overlying zone; and DELT, the time step of the calculation. An examination of the equations in Fig. 2(a) shows that the value of the inflow to the main aquifer, INF, is determined from a Darcy calculation in the vertical direction. The vertical head gradient in the overlying zone equals (WT - GW) and this acts over a vertical distance (WT - BASE). The fall in the water table in the overlying zone, FWT, depends on the difference between the water draining to the main aquifer, INF, and the recharge to the overlying zone, RECH; the fall is also proportional to the size of the time step.

5 Exploitation of alluvial aquifers 71 For situation (b) the vertical hydraulic gradient is unity because atmospheric conditions apply at the base of the overlying zone. In situation (c), the assumption was made that any recharge moves immediately through the overlying zone to be available to enter the main aquifer. In practice there would be a delay (Kruseman & de Ridder, 1990); nevertheless, the overall water balance is valid. This model can be used both to analyse pumping tests lasting for a number of days and to consider the long term aquifer response. Pumped well KAP/13 Sand x D, A 54.8 m 47.7 m D 2J\ 54.8 m X Sand 87.9 m o 10 Radial distance (m) Fig. 3 Details of tube-well and observation piezometers used for pumping test at KAP/ ANALYSIS OF PUMPING TESTS Three pumping tests were carried out in the Madhupur aquifer. One test was selected to illustrate the manner in which aquifer parameters can be estimated from the analysis of field results. The layout of the test site is shown in Fig. 3 and in Fig. 4 (a)-(c) results are presented of field drawdowns in the pumping tube-well, KAP/13, and in four observation piezometers. The tube-well penetrated 89.3 m below ground level (82.7 m below the rest water level); the upper 12.8 m was a clay layer with the remainder primarily of sand but with some clay zones; the slotted casing extended from 29.6 to 87.8 m below ground level. Before pumping, the rest water level (RWL) in the main tube-well was 6.6 m below the reference level. Pumping at a rate of 5141 m 3 d" 1 continued for three days and recovery was monitored for a further two days. Drawdowns during the pumping and recovery phases were monitored in: P the pumping tube-well, Fig. 4(a); SI an observation piezometer, 10 m from the tube-well, 11.8 m below RWL, Fig. 4(b);

6 72 M. M. Miah & K. R. Rushton (a) &4 Time (day) 1E I I I I 11 I I I I I I I 11 I I I I I I 111 M i l I I I I I I I il (b) 1 (0! 3.00 I Fig. 4 Pumping test at KAP/13: comparison between field and modelled results (a) in pumped well; (b) in observation piezometers at 10 m from tube-well; arid (c) in observation piezometers at 100 m from the tube-well.

7 Exploitation of alluvial aquifers 73 Dl an observation piezometer, 10 m from the tube-well, 54.8 m below RWL, Fig. 4(b); S2 an observation piezometer, 100 m from the tube-well, 11.8 m below RWL, Fig. 4(c); and D2 an observation piezometer, 100 m from the tube-well, 54.8 m below RWL, Fig. 4(c). The field results are represented in Figs.4(a)-4(c) by the discrete symbols while the lines represent model results which will be described later. Examination of the field results A careful examination of the field responses at the five locations provides important information about the aquifer response to pumping: - in the pumped tube-well there was a sudden fall in the water level of about 5 m which increases to 8 m during the three days of the test; the initial recovery to 4 m was rapid; the recovery was almost complete after two days; - at the observation piezometers 10 m from the pumped tube-well, the responses of the shallow (SI) and deep (Dl) piezometers were similar although during the pumping phase the drawdown in the deep piezometer was slightly larger suggesting that there are some low permeability zones in the main aquifer; the initial drawdowns were smaller than in the pumped tube-well although the general response during pumping and recovery was similar; and - for the observation piezometers at 100 m there was little difference between the shallow (S2) and the deep (D2) piezometers; however the response to pumping was slower than in piezometers SI and Dl yet after about 0.01 day (about 15 min) the drawdown approached 0.3 m. After three days the drawdown was increasing at a slower rate yet the drawdowns had not reached equilibrium. The recovery curve is similar to the pumping curve; after two days the residual drawdown was about 0.1 m. Estimation of aquifer parameters using the two-zone model These field drawdowns can be analysed using the two-zone numerical model. In that model the mesh spacing increases logarithmically from the pumped tube-well; time steps also increase logarithmically from 10" 7 day during each phase of pumping or recovery. The aim was to obtain the best match for the five time-drawdown curves both for the pumping and the recovery phases using a single model. Particular emphasis was placed on the observation piezometers at 100 m. Figure 4 shows the match between the field readings and the modelled results shown by the continuous lines. The parameters deduced from this analysis are as follows: overlying layer: vertical hydraulic conductivity =0.007 m d" 1 specific yield = 0.03

8 74 M. M. Miah & K. R. Rushton upper zone: horizontal hydraulic conductivity = 11.5 m d" 1 confined storage coefficient = well loss factor = 4.0 middle zone: vertical hydraulic conductivity = 0.01 m d ' equivalent thickness = 1.0 m lower zone: horizontal hydraulic conductivity = 11.5 m d" 1 confined storage coefficient = well loss factor = 3.0 The well loss factor represents the restriction on the movement of water through gravel packs and the well screen into the tube-well; the hydraulic conductivity adjacent to the tube-well equals the standard value divided by this factor. An important issue is the tolerances of these parameters. Particular attention needs to be paid to the vertical permeability and specific yield of the overlying layer. Table 1 lists the drawdowns in piezometer S2, which was 100 m from the pumped tube-well, at the end of the pumping phase (3 days) and recovery phase (5 days). Different values of the vertical hydraulic conductivity and specific yield were considered. Model results for the end of the pumping phase and after two days recovery are included in Table 1. Table 1 Sensitivity analysis of the overlying zone parameters; drawdowns in piezometer D2 Vertical hydraulic Specific Drawdown (m) Drawdown (m) conductivity (md 1 ) yield at 3 days at 5 days field The results in Table 1 demonstrate that the response in the distant observation piezometer was sensitive to the vertical hydraulic conductivity; if the vertical hydraulic conductivity was set too low the drawdowns did not level off at the end of the pumping phase; if they were set too high the levelling off was too rapid and the recovery was completed too soon. A vertical hydraulic conductivity of m d" 1 provided an adequate match between model and field results. Values of m d" 1 and 0.01 m d" 1 were not acceptable. The observation well responses were not very sensitive to the specific yield of the overlying layer as indicated by the final line of Table 1. The lower value of 0.03 was used for the predictive studies to represent the slow drainage that occurs from clay soils due to their low permeability. Although reliable values of certain aquifer parameters were deduced by matching

9 Exploitation of alluvial aquifers 75 the model to the pumping and recovery field results, the analysis would have been enhanced if piezometers had also been provided in the overlying lower permeability layer. REPRESENTATION OF LONG TERM RESPONSE The next stage in the study was to consider the response when this aquifer system is used to provide water for irrigation. The following assumptions were made which allowed the formulation of a problem to represent a typical long term response. For each year included in the simulation (the year started at the beginning of December) the recharge and abstraction patterns were idealised as follows: December-April: May: June-September: October-November: abstraction no abstraction, no recharge recharge no abstraction, no recharge. Further details are as follows (each year was assumed to consist of 12 months of 30 days): Abstraction rate: 4900 m 3 d 1 for 12 h, recovery for 12 h; continuing for days. Spacing of the wells: 700 m by 700 m (equivalent outer radius 395 m), meaning the intensity of irrigation was equivalent to 5 mm d" 1 (0.5 x 4900 m 3 d' m 2 ) which is suitable for growing rice (total 750 mm in 150 days). In practice there would be a variation in water demand during the land preparation and growing season. Recharge: the four recharge months were divided into ten day periods; total recharge (mm) for each ten day period was as follows: 0, 30, 0, 15, 0, 30, 30, 0, 30, 30, 15, 0; hence the total annual recharge was 180 mm. In estimating this recharge, note was taken of the high runoff which occurs during the intense monsoon rainfall. The effect of higher recharge intensities is considered later. Initial conditions: taken to be the same as for the pumping test with the initial saturated depth of the overlying clay layer of 6.2 m. Aquifer parameters: in general the parameters were the same as those deduced from the pumping test; during the analysis of the pumping test it was not possible to deduce a value for the specific yield of the main aquifer; from information in similar areas it is taken to be 0.15 (Kavalanekar etal., 1992; Walton, 1987) This standard example was called Example A. In the simulation, pumping and recovery phases occurred each day and the time step was reduced to 10' 7 at the start of each phase. Figure 5 indicates the response in the pumped tube-well during the first 150 days. The change in the pumping level between successive pumping and recovery phases was about 7.0 m; when the non-pumping level was about 6 m the maximum and minimum levels showed a change in slope due to the changing

10 76 M. M. Miah & K. R. Rushton conditions in the overlying layer from condition (a) to condition (b) of Fig. 2. The pumped drawdown after 150 days is m and the non-pumping level is m. Figure 6 shows the response for a complete year. At the end of the pumping season the pumped drawdown was 17.2 m; during the recharge period of June to the end of September, drawdowns recovered to 8.6 m and remained at this value during October and November when there was neither recharge nor abstraction. Results for five years of pumping are summarized in Table 2. The maximum pumped drawdown and final recovery level at the end of the first year are recorded in the column headed Example A beneath the heading drawdowns. In addition results are quoted at the end of the pumping phase and the end of the year for years 2, 3 and Time (day) J I L g "p & Fig. 5 Predicted response in tube-well over 150 days with pumping for 12 h and recovery for 12 h Time (day) J Fig. 6 Predicted response in tube-well for one year of operation.

11 Exploitation of alluvial aquifers During Year 5 the pumped drawdown reached 34.3 m with recovery to 23.8 m at the end of the recharge period. This rapid fall in groundwater level averaged more than 4.7 m per year which could mean that the aquifer resources would be exhausted within a decade. The fall in the pumped levels to 34.3 m during Year 5 would be likely to result in a deterioration in the tube-well performance. Three further situations were considered with the results presented in Table 2. For Example B: the spacing between the tube-wells was increased to 900 m which meant that the equivalent radius of the area supplying a single tube-well was 508 m. The abstraction rate remained at 4900 m 3 d" 1 which meant that the average intensity of irrigation was reduced from 5.0 to 3.0 mm d" 1 to compensate for the increased area from which the tube-well withdrew water. Even with this reduced abstraction rate the maximum pumped drawdown was 23.0 m compared to 34.3 m with the standard spacing and the average groundwater head fell a total of 13.7 m in five years, equivalent to 2.7 m each year. In Example C all parameters were the same as for Example A apart from the annual recharge which was increased from 180 to 300 mm year" 1. As indicated in Table 2 the fall in groundwater head during the five year period would be 19.9 m which is equivalent to 4.0 m per year. Only if the quantity of water moving through the overlying layer was equivalent to 750 mm in a year would no mining of the groundwater occur. Although the annual rainfall is in excess of 2000 mm in a year, the high intensity of much of the monsoon rainfall leads to very high runoff with the result that the quantity available to move through the overlying zone would never be as high as 750 mm in a year. For Example D, the effect of pumping continuously was investigated. Instead of pumping for 12 h with a recovery for 12 h, the pump operated continuously at half the original pumping rate. Due to the continuous pumping, the pumped drawdowns were less, 13.7 m compared to 17.2 m for the intermittent pumping during Year 1 Table 2 response to long term pumping for four different scenarios; the first five lines give the specified parameters and the remaining results are drawdowns in the pumped well in metres. Example no. Details of problem: Well spacing (m) Irrigation intensity (mm" 1 ) Abstraction rate (m 3 d" 1 ) Duration per day (h) Annual recharge (mm) Drawdowns (m) at: End of 1st year pumping, End of first year End of 2nd year pumping, End of third year End of 3rd year pumping, End of third year End of 5fh year pumping, End of fifth year A B C D

12 78 M. M. Miah & K. R. Rushton and 29.8 m compared to 34.3 m during the fifth year of pumping. However, whether the pumping was intermittent or continuous, the average drawdowns after five years remained at 23.8 m. CONCLUSIONS This study has shown that, when an alluvial aquifer is overlain by a less permeable layer, it is essential to consider the combined effect of the main aquifer and the overlying layer, both when analysing pumping tests and when predicting the long term behaviour of the aquifer system. A two-zone radial flow model was modified to represent the important mechanisms of a water table forming in the main aquifer leaving the overlying zone as a perched aquifer which still transmitted water. The two-zone model was used to analyse pumping tests in the Madhupur aquifer in Bangladesh. The model simulated successfully the pumping and recovery phases of the pumping tests and reproduced the main features of the pumped and observation piezometer responses. Values of many of the aquifer parameters were estimated but the analysis would have been enhanced if piezometers had been provided in the overlying layer. The original workers who carried out the pumping tests suggested that the aquifer would provide a high, sustainable yield. However, when the model, which includes the effect of the overlying layer, was used to examine the long term responses over a period of five years, questions were raised about the sustainability of the yield. Each year a typical pattern of pumping and recharge is considered. The responses were simulated for different values of irrigation intensity and annual recharge. For each of the four cases considered, the groundwater head within the main aquifer and the water level in the pumped well showed a rapid decline. The fall in pumping level would be likely to lead to a deterioration in the efficiency of the tube-wells. Consequently, as abstraction increases in this type of aquifer, it is essential that monitoring is carried out in the overlying zone as well as in the main aquifer to determine how rapidly the resource is being used. This method of analysis is being applied to other areas. However, even before a detailed study is carried out it is possible to conclude that mining of water from aquifers used for extensive irrigation is likely to occur unless there is unusually high recharge. For example, if the intention is to use the pumped water to irrigate a crop in the dry season which requires more than 700 mm of water, the annual recharge is unlikely to be this high and therefore mining of the groundwater is almost certain to occur. REFERENCES Ahmad, N. (1974) Groundwater Resources of Pakistan. Ripon Printing Press, Lahore, Pakistan. Banerji, A. K. (1983) Importance of evolving a management plan for groundwater development in the Calcutta Region of the Bengal Basin. In: Groundwater in Water Resources Planning, IAHS Publ. no. 142, Kavalanekar, N. B., Sharma, S. C. & Rushton, K. R. (1992) Over-exploitation of an alluvial aquifer in Gujarat, India.

13 Exploitation of alluvial aquifers 79 Hydrol. Sci. J. 37, Krishnasamy, K. V. & Sakthivadivel, R. (1986) Regional modelling of non-linear flows in a multi-aquifer system. UNESCO Regional Workshop on Groundwater Modelling, Roorkee, Kruseman, G. P. & de Ridder, N. A. (1990) Analysis and Evaluation of Pumping Test Data, Publication 47, International Institute for Land Reclamation and Improvement, Wageningen, The Netherlands. Mott MacDonald International Limited (1990) Report on exploration drilling in Kapasia Upazila, Working Paper 53, IDA Deep Tube-well II Project, Bangladesh Agricultural Development Corporation, Ministry of Agriculture, Government of Bangladesh. Ramnarong, V. (1983) Environmental impacts of heavy groundwater development in Bangkok, Thailand. In: International Conference on Groundwater and Man, Vol 2, Australian Government Publishing Service, Canberra, Australia. Rathod, K. S. & Rushton, K. R. (1991) Interpretation of pumping from two-zone layered aquifers using a numerical model. Ground Water 29, Walton, W. C. (1987) Groundwater Pumping Tests, Design and Analysis. Lewis Publishers, Michigan, USA. Received 2 January 1996; accepted 11 June 1996

14