Recharge estimates for an unconfined aquifer affected by surface mining and rehabilitation

Transcription

1 Hydrology of Warm Humid Regions (Proceedings of the Yokohama Symposium, July 1993). IAHS Publ. no. 216, Recharge estimates for an unconfined aquifer affected by surface mining and rehabilitation R. E. VOLKER Department of Civil and Systems Engineering, James Cook University, Townsville, Queensland, Australia 4811 M. R. CREES Australian Centre for Tropical Freshwater Research, James Cook University, Townsville, Queensland, Australia 4811 Abstract Field monitoring and parameter measurement together with numerical modelling of moisture movement in the unsaturated zone have facilitated the quantification of recharge to an unconfined aquifer in North Queensland, Australia. The land surface above the aquifer is disturbed by bauxite mining and results from the recharge simulation suggest that, although there are significant short term increases in recharge due to clearing and mining, the rates decrease again once mature vegetation is re-established. INTRODUCTION Underlying the Weipa Peninsula on Cape York in northern Australia is a relatively shallow aquifer which supplies all of the water resources for a bauxite mining operation and the associated urban development. Situated at approximately 12 S latitude, the area is subjected to intense but highly seasonal rainfall. Water levels in the aquifer rise relatively rapidly during the wet season and decline gradually under the influence of natural outflow and pumping over the following dry season. The major water withdrawal sites are located in an area bounded on three sides by the sea or saline estuaries so that salt water intrusion is a potential threat if water levels decline excessively. Consequently, yield from groundwater is particularly sensitive to quantity and regularity of recharge. Assessment of the long-term water supply able to be drawn from the aquifer is therefore inextricably linked with an accurate determination of recharge rates in both mined and unmined areas. The predominant recharge mechanism is infiltration of water through the highly permeable bauxite and ironstone layers below the soil horizon. The bauxite mining operation has a major impact on these overlying strata since it involves large-scale disturbance, removal of a significant amount of material, and replacement of overburden and introduction of different vegetation. Kaolin mining, a relatively recent activity at Weipa, also has a potentially significant impact because of the large-scale excavations and dewatering associated with it. Aquifer recharge is notoriously difficult to estimate accurately, although this is often conveniently ignored in groundwater modelling exercises; calibration is usually then undertaken by changing values of transmission and storage parameters and assuming recharge values are correct and not in need of adjustment. With a marked

2 482 R. E. Volker & M. R. Crées seasonal pattern of rainfall, as at Weipa, the recharge will suffer a similar fluctuation with time of year, and this adds to the problems of recharge estimation. The usual difficulty of estimating recharge under natural conditions is further exacerbated because the effect of the mining operations has not been quantified with relevance to recharge. THE GROUNDWATER SYSTEM Figure 1 shows the shape of the Weipa Peninsula under which there is a shallow unconfined or semi-confined sand aquifer and a much deeper confined aquifer which is part of the Great Artesian Basin. Only the shallow aquifer is considered here as it is the source of water for all industrial and urban operations. It extends approximately 40 km inland from the sea and its thickness varies from about 10 m near the axis of the Peninsula to less than 1 m near the Mission and Embley Rivers. The base of the aquifer is approximately 25 m above mean sea level in the east and drops to 15 m below in the west. From aquifer pump tests, Chapman (1963) estimated hydraulic conductivity of the sand at about 200 m per day and specific yield at Groundwater flow generally is from a central mound north and south to the Mission and Embley Rivers respectively, and west to Albatross Bay. Recharge to the shallow aquifer is by Fig. 1 Map of Weipa Peninsula showing locations of study sites.

3 Recharge estimates for an unconfined aquifer 483 infiltration of rainfall while discharge is to surface drainage, the sea, pumping bores and trenches. Mining temporarily eliminates vegetation and permanently removes the bauxite layer in the mined area so that any consequences for the recharge process and water balance of the aquifer need to be determined. The present emphasis in research on groundwater recharge in Australia is on understanding the impact of land use changes in areas where problems of salinity and waterlogging occur. The conditions that exist on the Weipa Peninsula are unusual because of the combination of high summer rainfalls, a shallow aquifer bounded closely by the sea on three sides, and a water table that rises almost to the surface in the wet season. Some studies in the southwest of Western Australia are relevant because, although the climate is different, there are lateritic profiles and a bauxite mining operation similar to Weipa. Examples of studies dealing with the effect of forest clearing on groundwater recharge are Peck & Williamson (1987), Hookey (1987), Sharma et al. (1987), and Schofield (1991). They document the significant increases in groundwater level brought about by the replacement of the native forest vegetation by pasture and crops. Recently Ruprecht (1991) reported groundwater levels under a mined area initially up to 3 m higher than in unmined control areas, but within 10 years of rehabilitation the differences had reduced to less than 0.7 m. Stream flows increased after mining by as much as 20% of rainfall before declining to almost pre-mining levels as the proportion of rehabilitated area increased. Although there has been limited work reported on groundwater recharge at Weipa, estimates based on measurements of water level rise have been made (for example, Chapman 1963). The uncertainty of the results is illustrated by the fact that estimates of the annual average recharge from previous studies have varied from Ml per annum to Ml per annum. There is no indication in these studies of attempts to evaluate the consequences of the mining operations for recharge. PROJECT STRATEGY The project was designed to increase understanding of the recharge processes so that better predictions of recharge quantities could be made and, in particular, to quantify the influence of the mining operation on recharge rates. Two sites were chosen to allow comparisons of water level response under unmined areas with, in one case an adjacent recently mined area and, in the other case an area of well established revegetation after mining many years before. Figure 1 shows the location of the study sites in relation to the Weipa Peninsula as a whole. They are identified by the shallow aquifer (SA) numbers of the observation bores (SA154 and SA155 at the eastern site and SA156 and SA157 at the western site). Two water level recorders (one on each observation bore) and one rainfall recorder were installed at each site. SA154 is located in an unmined area with natural vegetation while SA155 is in the corresponding mined and revegetated area. The ground surface at SA155 is approximately 2 m lower than the natural ground surface at SA154. At the other study site, bore SA156 is in the unmined area while bore SA157 is in a mined and revegetated zone. Unfortunately after being established in late 1988, the vegetation on half the area adjacent to SA156 was felled during the 1989/90 wet

4 484 R. E. Volker & M. R. Crées season followed by clearing and stripping during the middle of The vegetation on the remaining area was then felled early in the 1990/91 wet season. Consequently, the usefulness of the SA156/SA157 site was so significantly reduced that most of the analytical work comparing mined and unmined areas has been concentrated on the SA154/SA155 site and this is the only one considered further in this paper. In addition to water level and rainfall measurements at these sites, hydraulic properties of the material above the aquifer were measured both in situ and on samples removed and taken to the laboratory. A set of five access tubes for moisture content determination using a neutron moisture meter was established adjacent to each of the observation bores. Results from these were used to develop a water balance of the unsaturated zone as described by Crées & Volker (1993). The project included the development of a numerical model to simulate onedimensional unsaturated flow from the surface to the water table so that it could then be used to predict recharge for groundwater management considerations. MODELLING INFILTRATION THROUGH THE UNSATURATED ZONE The model used was SWIM (Soil Water Infiltration and Movement) developed by Ross (1990). It employs a finite difference solution of Richards' equation for one dimensional unsaturated flow. The solution algorithm represents a major advance over previous methods because a sinh transform allows the mass conserving water content formulation of the equation to be solved efficiently for all water content conditions including saturation. SWIM caters for runoff, surface storage, surface conductance, and évapotranspiration from up to four different types of vegetation and it provides for soil property variation with depth. Soil hydraulic properties determined during the testing program were used in the model. Rainfall and evaporation data obtained from the Bureau of Meteorology (see Fig. 1) were used as input to simulate the recharge between early 1990 and early 1991 for both SA154 and SA155. The model was calibrated by comparing the observed water levels and profile water contents with those predicted by the model, and adjusting the parameters until a match was obtained. Model development In developing the models of areas around SA154 and SA155 an appropriate soil profile composition was required together with hydraulic properties and boundary conditions. Following examination of drilling logs of bores relevant to the two sites, model profiles were described in terms of five components topsoil, bauxite, ironstone, clay and sand. Representative values of soil properties were determined for each of the profile components from the results of the testing programme. Drainage from the bottom of the model profile was estimated from the fall in water level observed toward the end of the dry season. Estimates of the vegetation parameters were made based on root depth analyses and using évapotranspiration values calculated from the water balance analysis (Crées & Volker, 1993).

5 Recharge estimates for an unconfined aquifer 485 The model of the unmined area (SA154) was calibrated by adjusting parameters to minimize the differences between model output and observed values of water levels and profile moisture contents. Adjusted parameter values were confined to within the ranges obtained in field tests. The profile for the model of the mined area (SA 155) was derived from that of the unmined area by removal of the bauxite layer. This model was then calibrated by adjusting only the vegetation parameters and comparing the model predictions with the observed water levels and profile moisture contents. Adjustment of initial estimates of soil property values during the calibration of the unmined area model gave the properties in Table 1 where K s is saturated hydraulic conductivity, 0 S is saturated moisture content. Drainage was included in the model by specifying a zero head bottom boundary and including an artificial retarding layer above the zero head boundary to simulate a head dependent loss rate from the profile. The conductivity of this retarding layer was determined from the drainage estimated from the dry season fall in water level. Two types of vegetation were modelled, a shallow rooted type to represent grass, and a deeper rooted type to represent forest. Values of vegetation parameters, such as a root depth constant and root length density, were determined by calibration. The calibration exercises and subsequent model production runs all used the time of peak ground water level early in 1990 as the start of simulation. This procedure reduces the uncertainty produced by unknown initial soil moisture conditions since it minimises the depth of unsaturated profile, and allows an assumption of very wet soil conditions above the water table to be used with confidence. Table 1 Calibrated values of hydraulic properties. Profile material K s (m s" 1 ) d s Topsoil (red earth) 1 X 10" Bauxite 1 X 10" Ironstone 3 X 10" Clay 1 X lu' Results Figure 2 shows the water levels predicted by the calibrated model of the unmined (SA154) profile compared with the observed water level (where AHD is Australian Height Datum). The recession of the ground water level is modelled very well but noticeable differences between modelled and observed levels occur when water levels start to rise during the wet season. The first difference is that the modelled response is much slower than the observed rise in level, and the second is that the modelled water level exceeds the observed level by about 1.5 m. The second discrepancy, the mismatch of highest water level, is brought about by surface runoff. A comparison of modelled and measured (by neutron moisture meter) water contents of the profile indicate that at the time of peak water levels there is an excess of about 180 mm of water from the model results. This compares well with the

6 486 R. E. Volker & M. R. Crées 15 SA S" 12 I < 11 E ^ 10 a> Days. _.Observed Modelled Fig. 2 Observed and modelled water levels for bore SA154. amount and timing of surface runoff estimated from a streamflow analysis and confirms that this is the source of the discrepancy. The model does not simulate runoff during this time because runoff is modelled as a function of the depth of surface storage and the model did not predict any surface storage. This indicates that either the modelled surface conductance should be reduced to allow ponding to occur, or that another mechanism such as subsurface lateral flow from the unmined area to the mined area occurs (and is not accounted for) when the water levels in the unmined area rise above the ground level in the mined area. The second discrepancy, the difference between the observed and modelled water level response indicates that the transport mechanism being modelled by SWIM does not allow sufficiently rapid movement of water through the profile at the start of the wet season. In SWIM, water movement is modelled as one-dimensional unsaturated flow through the soil matrix, assuming that conditions are uniform in the horizontal directions. Indications from the water balance analysis (Crées & Volker, 1993) are that flow through macropores in the soil structure contributes significantly to the total flow especially early in the wet season. The difference between the modelled and observed water level responses reinforces this suggestion that a quick-flow mechanism is acting because the simulated water levels lag behind the observed levels. In the next step, the calibrated model of the unmined area (SA154) was adjusted to simulate the changes made by the mining and rehabilitation process by removing the bauxite layer and reducing the thickness of the topsoil layer. This model of the mined area (SA155), with no alteration of parameters, then produced the results in Fig. 3 showing reasonable agreement between the modelled and observed water levels, although as with the model of the unmined area, the modelled response is slower than the observed. The extreme water levels are predicted accurately because, in this case, water levels rise above ground level and surface runoff occurs. The close agreement between modelled and observed values without the need to adjust vegetation parameters is in line with the findings of the water balance analysis

7 Recharge estimates for an unconfined aquifer SA Q CD 7 _l S 6 cd az^-^~ Days.--Observed Modelled Fig. 3 Observed and modelled water levels for bore SA155. (Crées & Volker, 1993) which showed that évapotranspiration in the rehabilitated area at SA155 has returned to levels close to those of the natural forest. The reasonably close correspondence between modelled and observed water level responses is significant because one possible effect of mining and subsequent rehabilitation was that remoulding of the soil structure and destruction of the macropore system would result in a reduction in quick-flow recharge. Since the observed water level response is more rapid than would occur from matrix flow mechanisms, either this has not happened or the macropore system has been reestablished since the area was rehabilitated. To test the transportability of model parameters and the profile modelling process, another site was modelled. This site, SA62, is one of several observation bores that have been monitored on a nominally daily basis since late Drilling logs for the site were used to devise the corresponding model layering. The calibrated parameters from the SA154 model were used to describe the hydraulic and vegetation properties. Drainage was modelled as a low conductivity layer above a zero head bottom boundary (as in the SA 154 model), and the conductivity of this layer was calculated from the observed drop in water level at the bore during the dry seasons. Figure 4 shows the comparison of modelled and observed water levels for SA62. The only adjustment made to the original configuration based on SA154 parameters was the total removal of vegetation. The model was originally formulated without any knowledge of site specific conditions; it was assumed that the area was representative of an unmined and naturally vegetated area and vegetation parameters were assigned from the SA154 model accordingly. After a number of model runs which showed no agreement between modelled and observed water levels, details of the actual conditions at the site were sought. It was then found that the area was cleared and free from any vegetation and hence the vegetation was removed from the model giving the results shown in Fig. 4. There are some differences in the magnitude of water level response to early wet season rain, but results differ from those at SA154 and SA155 in that modelled response occurs at about the same time as the actual response. This suggests

8 488 R. E. Volker & M. R. Crées SA62 26 S" E. 24 "g> 20 o _i œ Apr ct May Nov Jun Jul Jan Aug Mar-91 Date _ Observed Modelled Fig. 4 Observed and modelled water levels for bore SA62. that macropore quick-flow is not an active water transfer mechanism at this site, which is consistent with the area being cleared with a consequent disruption of the natural quick-flow pathways. Overall, the degree of agreement between modelled and predicted water levels suggests that the hydraulic property parameters are transportable. Recharge estimation The model was used to investigate the relationship between rainfall and the recharge to the aquifer. This required a consideration of the definition of recharge, and a determination of representative soil profiles and rainfall distributions. Care in defining recharge was necessary because of the potential use of the calculated values as input to a decoupled model of the groundwater system and because the water table rises close to the surface and inundates the region that would normally be considered as unsaturated in that model. For the purpose of these calculations, constant values of specific storage were assumed for each component of the profile and recharge was calculated from the volume of water required to saturate the profile between a lower and higher position of the water table. Three different representative rainfall distributions were defined. An average distribution applied the rainfall throughout the year according to the long term average monthly amounts, a lumped distribution applied the full year rainfall over a twenty day period commencing late January, and an intermediate distribution applied it over a 42- day period from mid-january to early March. Representative profiles to approximate the material above the aquifer sand were defined for various parts of the Peninsula. Figure 5 shows schematics of various land use types in conjunction with one representative profile. Details of the layer depths and water levels at the start of the simulation are also provided.

9 Unmined Natural U\ Recharge estimates for an unconfined aquifer 489 Unmined Cleared Mined Cleared Mined Revegetated _m -»- sx. Topsoil ; -s- Bauxite/Ironstone LEGEND: Clay f j Sand H Bottom Boundary Layer Hj Fig. 5 Schematic of subsurface profiles and land use types. Four profiles, four land use types, and six annual rainfall totals applied in two or three rainfall distribution patterns were modelled. In all 240 simulations were performed. Each simulation of recharge was for a period of 570 days commencing in late March when water levels were high. Monthly recharge estimates were calculated m o <l) It 01-1 r < ?<> U (20) -, S''" / ' ^ X / jftrc. -* (40) 50 I! Annual Rainfall (cm) i Unmined Unmined Mined Mined natural cleared cleared revegetated B ---A O --*-_. Fig. 6 Relation between annual recharge and rainfall for the average rainfall distribution.

10 490 R. E. Volker & M. R. Crées for one year commencing on day 191 (1 October) of the simulation. The initial water levels shown on Fig. 5 are the water levels at day 1 of the simulations. Simulation results for the profile shown in Fig. 5 and the average rainfall pattern have been used to derive estimates of recharge for different annual rainfall totals. These estimates are shown in Fig. 6 for the various land use types. A noticeable feature is the apparent upper limit to recharge in the cleared and mined cases. This limit corresponds to water levels reaching the ground surface so that further rainfall results in surface runoff and does not contribute to recharge. Simulations using the intermediate and lumped rainfall distributions indicate that this effect appears at lower cumulative rainfalls as the rainfall becomes more intense. Apart from this effect, there are no marked differences between the recharges for the natural landscape and for mined and revegetated areas although there are significant differences generated by clearing and mining before re-establishment of mature vegetation especially for lower annual rainfalls. Other representative profiles did not show any substantial change from this pattern. CONCLUSIONS Estimates of groundwater recharge on the Weipa Peninsula have been made by simulating infiltration of rainfall through the unsaturated zone. The results showed that there were no long term major changes to the recharge rates introduced by mining once vegetation was fully re-established. Significant short term changes in recharge resulted from clearing and from mining. Acknowledgements The authors thank their colleagues John Williams and Peter Ross from the Division of Soils, CSIRO for valuable advice. The research was supported by Comalco Aluminium Limited and the authors gratefully acknowledge the assistance of company staff in collection of data. REFERENCES Chapman, T. G. (1963) Groundwater hydrology of the Weipa Peninsula, North Queensland. /. Instn Engrs, Austral, July-August, Crées, M. R. & Volker, R. E. (1993) Water balance analysis of natural and rehabilitated areas at Weipa. To be presented at Int. Conf. on Environmental Management Geo-Water and Engineering Aspects (Wollongong, Australia). Hookey, G. R. (1987) Prediction of delays in groundwater response to catchment clearing. J. Hydrol. 94, Peck, A. J. & Williamson, D. R. (1987) Effects of forest clearing on groundwater. J. Hydrol. 94, Ross, P. J. (1990) Efficient numerical methods for infiltration using Richards' equation. Wat. Resour. Res. 26(2), Ruprecht, J. K. (1991) Hydrologie impact of bauxite mining and rehabilitation in South-West Western Australia. In: Proc. Int. Hydrology and Water Resources Symposium, Institution of Engineers, Australia. Schofield, N.J. (1991) Hydrological response to vegetation changes and its consequences in Western Australia. In: Proc. Int. Hydrology and Water Resources Symposium, Institution of Engineers, Australia. Sharma, M. L., Barron, R. J. W. & Williamson, D. R. (1987) Soil water dynamics of lateritic catchments as affected by forest clearing for pasture. J. Hydrol. 94,