EVALUATION OF GROUNDWATER FLOW PATTERNS IN FRACTURED ROCK AQUIFERS USING CFCS AND ISOTOPES
REPORT NO: 1009/1/03
March 2003

EXECUTlVE SUMMARY

Introduction

Fractured rocks are the major source of groundwater supplies in South Africa. Over 90% of the surface area of South Africa, groundwater occurs in secondary openings in so-called hard rocks. These openings occur in very irregular fashion and make prediction of aquifer properties very difficult. Reliable tools for resource evaluation in fractured rock aquifers are nevertheless needed for efficient management of these resources. The Water Research Commission identified hard rock (or fractured rock) aquifers as a theme of groundwater research requiring a high priority in terms of funding. The present project is therefore a part of the WRC Fractured Rock Research Programme.

The chlorofluorocarbon gases, CFC-11, CFC-12 and CFC-113, were developed during the 1930s. These gases are chemically stable, safe and have convenient boiling points resulting in widespread use in society. Used CFC gas accumulates in the atmosphere where it poses a serious hazard to' stratospheric ozone. This has led to successful international action to reduce global CFC emissions (Montreal Protocol). The known growth rates of CFCs in the atmosphere over the last fifty years, the rapid mixing in the world's atmosphere, their solubility in water and their good chemical stability have enabled this hazard to become a useful tool for hydrologists. This phenomenon is used to trace water movement in the oceans, in surface water and in groundwater and will likely remain useful for a few decades in the near future.

The development of a reliable sampling and analytical procedure for CFC in groundwater by the US Geological Survey has ensured wide application of this technique during the last decade. CFC applications in groundwater rest on the assumption that groundwater at the water table is in equilibrium with atmospheric air, including its CFC component, following the laws of solubility. Once water moves in the saturated zone below the water table, it will not be able to acquire or lose any additional CFC. The CFC quantity in the water will be characteristic of the atmospheric CFC level prevailing during the last contact with the atmosphere. This forms the basis of the concept of CFC dating of groundwater. The steep increase of atmospheric CFC levels with time ensures that dates can be well defined up to the mid 1990s. This is in contrast to tritium and radiocarbon where the input curve has become rather flat since 1963. The dates derived in this manner can be considered model recharge dates.

Earlier projects in the Table Mountain Group and Karoo Group sandstones (Weaver et al 1995, Weaver and Talma 1999) have shown that CFCs can be quite useful groundwater evaluation tools in South Africa. The existing techniques for groundwater dating, or generally, tracing water residence times underground, use radiocarbon (¹⁴C) and tritium (³H) which each have their advantages and disadvantages. These techniques have been in use in the southern African region since the late 1960s and have been proven immensely useful. The applications rely on specific flow models to interpret the isotope measurements and for fractured rock aquifers with their very random and irregular flow character, the applicability of such models are difficult to assess. The help of any additional data is therefore beneficial for accurate resource evaluation.

Objectives

The overall aim of the present project as formulated in the agreement between CSIR and WRC is:

• To develop a method of integrating and analysing groundwater age data provided by ¹⁴C, CFC- 11, CFC-12, CFC-113, tritium and other isotope analyses so that groundwater mixing ratios for fractured rock aquifers can be determined.
• To investigate the application of these mixing ratios to groundwater resource evaluation so that the reliability of these evaluations can be refined.

Project Approach

The first task was to identify and select suitable sites where case studies were likely to be successful within budget and time limits. The requirements for suitable study sites were formulated as:

• Fractured rock aquifers where it is probable that mixing of young and old water occurs
• The hydrogeology of the site should be reasonably well known.
• Some isotope and/or chemical data should be available.
• Groundwater recharge ages should be within the past 10-20 years, to make full use of the major CFC variations.
• A sufficient number of sampling points (boreholes) should be available for sampling.
• Established pumps in those boreholes would simplify the sampling procedure.

The project team contacted a number of hydrogeologists, reviewed existing reports and visited a few sites. Thirteen sites were evaluated and subsequently three sites were selected:

• Leeukuil, a small gneiss/granitic aquifer to the west of Pietersburg/Polokwane in Limpopo Province.
• A selection of 11 springs in the dolomitic areas of the Chuniespoort Group between Pretoria and Zeerust.
• Leeu Gamka in the Karoo, south of Beaufort West, where there is clear evidence of mixing between irrigation and groundwater in farmers' boreholes.

Water from a selection of boreholes from these sites was sampled and analysed for the tracers mentioned above, for stable isotopes and for chemistry. The data obtained were assessed in terms of description of the processes occurring underground and evaluation of the time scales involved.

Groundwater flow model

The interpretation of tracer concentrations in groundwater relies strongly on the input behaviour of the specific tracer that is employed, the perceived flow path of the specific tracer and the time ranges involved. The complexity of water flow and tracer behaviour and the availability of aquifer information require appropriate approaches. For the present project the simplest types of models, the lumped parameter models, were used. In these cases one assumes linear relations between tracer concentrations in the input (recharge) and in the output (groundwater) and that the aquifer properties can be described with only a few parameters that are applicable to the entire aquifer. This simplification is necessary when only a minimum of information about the aquifers is actually available.

The input functions of the tracers are the tracer concentrations during recharge and, before AD1950, these were essentially constant over time. For that period one can then simply apply conceptual models as analytical functions between in- and output values. Since 1950 however atmospheric concentrations have changed. ¹⁴C and tritium in recharge has increased due to the contamination from the atmospheric nuclear weapon tests between 1955 and 1963 and atmospheric CFC levels have steadily increased. These increases have greatly expanded the usefulness of tracers for groundwater evaluations, but necessitate a numerical rather than analytical approach to modelling.

Such a set of models was set up on a spreadsheet, "MRTMODELS", where the tracer inputs and numerical integration were calculated in annual steps. The inputs of the five tracers are required and the aquifer responses are simulated with a range of aquifer parameters that are presented graphically and numerically.

MRTMODELS presents the data for two lumped parameter models;

• The piston flow model describes the flow of water as that of a single parcel flowing along a flowline through the aquifer. The water sample will then present the tracer levels as they were during recharge subject only to decay in the case of the radioactive tracers. This model produces a single age for a sample, which is the flow time from recharge point to sampling point. This model is applicable to confined aquifers.
• The exponential flow model (or box model) is more generally applicable to phreatic aquifers. It assumes that the water at the sampling point is a mixture of contributions of water having a range of ages. The distribution of ages is an exponentially decreasing function with a characteristic time. The mathematics is similar to that of a mixed box model in which water with variable input functions is mixed and the mixture overflows the box. The tracer concentration in the mixture represents that of the mean residence time (MRT) of water in the box and is the characteristic time of the exponential distribution of input functions.

Each of the models produces relations between tracers as functions of age or MRT against which real- life measurements can be fitted to test the validity of the models. This was done in the case studies that were undertaken during this project.

Field study: Leeukuil

In this granite/gneiss aquifer groundwater occurs in the weathered overburden and the fracture base rock. Two parallel sub-catchments have been studied by sampling from existing pumped boreholes along the two drainage lines. Tritium and CFC both indicate significant quantities of pre-1955 recharge water in the boreholes. Halfway down both sub-catchments, nitrate levels increase due to seepage from manure heaps from the piggery that is located in the catchment. These pollutant waters are indicative of excessive local recharge and higher CFC levels.

The flow pattern resembles that of flow along a surface somewhat below the water table to which local seepages are added. The tracer data fit those of the exponential flow model and the binary flow model. The latter model appears to be more applicable for the samples that are contaminated by piggery waste.

Field study: Dolomite Springs

The springs in the Dolomites are the result of the compartmentalization of the aquifer by dolerites, thereby forcing the water to the surface. Each sample therefore represents its own catchment, or compartment. The aquifers consist of some karst erosion features at and below the water table and deeper fractures in the limestone. Selection of the springs for the present project was based on favourable sample conditions and to obtain a range of mean residence times.

While some springs deliver water to some extent contaminated (as also evident from the CFC levels), most are quite pristine. Most springs exhibit the tracer relations typical of the exponential flow model, while some show typical behaviour of mixing between two distinct younger and older water types. There may be an influence of ventilation of groundwater in the karstic features of the Dolomites. The presence of mine and other waste water pollution in some spring water is evident from both chemical anomalies and high CFC values. Mean residence times of spring water in the compartments are in the order of tens to hundreds of years.

Field study: Leeu Gamka

In this agricultural area along the Gamka river, water is extracted from the Beaufort sandstone underlying the alluvium upon which heavy irrigation takes place, mainly from sprinklers. In addition there is some water importation from the Leeu Gamka dam upstream. Groundwater is extensively recycled as can be observed from the higher water salinity and CFC downstream towards the south and the consistent high tritium values that are close to the present day rainfall levels.

The tracer patterning from the boreholes fits the exponential mixing model quite well. There appears to be slightly elevated CFC-12 levels and reduced CFC-11 levels. The water properties (chemistry and isotopes) can clearly be separated into sources of old 'inflow' water and mixtures with the younger recycled water.

Conclusions

The work of the present project has demonstrated that the piston-flow model is unlikely to be applicable to fractured rock aquifers. In all cases there is substantial mixing of water from various origins. Whether this mixing occurs in the fractures of the aquifers, or in the sampling boreholes is immaterial. The combined use of CFCs, isotopes and chemistry has shown that both the exponential mixing model and the binary mixing models are applicable for all three fractured rock aquifers, namely for the Dolomites, for a granite/granite gneiss aquifer and for a Karoo sandstone.

Significant CFC levels above the maximum atmospheric limit (100 pmf) indicates contamination of the groundwater. Different contamination events seem to generate characteristic CFC profiles. Combining CFCs with macro chemistry has proved very useful in detecting, or confirming, examples of low levels of pollution.

In general there seems to be a loss of CFC-11 in the groundwater compared to the other two CFCs. This can be due to the tendency of CFC-11 to be reduced in anoxic environments or be adsorbed on organic matter in the aquifer. CFC-12 and CFC-113 generally provide consistent patterns in groundwater.

A general algorithm whereby the flow regimes for all three aquifers can be evaluated from tracer data only, is not possible. Each site requires an approach that is suitable to the conceptional hydrogeology of the specific aquifer and its flow distribution pattern.

The overall aims of the project as formulated in the agreement between the WRC and the CSIR have been met.

Recommendations

1. The combined use of CFCs, isotopes and chemistry in groundwater enables one to conclude with some degree of certainty which mixing mode! is applicable to the study. Thus for a regional or local aquifer which has a high degree of importance and for which proper management is needed, we recommend that a program of investigation be carried out which includes the suite of sampling and data evaluation demonstrated in this project. Proper management of an aquifer implies that the flow system is understood, and this would include an understanding of mixing processes. The estimated cost for such an exercise is in the region of R100 000 which is acceptable in relation to the alternatives available and the potential applications of the water. We recommend that the combined use of CFCs, isotopes and chemistry be more commonly applied in Southern Africa.
2. The spreadsheet flow model, MRTMODElS, needs to be improved. One aspect is to incorporate the fact that recharge in the arid zone can vary from year to year and this should be incorporated as a separate data set.
3. CFCs can be used as an early warning system for detecting anthropogenic pollution of an aquifer. The present project has highlighted a few cases where CFCs are elevated, but other pollution indicators are barely above normal concentrations. For sensitive and strategically important water-supply sources it is recommended that regular sampling for CFCs are carried out in order to provide early warning of pollution.
4. Lack of knowledge of the regional distribution of tritium in modem and past rainfall has hampered the usefulness of tritium as a tracer for mixed water. An evaluation of the past distribution of tritium in rainfall is required.
5. In addition it is becoming necessary for more data to be made available of the present day distribution of tritium in rainfall. A sampling and analysis programme throughout the country is necessary, to supplement the data presently available.
6. An evaluation of the effective tritium content of recharge is required. This should be based on the rainfall data base and accommodate the selectivity of recharge events In our arid region
7. The tracers 3Hf3He and SF6 are less susceptible to sample contamination and have been used in some published investigations with great success. Particularly 3Hf3He is useful to assist in unravelling the complexities of the tritium input during recharge. It is recommended that these techniques be introduced in further studies of this nature.
8. More detailed numerical groundwater models than the lumped parameter models employed here, should be developed to incorporate these short-lived tracers. The first attempt by the ETH Group (Switzerland) in Botswana, seems promising and should be followed up with additional and appropriate case studies.