The sustainability of groundwater resources is important for the environment, the economy and communities where surface water is scarce. It is a hidden resource, but additional information can be extracted by combining groundwater measurements and lithological information with groundwater flow equations in groundwater models. The models convert data and knowledge about the groundwater systems into information, such as relative inflow and outflow rates and water-level predictions that can be readily understood by groundwater managers. ¶ The development of models to effectively inform groundwater management policies is, however, a complex task that presents a fundamental scientific challenge. This thesis presents methods and results for water balances calculated using groundwater flow models. Groundwater flow modelling methods and approaches are discussed, and their capabilities and limitations are reviewed. Two groundwater systems are studied for the Great Artesian Basin (GAB) and for the irrigation area near Bowen, Queensland. Three approaches to water balance modelling are applied in comprehensive model-development frameworks that take into account model objectives, data and knowledge availability and sensitivity analysis techniques. The three models show numerical methods of increasing complexity. The Bowen study area is well-suited to the least-complex method because data collection has been a priority there. As a contrast, the GAB is a large, poorly-monitored basin for which more knowledge of the groundwater system can be gained from its simulation by the steady state and transient groundwater flow models. The Bowen impact assessment model calculates dynamic historical water balances. The GAB aquifer models are high-complexity representations of the groundwater system that include predicted responses of the system to changes in hydrological conditions. These are comprehensive and well-documented attempts to model these systems. They provide a platform for scenario investigation and future improvements. ¶ Darcy’s Law was used in a GIS (Geographic Information System) to calculate dynamic water balances for an aquifer near the Queensland town of Bowen. This is the first time this approach has been applied to generate a complete groundwater balance. Over the period 1989-1997 the model estimates average total inflows to be 87 % groundwater recharge by rainfall and irrigation return flow, 12 % river recharge and 1 % inflow across the study area boundary. Outflows are estimated to be 66 % evapotranspiration, 28 % water bore discharge, 4 % discharge to the ocean and 2 % groundwater loss to rivers. Analyses show that evapotranspiration is the most uncertain parameter value. The GIS method was found to be useful for calculating water balances more accurately than analytical methods, because of their simplifying assumptions, and less time consuming than the more complex numerical models developed for the GAB aquifer. ¶ For the GAB, a steady state numerical model was developed and tested and predictive scenarios were run. The purpose of this modelling was both to gain a better understanding of the water balance of the GAB and to provide a tool that could predict water level recoveries under different bore rehabilitation scenarios. The model complexity is greater than in any previous numerical groundwater model of the GAB. In particular, the model uses more data, extends over a larger area and uses a generally finer discretisation than previous models. For the nearest surface artesian aquifer in 1960 the model estimates total inflows to be 60 % groundwater recharge and 40 % diffuse vertical inter-aquifer leakage. The model estimates outflows to be 53 % diffuse vertical leakage, 43 % water bore discharge, 3 % spring discharge and 1 % discharge to the ocean. Analyses show that the model is most sensitive to changes in horizontal hydraulic conductivity and recharge. The model-predicted heads match field measurements with a Scaled RMS error of 0.8 %, which is well within the guideline error of 5 %. The predictive scenarios show net vertical leakage into the aquifer decreasing and net vertical leakage out of the aquifer increasing, as bore flows are reduced. These estimates of inflows and outflows complement other studies of the Basin and add to our understanding of its hydrodynamics. In this way the water balance helps provide a sound basis for the development of GAB groundwater management plans and policies. Through its water level recovery predictions, the model has also been used to support the GAB Sustainability Initiative. ¶ A transient numerical model of the GAB was also developed and tested, and predictive scenarios were run. This model builds on the steady state model, and is more complex, with a calibration period (1965-1999) that is longer than in any previous GAB model. During calibration the model observations were expressed and weighted so that the minimisation of the objective function reflected the relative importance of the model’s potential uses, these being respectively: to simulate the impact of changing bore flows, to more generally inform water management plans and to provide an estimate of the water balance. It was found that the 1960 steady state assumption was not correct. Discluding anthropogenic discharge, the model is most sensitive to recharge and hydraulic conductivity. The model-predicted heads match field measurements with a Scaled RMS error of 2.7 %, which is well within the guideline error of 5 %, but the increased data requirements of the transient model highlighted deficiencies in the data available for the modelling. In particular, the uneven spread of the groundwater measurements over both time and space, the questionable accuracy of measurements from both high temperature and pressure bores, and corroded bores, and the type of discharge measured (for example, maximum yield or flow-as-found), became evident during the calibration of the model. Insights and the value of this work indicates for the first time that at the start of 2005 outflows were estimated to exceed inflows by 266 GL/year, or 62 % of total inflows, and, assuming that inflows through the aquifer’s boundary will not be reduced due to climate change, it will be possible to recover some of the lost groundwater pressure if all stock and domestic bores are rehabilitated and new extractions are limited. In this case the modelling estimates that inflows could exceed outflows by up to 40 % of total outflows.

16S rRNA gene analysis has shown that bacterial diversity in the GAB bores studied was limited to the genera Hydrogenobacter in the phylum Aquificae, Thermus in the phylum Deinococcus-Thermus, Desulfotomaculum in the phylum Firmicutes, the alpha-, beta- and gamma-classes of the phylum Proteobacteria and the phylum Nitrospirae. There was no clone closely related to members of the delta-proteobacteria and epsilon-proteobacteria classes detected. The number of bacterial strains directly isolated from the Fairlea and the Cooinda bores were far less than the numbers of distinctive phylotypes detected by the 16S rRNA gene characterisation. In addition none of the bacterial strains directly isolated from the water samples were represented in the 16S rRNA gene clone libraries. Similar discrepancies between the bacterial populations obtained from the 16S rRNA gene analysis and those obtained from direct isolation have been reported in the literature (Dunbar et al., 1999; Kampfer et al., 1996; Suzuki et al., 1997; Ward et al., 1998; Ward et al., 1997). However, in general, the phyla with which the isolates were affiliated were the same as those phyla to which the clones belonged. The environmental changes introduced (by bringing the artesian water up to the surface and exposing it to four types of metal coupons made of carbon steels identified by codes ASTM-A53B, ASTM-A53, AS-1074 and AS-1396 and commonly used in bore casings) led to changes in the bacterial community structures. In general, the species which proliferated in the communities before and after the changes were different. The diversity of the bacterial species in the community decreased following the environmental changes. Clones dominating the clone libraries constructed from newly established bacterial communities also differed from the clones dominating the libraries constructed from the bacterial communities which had existed naturally in the bores. These trends toward change in the bacterial communities were observed at both the Fairlea and the Cooinda bore sites. All four metal types incubated in the Fairlea bore water lost between 3.4 and 4.7% of their original weight. In contrast none of the metals incubated in Cooinda bore water lost weight. Clone library A1 showed that the natural population of the Fairlea bore was dominated by clone A1-3, which represented a novel species related to the isolate boom-7m-04. But after metal incubation (and recording of the metal weight loss), the bacterial community was dominated by clone PKA34B, which has a 95% similarity in its 16S rRNA gene sequence with Desulfotomaculum putei. Desulfotomaculum species are known to cause metal corrosion due to their byproduct H2S. But the low level of phylogenetic relatedness found does not provide enough information to speculate on whether the species represented by clone PKA34B is a member of the genus Desulfotomaculum or not. However, the fact that clone PKA34B dominated the PKA clone library by 50% makes the species it represents a suspected candidate likely to be involved with the metal weight loss at the Fairlea bore. In contrast, clone library 4381 showed that the natural population of the Cooinda bore was dominated by clone 4381-15 representing a species distantly related to a hydrogen oxidiser Hydrogenophaga flava (95% similarity). The dominating clone of the new community formed after metal incubation was clone COO25, which has 99% similarity with Thermus species that have not been reported to be involved with metal corrosion to my knowledge. In this project detection, identification and comparative quantification by 16S rRNA gene-targeted PCR probing with probes 23B and 34B were successfully developed for a Leptothrix-like species and for a Desulfotomaculum-like species represented by clones PKA23B and PKA34B respectively. This method of probing permits a fast, sensitive and reproducible detection, identification and at least a comparative quantification of the bacteria in the environment without the need for culturing. Therefore it is extremely suitable for use in bacterial population monitoring. PCR probing with the 34B probe has a potential commercial use as a means of screening for bores with a potential high risk of corrosion due to this Desulfotomaculum-like species. Direct isolation of bacteria from the GAB water has resulted in the isolation of seven strains from the Fairlea bore and eight from the Cooinda bore. Among these isolates, three novel strains were studied in detail. Reports on the characterisation of strain FaiI4T (T=Type strain) from the Fairlea bore (Kanso & Patel, 2003) and strain CooI3BT from the Cooinda bore have been published (Kanso et al., 2002). The data generated during this project add to our current information and extend our knowledge about the bacterial communities of the GAB's sub-surface environment. This information will provide a basis for further ecological studies of the GAB. Studies on involvement of certain groups of bacteria with the corrosion of metals used in bore casings could provide a foundation for further studies to develop maintenance and managing strategies for the GAB bores.