In the assessment of information requirements two different levels will be considered, namely

• to set up the Catchment Management Strategy (CMS), and
• to assess individual license applications.

The central focus of any CMS developed by a CMA is the water allocation plan. In order to develop a CMS, and consequently the water allocation plan, the CMA must first determine the allocatable quantity of water available in the WMA. The allocatable quantity of water could perhaps be defined as the amount of water available for use within a particular catchment or area at a certain level of risk. It is also necessary within the allocation plan to take into account the future water demand and constraints that may limit development. Therefore, when formulating the allocation plan, water managers need information on

• allocatable quantity of water;
• projected demands; and
• developmental constraints as shown in Figure 4.

• Allocatable quantity of water

To determine the allocatable quantity of water within a particular area, the CMA must first take into account criteria set out in the NWRS. Hence, before any water can be allocated to specific users the CMA must set aside specific quantities of water to fulfil the reserve quantity and quality requirements, international obligations, water users of strategic importance and future water needs and inter basin transfers set out in the NWRS. Once these have been taken into account, it is necessary to determine the actual level of water use in the catchment by different water users. In this regard an initiative is currently underway whereby all water users must register their water use.

For any CMA to institute a practical allocation plan it is necessary for the paper water to correspond to real water. This means that not only is accurate information on current water use required but that the modelling of such systems needs to reflect the reality in terms of both the temporal and spatial scale that water use is operated at and decisions are made.

The reserve is perhaps the most fundamental aspect of allocatable water determination. The reserve determination has implications for both the sustainability criteria and equity criteria in the NWA (1998) and, hence, deserves special mention in this analysis.

The reserve is defined as the quantity and quality of water required for basic human needs, as well as the quantity and quality of water required to sustain the aquatic ecosystem. The basic human needs reserve provides for the essential needs of individuals served by the water resource and includes water for drinking, for food preparation and for personal hygiene (NWA, 1998). The basic human needs reserve is essentially a function of the population and could be determined on an annual basis in the course estimate required for the implementation of the NWRS.

While the basic human need remains relatively constant, the environmental reserve fluctuates on a daily basis in some cases on a sub daily basis. In order to implement the reserve at a catchment and sub-catchment level the CMA will need water quantity and quality information on at least a daily basis from both its modelling and monitoring systems.

Allocatable quantity of water is often determined by means of systems yield, which is the amount of water that a catchment system is able to supply at a certain risk of failure. In terms of water availability in modelling the system the CMA may need more detailed information than just the systems yield as it may be critical to determine water availability at different points in the catchment. Hence, information is required for both individual and potential users as well as the system as a whole.

• Projected Water Demands

In order to develop a CMS and its associated allocation plan the CMA needs information on projected water demands in different sectors of the economy. Water demands are influenced by population and development trends, which are in turn influenced by socio-economic factors acting within and on the WMA. The demand projections are also influenced by the institutional arrangements in the WMA.

In assessing potential projected water demands it is important for the CMA to take into account outside institutional developments and it is hence necessary for the CMA to foster co-operative governance relationships with other government departments and parastatal organisations. Consultation with stakeholders is also essential in determining developmental plans.

Demographic information is needed to assess population trends. Population trends affect both the basic human needs reserve as well as developmental water demands within a WMA.

Socio-economic factors influence industrial, commercial and agricultural development in a WMA. These factors, combined with population dynamics, influence the quantity and distribution of water demand. Resource potential is on of the major overriding factors that will limit population growth and development in South Africa. Hence, information on the resource potential is critical in determining the water demand in a particular WMA.

• Developmental constraints

Developmental constraints are aspects that will restrict development in certain areas. These consist of environmental constraints that include conservation areas such as game parks as well as areas of high biodiversity. In such cases the loss of such areas could cause irreversible damage within the catchment area and significantly impact the sustainability.

The resource potential is another major constraint, where in the case of agriculture and forestry the biophysical potential of the area has a major impact on the types of crops that can be grown in particular areas. Resource potential is also limiting in other primary activities such as mining where the distribution of mineral deposits is a major limiting factor.

Other constraints that may inhibit development are infra-structural constraints such as access to efficient transport routes. Distance can also be a major constraint where distance from markets or mills can significantly increase transport costs and impacting on the economic viability of specific enterprises.

The equity, efficiency and sustainability criteria set out in the NWA (1998) can place major constraints on development. In terms of equity certain developments may only take place once the discrimination of the past has been redressed. The efficient, beneficial use of water is another factor that needs to be considered in undertaking a project or proposed development and can have a major impact on development if a water use is considered inefficient.

In terms of individual licenses the CMA needs to assess the following main criteria before it can issue a license:

• Is the application in compliance with the CMS and the NWRS?
• Does it fulfil the equity, efficiency and sustainability criteria set out in the NWA (1998)?
• Is the impact on other users, the water resource and the environment acceptable?

Each of these criteria is dealt with in more detail in the following sections.

• Is the application in compliance with the CMS and the NWRS?

The CMA water manager reviewing the various license applications needs to assess if the water license in question is in line with the both the NWRS and the CMS. Before issuing water use licenses the CMS and the NWRS should have been developed and the necessary information will be passed down to water resource managers assessing the licenses.

The CMS, which must be in compliance with the NWRS, should allow the manager to assess the license in terms of the following criteria set out in Section 27 of the NWA (1998), namely

• the class and resource quality objectives of the water resource;
• the strategic importance of the water use in question; and
• the water quality objectives and international obligations.

The information required to assess the licenses in terms of these criteria is therefore the same as those needed to develop the CMS, which has been discussed in Section 7.1 of this document.

• Does it fulfil the equity, efficiency and sustainability criteria set out in the NWA (1998)?

In addressing this question two aspects in Section 27 of the NWA (1998) are addressed which are

• the need to redress the results of past racial and gender discrimination; and
• the efficient and beneficial use of water in the public interest.

To assess the equity criteria, in terms of redressing the results of past racial and gender discrimination, information on the existing status quo in terms of catchment water use is needed. This aspect is already being addressed through the water registration process where all water users are being registered with the WARMS system. This aspect should also be provided for in the CMS to some extent, as certain developmental plans will already incorporate the redressing of past racial discrimination. The information requirements for this type of decision consist of socio-economic information as well as hydrological information in terms of actual and projected water use.

The efficient, beneficial use of water in the public interest incorporates both the efficiency and sustainability criteria. In order to determine efficient water use both hydrological and economic information are required. Efficiencies can then be determined in terms of the economic return per unit of water used. Feed forwards and feedbacks in the economy need to be taken into account when assessing the economic returns and also lead to a better assessment of the beneficial use of water.

Beneficial use of water has not only an economic implication, but social and environmental implications too. The water use needs to be sustainable so as not to reduce the water resource carrying capacity and reduce the resource potential thereby having a negative impact in the long term on society. Social impacts of water use also need to be considered in terms of jobs created and benefit to communities as a whole. Social, economic and environmental information are required to assess the efficient and beneficial use of water.

• Is the impact on other users, the water resource and the environment acceptable?

The impact information required by the water manager in a CMA is outlined in two different aspects of Section 27 of NWA, namely

• the socio-economic impact of
• water use or uses in authorised; and
• the failure to authorise the water use or users; and
• the likely effect of the water user to be authorised on the water resource and on other water users.

Modelling is thus required to provide social, economic, hydrological and environmental information to estimate the impact of a particular development might have. The modelling could consist of a suite of models that can take into consideration both hydrological impacts in terms of both quantity and quality, as well as social, economic and environmental impacts. The scale of information required would need to reflect the operations and decision framework of the development being considered and will need to be at least on a daily scale in terms of hydrology to get a accurate assessment of the water quantity and quality impacts. The next section addresses the modelling requirements for a DSS to support the CMS of a CMA.

Our philosophy in tackling this document is not how the current set of modelling tools and methodologies can service the needs of the NWA (1998), but rather a more rounded perspective of what tools and methodologies need to be put into place in order to service the water resource use and management community in implementing the NWA.

Models are useful in generating information about the water resource systems so that options can be considered and decisions taken to manage the resource and resolve conflict. In the NWA (1998) there is a strong emphasis on stakeholder participation and indeed a CMA is comprised of members representing different stakeholder groups. The process of developing allocation plans and allocating water to different individuals has become not only a scientific but also a social process. This aspect places new demands on models and tools used in the decision making process. Information generated from models now needs to gain the trust of the stakeholder community in order to be accepted in decision making.

Information generated from the modelling process will need to be credible, trusted, and promote shared understanding (Dent, 1999). "This in turn could promote transparency in the modelling and decision making process leading to the acceptance of decision and a shared understanding of their consequences" (Dent, 1999).

The processes, which yield this type of information, need to be replicable and consistent, offering regular, affordable and meaningful communication among stakeholders and their representatives. The processes should be open and transparent enabling implicit assumptions and mental models to be made more explicit. Processes and models used need to overcome the barriers to communication between stakeholders, which arise from geographic, disciplinary and institutional / organisational separation. While these are the generic requirements that all modelling tools need to fulfil, the next section will concentrate on more specific modelling requirements of a DSS required to fulfil the objectives of the SEA in helping CMAs develop and implement a CMS.

To set up an allocation plan several different levels of information are required. While the majority of the information required is in the form of raw data, such as landuse, water use, demographic information and biodiversity data (as described in Section 7.5), some information needs to be generated with the use of models, particularly in the case where planning projections are needed to expand beyond the catchments or WMAs current status.

While the discussion in this document is on a complete system concentrating on incorporating SEA principals and practice into CMAs the focus of this research is on the hydrological and economic components and while modelling requirements in terms of other aspects such as social and environmental will be mentioned the main focus will concentrate on hydrological and economic modelling.

The initial focus on developing a CMS and associated allocation plans is to determine the allocatable quantity of water. In order to perform this function it is necessary to determine

• the reserve estimate in terms of quantity and quality; and
• current water use status in the catchment in the catchment.

A complete allocation plan, however, must include projections on water demands into the future and prediction on how these demands are influenced by physical, environmental and social constraints. In the following sections a description of the modelling requirements needed to determine the allocatable quantity of water will be presented followed by those requirements needed to provide for future water demand projections.

The modelling requirements in terms of estimating the reserve are rather rigorous as the detail of the information required is high. While the information passed down from the NWRS in terms of reserve classification need not be rigorously defined, the information needed to implement the reserve needs to be far more detailed. For the purpose of implementing and determining the reserve water quality and quantity information is required on at least a daily basis and at a sub Quaternary Catchment scale. It is also necessary to estimate both the water quantity and quality from present conditions as well as those that would have occurred under relatively pristine conditions (Acocks? Veld Types or naturalised flows).

Models are also required to perform in a data poor environment where lack of streamflow data at gauging weirs is a major constraint in terms of model verification. This means that calibration modelling is not a viable alternative due to lack of gauging stations, while physically based process rainfall runoff modelling is able to provide both more reliable inputs at ungauged sites and for pristine conditions. Modelling needs to be able to incorporate operational hydrology to simulate the operating conditions in the catchment and will help in developing a strategy to implement and maintain the reserve.

The hydrological modelling requirements to assess the reserve can be summarised as follows:

7.3.4 Current status of water use in the catchment

The NWRS sets aside certain quantities of useable water for specific needs such international obligations, strategic water user, future water needs and interbasin transfers. The development of the NWRS requires broad level information, which is currently being provided by the WSAM. This model provides estimates on a yearly basis, which for the purposes of a CMA are too broad and need to be used as guidelines in developing and implementing the CMS.

Before allocation plans can begin it is necessary to model the catchment at current level of development and determine from this the actual quantity of water currently being used in the WMA. The current water use and the information provided from the NWRS will allow the CMA to determine the actual quantity of allocatable water left in the catchment. If, as has occurred in some catchments in South Africa, the allocatable quantity of water is less than the actual water use, the catchment is considered stressed and water supply needs to be augmented or the water demand need to be adjusted in order to bring it in line with the supply.

In modelling the actual quantity of water being used in the catchment cognisance needs to be taken of land and water use activities that impact the water resource. The model / models must also be able to account for the operational framework within the WMA, which incorporates all the water supply and demand systems within the catchment and how they are managed. In the past the estimates of actual quantity of water used were established using calibration models and statistical relationships derived from experimental results. The systems were modelled on monthly basis using system optimisation models, which could generate large sequences of stochastic results.

While these modelling techniques give a reasonable estimate of water use in a catchment or WMA, with the advent of the NWA (1998) and the requirements of the reserve to be operated at a daily scale as well as the more rigorous requirements on users to use water beneficially and efficiently, the information and associated modelling requirements have become more detailed.

Physical process based modelling systems are able to simulate current and past landuse, able to produce naturalised and denaturalised flows directly. In estimating the operational hydrology it is also necessary to simulate water use and management at the level of operational decisions, which is often less than a monthly time step where decisions are made on a daily to weekly basis. Physical process based modelling also has the advantage of representing the system more explicitly while the assumptions and results from empirical modelling are not explicit leading to lack of transparency and decreased bargaining power.

The modelling requirements to determine current land and water use can be summarised as follows:

7.3.5 Modelling projected water demands and developmental constraints

As has already been mentioned, the modelling of projected water demands requires demographic information, outside institutional plans, and developmental projections, which are inhibited by resource potential or have a direct impact on the resource potential. Developmental constraints include environmental constraints such as biodiversity value and conservation areas, biophysical constraints such as carrying capacity and production potential and social constraints such as equity considerations.

Development projections and demographic information can be obtained with simple population growth rate projection models which are superimposed over demographic trends in terms of rural urban migration patterns. Certain indicators could be used in collating population variables with water demands and industrial developments. In assessing outside institutional plans it is possible to identify specific areas of development and model the hydrological consequences. Hence, information on projected water demands can be determined from actual known consequences such as planned developments as well as projections connected with population growth and associated developments. All these factors are, however, limited by certain constraints and the resource potential.

Resource potential is essentially the potential that a natural resource has in supporting a specific type of activity. Resource potential in terms of agriculture is essentially a combination of different factors such as climate, altitude, slope and soils. Simple crop modelling could be combined with geographical information systems (GIS) to give estimates of biophysical landuse potentials. Yields from certain potential crops and livestock carrying capacity could also be modelled, which when combined with economic information will provide estimates of economic returns. Resource potential in terms of mining is a directly dependant on mineral deposit distributions, while certain industries may be constrained by lack of raw material in the near vicinity. These can all be mapped and determined combining simple modelling with GIS capabilities.

Constraints on the system, such as conservation areas and local zoning in terms of town planning, can once again be mapped using a GIS. Constraints, which may influence the economic viability of certain operations, such as distance from markets, and infrastructure can again be estimated using simplified algorithms combined with GIS data. Biodiversity values can also be calculated using a combination of indicators with different algorithms and GIS to map out areas of high and low biodiversity potential.

The hydrological impacts associated with the different water demand projections and landuse changes can be modelled using physical conceptual process based modelling systems where the system is not represented as empirical relationships but is represented explicitly as a set of physically based processes. The physically based hydrological modelling approach allows more flexibility in terms of planned scenarios as it can account for the unique and novel situations that arise. It is also necessary to incorporate operational systems modelling into the planning framework as it allows the modeller to assess the potential impacts of new and proposed schemes. Using the above approach future scenarios can be identified and built up and the interrelated hydrological, economic, social and environmental consequences can be identified.

In summary the modelling requirements associated with determining water demand projections are

7.3.6 Modelling required for the assessment of individual license applications

In terms of assessing any license application it is necessary to account for the factors outlined in Section 27 of the NWA (1998). In addition, the application is required to be in line with the CMS and the NWRS and, therefore, the modelling requirements need to be at least as rigorous as those required to develop the CMS and NWRS. In terms of individual license application it is necessary to estimate the hydrological, economic, social and environmental consequences of issuing or not issuing a specific license. It is only once all these factors have been taken into account that the license can be issued.

Impacts on other users can also be estimated by combining physically processed based models with operational modelling. The socio-economic impacts are, however, a little more difficult to estimate and require resource potential estimates and viability constraints in terms of transport and environmental, to be combined with hydrological modelling. The hydrological modelling in terms of the operational hydrology which includes all the impacts of structures and associated operating rules, needs to provide not only the impact of different land and water use options on the system as a whole but also needs to represent the impact on other individual users within the system.

The modelling requirements in terms of hydrological modelling for the assessing individual license application could be summarised as follows:

Modelling requirements in terms of social and economic modelling could be summarised as follows:

In the sections to follow several different models (both local and international) are analysed in terms of their ability to satisfy the above criteria. A brief discussion of the present approach used in water resources management in South Africa and some of the perceived shortcomings in this approach is initially provided. This is followed by a description of different hydrological water quantity models and the identification of their suitability in terms of the factors mentioned in the initial parts of the Section 7.3 and perceived shortcomings in each case. A summary table describing model suitability in terms of certain criteria specifically mentioned for this project will conclude the modelling section.

The discussion will begin with a review the current modelling methodologies used in water resource assessment.

In the past in South Africa a "horses for courses" approach has been adopted in water resource management and practice. The belief being that specific water management and implementation tasks require information at specific levels of complexity, such tasks should then be supported by methods and models of a level of complexity and with data requirements that are at the appropriate level of the information required (Gorgens, 1999).

This philosophy has led to the use a multiple model approach where each specific model is designed to accomplish a certain task. The result is that when performing a water situation assessment a large array of different models are used with each performing a specific task. These models are then fed into each other using a series linking approach. A series linking approach takes the output from one model once it has completed running and feeds it into another model to execute some other function. The modelling methodology described in stepwise sequence below is a generalisation of the methodology used by many water situation assessment studies and is similar to the methodology currently being adopted on the Breede River. The methodology usually incorporates the following sequence with a set of different models being used.

Calibration and statistical methods such as those described above are also becoming less attractive as they tend to be black box approaches, which do not make the assumptions explicit. The result is that, in general, such modelling efforts are less likely to have credibility in the stakeholder community who may not be able to understand the results produced. Whereas physically based process models have more complicated algorithms and are generally more time consuming to set up, the inputs and outputs are generally easier to understand as they represent real world quantities. Calibration models and statistical methods are, in general, situation specific and the results are non transferable to other areas or novel situations. This means that the testing of different scenarios and extending deriving estimates at ungauged sites can result in large inaccuracies with the use of calibration and statistical methods. In the next section of this chapter a more detailed perspective of different local and international models is provided.

This review of models is not intended to be exhaustive but concentrates on the main hydrological water quantity models currently available and in use in South Africa.

The WRYM model is capable of simulating a wide range of operating policies governing the allocation of water in a multi-purpose, multi-reservoir system. The WRYM model allows the user to define operating policies governing the allocation of water by altering the penalty structure associated with channels and reservoirs in the data sets used to run the model. The WRYM is based on the assumption that a flow network can represent a water resource system. The network can be analysed for each time period and solved using an efficient network solver (subset of linear programming techniques) with the careful selection of penalty structures. The network solver will solve a particular network problem using a minimum cost approach, where appropriate costs (penalty structures) are allocated to channels and reservoirs in such a manner as to define the relative "value" of water in each storage zone. The penalty structure is also selected to dictate the most attractive route (i.e. minimum penalty) for transferring the water from the storage zones to the demand centres.

MIKE BASIN is a network model, which provides a mathematical representation of the river basin encompassing the configuration of the main rivers and their tributaries, the hydrology of the basin in space and time as well as existing and potential major schemes and their various demands of water. Rivers and their main tributaries are represented by a network of branches and nodes; the branches represent individual stream sections while the nodes represent confluences, diversions, locations where certain water activities may occur, or important locations where model results are required (MIKE-BASIN, 2001). The model operates on the basis of a digitised river network generated directly on the computer screen in ArcView GIS.

Time series data of catchment constitute the basic input to the model, while additional input files define reservoir characteristics and operation rules of each reservoir, meteorological time series, and data pertinent to each water supply or irrigation scheme such as diversion requirements and return flows.

Model output comprises information on the performance of each individual reservoir and associated hydropower units, as well as other schemes with water demands, such as irrigation. For the entire simulation period, the magnitude can thus be illustrated. Moreover, time series of river flow at all nodes are simulated enabling the user to determine the combined impact of selected schemes on river flows. All results can be visualized in various formats, including animation.

The ACRU modelling system (Schulze, 1995), established at the University of Natal, is a deterministically based, physical-conceptual model revolving around a daily time step water budget. Internal state variables (e.g. soil moisture), model components (e.g. interception) as well as end-product model output (e.g. streamflow or sediment yield) have been widely verified under different hydrological regimes in Africa, Europe and the Americas.

Daily climate data is used in ACRU; however, more cyclic, less sensitive variables such as temperature or reference potential evaporation can be entered on a monthly level if daily values are not available. Fourier Analysis transforms these monthly inputs to daily inputs. The soil water and runoff regimes in the model are structured to be highly sensitive to land cover / use change. A database of inputs to ACRU at a Quaternary Catchment scale is already in place, which can be used to run the model. This structure that has been set up enables experienced users to run the model relatively quickly by changing input information in the database.

The VTI (Variable Time Interval Model) (Hughes and Sami, 1994, as cited in Jewitt and Görgens, 2000) is an extended version of the single flood event model (OSE2) (Hughes, 1984; Hughes, 1989, as cited in Jewitt and Görgens, 2000), which has been adapted into a continuous time series model.

The model uses a variety of parameters. Some of these parameters have physical meaning, others are estimated from physical indices and some are completely empirical. The complexity of the model algorithms has been limited to prevent the models information requirements becoming too limiting. This model allows for simulation at a variable time interval according to user selected rainfall thresholds. The model is operated from within the HYMAS software package, which offers several support facilities for results analysis.

VTI uses a semi-distributed approach as opposed to a grid square or slope element. The model components are thus based on simulating the integrated response over sub-area, rather than at a representative point. As a result many of the functions are of a distribution type, where probability distributions represent the internal sub-area variation in hydrological processes. Two basic runoff generation methods are available; the one is dependent on rainfall intensity and the infiltration characteristics of the catchment soils, while the other depends on the dynamic moisture status of the catchment or sub-areas. A third runoff generation function allows for the simulation of groundwater baseflow.

The Precipitation-Runoff Modelling System (PRMS) is a modular-design, deterministic, distributed-parameter modelling system developed to evaluate the impacts of various combinations of precipitation, climate and land use on streamflow, sediment yields and general basin hydrology. Basin response to normal and extreme rainfall and snowmelt can be simulated to evaluate changes in water balance relationships, flow regimes, flood peaks and volumes, soil-water relationships, sediment yields and ground water recharge. There is also a provision for parameter optimisation and sensitivity analysis within selected model parameters where both individual and joint effects on model output are evaluated. PRMS offers a flexible framework for continued model-system enhancement and hydrologic modelling research and development (USGS, 2001)

This is a useful physical process based hydrological model. There is, however, extremely little local expertise in the running of the model. The parameters are of a USA standard and would need to be transformed and tested in South Africa.

The Soil and Water Assessment Tool (SWAT) is river basin, or watershed, scale model developed for application in large, complex rural basins. SWAT is a continuous daily time-step model, which simulates the impacts of alternative land use management practices on surface and ground water, sediment and agricultural chemical yields in ungauged watersheds, over long periods of time. The model is capable of simulating hydrology, pesticide transport by runoff, percolate, and soil evaporation. In addition, SWAT allows for the simulation of nutrient cycling, erosion and sediment transport. Large river basins are subdivided into homogenous parts and each part is then analysed individually as well as its interaction with the whole.

The SWAT interface extracts spatially distributed parameters of elevation, land use, soil types and ground water table. The interface creates a number of input files for the basin and subbasins, including the subbasin routing structure file. Input consists of files, information from databases and information from a GIS interface. More specific information can be entered singly, for each area of the watershed as a whole. The SWAT-GIS linkage incorporates advanced visualization tools capable of statistical analysis of output data (Soil and Water Assessment Tool, 2001).

SWAT requires specific information about weather, soil properties and topography, vegetation and land management practices occurring in the watershed. The physical processes associated with water movement, sediment movement, crop growth, nutrient cycling and the like are directly modelled by SWAT using this input data. The benefits of this approach are that the relative impact of alternative impact data on water quality or other areas of interest can be quantified even in watersheds with no monitoring data (Arnold et al. 1993)

SWAT uses readily available inputs, is computationally efficient and enables users to study long-term impacts. A number of output files are generated by SWAT. The type of data stored in the file can group these files. Other than the standard output file, the files produced during a model run are formatted as spreadsheets to facilitate importation of the data into spreadsheet software.

The SWAT model is ideal for assessing land and water use impacts. It is a processed based model and requires easily understood input. The model was developed in the USA and there is some question as to its applicability to South African conditions. The model has been used by the CSIR to perform landuse impacts in South Africa. The model does not include operational hydrology and would need to be linked to a network simulation model to obtain yield and water availability estimates. It could be linked with models such as HSPF and MIKE BASINS to perform these components.

SAPWAT is a planning and management tool, which relies heavily on an extensive South African climate and crop database (SA Waterbulletin, 2000). One of the primary objectives of the SAPWAT development programme was to provide for the specific circumstances and requirements of emerging irrigation farmers and community gardens. The impact of irrigation practices and strategies on water budgets was an additional force driving the development of this model.

This model is general in its applicability, while it is able to simulate a variety of different management practices and irrigation methods. In addition, the effect of soil water management options such as deficit irrigation can be evaluated. SAPWAT also facilitates consultation and interaction with farmers and advisors.

SAPWAT is easy to apply from an operational point of view and demands minimum computer expertise (SA Waterbulletin, 2000). Moreover, this tool can be used in a variety of applications. SAPWAT principles have been recognised by the DWAF and incorporated in the irrigation inputs into the national water balance model. This model has also been indicated by the DWAF as the method for determining the annual irrigation requirement. "SAPWAT in the absence of general metering, enables all water use for irrigation to be quantified equally thereby ensuring a cost recovery in a fair and systematic manner" (SA Waterbulletin, 2000). The strength of SAPWAT lies in an extensive database that saves the user an enormous amount of time and effort. Furthermore, this model is designed to accommodate updated historic weather data to the present should the user require it. Although SAPWAT is not a real-time scheduling model it can be a valuable complement to instrumented soil water content methods.

The SAPWAT model is used to determine irrigated areas required demands and hence specify the allocations required by different users. It is not an overall catchment model and the model cannot estimate the impacts of a water use on other users within the catchment. If used with individual licensing applications it should be used in combination with other models that are able to determine impacts on specific users and the catchment as a whole.

In reviewing the different water quantity models available (Table 1) it was concluded that the new ACRU model, which is currently under development, fulfils the majority of the criteria required by the DSS. These include that is a physically based, daily time step model. Because the model is physically based it can test scenarios and be transferred to ungauged situations will reasonable confidence. The model is particularly suitable for land use impact studies. The daily resolution allows for IFRs which are assessed on a daily basis. The model needs to reflect the management level within the system where water users and water supply operators make decisions on daily to weekly time steps. The daily time step allows for the assessment of water quality issues, which can fluctuate on a daily and even sub-daily time step.

The new operational hydrological components currently being developed in the model allow for the assessment of different water use and water supply impacts. The model should be able to test the water availability yielded from the system as a whole as well as the water availability for individual users. The model has been developed under South African conditions and a pre-populated database exists for use with the Quaternary Catchments.

Table 1 Summary table of the attributes of models used by DWAF

UD = Under development

• This question is more concerned with the idea that using ACRU might introduce trends in the runoff that will produce skewed results in the stochastic sequences. The answer to this is simply that trends will not be introduced as the development levels will be held constant throughout the ACRU simulation period. The ACRU outputs can be linked into the stochastic streamflow simulation component of the WRYM to produce stochastic streamflow sequences to extend the streamflow records if the link to the WRYM is required.

• The ACRU model itself will not have a stochastic flow generator developed owing to the complexity in generating daily streamflow simulations. The daily stochastic generation process is complicated by the need to account for daily cross correlations between the different catchments that are being modelled. If a technique is developed that can account for the cross correlations on a daily basis then the stochastic generation of daily sequences is a possibility that could be explored.

4. Water Quality Considerations

In Section 137 of the NWA (1998) it is specified that monitoring systems must be established to assess, among other matters

• "the quantity of water in the various water resources;
• the quality of water resources;
• the use of water resources
• the rehabilitation of water resources;
• compliance with resource quality objectives;
• the health of aquatic ecosystems, and
• atmospheric conditions which may influence water resources."

Management of water quality issues occurs at a local scale and therefore water quality management plans, decisions and actions are generally at a sub-catchment scale. However, the decisions made have implications for downstream users (and upstream impactors). Therefore, alignment between sub-catchments within a WMA (and possibly between WMAs) is required.

• Significance of changes in water quality

Possible water quality problems include

• the presence of toxic substances such as aluminium, arsenic and mercury;
• changes in system variables such as salinity, temperature and dissolved oxygen;
• non-toxic inorganic substances such as total dissolved solids and total suspended solids; and
• the presence of nutrients such as nitrates, ammonium and inorganic phosphates

In terms of domestic water supply there are various short and long term health impacts of deteriorating water quality. Aesthetic impacts include changes in odour, taste and colour. Economic impacts include increased treatment costs, scaling, corrosion or deposition of sediments in distribution systems.

In industrial plants water quality concerns include damage to equipment, problems in the manufacturing process as well as impairment of product quality. Agricultural industries could experience decreased crop yield, impaired crop quality, impaired soil suitability and / or damage to irrigation equipment owing to decreased water quality. Livestock farmers also rely on a certain water quality for efficient operations.

The Constitution of South Africa regards a healthy environment as a basic human right. In order to provide a healthy aquatic environment water quality managers and politicians needs water quality data and information to base water management actions and decisions on. To secure the availability of such data and information the National Water Act requires that the Minister, as soon as practicable to do so, establish national water quality monitoring systems.

To address these aspects of water quality information is required on

• Natural attributes of the catchment which include

• soils;
• vegetation;
• geology; and
• sediment production potential.

• River information such as

• main stem rivers and tributaries ? volumes of water and locations;
• wetlands and estuaries;
• catchment boundaries; and
• instream flow requirements.

• Monitoring stations which could be

• water sampling sites at both rivers and effluent sites;
• flow gauging ; and / or
• rainfall stations.

• Infrastrucutre which includes

• storage dam dimensions;
• irrigation components;
• water transfer schemes; and
• water and wastewater treatment works.

• Demographic distribution in catchment
• Land use such as

• human settlements;
• commercial and industrial areas;
• irrigation activities;
• mining details;
• solid waste sites;
• nature reserves; and
• indigenous veld and forests.

Information on both point and non-point sources of pollution in a catchment is required. Point source pollution refers single source pollution, for example, from water works, industrial plants, sewage treatment sites or pipe outflows. Non-point source pollution refers to pollution that emanates from land use types, areas and activities that result in conveyance of pollutants is a manner other than through a discrete source. Non-point source pollution can result in diffuse and intermittent pollution over a widespread area. Alternatively it may be concentrated and associated with localized high activity areas such as mines, feedlots and landfills.

• Sources of information

For water quality assessment in a catchment, information is required both on the desired water quality levels as well as the present condition of the quality of water in a catchment.

Water quality information in a catchment is available from a number of sources. The Department of Water and Sanitation (DWAF) operates the most comprehensive water quality monitoring programme in South Africa. DWAF is, in terms of the NWA, responsible for the coordination, organization, control and further development of national water resource quality monitoring programmes. Monitoring, recording, assessing and disseminating information on water resources is critically important for achieving the objectives of the NWA (Section 137).

A number of national water resource quality monitoring programmes are already in place. The national chemical water quality programme has been operational for many years. A national biomonitoring programme is currently being implemented.

Much of the responsibility for reporting on surface water quality in South Africa rests with the Institute for Water Quality Studies (IWQS) of DWAF. The water quality samples analysed at IWQS are obtained from about 1750 surface monitoring stations around South Africa, at gauging weirs (1200), lakes (450), springs (50), water purification works (35) and canals (15).

Many stations are monitored at two-weekly intervals, but some are only monitored intermittently and others are intensively monitored for short periods during specific investigations. The main chemical analysis is for major inorganic ions, while a few samples are tested for trace metals and organic compounds. Results are stored locally on a Laboratory Information Management System (LIMS) and at DWAF head office on part of its mainframe Hydrological Information System (HIS). About 40 000 borehole sample analyses are stored in a separate database.

The framework for monitoring and assessment of water resources at DWAF includes:

• Maintaining and co-ordinating national and regional data and monitoring networks. Key activities include:

• National Water Quality monitoring programmes
• Maintain and assess current trophic status monitoring
• Radioactivity monitoring
• National biomonitoring programme

• Assessing monitoring and data collection services
• Maintaining and updating water resource and related databases, digital cartographic data and geospatial data systems
• Distributing, publishing and providing water resource data and information.

Other potential sources of water quality information in South Africa include water service providers such as water boards e.g. Umgeni Water, local authorities, metropolitan councils as well as research institutions, however, the key sources of water quality data and information need to be established for the catchment under investigation (DWAF, 2000; DWAF, 2001b; DWAF, 2001c) .

Information on the soils and general rainfall intensity of a catchment would allow modelling of sediment loads for example. Information on agricultural, industrial and mining water use is required to establish the amount of phosphorus and nitrate loading to model nutrient contents, for example, which have regional impacts.

Assessment techniques range from complex water quality process models for non-conservative variables, to simple mass balance models for conservative substances. Simple heuristic models based on previous studies may also prove to be valuable tools. In general, these assessment techniques should not focus on processes that lead to pollution, but rather on the impacts of pollution loads on water quality.

Configured and calibrated water quality predictive tools/ models can serve to

• indicate whether point or non-point source pollution is dominant, or which sub-catchments in a basin are dominant water quality load contributors, etc; this would help to prioritise certain types of management actions
• estimate water quality constituent loadings from a range of land uses and water uses that result in non-point source pollution, and indicate which non-point sources are dominant
• indicate the likely effects of pollution load increases or decreases on downstream water quality, or receiving waters
• simulate water quality constituents at key points in river-reservoir systems in response to particular system operating rules
• simulate water quality variables at points of concern for different future scenarios of land use and water use
• support prioritisation and appropriate selection of competing management options
• extend, infill or simulate time series of water quality variables at points of concern.

The temporal and spatial scales of water quality modelling need to be addressed. Ideally water quality status should be modelled on a daily basis owing to daily changes in water temperature and therefore water quality. In terms of spatial scales, modelling is best carried out at a catchment or sub-catchment scale. Sub-catchment scale modelling is required where instream flow requirements (IFR) need to be met and therefore modelling streamflows and water quality up until these IFR points is needed.

Catchment wide point and non-point source assessment techniques are also important in the formulation of source management objectives, as they indicate which sources are contributing to current pollution loads. These models may be quantitative or qualitative, and contribute to assessing the feasibility of pollution load reductions.

Streamflow modification associated with land use change is also a type of non-point source pollution, which can adversely affect the physical and biological integrity of surface waters. The management of non-point sources is complicated by the dispersed and variable nature of the impacts, being primarily driven by rainfall events. The potential lag between polluting activity and effect also complicates non-point source management. These characteristics obscure the impacts from different sources and restrict the opportunities for their measurement. Therefore, non-point source contributions can generally not be monitored directly, but have to be inferred by experience-based interpretation, mass balances against measured point source loadings, or simulation modelling. However, the importance of non-point source management in South Africa is increasing as point sources are better controlled and catchments are developed.

Regulations exist to control the quality of discharge from point source pollution outlets. However, the cumulative effect of pollution from point sources has resulted in degradation of the water quality in many South African rivers to the point of decreased fitness-of-use for specific water users. Consequently, the assessment of contaminant load contributions to streamflow and other water bodies originating from point sources is a prerequisite for understanding of water quality patterns and problems in catchments.

Other modelling issues include the importation of salts through inter-basin transfers and the need to predict the implications of future developments in a catchment on catchment water quality.

There are conflicting demands on the water source of a catchment to both supply water users with a desired level of water quality as well as to use the water source as a method of disposing of waste. It is necessary for stakeholders to express their water quality requirements in terms of "the impact of water quality on their livelihoods". These requirements are then compared to the present water quality status.

The models used in South Africa for water quality modelling can be divided into five types, viz. simple process models, detailed process models, system analysis models, daily reservoir hydrodynamics models and sub-daily river hydrodynamics models. Brief descriptions of some of the models that fall under these categories are provided.

• Simple process models

This is a coarse-scaled model for salinity production and transport in large multi-use catchments, specially designed to be driven by the same natural flows that drive the Water Resources Yield Model (WRYM) and Water Resources Planning Model (WRPM) system analysis models. WQT is used to determine salinity parameters, which are then input to the WRPM model for multiple stochastic optimisation runs in large river systems.

IMPAQ (Impoundment/river Management and Planning Assessment tool for water Quality simulations) ? monthly

This is a medium-to-fine-scaled model for salinity, sediment and phosphate production and transport in large multi-use catchments, specially designed to be driven by the same natural flows that drive the WRYM and WRPM system analysis models outlined in (iii) below. It has a washoff routine that uses SCS Curve Numbers to allow any mix of land-uses to affect sediment and phosphate production, which are derived from a combination of loading functions, potency factors and the USLE approach. Non-conservative processes are allowed to play a role in a channel transport module and a simple mixed reactor reservoir module. IMPAQ is used in conjunction with WRYM to generate very long sequences of monthly loads/concentrations of selected constituents in large river systems.

ACRU ? daily

This is a fine-scaled model for sediment and phosphate production from individual small catchments with a limited range of agricultural land-uses. It is driven by daily rainfall and uses soil-moisture budgeting according to a discretisation based on soil texture classes and agricultural practices. It is recommended to investigate localised impacts of land-use and their related management options.

This is a medium-scaled model for salinity production and transport in large multi-use catchments. It is built around the relatively black-box daily Pitman rainfall-runoff model, it allows urban washoff as well as operation of reservoirs, wetlands, and coarse irrigation activities. It is recommended as background support for WQT applications where certain parts of a multi-use catchment require more detailed treatment, or to assess salinity management options.

This is a fine-scaled model for salinity production and transport through formalised irrigation schemes and allows operation of supply reservoirs, river channel transport, diversion devices, primary and secondary canals, balancing dams, artificial drainage, groundwater variability and a wide range of irrigation practices. It is driven by daily rainfall and uses soil-moisture budgeting according to a discretisation based on soil texture classes, location on the landscape, and agricultural practices. It is recommended as support for any of the other models to assess irrigation impacts of large or multi-offtake irrigation schemes, or to examine management options for salinity control.

This is a medium-to-coarse-scaled model for production and transport of salinity, temperature, sediment and a range of non-conservative constituents in medium-to-large multi-use catchments. Its water quality chemical simulation components are comprehensive and it uses relatively black-box rainfall-runoff functions, different forms of hydrological channel routing and treats reservoirs as simple mixed reactors. It may be used to assess water quality outcomes of management and operational options in medium-to-large catchments.

• Detailed process models

Detailed process models incorporate sophisticated processes, such as adsorption-desorption, decay and plant uptake, into the simulation of contaminant movement and transformation in soil and water. These contaminant processes are integrated with relatively complex hydrological and sediment models.

• System analysis models

The following two models are used to optimise the allocation of water on a monthly basis throughout a large multi-use river system, according to a penalty structure, for a given time horizon of water demands and allowing stochastic variation.

This model is used to calculate the long-term yield from a specific flow series, to examine operating rules or to develop yield-reliability curves.

WRPM allows various sub-systems to support each other during deficit periods and is used as a planning tool to explore augmentation or restriction strategies.

A 1-D finite difference model using LaGrangian principles to simulate all energy and kinetic exchanges as well as salinity processes.

• Sub-daily river hydrodynamics models

Three one dimensional models have seen operational use in South Africa: MIKE11, ISIS and DUFLOW. All three models are based on a finite difference application of the full St Venant?s flow equations to a series of cross-sections of the river channel and flood-plain. A range of conservative and non-conservative water quality routines are incorporated in all three models.

• Results of water quality modelling

Following an analysis of water quality three cases could occur

• the water quality is significantly better than is required by the stakeholders (unstressed status);
• the water quality is close to that required by the stakeholders in which case it may be necessary to revise the water quality requirements or the need to discharge waste (threatened status); or
• the water quality is worse than what is required by the stakeholders in which case the implications to other stakeholders would require assessment (stressed status).

Results from water quality modelling can be displayed using a GIS thereby showing both the types of problems that occur in a certain area as well as the location of the water quality issues.

Alternatively a time series of results or exceedence diagram to illustrate the acceptability of the water quality can be displayed.

• Strengths

• Some systems (models and monitoring networks) are in place, however, these are currently far from ideal.

• Models or mathematical tools do go some way to address the problem of data shortages.
• Using long rainfall records for catchment modelling provides a useful method of extending or infilling streamflow and hence water quality records synthetically.
• Another method of dealing with data shortages is the use of simple empirical predictive tools based on the sequences of long sequences of streamflow information to make statistical estimates of water quality. Using simulation models usually requires more intensive data input than the simpler, more qualitative approaches.
• Models of differing complexity are available allowing simpler, more qualitative methods to be applied in catchments that are considered less stressed and more complex models to be applied in threatened or stressed catchments

• Weaknesses

• The present water quality databases in South Africa are limited in duration or spatial representativeness and often represent only intermittent samples.

• To date in many instances, monitoring points have been positioned to provide information on man-made impacts with little consideration being given to background or un-impacted state of a river or stream.

• Only a few models are available for water quality modelling. This is especially true of daily time step models.
• Most models that are used for water quality simulation are data intensive and are complex to set up.
• In some models the configuration and calibration of water quality simulation models for use in water quality assessments and investigation of management options requires quantified point sources as essential inputs. The current day point source waste discharges and historical waste discharge records or trends are required for proper calibration of the models over a representatively long time period. This information is generally not available in many parts of South Africa.
• Information on both point and non-point source pollution is needed for modelling. This information is often not available and non-point pollution is particularly difficult to monitor.

• Opportunities

• Advances in technology may provide a means to monitor water quality more effectively than in the past (e.g. the use of telemetry and data loggers)
• Continual advances in research allow more advanced water quality models to be developed.
• WMAs and CMAs should be empowered to address their specific catchment problems at a local level.

• Threats

• Ideally monitoring points should be distributed over the catchment to provide a balanced view of water quality changes.
• Many catchments are stressed in terms of water quality and lack both water quality data and catchment information. There could be a delay in obtaining a good monitoring network and database of water quality monitoring information for these catchments for management and modelling purposes.

The data requirements to establish the CMS and for individual license applications include

• water quantity and quality data (water quality data requirements have been addressed in Section 7.3) ;
• invariant data;
• variant data;
• ecological and environmental data; and
• economic information including projections of demographic trends and economic development and consequent water demands (cf. Sections 3.2.4.1 to 3.2.4.4).

The data may take the form of

• regular or intermittent time series;
• spatial coverages; or
• dimensions of infrastructure.

This data may be linked to

• points (e.g. monitoring stations, dams or weirs);
• river reaches; or
• polygons (e.g. land uses).

This following list of data requirements is a non-exhaustive list of the types of data required to assist in planning, design or systems operations by the CMA.

Invariant data implies data, which remains constant over time. Minimum invariant data requirements include

• Baseline land use information (e.g. Acocks? Veld Types)

• Precipitation data (preferably daily rainfall)
• Maximum and minimum temperature (either daily or monthly)
• Physical catchment attributes which include

• Location
• Area
• Altitude
• Catchment configuration

• Mean monthly reference potential evaporation
• Geology
• Topography
• Soils data which includes at a minimum

• depth; and
• texture class

• Cadastral boundaries such as

• provincial boundaries

• Natural features such as

• rivers

In addition to invariant data there is also the need for data, which does vary over time such as land use information. Therefore data is required on the location and nature of

• roads and transport routes;
• game parks and nature reserves;
• commercial forestry;
• dryland agriculture;
• irrigated agriculture;
• markets;
• dams;
• transfer schemes;
• rural and urban centres;
• biodiverse areas;
• population distributions and status;
• mining zones;
• wetlands;
• return flows from water treatment sites and irrigation schemes; and
• point and non-point source loadings

in the WMA concerned is required.

Land use and water use data is required for the CMS depending on the land use in the individual catchments being managed. If irrigation demand and supply routines are activated in a catchment then data is further required on

• mode of scheduling;
• source of the irrigation water;
• supply and abstraction losses;
• crop characteristics; and
• soil texture class.

If dams are present in the catchment then data is needed on

• dam capacity;
• surface area;
• dimensions;
• storage to discharge relationships; and
• evaporation amounts from the dam.

If water is abstracted from the dam information is required on

• the amount of water abstracted from the reservoir month-by-month for uses other than irrigation (i.e. urban or ecological).

If an inter or intra-catchment transfer is in place then additional data required would include

• the total volume of water to be pumped into the reservoir month-by-month from an outside catchment/subcatchment.

The amount of water that can potentially be pumped from the river, if off channel storage is available, depends on the

• river conditions;
• pumping capacity;
• number of pumps; and
• number of hours per day that can be used to pump water.

Data required on SFRAs would include

• the type of trees being planted.

Non-SFRAs would require month-by-month inputs on

• water use coefficients;
• interception loss by vegetation; and
• fraction of roots in the topsoil.

Ecological data requirements include

• class of the reserve;
• biodiversity data;
• operating rules;
• instream Flow Requirements (IFRs) at selected sites or reaches;
• environmental flow and tidal exchange requirements of estuaries; and
• aquatic biotic responses to flow changes due to upstream developments.

To make economic decisions data is also required on

• crop selling prices;
• distance to markets and mills;
• labour costs;
• transport costs;
• capital equipments costs;
• conversions costs from one land use to another;
• gross margins;
• enterprise margins;
• number of employees per activity; and
• water use charges.

Data is required on historical, present and projected water demand distribution in the catchments.

4. Basis and Requirements for a Common Database

There have been a number of developments in southern Africa over the past few decades that have contributed to the need for a more collaborative approach to some aspects of hydrological and water resource data analysis.

There have been some developments in the availability of hydrological and water resource modelling tools. There is now a diverse set of applicable models in practical use and the application of models is no longer confined to the groups that developed them. However, while most have built-in results analysis procedures, there has been little standardisation of approach with respect to the methods used to display and analyse the simulation results.

The recent proliferation of models simulating a number of different processes is a mixed blessing as is not only means that the user has to spend a large amount of time choosing between different models but must expend considerable effort amassing and manipulating huge quantities of data the model may require. If the user wants to perform an analysis of more than one process two or more models need to be coupled together, while the underlying assumptions may be somewhat incompatible, more frequently the data structures are so different that coupling requires extensive data conversion work. Extensive modelling data requirements coupled with inconsistencies in data storage within and between different organisation and state departments means that a large amount of a modellers time is spent, when performing catchment studies, on data management, collection and processing.

• As a consequence of the previous point, there are also quite large volumes of simulated data available, but in a variety of different data storage formats.
• Perhaps the most important consideration is that a broader group of scientists and engineers now require the information contained within hydrological time series. Many of these scientists are non-hydrologists who are not always familiar with the full range of methods that can be used to present hydrological data. A more standardised approach could possibly alleviate the confusion often caused by being faced with different methods of analysis and display.
• There is a trend toward viewing hydrological data in a spatial context and developing software products that allow spatial and time series data to be better integrated. This is essential for effective integrated catchment management.
• Some groups have had less than satisfactory experiences with the purchase and / or use of imported software, which has not been designed for the local situation.
• The funding agencies are under pressure to avoid financing duplication of research effort.
• One of the results of these points is that many different data sets and analytical / simulation tools are being used and that the outputs are being presented in many different ways. Modern software development tools make it relatively easy to generate programs that can produce impressive graphics displays, etc. and there is the real danger that several groups will follow independent paths toward creating products that satisfy only a limited range of users, but are largely incompatible with parallel developments. It therefore seems to be sensible to pool the available resources within South Africa and develop time series analysis and display software products that can be flexible enough to satisfy the majority of potential users requirements in an integrated package.

Thus there is a definite need for a common database, which can be accessed by multiple models. Input for the models and output from the models should be written in a common format, which can be accessed by different models as shown in Figures 5 and 6.

Görgens (1999) identifies the need for an integrated catchment modelling and information system. A number of models are interfaced with a common database, which is in turn linked with time series and GIS viewing facilities. A system of this nature will, according to Görgens (1999), allow for varying levels of input depending on the complexity of the problem. Examples of some systems that have been developed along these lines include HYMAS, ICIS, IMPAQ, BASINS and NWBM. These systems appear to have strengths in the visualisation software however, often are restrictive owing to their resolution or preference for a single modelling system.

If the system in Figure 5 were used then methods would have to be devised to convert the output from one model to input to the next for each of the models used. However, if the system set out in Figure 6 is used then only one method of input and output extraction from a common database is required. As can be seen from the two diagrams the efficiency in data manipulation using a common database is much enhanced, as data manipulation only needs to take place between the database and the models, which reduces the amount of effort spent on data manipulation.

Linking models in series is the most common type of model linkage. It involves taking output from the one model and using it as input to another model. The implication of this is that all calculations by the one model need to be complete before commencing the next model?s run. The usual requirement for linking models in series is the conversion of output from the first model from its unique format to the unique input format of the other model using a data transformation program. A time manager and a set of files that is accessible can facilitate communication by the models by all models.

A present there is no system that is able to link spatial data with time series data and a viewing platform in GIS. The database behind ArcView for example is a dBase file format to store and link data. This requires the data storage in a flat file format that allows for a large amount of redundancy and is not easily queried. A relational database has the ability to reduce redundancy and allow for easier query options to be developed. The ARC/INFO and ArcView systems are being developed to incorporate relational database facilities, these are, however, not available at present. It was therefore decided to review several different existing database applications that have been developed locally and internationally to assess how they have dealt with the problem of linking time series, attribute and spatial data together. It was also believed that substantial time and effort might be saved if and existing database could be used.

When assessing the option of using an existing database system it is necessary to weigh up

Neither option is particularly suitable, as both would require prohibitively large investments in terms of both time and resources. It was therefore decided that the option of choosing a database structure that had been locally developed and was in general use would be the most appropriate option. In this regard DWAF was approached to gain an idea of the types of systems they are using and how these systems could be adapted to a working version of a database that could link to the models that were required in the DSS. It was believed that by "piggybacking" on the protocols and standards developed in DWAF the setup time for the database would be decreased and a functioning system would be in place in the least amount of time. It was also thought that by using a system that conformed with the standards of DWAF the system would be generally better accepted by DWAF and would be used more readily in actual applications especially in terms of the SEA process that the DSS supports. It was also believed that the most recent technology in terms of database systems should be used, as that would limit the possibility of the system becoming outdated and redundant.

In several meetings and discussions with different persons in different departments in DWAF it became clear that there is no real consensus on the database structure to be used in DWAF itself. There are several sections in DWAF involved with data manipulation and management however there is no standard method of storage or manipulation and most of the different sectors use their own data storage formats and database packages. In DWAF the integrator has been given the task of standardising all the different data formats and incorporating them into a common database and data storage format. It also became clear when speaking with various role players in DWAF that none had a system that stored both simulated and observed data and that DWAF was mainly concerned with the collection and storage of observed data.

Several of the current databases used in DWAF at the moment are named below. This is not an exhaustive list as other sectors in DWAF may have another systems.

The integrator has set up a database architecture that needs to be conformed to in terms of database design and development. The data standards within the database architecture are not however explicitly defined. The Informix database system while being used by the integrator is not necessarily standard throughout DWAF. There also apparently seems to be some debate as to which is the best database package to use, in terms of Informix or Oracle.

It is clear from the discussions with DWAF and other organisations that the database issue needs to be given a lot more thought. It has been suggested that the consultants review some of the current databases used in DWAF at present and particularly the Regis system that is been used by the geohydrology section. It is, however, clear that the consultants must use the ArcView GIS for the spatial data storage component and link this to a relational database. At this stage several, options have been suggested in terms of which database to use and link to the ArcView system. The most promising link looks to be that which has been adopted by the Regis system which links an Oracle database to the ArcView GIS. It is hence this approach that is most likely to be suggested by the consultants in the design but is subject to further investigation.