Review of the water resource hydrological analysis decision support systems in South Africa Open Access

Hydrological analysis decision support systems (DSS) have been applied to manage water resource systems and catchments. The National Science and Technology Centre for Disaster Reduction in Taiwan developed a DSS that is geographic information system (GIS)-based to assist in managing emergency setting up response strategies (Hsu et al. 2011), and the Colorado State University, Middle Rio Grande Conservancy District, and the New Mexico Interstate Stream Commission developed DSS to help to improve the quality of making a decision and also to help water managers to match water deliveries to crop water requirements (Oad et al. 2006). It is vital that hydrological analysis DSS is enhanced and maintained to ensure the supply of water used for various purposes such as industries and household consumption. In South Africa, water resource planning DSS or models, such as the water resource yield model (WRYM) and water resource planning models (WRPM), are used to determine and manage the amount of water available for use within a system. DSSs were developed in the early 1970s, with the Pitman model's first development in 1973 and the WRYM later, and have continued to be enhanced and applied or used for use by water resource planners, scientists, and engineers (de Jager et al.; Coleman & Pitman 2008a; Hughes 2013; Pitman & Bailey 2021). The hydrological analysis DSS still relies heavily on monthly time step data as input, but the daily time step data are yet to be incorporated. The option for a choice used in either daily or monthly time steps is related to the complexity of the simulation procedure; the easier the DSS, the longer the time interval, and the more complex the DSS, the shorter the time intervals (Hughes 1992, 2013). However, more comprehensive research is required to optimize the process of obtaining satellite data for better robustness and reliability.

The current hydrological analysis DSS technology and policy perspective have been applied for a very long time, and there is a need to review and recommend an upgrade, considering artificial intelligence, cloud computing, and the Internet of Things among other key additions (Lee et al. 2018; World Economic Forum Report 2018; Xu et al. 2018; Du Preez & Sinha 2020). To supply water to a resource, water resources systems should be managed effectively within a preferred catchment with the required water, and the required water differs with time (Wurbs 2012; Wurbs et al. 2012). The intent of the water resources planning process in a catchment is to balance what is required and lost with the supply of water resources (Wurbs 2005; Seago 2016). This process involves several aspects that are briefly discussed below.

• (a) Water resources have the capability, how much they can supply, and at what quality. What are the water requirements, if there are users, what are their priorities, and what risk can they accept a non-water supply?

• (b) At any given time, the capacity of the water resource is directly proportional to how the resource is operated, which includes all the transfers, the user priority that changes at each level, and cost savings of operation, particularly the water quality not forgetting maintenance (Wurbs 2005; Coleman 2007; Seago 2016; Nkwonta et al. 2017).

• (c) There have been experiences where the resource requires interventions, especially during drought periods, or otherwise, and the need necessitates the water resource to be balanced with the requirements maintained. Different approaches have been taken, including infrastructure development possibility to increase the yield, resource reallocation, water reuse, and managing the activities within the catchments that affect the resource (Parker et al.; Draper et al. 2004).

• (d) Lastly, it requires monitoring of the water resources to ensure that the agreed allocations are adhered to.

The water resource planning process has different components applied to test several scenarios on how well the system will perform before the catchment's experience. The DSS is thus a tool applied to help test the impact of what it would be like if certain infrastructures were to be developed or supplied from the resource in a reliable manner for scientifically sound decisions to be followed.

Hydrological analysis DSS has been developed and enhanced for use or applied in existing catchments across the country and used with datasets. DSS is based on complex techniques and software development programming languages that help simulate the etiquette of water resource systems (Zagona 2001; Belachew & Mekonen 2014). Most DSSs provide sophisticated graphical user interfaces with menu-driven editors for entering and revising input data and displaying simulation results in tables, plots, and other features that allow reservoir and river system schemas to be created by connecting and selecting radio buttons’ (de Jager et al.; Draper et al. 2004; Seago 2016). All DSSs have assumptions and shortcomings that may include the following:

• (a) How to simplify the real-world challenge.

• (b) Which variable needs to be guesstimated?

• (c) Output depends on the quality of data.

• (d) Selection of different DSS to use depending on a specific need.

• (e) The best programming language and the quality of the team to ensure a development procedure.

DSSs have played an ‘increasingly important role in various aspects of water resource planning and management throughout the world over the past several decades’ (Wurbs 1998, 2020). Therefore, DSSs should be extensively tested, explicitly documented, and attainable. Wurbs (1998), Labadie (2004), Rani & Moreira (2010), and many others have reviewed the massive literature on modeling multiple-purpose river/reservoir system operations.

The hydrological analysis DSS uses a complex mathematical solver to solve water resource planning studies networks that are linked to different management and operational policies. The solver also helps in assessing the catchment's yield in short-term and long-term analyses (de Jager et al.; Coleman 2007; Coleman & Pitman 2008b; Nkwonta et al. 2017). The major strength of the DSS is the use of the basic building blocks used to combine the system network that shows the link between the different elements of the catchment and how they relate to each other from upstream to downstream elements or reservoirs (de Jager et al.; Wurbs 2005; Coleman 2007; Coleman & Pitman 2008b; Seago 2016; Nkwonta et al. 2017). In South Africa, hydrological analysis DSS is used to support the management of water and the amount of water available for use at a given point (de Jager et al.; Schmidt & Schulze 1984; Jeleni et al. 2011; Serrat-Capdevila et al. 2011; Odiyo et al. 2015; Nkwonta et al. 2017). The hydrological analysis DSS was developed in the early 1970s and has continued to be enhanced by water resource planners, scientists, and engineers.

The DSS uses hydrological monitoring data as input. Hydrological monitoring data, such as rainfall, streamflow, and evaporation, are as important as modeling DSS (de Jager et al.; Schmidt & Schulze 1984; Pegram & Pegram 1993; Odiyo et al. 2015; Nkwonta et al. 2017; Marques et al. 2022). Thus, one cannot look at the use of the DSS without considering hydrological data quality, quantity, and other datasets.

South Africa's Department of Water and Sanitation (DWS) Water Resource Modeling Framework (WRMF) DSSs include the rainfall DSS, hydrological analysis DSS, which includes the Rainfall–Runoff and WRYM, and the water resources planning model (WRPM). Some of the other DSS developed in South Africa and available to use include the Agricultural Catchments Research Unit (ACRU), developed and maintained by the University of KwaZulu Natal (2023) while Rhodes University developed the Water Quality Systems Assessment Model, and the Agricultural Research Council and South African Weather Services (SAWS) have also developed information dissemination interfaces to help share rainfall-related information. WRMF contains, besides DSSs, such as the rainfall DSS, hydrological analysis DSS, WRPM, other pre-processors, such as the instream flow requirements (Department of Water & Sanitation 2012a), daily diversion (Department of Water & Sanitation 2012b) and Stochastic Model of South Africa (STOMSA) (Department of Water & Sanitation 2012c). This review focuses on the hydrological analysis of DSS.

The Hydrological Analysis DSS incorporated Rainfall–Runoff DSS and related calibration functionality into the WRYM DSS (DWS 2020) to ensure that the implementation has the following advantages:

• (a) Reduced modeling steps and data conversion processes between the two DSSs.

• (b) Consolidation of simulation elements among the DSSs reduces duplication with the advantage of lower software maintenance costs.

• (c) The simulation capability of the catchment runoff processes in WRYM allows explicit and interactive modeling with other modules, including irrigation, wetlands, groundwater-surface water interaction, mine modules, and streamflow reduction activities (Coleman 2007; Nkwonta et al. 2017).

• (d) The development paves the way for integration with the rainfall stochastic model and the automation of several processes, such as calibration, climate change scenario analyses, and performing uncertainty analysis.

The development of the Hydrological Analysis DSS began in the early 1970s, with the Rainfall–Runoff also known as the Pitman model first developed in 1973 and the WRYM originally starting development around 1985 (Coleman 2007; Nkwonta et al. 2017). The hydrology of rainfall–runoff, or the Pitman model, is used to develop hydrological time series and module parameters required to calibrate simulated flows and produce naturalized flows, while the WRYM is used to determine the amount of water that can be obtained or made available to supply a certain water demand considering the reliability of water resource systems using historical and stochastic hydrology or runoff for constant catchment development scenarios without failure (de Jager et al.; Rugumayo & Kizza 2002; Makapela et al. 2015; Seago 2016). Failure in this context refers to the period or interval during which the reservoir drops to the minimum required level or yield demand (Rugumayo & Kizza 2002).

The volume or amount of water in reservoir storage is directly proportional to the demand imposed on it and the topography of the catchment area, including the streamflow and rain that falls within the resource (Sood & Smakhtin 2015a). For a reservoir to be reliable, it is required that through operations it indicates how it can meet the supply or the water demand without failure (Carty & Cunnane 1990; Koutsoyiannis 2004; Jager & Smith 2008). The probability that a certain reservoir on a resource can have a yield met without failure at any given target draft is very high if the nature of the operation is not reliable (Nagy et al. 2002). Hydrological analysis DSS is a monthly time step based on the need to prepare natural time-series data and other required data to determine the yield capacity of the water resource system (Seago 2016).

The Hydrological Analysis DSS uses the out-of-kilter algorithm built in the WRYM, the same algorithm that was proposed in the early 1960s by Minty and Fulkerson to solve flows in a network problem that may represent physical and nonphysical systems (Fulkerson 1961; Minty 1961; Comley 1996). The DSS for the Middle Rio Grande uses a mass balance equation and linear programming solver with an objective function to find an optimum water delivery schedule for the service areas in an irrigation system (Oad et al. 2006). The network elements that store and move water around are dams; channels for canals, tunnels, transfers, and rivers; and catchments. Minty and Fulkerson's Out-Of-Kilter algorithm states that if the flow of water between two nodes, A and B, denoted as F_ab is the flow from note A to node B is negative, i.e. F_ab < 0 then the flow is from node B to node A. There are benefits or costs incurred when moving or transferring water between two nodes in a network problem, the benefit is denoted as CostF_ab. The flow through a channel or canal must have a minimum flow of MinF_ab and a maximum flow of MaxF_ab defined as the canal has a limit on how much it can transfer (Fulkerson 1961; Minty 1961; Comley 1996).

The water resource network characterised by nodes, A and B with flow F_ab with costs and benefits CostF_ab incurred when water is transferred between them and defined the minimum flow MinF_ab and the maximum flow MaxF_ab, the following conditions were considered in the hydrological analysis DSS and WRYM (Fulkerson 1961):

• (a) If there is no cost to transfer water, then CostF_ab = 0;

• (b) If there is no minimum flow and flows move in one direction, then MinF_ab = 0;

• (c) If there is no maximum flow limit MaxF_ab = +∞. Option (c) was not implemented in the WRYM.

The complexity of a water resource network, combined with computational analysis, with the use of an Out-Of-Kilter algorithm, was a better implementable algorithm for hydrological analysis DSS or WRYM compared to the traditional linear programming algorithm. Thus, a set of flows that satisfy a given flow constraint in Equation (2) and conserve Equation (3) is called feasible (Fulkerson 1961; Minty 1961).

To determine the amount of water to be stored, the analyses begin first by determining the calibration of simulated flows and producing naturalized flows and then using historical and stochastic hydrological runoff to determine the volume of water that can be obtained and stored (Coleman 2007; Nkwonta et al. 2017).

WRMF is a graphical user-friendly interface DSS, with a substantially enhanced user interaction (Wurbs 2005; Coleman 2007; Seago 2016; Haasbroek 2020). WRMF provides model-assisted procedures with guidelines and training materials for model configuration and modification. The WRMF was developed using the DELPHI programming language and provides access to a water resource database to store and manage data, with the option to configure and display results within a Microsoft Windows environment (de Jager et al.). The user can create new study data or import other study data that have been previously installed, and the datasets can be stored or saved into the database and exported and shared with other modelers. Different scenarios can be created within each study based on requirements or demands.

To successfully set up the DSS, it is important to have a good understanding of the physical characteristics of the catchment, the location of the rainfall stations, streamflow gauges, and human-induced impacts on the stream inflow data (Coleman 2007; Wurbs 2011; Seago 2016; Nkwonta et al. 2017). Human-induced impacts include the collection and collation of either supply infrastructure and land-use information, such as water usage data, land-use data, infrastructure data, and return flow data (Coleman 2007; Makapela et al. 2015; Nkwonta et al. 2017; Haasbroek 2020). In each study, hydrological data needed to be collated and prepared for pre-processing functions, including rainfall, evaporation, and observed streamflow datasets. The sources of the hydrological analysis data used by the model were managed through the Hydstra database from the DWS, private companies, and individuals also have data used for DSS setup. The following steps must be considered for a successful water resource catchment study:

• (a) Do rainfall analysis that uses Rainfall DSS to produce a catchment rainfall file.

• (b) Do hydrology analysis that uses a Rainfall-Runoff DSS to produce naturalized flows, and the yield analysis uses the WRYM DSS to determine the amount of water available at a given point in the catchment (Coleman 2007; Jeleni et al. 2011; Nkwonta et al. 2017).

• (c) Do planning analysis that uses the WRPM DSS to ensure the availability of water resources that can support changes in water requirements (Coleman 2007; Coleman & Pitman 2008b; Nkwonta et al. 2017).

On the vertical axis, Figure 2 shows the type of input data required to execute or run the model throughout the water resource analysis, and the horizontal axis shows the type of DSS that uses the input data. The data quality and quantity applied in the study of water resources determine the most effective method for using the data. The data can be used as it is or extended to cover the missing month data in cases where a study has already been conducted (Steve et al. 2011; Haasbroek 2020). The procedure followed to complete a catchment study with no data or extend the hydrological and rainfall data will require the following:

• (a) Water resource boundary delineation;

• (b) Water resource rainfall records;

• (c) Water resource natural runoff records;

The delineation of catchment boundaries refers to the derivation of a catchment map that provides the size, shape, and characteristics of the study area. The catchment rainfall and natural runoff time-series datasets for each study area were collected from the data mostly from the DWS Hydstra database and other institutions that collect other required data that are useful to the studies (Seago 2016; Haasbroek 2020). Once the data are collated and examined for quality and gaps, the result of the data, representing the naturalized hydrological time series, the following need to be considered:

• (a) Select rainfall stations to be used to create catchment rainfall for the study area.

• (b) Patch source point rainfall stations that can be used for catchment rainfall record extension.

• (c) Complete the source calibration and naturalization process.

The following steps or procedures must be considered to generate and calibrate the natural runoff time series:

• (a) Use a catchment rainfall record.

• (b) Create a network configuration for hydrological calibration analysis.

• (c) Assess irrigation areas, farm dams, and possible growth assumptions. Irrigation areas must be added to the network diagram, and re-evaluation and correction are necessary.

• (d) Recalibrate data and where necessary, repeat the steps until the results are acceptable.

Some of the challenges encountered in the above process might be how the quaternary, tertiary, and secondary catchments are broken down into different categories, as well as the calculation of the catchment areas and the dependency on the number of gauging stations available to obtain information or data and, in some cases, the use of information or data used from adjacent stations with observed data (Coleman 2007; Seago 2016; Nkwonta et al. 2017).

In addition to evaporation data, hydrometeorological data and diffuse water use are required. The generated files contain information on monthly naturally simulated incremental runoff (measured in million m³), monthly rainfall (measured in mm), monthly diffuse water requirements (measured in million m³), and monthly streamflow reductions (measured in million m³) (Coleman 2007; Seago 2016; Nkwonta et al. 2017; Haasbroek 2020).

When undertaking a yield analysis, the user has three options for analyzing the data: historical, historic firm yield (HFY), and stochastic analyses. For historical (or single sequence) analyses, the yield of a system is determined by imposing a selected target water requirement (or ‘target draft’) on the system and evaluating the system modeled behavior of the system (de Jager et al.; Seago 2016; Haasbroek 2020). If the historical run type is selected, the DSS scenario is analyzed using the hydrological monthly time-series data files (de Jager et al.; Coleman 2007; Seago 2016; Nkwonta et al. 2017). The DSS allows a maximum of 10 target energy requirement values to be specified for the analysis in a single run. In addition to single sequence analyses, one can determine the HFY of the system by means of an iterative process, where a range of target drafts are imposed on the system and the yield (or supplied amount) is determined for each target draft. To determine the HFY, the user must define two target drafts: an upper value and a lower value, between which firm yield is expected to occur. However, in the case of a stochastic run, the model generates streamflow sequences stochastically as well as appropriate monthly target flows for diffuse water requirements and streamflow reductions.

The water resource system's yield is determined using either historic or stochastic runs, which may be determined in terms of the water resource capability or hydropower generation potential and depends on the scenario system being analyzed and the specific requirements of the user (de Jager et al.; Seago 2016). The water resource capability of a system provides a measure of its ability to supply water requirements over either the long or short term, at a fixed development level, and for a selected set of operating rules.

To determine the capability of a water resource system with hydropower generation, water resource systems must incorporate hydropower plants. For this purpose, the target energy requirement is imposed using the power supply channel type, and the flow released will satisfy the defined requirements that are determined. If multiple target requirements are imposed, the model undertakes a separate analysis for each of the target drafts, and results are produced for each, enabling the comparison of the associated results.

When the DSS runs or executes a stochastic analysis, and several target drafts are selected, the target draft with a sequence in which failure occurs or is recorded when analyzed is associated with the reliability of supply (de Jager et al.; Steve et al. 2011; Seago 2016; Haasbroek 2020). Figure 6 displays the results of the stochastic run. The assessment of a system's reliability characteristics is based on the analysis of a range of target drafts.

Detailed output information can be obtained from the DSS run for any of the active reservoirs and channels in the water resource system. The DSS can display system components for plotting line graphs, including average channel flows, in Figure 4, reservoir and system storage volumes in Figure 5, reservoir elevation levels in Figure 6, rainfall and evaporation losses from reservoir water surfaces, net incremental catchment runoff into the system at a particular point, and pumping energy results (Coleman & Pitman 2008b).

If a historical or single-sequence stochastic yield analysis is undertaken, plots may be in the form of a line graph, where a single line is used to represent the component in question. Line graphs can also be combined by plotting more than one system component on a single set of axes so that they can be viewed simultaneously for comparison.

Figure 7 displays such plots that are extremely useful for checking the interdependence of system components and related operating rules.

For multi-sequence stochastic sequences, the results are often shown in the form of ‘box-and-whisker’ plots in Figure 9. These results demonstrate an easier way to pick up the probability distribution, and it works even better if there is more than one probability distribution on a graph. The box and whisker display ‘the view of a probability distribution ‘bell’ curve indicating the locations of specified exceedance probabilities within that curve’ associated with the data being plotted.

There are two steps in hydrological analysis DSS testing: verification and validation (Stedinger & Taylor 1982; Pegram & McKenzie 1991; Turner 2014). Verification resamples the statistics of interest from the generated sequences to see that the DSS faithfully reproduces to a reasonable accuracy those in the historical sample, and the validation testing used in the water resource reliability calculations involves tests that were not used directly in the definition or setting up of the DSS. A set of data files with different demands and supplier information is compiled to test each version of the DSS. The purpose of verification and validation is to provide the opposite of the assumption in showing that the hydrological analysis DSS does not display the correct assumption or is unable to provide the expected results of the questions asked (Robinson 1997; Chwif et al. 2008). This ‘stems from the concept that there does not exist a DSS that perfectly represents reality’; thus, it is not possible to validate the DSS (Page et al. 1997; Sánchez 2006; Chwif et al. 2008; Sargent 2010, 2013). Essentially, the process of verifying and validating the DSS is to increase the chances that DSS users can have confidence in the DSS (Robinson 1997).

One of the standard procedures after implementing a new or modified system element in the DSS is to validate, verify, and test whether the basic input specification and the related system network definition provided are correctly implemented and accurately represent the intended configuration. The hydrological analysis DSS provides users with numerous automated DSS verification and testing datasets exemplary setups and features that assist in avoiding mistakes during the process of configuring system networks. The most important of these is the validate DSS data feature, which is activated by the user and outputs a detailed validation report including a list of warnings and errors related to the identified problems (de Jager et al.; Chwif et al. 2008; Turner 2014). This provides the user with the opportunity to address problems before attempting a DSS. Extreme discipline should always be employed, as users ultimately carry the responsibility for DSS results (Markosian et al. 2011; Turner 2014).

Many hydrology-related DSSs globally simulate similar hydrological processes, and they resemble similar objects and mathematical equations. The assumptions and details of the estimation approaches, different time steps, spatial resolutions, and simulations are different for each DSS (Lindström et al. 1997; Haddeland et al. 2011; Neitsch et al. 2011). Some DSSs, including the Soil and Water Assessment Tool (SWAT) (Neitsch et al. 2002), the Hydrological Simulation Program-Fortran (Bicknell et al. 1996), and the ACRU (Schulze & Lynch 1997), are applied at different time scales, water resource scales, and even larger basin scales. In principle, other DSSs can be applied on a larger global scale only if several data or information constraints are addressed prior to their occurrence in practice. Table 1 compares the different DSS with a focus on the different algorithms applied and the input dataset time steps (Nilebasin 2017; University of Kwazulu Natal 2023; Kim et al. 2008; Sileet et al. 2014; U. S. C. Stockholm Environment Institute (SEI) 2015; SEI 2016; Ayivi & Jha 2018; Suleman 2018; Tan et al. 2020; Wurbs 2021; Agarwal et al. 2023). Information on the developers of the DSSs in Table 1 compares the different DSSs' time steps, input data, algorithms applied, schematics or tools used, and how results are presented. Additional information on each DSS is supplied in Table 1.

The ACRU DSS has its ‘origins in a catchment evapotranspiration-based study carried out in Natal in the early 1970s. ACRU was derived from the Agricultural Catchments Research Unit of the Department of Agricultural Engineering of the University of Natal in Pietermaritzburg’, South Africa (University of Kwazulu Natal 2023). The ACRU DSS can be used for different analyses, such as hydrology design, crop yield, and irrigation demand. The DSS has most of its existence used monthly time step, but the latest version can now use daily time step. The three main object classes are the components used to represent physical entities, data, and processes.

The US Center's Stockholm Environment Institute initiated the Water Evaluation and Planning (WEAP) system, taking a water resource-integrated approach to planning and modeling water quality issues in rivers (S. E. I. U. S. Center; Seago 2016; Jayantari et al. 2019; Tena et al. 2021). Zambia, an African country, used the WEAP model to assess water resources in the Mutama-Bweengwa, Kasaka, and Magoye sub-catchments. The study examined different scenarios for water demand and potential future-related estimations using historical hydrometeorological data from 1951 to 2018. In addition to the application of WEAP in South Zambia, WEAP was also applied in Indonesia's Bali Province, Bangli District, in the Unda River Basin to evaluate the water availability and use the model for future scenarios to determine how to manage the planning of the resources best. The findings were mostly on the irrigation area, as the model results displayed a shortage between May and August.

Many agencies, including the Texas Water Resources Institute, worked together with Ralph Wurbs to develop the Water Rights Analysis Package (WRAP) model. The aim was to provide a water resource system with the capability to simulate the management of river systems and determine the reliability of the results. Modeling of the water resource system supports assessments of hydrology, water availability, and supply reliability for a given condition with the resource on a monthly time step (Wurbs 2021). The daily time-step module was recently developed, which can perform short-term reliability modeling with flow forecasting and routing (Wurbs & Lee 2011; Wurbs et al. 2012).

The Nile Basin Initiative developed the Nile Basin Decision Support System (NB-DSS), which relies on three components: the information system, the analytical part containing simulation and optimization tools, and a multi-criteria analysis tool (Digna et al. 2018; Kamel et al. 2019). DSS was applied to the Blue Nile Sub-basin to assess the positive and negative impacts of the planned water projects.

The SWAT model is a continuation of the United States Department of Agriculture's Agricultural Research Service experience with a complex model forecast of water and sediment circulation in ungauged basins (Neitsch et al. 2002; Neitsch et al. 2011; Devia et al. 2015). Makwana & Tiwari (2017) investigated the need for continuous daily streamflow based on rainfall in arid and semi-arid regions. In their study in the Limkheda watershed of Gujarat, India, the SWAT and neural networks (NNs) were compared to determine the performance of a continuous runoff simulation in a hilly and agricultural watershed. The results showed that SWAT displays a better estimate of the water balance than NNs.

The European Hydrological System (or Systeme Hydrologique Europeene, SHE) ‘was initiated as a collaborative venture in 1976 between the Danish Hydraulic Institute (DHI)’ in Denmark, the Institute of Hydrology in the United Kingdom, and a design and engineering consultancy, named SOGREAH in France (Refsgaard; Graham & Butts 2005; Devia et al. 2015). Prucha et al. (2016) used the MIKE SHE/MIKE11 model in South Africa's Mokolo River Basin to develop a flow system to simulate potential indicator inputs. Because the basin has less data, calibration was difficult. One of the conclusions drawn from the study was that the model and associated software contain certain features that make it relatively easy to establish a model set up for a particular catchment.

The results of research on alternative hydrometeorological and precipitation data show that satellite products have the potential to represent or be an alternative to the temporal and spatial variability of data in a water-resource catchment study (Maswanganye 2018). To use different remote-sensing products, attention needs to be paid to how the data are stored, the time zone used, and what elements were considered when producing specific data. MODSIM has been configured to provide hourly and daily real-time regulatory guidelines to automate control of gates in the Imperial Irrigation District system (Labadie 2006), while WRAP has a daily time step modeling capabilities ‘that include flow forecasting and routing and disaggregation of monthly flows to daily’ (Wurbs et al. 2012; Wurbs 2021). In addition, Srikanthan & Pegram (2009) performed work on daily time-step rainfall modeling, which is the daily version of the current monthly input catchment rainfall file for hydrological analysis DSS. The need to investigate or incorporate the use of daily time step data on the hydrological analysis, DSS should be researched for both satellite and ground-based data.

The work completed on the incorporation of Rainfall-Runoff and WRYM (DWS 2020) paves the way for integration with the rainfall daily stochastic model and the automation of several processes such as calibration, climate change scenario analyses, and performing uncertainty analysis. The future of water resource hydrological DSSs in South Africa should consider a scenario that applies satellite data (directly or derived) to be included when calibrating.

Substantial progress has been made in acquiring and understanding satellite-based data in the last decade (Sood & Smakhtin 2015a). Maintaining and including remote sensing technologies can help to clarify and bring changeable options associated with data inputs and observations. The future of water resources hydrological DSSs is largely associated with the development of satellite-based technologies and the public availability of satellite data with a list of bias correction factors. The current ground-based hydrology data measurements, as applied in the DSSs in South Africa, are seen to still be reliable and more accurate than the use of satellite-based measurements. The challenge here is that ground-based data do not cover the entire country's catchment, and thus there is a need for satellite-based data (Sood & Smakhtin 2015a, 2015b; Wurbs 2021).

DWS conducts dam surveys every 5 years of reservoir existence, and one of the reasons for the survey is to determine how much, if any, the size of the reservoir changes. Sedimentation plays a key role in the changes in the size of the reservoir that accumulates, the erodible materials that add to the surface area of the reservoir, and the sediment load that is supplied decreases with time over the duration of the event (White 2001; Rulot et al. 2012). South Africa is currently faced with land invasion and less agricultural practice in some areas, with farmers abandoning farms due to invasion, funding, and suppliers. The need to consider the computation of sediment yield is vital, a possible review of the existing algorithm or consider an alternative algorithm and an option to use spatial modeling can be of additional help if considered.

In addition to the 5-yearly survey results, in South Africa, the need to estimate long-term sediment yield has been applied for many decades in different parts of the globe to determine the size of the sediment storage pool rather than to estimate the size of the reservoir with sediment. However, DSSs that estimate long-term sediment yield may not, at least for now, be accurate for issues such as floods. Challenges will always occur when evaluating techniques for sediment yield, and the collection of data that will be used to determine the long-term trends in South Africa's sediment yields will be interesting to see how and where the data will come from (Garcia 2008). Two possible approaches for measuring sediment yield that are well known are ‘inspection of the sediments volume deposited in the reservoir and a continuous monitoring of fluvial sediment discharge.’ The first method is more accurate than the second, as reservoir construction eliminates missed or underreported events at streamflow gauge stations; however, the second method captures the description of both spatial and temporal patterns.

This article reviewed the hydrological analysis DSS's Out-Of-Kilter algorithm solver that has been used and the traditional monthly input calibration data. The monthly input data play a critical role in shaping effective water resources management decisions or strategies within the country's unique socioeconomic and environmental context. The monthly data are suitable for analyzing long-term patterns and strategies while it becomes a challenge to identify short-term trends and make possible quick tactical adjustments with daily or hourly input data. Although the choice of the algorithm seems to have been a suitable option for the DSS, an alternative algorithm might be of help or an alternative option especially in catchment that may not be suitable to apply the algorithm.

The catchment used at first to test the algorithms differs from many others in characteristics and thus the other algorithms could provide a different perspective. Both the algorithm and the monthly time steps have been well developed and applied in all catchments and the current technology will help provide new technology features, cloud service, and faster processing machinery. The incorporation of one or two alternative algorithms may improve the DSS performance and incorporation of the less time step datasets will help refine the output or information presented by it.

The DSS is data-driven, the more data input you have, the better or more refined the result. With refined daily data input, the information is likely to be different from the monthly input data. It is more appropriate to use a DSS with input data obtained from a large amount of historical data as it will take time for the algorithm to converge. However, when the data are less and its convergence is easy to obtain, daily base data DSS are then preferred.

Gerald de Jager and Moroka Letshela contributed to the review with their invaluable knowledge, experience, advice, and work. The Department of Water and Sanitation is thanked for providing information on the water resource models.

All relevant data are included in the paper or its Supplementary Information.

The authors declare there is no conflict.