THE DISA HYDROSALINITY MODEL
REPORT NO. 369/1/01
November 2000

EXECUTIVE SUMMARY

INTRODUCTION

The Breede River is one of South Africa's primary vine and deciduous fruit growing areas. Although irrigation occurs throughout the catchment, the greater portion of the irrigated lands, comprising some 45 000 ha, is situated between Worcester and Bonnievale. This irrigation occurs along the main stem of the Breede River, as well as along its tributaries. The majority of these lands are supplied with water from the Greater Brandvlei Dam as part of the Greater BrandvIei Dam Government Water Scheme (GBDGWS). Relatively small volumes of water are supplied from groundwater and small dams on tributary rivers. The bulk of the Brandvlei Dam water is either released into the Breede River channel from where it is diverted into canals or pumped, or released directly into the Le Chasseur canal for downstream distribution. The main channel of the Breede River, however, also acts as a drain for saline irrigation return flows, the effect of which requires amelioration in the form of freshening releases from Brandvlei Dam.

During the 1980s irrigated areas were expanding steadily, albeit at a relatively low rate. The political and constitutional reforms set in motion in the late 1980s created expectations of increased export opportunities for products from this area, leading to a distinct likelihood of increased irrigation expansion in the 1990s. Concerns arose that the concomitant increases in future saline return flows would render certain stretches of the Breede main channel unuseable as a supply conduit.

Against this background the then Department of Water Affairs (now known as the Department of Water Affairs and Forestry (DWAF)) appointed Ninham Shand in 1988 to develop and apply a computer model, capable of predicting the impact of irrigation development in the GBDGWS supply area on river flow and salinity. During 1990, the DISA (Daily Irrigation and Salinity Analysis) model was completed and implemented to examine the potential impacts of a number of planning scenarios for the GBDGWS supply area. Aspects addressed included future salinity and flow patterns in the Breede River, as well as requirements for future freshening releases from BrandvIei Dam.

In 1987, an intensive five year hydro-salinity field research programme to support the development of the DISA model was initiated and conducted by the Hydrological Research Institute (HRI) in the Breede River Valley. This research was based on an intensive monitoring programme at measurement points in the main river, tributaries, canals, deep and shallow boreholes, piesometers in alluvial soils, tensiometer fields and rainfall stations. By 1990, Government spending cuts and subsequent Departmental budgetary constraints necessitated a curtailment of some longer term projects, which included a premature end to the HRI's Breede River Salination Research Programme. The time table for this termination required the conclusion of all Breede River field studies, including intensive monitoring, by the end of the 1989/1990 irrigation season. This timetable meant that certain ongoing aspects related to the model development had to be regarded as provisional, as continued field research and monitoring was required for further model verification.

To ensure that the full potential of the Breede River Salination Research Programme be realised, an agreement was subsequently reached whereby the Water Research Commission would provide funding to ensure that certain actions relating to the original project specifications could be finalised. These included a further year of intensive monitoring, refinements to certain aspects of the DISA model, and effective technology transfer. This report describes the refined version of the DISA model, including model verification and sensitivity analyses. It also serves to upgrade the existing documentation, which is in the form of consulting reports to the DWAF, to a research oriented communication.

PROJECT AIMS

In January 1991, an agreement was reached whereby the Water Research Commission would fund a project to complete research related to the DISA model. The aims of this project, which culminates with the publishing of this report, were to:

1. Implement refinements to the DISA model, comprising the following:
   • Review of the representation of three important processes in DISA
     • the fate of irrigation percolate from the higher-lying terrace soils and the influence on this of artificial drainage.
     • the fate of canal leakages. - bedrock contributions to river and alluvial aquifer salinities.
     • active soil volumes associated with different irrigation techniques and soil types.
   • Incorporation of findings of various relevant field studies in the Breede River Valley funded by the Hydrological Research Institute (HRI) and the Water Research Commission (WRC).
   • Further verification of DISA using the most recent hydro-meteorological and hydro-salinity observations in the Breede River system.
2. Achieve effective technology transfer of the software and the supporting research findings, including upgrading the existing model documentation and the user manual.
3. Complete the curtailed hydro-salinity monitoring programme with observations during the period 1 January - 30 April 1991.

DEVELOPMENT OF THE DISA MODEL

Model characteristics

The following comments summarize the modelling philosophy which governed the development of the DISA model:

• A non-calibration approach was adhered to-configuration of the system and all physical and salinity parameters were based on the best information available from field work, maps, Landsat data, surveys and relevant theory.
• DISA allows the catchment to be divided into a number of homogeneous physiographic units, each of which comprises a part of the catchment with relatively homogeneous agricultural and hydrolological features.
• Each of the physiographic units is further divided into areas termed "return flow cells", on the basis of soil type. A distinction is made between terrace and alluvial return flow cells. Terrace cells occur in the higher-lying areas of the physiographical units, while alluvial cells occur in the lower-lying areas, particularly in the vicinity of river channels.
• DISA configures the physical water distribution system accurately at any desired scale, including reservoir releases, irrigated physiographic units, river diversions, canals, pump schemes, farm dams and rejects.
• DISA is structured to preserve daily water and salt (as Total Dissolved Salts - TDS) balances in all irrigated soil profiles as well as in all canal and river channel components in the system during the irrigation season only.
• Soil moisture budgeting occurs via layer capacity limits, not via mechanistic soil moisture flow equations.
• DISA allows routing of seepage from canals and farm dams to appropriate receiving points.
• DISA distinguishes between irrigated alluvial soils and higher-lying irrigated (terrace) soils and models the groundwater levels in the former dynamically, with recognition of the capillary fringe as a redistributor of water and salts between saturated and unsaturated soil layers.
• DISA treats non-alluvial groundwater as a regional influence and requires empirical estimates of groundwater flows and salinities as inputs to the surface water flow system.
• DISA does not account for dynamic salinity-related soil processes such as weathering, sorption, dissolution or deposition, because suitable quantification of these processes can currently only be achieved by calibration of black-box equations, which is contrary to the approach aimed for in this study. (During the final stage of model development, an attempt was made towards the development of a salt generation function for incorporation into the model. However, it was concluded that further research is still needed to successfully implement this concept.)
• DISA does not model rainfall-runoff processes in the catchment regions upland of the irrigated areas - the configuration is such that observed (or simulated by catchment modelling) flows and salt loads can be used to bring upland surface hydrosalinity influences to bear.
• DISA does not keep account of gypsum applications. In the Breede River system for example, data concerning this practice is almost non-existent, but it is surmised that the annual gypsum application is much less than 10 % of the total annual salt load applied in the form of irrigation water.

Model structure

The DISA model is imbedded in a user-friendly software environment, and was developed with the Turbo Pascal programming language. The software consists of four components:

• The Database Manager, which controls and manages the system database containing all the time series data that serve as external input to the catchment area being modelled.
• The System Configurator is used to configure the physical system in terms of modelling node types, node sequencing, process parameters and global parameters. It also controls the selection of nodes for producing graphical and numerical output.
• The System Model uses the time series database (in the Database Manager) and the system configuration file (in the System Configurator) to perform hydrosalinity simulations of all flow and salt transport through the conduits and soils of the irrigation system. Simulations are based on a daily time step.
• The Output Manager produces graphical output in the form of daily flow and salinity time series, and percentile (exceedance) curves, as well as numeric output files.

Data requirements

It can be appreciated that the modelling approach implemented by DISA, which is both physically based and highly organised in terms of the physical structure of the modelled area, is extremely data demanding. The data input to the model consists of two types, viz. time series data and physical data parameters.

Time series data serve as external input to the modelled catchment area. These include daily rainfall and evaporation records within the catchment, as well as daily flow (m³/s) and salinity (mg/1) records in rivers and tributaries entering and leaving the catchment

Every return flow cell as well as every sub-model in the water distribution network has to be defined in the model in terms of parameters or physical process constants. To obtain these parameters, a detailed knowledge of the system is required in terms of layout, land-use, soil characteristics, river channel, canal and farm dam geometry and abstraction rules implemented by farmers.

MODEL VERIFICATION

The final stage of model development involved the verification of the model to demonstrate the applicability and accuracy thereof. The model was verified on the Breede River system as well as the Vaalharts Irrigation system.

The Breede River System

Model verification for the Breede River system was completed in two phases, namely an internal verification and an external verification. Three database years were used, viz. 1985/86, 1986/87 and 1987/88. Verification of the model was extended to cover two additional years, namely 1988/89 and 1989/90. Due to gaps in either inflowing datasets or downstream datasets, verification on these two years was only partially successful. The internal verification was done by evaluating alluvial aquifer behaviour, with the groundwater levels in the alluvial return flow cells used as an indicator of "acceptable" behaviour. Groundwater levels in the aquifers behaved as expected, with simulated water levels fluctuating between approximately 300 mm below the surface and 500 mm above the bottom of the lowest layer. Salinity concentrations in a few aquifers were not in equilibrium. In line with the "no calibration" approach, it was decided that overall model performance was not adversely affected by this and that no rectifying measures were required.

External verification was evaluated by comparison of observed and modelled river flow and salinity time series at two different flow gauging stations. It was found that seasonal trends and magnitudes of river flow were simulated satisfactorily.

The Vaalharts Irrigation System

During 1995 the DISA model was applied to the Vaalharts Irrigation system. This was done to evaluate the applicability of DISA to an irrigation system situated in a summer rainfall region, as well as to evaluate the performance of various additional processes incorporated into DISA, e.g. surface runoff, artificial drainage and deep percolation.

Only an external verification was completed. A period of five seasons between October 1988 and April 1991 was considered and included three summer and two winter seasons. Observed and modelled river flow and salinity time series at a weir situated at the downstream end of the system were compared.

Overall, the time series and percentile curves indicated that during certain months, DISA underestimated the salt concentrations, while simultaneously overestimating flow. This was attributed to the fact that the simulated volume of tailwater (water which is not abstracted from the canal system for irrigation) is too high. As this water is relatively fresh, it leads to low TDS concentrations and an accompanying overestimation of river flow. This was caused by incorrect monthly abstraction volumes for irrigation. The situation was remedied by modification of the monthly distribution of irrigation supply, which led to a much better fit between observed and simulated flows and especially salinities.

This application of the DISA model confirmed that about 80% of the TDS load in the incoming irrigation water is retained in deep groundwater bodies underneath the Vaalharts Scheme.

SENSITIVITY ANALYSES

The DISA model was subjected to two series of sensitivity analyses performed during different stages of the model development. Five model components for which inadequate or no field data existed, were identified during the verification phase of model development. The need for sensitivity testing of a further five model components came to the fore during model application for a planning study of the Breede River System. The full set of components which were tested are:

1. The distribution of canal seepage between irrigated and non-irrigated alluvial soils.
2. The portion of groundwater outflow from terrace cells which is drained either by artificial means (surface ditches or pipes) or by natural collectors.
3. Capillary fringe depths associated with the various soil texture classes.
4. The Dry Evaporation Factor which expresses evaporation from bare soil as a fraction of A-Pan evaporation.
5. Evaporation depths associated with the various soil texture classes.
6. The reduction of canal seepage due to riparian vegetation growing along the canal system.
7. The sensitivity of model response to a change in soil texture classes of return flow cells.
8. The rate of aquifer recharge to the main river channel.
9. The sensitivity of model response to a change in estimated soil profile depths.
10. The performance of the model when model runs are extended over more than one irrigation season.

The configuration of the Breede River system was used to perform the sensitivity analyses. In order to assess model response, the return flow sub-model behaviour (internal) and model response at the downstream end of the modelled area (external) were monitored. Internal model response was evaluated by monitoring changes in the groundwater levels of four representative alluvial return flow cells, while external model behaviour was assessed by comparison of modelled outflows and associated salinities with observed values at a downstream weir.

Several scenarios were investigated for each component under consideration and where necessary, refinements were made to the relevant process variables.

CONCLUSIONS AND RECOMMENDATIONS

The DISA model provides a useful tool for anticipating the effects of various irrigation and operational planning scenarios on river salinity. The principal conclusions that were reached during model development and subsequent verification are:

• The modelling approach, whereby the catchment is discretized into a number of homogeneous units and return flow cells, which is overlain by a water distribution network, allows for a relatively simple configuration of all the primary components of an irrigation system.
• Due to the no-calibration approach which was adhered to, the DISA model requires detailed and extensive input data. During model verification it was evident that a lack of detailed data regarding certain aspects of system configuration, leads to inaccuracies in the model output. Critical data requirements include the definition of crop types, irrigated areas and irrigation volumes applied.
• The process routines which simulate the delivery of water and salts to the irrigated areas as well as root zone and delivery zone processes within the irrigated soil profile, provide a realistic representation of actual water and salt movement. This was confirmed during model verification for both the Breede and Vaalharts Irrigation systems, where seasonal trends in river flow and salinity were accurately predicted.
• Although an attempt was made to develop a salt generation function to account for the introduction of salt through the weathering of parent material, further refinement is required to simulate dynamic salinity-related soil processes more appropriately.

The extended surface and groundwater monitoring exercise, undertaken by MBB Consulting Engineers on behalf of Ninham Shand, and taken over from the Department of Water Affairs and Forestry's Hydrological Research Institute (HRI) in May 1990, was completed in April 1991.

The original dataset for 1990/91 contained a number of problems, the most notable of which was a downward drift in electrical conductivity (EC) readings, due to progressive under-registration of the reading instrument. Based on laboratory tests of the EC probes used in the monitoring, conducted by Prof H Moolman of the University of Stellenbosch, the original datasets were adjusted to compensate for the instrument error. Adjusted datasets were handed over to the HRI for archiving.

During verification of the DISA model, it became clear that further improvements could be made to either facilitate its application or improve its accuracy. These improvements include :

• Modifications to the system configurator to allow for the seamless simulation of successive irrigation seasons.
• Development of a salt generation function, which will allow the accurate simulation of dynamic salinity-related soil processes such as weathering, dissolution or deposition and as such, replenish the salt reservoir in the ground water body.
• Allowing for seepage in any direction between adjacent alluvial return flow cells, based on the difference in groundwater levels. (Presently, seepage is always in the direction of the river, irrespective of groundwater levels.)
• The development of a GIS-based interface for the model to allow the spatial input data to be easily manipulated and verified. (This aspect was addressed during a project by the Institute for Geographical Analysis at the University of Stellenbosch (Wolff-Piggot, 1994)).