DECLARATION 2: PUBLICATIONS
8. DISCUSSION, CONCLUSIONS AND RECOMMENDATIONS
8.8 Recommendations for Future Research
Based on the research gaps identified, investigations performed and results obtained in this research, through the various chapters, the recommendations for future research are summarised as follows:
(i) To further develop the CSM system defined and assessed in this research, and to establish a comprehensive useable product for practitioners to apply, the following recommendations identified from the literature review (Table 2.1 – Chapter 2) still need to be addressed: further development, assessment and inclusion of national stochastic rainfall generation and/or disaggregation techniques; compilation of the CSM system and additional developments into a user-friendly, simple, software tool that is attractive to consultants and government organisations (e.g. DWS, SANRAL), and provision of training courses, workshops and user manuals related to the software;
as well as continual updating, refinement and improvement of the approach including, for example, flood routing routines and flood forecasting. In addition, refinements and improvements to the final ACRU land cover classification should be considered in future research, particularly with regards to possibly explicitly representing the three forestry genomes typically cultivated in South Africa.
(ii) Difficulties associated with obtaining observed data for research catchments and the poor quality of climate and hydrological data in South Africa are highlighted in this research. Therefore there is an urgent need to collate, error check and standardise climate and hydrological data from various sources into a single and easily obtainable national database. If this is not performed timeously, this valuable data from research catchments that is irreplaceable will be lost.
(iii) Based on the results obtained in this research regarding the simulation of peak discharges it was identified that the Schmidt and Schulze (1984) estimated catchment lag time and synthetic rainfall distributions developed by Weddepohl (1988) are reasonable average estimates of these parameters required to estimate daily peak
discharges. It was noted, however, that these parameters vary signficantly from day- to-day and the peak discharge computation is sensitive to both these inputs.
Consequently, it is strongly recommended that methods to better estimate the distribution of daily rainfall and catchment lag time on a day-to-day basis be investigated and/or developed. It is also recommended that priority be given to the development of a methodology to more adequately estimate the distribution of daily rainfall, due to the sensitivity of the incremental UH approach to this input and relationships identified between rainfall intensity and catchment lag time, i.e.
suggesting that lag time may be adjusted based on rainfall intensity.
(iv) Based on the sensitivity of the CSM system to land cover and soils information, it is recommended that verification and/or recalibration of the CNs and associated ACRU QFRESP and SMDDEP parameter values for the land cover and soils combinations, listed in the updated SCS-SA and final ACRU land cover classifications, be considered in future research and in further development of the approach. Furthermore, an assessment of the impact of catchment area and slope on the parameterisation of the QFRESP parameter in particular should also be considered in future research. In addition, consideration of including the detailed land use management scenarios provided in the MUSLE Handbook should be considered in future research. Another possible consideration for future research is to re-look the SCS equations from 1st principles and develop improved equations. This, however, would likely be a significant undertaking and sufficient observed data would need to be sourced, if available, to validate and verify the approach.
(v) Linked to the previous point on the sensitivity of the CSM system to land cover and soils information, it is recommended that further refinement and improvement of default estimates of land cover and soils information be considered in future research, including further refinement of the SCS-SA soil group map developed by Schulze and Schütte (2018).
(vi) As already highlighted above, the Schmidt and Schulze (1984) estimated catchment lag time was identified to be a reasonable estimate, and superior to the SCS lag equation. The Ī30 parameter used in the Schmidt and Schulze (1984) lag equation in this research was obtained using the 2-year return period maximum 1-day rainfall calculated from the daily rainfall files used as input to the ACRU model, and applying a multiplication factor defined for each specific region. In future research a
comparison and assessment of the impact that different Ī30 estimates have on the Schmidt and Schulze (1984) estimated lag time and the consequent impacts on the simulated peak discharges is recommended, this may include Ī30 estimates derived from the gridded RLMA&SI values (Smithers and Schulze, 2000).
(vii) In terms of the SCS-SA MCM and JAM the CN adjustment is based on the limited rainfall data, spatial coverage, land cover and soils combinations, and ACRU modelling capabilities available during the 1980’s when these methods were developed (Schmidt and Schulze, 1987a). This point is particularly relevant to the SCS-SA JAM where a frequency analysis was conducted on simulated flows and consequently the method is completely dependent on the rainfall data available and used at the time. Consequently, the method was not recommended for estimating design stormflow beyond the 20-year return period. Therefore, an additional recommendation is to use the results from, and methodology applied in, this research to comprehensively update the SCS-SA MCM and JAM. In terms of the JAM this involves running the ACRU model with updated rainfall and climate data, i.e. with the extended records currently available, and land cover and soils combinations, and performing frequency analyses or alternatively extreme value analyses on the simulated flows. This will provide updated design stormflow and peak discharge values for defined homogeneous response zones, i.e. either the quaternary or quinary catchments. While performing these simulations, additional information such as simulated daily soil water deficits, i.e. provided as an optional output in the ACRU model, may be used to update the MCM method. This will involve performing a frequency analysis on the simulated daily soil water deficits, and using the 50th percentile soil water deficit to adjust the original average catchment CN (CN-II).
Additional experimentation may also be performed, e.g. using different soil water deficit percentiles for different return periods.
The recommendations for future research, in conjunction with the CSM system developed and proposed in this research, may be used to further develop a comprehensive useable CSM system for DFE in South Africa. A baseline comprehensive approach, however, has been defined, verified and proposed which may be applied for DFE in small catchments in South Africa.