• Tidak ada hasil yang ditemukan

DECLARATION 2: PUBLICATIONS

5. PERFORMANCE AND SENSITIVITY ANALYSIS OF THE SCS-BASED PEAK

5.5 Conclusions and Recommendations

within 1 hour and therefore it was a very intense event, and a value close to zero indicates that the rainfall was more uniformly distributed throughout the day, i.e. low intensity. There is some scatter around the relationship, which may be attributed, but not limited, to antecedent soil water conditions, however, there is a clear inverse relationship between rainfall intensity and lag time.

If a methodology to account for rainfall intensity on a day-to-day basis is developed and included within the ACRU model, relationships such as these may be useful to adjust estimated lag times based on the rainfall intensity. This is important since there is a relationship between the two and they both influence the simulation of the stormflow contribution to peak discharge.

Furthermore, it provides an objective approach to account for the variability in lag time from event-to-event.

Figure 5.8 Relationship between catchment lag time and rainfall intensity

The lack of reliable sub-daily rainfall data, particularly consistent and accurate short-duration rainfall data, was a significant challenge to this study. This resulted in the use of only two pilot study catchments. These catchments were selected since they were identified to have high quality data, with the short-duration rainfall stations being highly representative of the catchments. The analysis of these two catchments, however, produced consistent trends and successfully addressed the objectives of the study to: (i) investigate the simulation of the stormflow contribution to peak discharge in detail for two case study research catchments, (ii) compare the results obtained from application of the single UH approach and the incremental UH approach, (iii) compare the simulated results when estimated parameter inputs are replaced with observed data, and (iv) investigate if there is a relationship between the distribution of daily rainfall, i.e. rainfall intensity, and catchment lag time. Through these objectives the overall aim was achieved, i.e. to guide further research and identify priority components that have the most significant influence on the stormflow peak discharge computation, as summarised below.

The following conclusions based on the analysis of the results in this chapter have been drawn:

(i) Both the single and incremental UH approaches are sensitive to stormflow volume, and although the UQFLOW OTD is a reasonable estimate of the daily stormflow volume, it still tends to overestimate stormflow in general.

(ii) The single UH approach, which does not account for the distribution of daily rainfall, was particularly sensitive to the estimated lag time, which varies significantly from event to event.

(iii) The incremental UH approach is sensitive to both the estimated lag times and daily rainfall distributions used, which both vary significantly from event-to-event. Based on the results obtained for the two case study catchments, however, the incremental UH approach was identified to be more sensitive to the distribution of daily rainfall used.

(iv) When applying the incremental UH approach, and both the daily rainfall distribution and catchment lag time are incorrectly estimated, a compounding of the error obtained is observed.

(v) The Schmidt and Schulze (1984) lag equation was identified to provide a relatively good estimate of the average catchment response time, and although less satisfactory, the synthetic daily rainfall distributions provided a reasonable average representation of the typical rainfall distributions observed in the catchments.

(vi) The incremental UH approach provides more accurate peak discharge estimates compared to the single UH approach, i.e. both when using parameters obtained from observed events and when using estimated and synthetic information. The results are, however, much improved when using parameters derived from the observed data. This indicates the importance of accounting for the variation of daily rainfall distributions and catchment lag times on a day-to-day basis. Therefore, to improve on the results obtained from the incremental UH approach, methods to account for these variations need to be developed.

(vii) There is a relationship between catchment lag time and rainfall intensity.

Consequently, if regional relationships between rainfall intensity and lag time can be developed, adjustments to lag time estimates, such as using the Schmidt and Schulze (1984) estimate, may be made based on the rainfall intensity of the event for a specific day.

(viii) Lastly, the results highlight that accurate simulations of peak discharge may be obtained when applying both the single and incremental UH approaches when accurate inputs to the equations are used, therefore, validating that the model concepts and structure are reasonable to use in practice.

Based on these results the following recommendations are made for future research:

(i) To confirm that the incremental UH approach consistently produces superior results to the single UH approach, as identified in this chapter, i.e. the performance of the single and incremental UH approaches need to be assessed for all verification catchments used in the assessment of the CSM system developed in Chapter 4.

(ii) There is also a need to perform several additional sensitivity analyses on the CSM system developed, including the performance of the CSM system when only default datasets suggested to estimate soils and land cover information are used. In addition, the sensitivity of the approach to different lag time estimates, i.e. used to simulate the stormflow contribution to peak discharge, needs to be assessed.

(iii) Owing to the greater impact on the incremental UH approach to the sub-daily temporal distribution of daily rainfall identified in this chapter, as well as the relationship identified between the daily rainfall distribution and lag time, it is recommended that methods to account for the actual distribution of daily rainfall on a day-to-day basis be

prioritised in future research. This information may then be used to further improve the estimation of lag time and peak discharge on a day-to-day basis.

(iv) Linked to the previous point, further investigation of the links between rainfall intensity and catchment lag time is recommended, with the possibility of developing regionalised relationships for South Africa.

(v) Another aspect to consider, which was not applicable in this chapter, since the catchments were very small (approximately 1 km2), with rain gauges located within the catchments, is the spatial distribution of rainfall. As catchment size increases the distribution of rainfall over the catchment is non-uniform and varies from event-to- event. Therefore, it is recommended that methods to account for the spatial distribution of rainfall be investigated. It is also hypothesised that lag time may change as a function of the spatial distribution of rainfall, and therefore these considerations should also be included in further research.

Chapter 6 addresses Recommendations (i) and (ii) made above.

6. IMPACT OF MODEL CONFIGURATION AND PARAMETER