5. Hydrological Modelling of the Densu Basin
5.2 Sensitivity Analysis
5.2.4 Results and Conclusion
As a rule, high sensitivity of a parameter is only desirable if that parameter can be measured easily and has a very small range. Schultz (1995) suggested the following scale for parameter sensitivity in ACRU:
• extremely sensitive (E): the percentage change in the output (∆O%) is more than twice that of the input parameter being tested (∆I%), i.e ∆O% > 2(∆I%)
• highly sensitive (H): the output change is more than the input change, but by less than 200% , i.e. 2(∆I%) > ∆O% > ∆I%
• moderately sensitive (M): relative output changes less than the relative input change, but by more than 50% of the input change, i.e. ∆I% > ∆O% > 0.5(∆I%)
• slightly sensitive (S): output changes by between 10% and 50% of the input change, i.e. 0.5(∆I%) > ∆O% > 0.1(∆I%)
• insensitive (I): output changes by less than the 10% of the input change, i.e. ∆O% ≤ 0.1(∆I%).
Considering Figures 5.1-5.5 according to the above scale, it is apparent that for cumulative streamflow and cumulative baseflow, apart from FC1 and CAY which showed moderate sensitivity, the remaining parameters were either slightly sensitive or insensitive, although PO1 is moderately sensitive over some of its range. The 38% increase in rainfall amount gave a corresponding increase in simulated cumulative streamflow of 109.6% which is over 200% that of the input parameter and is therefore extremely sensitive. This is also corroborated by Angus (1989). In Table 5.2 the Densu Basin sensitivities are compared to Angus (1989) who also tested the sensitivity levels of ACRU model parameters on cumulative flows.
Table 5.2: Comparison of Densu basin sensitivities to Angus (1989)
From the Table of sensitivities it is apparent that the sensitivities seen in the Densu are lower than those of Angus (1989 in Schulze, 1995). It should be noted here that Angus (1989, in Schulze, 1995) undertook this sensitivity analysis at only one location i.e. Cedera in the KwaZulu Natal Midlands and further emphasised that because it was undertaken at only one location, the result can therefore not be used to represent the response of the output in ACRU in other regions. Some suggested possible reasons for the differences in sensitivities at the two locations could be the following:
Sensitivity of Parameter (Angus, 1989)
Sensitivity in the Densu Basin Parameter
Reduced Increased Reduced Increased FC1 DEPAHO, DEPBHO ABRESP BFRESP CAY ROOTA VEGINT SMDDEP COIAM - H H S S H S S H M - S S S S H S S S S M I I I I M S S I I S I I I I S I I I I
1. That the two sensitivities have been conducted at two different climatic regions i.e Kwazulu Natal in the semi arid climatic region as against the Densu catchment in the tropical region.
2. DEPAHO is a soil parameter given by the depth of the A horizon of soils. Further, the SMDDEP (effective depth of the soil from which stormflow generation takes place) parameter is known to be very shallow (0.1m) in semi arid regions (Schulze 1995) which are characterised by high intensity convective rainfall, so that the response to stormflow could be high (Schulze, 1995). In the ACRU model the default value of SMDDEP is taken to be the depth of DEPAHO if it is not known, since it is always ≤ DEPAHO (Schulze, 1995), meaning that the smaller SMDDEP the higher the stormflow and vice versa. Since the DEPAHO of the soils of the Kwazulu Natal province in the semi arid zone as indicated by Shulze (1995) are very shallow (0.1m) and that of the Densu catchment are around 0.3m, it is apparent that the smaller DEPAHO and SMDDEP requires less rainfall to bring the soil up to field capacity so as to generate stormflows and hence be more sensitive to the soils of semi arid regions than that of the tropical Densu catchment.
3. Comparing the vegetative covers of the Kwazulu Natal province with that of the Densu catchment, the vegetation (land cover) of the semi arid Kwazulu Natal is sparser than the tropical Densu which is nearly covered throughout the year. In its simulations ACRU is sensitive to seasonal above and below-ground vegetation/land cover changes (Schulze, 1995). CAY which is a vegetation dependent parameter is more sensitive when CAY is decreased than when it is increased in the Densu (Figure 5.5). In the ACRU model typical values of the crop coefficient (CAY) have to be specified month-by-month. If there is more than one homogeneous land cover/use zone specified for the catchment, then for each "zone" the percentage area covered by the zone is also given and the land use utility facility in the model will compute an area-weighted monthly CAY. The Densu weighted CAY’s were found to be higher than the Kwazulu Natal monthly values. As CAY’s have a direct relationship with the stage
of growth of crops and evapotranspiration, it implies that bigger CAY’s would result in higher potential evapotranspiration which may impact on the volume of streamflows depending on soil water availability. Thus with monthly CAY’s lower in values in the Kwazulu Natal province than in the tropical Densu catchment it is likely that the impact of decreased CAY’s on streamflows would be higher (meaning more is generated because there is less potential evapotranspiration) than that of the Densu Basin.
When the relative sensitivities of FC1 and CAY when using the Apan and
Hargreaves methods are compared in relation to cumulative streamflow, the level of sensitivities observed from use of the Hargreaves method for evapotranspiration tended to show higher sensitiveness to the input parameters than the Apan method of evapotranspiration. These slight changes in gradients were expected.
To conclude, the sensitivity analysis has demonstrated how cumulative total streamflow and baseflow respond to changes in ACRU parameters within the Densu Basin. This understanding is thus carried forward to inform the calibration where the most sensitive parameters in this basin notably CAY, VEGINT, ROOTA (the landuse parameters) and FC1 will be calibrated. COFRU
showed no influence in cumulative simulated streamflows for the sensitivity analysis however it was noted that it affected the shape of the hydrograph (recession section) when it is varied and calls for more examination during calibration.