INITIAL CONFIGURATION OF THE PITMAN MODEL FOR THE KAFUE

BASIN

For the most of the Kafue basin (see Figure 6.1), there are no special model configuration requirements. One minor exception to this is the need for a time series of transfer inflows to represent the mine waste pumping within sub-catchment C. Two major exceptions are the configurations required for the Lukanga swamps (from the outlet of J down to Q) and the sub-catchments and reservoirs from sub-catchment S downstream (which includes the Kafue flats).

6.1.1 Lukanga Swamps

The physical setting of the Lukanga swamps is described in section 5.6.1 and can be seen in Figure 5.6, while flow anomalies can be seen from Figure 5.28. Comparison of the observed monthly flows for the outlets of sub-catchments J and L suggests that flow to the swamps occurs in months with flow volumes greater than 800 Mm3. There appears to be very little flow out of the swamps back into the Kafue channel and this only occurs during wet years.

There are no model components that have been designed to cater for this type of process, although channel losses have been frequently simulated using ‘dummy’ reservoirs to represent either evaporative or channel bed seepage losses. The same approach has been used here to represent the storage and losses from the incremental catchment L. A reservoir has been established in the model for catchment L (Figure 6.2, Dam B, maximum volume = 7400 Mm3) with a large surface area to represent the swamp, through which all the runoff from catchment L is routed. However, the original spatial configuration of the model had flows from J being routed through L to Q. Establishing a reservoir at the outlet of catchment L would suggest that all the upstream flow of the Kafue would have to pass through the reservoir representing the swamps. To prevent this and to represent reality more accurately,

outflows from J have been re-routed into Q (bypassing L). In addition, the area of Q has been increased and L decreased, so that the part of L that does not contribute to the swamps bypasses the ‘dummy’ reservoir (Dam B).

Figure 6.1 Kafue basin showing the sub-catchments used for model calibration.

This type of approach cannot be used for the spillage from the main Kafue channel into the swamps and therefore a more complex and multi-step approach had to be adopted. The first step was to calibrate the outflows from J without a dam against the observed data at that point and then edit the simulated flows at the outlet of J, decreasing all the monthly flows with volumes greater than 800 Mm3 to a value of 800 Mm3. A relatively small ‘dummy’ reservoir (1 000 Mm3:Dam A in Figure 3.11) was then established at the outlet of J and the edited time series generated by the first step used as a high priority downstream flow requirement on this

dam. The surface area to volume ratio of the reservoir was set to a low value to minimise simulated evaporative losses, such that the downstream requirement would always be met. An ‘abstraction’ demand (1 000 Mm3) was then established to represent the outflow from the channel into the swamps. The ‘operating rules’ were set such that the reservoir would have to be 98% full for the complete demand to be met, while progressively lower demands were established for 90%, 80%, 60% and 40%. The effect of this conceptual design is that during inflows of less than 800 Mm3 the downstream requirement is the same as the inflows, the reservoir remains relatively empty and there are no abstractions. During higher flows the reservoir starts to fill and the abstractions increase. The objective of the calibration exercise was to set the different demand levels to values that gave a satisfactory pattern of spill volume. These are then added to the downstream requirement plus any outflows from L and Q to generate an acceptable time series of flows at the outlet of Q (calibrated against observed flows at this point). The calibration exercise involved two iterative steps, as the ‘abstractions’ from the reservoir at the outlet of J become transfer inflows into L. The modelling process is illustrated in Figure 6.2

The calibration process effectively prevents the observed flows at the outlet of L (which is on the main Kafue channel) from being used, as in the model these are only outflows from the swamps. However, after an examination of the observed flows at the outlets of J, L and Q, it appears that J and Q are more consistent with each other, while L appears to be somewhat anomalous. While there is no certainty that the model configuration is truly representative of the real processes, acceptable results have been obtained with parameter values that are consistent with calibrations for other parts of the Kafue catchment.

6.1.2 Itezhi-tezhi dam, Kafue Flats and Kafue Gorge dam

The configuration of the two dams is relatively straightforward. Itezhi-tezhi dam has been established at the outlet of sub-catchment S with a capacity of 4 925 Mm3 and an area of 475 km2 at full supply. The latter is higher than the stated area, as it was found that the evaporation losses appeared to be too low. The release estimates given in Table 5.3 (an annual total of about 7 100 Mm3) did not seem to match the pattern of observed outflows and were reduced to give an annual total of 4 925 Mm3. These were then reduced at dam volumes of 70% of capacity and lower.

Figure 6.2 Diagramatic representation of the model setup and calibration steps for sub- catchments J, L and Q.

Kafue Gorge dam has been established at the outlet of U with a capacity of 900 Mm3 and an area of 1 165 km2 at full supply. This conforms with the known facts about the reservoir. The area of sub-catchment U was set at 15 000 km2 (and the area of T reduced by the same amount) to represent that part of T that lies downstream of the main floodplain. This also allowed a separate ‘dummy’ dam to be placed at the outlet of T to represent the Kafue flats floodplain storage. The maximum release (at full supply) was set to 6 000 Mm3 y -1 (or 190 m3 s-1) and reduced when the dam reached 50% of capacity and lower.

The ‘dummy’ dam representing the Kafue flats has been established with a capacity of 6 500 Mm3 and an area of 3 250 km2 at full supply, decreasing linearly with volume. As this ‘dam’ is required to have continuous outflow and yet its volume and area should fluctuate seasonally, it is not possible to simulate downstream flows as spillage. It was therefore necessary to assess (using the observed outflows from the Kafue Gorge dam during wet years) the likely outflows from the floodplain in advance of modelling and set these as

Dam A

Step 1: Simulate flows from J

Dam B

J

L

Step 2: Edit Simulated flows from J to exclude peak flows greater than 800 Mm3 and set as downstream flow requirements for Dam A.

Q

Step 3: Calibrate the abstractions (inflows to swamp) from Dam A to approximately achieve the required flows at the outlet of catchment Q.

Step 4: Convert the abstractions from Dam A to upstream inflows into L and repeat steps 3 and 4 until the outflows from Q are a satisfactory representation of the observed flows.

‘releases’ for different storages in the ‘dam’. Table 6.1 lists the levels at which the different releases apply and it was found to be sufficient to establish the same ‘release’ volume for all months. This is because the reduced inflows during the dry season ensure that the ‘dam’ volume decreases and therefore the outflows decrease.

Table 6.1 Release operating rules for the dummy dam representing the Kafue flats floodplain.

Volume (% Capacity) > 90 89-75 74-50 49-25 24-10 <10 Release (Mm3 mnth-1) 2500 1995 1468 1012 744 506

Assessing the usefulness of these unusual approaches to simulating both the Kafue flats and the Lukanga swamps, without modifying the structure of the model, forms part of the assessment of its applicability to the Kafue basin. This issue will be discussed in more detail after the final calibration results are presented.