APPLICATION: FLOOD FORECASTING MODEL FOR KLODZKO CATCHMENT (POLAND) - Modelling Uncertainty in Flood Forecasting Systems - page 103

5.4.2 Results with 3 and 6 subperiods
A key issue in the application of this methodology is the selection of an appropriate
number of subperiods. Increasing the number of subperiods increases the number of
parameters for the search space and consequently the computational requirement. On the
other hand, intuitively, increasing the number of subperiods may widen the uncertainty
bounds in the output. To verify this, the results with 3 subperiods are compared with the
results with 6 subperiods. With 6 subperiods, given 9 subbasins, the number of
parameters increases from 18 (with 3 subperiods) to 45. To reduce the number of
parameters it is assumed that the temporal pattern is the same for all subbasins. This
assumption significantly reduces the number of parameters from 45 to 5 in the case of 6
subperiods and 18 to 2 in the case of 3 subperiods. It is however important to check the
effect of this assumption. Therefore, the 3 subperiods case is repeated applying this
assumption. The comparison of this result with the previous result for varying temporal
patterns over subbasins is presented in Figure 5.14. The differences are minor except for
forecasts at 84 and 87 hours. It is however important to notice that, in this particular
example, the same temporal patterns over the subbasins resulted in larger uncertainty
bounds than with varying patterns. The same temporal pattern is a special case of the
varying temporal pattern. Therefore, the uncertainty in the case of a varying pattern
should be larger than or at least equal to the case with same temporal pattern. This
indicates a problem in the scheme used for finding the minima and maxima. For a GA
scheme, given that other parameters are chosen appropriately, the solution of this
problem normally requires more function evaluations. In this example, the maximum
number of function evaluations is limited to 2000, which appeared not to be enough for
the varying pattern.
Figure 5.14. Uncertainty range in forecasted discharges (Q UB -Q LB ) with
reconstructed precipitation into 3 subperiods for: (a) ! -cut level 0, and
(b) ! -cut level 1.