APPLICATION: FLOOD FORECASTING MODEL FOR LOIRE RIVER (FRANCE) - Modelling Uncertainty in Flood Forecasting Systems - page 125

Figure 6.10. Standard deviations of forecast water levels given by Monte Carlo
(MC) method for uniform, triangular and normal PDFs.
Sensitivity of the FOSM method to perturbation ratio
It is important to note that the results obtained from both the FOSM and IFOSM methods
vary with the size of the perturbation (because the approximate function of the real
function depends on the size of the perturbation). A measure of the perturbation is the
perturbation ratio (PR), that is, the ratio of the actual perturbation to the standard
deviation of the parameter. The values shown in Figures 6.9(a) and (b) are the best
results on average for all forecast horizons, when taking the same PR for all forecast
horizons. The results can be further improved if different PRs are used for different
forecast horizons, but this yields an undesirable increase in computational cost.
Figure 6.11. Standard deviation vs perturbation ratio for FOSM method for
forecast water levels at 12 and 24 h.
Figures 6.11 and 6.12(a, b) show the variation of computed standard deviations on
12-and 24-h forecast water levels with respect to the PR applied for the FOSM and