4.1.1. Synthetic Flood Events

The ANN model is tested with the 60 synthetic flood events from the FloodEvac tool. The ANN model for the prediction of first intervals of flood events is set up for the duration of 3 h, 6 h, 9 h and 12 h, using the discharge within the same time as the model input. After this, the prediction of first intervals of flood events is evaluated by the RMSE with the testing dataset (event #121 to event #180). The averaged RMSE was calculated for different prediction times (3 h, 6 h, 9 h, 12 h) to quantify the prediction performance of each individual ANN. Table 1 shows the percentage of the accurate prediction ANNs, classified by RMSE of 0.2 m, 0.3 m and 0.4 m. In Table 1, if the error threshold is set to 0.3 m, the accuracy can be considered excellent with values above 80% for all the prediction durations.


**Table 1.** Number of wet grids and grid percentages of different large error thresholds for testing synthetic flood events (60 events, #121~#180).

Note: highlighted in gray are the percentages larger than 70%.

#### 4.1.2. Historical Flood Events

After testing with the synthetic events, the ANN model performance is further examined with the historical flood events. Thus, the historical events, the same as the synthetic events, are simulated by the FloodEvac tool for their inundation maps. Afterward, the grid RMSE is calculated for the evaluation of prediction accuracy on the three historical flood events of their first intervals. A value of 10 m<sup>3</sup> /s was selected as the forecast threshold to initiate the forecasts since this value is crossed before the beginning of the flooding in all three historical events. The forecast threshold is chosen slightly bigger than the average discharge of 9.2 m<sup>3</sup> /s of White Main [35] to avoid the low discharges from triggering flood warnings.

Historical flood events 2006

Figure 5 shows the discharge inputs for the historical flood event in 2006. The first 3 h, 6 h, 9 h, 12 h discharge curves are given by the trained ANN as in Chapter 4.1. Figure 6 compares the prediction of the inundation map of the first intervals of 3 h, 6 h, 9 h and 12 h with the inundation map from the hydraulic model of the historical flood event 2006. Table 2 shows the performance of the prediction for historical event 2006, evaluated by average RMSE for each individual ANN. As the forecast interval increases from 3 h to 12 h, the prediction accuracy drops, evaluated by grid percentages of RMSE.

**Figure 5.** Hydrographs of the flood event in 2006. Seven discharge curves of three rivers and four streams are shown in different colors. Time 0 marks the start of the prediction. The dash lines upon the discharge curves mark the different discharge sections for prediction inputs.

**Figure 6.** *Cont.*

**Figure 6.** Inundation maps from the prediction of water depths of the first intervals in flood event 2006. (**a**) ANN inundation map 3 h; (**b**) hydrodynamic inundation map 3 h; (**c**) ANN inundation map 6 h; (**d**) hydrodynamic inundation map 6 h; (**e**) ANN inundation map 9 h; (**f**) hydrodynamic inundation map 9 h; (**g**) ANN inundation map 12 h; (**h**) hydrodynamic inundation map 12 h.

**Table 2.** Numbers of wet grids and accurate grid percentage for event 2006. A wet grid is with the water level over 0.1 m; any water depth below this cutoff value is eliminated. Table shows grid numbers with a larger root-mean-square error (RMSE) and their percentages to the total wet grids.


Note: highlighted in gray are the percentages larger than 70%.
