Urban flooding occurs during heavy rains of short duration, so quick and accurate warnings of the danger of inundation are required. Previous research proposed methods to estimate statistics-based urban flood alert criteria based on flood damage records and rainfall data, and developed a Neuro-Fuzzy model for predicting appropriate flood alert criteria. A variety of artificial intelligence algorithms have been applied to the prediction of the urban flood alert criteria, and their usage and predictive precision have been enhanced with the recent development of artificial intelligence. Therefore, this study predicted flood alert criteria and analyzed the effect of applying the technique to augmentation training data using the Artificial Neural Network (ANN) algorithm. The predictive performance of the ANN model was RMSE 3.39-9.80 mm, and the model performance with the extension of training data was RMSE 1.08-6.88 mm, indicating that performance was improved by 29.8-82.6%.

In Korean metropolitan areas, the high density of residential and commercial buildings, highly impervious surfaces, and steep slopes contribute to floods that can occur within a short duration of heavy rainfall. To prepare for this, advance warning measures based on accurate flood alert criteria are needed. In our previous study, we demonstrated the applications of a Neuro-Fuzzy model that considersthe characteristics of the basin to predict flood alert criteria in areas with no damage. However, as the number of learning materials are low, at 27, the evaluation and verification of the model has not been sufficiently accomplished, and its application is limited. Therefore, in this study, we propose an improved model based on the initializing function of the Neuro-Fuzzy algorithm, the construction of training data, and preprocessing. Compared to the existing model, the improved model reduced the average error by 48.1%~65.4% and the RMSE by 50.7%~60.1%. The new model, when applied to actual floods, showed an improvement of 0.7%~19.1% in accuracy.