RAINFALL RUNOFF BENCHMARK TEST BY DEEP LEARNING MODEL USING URBAN MEDIUM AND SMALL RIVER BASIN DATASET

Abstract

 Urban floods occur frequently, and there are great expectations for the application of machine learning models that can be easily constructed to flood forecasting fields in urban medium and small size basin, which have complicated runoff mechanisms. Therefore, in this paper, we constructed a deep learning model using an actual river basin dataset of urban small and medium rivers, which was created at a short observation interval of 1 minute, which is different from that of large river basins. Benchmark tests were performed when the parameters were changed. An ANN model was also constructed for comparison with the deep learning model, and the performance was evaluated by a new index called the PD ratio, which is the number of parameters of the deep learning model and the number of observed data. As a result, it was found that the deep learning model is superior to the ANN model for the learning/verification flood at the same PD ratio, and in particular, the ANN model shows rapid fluctuations in the verification flood that do not follow the actual results.