Multi‐step‐ahead flood forecasting using an improved BiLSTM‐S2S model

Rainfall–runoff modeling is a complex hydrological issue that still has room for improvement. This study developed coupled bidirectional long short-term memory (LSTM) with sequence-to-sequence (Seq2Seq) learning (BiLSTM-Seq2seq) model to simulate multi-step-ahead runoff flood events. The LSTM Seq2Seq (LSTM-Seq2Seq) and multilayer perceptron (MLP) was set as benchmarks. results show that: (1) root mean absolute error reduced by approximately 19% up 27%, the Nash–Sutcliffe coefficient of efficiency improved 14% 34% 6-h-ahead prediction BiLSTM-Seq2Seq compared LSTM-Seq2Seq MLP; (2) good performance not only one-peak events but also multi-peak events; (3) can mitigate time-delay problem time lag shortened 39% 69% in comparison MLP. These suggest be mitigated BiLSTM-Seq2Seq, which excellent potential series predictions field.