Hydrological time series prediction using neural architecture search-enhanced dual-stage attention-based Bi-LSTM

ÇITAKOĞLU H., Apak S., ÇITAKOĞLU H., Sammen S. S., YURTSEVER A.

HYDROLOGICAL SCIENCES JOURNAL, vol.71, no.7, pp.1430-1448, 2026 (SCI-Expanded, Scopus) identifier identifier

  • Publication Type: Article / Article
  • Volume: 71 Issue: 7
  • Publication Date: 2026
  • Doi Number: 10.1080/02626667.2026.2637779
  • Journal Name: HYDROLOGICAL SCIENCES JOURNAL
  • Journal Indexes: Science Citation Index Expanded (SCI-EXPANDED), Scopus, IBZ Online, Compendex, Geobase, INSPEC
  • Page Numbers: pp.1430-1448
  • Keywords: dual-stage attention, Eastern Black Sea Basin, neural architecture search, streamflow prediction, water resource management
  • Istanbul University Affiliated: No

Abstract

A Neural Architecture Search-enhanced Dual Stage Attention Bidirectional Long Short-Term Memory (NAS-DSA-BiLSTM) model is proposed to capture nonlinear temporal hydrological patterns and is evaluated at three hydrological stations. The NAS-enhanced model achieved the best performance at Kaptanpasa, with the lowest MSE (11.249-16.224) and the highest R2 values, outperforming LSTM and DNN-LSTM models. Additional evaluations using NSE, sensitivity analysis, and Taylor diagrams confirmed stable and accurate predictions, particularly at Ulucami (NSE = 0.91) and Kaptanpasa (NSE = 0.80), demonstrating the model's robustness for streamflow forecasting. The proposed model improved MSE by 15-26%, demonstrating robust and reliable streamflow forecasting performance.