Prabaswara (2026) Analisis Komparatif Model Artificial Neural Network (ann) Dan Long Short-Term Memory (lstm) Untuk Prediksi Debit Sungai Dalam Mendukung Pengelolaan Sumber Daya Air Berkelanjutan

⚠️ Warning: This summary was generated from the abstract only, as the full text was not available.

Identification

• Journal: UMM Institutional Repository (University of Maine at Machias)
• Year: 2026
• Date: 2026-07-24
• Authors: Alifia Oriana Prabaswara
• DOI: None

Research Groups

Not specified in the provided text.

Short Summary

This study compares Artificial Neural Networks (ANN) and Long Short-Term Memory (LSTM) models for daily streamflow prediction in the tropical Welang River, finding that a 7-day look-back ANN model provides the best performance for short-term forecasts due to the watershed's short hydrological memory.

Objective

• To compare the performance of Artificial Neural Networks (ANN) and Long Short-Term Memory (LSTM) models for daily discharge prediction in the Welang River watershed to support Sustainable Water Resources Management.

Study Configuration

• Spatial Scale: Welang River watershed (tropical).
• Temporal Scale: Daily predictions; Data period from 2010 to 2025.

Methodology and Data

• Models used: Artificial Neural Network (ANN) with 7, 15, and 30-day look-back periods; Long Short-Term Memory (LSTM) with 30-sequence look-back.
• Data sources: Observation data from AWLR Dhompo (upstream) for daily discharge (5,479 data points). Data split 80% for training and 20% for testing.

Main Results

• The ANN model with a 7-day look-back period (ANN LB7) demonstrated the best performance for daily discharge prediction: R² = 0.9305, RMSE = 1.6447 m³/s, NSE = 0.9305, and MAPE = 21.58%. Training time was in seconds.
• The LSTM model performed similarly with R² = 0.9296 and RMSE = 1.6577 m³/s but required significantly longer training time (~7 minutes).
• ANN accuracy declined with longer look-back periods, indicating a short hydrological memory (≤7 days) in the Welang River watershed, with Mean and t-1 identified as dominant features.
• ANN LB7 is recommended for daily flood warnings and irrigation allocation, while LSTM is suggested for medium-term forecasts.

Contributions

• Provides a comparative analysis of ANN and LSTM models for daily streamflow prediction in a highly variable tropical watershed, a region often under-represented in such studies.
• Identifies the optimal model configuration (ANN with 7-day look-back) for short-term streamflow prediction in the Welang River, offering insights into the watershed's short hydrological memory.
• Offers practical recommendations for model application in Sustainable Water Resources Management, including daily flood warnings and irrigation allocation, leveraging upstream data for significant lead time in downstream mitigation.

Funding

Not specified in the provided text.

Citation

@article{Prabaswara2026Analisis,
  author = {Prabaswara, Alifia Oriana},
  title = {Analisis Komparatif Model Artificial Neural Network (ann) Dan Long Short-Term Memory (lstm) Untuk Prediksi Debit Sungai Dalam Mendukung Pengelolaan Sumber Daya Air Berkelanjutan},
  journal = {UMM Institutional Repository (University of Maine at Machias)},
  year = {2026},
  url = {https://openalex.org/W7219361103}
}