Enhancing Hydrological Modeling with Transformers: A Case Study for 24-Hour Streamflow Prediction DOI Creative Commons
Bekir Zahit Demiray, Muhammed Sit, Omer Mermer

et al.

EarthArXiv (California Digital Library), Journal Year: 2023, Volume and Issue: unknown

Published: Sept. 12, 2023

In this paper, we address the critical task of 24-hour streamflow forecasting using advanced deep-learning models, with a primary focus on Transformer architecture which has seen limited application in specific task. We compare performance five different including Persistence, LSTM, Seq2Seq, GRU, and Transformer, across four distinct regions. The evaluation is based three metrics: Nash-Sutcliffe Efficiency (NSE), Pearson’s r, Normalized Root Mean Square Error (NRMSE). Additionally, investigate impact two data extension methods: zero-padding persistence, model's predictive capabilities. Our findings highlight Transformer's superiority capturing complex temporal dependencies patterns data, outperforming all other models terms both accuracy reliability. study's insights emphasize significance leveraging deep learning techniques, such as hydrological modeling for effective water resource management flood prediction.

Language: Английский

Better Localized Predictions with Out-of-Scope Information and Explainable AI: One-Shot SAR Backscatter Nowcast Framework with Data from Neighboring Region DOI Creative Commons
Zhouyayan Li, İbrahim Demir

EarthArXiv (California Digital Library), Journal Year: 2023, Volume and Issue: unknown

Published: June 10, 2023

Synthetic Aperture Radar (SAR) provides 10-m weather-independent global Earth surface observations for various tasks such as land cover use mapping, water body delineation, and vegetation change monitoring. However, the application of SAR imagery has been limited to retrospective by a “first event then observation” rule. Recent studies have proven feasibility one-shot forecast backscatters using meteorological driving forces, soil moisture, geomorphic factors, previous images collected target area. Although approach is promising, spatial connectivity, more specifically, influence status surrounding areas on location yet be considered. To fill that gap, this study proposed two nowcasting frameworks can integrate precipitation moisture data from through aggregation (SA) processing series (SS), respectively. The catastrophic 2019 Central US Flooding was used case with goal predicting captured during event. results SA, SS, framework only considers localized input (S0) are compared against each other well benchmark performance created persistence assumption. Results show S0, SS outperform benchmark. In addition, considering neighboring contribute further improves prediction accuracy. Comparing gradients considering/not additional indicates alter model’s attention feature matrix. difference in between SA way information integrated also matters. methodology serve building block active usage forward-looking early flood warning response.

Language: Английский

Citations

2
Enhancing Hydrological Modeling with Transformers: A Case Study for 24-Hour Streamflow Prediction DOI Creative Commons
Bekir Zahit Demiray, Muhammed Sit, Omer Mermer

et al.

EarthArXiv (California Digital Library), Journal Year: 2023, Volume and Issue: unknown

Published: Sept. 12, 2023

In this paper, we address the critical task of 24-hour streamflow forecasting using advanced deep-learning models, with a primary focus on Transformer architecture which has seen limited application in specific task. We compare performance five different including Persistence, LSTM, Seq2Seq, GRU, and Transformer, across four distinct regions. The evaluation is based three metrics: Nash-Sutcliffe Efficiency (NSE), Pearson’s r, Normalized Root Mean Square Error (NRMSE). Additionally, investigate impact two data extension methods: zero-padding persistence, model's predictive capabilities. Our findings highlight Transformer's superiority capturing complex temporal dependencies patterns data, outperforming all other models terms both accuracy reliability. study's insights emphasize significance leveraging deep learning techniques, such as hydrological modeling for effective water resource management flood prediction.

Language: Английский

Citations

0