Authors | Kushagra Kulshreshtha1, Nimesh Raj2, Sohini Chowdhury3, Yatika Gori4, Anurag Shrivastava5 , A. Kakoli Rao6, Akhil Sankhyan7 , P. William8 |
Affiliations |
1Institute of Business Management, GLA University, 281406 Mathura, Uttar Pradesh, India 2Centre of Research Impact and Outcome, Chitkara University, 140417 Rajpura, Punjab, India 3Centre of Research Impact and Outcome, Chitkara University, 140401 Rajpura, Punjab, India 4Department of Mechanical Engineering, Graphic Era Deemed to be University, Dehradun, India 5Saveetha School of Engineering, Saveetha Institute of Medical and Technical Sciences, Chennai, TN, India 6Lloyd Institute of Engineering & Technology, Greater Noida, India 7Lloyd Law College, Greater Noida, India 8Department of Information Technology, Sanjivani College of Engineering, Kopargaon, MH, India |
Е-mail | kushagra.kulshrestha@gla.ac.in |
Issue | Volume 16, Year 2024, Number 6 |
Dates | Received 25 August 2024; revised manuscript received 14 December 2024; published online 23 December 2024 |
Citation | Kushagra Kulshreshtha, Nimesh Raj, et al., J. Nano- Electron. Phys. 16 No 6, 06021 (2024) |
DOI | https://doi.org/10.21272/jnep.16(6).06021 |
PACS Number(s) | 92.40. – t, 92.70.Np |
Keywords | Daily streamflow, Hydrological model, CNN-LSTM, WRF. |
Annotation |
Daily streamflow prediction in data-sparse watercourses is significant for efficient water resource management and climate change variations. Especially in areas with sparse observational data, the geographical and temporal complexity of hydrological systems presents a significant challenge for traditional hydrological models. In this research, we offer an innovative approach for improving daily streamflow predictions by integrating the Convolutional Neural Network-Long Short-Term Memory (CNN-LSTM) architecture with physical processes and utilizing the Weather Research and Forecasting (WRF) model for the hydrological process. The objective is to improve the WRF model's capability of capturing the complex interactions between weather conditions and streamflow dynamics by combining this deep learning framework with the physical processes that define the model. For a more precise depiction of the hydrological system, the WRF model, well-known for its high-resolution atmospheric simulations, provides fine-grained meteorological inputs. The performance of the suggested method is evaluated using RMSE (5.14), MAE (6.85), MEDAE (5.97) as well as R2 (12.05) metrics and they are compared to existing methods. The combination of CNN-LSTM and WRF offers a promising path for improving the accuracy and reliability of hydrological models, which is critical for informed decision-making in water resource management and climate resilience. |
List of References |