We found a match
Your institution may have rights to this item. Sign in to continue.
- Title
Rainfall variability over multiple cities of India: analysis and forecasting using deep learning models.
- Authors
Panda, Jagabandhu; Nagar, Nistha; Mukherjee, Asmita; Bhattacharyya, Saugat; Singh, Sanjeev
- Abstract
India being an agrarian economy, rainfall is an essential component that directly or indirectly influences agricultural produce. With the increasing impacts of the changing climate scenario, it is anticipated that in the near future, frequent and extreme rainfall episodes will trigger events like severe floods, landslides, etc. Therefore, it is extremely important to make precise predictions so that the intensity of the impacts on life and property can be reduced. In recent times, with the advancement of AI/ML applications, it has become popular in weather and climate sciences. The current work uses 121 years of rainfall data for analysis and prediction purposes, where deep learning (DL) approaches like LSTM (Long Short Term Memory), BiLSTM (Bi-directional LSTM) and GRU (Gated Recurrent Unit) have been adopted. The long-term rainfall analysis and prediction over selected smart cities of India is based on their location in the homogenous monsoon regions. The results obtained from the models indicated that for univariate forecasting, the overall performance of BiLSTM is better than others for most cities considered, while GRU predicted well for places with higher rainfall variability. In the multivariate approach, LSTM's performance is superior. Therefore, a combination of BiLSTM and GRU might offer a better result in the univariate approach, or an advanced version of LSTM might enrich the outcomes in the multivariate approach for rainfall analysis and forecasting.
- Subjects
INDIA; DEEP learning; CITIES &; towns; LANDSLIDES; LONG-term memory; CLIMATOLOGY; RAINFALL; SMART cities
- Publication
Earth Science Informatics, 2024, Vol 17, Issue 2, p1105
- ISSN
1865-0473
- Publication type
Article
- DOI
10.1007/s12145-024-01238-1