01-05 December 2025
INCOIS, Hyderabad, India.
| Abstract Submission No. | ABS-05-0056 |
| Title of Abstract | Downscaling Sea Surface Temperature for Extreme Events with Residual Corrective Neural Networks |
| Authors | Onkar Jadhav*, Tim French, Ivica Janekovic, Nicole Jones, Matt Rayson |
| Organisation | The University of Western Australia |
| Address | 63 adelaide tce Crawley, WA, Australia Pincode: 6004 E-mail: onkar.jadhav@uwa.edu.au |
| Country | Australia |
| Presentation | Oral |
| Abstract | The large-scale oceanic and atmospheric forecasts provided by global climate models lack sufficient resolution to accurately capture the response of the coastal ocean. Dynamical downscaling is computationally prohibitive, especially when applied to the large coastlines, like Western Australia and to many climate ensembles. Therefore, this work presents a statistical downscaling of sea surface temperature (SST) from the seasonal coupled ocean-atmosphere forecast system (ACCESS-S2) that is based on machine learning techniques. This study introduces a novel methodology that combines a convolutional neural network called U-Net, known for its ability to effectively capture spatial features, for coarse-scale predictions with a residual corrective neural network (RCNN) that iteratively refines these predictions toward high-resolution SST. The target high-resolution SST fields are obtained from the Regional Ocean Modeling System (ROMS). The RCNN incorporates dynamically adjusted residual corrections proportional to residual scales, ensuring stable and multi-scale refinement. Through iterative residual corrections, the model shifts its focus from coarse corrections in early steps to fine-grained refinements in later steps, capturing both broad trends and localized features, such as eddies and fronts. The developed framework efficiently downscales the SST along the west coast of Australia. A specific case study involving the 2011 marine heatwave event shows that the RCNN can sharpen the SST predictions of ACCESS-S2 and determine temperature anomalies during the heatwave efficiently. By applying the developed framework, this work achieves a balance between computational efficiency and the accuracy required for capturing local oceanic variations, ultimately improving forecasting capabilities for coastal management and marine ecosystem studies. |
| Are you part of IIOE-2 endorsed project | yes |
| Endorsed Project Number | IIOE2-EP58 |
| Keywords | Sea Surface Temperature, Statistical Downscaling, Marine Heatwaves, Extreme Events, Residual Corrective Neural Network |
| For Awards | no |