01-05 December 2025
INCOIS, Hyderabad, India.
| Abstract Submission No. | ABS-05-0220 |
| Title of Abstract | Rapid climate downscaling using deep learning to inform climate vulnerability assessment of a marine World Heritage Area in the Indian Ocean |
| Authors | Chaojiao Sun, Ajitha Cyriac, Jim Greenwood, Jess Melbourne-Thomas, Damian Thomson, Marian Sheppard, Russell Babcock, Dirk Slawinski, John Taylor, Nick D'Adamo* |
| Organisation | University of Western Australia |
| Address | CSIRO Environment Crawley, WA, Australia Pincode: 6009 E-mail: chaojiao.sun@csiro.au |
| Country | Australia |
| Presentation | Oral |
| Abstract | To help evaluate climate vulnerability of marine values of the Ningaloo Coast World Heritage Area and the Ningaloo Marine Park, five best performing CMIP6 climate models for the Australian Region [Grose et al., 2023] have been chosen to produce high resolution climate projections of sea-surface temperature (SST) using machine learning, which is capable of rapidly downscaling climate models for different Shared Socioeconomic Pathways (SSPs) to inform a range of likely outcomes. First, biases in SST from these climate models were assessed and corrections made using 30 years of satellite SST observations [Skirving et al., 2020] using a quantile delta mapping method [Cannon et al., 2015]. The analysis shows all five models to have a warm bias in SST of ~1°C. A deep learning model was then trained to downscale the SST projections from 100 km resolution to 10 km using the Australian BlueLink OFAM model simulations [Cyriac et al., 2025]. The downscaled ensemble mean of the five climate models at 10 km resolution show that SST at the Ningaloo Coast of the study region is projected to warm by about 0.5°C under a global warming level (GWL) of 1.5°C, increasing to ~1°C under a GWL of 2°C. At a GWL of 3°C, the warming would exceed 1.5°C. As global warming levels (GWL) approach 3°C, the degree heating week (DHW), a measure of cumulative heat stress, is projected to increase substantially. This increase is a major concern because higher DHW values are directly linked to a greater risk of severe coral bleaching and widespread mortality. On the other hand, sea level along the west coast of Australia will remain slightly below the current level under the high emission scenario (RCP8.5) by the end of 21st century, with likely very little change under a moderate emission scenario [Zhang et al, 2017]. |
| Are you part of IIOE-2 endorsed project | no |
| Keywords | climate downscaling, deep learning, climate vulnerability, world heritage areas |
| For Awards | no |