notesum.ai

Published at November 22

Resolution-Agnostic Transformer-based Climate Downscaling

cs.CV

cs.AI

Released Date: November 22, 2024

Authors: Declan Curran, Hira Saleem¹, Flora Salim, Sanaa Hobeichi²

Aff.: ¹School of Computer Science and Engineering, University of New South Wales, Sydney, New South Wales, Australia; ²ARC Centre of Excellence for the Weather of the 21st Century and Climate Change Research Centre, University of New South Wales, Sydney, New South Wales, Australia

Arxiv: http://arxiv.org/abs/2411.14774v1

Model	RMSE $\downarrow$	PSNR $\uparrow$	SSIM $\uparrow$	Carbon $(kgCO_{2})$
Bilinear Interpolation	0.138	17.269	0.992	0.002
ResNET	1.082	1.647	0.984	0.002
ResNET - Modified Loss	3.199	-9.807	0.755	0.003
Earth ViT	0.031	30.221	0.973	0.040
Earth ViT (TRAINED ON BARRA)	0.031	30.229	0.964	0.194
Earth ViT - Modified Loss	0.031	30.375	0.983	0.043
Earth ViT - Modified Loss (TRAINED ON BARRA)	0.029	30.690	0.969	0.173