POSTER Session 4

CONTEXT AWARE DIFFUSION MODEL FOR SEA SURFACE TEMPERATURE RECONSTRUCTION

Weather and climate forecasts use the distribution of sea surface temperature (SST), a critical factor in atmosphere-ocean interactions. Infrared sensors, which use channels with wavelengths ranging from about 3.7 µm to 12 µm, are used to produce high spatial resolution SST data. However, SST retrieved from infrared sensor-based satellites contains noise and missing areas due to cloud contamination. Therefore, when reconstructing SST under clouds, it is also necessary to consider observational noise. In this study, we present a conditional diffusion probabilistic model (DDPM)-based SST reconstruction method that can also reduce observational noise. The proposed approach reconstructs SST under cloud regions from satellite SST observations obtained with the visible infrared imaging radiometer suite (VIIRS), using optimum interpolation sea surface temperature (OISST) as a background. Our model was conditioned on the denoising step of the DDPM in order to reduce observation noise. To evaluate the effectiveness of our method, a reconstruction using a fixed mask was performed with 10,578 VIIRS data from 2022. The result showed that the average mean absolute error and root mean squared error were 0.23°C and 0.12°C, respectively. When evaluated with buoy data, the proposed model exhibited better accuracy compared to other numerical models and followed the overall spatial distribution of OISST under the large cloud cover. This study suggests the possibility of a deep learning model for generating high-resolution, gap-filled SST data on a global scale.

*Minki Choo, Ulsan National Institute of Science and Technology, Republic of Korea, [email protected], https://orcid.org/0000-0002-0600-7065

Jungho Im, Ulsan National Institute of Science and Technology, Republic of Korea, [email protected]

Sihun Jung, Ulsan National Institute of Science and Technology, Republic of Korea, [email protected]

Code of Conduct
Privacy Policy