Earth and Space Science Environmental Science (miscellaneous)

Spatiotemporal Mapping of Sea Surface Height and Temperature Fields From Satellite Observations Using Score‐Based Data Assimilation

Tatsu Monkman, Dhruv Balwada, Spencer Jones, Scott Martin, K. Shafer Smith
Wiley · 2026-11-01 · DOI 10.1029/2026ea005326

Abstract

Abstract Submesoscale flows at the ocean surface play a crucial role in the Earth system but are difficult to observe on a global scale. Modern satellite observation systems allow us to partially bridge this gap by capturing fields that are dynamically linked to the surface flow such as sea surface height anomaly (SSHA) and sea surface temperature (SST) at submesoscale resolution. However, these high‐resolution observations contain frequent temporal and spatial gaps, which create challenges for traditional reconstruction methods that depend on SSHA and SST data on a uniform grid. To address this, we apply diffusion models guided using the score‐based data assimilation (SDA) algorithm to generate plausible time series of SSHA and SST evolution over small regions of the ocean surface at submesoscale resolution based on spatially and temporally sparse satellite observations. Our procedure is based on training unconditional generative diffusion models to predict the time series evolution of satellite‐observable surface fields using MITgcm llc4320, a high resolution numerical ocean model, as a “prior” field, and steering the generative process via posterior sampling. Next, we use the SDA algorithm to extend our generative models’ receptive field to encompass observations spaced over arbitrarily long time series. We test our models” performance over a wide range of dynamical regimes and observational conditions in an idealized setting and apply our models to real Surface Water and Ocean Topography SSHA data and Advanced Clear‐Sky Processor for Ocean SST fields as proof of concept and motivation for future generative data assimilation efforts.

Read at publisher →