Satellite observations are often incomplete due to cloud cover, resulting in missing ocean data. To address this, we developed T-DINEOF (Data Interpolating Empirical Orthogonal Function), a reconstruction method that simultaneously estimates sea surface temperature, chlorophyll concentration, and wind conditions by learning relationships among variables. Results show that T-DINEOF improves reconstruction accuracy, especially in regions with sparse data or weak correlations, providing more reliable ocean information for environmental monitoring.

Strong motions caused by surface waves can set the material at the bottom in motion. How strong the wave motions need to be depends on the bottom type, for example mud or sand. We estimated how often waves can lift particles from the bottom by comparing wave model results to sea floor samples in the laboratory. We included the effect of seasonal biological activity, which was found to be significant. 

This study reveals the structure and alteration of a narrow freshwater-driven ocean low-density filament and its evolution through tidal modulationa an dlocale filament kinematics. Using multi-platform in situ observations, it is shown that these submesoscale features can rapidly being altered and reshape in near-surface waters, influencing how energy and heat alter just below the ocean-atmosphere interface.

We introduce TS (Temperature-Salinity)-Cast, a novel deep neural network that reconstructs subsurface thermohaline structures from satellite observations. Validated against independent time-series data, TS-Cast achieves root mean squared errors of < 1 °C and < 0.1 psu in the upper 500 m of the Kuroshio Extension, comparable or surpassing data-assimilated numerical models. Critically, we demonstrate that the physical limitations of the input satellite data fundamentally constrain the model's predictive skill.