Mapping coastal currents in midlatitudes and the Arctic

PI:  James Carton, University of Maryland College Park
Start Year: 2025 | Duration: 4  years
Partners: NOAA, NASA

Project Abstract: 

Central objectives: Community-standard data-driven sea level and associated surface current products available now are designed for the open ocean and do not resolve the fine O (10-50 km) spatial scales and short O(1-10dy) time-scales needed in high human and economic impact coastal zones and the Arctic.  This project will conduct research with the goal of improving sea level and surface current products to extend into these areas and thus explore the limits of space- and time-resolution.  The project will use the resulting software to produce   global daily gridded sea level, surface current and uncertainty estimates 1993-2024.  Emphasis will be placed on realistic evaluation of uncertainty in both sea level and surface current. Methods/techniques: Two approaches are presented. Approach 1 uses multiscale 3DVar built on the community-standard Joint Center for Satellite Data Assimilation JEDI framework, which ingests along-track nadir altimetry and ancillary information and produces uniformly gridded daily sea level and current estimates. These can be compared to independent observations such as HF RADAR currents, drifter tracks, and tide gauges. The prespecified error covariances are adjusted to improve spatial and temporal resolution. Results from the ongoing ‘Ocean Data Challenges’ community effort suggest that machine learning-based algorithms are developing the potential to improve on linear estimators such as that of Approach 1. In our second approach we will explore this potential by adding a convolutional neural network layer to extract additional information from the observation-analysis misfits resulting from Approach 1. Because this builds on Approach 1 we allow simple ocean dynamics to inform the machine learning.  This software structure clarifies the contribution of the neural network and to makes it easier to interpret the origins of various features in the output fields.  Uncertainty is explored through a series of data withholding and parameter varying experiments as well as by comparison to independent observations and exploration of dynamical balances. Perceived significance: Populations living in coastal regions depend on maritime operational and commercial activities such as fisheries, weather forecasting, and debris tracking, all of which require knowledge of nearshore mesoscale currents and their seasonal and decadal climatology.  This need is highlighted in the most recent NAS Space Studies Board which ranks sea level and surface current as targeted environmental variables. By using continuous reference mission sea level and ancillary nadir sea level over the full 40+ year record to map surface currents in coastal and high latitude zones and their connection to the global context, this proposal directly addresses item 2.2:  Regional and coastal dynamics.