General comments :

This document describes the 3D ocean current dataset developed in the framework of the World Ocean Circulation project (ESA funded) and named  WOC-NATL3D.

The dataset extends over the North Atlantic, from the surface to 1500 m depth at a 1/10° spatial resolution on a daily basis. The dataset contains horizontal and vertical velocity, as well as temperature and salinity fields from 2010 to 2019.

A deep learning algorithm is used to compute 3D T/S fields which then feeds a diabatic quasi-geostrophic equation together with ERA5 fluxes and horizontal Ekman currents estimates.

This product has been developed to fit the purpose of fishery applications that need higher spatial resolution vertical velocity.

This article is well written, pleasant to read and well structured. However, we are still left wanting more information about validation. It does not clearly answer the question of the validity of the WOC-NATL3D vertical velocity:

• Some figures request more discussion in the text. The authors referred to publications describing the method, but more details could have been provided in the text for a better comprehension.
• All the validation is performed on current intensity. There is no indication on the validation of direction of the current, neither on WOC-NATL3D nor the other reanalysis dataset. Vertical velocity is related to divergence of the horizontal flow but no validation of the zonal nor meridional component is shown here. Can the authors provide some justification for this choice or provide some more validation results?

More specific comments are listed in the supplement file.

Review of “3D reconstruction of horizontal and vertical quasi-geostrophic currents in the North Atlantic Ocean” by Sarah Asdar, Daniele Ciani , and Bruno Buongiorno Nardelli

The authors present a 1/10° data-driven data set of 3D ocean currents, as well as of temperature and salinity in the upper 1500 meters of the North Atlantic subtropical gyre between 20°N and 50°N, WOC-NATL3D. The data set covers the period from 2010 to 2019 with daily resolution.

The product is based on a diagnostic tool originally developed for a global  product (OMEGA3D) by Bruno Buongiorno Nardelli (2022). The method is based on the the quasi-geostrophic omega equation. A deep learning technique is used to obtain the fields from Argo profiles, altimetry, SST and SSS. Also used are ERA5 air-sea fluxes and modelled Ekman currents from Copernicus.

Both products, WOC-NATL3D and OMEGA3D, are supposed to better reproduce drifter observations when compared to reanalysis products. WOC-NATL3D aims to improve accuracy near the surface, in particular by using the modelled Ekman currents. Two reanalysis products (SODA and GLORYS) as well as drifter and altimetry data are used for evaluation.

The article is written well and comprehensibly and also well structured. 

Comments:

• Evaluation of the vertical velocities (section 3.1) is quite limited. I find it understandable that no comparisons with direct measurements are possible. However, an estimation of the uncertainty of the vertical velocities is desirable.
• I understand that the SODA data set was selected for comparison because vertical velocities are rare in reanalysis products. With GLORYS a second reanalysis product was selected for comparison, is there a justification for this choice?
• The labels and titles of the figures are in a small font size. The subscript letters in the titles in particular are difficult to read on a printout.

More specific comments can be found in the supplement.