The authors have proposed a deep learning-based model for extracting surface-intensified eddies (SEs) and subsurface-intensified eddies (SSEs) using satellite-derived sea surface height and sea surface temperature data from 2000 to 2015 in the North Indian Ocean (NIO). Additionally, the study integrated ocean color remote sensing data to investigate the differential impacts of SEs and SSEs on chlorophyll-a (Chl-a) concentration in the two basins of the NIO: the Arabian Sea and the Bay of Bengal, known for their varying biological productivity. The authors have successfully linked the observed differences in Chl-a anomalies between SEs and SSEs to the contrasting subsurface structures revealed by Argo profiles. Overall, the manuscript provides an effective deep learning-based method for extracting SSEs solely from remote sensing data and contributes to a deeper understanding of the complex interactions between eddy dynamics and biogeochemical processes.

The manuscript is generally well written. But moderate revisions should be made before publication. My concerns and comments are detailed below.

- Consider adding information on the formation mechanism of SSEs in the regions to provide readers with a more comprehensive understanding of SSEs. What’s the dominant mechanism of SSE generating in NIO? What cause the different chlorophyll features between SEs and SSEs? If necessary, please give explanations with Argo/BGC-Argo results.

- The manuscript concluded that SSEs account for nearly 50% of the total eddies, which needs further consideration. The sea surface temperature can be easily disturbed by environment, such as wind speed. Therefore, identifying SSEs according SSTA<0 (or SSTA>0) may increase the noises from low-energy eddies. It is suggested to set threshold for SSEs identification, such as amplitude, lifetime, which should increase the accuracy of SSEs identification.

- The paper proposes an identification method for SEs and SSEs using deep learning, along with validation and analysis of their temperature and chlorophyll characteristics. Consider refining the title to align more accurately with the manuscript's content.

- Line 104: Reword 'as described by Assassi et al. (2016)' to 'as described in the study by Assassi et al. (2016)'.

- Line 114-117: Has the sign of SSρ/SSHA been previously used as an indicator to distinguish SEs and SSEs in any studies? Please provide references if available.

- Line 145: Please include the formula for the dice loss function.

- Line 146: What is the specific definition of accuracy for the DL-based model? Clarify this point.

- Line 167: Provide detailed information on the inversed distance weighting interpolation method.

- Line 258: Revise 'to accurately determine the most intense core's location' to 'to determine the location of the most intense core accurately.'

- Figure 5: The Chl-a anomalies induced by SSAEs and SSCEs displayed in the current color bar are not easily discernible. It is recommended to modify the color bar to enhance the visibility of the differences.

This is an important work. The major contributions of the authors are 1) developing Deep Learning methods to identify surface-intensified and sub-surface-intensified eddies using satellite observations and 2) gaining new understanding on the role of eddies played in spatiotemporal variations of ocean primary productivity in the Northern Indian Ocean. Thus, this reviewer recommends the submission to be accepted for publication.