Motivated by the desire to identify and quantify the interaction of “coastal” water with the adjacent deep Levantine Sea, the authors develop a pattern detection model using deep learning applied to ocean color data to track the interactions between the coastal and offshore water. They choose to use ocean color because using satellite altimetry is prone to errors near the coast. The authors discover that pinching-off events that send coastal water offshore occur more frequently to the north along the coast where the coastal flows become more unstable. They also consider the temporal evolution, persistence, spatial scale and year-to-year variability of these features that transport coastal water seaward. Overall, this is a nice application of a novel technique, and the authors do a nice job of analyzing the results to gain insight into the connection of the coastal and adjacent deep sea. I recommend that this paper be published in the “Ocean Science” after minor revision.

Suggestions to improve the manuscript:

1. I agree with the authors that satellite altimetry is likely to miss small-scale features, especially near the coast but I do not think the right-hand panel in Figure 2 makes the case well. First, the text says “velocity field” while the figure shows vorticity. Second, have the chlorophyll and altimetry fields been closely matched in time? The altimetry field has the correct sized positive and negative vorticity features that, if simply displaced by a mismatch in time, could match up with drifter field (see my comment that a symbol is needed at the start of the drifter track so the reader can tell the direction of flow). Please clear up the use of the altimetry data.
2. More needs to be said about choosing the region of the Nile and eastern part of the Levantine Sea to train the learning procedure. This region should be indicated on one of the maps. What are the consequences if a larger or different region are chosen for training? What biases or errors might be introduced by the choice of training region?
3. Lines 130-131 present the model performance. Please state whether these are “good” scores, perhaps by giving typical values for a good performance from other studies. The 3% error is indeed impressive.
4. Section 4, figure 7: how is the dark blue mask along the coastline chosen? It evidently masks out the coastal water that is on the continental shelf (inshore of some isobath? 200 m?), so that only the coastal water making it into the deep ocean is highlighted in color. Does the sentence “The along-slope coastal circulation has been removed from the analysis to isolate and highlight the deviations or pinching-off events” have something to do with the mask?
5. In the paragraph about potential instability of the along-coast flow (lines 157-164), a simple explanation is that when the coastal current is over sloping bottom topography the flow is more stable as water parcels tend to follow isobaths. So downstream of where the continental shelf narrows (the start of the eastern block), more of the flow is over the slope and stratification isolates it from the stabilizing influence of bottom topography. Offshore flow can also be caused by flow-topography interaction. I cannot tell from Figure 1 if there are coastline or bottom topographic features that might promote offshore flow. The authors should comment about this possibility.
6. Section 4.2 about spatial scales and temporal persistence is interesting. Regarding the spatial scale, how many eddies based on an estimate of the local internal Rossby radius of deformation can “fit” along the coast in the eastern block? This would help explain how many different structures might be observed.
7. I am not sure of the purpose of section 4.3 and suggest it could be omitted.

Minor/editorial comments:

1. Figure 1 caption: define AW; ShE not She
2. Figure 2: put a mark on the start location of the drifter
3. Figure 8 caption: “… (orange) blocks from 2003 to 2023.”
4. Figure 13: how was the top-to-bottom order of the seasons chosen? Not by total, since spring is the least of all. Why not order them by season: fall, winter, spring, summer?
5. Figure 14: “… chlorophyll images are overlaid by the average …”