The paper introduces an underwater camera-trapping device (UCT) and discusses/analyses corresponding data, in particular on newts, from a survey employing the camera.

The title suggests that the camera is suitable for freshwater monitoring (ie general) but then the focus is almost exclusively on newts. I suggest that either the title should be adjusted, or the writing and examples in the paper, or a combination of the two.

Authors refer throughout to species “hardly” surveyed. Do they mean species that are difficult/challenging to survey? Because hardly surveyed can be interpreted as species that are rarely surveyed, which is not the case of course for newts. There is a growing literature looking at population dynamics and behaviour that goes back a long way and a growing literature on image analysis in amphibians. For example:
https://doi.org/10.1111/1365-2656.13208
Daily and seasonal activity patterns are already well-known through these and other studies.

Top of page 6: this paragraph makes quite strong claims, eg most existing approaches focus on estimating the number of adults, or provide information of limited usefulness etc but has just one reference.

First paragraph of results: challenges that need to be discussed: bug fixing, size of corresponding data, missing data; these should be discussed much more in detail, as they are crucial points for employing the camera for monitoring in practice.

This is not a research paper of course, but why was the choice made to only consider full operating days and not account for the exposure rate?

First paragraph of Section 3.2: more discussion about how the 1000s of photographs need to be sorted to species and then managed/uploaded for further identification at the individual level is needed. Again, in practice, this is a major challenge for field employment of this camera.

The trial discussed in the paper generated over 98K images/videos and 21K of sequences to analyse. How did the authors analyse such an enormous data set of images and how long did it take? Did they download the lot into Amphident software? An AI system to interrogate such a data set is needed if this is to be of practical use.

End of first paragraph of Section 4.1.2: it would be interesting to compare with studies that also collect CR data, and there exist several such long-term studies. Has there been an attempt to compare results?

Line 258 discusses an interesting research direction on monitoring larvae: the challenge for tracking larvae and juveniles is not about a method to catch them in ponds as they enter bottle traps, but more due to the fact they may disappear on land for 1-3 years during which their belly pattern changes. By the time they return (if they return) their patterns have changed, making it difficult to match them to earlier detections. How is this UCT helping with this challenge?

Related to above, line 275 mentions limitations on storage capacity, power consumption and maintenance: how serious are these limitations at this stage and how likely are they to be overcome for this UCT to become widely employed?

Another major issue is about study design. Landscape-level studies deploying 100s of bottle traps and/or eDNA samples across multiple microhabitats within a pond are now becoming more common. So the UCT may be useful for small ponds where limited replication is needed across microhabitats but the cost and logistics of replicated sampling in larger ponds and multiple water bodies may be prohibitive. This aspect (of cost to purchase the camera and logistically for transporting and deploying the UCT across many different locations) needs to be discussed. For example, the paper does not discuss weight of the equipment or power supply, so portability and deployment in remote locations for long periods is key.

Several images/plots missing towards the end of the document

The authors have done a thorough revision of their manuscript and the provided response is transparent and fair, so I do not have any further comments.