General comments

Lane et al. (2023) presents the initial global geospatial dataset of non-floodplain wetlands by integrating the global floodplain and wetland datasets. The floodplain dataset is produced utilizing an existing algorithm and the MERIT Hydro dataset, whereas the wetland dataset is resampled and modified with greater precision, incorporating previous 500 m CW-WTD data and the GSW and CCI datasets. The authors evaluate the datasets in 21 CONUS watersheds, examining their locations and extents. The study provides valuable insights for hydrological and biogeochemical scientists investigating the impacts and functions of (eco)hydrological cycles. Nonetheless, some aspects require further explanations/modifications in the main text. So, I recommend major revision. Please find my comments in the following.

Major Comments

• What led to the authors' selection of Woznicki et al.'s (2019) machine learning-based floodplain dataset for validating their GFPlain90 floodplain? Are there alternative floodplain datasets available for cross-validation? Furthermore, why did the authors limit their watershed-scale comparison of their global wetland product versus the NLCD wetlands to only 21 watersheds? If feasible, it would be more beneficial to provide explicit validations for as many watersheds as feasible to demonstrate the accuracy of the wetland product. Why did the authors opt to compare their wetland and non-floodplain wetland products to the NLCD wetlands? To better understand the breadth of available regional and global wetland and floodplain datasets, I recommend that the authors include a table summarizing such datasets in their work.
• It is recommended that the authors add a section to thoroughly discuss the uncertainties and limitations associated with their proposed dataset. For instance, the authors mentioned that the various input maps capture different types of wetlands (Lines 241-246). Additionally, it is worth discussing why the CW-WTD dataset is used as the benchmark dataset for mapping the wetland data, considering that it has a low Pearson correlation value (r=0.34) with other wetland datasets such as GLWD and Hu et al. (2017b). Is it possible to use other input datasets? A table that outlines available alternative inputs/benchmark validations for the products proposed in this study could be included.
• There appears to be a lack of clarity regarding the methodology employed for generating the global wetland data. The rationale for not directly utilizing the original 30 m GSW, 300 m CCI, 500 m GIEMS-D15, and 1 km GDW for creating the global wetland data remains unclear. It is noteworthy that resampling the 500 m CW-WTD product, which was originally based on the 30 m GSW, 300 m CCI, 500 m GIEMS-D15, and 1 km GDW, to 30 m and subsequently incorporating any detected wetland pixels from the resampled 30 m CCI data and inundated pixels from the 30 m GSW data, may introduce discrepancies as compared to directly utilizing the original GSW, CCI, GIEMS-D15, and GDW for wetland mapping. Moreover, it is apparent that the current methodology potentially utilizes the information from both GSW and CCI datasets twice, though this may be open to interpretation. Nonetheless, it is acknowledged that this may require further clarification, and any potential misunderstandings on the author's part are sincerely regretted.

Minor comments

The subscripts in Table 1 require further clarification as they are not clearly understandable.

In Section 2.4.2, it would be preferable to have a table containing all these equations for the sake of conciseness.

Line 358: Is it necessary to include citations for all 7 references to define the Hit Rate? It would be more appropriate to include the most relevant reference for the definition in the article and please balance the number of citations throughout the paper.

Line 189: Please clarify the upscaling parameters.

Line 581, not clear “global population of wetlands"

Line 663: Please exclude this reference that is still under review.

Instead of presenting the results in similar tables (tables 2-5), it would be more engaging to include some figures that can get the reader's attention.

The statement in lines 261-263 about not losing any data when resampling to a finer resolution needs further clarification.

The methodology section in 2.2 about the global wetland data could benefit from further clarification. To improve the clarity, it is recommended to introduce the CW-WTD dataset first, followed by the description of the other datasets (RFWs and GDWs). This may help the reader better understand the analysis approach and the role of each dataset in the study.

This paper discusses the development of a publicly available Global NFW dataset, which provides a foundation for global non-floodplain wetland functional assessments. The dataset reveals that the majority of the globe's wetlands likely occur outside of river floodplains and coastal habitats, and is estimated to include over 16 million km2. The dataset will facilitate the inclusion of non-floodplain wetlands in hydrological, biogeochemical, and biological model development, and advance wetland conservation and resource-management goals. Overall, the manuscript is well-written and contains extensive discussion on the procedure followed by the authors. There are two main steps. First the identification of wetlands and floodplain at a global scale, and secondly the derivation of non-floodplain wetlands by subtraction. These two datasets (i.e. floodplain, and wetlands) had been already introduced in the past and rely on a mix of observations and algorithms. The wetland dataset is resampled from the original 500m to 30m, but there is not a discussion on how this might affect the results (the masking in particular). A second comment is about the temporal coverage of the Global NFW dataset, which I couldn't find in the main text. Then some more concerns are about the second part of the paper on the validation. Table1 is not clear to me, and perhaps could be removed. The metrics (i.e. Eqs. (1)-(8)) could be perhaps moved to the Appendix since are quite standard evaluation scores. Again there is a resampling involved in order to allow the validation and I would like the authors to discuss how this could impact the metric scores. It should be stressed that the datasets used for validation are not really ground truths since are themselves derived with models.  Finally, the authors report quite a significant discrepancy with respect to previous estimates of regional non-floodplain wetlands and I think it would deserve further justification and discussion. Some potential applications of the dataset (e.g. for hydrological model improvements) are mentioned at the end, but I would suggest to also discuss some possible use cases and key breakthroughs that the proposed dataset might be useful for.