Overall, this paper is a useful addition to the literature. Better characterizing greenhouse gas fluxes from tropical soils, and identifying drivers, is a timely and important research question. The combination of automated chambers and eddy-flux, especially for N2O, is very novel for tropical forests. Overall, given the well-known spatial and temporal heterogeneity of N2O fluxes even in much more homogenous ecosystems, I am not surprised that fluxes were not well explained by simple environmental variables, even with the data density of this paper. This is a dataset that will certainly be of interest to many people. Several of the methodological lessons learned (such as lack of storage of N2O and Ch4 under the canopy, revealed by vertical profile measurements) are also likely to be of use to other researchers.

I have one significant concern that I think merits serious consideration- the choice to exclude high fluxes from analysis (and indeed not to present them at all, making it very difficult to judge how important they might be). Specifically, the authors did not present (or include in analysis) any fluxes outside the 5^th-95^th percentile (per line 248), even after rigorous data cleaning steps that should have weeded out any anonymously high fluxes that were methodological artifacts. I realize that it is very difficult to scale rare, high fluxes without very good estimates of their probability. However, I do not thing that dismissing them entirely makes any sense, and no specific rationale or citation was given for the choice. Rare, very high N2O fluxes are not at all uncommon in tropical forests in my personal experience, but rarely do we have the data density (as we do here!) to judge their potential importance. These excluded fluxes- depending on their magnitude, could be potentially important for net ecosystem emissions, especially because they’re probably quite skewed- extreme production events could be somewhat common but extreme consumption events likely are not.

To summarize- I am not necessarily suggesting that all data needs to incorporated into scaling, but I would strongly suggest 1) presenting the relative magnitude of the excluded fluxes compared to the data that was included. Were they common and somewhat high? Or more rare and extremely high? 2) At least conducting some sort of sensitivity of means, medians etc to the inclusion or exclusion of these omitted values (including perhaps making a range of assumptions about their probability, in the case of very high outliers). I also would hazard, and add caveats, against comparisons with any other rate fluxes from tropical sites that may indeed have included hot spots and hot moments in their scaling efforts. Overall it seems counterintuitive to highlight the heterogeneity of soil GHG fluxes and then ignore a potential large fraction of the variation. 

Finally, data should be posted in an accessible database online in keeping with the specifications of this journal (rather than ‘on request’).

Tropical forest in general are still largely understudied and to understand possible future pathways of these extensive forests, first the current state and drivers need to be understood, which is where studies like this come in play. Long term data sets of ecosystem and soil fluxes especially of non-co2-greenhouse gases are rare. This study therefore delivers an important and useful contribution to the study of tropical forests. The combination of Eddy Covariance (EC) and automated soil chamber measurements for N2O and CH4 in a tropical forest is novel and certainly not trivial to accomplish. The highly variable fluxes with a significant seasonal effect is as expected from N2O and CH4 fluxes in tropical forest.

I agree with referee #1 that the removal of flux values outside the 5th - 95th percentile flux range is a concern. I think it at least needs a strong reasoning and explanation in the text as why the authors chose to do this and which values were removed. Also the sporadic high CH4 emissions and the many N2O uptake events are not discussed, even though they are of interest.

In the attachment I add some comments, a few questions and some suggestions to improve readability and/ or interpretation.