The authors present a compelling study and a timely article on the challenge of using thermal neutron detectors for environmental research. The authors show both extensive modeling results and field observations. Some limitations of the study are acknowledged as a lack of field soil moisture data in the peatland but do not greatly impact the main results and conclusions. The authors also discuss and seem to solve some of the long-standing questions in the CRNS community about how best to estimate the coefficient(s) in the calibration function. The manuscript is well written and suitable for HESS. I have some moderate comments and minor edits needed before publication.

Moderate Comments

1. The figure legends and axis labels are too small. Please enlarge before publication.
2. L444. The authors argue that the thermal neutron footprint may be significantly deeper than the epithermal range of ~30-40 cm. If this is indeed the case additional profile sampling of soil chemistry is needed by the community in order to understand the distribution of trace elements (e.g. Gd and B) that may greatly impact the thermal neutrons. In particular, as the neutrons interact with more soil horizons (beyond the O and A typical for epithermal) soil chemistry may play a greater role. The authors point this out a little but should highlight the need by the community to sample more soil horizons for relevant epithermal and thermal neutron soil chemistry. Avery et al. 2016 and others have presented a nice lattice water dataset for the top 30 cm but it seems the community needs to expand this effort across CRNS sites and more soil horizons.
3. L475. The authors show the heteroscedasticity effect from local and far-field soil moisture changes on the thermal and epithermal scatterplots nicely in Fig 8A. Without the soil moisture data in the peatlands, the conclusion is somewhat more speculative based on GW depth but still compelling. However, additional CRNS sites with large-scale irrigation (60 ha) from center pivots may confirm this effect (CRNS sites exist in NE, KS, and IA in the USA with center pivots). As the center pivots water in pie slices over 48-72 hours they will create this near and far-field effect, particularly when compared against a rainfall event on-site. The authors could mention this experiment as future work needed by the CRNS community to help confirm the conclusions here.

Minor Comments

L 263. Need space “time seriesNET”

Figure 3. check legend and data time series in panel B? No red squares etc.

L538. Also, watch out for overfitting if only calibration data is available on campaign days. The CRNS community has found ~3 calibration dates are needed for robust N0 estimation. For 4 parameters you may need 12 or more calibration sampling days. That is a lot of digging :). It is already challenging to calibrate on multiple days if you have several CRNS sites.

There are three main result in the paper, (i) the simulations of footprints for thermal and epithermal neutrons, and how this feeds through to the measured neutron count for heterogeneous soil, (ii) measurements and simulation showing that the scatter plot of thermal and epithermal neutron counts can be used to characterise soil heterogeneity, and, (iii) the testing of different methods for calculating VWC. I found all these points interesting, relevant and suitably novel. Apart from clarification on Figure 6, I see no obvious problems - the rest of my points are suggestion or easily implemented clarifications.

Fig 6a: Why is there so little change in the R86 footprints in this figure (as measured by first soil contact) compared to Fig 3a? In the case of simulation set 3 (dashed lines Fig 6a) far field soil moisture drops from 0.7 to 0.2 which is very similar to the drop from 0.7 to 0.1 in Fig 3, but in Fig 6 we also have that the near field soil moisture is decreasing. Surely there should therefore be a larger change in the various footprints in Fig 6? In fact R86 actually decreases slightly in simulation set 2 between scenario 1 (wettest) and scenario 6 (driest)! Is there a mistake? Have missed something important?

Footprints: One question that remains in my mind is the practical relevance of simulated R86 and D86 footprints. This is especially the case for the thermal neutrons where the authors explicitly consider different definitions for the distance travelled by an individual neutron. But even for an epithermal neutron a choice is made to measure distance from the first interaction with the soil, rather than for example some weighted average of the distances from all interactions with the soil. This isn’t a criticism particular to this manuscript - it is a general practice when simulating R86 for epithermal energies.

One might hope that the R86 footprint would approximately have something like the following property,

N = 0.86*(p1*N1 + (1-p1)*N2) + 0.14*(p2*N1 + (1-p2)*N2 )

where N is the counts detected at the detector, N1 and N2 are the counts that would be detected if the entire area was mineral soil or peat soil respectively (with their own VWC), and p1 and p2 are the proportion of the landscape from within or outside of the R86 distance respectively that is mineral soil. This kind of reasoning is already alluded to around lines 425. But perhaps this can be quantified maybe using something like the equation above? Perhaps p1 and p2 could be estimated? N1 and N1 could be added to Fig 4? Would similar hold for the both thermal and epithermal footprints? One could even envisage using the above equation as a definition for a footprint radius if a simplified circular geometry (p1=1,p2=0) was employed for the mineral soil. In any case extra discussion would be helpful.

Fig 3: Perhaps a comment on why R86 for thermal neutrons as measured from the first soil contact isn’t in fact larger than R86 for the epithermal neutrons. I could imagine that as an epithermal neutron undergoes further collisions it will eventually reach thermal energies and will had further opportunity to travel from its initial soil contact – although I appreciate the picture is not be as simple as this.

Equation 6: This is an equally weighted normalised average of NT and NE. But its not clear at this point why this is done. It is explained that this combination makes the response have a “shallower slope” than NT, but one normally expects reduced sensitivity to be a bad thing! Perhaps the actual reason is a compromise between having a footprint more representative of the location in which the soil moisture sensors are installed (NT), and the better sensitivity of NE? There is additional explanation around line 545. Also, when using this “alternative approach 2” perhaps one needs to recalibrate the parameters a0, a1, a2, as I believe the original choice of these was made for the epithermal neutrons?

Minor corrections/suggestions:

Fig 3b: There’s a problem with the legend.

Lines 16 and 84: “spatial discretization” is supposed to be “spatial disaggregation”? Also line 84 could be clearer.

Section 2.2.1: Some of the details in this section are general to all simulations (e.g. the detector radius, the energies of the thermal/epithermal neutrons) and would therefore be better in section 2.2.

Fig 5: I can’t really see the reason to show both the blue lines and the green lines – they sum to 1.

Fig 6: Could be more easily understood if x-axis labelled by near field soil rather than scenario. This is especially because when reading the x-axis left to right it becomes drier which is the opposite way around compared to Fig3.

Fig 9: I can understand the authors might be pleased with this figure but why not simply add the simulation points to Fig 8a instead.

Line 433: Add reference to the Figure.

Line 272: I think the bandwidth should have time/frequency units? Partly I ask because, I think that if the smoothing is too intense your residual “noise” will actually contain some of the soil moisture signal. I therefore want a rough idea how much smoothing occurred. Not that I think excess noise causes a problem, given the result stated on line 512. And I never doubted that the different approaches where significantly different.

Overall, I believe this to be a well written report and a valuable area of research within the scope of HESS. The authors have tackled an important aspect of CRNS measurements that defines potential uncertainties of the sensor and suggests ways to improve its accuracy.

Fig 8 is powerful in demonstrating how this ratio between epithermal and thermal neutrons can change at sites with known heterogeneous soil moisture dynamics. I can see this being a useful metric to describe CRNS site heterogeneity in a simple way that can help a user understand possible site-specific impacts on soil moisture dynamics. I wonder if we should reconsider sensor footprint size once the Spearman rank correlation coefficient falls below a certain value?

Overall a good piece of work – comments below.

Moderate Comments:

L55: The authors rightly point out here that there are methods using the ratio of epithermal and thermal neutrons to estimate biomass in the sensor footprint (e.g. Tian et al., 2016). The site description seems to suggest that there is a uniform (spatial) biomass at the test site, it would be better to explicitly state this if true. The literature has shown the ratio of thermal and epithermal neutrons can be influenced by biomass changes (although research tends to be looking at this temporally rather than spatially), so knowing that biomass is spatially uniform at the site would be beneficial. On this point I feel a bit more discussion on possible impacts of spatially diverse biomass would make the paper more robust, considering research has shown biomass to impact neutron ratios too. The limitations are touched upon (L586) but an expansion on this (hypothesis of influence, future research ideas?) would benefit the paper.

L141: Three simplifications in the model are outlined here. A brief expansion on the impact the authors predict this may have on the simulation would be a benefit to the reader.

Minor Comments:

L60: Needs re-wording as it sounds a bit confusing currently: perhaps something like “However, the integration radius of thermal neutrons at the CRNS sensor can be expected to be much smaller (a footprint of approx. 35m)”

L109: Write the actual value for the material density of quartz next to the description.