esurf-2020-72
The effects of topography and soil properties on radiocesium concentrations in forest soils in Fukushima, Japan

This manuscript received three reviews and I have read it in detail myself. Two reviewers were broadly supportive, but reviewer two was sceptical, stating that the motivation of the study was poor. Reviewer one, was positive, though makes a number of important points, and doesn’t comment explicitly on the statistical methods. Reviewer 3 comments critically about the vertical resolution in relation to the samples collected.

Whilst this is undoubtedly an interesting problem and data set I have to agree with reviewer two that the motivation for the study would need to be much more clearly conveyed if this is to be published. An interesting data set and statistical analysis is not sufficient for publication. The bar is high in terms of how this needs to be presented if this is to be understood by the broad audience of ESURFD. In the authors response, I don’t think they addressed sufficiently the comments of R2. It is not reasonable to assume that readers will be familiar with all the statistical methods as well as the radiocesium literature.

In summary, this manuscript takes a statistical approach to determine whether geomorphological parameters influence the accumulation of radiocesium following a major nuclear disaster. A number of samples were collected over a 3 year period following the disaster from a relatively restricted area, over a relatively narrow range in altitude. The radiocesium was determined and a number of statistical models are used to test whether there are links with key topographic features such as elevation, slope, curvature and vegetation type. It is never really laid out to the reader why 137Cs might correlate with such topographic parameters (the list on pp17 is the best but it should have come much sooner). What are the underlying physical/chemical/biological mechanisms (they get lost in the text and should be stated right at the beginning). On the one hand altitude might relate to the original deposition of 137Cs (orographic effect), and vegetation type (and elevation) might relation to physical/biological/chemical processing of 137C post deposition. Why should slope and curvature matter (this is briefly stated but not elaborated)? If these are the key parameters that control the accumulation (retention) of 137Cs in the environment then some clear figures with the data are required to show this both in an individual and multivariate manner. The current figures do not do a sufficient job of conveying a clear message from the data and are set out in a disorganised manner. There are too many figures for the message that is being conveyed. There would be clearer, simpler more concise and convincing methods to elucidate a complex yet fascinating dataset.

There are a series of major issues that would need to be addressed if this were to be considered further for publication. Based on the author response to the initial reviews I’m not convinced that such a whole-scale rewrite is possible. There are organisational, methodological and grammatical issues which must all be addressed. As is, this does not read like a polished paper, but more like a detailed thesis or specialist report. Unfortunately, in the version I read, the figure numbering appears to have gone wrong which doesn’t help. It may be, that it would be more appropriate in a specialised journal.

Specific comments (not exhaustive)
1. Generalized additive models (GAM)
The statistical models (for example the Generalized additive models (GAM)) seem complex for what the data shows. The GAM method is not well explained. What exactly is being minimised to generate the residuals? I’m not sure the text and figures do justice to the data? A large number of the figures (and there are too many) show deviance or residuals. It is difficult to ascertain from these figures the validity of the proposed covariation or not. I don’t think it is sufficient to say that the “gam.check ()” function was used. If readers aren’t familiar with R or this particular function in R, this will convey very little.


There is no obvious or particularly intuitively relationships between 137-Cs and topographic parameters. For example, in section 6.2 accumulation patterns on a simple representative hillslope, 3 figures are shown. The first appears to show the relationship between elevation and 137-Cs. It is clear that there is no simple relationship. Perhaps this is why a complex GAM is required. What is the trend line on this diagram? What is the simple representative hill slope? The second figure appears to show slope vs elevation, but how does this relate to 137Cs?


Perhaps clearly subheadings should guide the reader through the data analysis in a more systematic way such as:
I) Elevation
II) slope
III)soil water content
IIII) density

At the moment, the data description and discussion appear to be driven by the statistics (four tests), rather than clearly testing a hypothesis with a clear physical or geomorphological basis. If the 4 tests were retained, they should be laid out much more clearly so the reader can follow. An example of how I was unable to follow even the Single parameter GAM results is given below:
Page 28:
As I understand section 6.2, it is examining individual parameters as a controlling factor on 137-Cs. It is a pity that it is not possible to indicate the lack of correlation (as I understand it) with a figure. However, what I really don’t understand is the following sentence:
“None of the single parameters in either table returned deviance-explained percentages above 36 % (which was found via a model including water content).”

I don’t understand how the water content is being modelled at this point. How the single parameter GAM works has not been well enough explained. Being single parameter, I had clearly wrongly assumed that it meant it was dealing with one parameter at a time.

2. Section 4.1 and 4.2: DEM work
This section seems to deal with extracting key parameters from the different resolution DEMS. However, it also seems to mix in 137Cs measurements (line 281). This doesn’t seem like the right place to know the units of 137Cs. The work on the DEM resolutions is clearly important, but equally it is a distraction from the main message of the manuscript. Can section 6.3.1 move to an appendix? Equally, the organisation seems to be confused. Later on in the manuscript, Fig 15 (Slope degrees of sampling points along elevation (m)) seems to be entirely related to the DEM work and have nothing to do with radiocesium. There are clearly major organisational problems.





3. Figures
There seems to be a major problem with the numbering of the figures.
Figures should be understandable just by looking at the figure and the caption. For example, Fig 12 (pp36). What are the solid red lines. In the caption, it refers to top row graphics and bottom row graphics. What are these?

Fig 4: It is a shame that there is nothing higher resolution available than this. As is, it is not very satisfactory. Could this not be incorporated into fig 5, plotting the sample locations in 3 different colours (if I have understood the figures)?

On Fig. 15, what are the red dots. It is not clear from the Fig or the caption.

Line 432: “In Fig. 13, both trend lines of Cs-137 in Bq m-2 values show upward trend along elevation.” Fig 13 appears to show depth vs 137Cs and COV.

Figure 3. An aerial views of the study site, facing northwest.
This would be fine in a Phd thesis, but is not appropriate in a paper. In addition, it does not appear to be an aerial view as is stated.

Figure 6: The low water content appears to contradict the 100% statement (referred to below). There appear to be a large number of outliers on the high side. Would this be best shown as a histogram? It is hard to visualise how much data is in the box.

Figure 8: Nice to see a photo, but is it needed? What does the photo convey scientifically? This would be fine in a thesis, but not a paper.

Can figures 9 and 10 be combined with fig 8 that comes after 9 and 10? Or combine 8, 12 and 13, the interesting depth figures.

Fig.17: Hard to know what it means.

Can Fig. 9, 10 and 20 be combined? (Pp30 and 31)

The Fig on pp35 appears to be very useful, comparing model predicted 137Cs with data. A pity there is no fig number or caption.


4. Writing style and conciseness:
There are numerous examples where the writing style does not have the conciseness required for a scientific publication.
Avoid phrases like “This section displays” and make a point directly.

Other examples (non-exhaustive):

“Literature review” Seems like an odd heading for a scientific paper. Sounds more like a thesis
“When driving into the study site region, visitors see hills and mountains covered with forests. Winding, narrow roads connect”. Not appropriate for a concise scientific paper
“natural” or “native”. Choose one option
Line 140: Land use comments: This is not the right place for this comment. This section is about the study site. If and use matters, then include it. If not, then say it doesn’t
Line 157: perhaps change to “to confirm anomalies observed in previous years…”
Line 159: change to “but mostly on the south-west side due to accessibility”
Liner 190: Particles larger than 2mm were removed with a sieve
Line 198: The unit of measurement is the kiloelectron volt (KeV), equivalent to the kinetic energy gained by an electron falling through a 1 volt potential.
Line 215: “Approximately less than”. Just less than
Line 217. Missing full stop and space.
Line 226: “Only some of the samples were tested for texture because of time and human resource constraints during the limited lengths of the first author’s stays in Japan.” Remove. Informal detail not needed.




Other
Line 433: Camelback. This is not scientific terminology. The manuscript goes to a great effort to produce a set of descriptive statistics to define relationships between parameters. This shouldn’t be undermined by colloquial vocabulary.

Center of gravity (COG) depth (cm): What is this?

“Migration head depths” are not defined. This appears to be first mentioned on line 413.


Soil properties are discussed, but not in great detail. What about the cation exchange capacity of the soils for example that Cs has a very high affinity for. Presumably the cation exchange capacity is thought to be constant in these soils. Is this also true with depth. It seems to me that one major variable that has been overlooked in this analysis is the chemistry. How much of the Cs is partitioned between aqueous Cs and the soil?

How can the average water content>100%. This was also asked by reviewer 1. I don’ understand the author response. How can the moisture be in excess of 100%. I must be mis-understanding the definition.

Dear Dr Yasumiishi,
We have now received three reviewers reports in addition to my own detailed comments on this manuscript. This is clearly an interesting problem but requires some substantial modifications for it to be accepted for publication. These modifications are more significant than the revisions that were made after review. Given the nature of these revisions, your manuscript will most likely require an additional round of review.
Best regards
Ed Tipper

The authors have worked really hard to take on board previous comments and it really shows. The result is a coherent, interesting manuscript that is now well on the road to being published.

The premise of the manuscript is now clear; to model statistically a multivariate dataset. The GAM is generally well explained and understandable to the non-specialist. Model A seems to go a reasonable job at capturing the key aspects of the dataset.

One thing that I think would prove a very useful additional is a small section explaining why the authors think water content, bulk density and elevation are the critical parameters from a mechanistic point of view. Why is density more important than slope or wetness for example?

In addition to the comment above, I have annotated the pdf with a series of minor comments that the authors should find very straightforward to address.

The paper is now essentially ready for publication. The only suggestion I would have is that the authors include a sentence near the end of the discussion along the lines of that which is included in their response letter in response to the following point:
"The referee asked that we include thoughts as to why water content, bulk density, and elevation are important parameters for consideration. As noted herein, the statistical analyses employed are empirically-based approaches to derive predictive relationships amongst the available data. That is, the results are wholly dependent on the limited data collected. These analyses do not provide a mechanistic explanation as to why specific parameter combinations were shown to be more important. The discussion of these results sought to emphasize both the power and limitations of this approach while minimizing any discussion that would be considered speculative."

I think this answer is satisfactory, but that a version of this be included for the benefit of all readers.

Dear authors,

I am pleased to confirm provisional acceptance of your manuscript for publication in esurf based on the recommendation of the reviewers and Associate Editor (AE), Ed Tipper. Thank you for the effort you put into revising your manuscript; it has improved a great deal as a result and should provide a valuable resource for those interested in radionuclide contamination as well as for researchers using radionuclides in geomorphology.

Before final acceptance, please add a version of the text suggested by the AE related to the statistical rather than mechanistic nature of your analysis. Also, please note some typos and wording suggestions on the attached file. Importantly, I also found that the font size on many figure axis labels was too small to be readable. Please review all of the figures and make sure that you increase font sizes so that all text will be readable in print layout.

Once you have completed these changes, you should be able to upload your final manuscript files for typesetting.

Thank you for choosing esurf for your work!

Josh West