7
$\begingroup$

This might sound like a silly question to many of you, but I am addressing the comments of a referee regarding a study I'm trying to publish and I remained a little puzzled about this issue.

Let's assume I have set up a simplified model that processes biophysical and socioeconomic data to analyze an issue attaining water management. Biophysical data (mainly topographic, climatic, and hydrological) are available at relatively high resolution (0.25 or 0.5 degree), while socioeconomic data are available at country resolution. The model was set up to run at gridcell resolution (0.25 degree), but, of course, all the cells of the same country share the same values of the socio-economic variables. However the process is run for each individual cell and the results are different from cell to cell, even in the same country.

In my manuscript, I claim that the resolution of my study is a quarter degree. This sentence is contested by a referee that states: "The resolution of the results of a given study is actually defined by the coarsest dataset that the study uses (not the finest). Part of the data used are available only at country level, so the spatial resolution of their results is at that level, not finer."

Now, I understand that coarse biophysical data (precipitation, temperature, land use, ....) should be eventually downscaled to be used in a gridded modeling approach, but what about variables like "corruption" or "quality of the institutions"? Does a downscaling of such variables make any sort of sense? What finally determines the spatial resolution of a study?

Thank you

$\endgroup$
  • 5
    $\begingroup$ While the reviewer makes a good point that is difficult to argue against, I think you could get around it by being more specific in the way you describe your resolution; as you have done in the question. I think at that point you are allowing the reader (not the reviewer) to make the judgment on the utility of the results. $\endgroup$ – arkaia Jan 8 '18 at 17:10
  • $\begingroup$ To what degree do the results of your study depend on the socioeconomic data? It could be negligible, or it could be critically important: e.g. trying to get a meaningful answer for say California by averaging the socioeconomics of East LA, Bakersfield, Palo Alto, and Alturas would not be expected to produce useful results. $\endgroup$ – jamesqf Jan 8 '18 at 20:26
  • $\begingroup$ You might consider an approach to resolve your socio-economic data finer. Maybe you can distribute on the base of population density or of urban/rural areas (land use). $\endgroup$ – daniel.neumann Jan 8 '18 at 21:31
4
$\begingroup$

It sounds to me like the reviewer is fixated on a 'rule' they encountered in grad school, but the rule ignores some real-world nuance (as rules do).

Quick caveat: my answer isn't definitive, it's more of an opinion, but it was too long for a comment.

It seems intuitively true that some phenomena just are smooth, so the resolution of the data is not the limiting factor in an analysis of resolution. I'm thinking of things like maps of average seasonal air temperature (say). Certainly, you can combine that with higher-resolution data and still claim to have a high-resolution analysis.

Other data may not be implicitly smooth, but it's interesting to aggregate the data for the purposes of generalization.So the quality of academic institutions obviously varies on a per-institution basis (or per teacher, if you want to get silly about it), but it would be crazy to interpolate data like that and, since it's reasonable to think there might be national-level influence (e.g. in funding for education), it only makes sense to aggregate it that way. Again, I think it's reasonable to argue that you can combine data like this with higher-resolution data and have a high-resolution analysis overall.

Still other data may simply be oversmoothed, or smoothed on the wrong basis. Temperature data by country, or geological data on a 100-km grid... Then I'd say those are limiting the resolution of any analysis you do.

In conclusion...

I can't think offhand of a way to formalize all this, and don't know of any research around it (though I'd bet on its existence). Assuming you don't want to go down the research rabbit hole, or have a protracted argument with the reviewer, my advice is to respond to the objection by using more specific language. Instead of generalizing, specify the resolutions of the various datasets involved and mention where it might be possible to improve the resolution.

$\endgroup$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.