Implications for Tracking SDG Indicator Metrics with Gridded Population Data

Round 1

Reviewer 1 Report

Combining with natural disaster cases, I find it interesting and innovative to compare five existing population data products. Unfortunately, there is no population data to do the validation. Although there are many global grid population data products, there are many differences among these products, and they are not consistent in different regions. This paper selects three case areas, compares the differences of different population grid data products in the three case areas, and carries out discussion and analysis aim to SDGs 11.5, and discusses the exposure of disaster affected population in the three case areas under different data. This study points out the problems existing in the use of current population data, and puts forward how to reasonably evaluate the affected population in the absence of field verification data. There are few problems:

1.“To identify urban versus rural population estimates across the five gridded population products, we use an urban-rural binary land cover classification derived from MODIS data – the MODIS global urban extent product (MGUP) . ” and have the conclusion “Vector based administrative-level population fails to disaggregate population at the spatial scales requisite to identify where people actually live on the planet and thus fail to provide use information for the delivery of services required to achieve the SDGs.” However, the quality of this dataset is not good enough for the urban area analysis. It should be clear that it might not only the problem of the population data, but also the urban area data.

2.Figure 5 make me very confused, after reading the manuscript carefully, here is the suggestion for it: (1) In the legend of Figure 5, “GHSL-15” should be “GHS-15” ; (2)“Intesnity”should be “seismic intensity”;(3) why there is (a),(b),(c)? It's not shown in Figure 5. (4) The description of total population and rural population in the figure is not very clear and easy to be confused.

3. in Table2 and Figure 8: what does admin 4 and admin 3 meaning? What is the difference？is it the different level of administrative? It would be better to make it clear.
4. if possible, it should use the census data in these three cases to compare with the results of the five gridded population datasets.
5. Unfortunately, the paper points out that there are some differences and problems in population data, which need to be paid attention to in relevant applications, and the authors tried many ways to reduce the errors which come from the data. But it is hard to draw the final conclusion.

Author Response

Please see the attachment.

Author Response File: Author Response.pdf

Reviewer 2 Report

The paper "Implications for tracking SDG indicator metrics with gridded population data" aims on analysis how grid method on population datasets impacts SDGs to advance the “fitness for use”. The paper relies on the topic of Sustainable Development Goals (SDG) which is quite popular and modern topic nowadays in both academic and non-academic research.

I would like to congratulate authors to an excelent paper. 
The paper is well structured, it follows an academic structure, it combines.
The workflow and methodology is not unique since grid method is used routinely in the field of GIS by researchers. On the other hand the combination with SDG and nature hazards provides extremly interesting results. Since results are available for public, I do not have any significant negative comments (see three comments below).

I miss just a brief introduction about grid method - it is an appropriately chosen method for this purpose, but can you add one paragraph about the method in general + benefits + comparison (or the reason why you use grid instead some of) another methods, e.g. heat-maps

Fig 1 - add a scale to a smaller maps

Fig 2 - use the same scale of all maps/countries when it aims on compariosn

Author Response

Reviewer 2

The paper "Implications for tracking SDG indicator metrics with gridded population data" aims on analysis how grid method on population datasets impacts SDGs to advance the “fitness for use”. The paper relies on the topic of Sustainable Development Goals (SDG) which is quite popular and modern topic nowadays in both academic and non-academic research.

I would like to congratulate authors to an excellent paper.

The paper is well structured, it follows an academic structure, it combines.

The workflow and methodology is not unique since grid method is used routinely in the field of GIS by researchers. On the other hand the combination with SDG and nature hazards provides extremly interesting results. Since results are available for public, I do not have any significant negative comments (see three comments below).

We are thankful for your review and encouraged by your positive critique. We have updated the text and figures per your recommendation.

I miss just a brief introduction about grid method - it is an appropriately chosen method for this purpose, but can you add one paragraph about the method in general + benefits + comparison (or the reason why you use grid instead some of) another methods, e.g. heat-maps

We appreciate your recommendation and we have added a sentence to our introduction to clarify why we focus on gridded population products. Per recommendation of Reviewer 3, we have opted to minimize over-explanation and technical explanation of each gridded population product. We point readers in Section to 2.1 to two excellent recent publications that systematically review and explain gridded population products in more detail, as well as www.popgrid.org. 

Fig 1 - add a scale to a smaller maps

We have updated Figure 1 per your recommendation.

Fig 2 - use the same scale of all maps/countries when it aims on comparison

Thank you for this recommendation. We, however, have retained the original Figure because if we set each panel to the same spatial resolution, MMZ will be twice as big as the other panels. We did update the caption to note the different scales of the panels.

Reviewer 3 Report

This manuscript describes the comparison of various sources of gridded population data applied to three locations affected by natural hazards: earthquakes in Nepal, hurricanes in Mozambique/Malawi/Zimbabwe, and floods in Ecuador. The goal is to identify the variability inherent with these population data as applied to SDG 11.5.

I found the level of technical details on remote sensing analysis too extensive and unnecessary for this cross-disciplinary journal and the main goal of the manuscript: identify the shortcomings of population estimates based on EO sources. Technical terms (e.g., gridded population, susceptibility decile) and acronyms that are not known to non-remote sensing/geography specialists (e.g., LMIC, MGUP) should be properly defined and spelled out more frequently to avoid having to scroll back to the first mention. The level of detail in sections 2 and 3.1 is excessive and appropriate for a technical journal, but distracting from what is the main purpose of this manuscript submitted to this journal. This lengthy and detailed comparison of different gridded population data sources shows that there is variability, sometimes a significant one, although I do not know if the term has a statistical meaning in this study since the comparison is based on visual observations of bar graphs. 

In the discussion, the authors recommend using multiple sources and share their code and datasets but did not convince me that this level of detail is really necessary from the point of view of researchers, stakeholders, and decision-makers who have the specific goal to assess and minimize the risk to populations based on specific natural hazards. 

In the discussion section, we are introduced to ‘constrained’ and ‘unconstrained’ gridded population products, and the pros and cons of each one, and it seems that one unconstrained product is better than the others. Are the products the authors used constrained or unconstrained? Why using multiple ones when there is one that is better than others? 

A further recommendation in the discussion is to include the uncertainty in the population data used by giving ranges: the authors do so for specific locations and quantified the potential differential financial cost associated with e.g., Cyclone Idai in MMZ . Still, after this detailed technical evaluation, I am left wondering: from their expert point of view, is there a ‘better’ product for one hazard, or type of setting (urban, rural, coastal, mountainous)? They all seem to give incorrect estimates based on different regions (e.g., urban vs rural) or hazard type: so how does this paper help researchers and stakeholders in tracking SDG 11.5? I would recommend that the authors significantly simplify and shorten the paper and focus on the potential usefulness of this study for the endusers that it is focused on.

There are some grammatical errors throughout the manuscript, e.g. in the caption of Fig. 2 (Spatial agreement of if a pixel is inhabited?) and Fig. 6 (wind speed and not winds speed, also in lines 431 and 438), line 434 (‘exposed’ and ‘population exposures’), and line 527 (case studies). The caption of Fig. 5 mentions a, b, and c but there are no letters in the figure; the label of the x-axis is also misspelled.

Author Response

Please see the attachment.

Author Response File: Author Response.pdf