Next Article in Journal
Quantifying Citrus Tree Health Using True Color UAV Images
Previous Article in Journal
A New Set of MODIS Land Products (MCD18): Downward Shortwave Radiation and Photosynthetically Active Radiation
Previous Article in Special Issue
Human Lights
Open AccessFeature PaperLetter
Peer-Review Record

Nighttime Lights and Population Migration: Revisiting Classic Demographic Perspectives with an Analysis of Recent European Data

Remote Sens. 2020, 12(1), 169; https://doi.org/10.3390/rs12010169
Reviewer 1: Anonymous
Reviewer 2: Anonymous
Reviewer 3: Anonymous
Reviewer 4: Anonymous
Remote Sens. 2020, 12(1), 169; https://doi.org/10.3390/rs12010169
Received: 27 November 2019 / Revised: 30 December 2019 / Accepted: 31 December 2019 / Published: 3 January 2020
(This article belongs to the Special Issue Advances in Remote Sensing with Nighttime Lights)

Round 1

Reviewer 1 Report

My previosu comments and suggestions have been taken into account by the author. Many Thanks.

I have no additional issues about this paper.

Author Response

No further reply. 

Reviewer 2 Report

Thanks for adding all my suggestions.

Author Response

No further responses. 

Reviewer 3 Report

This version responds adequately to earlier my earlier criticisms.

Author Response

No further response. 

Reviewer 4 Report

Dear authors,

 

Your paper on “Nighttime Lights and Population Migration: Revisiting Classic Demographic Perspectives with an Analysis of Recent European Data” is relating VIIRS nighttime light images to the migration process in the European Union. In general, there is not much work done within Europe to relate NTL image and demographic trends, therefore, I found the paper interesting. However, the following major problems are presently found in your paper:

 

Abstract: the abstract would require a substantial revision. The abstract should be clearer on the model used - what is the dependent and independent variable(s) and what are the numeric results!?

 

Within the introduction, it is not sufficiently clear why you do this analysis within Europe – and at which level (scale) is it done - what are the data gaps you aim to address. In general, in Europe data are available.

 

In addition, one part of Ravensteins theory is that cities growth more on migration than on natural growth. This could be added.

 

Section 2.2 and in general: data are not sufficiently described and methods are not at all described. What are the dependent and independent variables? Why are national boundaries important and not only the regional levels? In Europe migration happen to differ from regions to other regions not because of countries.

 

Section 2.2 and results in the methodology section NUTS 3 is stated to be the analysis unit but in the result section, everything is on countries?

 

Section 2.2. and results: for the regression model, did you check whether the relations are linear?

 

Line 263: what are pooled country data?

 

Line 268: Please keep in mind that light is not driving migration in Europe.

 

Author Response

Please see the attached responses

Author Response File: Author Response.docx

This manuscript is a resubmission of an earlier submission. The following is a list of the peer review reports and author responses from that submission.


Round 1

Reviewer 1 Report

1) The second paragraph of the Introduction needs clarification.

2) Multicollinearity leads to inefficient, not biased, estimates.

3) The arguments made by the author against using a pooled analysis are countered by the pooling process. The use of country dummies controls for idiosyncratic variations among them.

4) Adjusted R-squares are not useful here.

5) The individual country models do not appear to account for spatial autocorrelation among the observations. This is a fatal flaw in the methodology. Unless space is directly considered we cannot trust the standard errors of the estimates (again an effective inefficiency problem). Further, the lack of incorporating space in the model can even lead to incorrect signs on the estimates, which the author simply ignores in the current analyses.  The author needs to use either a spatial lag or spatial errors regression method in order to make this paper interesting. Otherwise the readers cannot be confident in the results or conclusions of the analyses. 

 

Author Response

1) The second paragraph of the Introduction needs clarification.

---Paragraph has been revised and clarified.

2) Multicollinearity leads to inefficient, not biased, estimates.

---The section on multicollinearity is revised.

3) The arguments made by the author against using a pooled analysis are countered by the pooling process. The use of country dummies controls for idiosyncratic variations among them.

---A model using pooled country data only estimates one coefficient of lights – the averaged coefficient for all countries, even if country dummies are included. As we know, human migration is a country-specific phenomenon, strongly influenced by history, culture, government policy, political environment, and even national sentiment (Zimmermann, K. F., Bauer, T. K., & Lofstrom, M. 2000).  Thus the analysis by country is more appropriate and can provide more information on country-level parameters.

4) Adjusted R-squares are not useful here.

---Adjusted R-squared indicates the model fitness. Usually SER is reported. However, it is difficult to compare SER across countries, as total migration population vary substantially by countries. It is easier to compare adj R2 in prediction models.

5) The individual country models do not appear to account for spatial autocorrelation among the observations. This is a fatal flaw in the methodology. Unless space is directly considered we cannot trust the standard errors of the estimates (again an effective inefficiency problem). Further, the lack of incorporating space in the model can even lead to incorrect signs on the estimates, which the author simply ignores in the current analyses.  The author needs to use either a spatial lag or spatial errors regression method in order to make this paper interesting. Otherwise the readers cannot be confident in the results or conclusions of the analyses. 

---As I explained above, individual country analysis is appropriate, as migration is a highly country-specific phenomenon. When the assumptions of OLS regressions are met, and sample size is large enough, the estimates and the sign of coefficients are correct. In the case of multicollinearity, if it is not severe, the standard errors of the coefficient estimates are still correct. That is why in the second half of the paper, the separate models for lights and population are presented.

Most studies that introduce lights into a new field, as this study attempts to do with migration, start with standard linear regression models. I agree with the reviewer that including spatial models has merits, yet it should be done in a more advanced phase of investigation and cite more appropriate literature.  The primary goal of this paper is to introduce traditional migration theories to the RS audience, and to test whether lights can provide new info to these classic ideas. Spatial analysis on migration is a topic that needs a full new paper to discuss. This is beyond the scope of current manuscript.    

I suggested that using spatial models should be the focus of future investigation in discussion.   

Author Response File: Author Response.docx

Reviewer 2 Report

The author has presented an interesting study of relating the VIIRS monthly and annual nighttime data with population migration in the EU countries. The study is interesting and innovative. I have a few comments in the attached pdf. 

Comments for author File: Comments.pdf

Author Response

---The MS is revised based on all suggestions in the review.

Line 207: I don't understand the data under these two column headings. The 'Lights' column is not the sum of the annual lights of 2015 and 2016? What is the 'Year 2016' column then? Please clarify.

---The lights data is annual lights for year 2015 and 2016. Because the model also controls the year effect, and year 2015 is treated as the reference year, the coefficient of year 2016 is estimated and reported in the table. The negative value in column Year 2016 means that, everything else equal, the migration population size in 2016 is lower than that in 2015 for a particular country. The positive value means that, everything else equal, migration population in 2016 is higher than migration in 2015 for that particular country.

Reviewer 3 Report

Dear author,

the topic is very interesting but feels that the analysis done to the images is too naif for the complexity of the problem. The first problem that the author should consider is that during the period of the analysis, there are physical changes on the light technologies that are not connected to migration flows that will impact the date in a dramatic way.

The change to LEDs produces a decrease of lights that is not real but will impact your data.

You can find more about this here: Kyba, C. C., Kuester, T., De Miguel, A. S., Baugh, K., Jechow, A., Hölker, F., ... & Guanter, L. (2017). Artificially lit surface of Earth at night increasing in radiance and extent. Science advances3(11), e1701528.

Román, M. O., Wang, Z., Sun, Q., Kalb, V., Miller, S. D., Molthan, A., ... & Seto, K. C. (2018). NASA's Black Marble nighttime lights product suite. Remote Sensing of Environment210, 113-143.

To avoid this problem, instead of using the raw information, might be better to use the lighted area that the full intensity of the images.

Also, is known since a long time that GDP relationship with lights do not work in Europe, and cultural factors can be more important:

Falchi, F., Furgoni, R., Gallaway, T. A., Rybnikova, N. A., Portnov, B. A., Baugh, K., ... & Elvidge, C. D. (2019). Light pollution in USA and Europe: The good, the bad and the ugly. Journal of environmental management248, 109227.

Levin, N., & Zhang, Q. (2017). A global analysis of factors controlling VIIRS nighttime light levels from densely populated areas. Remote sensing of environment190, 366-382.

Kyba, C., Garz, S., Kuechly, H., de Miguel, A., Zamorano, J., Fischer, J., & Hölker, F. (2015). High-resolution imagery of earth at night: New sources, opportunities and challenges. Remote sensing7(1), 1-23.

Bennie, J., Davies, T. W., Duffy, J. P., Inger, R., & Gaston, K. J. (2014). Contrasting trends in light pollution across Europe based on satellite observed night time lights. Scientific reports4, 3789.

Sánchez de Miguel, Alejandro (2007) Differential Photometry Study of the European Light Emission to the Space. In World Conference in Defence of the Night Sky and the Right to Observe the Stars, 20-23 april 2007, La Palma.

Also, you have to notice that countries like Iceland and Finland have a significant contribution of Aurora lights.

Specifically, about the culture, I recommend you to read the Annex of: Alejandro Sánchez de Miguel. (2018, June 14). Spatial, Temporal and Spectral Variation of Light Pollution and its Sources: Methodology and Results (Version 2). Familia Sánchez de Miguel, Madrid. https://doi.org/10.5281/zenodo.1289933

Román, M. O., & Stokes, E. C. (2015). Holidays in lights: Tracking cultural patterns in demand for energy services. Earth's future3(6), 182-205.

In my opinion, the main problem of the article is that the author needs to make a more detail analysis or assume that cultural factors are more important than inmigration in this case.

Also, I feel that some plots as an example are missing. Remember, that there is no correlation with GDP is a relevant thing. Because even if is not the typical thing to say, it is the truth and needs to be shown.

I have to congratulate the author because is a brave action nowadays to submit results that are not exactly what the community expects. I have to encourage the author to follow that line.

The title of the article, in my opinion, does not reflect the content of the article, so I suggest make sexier by adding a question or critical analysis.

Something like: Nighttime Lights and Population Migration: Critical analysis on European data.

Reviewer 4 Report

The paper deals with the use of satellite-derived nighttime Lights for investigating population migration in european countries.  The paper is surely of interest for the Remote sensing community and is well written and organized.

However, I would like authors make some moderate revisions before publication, according to the following comments/suggestions.

The introduction and theoretical background sections need to be enriched with more citations. For example, the following lines require appropriate citations of background or previous papers/works: page 1, Lines 22-24; page 1, lines 25-27; page 1, line 28; page 1, lines 38-44; page 2, lines 48-55; page 2, lines 61-65. - The main objective of the work need to be better identified and declared: it is not clear to me if the potential of nighttime lights, investigated here, is for "predict" or "estimate" population migration. Authors alternatively use the words "predict" (e.g. page 1, line 32) or "estimate" (e.g. page 2, line 56) or "understand" (e.g. page 3, line 116) which all have different meaning actually. Thus, please clarify.     - If "prediction" if the actual potential of nighttime lights, authors have to better explain how nighttime lights can actually "predict" net migration. If nighttime lights changes refer to a well defined period of time (2015-2016) net migration here reported to what time period refer to? A previous one? This is not clear reading the paper. Please clarify this aspect.  - Finally, concerning achieved results, I found that correlation coefficients exploit a very fragmented, heterogeneous and unpredictable behaviour (at country level) of lights and other predictor variables. In particular, for some countries, lights and population behave similarly, but this is not for others and the same consideration can be done for other models (e.g. lights and GDP). Such a variability (not fully explained by authors) should suggest a more cautious interpretations of the results that, in my opinion require additional investigation (at least looking at a longer time period). 

Author Response

The paper deals with the use of satellite-derived nighttime Lights for investigating population migration in european countries.  The paper is surely of interest for the Remote sensing community and is well written and organized.

However, I would like authors make some moderate revisions before publication, according to the following comments/suggestions.

---The introduction and theoretical background sections need to be enriched with more citations. For example, the following lines require appropriate citations of background or previous papers/works: page 1, Lines 22-24; page 1, lines 25-27; page 1, line 28; page 1, lines 38-44; page 2, lines 48-55; page 2, lines 61-65.

New references are added in the suggested places.

- The main objective of the work need to be better identified and declared: it is not clear to me if the potential of nighttime lights, investigated here, is for "predict" or "estimate" population migration. Authors alternatively use the words "predict" (e.g. page 1, line 32) or "estimate" (e.g. page 2, line 56) or "understand" (e.g. page 3, line 116) which all have different meaning actually. Thus, please clarify.    

---To keep the consistency of wording, “estimate” is replaced by “predict” in appropriate places, such as in page 2 line56. In most places, “estimates” refers to regression parameter estimates, or the coefficients. Thus it does not contradict with prediction models.    

The word “understand” is kept in the text, as it refers to the improvement of our general understanding about migration. It does not contradict with “predict”.

 - If "prediction" if the actual potential of nighttime lights, authors have to better explain how nighttime lights can actually "predict" net migration. If nighttime lights changes refer to a well defined period of time (2015-2016) net migration here reported to what time period refer to? A previous one? This is not clear reading the paper. Please clarify this aspect. 

---Lights is measured as averaged annual lights. It does not measure changes in lights between years. Net migration also refers to migration population during the same year as the lights measured, so does the total population and GDP per capital. These measures are explained in Data and Methods sections. This study is cross-sectional analysis, therefore, all variables are consistently measured in the level value, but not in growth rates or year-to-year differences. The word “predict” is also appropriate, as it predicts the cross-sectional difference, but not growth overtime. The recommendation for using time-series data (or growth rates) in future studies has been added in the Discussions.

- Finally, concerning achieved results, I found that correlation coefficients exploit a very fragmented, heterogeneous and unpredictable behaviour (at country level) of lights and other predictor variables. In particular, for some countries, lights and population behave similarly, but this is not for others and the same consideration can be done for other models (e.g. lights and GDP). Such a variability (not fully explained by authors) should suggest a more cautious interpretations of the results that, in my opinion require additional investigation (at least looking at a longer time period). 

---I agree. The large variation in the results across countries is emphasized in Discussion. This has been known in the field demography, that is, migration is very country-specific, depending on culture, history, government policy, etc. The most commonly used predictors are population and economic variables so far. This study indicates that, at the least, nighttime lights have similar predicting power for many countries. And just like population and GDP, this is not consistent across all countries. The suggestions for using longer time periods, and incorporating other variables are included in Discussions.

Round 2

Reviewer 1 Report

minor point - Adjusted R-squared values are useful for comparison of different models, with different numbers of explanatory variables, for one country, as opposed to comparison of models across countries that each have the same number of variables.

minor point - individual country results are estimated in pooled models using interaction terms.

major point - The author replies that OLS requirements are met, but we do not know that without a spatial test of the independence of the residuals. Spatial observations in a regression analysis frequently yield autocorrelated residuals (spatial autocorrelation) in the same way regressions over temporal units of observation frequently yield autocorrelated residuals (serial autocorrelation). Using explicitly spatial regression models (spatial errors or spatial lag) allows us to have practical confidence in the significance of the estimates - and even their signs. 

Back to TopTop