Empirical Modelling of Public Lighting Emission Functions

Round 1

Reviewer 1 Report

1. As we can see, there are many simplicities and compromises in terms of the data processes; this brings a lot of misunderstanding and ambiguity for the readers to duplicate the results shown in this paper. It suggests attaching more details as an appendix in terms of how the parameters are set both for the models and software.

 

2. Would a workflow diagram be better for readers to understand the overall methods?

3. Ground truth of the interpolated DTM (in the line: 158-159)

This paper applies the street vectors from OpenStreetMap to generate the DTM raster maps. However, the accuracy of the generated DTM is concerned. It seems that this is a compromise due to the unavailability of the DTM in Dublin. But this may lead to suspicion in terms of the inaccuracy towards the interpolated terrain elevation. It suggests that the acquired DTM should be verified by the control points of DTM/DEM in the Dublin area, or the DTM could be interpolated from these control points. Both ways may improve the accuracy or confidence of this paper.

Author Response

1. I agree that there are many details. I endeavoured to simplify the discussion to the points that were relevant to the immediate topic, but will insert more details in an Appendix.
2. I will insert a flowchart of the work.
3. I have put in a statement regarding the accuracy of the resulting DTM. Over the whole of the road network the offset of the resulting difference map (DSM-DTM) had a median value of 0.13m which we require as acceptable. Note that this is to be expected as there are some road sections which passed under structures and the presence of traffic will tend to raise the LiDAR DSM values locally, but this will not be reflected in the derived DTM which is a global solution.

Reviewer 2 Report

Thank you for a clear and understandable paper.   I will suggest a flowchart of the method and testing to make it easier for the reader to flow and for someone to reproduce your work.  

No other comments. 

Author Response

I will insert a flowchart of the work.

Reviewer 3 Report

Dear Author,

your study is really interesting and innovative. It describes a model for city emission function the calculation of which are based on on-the-field measurements, GIS methods, LiDAR datasets and remote sensing databases. The concept is well written and gives a well usable approach of light pollution modelling. Please find my comments and questions below grouped by the sections of the manuscript.

Introduction:

In the introduction, you mention some paper regarding light emission modelling. There are some recent studies modelling radiation transfer which are also worth mentioning, like:

• Liu et al (2020): A Spectral Data Compression (SDCOMP) Radiative TransferModel for High-Spectral-Resolution Radiation Simulations.J. Atmos. Sci., doi:10.1175/JAS-D-19-0238.1.
• Kyba et al. (2021): The benefit of multiple angle observations for visibleband remote sensing using night lights. doi:10.1002/essoar.10507575.1.
• Kocifaj (2018): Multiple scattering contribution to the diffuse light of a night sky: A model which embraces all orders of scattering.J.Quant. Spectrosc. Radiat. Transf., doi:10.1016/j.jqsrt.2017.11.020.
• Noebauer-Sim (2019): Monte  Carlo  radiative  transfer. Living  Reviews  in  Computational  Astrophysics, doi:10.1007/s41115-019-0004-9.

The use and application of Suomi VIIRS DNB database should be introduced.

You shouls also mention, what advantages GIS tools have in modelling.

Data and methods:

Digital Mapping Data:

• The typical values for street widths and building heights are mathematical averages or some statistical parameters (e.g. mode or median)?
• What is the relevancy of the 2015 LiDAR data you use? Where there any significant changes in the studies areas since then?

Streetlight information:

• What online reference databases were used?
• How did the applies lighting design software model lighting on the surface?
• Typo at line 126: term "software" is doubled.

Processing

Generation of the zenith-directed diffuse emission:

• Were DTM (Digital Terrain Model) data also taken from the data.gov.ie site?
• You mention that there were no DTM available for the Dublin area. Were there any data for the studied Irish towns, though?
• At line 161: "beneath each light", you mean each lamp pole?

Calculating the diffuse emission for other angles:

• The mathematical method behind the mentioned hillshade tool lacks some explanation.

Results and discussion

Angle-dependent light emission:

• line 244: "asymmetry" should be "asymmetric"

Comparison with Garstang and New World Atlas models

• Regarding that you use New World Atlas and Garstang models for comparison, their basic idea should be summerized in the introduction.
• Why are data for Ballindangan above the line of Garstang model?

References

• Ref. nr. 6: typo: the model name "Illumine" should be "Illumina"

Figures

• Figures 2, 4, 7, 8 should be given a clear reference in the main text of the manuscript.
• Figures 3, 4, 5: What is the reason for Dublin test are having emission values ordered as triplet of dots in the azimuthal angle while for Ballindangan the distribution is uniform?
• Figures 6, 7, 9: Please add labes to the axes of the graphs for better understanding.

Best regards.

 

Author Response

Introduction

Introductory references: I examined the suggested list of references and will include the Kyba et al. (2021) reference [on which I am a co-author] as it is germane to the topic of the directional light emission and its components. The other references are more technical desriptions of radiative transfer methods dealing with scattering and photon transport over many kilometres. That work is beyond the scope of the current paper, which deals with the local origin of the emission, i.e. the emission function so I do not think that there is a requirement to include them here.

VIIRS DNB: I will insert a short section referring to VIIRS DNB data and its applications.

GIS tools: I will elaborate on my mention of the desirability of using GIS-based tools for this work.

Data & Methods

Digital mapping data: I will clarify the situation with regards to building heights and road weights as there is a variation across the Dublin City data.

Date of datasets: Thank you for pointing this out as I wished to say that the lighting database and DSM data are near-contemporaneous - I will point this out.

On-line reference databases: I will clarify this: the LiDAR data is available at the data.gov.ie site referenced in the paper. The lighting databases are not publically available and were provided to me by the authorities directly as noted in the Acknowledgements, but I do not have permission to distribute them. I have clarified this in the text.

Surface lighting calculation: I will clarify this in the text.

Processing

DTM data: For the Cork data both DSM and DTM data are available in a zipped package. I will clarify this in the text.

With regards to the hillshade tool a complete description of its operation (or other packages in the software) is beyond the scope of the text but I will add a link to a description. The reason I use both R and QGIS is that their packages have freely available source code, so users can clearly see how the work. My approach is a guide, but it would be possible, for instance, to write Python scripts to run within either R or QGIS.

Results & Discussion

I will add a summary of the Garstang and New World Atlas models.

The emission from Ballindangan is heavily influenced by the direct emission from older LPS lanterns: I will emphasise this point in the caption.

References & Figures

Reference to Illumina corrected (thanks for reading the references!)

Figures 2,4,7,8: Given the flow of the text and the proximity of the figure, I didn’t reference the figures explicitly, but will do so.

Figures 3,4,5: The Dublin results included some more detail through the addition of extra azimuthal angles, including the 112 degree and 292 degree azimuths mentioned later. I have removed these from the figure as they are not essential.

Figures 6,79: As stated in the caption to Figure 6 these are results (in cd) normalised to the zenithal emission for easy comparison between locations with very different total outputs, and so is unitless. I believe that it is clear that this is the case for the other figures as well.

Reviewer 4 Report

The research made in this work is very useful to light pollution researchers and has no critical shortcomings. The paper is well organized and written. I have no real objection to accepting the paper, except for a minor comment below.

The CCEF models the angular emission analytically, Eqs. 54-56 in ref. #16. Since the mean building roof heights and the streetlight information are both available including the manufacturer’s photometry for the luminaires, the author may estimate the CCEF parameters that best match the experimental data obtained in the Dublin area. This would be of significant progress to compare a new theoretical model for CEF against the field observations in an area with lighting inventory known.

Author Response

Thank you for your comments.

I agree that it would be useful to have a detailed model using the Kocifaj (2018) model and have also obtained the executable and sample data in order to attempt this. I didn’t have time to complete this prior to submission, but have been endeavouring to obtain results using a streetlight-only model, in particular to obtain the dip in the emission towards low zenith distances which could approximate what I see in my data, but without success. I included the figure from the Los Mochis paper to show that similar distributions might be seen in other locations and via the analytical model (see their Figure 4), but I think that a more detailed comparison with my own results must await further discussions and a future paper.  I will insert a brief section of text describing my attempts, and also the issue where there are large green areas and irregular spacing and sizes of building. Many small Irish villages and small towns consist of a linear development with houses either side of a single street, so the simple gridded city model is unsuitable.

Taking Dublin City as the most likely to suit the analytical approach, there are approximately 2100 buildings in the 1.5 km² Dublin test area: if I use these numbers, then the average building is 700 m² in area, whereas the typical Georgian buildings are of 130 m² with the largest shopping centre of 12,000 m². Neither of these buildings sits on a filled square grid where building spacing width-wise is similar to the street width and a simple arithmetic mean does not seem to be appropriate.

I do, however, think that a direct comparison between approaches is useful to validate one or the other, but more time would be required to get this right.

Round 2

Reviewer 1 Report

The paper looks much improved. Thanks.