Next Article in Journal
Tracking a Surrogate Hazardous Agent (Rhodamine Dye) in a Coastal Ocean Environment Using In Situ Measurements and Concentration Estimates Derived from Drone Images
Next Article in Special Issue
Spatiotemporal Characterization of the Urban Expansion Patterns in the Yangtze River Delta Region
Previous Article in Journal
Sea Surface Salinity Subfootprint Variability from a Global High-Resolution Model
Previous Article in Special Issue
Understanding Urban Growth in Beijing-Tianjin-Hebei Region over the Past 100 Years Using Old Maps and Landsat Data
 
 
Search for Articles:
Title / Keyword
Author / Affiliation / Email
Journal
Article Type
 
 
Advanced Search
 
Section
Special Issue
Volume
Issue
Number
Page
 
Journals
Remote Sensing
Volume 13
Issue 21
10.3390/rs13214413
remotesensing-logo
Submit to this Journal Review for this Journal Propose a Special Issue
► Article Menu

Article Menu

share Share announcement Help format_quote Cite question_answer Discuss in SciProfiles thumb_up
...
Endorse textsms
...
Comment
first_page
settings
Order Article Reprints
Font Type:
Arial Georgia Verdana
Font Size:
Aa Aa Aa
Line Spacing:
Column Width:
Background:
Article

A New Approach to Identify On-Ground Lamp Types from Night-Time ISS Images

by
Natalia Rybnikova
1,2,3,*,
Alejandro Sánchez de Miguel
4,5,
Sviatoslav Rybnikov
6,7 and
Anna Brook
8
1
Department of Mathematics, University of Leicester, Leicester LE1 7RH, UK
2
Department of Natural Resources and Environmental Management, University of Haifa, Haifa 3498838, Israel
3
Department of Geography and Environmental Studies, University of Haifa, Haifa 3498838, Israel
4
Environment and Sustainability Institute, University of Exeter, Penryn Campus, Penryn TR10 9FE, UK
5
Department Física de la Tierra y Astrofísica, Instituto de Física de Partículas y del COSMOS (IPARCOS), Universidad Complutense de Madrid, 28040 Madrid, Spain
6
Institute of Evolution, University of Haifa, Haifa 3498838, Israel
7
Department of Evolutionary and Environmental Biology, University of Haifa, Haifa 3498838, Israel
8
Center for Spatial Analysis Research (UHCSISR), Spectroscopy & Remote Sensing Laboratory, Department of Geography and Environmental Studies, University of Haifa, Haifa 3498838, Israel
*
Author to whom correspondence should be addressed.
Remote Sens. 2021, 13(21), 4413; https://doi.org/10.3390/rs13214413
Submission received: 30 September 2021 / Revised: 28 October 2021 / Accepted: 30 October 2021 / Published: 2 November 2021
(This article belongs to the Special Issue Remote Sensing Image and Urban Information Visualization)
Browse Figures
Versions Notes

Abstract

:
Artificial night-time light (NTL), emitted by various on-ground human activities, has become intensive in many regions worldwide. Its adverse effects on human and ecosystem health crucially depend on the light spectrum, making the remote discrimination between different lamp types a highly important task. However, such studies remain extremely limited, and none of them exploit freely available satellite imagery. In the present analysis, the possibility to remotely assess the relative contribution of different lamp types into outdoor lighting is tested. For this sake, we match two data sources: (i) the radiometrically calibrated RGB image provided by the ISS (coarse spectral resolution data), and (ii) a set of in situ measurements with detailed spectral signatures conducted by ourselves (fine spectral resolution data). First, we analyze the fine spectral resolution data: using spectral signatures of standard lamp types from the LICA UCM library as endmembers, we perform an unmixing analysis upon NTL in situ measurements; by this, we obtain the estimates for relative contributions of the standard lamp types in each examined in situ measurement. Afterward, we focus on the coarse spectral resolution data: by using various types of statistical models, we predict the estimated relative contributions of each lamp type via RGB characteristics of spatially corresponding pixels of the ISS image. The built models predict sufficiently well (with R2 reaching ~0.87) the contributions of two standard lamp types: high-pressure sodium (HPS) and metal-halide (MH) lamps, the most widespread lamp types in the study area (Haifa, Israel). The restored map for HPS allocation demonstrates high concordance with the network of municipal roads, while that for MH shows notable coincidence with the industrial facilities and the airport.

1. Introduction

Artificial night-time light (NTL), emitted by various on-ground human activities, becomes further intensive in many countries, making the world brighter [1,2,3]. At that, there exist vast inequalities in the spatial distribution of NTL intensities: Thus, night light flux per capita in the USA on average is three times more than in Europe; across US counties, there exists a 16,000-fold difference between the most and the least light-polluted ones [4]. On the global scale, the most polluted countries (in terms of the populations living under polluted skies) are Saudi Arabia, South Korea, Argentina, Canada, Spain, the US, Brazil, Russia, Japan, and Italy [3]. Dynamics of NTL levels might be used for monitoring globalization and urbanization processes, in addition to previously used GDP, CO2 emissions, and total and urban population data [5].
At the same time, a huge amount of empirical evidence has been accumulated for the adverse effects of NTL on both humans [6,7,8,9] and ecosystems [10,11,12,13]. An especially serious concern is the long-term cumulative effects of NTL, which remain almost unexplored [14]. Moreover, NTL also frustrates professional astronomical observations [15,16,17]. With respect to the above-mentioned challenges, the necessity to regulate light pollution becomes furthermore recognized [18,19,20]. The adverse effects of NTL are known to depend crucially on the light spectrum [21,22]. For instance, the short-wavelength light stronger suppresses melatonin production and distorts circadian rhythms in mammals [9,23,24], while the long-wavelength light more strongly disrupts the magnetic orientation of migratory birds [25]. To date, the detailed spectral signatures are available for the majority of common lighting sources (see, for example, the LICA-UCM database [26]). Discriminating between different lamp types based on satellite imagery would allow for a more fine-tuned analysis of adverse health effects associated with NTL. In turn, this would contribute to elaborating more precise policies for diminishing light pollution.
In the meantime, studies aiming to discriminate between different lamp types remain extremely limited. In two of them, the authors tested the principal possibility to identify lamp type, proceeding from corresponding spectral signature, either detailed or aggregated. In the first study, Elvidge with co-authors analyzed spectral signatures of 43 different lamps representing nine the most widespread lamp types, using ASD spectroradiometer, implying measuring the signatures from 400 to 2500 nm with 10 nm bandwidth [27]. They showed that discriminant analysis correctly classified all lamp types when based on their detailed spectra. The authors also succeeded to find a minimum set of broad bands ensuring sufficient classification quality: under blue, green, red, and NIR bands (a slightly modified set represented on the Landsat Thematic Mapper), only 4.7% of the lamp types were classified incorrectly. In the second study by Sánchez de Miguel and co-authors, it was demonstrated that main lamp types can often be separated using color–color diagrams with G/R and B/G ratios as the two coordinates [28]. Accordingly, the RGB bands corresponded to Nikon D3s camera sensors, used by astronauts in the ISS. However, in both mentioned studies, the proposed lamp-type discriminating methods, although being based on spectral bands of existing satellites, were not tested on real imagery [27,28].
Puschnig with co-authors reported on spectroscopic observations of night sky spectra over Vienna, which were performed from the North Dome of the Vienna University Observatory [29]. The authors used an 80 cm Cassegrain telescope, a DSS-7 spectrograph, and a ST-7 CCD camera with a wavelength coverage of ~450–850 nm. As revealed, the strongest spectral line was due to the most common type of residential lighting type (i.e., a fluorescent lamp) and was located at 546 nm; the second-strongest line at 611 nm was revealed as a composition of fluorescent and high-pressure sodium lamps, used to illuminate main streets.
To the best of our knowledge, the only explicit test of such a kind was performed by Hale with co-authors for Birmingham, UK [30]. The authors analyzed a one-meter aerial image of the city and a layer reporting location and type of lamp. They succeeded in classifying four main lamp types with high accuracy (7.5% error), based on three focal statistics: B and G/R ratio for pixels up to 1 m from the lamp center and the maximum averaged RGB level for pixels between 2 and 4 m from the lamp center. At that, RGB bands corresponded to those of the Nikon D2X digital camera. This study, however, benefits from high-resolution aerial images available only for limited sites and are typically costly.
In the present study, we test for the possibility to identify on-ground lamp types from freely available satellite imagery of relatively coarse spatial and spectral resolution. For the study area of Haifa, Israel, we match two NTL data sources: (i) the radiometrically calibrated RGB image provided by the ISS (coarse spectral resolution data), and (ii) a set of in situ measurements with detailed spectral signatures conducted by ourselves (fine spectral resolution data). Since the ISS image is of ~20-m spatial resolution, each pixel may represent a mixture of lamps. Thus, as the first step, we analyzed fine spectral resolution data: in situ measurements were subjected to unmixing analysis, with spectral signatures of the standard lamp types (from the freely available LICA UCM library [26]) used as endmembers. The unmixing analysis allowed us to estimate the relative contribution of different lamp types in each examined in situ measurement. Afterward, we switched to the analysis of the coarse resolution data: using various types of statistical models, we predicted the estimated relative contributions of each lamp type via RGB characteristics of spatially corresponding pixels of the ISS image. Finally, we applied the built models to restore the maps of the relative contribution of certain lamp types into outdoor lighting in the study area.

2. Materials and Methods

The study design comprises four main stages: (i) matching ISS RGB image with the in situ measurements—to select the representative ground truth observations; (ii) unmixing upon detailed spectral signatures of the selected representative in situ observations—to estimate the relative contributions of the standard lamp types; (iii) building statistical models—to predict the relative contributions via RGB characteristics of the corresponding pixels from the ISS image; and (iv) applying the models to all the pixels from the ISS RGB image—to restore the allocation of major lamp types in the study area. The methodological scheme of the study is reported in Figure 1 and described in detail in subsections below.

2.1. Data Sources

In the present analysis, two NTL data sources were used: (i) the radiometrically calibrated RGB image provided by the ISS (coarse spectral resolution data), and (ii) a set of in situ measurements with detailed spectral signatures conducted by ourselves (fine spectral resolution data).
The ISS-produced NTL image of Haifa (ISS045-E-148262) was taken on 29 November 2015, with the Nikon D4 DSLR camera [31]. The image was georeferenced and radiometrically calibrated by SAVESTARS Consulting SL [32] according to the procedure reported in [33].
In situ NTL measurements were performed in March 2015 with the Konica Minolta CL-500A spectrometer. Each of the 610 measurements reports spectral irradiance (w/m2) at a 1-nm pitch from 360 to 780 nm [34]. Figure 2 reports the original and the radiometrically calibrated RGB images from the ISS, overlaid with the in situ measurements localities.
Additionally, for unmixing analysis, we used spectral signatures of standard lamp types, taken from the freely available LICA UCM library [26]. We examined spectral signatures of 81 lamps representing seven (out of nine reported by Elvidge et al. [27]) of the most popular lamp types: fluorescent (FL), metal-halide (MH), high-pressure sodium (HPS), low-pressure sodium (LPS), incandescent and halogen (I&H), mercury-vapor (MV), and light-emitting diodes (LED). Two other widespread lighting types—liquid-fuel, and pressurized-fuel lamps—were omitted due to data unavailability.

2.2. Selection of the Representative Ground Truth Observations

Proceeding from the available data on the relatively coarse spatial resolution of ISS image, with each pixel reporting emissions from multiple light sources, and simultaneously given the sporadic point-wise available in situ NTL measurements, we selected among 610 observations only those which in some sense coincided with corresponding pixels in the ISS image. Since each pixel might be characterized by RGB radiances only, we first simulated the radiances of synthetic RGB bands of in situ measurements as if they would appear on the ISS sensors of Nikon D4 DSLR camera and then chose the observations with similar to the corresponding ISS imagery pixel (in terms of Euclidian distance in the coordinate system, represented by B/G and G/R ratios) RGB characteristics.
To simulate the radiance R (of either red, green, or blue band), we used the augmented equation reported by Sanchez de Miguel et al. [28]:
R = ∫ 0 ∞ ϕ ( λ ) T ( λ ) A ( λ ) d λ ∫ 0 ∞ ϕ A B ( λ ) T ( λ ) A ( λ ) d λ ,
where
  • ϕ(λ) = spectrum of the lamp under analysis;
  • T(λ) = spectral sensitivity of the synthetic band (in the present analysis, we used spectral responses of Nikon D4 Electronic Still Camera—the one that produced the image under analysis,—reported in [28])—see Figure 3a;
  • A(λ) = atmospheric transmittance (in the present analysis, we applied the MODTRAN® computer code [35] for the simulation of atmospheric transmittance of light emissions through the atmosphere over the study area at the time when in situ measurements were performed. MODTRAN code addresses the following issues: the effects of molecular and particulate absorption/emission and scattering, surface reflections and emission, solar/lunar illumination, and spherical refraction [35])—see Figure 3b;
  • ϕAB(λ) = reference spectrum of AB magnitude system, defined for a source of constant spectral density flux of 3631 Janskys across the spectral range of the band [28]—see Figure 3c.
As a measure of similarity between RGB radiances, reported by pixels of calibrated ISS image, and corresponding in situ measurements, we used Euclidian distance in the coordinate system, represented by B/G and G/R ratios. Given the variance of such a distance (dmax = 2.00), we settled the threshold of d < 0.2.

2.3. Estimation of the Relative Contributions of the Standard Lamp Types into Spectral Signatures of the in Situ Observations

The detailed spectral signatures of the representative in situ measurements (see Section 2.2) were subjected to unmixing analysis. As the endmembers (i.e., spectra of pure ‘materials’—see [36]), we used the detailed spectral signatures of the standard lamps from the LICA-UCM library. The end-members’ signatures are shown in Figure 4. For unmixing analysis, we used the FNNLS algorithm [37], implemented in MATLAB v.R2020b. The algorithm returned the percentages of all endmember lamps in each of the pre-selected in situ measurements. The obtained percentages were then aggregated within lamp types, and the sums were normalized to the unit. The obtained estimates are hereafter referred to as relative contributions.

2.4. Prediction of the Relative Contributions via RGB Characteristics of the Corresponding Pixels from the ISS Image

The relative contribution of each lamp type, obtained in the unmixing analysis (Section 2.2), served as the dependent variable in a set of statistical models. As the explanatory variables, we tried different characteristics of the pixels of the calibrated ISS image: (i) radiance in the red, green, and blue bands per se; (ii) their ratios (G/R and B/G, or GG/RB ratio), and an additional derivative characteristic describing the pixel’s ‘proximity’ to the lamp type in question. This distance was included in the models since we found that different lamps within each of the lamp types tend to lie along straight lines in the G/R, B/G coordinate plane (see Figure 5). We tried several formalizations of such a distance: (i) the ratio between B/G and G/R of the pixel, as a measure of the line’s slope; (ii) Euclidean distance from the pixel to the line representing the lamp type in question; and (iii) Mahalanobis distance from the pixel to the cloud representing the lamp type in question, which accounts for both the center of mass and the direction of the cloud [38].
We examined statistical models of several classes: multiple least-square linear regression, neural network (perceptron), and random forest. The general idea behind the multiple least-squares linear regression is fitting the observations (each represented by a point in N-dimensional space with (N-1) number of predictors and one dependent variable) by a linear relationship, represented by an (N-1)-dimensional linear surface, or hyperplane, by minimizing the sum of squared errors between the actual and estimated over this hyperplane levels of the dependent variable. In addition to the linear approach, a multi-layer neural network (perceptron)—a hierarchical model consisting of input, output, and several hidden layers of neurons, weighting, and summarizing the inputs from the previous layer and passing them to the neurons of the next layer—was treated. We also tested the random forest approach [39], which implies building an ensemble of decision trees, each ‘voting’ for a certain class or level of the dependent variable, with subsequent averaging of the estimates across all the decision trees.
All models were run in ORANGE v.3.28 with the default settings. Specifically, linear regression was applied with intercept and without regularization. Like a neural network, we used a multi-layer perceptron with backpropagation; the model parameters were the following: number of neurons in the hidden layer is 100, number of hidden layers is 1, the activation function is ReLu, solver for weight optimization is a stochastic gradient-based optimizer, L2 penalty parameter is 0.0001, and the maximal number of iterations is 200. In random forest models, the number of trees was settled to 10, an arbitrary set of attributes and limit depth of individual trees were left unchecked, and subsets smaller than 5 were required not to be split. We conducted a five-fold cross-validation procedure, which implies: (i) splitting the data into five nearly equal subsets, (ii) using each of them as the testing set while using the other four as a training set, and (iii) averaging the performance scores from these five iterations separately across training and testing sets. The input database used in the analysis is available from the authors upon request.

2.5. Restoring the Allocation of Major Lamp Types in the Haifa Area (Israel)

Finally, the best-performing models (Section 2.4) were applied to the radiometrically calibrated ISS image (see Section 2.1), and estimates for different lamp type relative contribution into light emissions from the study area were obtained and depicted.

3. Results

Among all initially available in situ measurements, we chose only the set of representative measurements—those deviating from the corresponding pixels of the radiometrically calibrated ISS image by less than 0.2 in terms of the Euclidian distance in the G/R, B/G coordinate plane (see Section 2.2). Overall, we obtained 196 measurements; their main characteristics are reported in Figure 6.
The detailed spectral signatures of the obtained 196 representative in situ measurements were then subjected to unmixing analysis (see Section 2.2). The results of the unmixing, i.e., the relative contributions of the seven endmember lamp types, are reported in Appendix A Table A1. In the table, we also report, for each in situ measurement, the highest correlation between its detailed spectral signature and the signatures of the lamps from the LICA-UCM library. As one can see from the table, the measurements with the predominant relative contribution of a certain lamp type (marked grey) simultaneously demonstrated a high correlation with this lamp type. Overall, HPS lamps were the most widespread in Haifa in 2015 (with an average relative contribution of 32.8%), followed by MH lamps (22.0%), while LED lamps were considerably less frequent (15.6%).
Table 1 reports the performance of three alternative statistical models (linear regression, neural network, and random forest) developed to explain the variance of each lamp type relative contribution, under various sets of predictors, for the training and the testing sets (see Section 2.4). As can be seen from the table, the performance is higher for the models with G/R and B/G ratios as predictors, compared to those with R, G, and B bands per se. We suppose that color ratios work better for predicting the relative contribution of a certain lamp type since it has a certain spectral signature. That is, the lamp of a certain type might be brighter or darker, and thus its aggregated R, G, and B levels per se will differ, but the ratios between those bands will remain the same. Indeed, in the considered set of observations, the variability of R, G, and B levels per se is much higher than the variability of their ratios (coefficients of variation for RGB channels are 0.72, 0.75, and 0.90, respectively, while for G/R and B/G ratios the corresponding numbers are 0.31 and 0.35). For all dependent variables and all predictor sets, random forest models perform generally better than either linear regression or neural network models. For three lamp types (HPS, MH, and MV), random forest models containing G/R and B/G ratios as predictors demonstrate high performance both for the training and the testing sets: R2 > 0.43 (see Table 1, predictor sets 2nd–4th).
Since models with the predictor sets 2nd–4th demonstrate similar performance, and proceeding from consideration of calculation simplicity, we run random forest models, built upon the whole set of observations, with G/R, B/G, and GG/RB ratios as predictors upon all pixels of the calibrated ISS image (see Figure 1b). Figure 7 reports resulting maps for two of the most frequent lamp types (HPS and MH) prevalent in Haifa. As can be seen from the figure, relatively higher contribution of HPS lamps in the outdoor lighting in Haifa coincide with the spatial pattern of municipal roads (see Figure 7a), while the pattern for MH lamps is more local and site-specific (see Figure 7b). In some regions of interest, HPS and MH lamps contribute to the outdoor lighting reversely. For example, in the Haifa Bay area, the impact of HPS lamps is pronounced, while MH lamps are absent (see Figure 7c); in contrast, in the areas hosting Khof Shemen industrial zone and Haifa airport, MH lamps are widely used while HPS lamps are absent (see Figure 7d).

4. Discussion

In the present study, we tested for the possibility to identify on-ground lamp types from freely available satellite imagery of relatively coarse spatial and spectral resolution. To this end, we conducted a series of in situ NTL measurements in Haifa, Israel and combined these data with a radiometrically calibrated NTL image of the city taken from the ISS. Since the ISS image is of ~20-m resolution, each pixel likely represents light emission from a mixture of lamps. By applying unmixing analysis to the detailed spectral signatures of in situ NTL measurements, we estimated the relative contributions of different lamp types in each analyzed observation. Then, we tried to train statistical models to predict these relative contributions based exclusively on the ISS image, which represents RGB bands.
As our analysis indicates, contributions of two of the most widespread lamp types in the region, HPS and MH lamps, could be successfully predicted by random forest models. Using them, we restored HPS and MH lamps relative contribution to outdoor lighting of the whole Haifa area. The obtained HPS map demonstrated high concordance with the network of municipal roads (confirming the result obtained by Puschnig with co-authors for Vienna [29]), while the MH map showed notable coincidence with industrial facilities and the airport (see Figure 7).
In the developed models, we used three explanatory variables. Two of them, G/R and B/G ratios, are similar to those previously used by Sánchez de Miguel [28]. An additional informative predictor described a ‘proximity’ of the analyzed pixel of the ISS image to the lamp type in question. Interestingly, it turned out that lamps of the same type form clear-cut line segment-shaped loci in the G/R, B/G coordinate plane (see Figure 5). In this respect, we tried three alternative formalizations for the above-mentioned proximity: GG/RB ratio of the pixel, its Euclidean distance to the line, and its Mahalanobis distance to the locus. Without this additional explanatory variable, the models’ performance was somewhat worse (with adjusted ΔR2 up to −0.1, depending on the model type).
It should be mentioned that both Elvidge et al. [27] and Hale et al. [30] reached remarkably high classification accuracy (with errors of 4.7% and 7.5%, respectively), presumably due to either laboratory measurements [27] or high-resolution aerial imagery [30], incompatible with mixed pixels and used in the present analysis. However, their results cannot be directly compared with those presented here since we solved regression rather than classification problem, implying continuous rather than binary dependent variable. Yet, it seems reasonable to expect a better performance of our models if they were based on an image of better spectral (like in [27]) and/or spatial (like in [30]) resolution. Again, as mentioned above, we used the ISS image intentionally—given its free availability for many geographical sites. We think that our results argue for the principal possibility to assess the lamp type composition of outdoor lighting from the color satellite imagery.
Some limitations and perspectives of the study should be mentioned. First, proceeding from the available data, we did not succeed to obtain a reasonable spatial pattern for LED lamps’ relative contribution in the outdoor lighting in Haifa. A trivial reason may be the insufficiency of the used data (small sample size, low prevalence of LEDs in the studied region in 2015). It also seems possible that the ISS imagery does not allow one to discriminate between LEDs due to their high spectral variability and, therefore, overlap with some other lamp types occurs (which becomes even more pronounced after considering the reflectance of the ground), such as MH lamps, in the B/G, G/R space [28]. It should be mentioned that Elvidge et al. [27] successfully discriminated between LED lamps from other types since they used almost non-overlapping red, green, blue, and NIR bands. Nowadays, LEDs’ popularity is growing rapidly, mainly due to their versatility and energy-saving potential [27,28,43], and some precedents of total LED-based street lighting already exist [28,44]. At the same time, LEDs’ primary emission peak, ~450–460 nm [27], radically distorts circadian rhythms and suppresses melatonin production in humans, contributing to sleep disorders [45], obesity [7], hormone-dependent cancers [9], and other diseases. Thus, further analysis is needed to explore the possibility to identify LED lamps from the ISS-provided imagery. It seems promising to exploit, in addition to ISS imagery, VIIRS-provided panchromatic data, which covers also NIR diapason on night-time light. To obtain spatially upscaled panchromatic data, it seems rational first to teach some models (such as linear regressions, non-linear kernel regressions, random forest models, etc.) to associate between downscaled (to the resolution of ~742 m) ISS-provided RGB data and actual panchromatic data, and afterward to apply the best-performing model to predict panchromatic lights from the levels of actual RGB data (of ~20 m resolution). Alternatively, or in addition to, land-use data of fine spatial resolution, obtained from aerial imagery and LiDAR data (see for example [46,47,48]), might be used as an additional predictor for better discriminating between different lamp types. It is important to note that, given the good-performing models for all lamp types, a detailed spectrum might be easily restored for each pixel for the whole study area: for this sake, we should sum up the spectra of corresponding lamp types, weighted for their relative contributions. By this, we would be able to predict night-time hyperspectral information, which is currently not available from any of the existing satellites.
Second, the acquisition time of the used ISS-provided image and the in situ NTL measurements do not coincide perfectly, which may cause some inaccuracy in the herein obtained estimates. However, weather conditions in Israel are rather similar in March (when the in situ measurements were conducted) and in November (when the ISS image was taken): the average low-temperatures are ~11 °C in March and ~14 °C in November; the average sunshine is ~7 h for both months; and it is raining on average for 8.6 in March and 8 days in November [49]. Additionally, our sample did not include observations from residential areas, in which brightness may vary during the night [50]; instead, it included observations along major roads, entertainment areas, hospitals, and high-tech enterprises—that is, represented by streetlights of nearly-constant brightness during the night.
Finally, further testing of the proposed approach is required. For this sake, other ISS-provided images, which underwent radiometric calibration procedure [32,33], might be used. Given that cities might substantially differ in terms of building typologies and relationships between building heights and street widths [51], further research is needed to test the applicability of the obtained models to other geographical regions.

Author Contributions

Conceptualization, N.R., S.R. and A.B.; methodology, N.R., A.S.d.M. and A.B.; software, N.R. and A.S.d.M.; validation, N.R.; formal analysis, N.R.; investigation, N.R., S.R. and A.B.; resources, N.R.; data curation, N.R.; writing—original draft preparation, N.R.; writing—review and editing, A.S.d.M., S.R. and A.B.; visualization, N.R.; supervision, A.B.; project administration, N.R.; funding acquisition, N.R., S.R., A.S.d.M. and A.B. All authors have read and agreed to the published version of the manuscript.

Funding

This research was supported by the Council for Higher Education of Israel and by the Cities at Night Project.

Institutional Review Board Statement

Not applicable.

Informed Consent Statement

Not applicable.

Data Availability Statement

Data and the codes used in the analysis are available from the corresponding author upon request.

Acknowledgments

The authors thank three anonymous reviewers for highly valuable comments, which helped to improve the manuscript and make it more comprehensive to the reader.

Conflicts of Interest

The authors declare no conflict of interest. The funders had no role in the design of the study; in the collection, analyses, or interpretation of data; in the writing of the manuscript; or in the decision to publish the results.

Appendix A

Table
Table A1. Unmixing results: The relative contribution of each lamp type into the detailed spectral signature of in situ observation (see text for explanation).
Table A1. Unmixing results: The relative contribution of each lamp type into the detailed spectral signature of in situ observation (see text for explanation).
In Situ Obs. NumberRelative Contribution into In Situ Observation Detailed Spectral Signature of Standard Type LampsMax Correlation between Detailed Signatures of In Situ Obs. and Lamps
CFLMHHPSLPSIncandMVLEDCorrelationLamp Type
5040.0850.6590.0670.0600.0140.0000.1140.720LED
1850.1970.1760.3050.0800.0300.0000.2130.911LED
5660.0220.0580.2010.0440.5680.0090.0970.957Incand
4670.0000.0240.7720.0540.0000.0000.1500.945HPS
2380.1940.0720.3230.0000.0080.3070.0950.634MV
2310.1950.0580.2650.0000.0000.3890.0930.654MV
970.0000.0650.6640.1690.0000.0000.1010.937HPS
4650.0000.0240.7600.1030.0000.0000.1130.930HPS
2860.0830.4650.3360.0000.0000.0000.1160.747LED
5700.0970.2870.0520.0590.0490.0000.4560.946LED
1120.0400.3050.3720.2430.0130.0000.0260.806HPS
5650.0000.0190.1510.0270.6380.0020.1640.979Incand
4310.0000.0560.6710.1510.0360.0000.0870.889HPS
4660.0000.0240.7550.1000.0000.0000.1210.932HPS
4330.0000.0850.6290.1190.0000.0000.1660.913LED
1070.0000.0760.5950.2980.0000.0000.0310.891HPS
4510.0740.5430.2720.0460.0000.0000.0660.777MH
10.2470.0980.1610.0000.0000.3860.1080.670MV
2500.2370.0420.1410.0000.0020.4800.0980.663MV
4320.0000.0680.6080.0750.0620.0000.1870.936LED
4300.0160.0070.7290.0000.0520.0000.1950.938HPS
1030.0000.0840.6580.2040.0000.0000.0540.922HPS
680.0080.3270.4120.1620.0000.0000.0910.782HPS
360.0000.1570.6460.1040.0000.0000.0930.935HPS
4530.0770.5550.2790.0250.0000.0000.0650.776MH
1680.0250.3580.4780.0450.0000.0000.0940.849HPS
400.0000.0580.6740.1760.0000.0000.0920.935HPS
640.0210.3450.4110.1480.0000.0000.0740.741HPS
4400.0070.1840.5840.1070.0000.0000.1180.898HPS
4520.0780.5510.2690.0410.0000.0000.0620.775MH
2640.0200.0300.0290.0040.8730.0000.0440.997Incand
2560.0000.7390.0000.0000.1210.0000.1400.832MH
2610.3920.1520.2350.0450.0260.0000.1490.843LED
3970.0680.3820.1680.0600.0200.0000.3010.852LED
1090.0000.0810.5860.3110.0000.0000.0230.885HPS
4450.0360.5110.2940.1160.0080.0000.0350.803MH
380.0200.0920.6040.1340.0000.0000.1500.928LED
4500.0760.5700.2660.0100.0000.0000.0790.762MH
5280.1240.1540.3840.0680.0000.1550.1160.725LED
1040.0000.0860.6140.2680.0000.0000.0320.900HPS
5590.5160.0360.1610.0000.0000.2030.0850.789CFL
5900.1230.6040.0000.0690.1090.0000.0950.838MH
290.1140.0900.5250.0940.0020.0000.1740.930LED
350.0000.0380.7580.0870.0000.0000.1160.935HPS
6080.0880.5520.0000.0200.0920.0000.2480.881LED
1870.0040.1230.5740.1510.0080.0000.1400.927HPS
310.0550.0600.6400.1520.0000.0000.0930.926HPS
1290.0000.0520.6930.1810.0000.0000.0740.931HPS
5080.4090.4090.0440.0000.0000.0250.1120.714CFL
2600.0340.6750.1490.0860.0000.0000.0570.752MH
5580.4500.0140.1670.0000.0000.2770.0920.759CFL
5290.1070.1950.4250.0900.0000.0720.1100.735LED
5560.5320.0380.1530.0000.0000.1980.0790.797CFL
5640.0000.0450.2390.0600.4550.0000.2020.934Incand
5180.0130.0270.1030.0450.0490.0000.7620.983LED
4950.6410.2210.0000.0210.0000.0000.1170.768CFL
690.0860.0470.6470.0800.0000.0000.1390.937LED
4290.0000.0450.6880.0000.0700.0000.1970.930HPS
1010.0000.0710.6430.1790.0000.0000.1060.934HPS
2910.0930.2630.2960.0000.0050.1550.1890.725LED
1020.0000.0800.6100.2760.0000.0000.0350.899HPS
1390.0000.1950.5160.2180.0000.0000.0710.885HPS
1380.0000.1390.5540.2450.0000.0000.0610.890HPS
5090.1130.6020.1260.0000.0000.0430.1150.716MH
1000.0000.0860.5470.3670.0000.0000.0000.841HPS
6070.0530.5540.0000.0000.1110.0000.2820.887LED
5260.7230.2040.0000.0190.0000.0000.0540.828CFL
4490.0630.5240.2660.1120.0000.0000.0350.824MH
5170.0380.1840.4920.2190.0000.0000.0670.840HPS
5200.0020.0000.0040.0000.0570.0180.9190.984LED
3910.0000.0810.5880.2910.0000.0000.0400.892HPS
460.2190.0930.0980.0000.0000.0000.5900.954LED
5230.0160.1330.4390.3730.0000.0000.0400.766HPS
5190.0280.0020.0090.0000.0600.0210.8800.984LED
840.0770.0910.6650.0000.0230.0000.1440.949HPS
1930.1010.1020.4830.1630.0000.0000.1510.906LED
5890.0320.6450.0000.0000.1790.0000.1440.810MH
5240.0320.1270.3650.2520.0000.0000.2250.858LED
5910.4740.4190.0000.0180.0250.0000.0640.747CFL
5630.0000.0390.1540.0730.6300.0000.1030.971Incand
4740.0620.4650.2700.1380.0000.0000.0650.756MH
160.0280.0130.1890.0210.0280.0190.7010.977LED
410.0000.0120.7770.0850.0000.0000.1260.942HPS
3430.7760.1930.0000.0110.0000.0000.0200.837CFL
320.0000.1020.5320.3660.0000.0000.0000.832HPS
2730.0150.3340.4700.0550.0000.0000.1260.880LED
2080.0050.1080.3440.1500.0320.0000.3610.965LED
980.0860.1120.5050.2480.0000.0000.0490.881HPS
2280.0310.1870.5250.1470.0000.0000.1100.911HPS
2450.2460.0390.1490.0000.0090.4620.0950.671MV
4460.0230.5370.2930.1260.0000.0000.0210.820MH
1730.0800.2760.0490.0000.0000.3240.2720.634MH
6050.0820.5480.0000.0170.0910.0000.2620.893LED
990.0550.1150.4800.3480.0000.0000.0020.811HPS
6010.7680.1690.0000.0000.0000.0000.0630.859CFL
180.1650.1030.3750.0000.0040.2230.1290.669LED
5550.6090.0690.1230.0000.0000.1260.0730.818CFL
5300.4730.0790.0800.0300.0000.0620.2750.772LED
2630.0330.0390.1300.0550.6570.0000.0850.977Incand
3270.0000.7760.0000.0740.0000.0000.1500.812MH
6000.7670.1650.0000.0000.0000.0000.0680.856CFL
5570.3050.0150.1910.0000.0000.4030.0870.678MV
2060.0150.0870.2500.1110.0040.0000.5330.971LED
5620.0000.0580.2010.1060.5270.0000.1080.936Incand
2650.0240.0150.0380.0130.8950.0020.0120.999Incand
3720.0930.0760.1600.0390.5250.0000.1060.957Incand
6040.0320.5250.0000.0000.1150.0000.3280.916LED
60.0990.6120.2000.0270.0040.0000.0590.776MH
6060.0470.5580.0000.0000.1140.0000.2820.885LED
450.4070.1390.2950.0440.0000.0000.1150.827LED
2660.0520.0310.0320.0100.8430.0000.0320.998Incand
390.0000.0520.6730.1790.0000.0000.0960.935HPS
150.0190.0140.0890.0130.0370.0220.8050.976LED
510.0850.1840.4230.0070.0000.1020.1990.775LED
5830.0940.5730.2170.0000.0000.0000.1150.729MH
5840.0980.6170.2060.0000.0000.0000.0780.776MH
420.0020.1970.6200.0600.0000.0000.1210.893HPS
2920.0520.1930.3170.0000.0000.2830.1550.604MV
4610.0370.5960.1050.1620.0000.0000.1000.826MH
4540.0720.5400.2820.0470.0000.0000.0600.772MH
4390.0110.0810.6190.1460.0010.0000.1410.917HPS
4710.0570.4690.2710.1410.0000.0000.0620.772MH
630.0190.4040.4170.1200.0000.0000.0410.810HPS
2070.0190.0910.2380.1040.0340.0000.5140.975LED
4680.0000.0310.7430.0910.0000.0000.1360.933HPS
5850.1510.4360.1870.0000.0000.0780.1480.652CFL
5990.7570.1650.0000.0000.0000.0000.0780.853CFL
1970.1140.1380.4150.2100.0000.0000.1230.887LED
5710.0320.4500.1160.1180.0850.0000.1980.891LED
4340.0000.0800.6870.1310.0000.0000.1010.928HPS
4480.0630.5010.2750.1280.0000.0000.0340.807MH
900.1630.1540.4410.0440.0000.0000.1980.904LED
3990.0600.3080.1580.0600.1160.0060.2920.899LED
660.2970.1900.2480.0770.0000.0000.1880.810LED
3750.0430.2490.4660.1000.0000.0000.1420.926LED
1050.0000.0710.6730.1410.0000.0000.1150.933HPS
3090.0030.2430.5830.0490.0000.0000.1220.913HPS
3490.3920.1810.1100.0000.0010.0000.3150.815LED
1980.0100.1190.5350.3100.0000.0000.0260.867HPS
4750.0550.4880.2550.1420.0000.0000.0600.780MH
2670.0270.0270.0290.0100.8720.0000.0350.998Incand
70.1950.4930.0880.0000.0290.0680.1260.677LED
2270.0060.1280.6060.1590.0000.0000.1010.919HPS
1990.0030.1110.5490.3120.0000.0000.0250.871HPS
2620.6790.1750.0210.0290.0000.0000.0960.791CFL
2440.2390.0420.1470.0000.0480.4320.0930.672MV
1890.0990.0410.6270.0420.0000.0000.1910.922LED
50.0410.5690.2740.0840.0000.0000.0320.817MH
520.1250.2290.2760.0550.0000.0990.2170.832LED
3760.0600.2510.4490.0970.0000.0000.1420.930LED
5450.0240.1390.5230.2970.0000.0000.0180.829HPS
5270.1700.1670.4000.1320.0000.0000.1310.839LED
4050.0630.5410.1620.1310.0000.0000.1020.774MH
3510.1270.0650.6560.0040.0000.0000.1470.941HPS
440.0140.1090.6430.0060.0000.0490.1780.886HPS
1260.0000.0470.7220.1030.0000.0000.1280.940HPS
2090.1530.1150.4360.1950.0000.0000.1000.884LED
3980.0610.3210.1590.0570.0790.0000.3230.890LED
3500.5810.1890.1380.0000.0000.0000.0920.770CFL
4620.0410.6100.1070.1430.0000.0000.0990.828MH
2410.2300.0390.1510.0000.0000.4840.0960.671MV
5210.0230.0070.0280.0040.0540.0320.8510.985LED
4040.0540.5720.1280.1460.0000.0000.1000.816MH
5690.2380.0630.0000.0000.0000.0000.6990.950LED
3520.0520.0580.1480.0000.0050.0000.7370.976LED
20.2160.2960.0610.0000.0000.2420.1850.648CFL
2840.0830.5310.2620.0000.0000.0000.1240.719MH
4470.0350.5150.2830.1310.0110.0000.0250.802MH
4980.5860.2570.0000.0000.0000.0000.1570.760CFL
1080.0000.0580.6420.1410.0000.0000.1590.928LED
1960.0640.0980.5770.1320.0000.0000.1290.925HPS
5400.1830.0270.1380.0000.0210.3990.2320.697LED
950.0680.0770.6790.0060.0000.0000.1700.940HPS
850.0160.1350.6110.0540.0100.0000.1750.938LED
1410.0000.2880.4620.1260.0000.0000.1240.877HPS
1130.0410.4820.2340.1890.0250.0000.0280.838MH
3070.0200.3960.3810.0910.0370.0000.0750.861LED
370.1860.1060.4580.1460.0000.0000.1040.733LED
4770.1330.0250.1930.0000.0000.5550.0950.689MV
4760.1330.0260.2000.0000.0000.5490.0910.686MV
130.1470.4440.1520.1740.0470.0000.0370.886MH
1430.0360.0910.6000.2330.0000.0000.0400.878HPS
5420.1810.0270.1390.0000.0330.3910.2300.701LED
4370.0000.1050.6010.2180.0000.0000.0770.879HPS
280.0020.0790.6410.1650.0000.0000.1130.928HPS
5780.1170.5490.0490.0100.0720.0000.2030.850LED
5800.0920.5830.1620.0000.0220.0000.1430.806LED
5820.1110.5490.1780.0000.0000.0180.1450.684MH
5410.1740.0290.1270.0000.0350.3710.2640.755LED
2000.0000.1140.6070.1690.0000.0000.1090.922HPS
4120.0000.5370.1550.0970.0590.0000.1520.728LED
2680.0060.0160.2610.0000.0080.0000.7100.990LED
4350.0000.0810.6060.2080.0260.0000.0800.873HPS
2180.0510.3540.3330.1210.0150.0000.1260.893LED
5610.0000.1030.3730.2000.2300.0000.0950.826LED
5340.1000.1090.4570.2170.0000.0630.0550.795HPS
Avg.0.1110.2200.3280.0880.0520.0440.156

References

  1. Tsao, J.Y.; Waide, P. The world’s appetite for light: Empirical data and trends spanning three centuries and six continents. LEUKOS-J. Illum. Eng. Soc. N. Am. 2010, 6, 259–281. [Google Scholar] [CrossRef] [Green Version]
  2. Kyba, C.C.M.; Kuester, T.; De Miguel, A.S.; Baugh, K.; Jechow, A.; Hölker, F.; Bennie, J.; Elvidge, C.D.; Gaston, K.J.; Guanter, L. Artificially lit surface of Earth at night increasing in radiance and extent. Sci. Adv. 2017, 3, e1701528. [Google Scholar] [CrossRef] [Green Version]
  3. Falchi, F.; Cinzano, P.; Duriscoe, D.; Kyba, C.C.M.; Elvidge, C.D.; Baugh, K.; Portnov, B.A.; Rybnikova, N.A.; Furgoni, R. The new world atlas of artificial night sky brightness. Sci. Adv. 2016, 2, e1600377. [Google Scholar] [CrossRef] [Green Version]
  4. Falchi, F.; Furgoni, R.; Gallaway, T.A.A.; Rybnikova, N.A.A.; Portnov, B.A.A.; Baugh, K.; Cinzano, P.; Elvidge, C.D.D. Light pollution in USA and Europe: The good, the bad and the ugly. J. Environ. Manag. 2019, 248, 109227. [Google Scholar] [CrossRef]
  5. Balsa-Barreiro, J.; Li, Y.; Morales, A.; Pentland, A. “Sandy” Globalization and the shifting centers of gravity of world’s human dynamics: Implications for sustainability. J. Clean. Prod. 2019, 239, 117923. [Google Scholar] [CrossRef]
  6. Kloog, I.; Haim, A.; Stevens, R.G.; Portnov, B.A. Global co-distribution of light at night (LAN) and cancers of prostate, colon, and lung in men. Chronobiol. Int. 2009, 26, 108–125. [Google Scholar] [CrossRef]
  7. McFadden, E.; Jones, M.E.; Schoemaker, M.J.; Ashworth, A.; Swerdlow, A.J. The relationship between obesity and exposure to light at night: Cross-sectional analyses of over 100,000 women in the Breakthrough Generations Study. Am. J. Epidemiol. 2014, 180, 245–250. [Google Scholar] [CrossRef]
  8. Garcia-Saenz, A.; Sánchez de Miguel, A.; Espinosa, A.; Valentin, A.; Aragonés, N.; Llorca, J.; Amiano, P.; Martín Sánchez, V.; Guevara, M.; Capelo, R.; et al. Evaluating the Association between Artificial Light-at-Night Exposure and Breast and Prostate Cancer Risk in Spain (MCC-Spain Study). Environ. Health Perspect. 2018, 126, 047011. [Google Scholar] [CrossRef]
  9. Haim, A.; Portnov, B.A. Light Pollution as a New Risk Factor for Human Breast and Prostate Cancers; Springer: Dordrecht, The Netherlands, 2013; ISBN 9789400762206. [Google Scholar]
  10. Longcore, T.; Rich, C. Ecological light pollution. Front. Ecol. Environ. 2004, 2, 191–198. [Google Scholar] [CrossRef]
  11. Hölker, F.; Wolter, C.; Perkin, E.K.; Tockner, K. Light pollution as a biodiversity threat. Trends Ecol. Evol. 2010, 25, 681–682. [Google Scholar] [CrossRef]
  12. Gaston, K.J.; Bennie, J.; Davies, T.W.; Hopkins, J. The ecological impacts of nighttime light pollution: A mechanistic appraisal. Biol. Rev. 2013, 88, 912–927. [Google Scholar] [CrossRef]
  13. Owens, A.C.S.; Cochard, P.; Durrant, J.; Farnworth, B.; Perkin, E.K.; Seymoure, B. Light pollution is a driver of insect declines. Biol. Conserv. 2020, 241, 108259. [Google Scholar] [CrossRef]
  14. Lyytimäki, J. Towards eco-efficient and enjoyable lighting. Br. GSDR 2015. [Google Scholar]
  15. Riegel, K.W. Light pollution. Science 1973, 179, 1285–1291. [Google Scholar] [CrossRef]
  16. Zhang, Q.; Wang, P.; Chen, H.; Huang, Q.; Jiang, H.; Zhang, Z.; Zhang, Y.; Luo, X.; Sun, S. A novel method for urban area extraction from VIIRS DNB and MODIS NDVI data: A case study of Chinese cities. Int. J. Remote Sens. 2017, 38, 6094–6109. [Google Scholar] [CrossRef]
  17. Kyba, C.C.M. Is light pollution getting better or worse? Nat. Astron. 2018, 2, 267–269. [Google Scholar] [CrossRef]
  18. Falchi, F.; Bará, S. A linear systems approach to protect the night sky: Implications for current and future regulations. R. Soc. Open Sci. 2020, 7, 201501. [Google Scholar] [CrossRef]
  19. Morgan-Taylor, M. Regulating Light Pollution in Europe: Legal Challenges and Ways Forward; Routledge: New York, NY, USA, 2015. [Google Scholar]
  20. Cho, S.; Lee, M.; Choi, H.; Korean, H.K.-J. Calculation and Regulation Proposal of Light Pollution from Road Lightings. J. Korean Inst. Illum. Electr. Install. Eng. 2011, 25, 21–26. [Google Scholar] [CrossRef]
  21. Gaston, K.; Duffy, J.; Gaston, S.; Bennie, J.; Oecologia, T.D. Human alteration of natural light cycles: Causes and ecological consequences. Oecologia 2014, 176, 917–931. [Google Scholar] [CrossRef] [Green Version]
  22. Schroer, S.; Hölker, F. Impact of Lighting on Flora and Fauna; Springer: Berlin, Germany, 2016. [Google Scholar] [CrossRef]
  23. Lockley, S.; Brainard, G.C.; Czeisler, C.A. High sensitivity of the human circadian melatonin rhythm to resetting by short wavelength light. J. Clin. Endocrinol. Metab. 2003, 88, 4502–4505. [Google Scholar] [CrossRef]
  24. Hatori, M.; Gronfier, C.; Van Gelder, R.N.; Bernstein, P.S.; Carreras, J.; Panda, S.; Marks, F.; Sliney, D.; Hirota, T.; Furukawa, T.; et al. Global rise of potential health hazards caused by blue light-induced circadian disruption in modern aging societies. Npj Aging Mech. Dis. 2017, 3, 9. [Google Scholar] [CrossRef] [Green Version]
  25. Wiltschko, W.; Munro, U.; Ford, H.; Wiltschko, R. Red light disrupts magnetic orientation of migratory birds. Nature 1993, 364, 525–527. [Google Scholar] [CrossRef]
  26. Tapia Ayuga, C.; Sánchez de Miguel, A.; Zamorano Calvo, J. LICA-UCM lamps spectral database. Ene 2017, 8, 29. [Google Scholar]
  27. Elvidge, C.D.; Keith, D.M.; Tuttle, B.T.; Baugh, K.E. Spectral identification of lighting type and character. Sensors 2010, 10, 3961–3988. [Google Scholar] [CrossRef] [PubMed]
  28. Sánchez de Miguel, A.; Kyba, C.C.M.; Aubé, M.; Zamorano, J.; Cardiel, N.; Tapia, C.; Bennie, J.; Gaston, K.J. Colour remote sensing of the impact of artificial light at night (I): The potential of the International Space Station and other DSLR-based platforms. Remote Sens. Environ. 2019, 224, 92–103. [Google Scholar] [CrossRef]
  29. Puschnig, J.; Posch, T.; Uttenthaler, S. Night sky photometry and spectroscopy performed at the Vienna University Observatory. J. Quant. Spectrosc. Radiat. Transf. 2014, 139, 64–75. [Google Scholar] [CrossRef] [Green Version]
  30. Hale, J.D.; Davies, G.; Fairbrass, A.J.; Matthews, T.J.; Rogers, C.D.F.; Sadler, J.P. Mapping Lightscapes: Spatial Patterning of Artificial Lighting in an Urban Landscape. PLoS ONE 2013, 8, e61460. [Google Scholar] [CrossRef] [PubMed] [Green Version]
  31. Search Photos. Available online: https://eol.jsc.nasa.gov/SearchPhotos/ (accessed on 7 April 2020).
  32. Home-Savestars Consulting S.L. Available online: http://www.savestarsconsulting.com/en/index/ (accessed on 6 May 2021).
  33. Sánchez de Miguel, A. Colour remote sensing of the impact of artificial light at night (II): Calibration of DSLR-based images from the International Space Station. Remote. Sens. Environ. 2021, 264, 112611. [Google Scholar] [CrossRef]
  34. Illuminance Spectrophotometer CL-500A. Available online: https://www.konicaminolta.com/instruments/download/catalog/light/pdf/cl500a_catalog_eng.pdf (accessed on 6 May 2021).
  35. MODTRAN®. Available online: http://modtran.spectral.com/ (accessed on 13 June 2021).
  36. Shi, C.; Wang, L. Incorporating spatial information in spectral unmixing: A review. Remote Sens. Environ. 2014, 149, 70–87. [Google Scholar] [CrossRef]
  37. Bro, R.; De Jong, S. A fast non-negativity-constrained least squares algorithm. J. Chemom. 1997, 11, 393–401. [Google Scholar] [CrossRef]
  38. Mahalanobis, P. On the generalized distance in statistics. 1936. [Google Scholar]
  39. Breiman, L. Random forests. Mach. Learn. 2001, 45, 5–32. [Google Scholar] [CrossRef] [Green Version]
  40. EarthExplorer. Available online: https://earthexplorer.usgs.gov/ (accessed on 22 June 2021).
  41. Landsat Normalized Difference Vegetation Index. Available online: https://www.usgs.gov/core-science-systems/nli/landsat/landsat-normalized-difference-vegetation-index?qt-science_support_page_related_con=0#qt-science_support_page_related_con (accessed on 22 June 2021).
  42. Normalized Difference Moisture Index. Available online: https://www.usgs.gov/core-science-systems/nli/landsat/normalized-difference-moisture-index (accessed on 22 June 2021).
  43. Schubert, E.F.; Kim, J.K. Solid-state light sources getting smart. Science 2005, 308, 1274–1278. [Google Scholar] [CrossRef]
  44. Kyba, C.C.M.; Ruby, A.; Kuechly, H.U.; Kinzey, B.; Miller, N.; Sanders, J.; Barentine, J.; Kleinodt, R.; Espey, B. Direct measurement of the contribution of street lighting to satellite observations of nighttime light emissions from urban areas. Light. Res. Technol. 2020. [Google Scholar] [CrossRef]
  45. Czeisler, C.A. Perspective: Casting light on sleep deficiency. Nature 2013, 497, S13. [Google Scholar] [CrossRef] [PubMed]
  46. Buján, S.; González-Ferreiro, E.; Reyes-Bueno, F.; Barreiro-Fernández, L.; Crecente, R.; Miranda, D. Land Use Classification from Lidar Data and Ortho-Images in a Rural Area. Photogramm. Rec. 2012, 27, 401–422. [Google Scholar] [CrossRef]
  47. Hermosilla, T.; Ruiz, L.; Recio, J.; Balsa-Barreiro, J. Land-Use Mapping of Valencia City Area from Aerial Images and LiDAR Data. In Proceedings of the GEOProcessing 2012: The Fourth International Conference in Advanced Geographic Information Systems, Applications and Services, Valencia, Spain, 30 January–4 February 2012; pp. 232–237. [Google Scholar]
  48. Megahed, Y.; Shaker, A.; Sensing, W.Y.-R. Fusion of Airborne LiDAR Point Clouds and Aerial Images for Heterogeneous Land-Use Urban Mapping. Remote Sens. 2021, 13, 814. [Google Scholar] [CrossRef]
  49. Haifa, Israel-Detailed Climate Information and Monthly Weather Forecast|Weather Atlas. Available online: https://www.weather-atlas.com/en/israel/haifa-climate#climate_text_4 (accessed on 26 October 2021).
  50. Bará, S.; Rodríguez-Arós; Pérez, M.; Tosar, B.; Lima, R.C.; Sánchez de Miguel, A.; Zamorano, J. Estimating the relative contribution of streetlights, vehicles, and residential lighting to the urban night sky brightness. Light. Res. Technol. 2019, 51, 1092–1107. [Google Scholar] [CrossRef] [Green Version]
  51. Hermosilla, T.; Palomar-Vázquez, J.; Balaguer-Beser, Á.; Balsa-Barreiro, J.; Ruiz, L.A. Using street based metrics to characterize urban typologies. Comput. Environ. Urban Syst. 2014, 44, 68–79. [Google Scholar] [CrossRef] [Green Version]
Remotesensing 13 04413 g001 550
Figure 1. Methodological scheme of the study.
Figure 1. Methodological scheme of the study.
Remotesensing 13 04413 g001
Remotesensing 13 04413 g002 550
Figure 2. Georeferenced ISS-produced (taken on 29.11.2015 by Nikon D4 DSLR camera; ID ISS045-E-148262) (a) and radiometrically calibrated (b) night-time image of the Haifa region. In situ measurements (n = 610), performed in March 2015 via Konica Minolta CL-500A spectrometer, are marked by red dots in the right sub-figure.
Figure 2. Georeferenced ISS-produced (taken on 29.11.2015 by Nikon D4 DSLR camera; ID ISS045-E-148262) (a) and radiometrically calibrated (b) night-time image of the Haifa region. In situ measurements (n = 610), performed in March 2015 via Konica Minolta CL-500A spectrometer, are marked by red dots in the right sub-figure.
Remotesensing 13 04413 g002
Remotesensing 13 04413 g003 550
Figure 3. Graphical illustration of the constituents for synthetic bands’ radiance computation: spectral sensitivity of RGB bands of Nikon D4 DSLR camera (a), atmospheric transmittance (b), and reference spectrum of AB magnitude system (c).
Figure 3. Graphical illustration of the constituents for synthetic bands’ radiance computation: spectral sensitivity of RGB bands of Nikon D4 DSLR camera (a), atmospheric transmittance (b), and reference spectrum of AB magnitude system (c).
Remotesensing 13 04413 g003
Remotesensing 13 04413 g004 550
Figure 4. Spectral signatures of selected representatives of each lamp type used in the analysis. Note: The following lamps’ signatures are depicted: ‘CFL’ = Compact Fluorescent Lamp of 2776K; ‘CMH’ = Ceramic Metal Halide of 2829K; ‘HPS’ = High-Pressure Sodium of 2005K; ‘LPS’ = Low Pressure Sodium of 1701K; ‘Incand’ = Incandescent Tungsten of 2805K; ‘MV’ = Mercury Vapor of 4717K; and ‘LED’ = LED of 3033K. At that, for lamp types under analysis, the temperatures varied in the diapason of 2478-5576K for ‘CFL’; 2378–6075K for ‘MH’; 1732–2176K for ‘HPS’; 2021–2911K for ‘Incand’; 4075–4741K for ‘MV’; and 1833–9089K for ‘LED’.
Figure 4. Spectral signatures of selected representatives of each lamp type used in the analysis. Note: The following lamps’ signatures are depicted: ‘CFL’ = Compact Fluorescent Lamp of 2776K; ‘CMH’ = Ceramic Metal Halide of 2829K; ‘HPS’ = High-Pressure Sodium of 2005K; ‘LPS’ = Low Pressure Sodium of 1701K; ‘Incand’ = Incandescent Tungsten of 2805K; ‘MV’ = Mercury Vapor of 4717K; and ‘LED’ = LED of 3033K. At that, for lamp types under analysis, the temperatures varied in the diapason of 2478-5576K for ‘CFL’; 2378–6075K for ‘MH’; 1732–2176K for ‘HPS’; 2021–2911K for ‘Incand’; 4075–4741K for ‘MV’; and 1833–9089K for ‘LED’.
Remotesensing 13 04413 g004
Remotesensing 13 04413 g005a 550Remotesensing 13 04413 g005b 550
Figure 5. B/G vs. G/R ratios of the simulated synthetic bands: Fluorescent (a), Metal Halide (b), HPS (c), Incandescent (d), Mercury Vapor (e), and LED (f) lamps. Note: LPS lamp, reporting G/R = 0.18 and B/G = 0.02, is the only representative of the type and is not depicted in the figure.
Figure 5. B/G vs. G/R ratios of the simulated synthetic bands: Fluorescent (a), Metal Halide (b), HPS (c), Incandescent (d), Mercury Vapor (e), and LED (f) lamps. Note: LPS lamp, reporting G/R = 0.18 and B/G = 0.02, is the only representative of the type and is not depicted in the figure.
Remotesensing 13 04413 g005aRemotesensing 13 04413 g005b
Remotesensing 13 04413 g006 550
Figure 6. Main characteristics of selected representative ground truth measurements: localities (marked as red dots) (a), G/R and B/G ratios (b), and Euclidian distances distribution from the corresponding ISS pixel in G/R, B/G coordinate system (c).
Figure 6. Main characteristics of selected representative ground truth measurements: localities (marked as red dots) (a), G/R and B/G ratios (b), and Euclidian distances distribution from the corresponding ISS pixel in G/R, B/G coordinate system (c).
Remotesensing 13 04413 g006
Remotesensing 13 04413 g007a 550Remotesensing 13 04413 g007b 550
Figure 7. Model-anticipated relative contribution of High-Pressure Sodium (a,c) and Metal Halide (b,d) lamps in the outdoor lighting in Haifa area (a,b) and selected ROIs (c,d): “1” = Haifa Bay Area; “2” = Khof Shemen Industrial Zone; “3” = Haifa Airport. Note: Water and vegetation masks (marked, respectively, blue and green) were produced from Landsat 8 OLI/TIRS C2 L2 dataset image (path 174 row 037) acquired on 8 June 2015 [40]. For vegetation mask, normalized difference vegetation index (NDVI) was calculated as NDVI = (Band5 − Band6)/(Band5 + Band6) [41], and levels in 0.22–0.25 and >0.25 diapasons were marked as light and dark green. For the water mask, normalized difference moisture index (NDMI) was calculated as NDMI = (Band5 − Band4)/(Band5 + Band4) [42].
Figure 7. Model-anticipated relative contribution of High-Pressure Sodium (a,c) and Metal Halide (b,d) lamps in the outdoor lighting in Haifa area (a,b) and selected ROIs (c,d): “1” = Haifa Bay Area; “2” = Khof Shemen Industrial Zone; “3” = Haifa Airport. Note: Water and vegetation masks (marked, respectively, blue and green) were produced from Landsat 8 OLI/TIRS C2 L2 dataset image (path 174 row 037) acquired on 8 June 2015 [40]. For vegetation mask, normalized difference vegetation index (NDVI) was calculated as NDVI = (Band5 − Band6)/(Band5 + Band6) [41], and levels in 0.22–0.25 and >0.25 diapasons were marked as light and dark green. For the water mask, normalized difference moisture index (NDMI) was calculated as NDMI = (Band5 − Band4)/(Band5 + Band4) [42].
Remotesensing 13 04413 g007aRemotesensing 13 04413 g007b
Table
Table 1. The performance of models (in terms of R2) with alternative predictor sets, for the training and the testing sets (dependent variable = relative contribution of a certain lamp type).
Table 1. The performance of models (in terms of R2) with alternative predictor sets, for the training and the testing sets (dependent variable = relative contribution of a certain lamp type).
Model
Type		SubsetLamp Type
HPSMHLEDCFLLPSIncandMV
Predictor set 1: R, G, B
Linear regressionTraining0.340.230.030.060.160.090.17
Testing0.280.18−0.110.000.070.030.13
Neural networkTraining0.390.07−0.44−0.32−0.96−0.32−0.58
Testing0.34−0.09−1.27−0.46−1.72−0.48−0.96
Random forestTraining0.880.860.770.750.790.810.82
Testing0.460.40−0.170.190.000.250.39
Predictor set 2: G/R, B/G, GG/RB
Linear regressionTraining0.520.330.020.110.210.090.22
Testing0.500.29−0.080.060.080.000.17
Neural networkTraining0.530.390.040.140.190.170.31
Testing0.470.32−0.120.04−0.020.020.21
Random forestTraining0.870.840.740.720.800.680.83
Testing0.550.47−0.030.120.22−0.160.53
Predictor set 3: G/R, B/G, Mahalanobis distance
Linear regressionTraining0.530.350.030.110.210.100.34
Testing0.500.30−0.080.070.10−0.040.30
Neural networkTraining0.240.400.050.130.130.130.27
Testing0.190.32−0.160.01−0.04−0.050.20
Random forestTraining0.870.850.740.750.780.810.84
Testing0.530.470.130.200.190.380.43
Predictor set 4: G/R, B/G, Euclidean distance to the line
Linear regressionTraining0.520.340.050.110.210.110.29
Testing0.500.29−0.070.070.10−0.120.24
Neural networkTraining0.260.360.060.110.130.130.29
Testing0.220.28−0.25−0.03−0.04−0.100.22
Random forestTraining0.860.840.730.750.780.810.85
Testing0.490.450.060.190.190.380.53
Note: The lamps are sorted from left to right in descending order of their relative contributions (see results of unmixing analysis, reported in Appendix A Table A1).
Publisher’s Note: MDPI stays neutral with regard to jurisdictional claims in published maps and institutional affiliations.

Share and Cite

MDPI and ACS Style

Rybnikova, N.; Sánchez de Miguel, A.; Rybnikov, S.; Brook, A. A New Approach to Identify On-Ground Lamp Types from Night-Time ISS Images. Remote Sens. 2021, 13, 4413. https://doi.org/10.3390/rs13214413

AMA Style

Rybnikova N, Sánchez de Miguel A, Rybnikov S, Brook A. A New Approach to Identify On-Ground Lamp Types from Night-Time ISS Images. Remote Sensing. 2021; 13(21):4413. https://doi.org/10.3390/rs13214413

Chicago/Turabian Style

Rybnikova, Natalia, Alejandro Sánchez de Miguel, Sviatoslav Rybnikov, and Anna Brook. 2021. "A New Approach to Identify On-Ground Lamp Types from Night-Time ISS Images" Remote Sensing 13, no. 21: 4413. https://doi.org/10.3390/rs13214413

Note that from the first issue of 2016, this journal uses article numbers instead of page numbers. See further details here.

Article Metrics

Back to TopTop