Light At Night: some Sensors and some Applications

Transcription

1 Light At Night: some Sensors and some Applications Marina V. Compagnucci Master in Emergency Early Warning and Response Space Applications Seminar, September 2 nd, 2010

2 Contents Abstract Introduction Night-Time Light Sensors SAC-C satellite and the onboard HSTC AVIRIS The DMPS-OLS sensor The ideal Night-time light sensor: NightSat DMPS-OLS types of products Frequency detection (Stable lights) Radiance calibrated Average Digital Number Differences among the three data sets Spatial and Temporal properties of the DMSP-OLS data set Spatial Characteristics Coarse spatial resolution Large Overlap between pixels Errors in the geolocation Temporal Characteristics Some applications for Night-time Light products Urban Extent Population Mapping Economy and Energy Consumption Economy Energy consumption Fisheries Protected areas Night-time Light and Breast Cancer Conclusions Bibliography

3 Abstract Light measured at night is an unmistakable sign of humankind presence on Earth. It points out human activity, be it on land or in water. This data coming from various sensors offers information about human settlements and their development, shipping fleets, and epidemiology. Likewise it shows how mankind can affect their own surroundings and the possible consequences on them, for instances protected areas near urban areas. Light is not only seen as a pollutant but also as a proxy for mapping economy, resource use and urban extension. This works aims to present the sensors that measure Night-time Light, their products and some of the various and creative uses of this data.

4 Introduction The presence of light across the Earth s surface provides some of the most visually stunning and thought provoking scenes from space. In this way, the human occupation footprint is uniquely visible from space in the form of artificial night lighting ranging from the burning of the rainforest to massive offshore fisheries to the omnipresent lights of cities and towns and related connecting road networks (Aubrecht et al. 2008, Doll 2008). The discovery that lights could be observed at night from a sensor that was initially conceived to observe clouds at night is one of the most fortuitous unforeseen benefits to have come from remote sensing technology. Figure 1: The World Atlas of the Artificial Night Sky Brightness Cinzano et al. (2001) once said that light pollution is the alteration of the ambient light levels in the night environment produced by man-made light. Light pollution is a broad term referring to excessive or obtrusive artificial light caused by bad lighting design. It includes such things as glare, sky glow, and light trespass. Excessive and misdirected light from streetlights, homes, and towns not only interferes with wildlife, stargazing, sleep habits, and professional astronomy, but it also wastes a vast amount of energy (Gallaway et al., 2010). Light pollution, a problem that affects almost any urban areas, is produced by a large number of lighting sources, which spill light into the sky. Due to the presence of dust and aerosols in the atmosphere the light is scattered, brightening the sky (Cinzano et al., 2001). One of the effects of the brightened sky is that stars and other astronomical objects, that are relatively faint, are lost in the background glow. While most people have a sense that artificial lighting can interfere with birds and insects, the effects are far more common, widespread, and serious than commonly realized. Light pollution does substantial damage to wildlife, aesthetics, and even to human health. Mammals, birds, amphibians, insects, fish and even plants are all affected

5 by light pollution. Light pollution disrupts feeding, reproduction, sleeping and migration. Indeed, problems from light pollution are so pervasive that unless we consider protection of the night, our best-laid conservation plans will be inadequate (Rich and Longcore, 2006). For example, light pollution disrupts the migration patterns of nocturnal birds and can cause hatchling sea turtles to head inland, away from the sea, and be eaten by predators or run over by cars (Salmon et al., 2000). Human physiology is not immune to the problem of light pollution. Davis et al., (2001) have concluded that there is an increased risk of breast cancer in women due to lower levels of melatonin production that results from light pollution. Ostensibly, light pollution keeps people from falling into a deep sleep, which causes their bodies to decrease the production of melatonin.. Light pollution also interferes with both professional and amateur astronomy by reducing the visibility of galaxies, nebulae, and other celestial objects. One important issue with observing artificial night lighting from space that needs to be addressed is a phenomenon known as skyglow. Even in its pristine state the night sky is not completely dark. Some light comes from the stars, some from sunlight scattered by space dust in the plane of the solar system, and some from atmospheric gases subject to radiation and particle fluxes mostly from the sun (Clark, 2008). This is called natural skyglow. Light emitted from human settlements in the atmosphere is refracted or scattered by air and water molecules and suspended particles (atmospheric aerosol) caused by dust, pollen, salt from sea spray, and waste products from industry. Artificially illuminating the sky over great distances this is called artificial skyglow. According to Clark (2008) the total artificial light flux emitted by a city tends to be proportional to the product of two quantities, (1) the number of light sources and (2) their mean output of light. Related to a growing economy and urban population growth typically both of these quantities increase over time. Many people assume artificial light provides safety and improves visibility. However, a large portion of lighting does neither. Lighting that is overused, misdirected, or otherwise obtrusive is simply pollution. This works aims to present the sensors that measure Night-time Light, their products and some of the various and creative uses of this data as well as their limitations.

6 Night-time Light Sensors SAC-C satellite and the onboard High Sensitivity Technology Camera (HSTC) The HSTC is a camera travelling onboard the SAC-C, the first Argentinean satellite scientifically used, launched on This camera has a spatial resolution of 300 m, the swath is 700 km and a spectral coverage between 450 and 850 nm. It operates during the night overpass (22:30 local time). The purpose of this instrument is to measure light use and misuse in human settlements, monitor thunder storms as well as forest fires (Figures 2 and 3). It is also used to study dynamic and evolution of polar auroras (extracted from: ). Figure. 2: Buenos Aires city Night-time light detection by SAC-C.

7 Figure. 3: Santiago de Chile and Mendoza cities Night-time light detection by SAC-C. AVIRIS In the absence of space borne sensors, researchers have used sensors mounted on aircraft to fly high altitude missions. The Airborne Visible/Infrared Imaging Spectrometer (AVIRIS) is one such sensor that may be used to acquire high-resolution data over individual cities at night. The AVIRIS sensor is a hyperspectral imaging system that senses in 224 very narrow bands (~10nm) from μm. This additional data source offers not only the advantage of an enhanced spatial resolution, but also of enhanced spectral resolution too. AVIRIS data could address this issue. A test flight over Las Vegas in 1998 suggested that there are distinctive spectral signatures over the city (Elvidge and Jansen, 1999; Doll, 2003). Combining these two data sources would be of use to help understand what the DMSP-OLS data is really showing at the small scale, and therefore aid the assumptions one makes in macro-scale models using nighttime imagery. There are various types of lighting used in cities. Each has distinct spectral characteristics depending on the element used. Commonly used types of high intensity discharge lights are high pressure sodium used for street lights, mercury vapour and metal halide used in lighting car-parks and sports stadiums. Mapping spectral patterns over cities could help to identify patterns of residential, commercial and industrial land-use (Elvidge and Jansen, 1999). This could be one way of filtering out the population component if concerned with assessing areas of high economic activity.

8 The DMSP-OLS sensor The Defense Meteorological Satellite Program, (DMSP) is the meteorological program of the US Department of Defense, which originated in the mid-1960s with the objective of collecting worldwide cloud cover on a daily basis (Kramer, 1994). Orbiting the Earth 14 times a day means that global coverage can be obtained every 24 hours. It has incorporated the Operational Linescan System (OLS) instrument, having two broadband sensors, one in the visible/near-infrared (VNIR, μm) and thermal infrared ( μm) wavebands. The OLS is an oscillating scan radiometer with a broad field of view (~3,000km swath) and captures images at a nominal resolution of 0.56km (see Table I for details). Low-level light amplification in the visible channel is facilitated through the use of a photomultiplier tube (PMT) so clouds illuminated by moonlight at night can be observed. The boost in gain enables the unique capability of observing lights present at the earth s surface at night. Most of the lights are from human settlements (Elvidge et al. 1997) and ephemeral fires (Elvidge et al. 2001). Furthermore gas flares and offshore platforms as well as heavily lit fishing boats can be identified. NOAA-NGDC archives the long-term DMSP data from 1992 to present. Although this was done with the initial aim of producing night-time cloud imagery on which to base short term cloud cover forecasts, a fortuitous unforeseen benefit was also discovered: city lights, gas flaring, shipping fleets and biomass burning can also be detected in the absence of cloud cover (Elvidge et al., 1997, Croft, 1978). Type OLS Oscillating Scan Radiometer Sensor Photo Multiplier Tube (PMT) Satellite NOAA-DMSP, sun-synchronous polar orbit Altitude and orbital period 830 km, 101 min Spectral range nm nm Spatial resolution 2.8 km at nadir (on-board averaging of 5 5 blocks at 0.56 km) Swath 3000 km Dynamic range 10 9 W cm 2 sr 1 nm 1 range of Table I. Main properties of the DSMP-OLS instrument (Extracted from: Barducci et al, 2006 Hyperspectral remote sensing for light pollution monitoring)

9 The ideal Night-time light sensor: NightSat The spatial resolution is recommended to be 25-50m. Based on experiments resampling the 1.5m Cirrus imagery, this was determined to be the maximum resolution permissible for delineating primary night-time lighting patterns. At this resolution and a swath of 80-90km, there would be a revisit period of ~30 days, at the equator yielding 12 views per year. The overpass pass time would be 9.30pm local time to provide the temporal consistency for change detection. As with DMSP, cloud and fire screening would be done with a separate thermal band. A key feature would include on-board calibration or a repeatable procedure for calibrating sensor data to radiance units and allow comparisons over time and between future sensors. There are essentially three types of lights which are detected: Flames such as lanterns and gasflares; Incandescent, where light is produced from a heated filament; and Vapour lamps where lighting is generated by electrically charged gasses such as mercury, sodium and neon. Incandescent and vapour lamps are most common for outdoor lighting. Each type of light has a distinctive spectral signature, which would be detectable if the new satellite had four band multispectral sensors to define the predominant type or mixture of lighting present (Figure 4). The ability to distinguish different types of lighting will have benefit for a number of applications. Classifying urban land use the use of different types of light is one promising area, especially as lighting practices tend to be homogeneously determined at some municipal, regional or even national scale. For ecological applications, the presence of certain wavelengths determines whether species will respond to lights or not. Sea turtle nesting and seafinding behaviors are not affected by lights with only yellow wavelengths (Salmon 2006), whilst salamanders and some birds show difficulty to navigate under certain lighting conditions. Some salamanders are unable to navigate properly under yellow light, while insects are attracted to short, ultraviolet light (Wiltschko and Wiltschko, 2002). Figure. 4: Field spectra of four different types of nocturnal lighting(extracted from Doll,2008)

10 DMSP-OLS types of products At the time of writing, there are currently three processed versions of night-time light data sets products which are released by NOAA-NDGC. There are three different types of imagery associated with the DMSP-OLS data set. Frequency detection (Stable lights) Radiance calibrated Average Digital Number Frequency detection (Stable lights) Whilst lunar illumination was crucial to imaging clouds at night, it is a hindrance to observing light sources from the ground due to the reduced contrast of light sources from the ground. Other hindrances include glare from scattered sunlight and bad scan lines. Filtering out bad scan lines (defined as 10 consecutive lights with no lights above of below) also removes lit pixels caused by lightning (Elvidge, 2001). Over the sixmonth period a temporal composite was built up of cloud free images of the earth at night. Compositing not only allowed clouds to be excluded, but also facilitated the analysis of stable lights. The presence of stable lights is important in distinguishing different light sources (e.g. city lights, shipping fleets or forest fires). However, the variation in brightness between orbits means that it is not possible to establish a single digital number (DN or at sensor radiance) threshold for identifying VNIR emission sources (Elvidge et al. 1997). To overcome this, an algorithm was developed to automatically detect light using a nested configuration of 200x200 and 50x50 pixel blocks. The light-picking algorithm applies a threshold to the central 50x50 pixel block based on the histogram of the surrounding 200x200 pixel block. Using this detection algorithm, the pixel value is assigned according to the percentage of times light was detected during cloud-free overpasses. Analysing the temporal frequency and stability of lights can help to distinguish their most likely source. City lights can be identified because they are temporally stable. However, forest fires can also be identified due to their location and lack of temporal stability over the compositing period. Through this process, the global night-time light composite can be filtered into a variety of different products: lights from human settlements and industrial facilities (city lights) fires gas flaring shipping fleets One issue with this data set is that certain areas of the globe receive more cloud-free views than others. This creates problems for the fire product, which often occurs in cloud-covered tropical. It should be noted that NOAA-NDGC do not feel 6 months

11 worth of data was sufficient to fully discriminate between stable lights and fire. This is currently being investigated using dedicated fire products from other satellites such as MODIS, part of NASA s Earth Observation System. One of the biggest problems encountered with this first version of night-time lights was low-light level pixel saturation. Radiance calibrated The problems of relatively low-level pixel saturation from the 6-bit sensor over bright urban areas led to the experimentation and ultimate production of a new low-gain data set. by varying the gain of the sensor The thresholding technique used to create the stable lights data set was found to perform poorly at identifying diffuse lighting, which is often dim and spatially scattered across the landscape The range was made deliberately ample on either side to allow for any future variations in gain. Since the DN variation is a physically meaningful quantity as opposed to a litfrequency observation, this makes it a flexible data set for use in a variety of modelling schemes subject to finding appropriate relationships between radiance and the parameters of interest. Average Digital Number The latest and now most extensive release of night-time light data comes in the form of average Digital Number (DN) values. The data was processed to use the high quality visible band data. Pixels were screened to remove those with lunar illumination, glare, bad scanlines and lightning and other marginal data (Elvidge et al., 2001). This has recently been extended to a full archive of data from every sensor for every year. This facilitates the analysis of changing lighting patterns in the following ways: The appearance of new light sources The disappearance of light sources The expansion and contraction of light sources Positive and negative changes in the brightness of lights.

12 Differences among the three data sets When choosing which data sets to use, it is of great importance to keep in mind the main scope of the research or even the field of study in which the results will be involved. A visual comparison is presented below in Figure 5 to gain an appreciation of the differences between the three data sets described above. It is apparent the initial stable lights product has far less variation than the other two and the imagery saturates very rapidly at the maximum 100 percent frequency detection value. The stable lights product has a large number of pixels taking the highest range of values. The distribution of values is spread more evenly in the radiance-calibrated version with the majority of pixels in the low range and only very few at the brightest radiance values indicating sensor saturation (even with the gain turned down) (Doll, 2008). Figure. 5: Comparison between the three different data sets over a portion of New England (from left to right: Stable Lights, Radiance Calibrated, and Average DN). (extracted from Doll, 2008)

13 Spatial and Temporal properties of the DMSP-OLS data set There are both spatial and temporal properties of the DMSP-OLS data set which affect the efficacy of the data set for its range of applications. Essentially these relate to changes in the spatial extent of lit areas and variations in the brightness of a pixel over time. Depending on the nature of the application at hand, the relative importance of spatial extent versus information content (DN) will vary. Understanding how lights behave in space and time will lead to the sound scientific use of the data set and minimize misinterpretation of the results. Spatial Characteristics The principal spatial consideration to bear in mind when working with night-time light imagery is the extent to which the spatial area depicted on images matches the true extent of lit area on the ground. Imagery from the DMSP-OLS satellite has a tendency to overestimate this parameter, an effect generally referred to as blooming (and more recently overglow ) in the literature. Small et al., (2005) cites three main reasons for this phenomenon, which are discussed briefly. Coarse spatial resolution Although the DMSP-OLS sensor has a nominal resolution of 1km, this has been resampled from the 2.7km native resolution of the sensor. An inherent feature of satellite imagery is that it will generalize ground based features to a single DN or radiance value. In the case of night time light imagery, this manifests itself as pixels appearing lit, when the light source is not being emitted over the entire pixel area. Large Overlap between pixels A feature of the data acquisition process is that there is a large overlap (some 60%) between pixels. This means that light observed in one location has the chance to be recorded in more than one pixel. This can contribute to a larger lit-area being detected than is actually the case. Errors in the geolocation Errors are inherent in the projection process. Data is recorded in arrays, the spatial position of these data values are calculated from the navigation data onboard the satellite. These values are then projected onto a 1km grid. The grid itself is an approximation of the Earth s surface corrected for the topographic variation. Each transformation introduces errors into the process. To this a fourth factor, the atmospheric water vapour content can be added. Lights can appear dimmer and more spatially diffuse where thin clouds are present, which is consistent with similar effects of image quality reduction for other optical (or passive ) sensors. The combined effect of these factors ultimately results in a general overestimation of area, which can be deceiving due to the visually stunning nature of the data set. Figure 6 illustrates the blooming effect, and also shows different sources of light that can be observed from the DMSP- OLS sensor. It is apparent that cities appear lit, but so

14 too do traffic on unlit sections of highway and areas that are unlit but which are affected by overglow. To correct this effect, thresholding (excluding values below a certain value) has been used to reduce the area of the lights. However, it soon becomes apparent that there is no single threshold that can be applied which would match the urban delimitation for all cities. In particular, thresholding large urban areas tend to result in the attenuation of lights associated with smaller settlements. The implication of this finding is that a range of thresholds needs to be applied depending on the size of the settlement involved (Small et al., 2005) Figure. 6: combined effect of the three previous factors ultimately results in a general overestimation of area (extracted from Doll, 2008). Temporal Characteristics Little work has been done regarding this feature, but that published suggests that low level saturation pixel, prevent city centre analysis, but it does allow investigation of the spatial expansion of lights in peri-urban areas. Temporal changes involve constructing tri-band red, green, blue false color composites with a different year for each channel. The convention has been to put the 1992 year through the red channel, 1998 in the blue, and 2003 in green. Superimposition of these channels can reveal whether lighting has been lost (red hues), gained (green hues) or emerged then disappeared (blue hues). This is most striking in places which have undergone massive economic/political change such as the countries of Eastern Europe following the fall of communism and the transition to free market economies. In Figure 7, we see a temporal color composite, showing that the former Soviet republics of Ukraine and Moldova are dominated by red hues indicating lights were most prevalent in 1992 and then declined in 1998 and This is sharply contrasted by Poland and Romania to the west whose greener and bluer hues indicate spatial expansion and brightening of lights.

15 The problem with the three color composites over long time periods is that they come from sensors on board different satellites and there is no internal or cross calibration between them. For practical purposes this means that we cannot say with any certainty whether changes in the brightness of lights are due to changing ground conditions or to changes in the sensor over time (Doll, 2008). Poland Ukraine Moldova Romania Figure. 7: temporal color composite, showing that the former Soviet republics of Ukraine and Moldova are dominated by red hues( lights were most prevalent in 1992). This is contrasted by Poland and Romania to the west whose greener and bluer hues indicate spatial expansion and brightening of lights. (Extracted from Doll, 2008) Given the range of variations that can occur with the night-time lights data set, any application will need to take into account the limitations of using this data source. For applications where light will be used solely as a delimiter of urban extent, then considerations of blooming (overglow) are most pertinent. Of the two phenomena, overglow is currently the best understood and strategies exist in the literature (Small et al., 2005) to account for its effects.

16 Some applications for Night-time Light products To shed new light on prevalent problems. Brief comments on: cancer, fisheries, poverty, urban extent, energy consumption and protected areas. Urban Extent Global land cover maps are spatial classifications of the Earth s surface. They are traditionally focused on the major vegetated biomes and cropland areas, with urban areas being the residual. This tends to underestimate urban area. By contrast night-time light imagery explicitly maps lit areas, however the overglow characteristic, means that the resulting maps tend to overestimate urban extent. Doll and Muller (1999) found that unfiltered night-lights covered 20 times as much area at the continental level compared to urban delineation of the Digital Chart of the World data set. Considering artificial skyglow entails that the DMSP satellite sensors record much larger areas than just the immediate location of the lighting sources. Using satellite observed nighttime lights for delineating urban areas (Small et al., 2005) and approximating impervious surfaces (Elvidge et al., 2007) requires eliminating skyglow from the data, i.e. by applying thresholds to the digital number values. Previous studies of DMSP-OLS stable night lights have shown encouraging agreement between temporally stable lighted areas and various definitions of urban extent. However, these studies have also highlighted an inconsistent relationship between the actual lighted area and the boundaries of the urban areas considered. Applying detection frequency thresholds can reduce the spatial overextent of lighted area ( blooming ) but thresholding also attenuates large numbers of smaller lights and significantly reduces the information content of the night lights datasets. This suggests that night lights could provide a repeatable, globally consistent way to map size and spatial distributions of human settlements larger than some minimum detectable size or brightness (Small et al, 2005).

17 Population The night-time satellite sensor data provided by the DMSP/OLS have been used for global/continental urban mapping, showing linear relations with other socio-economic variables such as population, Gross Domestic Product, and electrical power consumption Researchers used DMSP-OLS data to create the first quantitative and accurate depiction of the artificial brightness of the night sky and make it available to the scientific community and governments. This data is particularly valuable because of the singular lack of data on light pollution. Direct measures of light pollution are ad hoc, groundbased measures are sporadic and limited (Cinzano et al., 2001). This lack of direct data has forced researchers to rely almost exclusively on populationbased models of light pollution. Indeed, there is a very strong connection between population and light pollution. Nevertheless, the apparent proportionality between population and sky glow breaks down going from large scales to smaller scales and looking in more detail (Cinzano et al., 2001). Amaral et al, (2005) explored the potential of the DMSP sensor data for regional studies analyzing the correlation between DMSP night-time light foci and population (Figure 8), and the correlation between DMSP night-time light foci and electrical power consumption in the Amazonia. It was found that the night-time light foci were related to human presence in the region, including urban settlements, mining, industries, and civil construction. Thus the DMSP/OLS imagery can be used as an indicator of human presence in the analysis of spatial temporal patterns in the Amazonia region. These results are very useful considering the continental dimension of Amazonia, the absence of demographic information between the official population census, taking place every 10 years, and the dynamics and complexity of human activities in the region. Therefore DMSP night-time light foci are a valuable data source for global studies, modeling, and planning activities when the human dimension must be considered throughout Amazonia. Figure. 8: Linear relation between DMSP night-time light area and the urban population for municıpios of the state of Para, Brazil.(extracted from Amaral et al., 2005)

18 Mapping Economy and Energy Consumption The light pollution data used in this paper are remote sensing data from satellite observations. The raw data are from the Defense Meteorological Satellite Program (DMSP) Operational Linescan System (OLS). Economy Night-time light remote sensing data has been shown to correlate with national-level figures of Gross Domestic Product (GDP). Night-time light imagery was found to correlate with Gross Regional Product (GRP) across a range of spatial scales. The radiance-calibrated dataset is usually used in place of the time frequency composite data (Elvidge et al., 1997) employed on previous researchs. The radiance calibrated data set facilitates the investigation of the relationship between brightness of the lights and GDP rather than lit area. The first ever global map of GDP produced using a country level lit area GDP relationship, was done by Doll (2003). The importance of scale as a concept is central to developing an understanding of human-environment interactions. While scale can have spatial, temporal, quantitative or analytical dimensions, the diversity of disciplines incorporated into the Human Dimensions of Global Change may only use a sub-set of these domains to understand their subject. The previous section made reference to the mismatch in population and radiance at fine scale resolutions. Given the coincidence of brightest lights and downtown areas which are nodes of economic activity, an obvious extension of the application of night-time lights is to map economic activity. Generally, the maps obtained during these studies, use only light to distribute economic activity (Figure 9). This is a reasonable assumption to make in developed countries where industry and service sectors can comprise over 90% of the economy. Although agricultural productivity is spatially more widespread it is represented in these maps as nodes i.e. the map records the agricultural activity in the towns which emit light, not in the fields where crops are being grown. This is an important caveat to the maps described here and one which would be the first item to address when improving these maps and extending them into the developing world where agricultural comprises a larger section of the national economy (Doll et al, 2006). An inverse or complementary application of night-time light data is to identify the location of the poor through the absence of light. However, some cultural differences in light use, could have led to erroneous interpretation results.

19 Figure 9: Map of estimated economic activity based on DMSP-OLS radiance-calibrated nighttime lights for 11 countries in the European Union (extracted from Doll et al, 2006).

20 Energy consumption Economic studies quantifying night-time light damage are only now beginning, mainly due to understanding that light pollution wastes energy. Accordingly, poor lighting design contributes to increased carbon dioxide emissions and global warming. Electricity needlessly being generated at a cost of $6.9 billion a year in the United States. Furthermore, this unnecessary electricity usage generated an additional 66 million metric tons of CO2 (Doll et al., 2006). With all of these, it is very often the case that the good in question becomes problematic when it is found in the wrong location or in the wrong amount, or when it affects the wrong population. We might argue that the good becomes a pollutant when its effects are something other than its intended purpose. Similarly, for humans, light that improves visibility is a good. However when lighting causes glare, or deepens shadows, or washes out the stars, this reduces visibility. Then, such light is light pollution. Neon lights might improve the visibility of a sign or a storefront. However, a thousand such displays merely add to the clutter and reduce the visibility of any individual sign. (Gallaway et al., 2010). A research group combined unique remote sensing data on light pollution with economic data from the World Bank to estimate fractional logit regression light pollution models (Gallaway et al., 2010), showing that population, as measured by the percent of the population living in urban areas, remains an important explanation for the existence of light pollution. However, real per capita GDP also tends to be a highly significant variable in explaining the percent of a country's population affected by different levels of light pollution. The relationship between income and light pollution is non-linear, since other economic factors such as foreign investment and land use patterns also tend to be significant. Quantifying the link between real GDP and various levels of light pollution across the globe is a significant first step in correcting economists' neglect of this important environmental issue.

21 Fisheries Illex argentinus, the Argentine short-finned squid, is an important species within the Patagonian shelf ecosystem, where it supports a major multi-national fishery. The fishing fleet operating in this region is comprised of jigging vessels which attract squid using powerful incandescent lights. These fishing lights are detectable in remotely sensed satellite imagery which makes the fishery unusually amenable to a large-scale analysis of its spatial dynamics. Using fishing light imagery allows a synoptic view of effort in both regulated and unregulated fisheries, and has been used as a method of detecting and quantifying fishery activity in a number of locations around the world. Long-term inter-annual variability in fleet distribution and extent was examined using imagery from the Defense Meteorological Satellite Program-Operational Linescan System (DMSP-OLS) for the period , by Waluda et al (2010). The fishery was found to occupy a wide area across the shelf and slope, with regions of consistent fishing activity observed on the high seas (45 47 S) and to the north of the Falkland Islands (Malvinas). Distribution of the fishery over the 13-year study period was variable, with 28% of the fished area occupied in 1 2 years, and 7% of the area occupied in years. Annual catch levels were positively associated with the extent of the area occupied by the fleet. Higher catches corresponded to the fishery occupying a wide latitudinal range, whereas lower catches were observed during 2004 and 2005 corresponding to a contraction of the fishery away from the south of its range. In years of very high catches, fishing took place along almost the entire latitudinal range of the species. Due to the intensity of fishing, changes in the distribution of the fleet can reflect shifts in the distribution of I. argentinus; this has potential for the long-term monitoring of this highly variable squid fishery (Waluda et al, 2010). The variable distribution of the fleet over the 13 years of the study is most likely to be related to shifts in ocean dynamics. I. argentinus has been shown to be associated with thermal gradient regions occurring between different water masses (for example at the interface between the Falkland current and Patagonian shelf waters, or between the Brazil and Falkland currents), which can vary widely in location and extent from year to year. The distribution of the fleet is therefore likely to be indicative of feeding aggregations of squid occurring at these fronts. An added complication is that squid schools may occur at different depths dependent on water temperature variability (Bazzino et al., 2005), which will further contribute to variability in fleet extent but cannot be directly assessed using remotely sensed satellite imagery.

22 Protected areas The following paragraphs try to contribute with insights about a novel point of view regarding global assessment of light pollution impact on protected areas. Most applications related to ecological impacts of artificial night lighting focus on adverse effects on light-sensitive ecosystems or species, such as coral reefs (Aubrecht et al. 2008), sea turtles (Salmon, 2006), and migrating birds (Montevecchi, 2006). As far as coral reefs concern, they are of great importance for a number of reasons: they are areas with remarkable biodiversity; are important for coastal protection; they provide people with seafood and new medicines; and they have a great recreational value. Corals and coral reefs are extremely sensitive; slight changes in the reef environment may have detrimental effects on the health of entire coral colonies (Aubrech et al.; 2008). Corals are highly photosensitive many species synchronize their spawning through detection of low light intensity from moonlight and coral reef structure is strongly influenced by illumination. Other marine invertebrates in coral communities synchronize reproduction by monthly patterns of lunar illumination (Bentley et al., 2001). Such extensive structuring of this community by light is undoubtedly disrupted by artificial lighting, which has no ecological analogue moonlight, starlight and bioluminescence are the only sources of light to which marine organisms are adapted (Hobson et al., 1981). Aubrecht et al. (2008) calculated a lights proximity index (LPI) assuming that the nearer a coral reef is located to an artificial night lighting source the greater is its potential endangerment from direct and indirect impacts (Figure 10). Figure 10: The area of Puerto Rico was chosen to show coral reefs being at high risk by artificial night lighting caused by development (left part). There are many reefs within a 25 km radius of cities and towns having high LPI values due to their close proximity to the lighting sources (which are shown as reference on the right). Reefs in regions around big cities such as the capital, San Juan, especially show particularly high LPI values and the corresponding red colour in the image (extracted from Aubrecht et al., 2008).

23 As the previously cited researches with coral reefs have demonstrated, that system is highly susceptible to light interference, hence on a different approach as when dealing with urban areas and their extent, when ecological issues are the ones being analyzed, skyglow is a significant factor of light pollution as already very low light intensities alter the natural environment. That is the reason why, when modeling ecological impact of artificial night lighting skyglow is considered to be an important contribution (Aubrecht et al., 2008). Following a global assessment of the degree to which each country s total land area is legally protected, light pollution impact and approximated human influence were calculated. To delineate the protected areas worldwide, data from the 2007 Annual Release of the World Database was used provided by UNEP-WCMC (United Nations Environment Programme-World Conservation Monitoring Centre) In the study carried out by Aubrecht et al., 2010, neither marine protected areas nor historical, archaeological, or cultural sites, nor those areas that were listed as proposed but not yet designated, were not considered for the analysis. To carry on with the research, resources such as nighttime lights data acquired by the U.S. Air Force Defense Meteorological Satellite Program (DMSP) Operational Linescan System (OLS) and the WDPA data (provided online for download as ESRI shapefiles and consist of both polygon and point features) were acquired (Figures 11 and 12). Following what was described in section DMSP OLS products; to identify the best nighttime lights data for creating an annual composite Aubrecht et al., 2010 followed the next standards: Only the center half of the orbital swath was used (best geolocation and sharpest features) Sunlight and moonlight were not present No solar glare contamination was allowed Only cloud-free images were used (based on thermal detection of clouds) Two different approaches were used to relate the areal distribution of artificial night lighting to the areas under protected status, by means of two new indices resulted from combining the global protected area distribution data and nighttime lights data: PALI, a Protected Area Light Pollution Index (report the proportion of protected areas per country being directly affected by light pollution) PAHI, a Protected Area Human Impact Index. The last one refers to a binary lights data set in which the focal neighborhood operator assigned the value 1 to each pixel within a 5px radius circle around a lighting source, whereas all pixels further away we classified as 0 (not affected). The result was linked with a country-protected area With the first one, the direct impact of lighting is evaluated, hence referred to as light pollution, i.e.the direct spatial overlap between satellite observed nighttime lights as derived from DMSP-OLS and protected areas as delineated in the WDPA. The second approach, considers that DMSP nighttime lights data, are an excellent proxy measure for human activities that impact neighboring areas. This results in having

24 pixels within a 5 pixel radius (about 5 km at the equator) of the actual lit area identified as being potentially at risk. The results indicate that regions in Europe and Asia Minor, the Caribbean, South and East Asia as well as in the Eastern part of the United States are most affected. Introducing aggregated data on biomes reveals that temperate broadleaf and mixed forests suffer the biggest impact both in terms of general light pollution as well as lighting in protected areas. Figure 11: Data from DMSP-OLS, nighttime lights of the world, sample figure (extracted from Aubrecht et al., 2010). Figure 12: WDPA data (extracted from Aubrecht et al., 2010 ).

25 Night-time Light and Breast Cancer Recent studies of shift-working women have reported that excessive exposure to light at night (LAN) may be a risk factor for breast cancer. However, no studies have yet attempted to examine the co-distribution of LAN and breast cancer incidence on a population level with the goal to assess the coherence of these earlier findings with population trends. Coherence is one of Hill s criteria for an inference of causality. Nighttime satellite images were used to estimate LAN levels in 147 communities in Israel. Multiple regression analysis was performed to investigate the association between LAN and breast cancer incidence rates and, as a test of the specificity of our method, lung cancer incidence rates in women across localities under the prediction of a link with breast cancer but not lung cancer. After adjusting for several variables available on a population level, such as ethnic makeup, birth rate, population density, and local income level, a strong positive association between LAN intensity and breast cancer rate was revealed ( p, 0.05), and this association strengthened ( p, 0.01) when only statistically significant factors were filtered out by stepwise regression analysis. Concurrently, no association was found between LAN intensity and lung cancer rate. These results provide coherence of the previously reported case-control and cohort studies with the co-distribution of LAN and breast cancer on a population basis. The analysis yielded an estimated 73% higher breast cancer incidence in the highest LAN exposed communities compared to the lowest LAN exposed communities (Figure 13) (Kloog et al, 2008). Figure 13: Hotspot analysis of breast cancer rates. Note: Red circles mark clusters of adjacent localities with significantly high rates of cancers (relative to the global mean), while green circles mark geographic clusters of localities with significantly low cancer rates.

26 Breast cancer incidence increases rapidly as societies industrialize. Many changes occur during the industrialization process, one of which is a dramatic alteration in the lighted environment from a sun-based system to an electricity-based system. Increasingly, the natural dark period at night is being seriously eroded for the bulk of humanity. Based on the fact that light during the night can suppress melatonin, and also disrupt the circadian rhythm, it was proposed in 1987 that increasing use of electricity to light the night accounts in part for the rising risk of breast cancer globally. Predictions from the theory include: non-day shift work increases risk, blindness lowers risk, long sleep duration lowers risk, and population level community nighttime light level codistributes with breast cancer incidence. Thus far, studies of these predictions are consistent in support of the theory. A new avenue of research has been on function of circadian genes and whether these are related to breast cancer risk. In particular, a length variant of Per3 (5-VNTR) has been associated with increased risk in young women, and this same 5-VNTR variant has also been found to predict morning diurnal type and shorter sleep duration compared to the 4-VNTR variant. An important question is how an effect of light-at-night (LAN) exposure on breast cancer risk might be modified by polymorphisms and/or epigenetic alterations in the circadian genes, and conversely whether light-at-night exposure (e.g., shift work) can induce deleterious epigenetic changes in these genes (Stevens, 2009).

27 Conclusions Light at night is a visible evidence of our activities, our change in habits and the way we change the surroundings. As appreciated most Light at Night-Time research, uses data obtained from the Defense Meteorological Satellite Program Operational Linescan System (DMPS-OLS) but there are some that are incurring in the use of hyperspectral instruments to investigate light pollution. Night-time light is an excellent proxy for human activities as well as for human settlements. Hence, it is a reliable tool when dealing with the lack of data on extended territories or with little accessibility, such as Amazonia. It can also provide information on resources use or misuse, for instance electrical power utilization. When light passes from benefit to a pollutant, it becomes a serious problem with implications for wildlife, human health, scientific research, energy consumption, global warming, and the unchanging pastime of observing the night sky. Pristinely dark skies are very scarce in the developed world and most of the world's population lives under skies with at least some light pollution. Recognizing the importance of a dark period during the day should help us understand the weight of the decisions made for example when planning human settlements near protected areas, and also allows identify those ones that may require additional resources for management owing to their proximity to urban areas. Scotobiology, a science recently initiated as a branch of chronobiology, is bringing about evidence supporting the need for that period of darkness, mainly in more photosensitive ecological systems, such as corals and coral reefs. An interesting and novel point of view is introduced when considering Light at Night Time as a human risk factor, much more when referring to breast cancer. There is huge amount of previous evidence mentioning some possible reasons for this relationship, such as gene mutation, melatonin circadian rhythm disruption and others. Currently there are some groups studying the likely connection between Light at Night time and prostate cancer. Everything previously cited is true, if the advantages in addition to the disadvantages of this data are acknowledged. In that way, the results offered by it use can be properly interpreted. Facors to take into account are among others, scale of the research s question, which product to use to asses in the most fitting way that question, as well as the field involved in the decision making. Light at night has many other uses than those cited in this seminar, for example ice and snow detection. It can also detect auroras, green house emissions and could be involved in disasters managements such as hurricane and forest fires. It could be also involved in the control of fishing fleets and their activities areas.