Modeling artificial light viewed by fledgling seabirds

Artificial light is increasing in coverage across the surface of our planet, impacting the behavioral ecology of many organisms. Attraction to sources of artificial light is a significant threat to certain fledgling shearwaters, petrels (Procellariidae), and storm-petrels (Hydrobatidae) on their first nocturnal flights to the sea. Disorientation by light can cause these birds to crash into vegetation or manmade structures, potentially resulting in death from physical injury, starvation, dehydration, predation by introduced predators, or collisions with vehicles. We developed a GIS-based method to model the intensity of artificial light that fledgling procellariids and hydrobatids could view en route to the ocean (to estimate the degree of threat that artificial light poses to these birds) and present two models for the island of Kauai as examples. These models are particularly relevant to the federally threatened Newell’s Shearwater, or `A`o (Puffinus newelli ), of which .30,000 fledglings have been collected in response to disorientation by lights on Kauai during the past 30 years. Our models suggest that there are few to no portions of Kauai from which young birds could fledge and not view light on their post-natal nocturnal flights, which is concerning given evidence of a Newell’s Shearwater population decline. In future work using this technique, night light intensity layers could be altered to model the effects of modified coastal light conditions on known and potential procellariid and hydrobatid breeding locations. Furthermore, certain methods presented herein may be applicable to other seabirds and additional taxa in which attraction to anthropogenic light poses a serious threat, including migratory passerines and hatchling marine turtles. Components of this modeling approach could potentially be used to spatially estimate effects of other point-source threats to ecological systems, including sound and air pollution.


INTRODUCTION
Artificial light is increasing rapidly across the surface of our planet (Cinzano et al. 2001) and its impacts on organism ecology include numerous examples of interference with the typical behaviors of certain animals (Longcore and Rich 2004).Fledglings of certain small-to medium-sized shearwaters, petrels (Procellariidae), and stormpetrels (Hydrobatidae) are attracted to, and disoriented by, sources of anthropogenic light on their post-natal nocturnal flights to the ocean (Hadley 1961, Harrow 1965, King and Gould 1967, Imber 1975, Reed et al. 1985, Telfer et al. 1987, Le Corre et al. 2002, Rodriguez and Rodriguez 2009, Miles et al. 2010), though the reason for this behavior remains largely unknown.This disorientation can cause them to fall to the ground following exhaustion and/or crashing into manmade structures and vegetation (a phenomenon termed ''fallout'').Once grounded, they are vulnerable to starvation, dehydration, predation by introduced mammals, and collisions with vehicles, which can result in large numbers of injured and dead individuals (Reed et al. 1985, Telfer et al. 1987, Le Corre et al. 2002, Rodriguez and Rodriguez 2009).Previous studies of light attraction exhibited by species at Reunion Island (Le Corre et al. 2002), in the Canary Islands (Rodriguez and Rodriguez 2009), and in Hawaii (Ainley et al. 2001;Griesemer and Holmes, in press) suggest that light-induced fledgling mortality can affect recruitment and potentially play a significant role in population decline.
One of the most familiar examples of fallout occurs each autumn on the island of Kauai, Hawaii, as fledgling Newell's Shearwaters (`A`o [Puffinus newelli]), Hawaiian Petrels (`Ua`u [Pterodroma sandwichensis]), and Band-rumped Storm-Petrels (`Ake`ake [Oceanodroma castro]) take their maiden flights to sea (Reed et al. 1985, Telfer et al. 1987, Ainley et al. 2001).Between 1978 and2009, more than 30,000 Newell's Shearwater fledglings were collected as fallout birds, in addition to over 300 Hawaiian Petrels and 20 Band-rumped Storm-Petrels (State of Hawaii, unpublished data).Notably, many fallout fledglings are never recovered, either due to predation and scavenging by introduced predators or birds simply landing in areas unlikely to be visited by the public.Consequently, the .30,000Newell's Shearwaters collected have been estimated to represent from 93% to as little as 50% of actual fallout, suggesting ;32,250-60,000 fledglings may have actually been grounded due to the effects of anthropogenic light (Ainley et al. 2001).Kauai residents are encouraged to assist fallout birds by delivering them to aid stations for a veterinary examination, after which they are released if deemed to be in physical condition appropriate for fledging (Reed et al. 1985, Telfer et al. 1987, Rauzon 1991).Many individuals are located, banded, and released each season, but their fate after release is unknown (Duffy 2010).
Conservation efforts for seabirds would benefit significantly by using a spatially-explicit model to estimate the degree to which young birds fledging from regions of known and potential breeding habitat may be affected by anthropogenic light.A map of stationary night light intensity is insufficient, however, as it does not account for fledgling movement.Such a model must account for the total intensity of light that can be viewed along a fledgling's entire path to the ocean because young birds fledging from dark breeding sites (in which no light can be viewed) may encounter artificial light subsequent to fledging and become disorientated.
We developed a GIS-based method to estimate the total intensity of artificial light that young procellariids and hydrobatids, fledging from any terrestrial region of interest, could potentially view along a least-cost path (with respect to topography) on their initial nocturnal flights to sea.Because an unknown threshold of light intensity may exist for attraction by birds to be exhibited (making the exact manner in which they would respond to various light intensities unknown), our model birds traveled along leastcost paths based on topography, accumulating values of viewable light intensity along those paths, versus following paths based on movement toward light.Two models, developed using 2009 artificial light data from a satellite image, are presented for the island of Kauai and their relevance to the Newell's Shearwater is emphasized.We compare the proportion of twodimensional surface area on Kauai covered by different categories of light intensity, including values of no light, calculated from four GIS layers: an artificial light intensity layer obtained from a satellite image, a layer accounting for the intensity of light that can be viewed from each island location, and the two models that account for light viewed along hypothesized fledgling flight paths.This allowed a comparison of the coverage of three types of dark area on Kauai: area with no night light output, area from which artificial light could not be viewed, and area from which birds could fledge and not view artificial light along their flight paths.In addition, we present the island-wide pattern of Newell's Shearwater fallout on Kauai that occurred from 1998 to 2008 and compare it to our models.We then discuss possible applications of this modeling approach for taxa other than seabirds for which light is a significant threat, as well as its potential use for modeling other threats to ecological systems.

METHODS
We used ArcGIS 9.3.1 (ESRI, Redlands, CA) and digital layers of the island of Kauai to develop two models.One model represents the total intensity of artificial light a fledgling procellariid or hydrobatid could view along the topographically least-cost path from every location on the island to the coastline (the 'island' model) and the other represents these same paths to the coastline extended to 10 km offshore (the 'extended flight' model).This required the extension of a Digital Elevation Model (DEM) past the coastline (i.e., converting ocean to land), a layer of the total intensity of artificial light viewable from each pixel of this extended DEM (a viewable light intensity layer), and two models that summed all pixels of viewable light along topographically least-cost paths from the destination (the coastline for the 'island' model and 10 km past the coastline for the 'extended flight' model) back to each starting location (each DEM pixel).Fig. 1 outlines the steps involved in the development of the layers and models produced in this study.To provide repeatable steps for future use, the GIS-based processes are described in substantial detail in the Appendix.

Model assumptions
Both models required a set of assumptions concerning the flight behavior of fledgling procellariids and hydrobatids and if they can view light originating from the land once at sea.Our 'island' model included four assumptions.First, fledglings followed the path of least Fig. 1.Flowchart of GIS layers developed in this study and the steps involved in their production: (1) a Digital Elevation Model (DEM) of Kauai was extended 11 km beyond the coastline to produce an extended DEM, (2) all pixels of the extended Kauai DEM ,100 m in elevation were reclassified to 100 m so that fledgling seabirds flew straight paths to their destination once at or below 100 m, (3) the reclassified extended DEM was clipped to match the extent of the original Kauai DEM, (4) pixels of an artificial night light layer (from a satellite image) were converted to points raised 100 m above ground level to account for height of fledgling flight, (5) the extended Kauai DEM and light point layer were used to develop a layer of the total intensity of light that could be viewed from each extended DEM pixel (i.e., an extended viewable light layer that accounts for light that can be viewed offshore), ( 6) the extended viewable light layer was clipped to match the extent of the original Kauai DEM, (7) to develop the 'extended flight' model, both the reclassified extended Kauai DEM and extended viewable light layer were clipped to 10 km beyond the coastline and pixel values from the 10 km extended viewable light layer were summed along topographically least-cost paths from each fledgling destination (10 km beyond the coastline) back to all possible starting locations (every pixel of the 10 km reclassified extended DEM), and (8) to develop the 'island' model, pixel values from the viewable light layer were summed along topographically least-cost paths from each fledgling destination (the coastline) back to all possible starting locations (every reclassified DEM pixel).
topographical resistance (i.e., major drainages) from their natal sites to the ocean, as suggested for fledgling movements based on observed concentrations of fallout birds on Kauai (Telfer et al. 1987, Podolsky et al. 1998).Second, fledglings viewed light from 100 m above ground level, though this height differs from 100 m for sloped terrain because of the manner in which viewing light above ground level was modeled in this study.Third, once fledglings descended to, or were already at or below, 100 m in elevation they flew a straight path to the ocean, avoiding geographic features .100m in elevation along their path.This assumption allowed birds to discontinue following rivers once they descended to 100 m, which is likely more realistic based on personal observations of adults flying straight paths over lowland regions near the coast when returning to breeding sites.And finally, fledglings no longer viewed artificial light when they traveled beyond the coastline.Our 'extended flight' model included the first, second, and third assumptions from the 'island' model plus an additional one, that fledglings could view light emanating from the island until they traveled to 10 km beyond the coastline.

Viewable artificial light intensity layer
A layer of stable average artificial night light intensity from 2009 for the earth (developed from a satellite image with 911.25 3 911.25 m resolution; pixel values ranging from 0 to 63 relative units) was obtained from the National Geophysical Data Center (hhttp://www.ngdc.noaa.gov/dmsp/downloadV4composites.htmli) and the island of Kauai was clipped for further analysis (Fig. 2).Light originating from the sun, moon, aurora, and ephemeral sources (e.g., wildfires) were not included in this light intensity layer.All data included originated from sources of artificial light, including persistent sources such as gas flares, on cloud-free nights.Because these light intensities represent an average over an entire year, using this layer required the assumption that yearly average light intensities are a good approximation of light conditions on any night during the fledging period (e.g., October-November for Newell's Shearwater; Telfer et al. 1987).Light intensity pixels with values from 1-63 were then converted to points and the point layer was clipped so that all light points fell within the island boundary.
A previous study estimated that adult Newell's v www.esajournals.orgShearwaters on Kauai flew at a mean height of ;125 m (ranging from 8 to 750 m) above ground level with considerable variation among sites (Day and Cooper 1995).One limitation of the GIS methods we used is that, at each pixel on our extended DEM, fledglings remained at ground level; therefore, to account for height of both fledgling flight and light sources while developing this viewable light intensity layer, we raised the height of all light points to 100 m above ground level.For birds at ground level at the same elevation as light sources, raising the height of light sources above ground level allows birds to view light in an identical manner to raising birds the same height while keeping light sources at ground level; however, this relationship is not identical when birds and/or light sources are on sloped terrain.Therefore, this approach is considered an attempt to approximate the manner in which fledglings view light sources from realistic flight heights (Day and Cooper 1995) above the ground.We then generated a raster layer highlighting which light points were viewable from each pixel of the extended DEM of Kauai, correcting for the curvature of the earth.Because of software limitations, multiple output layers were produced; all output layers were summed to generate a layer of the total intensity of artificial light viewable in 3608 from each pixel of the extended DEM (i.e., a viewable light intensity layer; Fig. 3).

Input layers for models
As input for the final model step, we reclassified the extended DEM of Kauai so that no pixel values were ,100 m to achieve our assumption that fledglings fly a straight path to the ocean once they descend to, or are already at or below, 100 m in elevation.Our two models required clipping smaller layers from the reclassified extended DEM and the viewable light intensity layer.For the 'island' model, we clipped the extended DEM and light intensity layer to the size of the island of Kauai.For the 'extended flight' model, we clipped the extended DEM and light intensity layer to 10 km beyond the coastline of Kauai.

Development of final model layers
We used several tools from the TauDEM Version 5.0 toolset (Tarboton 2010) to develop the final layer for both models.We raised pits (low-elevation pixels that are completely surrounded by higher-elevation pixels and interfere with flow paths) in the DEMs for both models and created a flow direction raster from these pitremoved DEMs.We then used both the flow direction and viewable artificial light intensity layers to generate a layer of the intensity of all light viewable along the topographically leastcost path from every DEM pixel on the island either to the coastline (for the 'island' model; Fig. 4) or to 10 km past the coastline (for the 'extended flight' model; Fig. 5).To accomplish this, the software generated topographically least-cost paths from each DEM pixel to the destination (the coastline for the 'island' model and 10 km past the coastline for the 'extended flight' model) and summed pixels of viewable light along these paths from the destination back to all possible starting locations (each DEM pixel).This included light behind the flight direction of the bird; therefore, pixel values of this layer should be considered the total possible intensity of light that could be viewed along the path.In addition, the same individual lights are viewed in a series of consecutive pixels along a flight path, compounding their contribution to final cumulative light values.We consider this representative of the potential continual influence of particular light sources on the probability of attracting and disorienting fledglings.

Model limitations
Limitations of our modeling method include its lack of account for light attenuation (the decay in light intensity with increasing distance from the source), potential effects of wind speed and direction on fledgling movement patterns, effects of temporary weather conditions (e.g., local cloud cover and precipitation) on viewable light, or variation in the lunar cycle (which is known to affect fallout rates; Reed et al. 1985, Telfer et al. 1987, Ainley et al. 2001, Rodriguez and Rodri-Fig. 4. Map of Kauai (10 3 10 m resolution) illustrating the total intensity of artificial night light (for 2009), in relative units, that a fledgling procellariid or hydrobatid could potentially view if it followed the topographically least-cost path from any point on the island until descending to 100 m above sea level (asl), at which point it flew a straight path toward the ocean (avoiding obstacles .100m asl and only viewing light until it reached the coastline).
v www.esajournals.orgguez 2009, Miles et al. 2010).In addition, the light intensity pixels from the 2009 layer are 911.25 3 911.25 m in resolution, and each pixel was converted to a single point 100 m above ground level at the center of the original pixel; therefore, some variability in the horizontal and vertical distribution of lights was undoubtedly lost.In reality, individual lights in some locations could be distributed such that additional pockets of dark space may exist, through which some fledglings could fly to sea without viewing artificial light.Additionally, the 2009 night light layer contains artificial sky lighting (i.e., additional glow caused by refraction by water and dust molecules suspended in the air), which is most noticeable in proximity to cities (Elvidge et al. 2007), and this may inflate light intensity values to some extent in areas surrounding urban sites on Kauai.

Light intensity summary
From the 2009 night light intensity layer, we noted the single highest light intensity pixel value for Kauai and for the remainder of the Hawaiian archipelago for comparison.We also compared the proportion of two-dimensional surface area on Kauai covered by different categories of total light intensity calculated from the 2009 artificial light intensity layer, the 2009 viewable light intensity layer, the 'island' model, and the 'extended flight' model.We reclassified the pixel values for all four layers into groups as follows: 0, 1-250,000, 250,001-500,000, etc.; the last category contained values from 1,750,001 to 5,611,830.We then divided the number of pixels within pixel categories by the total pixels for the island, yielding the proportion of the island covered by the different light intensities.The category containing only values of zero allowed us to compare the proportion of area on Kauai covered by three types of dark area: area with no night light output (from 2009 artificial light intensity layer), area from which artificial light could not be viewed (from the viewable light intensity layer), and areas from which birds could fledge and not view artificial light along v www.esajournals.orgtheir paths (from both the 'island' and 'extended flight' models).

RESULTS
Light intensity values for Kauai from the original 2009 night light layer range from 0 to 55 (relative units; Fig. 2).The highest light intensity value for Kauai within a single 911.25 3 911.25 m pixel (¼ 55) approaches the highest value for the entire planet (¼ 63), which is recorded for many major metropolitan areas of the earth (e.g., Tokyo, Los Angeles), including Honolulu on the island of Oahu (the nearest main Hawaiian island southeast of Kauai ).Fig. 3 displays the intensity of light viewable from each location (i.e., DEM pixel) on Kauai, as well as over the ocean near the shoreline, and demonstrates the pervasiveness of artificial light in the interior of the island not illustrated by the original satellite image.Viewable light intensity values in Fig. 3 range from 0 to 5957 relative units; the highest value of viewable light intensity occurs on the island.
Summing viewable light values along topographically least-cost paths hypothesized for post-natal fledgling flights yielded cumulative viewable light values ranging from 0 to 1,575,440 relative units for the 'island' model (Fig. 4) and from 85,460 to 5,611,830 relative units for the 'extended flight' model (Fig. 5).The intensity of artificial light progressively increased, both in value and proportion of Kauai covered (Table 1), from the original 2009 night light layer to the 'extended flight' model (in which birds could view light along paths to 10 km past the coastline).Notably, with respect to fledgling seabirds, the 'island' model yielded only 3.2% of the island's total two-dimensional surface area ''unaffected'' by artificial light (i.e., that from which young birds could fledge and not view artificial light) and the 'extended flight' model yielded no ''unaffected'' portions of the island (Table 1).

DISCUSSION
Our model results highlight that fledging shearwaters, petrels, and storm-petrels on Kauai are likely exposed to artificial light beyond regions depicted by a satellite image alone.The practical value of this modeling effort lies in allowing managers to assess potential light exposure, and hence risk, to these fledglings from different colony sites on Kauai.The exact manner in which birds would respond to this light is unknown, however, as a threshold of intensity may be required for attraction and disorientation to occur.Given that the terrestrial activities of procellariids and hydrobatids (i.e., flights over land) are nocturnal and that the eye of the Manx Shearwater (P.puffinus; a close relative of the Newell's Shearwater) was shown to be adapted for nocturnal vision (Martin and Brooke 1991), faint light, as perceived by strictly diurnal animals, could be amplified when viewed by these fledglings, making even lowintensity light sources possible threats.
Our 'island' model, incorporating realistic assumptions concerning fledgling movement to Table 1.The proportion of two-dimensional surface area on Kauai covered by artificial light intensity categories from four layers: a 2009 night light intensity layer (obtained from a satellite image), a layer of light intensity viewable from each island location (developed using the 2009 night light layer), and our 'island' (representing the total possible light intensity fledglings could view along topographically least-cost paths to the coastline) and 'extended flight' models (representing the total possible light intensity fledglings could view along the same paths as in the 'island' model extended to 10 km offshore).
the shoreline, suggests that there are very few regions on Kauai from which young procellariids and hydrobatids could successfully fledge without potentially viewing artificial light along their paths to the ocean.Our 'extended flight' model, however, which allows birds to view light offshore, suggests that there are no such areas on the island.The large increase in minimum and maximum cumulative viewable light intensity yielded by the 'extended flight' model, compared to the 'island' model, arose from the expanded visual field birds experienced once over the ocean.This allowed fledglings to view distant lights along the coastline not previously visible when traveling along their terrestrial paths.Though it is not known how these birds respond to viewing lights on the land once they are at sea, previous authors suggested that birds can be attracted by light back to the shore (Podolsky et al. 1998), making a model with this assumption likely more appropriate for estimating the cumulative threat posed by artificial light.Notably, the north shore region of Kauai  Kauai (Griesemer and Holmes, in press), and given that fallout can hinder population growth (Ainley et al. 2001), minimization of light attraction will play a crucial role in the recovery of this species, particularly on the north shore.In addition, though lesser numbers of Hawaiian Petrels and Band-rumped Storm-Petrels are found annually, any reduction in successful fledging of these long-lived seabirds with low fecundity could have significant long-term impacts on their populations, especially in concert with the continuous threat of predation by introduced mammals.
Researchers and managers may use the layers stemming from our models as a first step in assessing the level of impact of artificial light on individual breeding colonies of the Newell's Shearwater on Kauai.Examining light pixel values from original satellite images (e.g., Fig. 2) within areas will be valuable for managers because it provides an assessment of risk to birds at sites of light sources.By comparison, modeling viewable light and fledgling movement as we have done in this study allows managers to assess the threat that light poses to birds originating from known breeding colonies, as well as regions where breeding activity is suspected.Summing light pixel values from original satellite images within individual watersheds can also provide a simple and rapid assessment of potential risk to fledglings that does not require complex modeling.Summing light by watershed, however, does not account for light that can be viewed originating from neighboring watersheds, and thereby may depict some watersheds as ''dark'' when they may contain fledglings that could view light along their flight paths.Risk assessment at the colony scale will become increasingly important for managers of threatened and endangered burrowing shearwaters and petrels on Kauai and other islands when determining where to allocate limited resources to protect birds.
The modeling method described here can be applied to related taxa in other locations.The 2009 night light layer we used is available for the entire planet from the National Geophysical Data Center, providing current artificial light data for other locations, particularly islands, where the threat of light attraction and fallout is significant.As light layers from this source are available for a period spanning 18 years (1992)(1993)(1994)(1995)(1996)(1997)(1998)(1999)(2000)(2001)(2002)(2003)(2004)(2005)(2006)(2007)(2008)(2009), light conditions could be modeled as well to investigate changes during the past two decades.In addition, our assumptions could be modified to incorporate behavioral information specific to other seabird species (i.e., varying the height of fledgling flight and the elevation at which fledglings fly straight paths to the ocean).Light intensities could also be altered on base satellite images to examine how proposed artificial light reduction goals or projections of future night light output might affect particular regions and breeding sites after accounting for seabird movement.Light attenuation could be accounted for as well through use of an attenuation formula, a map of two-dimensional distance from each light source (used as a proxy for actual three-dimensional distance from light sources), and the 'Raster Calculator' to adjust light intensities as they decay with distance.However, this may require each light point, which could range from hundreds to thousands, to be modeled individually.
Methods presented herein could also be used to estimate the threat of light altering the behaviors of non-seabird taxa, particularly migratory passerines and hatchling marine turtles, traveling through or breeding in human-populated regions (e.g., Cochran and Graber 1958, Witherington 1992, Salmon et al. 1995, Jones and Francis 2003, Bird et al. 2004, Tuxbury and Salmon 2005, Baker and Richardson 2006, Stone et al. 2009, Keenan et al. 2007, Kempenaers et al. 2010).Maps of viewable light intensity would provide such an estimate in some cases and could be developed using light layers at resolutions finer than the layer used to develop our models (i.e., those displaying individual light sources).Furthermore, elements of the technique described here (or alterations thereof ) could be used to spatially model other point-source threats to ecological systems, such as sound and air pollution.Movements that differ from the downhill, least-cost paths employed in our study (e.g., long-distance migration routes of passerines) could be modeled as well, but this would require different methods for the final step of model development than those presented in this paper.Spatially assessing threats at large scales may require developing multiple layers due to computer and software limitations.

Summary of GIS methods
We used ArcGIS 9.3.1 (ESRI, Redlands, CA) and digital layers of the island of Kauai to develop two models.One model represents the total intensity of artificial light a fledgling procellariid or hydrobatid could view along the topographically least-cost path from every location on the island to the coastline (the 'island' model) and the other represents the same paths to the coastline extended to 10 km offshore (the 'extended flight' model).The tool used for the final step of model development requires that its input layers have completely overlapping extents; however, preliminary development of one of the model input layers resulted in the two input layers for the final model step having slightly non-overlapping extents.To correct for this, we first extended a Digital Elevation Model (DEM) of the island of Kauai 11 km past the coastline, developed a layer of artificial light intensity viewable in 3608 from each pixel of this extended DEM, and clipped smaller layers for the 'island' and 'extended flight' models from the larger 11 km extended layers, snapping each clipped viewable light intensity layer to its companion clipped DEM to achieve complete overlap.Below, we describe the GIS-based model-building process in substantial detail to provide repeatable steps for future use.

Extended DEM
We extended a DEM of Kauai (10 3 10 m resolution) from the Hawaii Coastal Geology Group (hwww.soest.hawaii.edu/coastsi)11 km from the coastline (i.e., which converted ocean 11 km offshore to land).First, we reclassified the DEM so that all numeric pixel values were converted to values ¼ 0 and all 'NoData' pixels remained 'NoData' pixels.This reclassified DEM was converted to a polygon using the 'Raster to Polygon' Tool.A buffer of 11,000 m surrounding the island polygon was then created using the 'Buffer' Tool and converted to a raster layer (10 3 10 m resolution) with pixel values ¼ 0 using the 'Feature to Raster' Tool.We combined the original DEM and new raster layer using ArcCatalog to create an unmanaged raster catalog containing both layers; the 'Raster Catalog to Raster Dataset' Tool was then used to combine these layers, extending the coastline of the Kauai DEM by 11 km.

Viewable artificial light intensity layer
A layer of stable average artificial night light intensity from 2009 for the earth (obtained from a satellite image with 911.25 3 911.25 m resolution; pixel values ranging from 0 to 63 relative units) was obtained from the National Geophysical Data Center (hhttp://www.ngdc.noaa.gov/dmsp/downloadV4composites.htmli).We projected the layer in North American Datum (NAD) 1983 Zone 4 using the 'Project Raster' Tool and clipped out the island of Kauai using the 'Clip' Tool and a shapefile of the boundary of Kauai obtained from the USGS Hawaii Data Clearinghouse website (hhttp://hawaii.wr.usgs.gov/i)(Fig. 2).This clipped layer was reclassified so that pixel values ¼ 0 were converted to 'NoData' and all other pixels were equal to their original light intensities.The light intensity pixels with values from 1 to 63 were then converted to points using the 'Raster to Point' Tool and this point layer was clipped using the Kauai boundary shapefile so that all light points fell within the island boundary.To account for the height of both fledgling flight and light sources while developing this viewable light intensity layer, we raised the height of all light points to 100 m above ground level by creating an 'OFFSETA' field within the attributes table of the light point layer and setting values ¼ 100 using the 'Field Calculator'.
We then used the 'Observer Points' Tool and the 11 km extended Kauai DEM, correcting for the curvature of the earth (with a refractory coefficient ¼ 0.13), to generate a raster layer highlighting which light points were viewable from each extended DEM pixel.Because of the inability of the 'Observer Points' Tool to handle a large number of points, we segmented the light

Fig. 2 .
Fig. 2. Map of Kauai illustrating the stable average artificial night light intensity, in relative units, for 2009.Light intensity pixels (911.25 3 911.25 m in resolution) are not in complete overlap with the island boundary (outlined in black).

Fig. 3 .
Fig.3.Map of Kauai (10 3 10 m resolution) and nearby ocean illustrating the total intensity of artificial light (for 2009), in relative units, viewable in 3608 from each island and ocean pixel (i.e., from each extended DEM pixel).Light viewable from the ocean is included to show light intensity near the coastline that fledgling procellariids and hydrobatids could potentially view past the island boundary (outlined in white).The highest viewable light intensity value occurs on the island.

Fig. 5 .
Fig. 5. Map of Kauai (10 3 10 m resolution) illustrating the total intensity of artificial night light (for 2009), in relative units, that a fledgling procellariid or hydrobatid could potentially view if it followed the topographically least-cost path from any point on the island until descending to 100 m above sea level (asl), at which point it flew a straight path to 10 km beyond the coastline (avoiding obstacles .100m asl and only viewing light until it traveled 10 km beyond the coastline).
contains the highest number of Newell's Shearwater fallout records summed from 1998-2008 (Fig.6), a time during which the approximate geographical distribution of artificial light, relative to other portions of the island, did not change compared to 2009.This region, however, did not yield the highest light intensity values from the 2009 night light satellite layer, the viewable light layer, or the 'island' and 'extended flight' models.Most remaining Newell's breeding sites are known from the northwestern portion of the island (Hawaii Division of Forestry and Wildlife, unpublished data).Thus, the high volume of fallout on the north shore is expected to be a function of high fledgling productivity within northwestern watersheds and, therefore, a large number of fledglings from northwestern colonies being exposed to artificial light originating from the north shore (i.e., north shore lights viewed from the ground or air surrounding northwestern mountain peaks or once offshore north ofKauai [Fig.3]).The Newell's Shearwater is currently exhibiting a population decline on

Fig. 6 .
Fig. 6.Newell's Shearwater fledgling fallout summed by sector on Kauai from 1998-2008 (Hawaii Division of Forestry and Wildlife, unpublished data).Fledglings without a sector location identified are not included.The number of fledglings is labeled in sectors with .100known fledgling recoveries.All fallout data were obtained from Hawaii Division of Forestry and Wildlife.Data were collected by Hawaii Division of Forestry and Wildlife from 1998-2005, by Kauai Island Utility Cooperative from 2006-2007, and by Kauai Humane Society in 2008.