Quantifying wedge-tailed shearwater (Ardenna pacifica) fallout after changes in highway lighting on Southeast O'ahu, Hawai'i.

Attraction to artificial light at night (ALAN) poses a threat to many fledgling seabirds leaving their nests for the first time. In Hawai'i, fledgling wedge-tailed shearwaters disoriented by lights may become grounded due to exhaustion or collision, exposing them to additional threats from road traffic and predation. While the timing and magnitude of shearwater fallout varies from year to year, little is known about how changing lighting and environmental conditions influence the risk of grounding for this species. We analyzed 8 years (2012-2019) of observations of road-killed shearwaters along the Kalaniana'ole Highway on O'ahu to quantify the timing and magnitude of fallout during the fledging season (November-December). Our goal was to compare fallout before (2012-15) and after (2016-19) a transition in highway lighting from unshielded high-pressure sodium (HPS) to full-cutoff light-emitting diode (LED) streetlights. To detect the shearwater response to the lighting regime, we also accounted for three potential environmental drivers of interannual variability in fallout: moon illumination, wind speed, and wind direction. The effects of these environmental drivers varied across years, with moon illumination, wind speed and wind direction significantly affecting fallout in at least one year. Altogether, the interaction between moon illumination and wind speed was the most important predictor, suggesting that fallout increases during nights with low moon and strong winds. The lack of an increase in fallout after the change from HPS to shielded 3000K - 4000K LED streetlights suggests the new streetlights did not worsen the light pollution impacts on wedge-tailed shearwaters on Southeast O'ahu. However, due to potential species-specific disparities in the behavior and light attraction of petrels, similar studies are needed before energy saving LED lights are implemented throughout the Hawaiian archipelago.

Introduction

Light pollution is a concern for burrow-nesting seabirds globally, with documented impacts on over 50 species of shearwaters, petrels, and puffins [1]. While coastal light pollution can disrupt adult seabirds provisioning their chicks on colonies [2-4], fledglings consistently account for the majority (68% - 99%) of the grounded specimens [1]. Fledgling seabird “fallout” occurs when chicks leaving their nests are disoriented by onshore lighting and become stranded on land instead of flying out to sea [1]. The magnitude of fallout is likely influenced by the number of chicks fledging, the prevailing environmental and celestial conditions [4, 5], and the features of anthropogenic lights, which vary as a function of light fixture design and bulb type [6, 7]. To gauge the effectiveness of light pollution mitigation measures, wildlife managers need to understand the influence of these biological and environmental drivers on the timing and magnitude of fallout.

A conceptual model to explain fallout involves fledging seabirds being drawn toward well-lit coastal areas, especially in the absence of moonlight [4, 5, 8–11] and when strong winds are directed toward shore [10, 11]. Birds are affected by bright light sources from vessels at sea and urbanized areas on shore, including streetlights and sports fields [1, 6, 12–13]. Moreover, collisions with powerlines and other structure can lead to injury and grounding [14, 15]. While our understanding of the environmental drivers of fallout is growing, the influence of specific design features of anthropogenic light sources remains understudied. In particular, lamp color and directionality are two key streetlight features that can affect fallout [7, 16]. Spurred by efforts to improve energetic efficiency, many cities are replacing yellow high-pressure sodium (HPS) lightbulbs commonly used in streetlights with white light-emitting diode (LED) bulbs [17-19]. Although LED bulbs decrease electricity consumption and maintenance costs, these benefits could be costly to wildlife, as shearwaters may be more sensitive to LED lights [7]. A study on the visual perception of Wedge-tailed Shearwaters (Ardenna pacifica–previously Puffinus pacificus) showed that they experience maximum light absorption of the wavelengths emitted by white LED lights (406–566 nm) and have lower absorption of the wavelengths emitted by HPS lights (560–620 nm) [20]. Moreover, a field-based experiment in Australia showed that Short-tailed Shearwaters (Ardenna tenuirostris) show increased attraction to LED lights over HPS lights, although the difference was not statistically significant [7].

Mitigation measures often target light directionality, whereby streetlights are shielded through the use of a “full-cutoff” design, which inhibits light emission above the horizontal plane of the fixture. This approach, when applied to HPS lights, reduced Newell’s Shearwater (Puffinus newelli) fallout on Kauai (Hawaiʻi) [16]. Although mitigation is being addressed through shielding, the common use of optimized LEDs with broad spectra and Correlated Color Temperature (CCT) greater than the maximum recommended value for wildlife (2200 K) may be a cause for concern [17]. While modern LED lights possess the flexibility to give off a range of low to high CTT, short-wavelength light with high CCT is a common choice because of its efficiency [19]. The effectiveness of light shielding coupled with the use of broad spectrum, high CCT LEDs is unknown.

On the island of Oʻahu (Hawaiʻi), Wedge-tailed Shearwaters (hereafter referred to as WTSH) experience fallout during the annual fledging season (November-December) [21, 22]. A three-year study in the early 1990s, revealed that hundreds of chicks become grounded every autumn, with the number varying widely from year to year [22]. Starting in 2002, U.S. Fish and Wildlife Service initiated a program of opportunistic road surveys of the southeast section of Oʻahu during the fledging season, which documented a fallout hotspot in the town of Waimānalo, within 5 km from two WTSH colonies on offshore islets [21].

While there is evidence of interannual variability in WTSH fallout, little is known about the influence of environmental (weather and oceanographic conditions) and biological (breeding population size and reproductive success) drivers. To date, only one study has investigated the environmental drivers of WTSH fallout, by comparing a “wreck” year of unusually high fallout (1994), when WTSH groundings increased ten-fold from the two “normal” years prior [22]. This study suggested that anomalous southerly winds likely carried fledglings inland rather than out to sea and scattered them throughout the windward coast of Oʻahu. While the southerly winds help explain why many birds were found inland, it is unclear to what extent low ocean productivity during the breeding season and unusual weather conditions during the fledging period caused the high fallout observed that year.

Over a decade later, Friswold et al. (2020) documented an increasing trend in annual fallout numbers between 2003 and 2010, and a two-year cycle of alternating years of high and low fallout. Subsequently, an unusually large fallout event in 2011 was documented during a La Niña year with high ocean productivity [23]. These results are suggestive of the potential influence of breeding population size and reproductive success on fallout.

In 2012, we began conducting systematic road surveys along a 17.3-km section of the Kalanianaʻole Highway to document WTSH fallout. In 2016, the Hawaiʻi Department of Transportation changed the streetlights on Oʻahu’s major roads from unshielded 2200 K HPS lights to shielded 3000–4000 K LED lights, where Kelvin (K) is a unit of measurement for CCT. Lower CCT indicates a warm yellow-orange appearance whereas higher CCT indicates cool blue light [18]. The shift in lights halfway through our study provided a unique opportunity to compare WTSH fallout under different street lighting conditions. To this end, we continued conducting standardized surveys following the established protocol through 2019 and analyzed an 8-year time series with four years before (2012–15) and four years after (2016–19) the change in lighting. This is the first study to compare changes in seabird groundings in response to HPS versus LED streetlights, by repeatedly surveying a fallout hotspot during the fledging season.

The goal of this study is to quantify the magnitude of WTSH fallout under two contrasting lighting regimes, to inform future coastal development and management of light pollution. Although shielding of the LED streetlights may reduce initial WTSH attraction, we predicted that disorientation caused by high intensity/shorter wavelength lights would outweigh the benefits of shielding. Thus, we expected an increase in fallout after the installation of LED streetlights (2016–2019). To detect the fallout response to the lighting regime, we also accounted for three potential environmental drivers: moon illumination, wind speed, and wind direction. Because WTSH rely on wind to take flight and may become disoriented in the absence of moonlight, we predicted higher fallout during windy nights of low moon illumination. In particular, due to the location of our study area, southwest from two breeding colonies, we anticipated that strong northeasterly winds would drive the fledging birds towards shore.

Methods

Study area

This study focuses on the southeast section of Oʻahu, where a two-lane coastal highway runs through a rural and developed landscape (Fig 1). The survey route was illuminated with HPS streetlights until 2016, when the Hawaiʻi Department of Transportation transitioned to LED streetlights. The CCT of the LED streetlights is 3000 K on sections of the highway directly adjacent to the ocean, whereas inland lights are 4000 K.

Fig 1

Map of the study area in southeast Oʻahu.

Blue line shows the survey route, and crosses indicate the start and end points. Black dots indicate WTSH breeding colonies. (PI = Popoiʻa Island, MOK = Mokulua Islands, MI = Mānana Island, KI = Kāohikaipu Island, FSP = Freeman Seabird Preserve). Star marks the location of the Sea Life Park seabird rehabilitation center. Inset maps show the island of Oʻahu, and the main Hawaiian Islands. Map features are overlaid on an ArcGIS Pro Software Version 2.5 base layer [24].

The WTSH breeding colonies of Mānana Island and Kāohikaipu Island, where approximately 25,000 and 800 chicks were counted in 2019, are located 1.3 and 0.7 km offshore of our study area, respectively [25]. Three additional WTSH colonies on offshore islets (Mokulua Nui, Mokulua Iki, and Popoiʻa) lie approximately 6 km north of the study area (Fig 1), with 2019 chick count estimates of 3,500, 5,000, and 900, respectively [25].

Weather patterns on windward Oʻahu are dominated by the northeast trade winds, which typically persist for 1 to 2 weeks, interspersed with no-wind periods or southerly storms. Peak wind speeds occur in the afternoon, with lower wind speeds at night [26].

Intake records

Members of the public deliver grounded WTSH to Sea Life Park (SLP), a marine life center located along our survey route in Waimānalo, for rescue and rehabilitation. SLP intake records, involving the daily number of rescued WTSH chicks, have been used to document the island-wide temporal variability in WTSH fallout during the fledging season and from year to year [21, 22]. To provide a broader context for our localized surveys of a known WTSH fallout hotspot, we compared the timing and the magnitude of annual fallout documented in the SLP intake records and our surveys.

Road surveys

We used a time series of standardized road surveys along a 17.3-km stretch of the Kalanianaʻole Highway, starting at the Olomana Golf Club, running through Waimānalo, and ending at the Koko Marina Center (Fig 1). While this survey route is a subset of the area surveyed by USFWS from 2002–2010, it encompasses the main WTSH fallout hotspot in Waimānalo [21]. We conducted morning surveys by car, every 3 days, throughout the WTSH fledging season (November 6 –December 21). We began surveys at sunrise (6:15–7:15 AM) and drove the route once in each direction, at speeds between 25–35 mph, while visually searching for dead birds in each lane and along the shoulder. Since these surveys were conducted in the morning, likely a full 12 hours after fledging time, almost all the birds we observed were deceased. In 8 years of surveys, we observed 2 live birds, which were brought to SLP for rehabilitation and not counted in our analysis. All dead birds sighted while driving were included in the surveys, even if they were found on the shoulder, the median, or off the road.

Upon encountering a carcass, we recorded its position on the road, location (latitude and longitude coordinates from a hand-held Garmin e-trex GPS unit), nearest street address, and nearest utility pole using their unique id tags. We also took photos of each WTSH we encountered, showing diagnostic identification features (head and feet).

Environmental variables

We related WTSH fallout to two publicly-available environmental datasets: (i) wind speed (knots) and wind direction (degrees) recorded on Moku Loʻe (Kaneohe Bay) and provided by PacIOOS [27], and (ii) the lunar cycle, quantified using the percent of the lunar disk that was illuminated each night, from the U.S. Naval Observatory [28].

Because WTSH fledge during the night, we averaged the hourly wind data every night (18:00–6:00 local time). To match our surveys to the preceding environmental conditions, we related the number of grounded WTSH documented during a given road survey to the average wind speed (knots), wind direction (degrees), and lunar disk illumination (%) from the three nights prior.

Data analysis

We analyzed fallout across and within years using generalized linear models (GLM) built with R version 3.5.1 and the stats and MASS packages [29]. We developed and fitted nine separate models: a full model (involving all study years) quantified interannual variability, and eight yearly models visualized the interannual differences documented by the full model.

We ran all models using both Poisson and negative binomial distributions. Because the Poisson assumes that the variance equals the mean, the negative binomial is more appropriate whenever there is overdispersion [30]. We used the Akaike Information Criterion corrected for small sample size (AICc) to select the best-fitting distribution for each model [31].

Multi-year model of WTSH fallout

We related the number of WTSH observed during 128 surveys of the entire study area (16 per year times 8 years) to the light regime (unshielded HPS / shielded LED), year (2012–2019), and four environmental variables: moon illumination (% lunar disk illuminated), average wind speed (knots), average wind direction (degrees), and Julian date (the number of days since the beginning of the year).

We used multi-model inference to test all possible combinations of these six explanatory variables. Whenever two of three potentially interacting variables were included in a model, we also considered their interaction (‘moon*date’, ‘moon*wind speed’, and ‘wind speed*date’). We used the AICmodavg package [29] to assess the model fit using AICc, which prevents over-fitting by penalizing models for each additional variable [31]. AICc assigns a value to each model using the formula, AICc = -2log(L)+2K+(2K(K+1)/(n-K-1)), where K is the number of parameters, n is the sample size, and L is the maximum likelihood of obtaining the given results with K parameters. We used Akaike weights (wi) to calculate the likelihood of each model as follows: where the numerator is the model likelihood with Δi showing the change from the lowest AICc model to the given model, and the denominator is the sum of all relative weights, as determined by Δr, the change in each contending model from the lowest AICc model. The lowest AICc value indicates the model that best describes the patterns in the observed data without over-fitting [31, 32].

To test the influence of the streetlights, in the context of interannual variability, we built two complementary sets of full models that either included “light regime” (comparing two groups of years: 2012–15 vs 2016–19) or included individual “years”, regardless of their “light regime”. This resulted in a total of 106 models: 36 included “light regimes”, 36 included “years”, and 34 included neither. Individual models ranged from having one to eight predictors (five variables and three interactions) (S1 Table). Following Michael et al. (2014), we assessed the importance of each variable in terms of their scaled average weight, calculated using the models where those variables were included.

Yearly models of WTSH fallout

We related the number of WTSH observed during 16 surveys of the entire study area (every three days during a single year) to the four aforementioned environmental variables: moon illumination, average wind speed, average wind direction, and Julian date (S1 File). We did not consider variable interactions, and calculated pseudo R-squared values based on the standard errors using the ‘rsq’ package [29].

Results

Fallout records

To interpret our road surveys in a broader context, we compared the number of grounded WTSH we documented along the SE corner of Oʻahu with the SLP intake records, which provided an island-wide measure of fallout timing and magnitude. The SLP intake records of fledging chicks spanned from November 2 to January 5, and our observations of grounded shearwaters along the Kalanianaʻole Highway spanned from November 6 to December 21. Overall, only 2.3% of the SLP intake records fell outside of our road survey period (November 6 –December 21), with yearly proportions ranging from 1.3% to 6.3% (S1 Table).

The total number of rescued WTSH brought into SLP yearly across the 8-year study varied by nearly an order of magnitude, ranging from 74 to 525 birds per year, with an average of 226.1 +/- 170.6 S.D. (median = 159.5) (S1 Table). The number of WTSH carcasses observed on the survey route per year also varied widely, ranging from 7 to 60 birds, with an average of 24.1 +/- 18.7 S.D. (median = 17.5) (S1 Table). There was a positive correlation between the yearly number of road-killed birds (our surveys) and rescued birds (SLP records), with 2012 and 2016 standing out as high-fallout years (r2 = 0.85, df = 6, p < 0.01) (Fig 2). There were 469 rescued birds in 2012 and 525 in 2016, with both years exceeding the median by over 300 birds. Likewise, there were 60 road-killed birds in 2012 and 45 in 2016, compared to the median of 17.5 birds. The lowest numbers of rescued and road-killed birds occurred in 2018, with 74 and 7 birds respectively.

Fig 2

Fallout comparison between Sea Life Park intake records and road surveys.

Scatterplot showing the total rescued WTSH per year from Sea Life Park intake records versus the total WTSH carcasses documented per year during road surveys (r2 = 0.85).

Fallout modeling

Over the 8-year study, the number of grounded WTSH observed per survey ranged from 0 to 10, with an overall average of 1.5 +/- 2.2 S.D. (median = 1) (Fig 3). Moreover, to account for the large proportion (46%) of absences (0 WTSH detected during a road survey), we fitted the fallout count data to Poisson (1< VMR < 2) and negative binomial (VMR > 2) distributions. We developed and fitted a full model and eight single-year models.

Fig 3

Boxplots of WTSH carcasses observed during road surveys.

Distribution (5, 25, 50, 75, 95 percentiles) of the number of grounded WTSH observed each study year (n = 16 yearly surveys). Dots indicate outliers.

Because 9 different model formulations were required to achieve an AICc weight of 0.90, this model set was used to ascertain the importance of the driver variables. Of the 9 variables tested, only the interaction between moon illumination and wind speed (moon*wind speed) achieved a scaled average weight >1 and was thus deemed an “important” variable (S2 Table). Moon, wind speed, and year all had weights of 1, because they contributed an average amount to each model’s weight. Date, wind direction, (moon*date), and (wind speed*date) had weights < 1, and contributed less than the average variable to each model’s weight. Light regime had a weight of 0, and did not appear in any of the models required to achieve the AICc weight of 0.90.

The overall best-fitting model had a weight of 0.37 and included four explanatory variables: moon, wind speed, year, and the interaction of moon and wind speed (moon*wind speed) (Table 1). All variables in this model were significant, except wind speed and year 2016 (not significantly different from 2012). The negative coefficient for the moon variable in this model (-2.9) indicates that, across the 8-year period, fewer birds were grounded when a greater percentage of the lunar disk was illuminated. All years except for 2016 were significantly different from the reference year (2012). The interaction between wind speed and moon had a positive coefficient (+0.23) suggesting that fallout was higher during periods of lower lunar illumination and higher wind speed (Fig 4).

Table 1

Full model output.

Explanatory variable	Estimate	S.E.	Z-value	p-value
Intercept a	2.136	0.580	3.709	<0.001
Wind Speed	-0.027	0.046	-0.576	0.565
Moon	-2.884	0.827	-3.485	<0.001
Year2013	-1.711	0.447	-3.825	<0.001
Year2014	-1.685	0.406	-4.148	<0.001
Year2015	-1.490	0.365	-4.082	<0.001
Year2016	-0.373	0.324	-1.148	0.251
Year2017	-1.661	0.391	-4.247	<0.001
Year2018	-2.492	0.483	-5.153	<0.001
Year2019	-1.034	0.366	-2.824	0.005
Wind Speed*Moon	0.225	0.083	2.710	0.007

GLM results from best-fit full model, following a negative binomial distribution. Bold font denotes significance at alpha < 0.05.

a Reference year (intercept) is 2012.

Fig 4

Fallout as a function of moon illumination and wind speed.

Scatterplot of the number of grounded WTSH observed per survey, in relation to wind speed, and moon illumination. Open circles indicate the presence of fallout, with the increasing radius ranging from 1 to 10. Small solid dots indicate the absence of fallout (0 birds).

In addition to the interannual variability in the number of WTSH observed during road surveys (Fig 3), the temporal aggregation of fallout across surveys was also highly variable, as evidenced by the varying dispersion (variance to mean ratio, VMR) observed yearly (1.02–4.03) (Table 2)). Moreover, due to the large proportion (46%) of absences (0 WTSH detected during a road survey), fallout counts followed a Poisson distribution (1< VMR < 2) in every year, except for 2019 (VMR = 4.03), when the negative binomial model yielded a lower AICc value.

Table 2

Yearly model output.

Year	Dis.	VMR	Pseudo Adj. R2	Estimate					p-value
				Int.	WS	WD	Moon	Date	Int.	WS	WD	Moon	Date
2012	P	2.11	0.41	-29.293	-0.744	-0.032	-0.065	0.137	0.096	0.236	0.408	0.071	0.054
2013	P	1.50	0.51	-28.120	0.684	0.029	-4.944	0.065	0.146	0.028	0.016	0.006	0.211
2014	P	1.02	0.64	3.055	0.393	0.003	-1.659	-0.020	0.751	0.045	0.782	0.144	0.456
2015	P	1.14	0.33	-3.680	0.03	-0.001	1.993	0.008	0.575	0.818	0.902	0.006	0.720
2016	P	3.89	0.50	-0.169	0.268	0.002	0.439	-0.006	0.979	0.034	0.816	0.507	0.777
2017	P	1.20	0.18	8.690	0.099	-0.002	-0.854	-0.027	0.291	0.273	0.638	0.291	0.309
2018	P	1.20	0.57	-20.160	-0.190	-0.001	-1.592	0.064	0.183	0.546	0.959	0.261	0.178
2019	NB	4.03	0.26	42.940	-0.100	-0.022	-3.305	-0.115	0.056	0.543	0.196	0.004	0.055

GLM output of annual fallout models, based on 16 surveys (Nov. 6 –Dec. 21) and clumped data distributions (P = Poisson, NB = negative binomial), as evidenced by the variance to mean ratio (VMR). In addition to the intercept (Int.), four explanatory variables were considered: wind speed (WS), wind direction (WD), lunar illumination (Moon), and Julian Date (Date).

a The bold font denotes significance at alpha < 0.05. and the pseudo adjusted R-squared quantifies the model fit.

Overall, the yearly models explained a wide range of the variation in fallout throughout the fledging season, with their pseudo R2 values ranging from 18% (2017) to 64% (2014). Moreover, different variables were significant in different years (Table 2). Surprisingly, the influence of moon illumination was not consistent across our study, with a significant effect in three years: it was negative twice (2013 and 2019), and it was positive once (2015). Wind speed had a significant positive effect in three years (2013, 2014, and 2016), whereby higher wind speeds led to more fallout. Wind direction had a significant positive effect once (2013), whereby wind blowing from the southwest led to more fallout. Julian date was never significant, suggesting that fallout was variable throughout the survey period (November 6 –December 21).

Overall, while fallout was explained well (pseudo R2 ≥ 0.5) by wind speed alone in 2014 and 2016, it was explained moderately well (pseudo R2 ≥ 0.3) by moon illumination alone in 2015 and 2019. In 2013, about half of the fallout variation was explained by a combination of wind speed, wind direction, and moon illumination. In three years (2012, 2017, and 2018), fallout was not significantly explained by any of the predictors.

Two years (2012 and 2016) showed significantly higher fallout compared to the other study years (Fig 3) and together accounted for 55% of the WTSH found during road surveys. Those same years were also responsible for 55% of all rescued birds brought to SLP, within the timeframe of this study. While none of the predictor variables were statistically significant in 2012, moon and date were marginally significant (0.10 < p < 0.05) (Table 2). The highest yearly fallout occurred in 2012, when 60 WTSH were grounded during an early new moon period (Julian days: 317–326, November 12–21), and a later one (Julian days: 344–353, December 11–18) (Fig 5), both of which were accompanied by strong winds (Fig 6). In 2016, a new moon period occurred in the middle of the fledging season, leading to a single peak in fallout (Fig 5), which coincided with a period of high wind speeds (> 12 knots), increasing the number of birds grounding at this time (Fig 6).

Fig 5

Moon illumination and fallout.

Time series of the number of grounded WTSH observed per survey (bars) and nightly lunar illumination (back line). The dates on the x-axis indicate the survey days.

Fig 6

Wind speed and fallout.

Time series of the number of grounded WTSH observed per survey (bars) and nightly wind speed (back line). The dates on the x-axis indicate the survey days.

Discussion

Timing and magnitude of WTSH fallout

The strong positive correlation between the yearly numbers of grounded WTSH found during our road surveys and rescued WTSH brought to SLP suggests that our small-scale surveys of a fallout hotspot are indicative of island-wide fallout trends on Oʻahu. Both the rescue records and the road surveys documented the highest fallout in 2012 and 2016, and the lowest fallout in 2018. Moreover, only 2.3% of the WTSH brought to Sea Life Park during the fledging season between 2012–2019 fell outside of our study period (November 6 –December 21), suggesting that our survey window captures most of the fledging season fallout.

Interpretation of model results

Our hypothesis that the LED streetlights would increase shearwater groundings due to higher sensitivity to shorter wavelengths was not supported, as the light regime was not selected as a significant predictor variable in any of the top models. It is possible that shearwater visual perception of LED lights was in fact greater, but shielding reduced initial attraction, thus balancing out overall fallout. However, even if this were the case, our analysis could not distinguish between these two factors, because the changes in bulb type and shielding were not independent. Nonetheless, this finding has useful implications for resource managers since LED lights are a common replacement for HPS lights in Hawaiʻi and elsewhere. While we encourage managers to seek lighting adjustments that will mitigate fallout, our study shows that the change in streetlights from unshielded HPS to shielded 3000 K– 4000 K LED did not exacerbate this problem for WTSH on Oʻahu.

It is possible that, even if there was an effect of light type on WTSH fallout, its influence was marginal compared to the effect of the other environmental drivers. In particular, the higher than average variable weight of the interaction between moon and wind speed suggests that fallout is a dynamic process, driven by the synergy of low moon illumination and strong winds, more so than by moon or wind alone (Fig 4, S2 Table). While previous studies have identified the importance of moon and wind, this is the first time their interaction has been considered.

This significant interaction underscores a conceptual model, whereby wind speed determines the magnitude of fledging birds departing their colonies, and the moon illumination determines the attraction of those fledglings toward onshore lighting. This conceptual model can explain why years like 2016, when the peak of the fledging season coincided with a new moon and high wind speeds, have greater fallout.

While the interaction of the lunar illumination and the wind speed was critical, the timing of these variables, captured using their interactions with date, were less important. Together, these results suggest that, within the time frame of our study, fallout is most dependent on the temporal overlap of low moon illumination and high winds, rather than on their specific timing.

The influence of the four predictor variables (wind speed, wind direction, moon, and Julian date) were not consistent across all 8 study years. While most previous studies have found strong negative relationships between moon illumination and fallout, our yearly models only documented this pattern in 2 years [4, 8, 10, 11, 33]. One possible explanation for this result could be a mismatch between moonrise / moonset times and WTSH fledging. We used an average lunar disk illumination for the three nights prior to each survey, assuming that this would be representative of visible moonlight while WTSH were fledging. However, if most birds fledge shortly after sunset, before the moon rises during waning moon phases, conditions will resemble a new moon [16, 34]. Similarly, if island topography or clouds obscure a rising moon from a given natal colony, the navigational benefits provided by the lunar disk could be compromised until the moon rises over obscuring landscape features. The peak fledging times of WTSH are unknown but could be useful to improve our understanding of the influence of moonrise / moonset times on fallout. Furthermore, the lack of strong lunar trends in the yearly models could be due to small sample sizes, with each year only involving 16 surveys.

Contrary to Rodríguez et al. (2014), who documented a significant increase in fallout as the fledging season progressed, our yearly models did not find a date effect. However, the likelihood of finding an effect of date depends partly on the timing and the duration of the study period. While opportunistic studies using intake records and citizen-science programs sample a wider temporal window, spanning before and after the fledging season, our surveys spanned a narrow temporal window during the WTSH fledging season. Thus, our results suggest that, due to interannual variability in the timing and magnitude of fallout, on average it is distributed evenly throughout our study period (November 6 –December 21). Fledging primarily occurs during this 6-week period, and is likely modulated by a variety of factors, including breeding phenology, chick development, and environmental conditions [10, 33, 35].

The positive relationship between wind speed and fallout is likely related to the fledglings using this environmental cue for fledging and relying on wind to take flight. One possible explanation for why we observed more fallout with higher windspeeds is that intermediate to strong windspeed enable WTSH to take flight, however a lack of flight experience and muscle development may make it difficult for fledgling birds to navigate in strong winds, thus leading to more fallout under higher wind speeds [36, 37].

Although the model results suggest that wind speed is more important than wind direction, an exception to this general pattern was observed in 2013, when peak fallout coincided with a period of moderate to weak southerly winds (S1 Fig). A previous study on Oʻahu suggested that winds from the southeast were more common during a year of very high fallout and hypothesized that birds were advected to the northwest and deposited along the entire windward coast of the island [22]. The lack of significance of wind direction in years other than 2013 may be related to the prevailing wind patterns on windward Oʻahu, which rarely switch from the northeast direction, thus limiting the comparison of different wind directions. Furthermore, because south-westerly winds during the study are characterized by lower speeds, any potential influence of direction is not independent from the wind speed effect discussed previously (S2 Fig).

Previous studies show an increase in fallout when prevailing winds are directed toward brightly lit coastal areas [10, 11]. Yet, the influence of wind direction is difficult to interpret since headings are circular (0–360 degrees) and should be carefully considered on a case-by-case basis. Because the prevailing winds in our study area are the northeasterly trade winds or southern (Kona) storms, these bi-directional wind headings facilitated the analysis and simplified the interpretation. With the exception of 2013, peak fallout occurred during trade winds (Fig 6).

Although our study did not show a significant effect of Julian date on fallout, previous road surveys from Oʻahu spanning ten years (2002 to 2010), revealed that November 25 was the peak fallout date, with 67% of the grounded WTSH found during a one-week period (21–27 November) [21]. We hypothesize that the timing of the moon phase, in relation to this peak fallout period could explain the interannual relationship of moon illumination and fallout. Namely, higher fallout occurs in years when the new moon overlaps the peak fallout week (21–27 November).

In 2015, the full moon occurred on November 28, whereas in 2013 and 2019 the moon phase was closer to a new moon on that date. It appears that when peak fallout coincides with a new moon, a single fallout peak occurs, thus causing a negative correlation with moon illumination. However, if peak fallout coincides with a full moon, the unimodal pattern breaks down, resulting in two smaller fallout peaks. Previous work yielded a quadratic relationship between the timing of the full moon and the number of Newell’s shearwater (Puffinus newelli) fallout, with fewer total groundings when the full moon occurred during the middle of the month [5]. When we replicated this analysis for WTSH, the quadratic model was not significant (R2 = 0.038, F2,5 = 1.140, p = 0.39), suggesting that annual fallout did not follow the same pattern with the timing of the full moon. Although, other variables such as the timing of moon rise, cloud cover, and topography blocking the moon were not taken into account and may play a role in the moon’s influence on fallout. Demographic factors, involving the size of the breeding population and the reproductive success likely influence the yearly supply of fledging chicks [5].

Implications for fallout mitigation

Our results are reassuring because they suggest that the shielded LED streetlights did not increase WTSH mortality due to fallout, as we hypothesized. Given the strong correlation between the dead birds observed in our road surveys and the live birds brought to SLP, there is no evidence suggesting that the shielded LED streetlights impacted the number of birds affected by fallout overall. However, because these new lights did not reduce fallout, wildlife managers may consider modifications such as dimming, wavelength alteration or motion sensors, to mitigate negative impacts to fledging WTSH on Oʻahu [6, 19].

A recent survey of lighting experts suggests that while LEDs can be adjusted to reduce light pollution and minimize wildlife impacts, yet municipalities rarely capitalize on those benefits [19]. For instance, although new-technology LED streetlights can filter out lower wavelengths [17], full spectrum white LED lights maximize brightness, and are commonly chosen to replace HPS streetlights. Furthermore, LEDs come in a variety of CCTs with options as low as 2200 K, the maximum temperature experts recommend for wildlife [17]. However, municipalities commonly implement 3000–5000 K LED streetlights because of their efficiency for human use [19]. Future studies should compare different LED lighting options in areas where seabird fallout occurs to determine the characteristics that best mitigate negative impacts to seabirds and other wildlife.

While it may be unfeasible to reduce light pollution wherever fallout occurs, areas near breeding colonies could be targeted for localized management [21]. In addition to diminishing light pollution during the fledging season, we also encourage community-based rescue efforts for WTSH to target fallout hotspots on Oʻahu on nights with low moon illumination and strong winds. Further documentation of fallout hotspots could help guide lighting management and rescue efforts throughout the Hawaiian Islands.

Finally, predictive fallout models are limited by the lack of comprehensive annual population estimates, which might have explained some of the interannual variation in the number of grounded birds. Thus, annual WTSH breeding population sizes and reproductive success would likely improve our understanding of fallout interannual variability and trends in Hawaiʻi. The findings and conclusions in this article are those of the authors and do not necessarily represent the official views of the U.S. Fish and Wildlife Service.

Annual data from sea life park and road surveys.

Comparison of annual WTSH fallout magnitude (total number of grounded birds) and timing (date ranges) from Sea Life Park intake records and road surveys (this study). Summary statistics (mean, median, and range) refer to the number of grounded birds encountered yearly, based on 16 standardized surveys spanning November 6 to December 21.

(TIF)

Click here for additional data file.

Variable importance in AICc analysis.

Scaled average variable weights. (>1 values indicate greater than average weight when variable was included in model; weights = 1 are average, weights <1 less than average).

(TIF)

Click here for additional data file.

Wind direction and fallout in 2013.

Wind direction and fallout during the 2013 fledging season. Black line is wind direction and white bars are number of birds per survey.

(TIF)

Click here for additional data file.

Wind speed and wind direction.

Scatterplot of wind speed and wind direction during the fledging seasons 2012–2019 (R2 = 0.71).

(TIF)

Click here for additional data file.

WTSH fallout data from road surveys.

Data from road surveys (total = 128) including variables year, Julian date, moon illumination (%), wind speed (knots), wind direction (degrees), light regime (HPS = high pressure sodium, LED = light emitting diode), and number of grounded WTSH observed.

(XLSX)

Click here for additional data file.

1 Nov 2021

A rebuttal letter that responds to each point raised by the academic editor and reviewer(s). You should upload this letter as a separate file labeled 'Response to Reviewers'.

A marked-up copy of your manuscript that highlights changes made to the original version. You should upload this as a separate file labeled 'Revised Manuscript with Track Changes'.

An unmarked version of your revised paper without tracked changes. You should upload this as a separate file labeled 'Manuscript'.

If you would like to make changes to your financial disclosure, please include your updated statement in your cover letter. Guidelines for resubmitting your figure files are available below the reviewer comments at the end of this letter.

If applicable, we recommend that you deposit your laboratory protocols in protocols.io to enhance the reproducibility of your results. Protocols.io assigns your protocol its own identifier (DOI) so that it can be cited independently in the future. For instructions see: https://journals.plos.org/plosone/s/submission-guidelines#loc-laboratory-protocols. Additionally, PLOS ONE offers an option for publishing peer-reviewed Lab Protocol articles, which describe protocols hosted on protocols.io. Read more information on sharing protocols at https://plos.org/protocols?utm_medium=editorial-email&utm_source=authorletters&utm_campaign=protocols.

We look forward to receiving your revised manuscript.

Kind regards,

Vitor Hugo Rodrigues Paiva, Ph.D.

Academic Editor

PLOS ONE

Journal Requirements:

When submitting your revision, we need you to address these additional requirements.

1. Please ensure that your manuscript meets PLOS ONE's style requirements, including those for file naming. The PLOS ONE style templates can be found at

https://journals.plos.org/plosone/s/file?id=wjVg/PLOSOne_formatting_sample_main_body.pdf and

https://journals.plos.org/plosone/s/file?id=ba62/PLOSOne_formatting_sample_title_authors_affiliations.pdf

2. Thank you for stating the following financial disclosure:

“This work was supported by Experiment.com (Blinded by the light: reducing shearwater deaths along a coastal highway in O 'ahu, Hawai'i) and The Eppley Foundation for Research (Blinded by the Light: Shearwater Deaths Along a Coastal Highway in O'ahu)”

Please state what role the funders took in the study.  If the funders had no role, please state: "The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript."

If this statement is not correct you must amend it as needed.

Please include this amended Role of Funder statement in your cover letter; we will change the online submission form on your behalf.

3. In your Data Availability statement, you have not specified where the minimal data set underlying the results described in your manuscript can be found. PLOS defines a study's minimal data set as the underlying data used to reach the conclusions drawn in the manuscript and any additional data required to replicate the reported study findings in their entirety. All PLOS journals require that the minimal data set be made fully available. For more information about our data policy, please see http://journals.plos.org/plosone/s/data-availability.

Upon re-submitting your revised manuscript, please upload your study’s minimal underlying data set as either Supporting Information files or to a stable, public repository and include the relevant URLs, DOIs, or accession numbers within your revised cover letter. For a list of acceptable repositories, please see http://journals.plos.org/plosone/s/data-availability#loc-recommended-repositories. Any potentially identifying patient information must be fully anonymized.

Important: If there are ethical or legal restrictions to sharing your data publicly, please explain these restrictions in detail. Please see our guidelines for more information on what we consider unacceptable restrictions to publicly sharing data: http://journals.plos.org/plosone/s/data-availability#loc-unacceptable-data-access-restrictions. Note that it is not acceptable for the authors to be the sole named individuals responsible for ensuring data access.

We will update your Data Availability statement to reflect the information you provide in your cover letter.

4. We note that Figure 1 in your submission contain map images which may be copyrighted. All PLOS content is published under the Creative Commons Attribution License (CC BY 4.0), which means that the manuscript, images, and Supporting Information files will be freely available online, and any third party is permitted to access, download, copy, distribute, and use these materials in any way, even commercially, with proper attribution. For these reasons, we cannot publish previously copyrighted maps or satellite images created using proprietary data, such as Google software (Google Maps, Street View, and Earth). For more information, see our copyright guidelines: http://journals.plos.org/plosone/s/licenses-and-copyright.

We require you to either (1) present written permission from the copyright holder to publish these figures specifically under the CC BY 4.0 license, or (2) remove the figures from your submission:

a. You may seek permission from the original copyright holder of Figure 1 to publish the content specifically under the CC BY 4.0 license.

We recommend that you contact the original copyright holder with the Content Permission Form (http://journals.plos.org/plosone/s/file?id=7c09/content-permission-form.pdf) and the following text:

“I request permission for the open-access journal PLOS ONE to publish XXX under the Creative Commons Attribution License (CCAL) CC BY 4.0 (http://creativecommons.org/licenses/by/4.0/). Please be aware that this license allows unrestricted use and distribution, even commercially, by third parties. Please reply and provide explicit written permission to publish XXX under a CC BY license and complete the attached form.”

Please upload the completed Content Permission Form or other proof of granted permissions as an "Other" file with your submission.

In the figure caption of the copyrighted figure, please include the following text: “Reprinted from [ref] under a CC BY license, with permission from [name of publisher], original copyright [original copyright year].”

b. If you are unable to obtain permission from the original copyright holder to publish these figures under the CC BY 4.0 license or if the copyright holder’s requirements are incompatible with the CC BY 4.0 license, please either i) remove the figure or ii) supply a replacement figure that complies with the CC BY 4.0 license. Please check copyright information on all replacement figures and update the figure caption with source information. If applicable, please specify in the figure caption text when a figure is similar but not identical to the original image and is therefore for illustrative purposes only.

The following resources for replacing copyrighted map figures may be helpful:

USGS National Map Viewer (public domain): http://viewer.nationalmap.gov/viewer/

The Gateway to Astronaut Photography of Earth (public domain): http://eol.jsc.nasa.gov/sseop/clickmap/

Maps at the CIA (public domain): https://www.cia.gov/library/publications/the-world-factbook/index.html and https://www.cia.gov/library/publications/cia-maps-publications/index.html

NASA Earth Observatory (public domain): http://earthobservatory.nasa.gov/

Landsat: http://landsat.visibleearth.nasa.gov/

USGS EROS (Earth Resources Observatory and Science (EROS) Center) (public domain): http://eros.usgs.gov/#

Natural Earth (public domain): http://www.naturalearthdata.com/

5. Please include a copy of Table A1 which you refer to in your text on page 11.

6. Please upload a copy of Supporting Information Figure S1 and S2 which you refer to in your text on page 18 and 19.

[Note: HTML markup is below. Please do not edit.]

Reviewers' comments:

Reviewer's Responses to Questions

Comments to the Author

1. Is the manuscript technically sound, and do the data support the conclusions?

The manuscript must describe a technically sound piece of scientific research with data that supports the conclusions. Experiments must have been conducted rigorously, with appropriate controls, replication, and sample sizes. The conclusions must be drawn appropriately based on the data presented.

Reviewer #1: Yes

Reviewer #2: Yes

Reviewer #3: Yes

**********

2. Has the statistical analysis been performed appropriately and rigorously?

Reviewer #1: I Don't Know

Reviewer #2: Yes

Reviewer #3: Yes

**********

3. Have the authors made all data underlying the findings in their manuscript fully available?

The PLOS Data policy requires authors to make all data underlying the findings described in their manuscript fully available without restriction, with rare exception (please refer to the Data Availability Statement in the manuscript PDF file). The data should be provided as part of the manuscript or its supporting information, or deposited to a public repository. For example, in addition to summary statistics, the data points behind means, medians and variance measures should be available. If there are restrictions on publicly sharing data—e.g. participant privacy or use of data from a third party—those must be specified.

Reviewer #1: Yes

Reviewer #2: Yes

Reviewer #3: Yes

**********

4. Is the manuscript presented in an intelligible fashion and written in standard English?

PLOS ONE does not copyedit accepted manuscripts, so the language in submitted articles must be clear, correct, and unambiguous. Any typographical or grammatical errors should be corrected at revision, so please note any specific errors here.

Reviewer #1: Yes

Reviewer #2: Yes

Reviewer #3: Yes

**********

5. Review Comments to the Author

Please use the space provided to explain your answers to the questions above. You may also include additional comments for the author, including concerns about dual publication, research ethics, or publication ethics. (Please upload your review as an attachment if it exceeds 20,000 characters)

Reviewer #1: This manuscript reveals a novel analysis of the effect of moon, wind speed and wind direction on the fallout of wedge-tailed shearwaters. Additionally, there was a transition from high-pressure sodium lamps to LED lamps on the route of the regular fallout survey. This provided a unique opportunity to study the effect of the change in streetlamps on the number of grounded birds. The study reveals that while the transition from HPS to LED streetlamps did not influence the fallout, the interaction between moon and wind speed did. While the data and analysis provide a new insight into the grounding of burrow-nesting shearwaters, I believe that the manuscript itself could be improved. Firstly, I think that moving sections of full model before the yearly models could improve the focus of the study. Furthermore, the authors could consider including the timing of moonrise, moonset, as well as cloud cover in their model. Finally, I include minor points in the attachment that could improve clarity and readability of the manuscript. My comments include first words of each line I referred to, since no line numbers were provided.

Reviewer #2: Overall it is an important paper that provides further evidence for effects of drivers in fallout events, crucially it provides initial evidence for the effect of altering urban lighting.

The text is very well written and clear without the need for much correction. I found the methods section too detailed and would suggest compressing some information and maybe place it in a supplement material annex.

Both the results and the discussion highlight the environmental aspects of the analysis while the LED is nearly absent from the results and, albeit well discussed, does not have a prominent placement in the discussion. I would suggest rearranging the paper in a way that highlights the LED vs HPS change as this is a most crucial information for current light pollution mitigation efforts. I would also suggest that the authors increase the analysis of the data by including if possible the total fallout per year registered by the SLP which correspond to the survey area, thus increasing the scope of light regime change analysis. You provide the totals in Fig.2 it would be possible to add this data to the light regime comparison? I understand that by drawing a parallel between SLP fallout records and your survey dead birds you are showing that your data is potentially indicative of the overall fallout and I agree, however I feel the analysis and results feel flat and could be worked on to present a more robust evidence.

Personally I would also suggest changing the title, moon and wind effects on fallout are known and not novel, their interaction is expected, on the other hand LEDs effects on fallout are unknown, understudied and needed!

As a final note the authors should take extra care when presenting manuscript for assessment, you are missing Figure 4, the lines are not numbered and the table was outside the page bounds so unreadable in the pdf version. I added line to the word version so that my comments were easier to follow.

Major changes:

> the authors collected data on utility poles (lighting systems) nearest the grounded bird. They could present an analysis of this data, i.e. was it possible to identify specific areas within the transects or fallout was widespread across it? Does this coincide with the previous research of Friswold et al 2020 (so data from 2002-2010) ? it could be interesting to discuss the implications of, after the change in lighting systems, the locations for fallout remain the same or change, especially in relation to the colonies you identify in line 138-142.

> The authors only used data of dead birds. Were live birds observed during the transects and if so were they added to the rescue center tally? I miss some discussion regarding the dead versus live birds in relation to the surveys. I understand that the authors provided a parallel between the two datasets by comparing proportions of live (rescue center data) and dead (this study), but if possible it would be interesting to include the live rescued birds in the full models.

For instances the change to LED did not provide increased deaths however is it possible to evaluate its effect on the total fallout numbers provided by the rescue center? I understand that it might not be possible to confirm the location of fallout for all birds and that the rescue center possibly obtains birds from outside the survey area, however if it were possible to include the birds that have been rescued within the survey area it would greatly improve the results and further increase the impact of this work. Thus discussion not only the overall negative effect of the light pollution (fallout) and the ‘no-effect’ of the light change as well as the mortality associated with both.

> Figure 4 is missing. If possible maybe combine the two time series graph (I understand this will generate three y scales but perhaps there is a way to illustrate the moon illumination as well).

> reduce the size of paragraph 5 in the introduction

> methodology could be shortened and the information placed in a supplementary material, for example the passages pertaining to the specificity of Poisson distributions and the AIC (lines 192-195; 211-221)

> If possible include 95% Confidence intervals in the results from your models. I fell it would provide a more cohesive interpretation of the results.

Minor changes:

I have added line numbers to the text version of the ms as they were missing.

> Carefull with the use of abbreviations such as HDOT. PLOS ONE guidelines state ‘Do not use non-standard abbreviations unless they appear at least three times in the text.’

> in the CCT mentions throughout the text, remove the degree symbol before K (absolute temperature Kelvin is not used with the degree symbol)

> line 43. Remove comma after petrels: ‘petrels and puffins’

> line 50. ‘conditions’ is repeated

> line 60. Add recent reference regarding powerline collisions Travers et al. 2021 Avian Conservation Ecology

> line 79. I disagree that a CCT >2700K is high. I would substitute by ‘recommended’ for example. Organism friendly.

> line 292 and 293. Revise text. Two sentences are unconnected, ‘yet’ followed by ‘however’ is confusing.

> Table 1. Correct to ‘mean ratio (VMR). In’

> Figure 3. Perhaps it would be appropriate to add a line plot with total of rescue birds from the rescue centre.

> Figure 5. I find this graph particularly useful for the wind and moon integrated evaluation. The smaller circles represent 0 fallout? I would substitute all 0 fallout records with a different symbol, for example an ‘x’ or a ‘+’. As it is its not possible to discern surveys with no fallout or surveys with low fallout. I wonder if using two colors to represent the two light regimes would also be usefull in this graph.

Reviewer #3: This opportunistic study provides much needed information on how changing light characteristics influences (or in this case, failed to influence) stranding rates of shearwaters vulnerable to fallouts, in addition to investigating more thoroughly the influence of environmental conditions, including moon phase and wind conditions. The manuscript is very clearly written and well laid out. The statistical analyses appear sound (however, please note that this is not my area of expertise) and results are clearly reported, followed by a well-referenced and interesting discussion. Overall, I very much enjoyed reading this manuscript and only have minor edits, as follows:

Page 4, last paragraph: “wedge-tailed shearwaters” should be capitalized.

Page 5, second line: USFWS should be spelled out the first time this acronym is used.

Page 6, second paragraph, last sentence: The prediction that northeasterly winds would lead to higher fallout rates goes against the observation that more birds were found stranded during anomalous southerly winds in 1994, as stated on page 5, second paragraph. Please clarify. Also, it would be useful to add an explanation of why >8 mph winds was chosen as “strong”. For example, in other parts of the world, winds under 13 mph would be classified as low, so it would be good for the reader to have a better understanding of what typical wind speeds are for this particular region. Also, for consistency, it would be better to present all wind speeds as knots, rather than a combination of mph and knots.

Table 1: the p-values for moon and date were cut-off from the table in the produced document.

While the Figure 4 caption is present in the text, the actual figure is missing from the document.

**********

6. PLOS authors have the option to publish the peer review history of their article (what does this mean?). If published, this will include your full peer review and any attached files.

If you choose “no”, your identity will remain anonymous but your review may still be made public.

Do you want your identity to be public for this peer review? For information about this choice, including consent withdrawal, please see our Privacy Policy.

Reviewer #1: Yes: Martyna Syposz

Reviewer #2: No

Reviewer #3: No

[NOTE: If reviewer comments were submitted as an attachment file, they will be attached to this email and accessible via the submission site. Please log into your account, locate the manuscript record, and check for the action link "View Attachments". If this link does not appear, there are no attachment files.]

While revising your submission, please upload your figure files to the Preflight Analysis and Conversion Engine (PACE) digital diagnostic tool, https://pacev2.apexcovantage.com/. PACE helps ensure that figures meet PLOS requirements. To use PACE, you must first register as a user. Registration is free. Then, login and navigate to the UPLOAD tab, where you will find detailed instructions on how to use the tool. If you encounter any issues or have any questions when using PACE, please email PLOS at figures@plos.org. Please note that Supporting Information files do not need this step.

Submitted filename: Review Urmston et al.docx

Click here for additional data file.

28 Dec 2021

RESPONSES TO REVIEWER COMMENTS

Journal Requirements

Comment 1: Please ensure that your manuscript meets PLOS ONE's style requirements, including those for file naming. The PLOS ONE style templates can be found at

https://journals.plos.org/plosone/s/file?id=wjVg/PLOSOne_formatting_sample_main_body.pdf and

https://journals.plos.org/plosone/s/file?id=ba62/PLOSOne_formatting_sample_title_authors_affiliations.pdf

Response: We have formatted the ms according to the journal requirements.

Comment 2: Thank you for stating the following financial disclosure:

“This work was supported by Experiment.com (Blinded by the light: reducing shearwater deaths along a coastal highway in O 'ahu, Hawai'i) and The Eppley Foundation for Research (Blinded by the Light: Shearwater Deaths Along a Coastal Highway in O'ahu)”

Please state what role the funders took in the study. If the funders had no role, please state: "The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript."

If this statement is not correct you must amend it as needed.

Please include this amended Role of Funder statement in your cover letter; we will change the online submission form on your behalf.

Response: We provided an amended Role of Funder statement in our cover letter, to reflect that this work was supported by Experiment.com (Blinded by the light: reducing shearwater deaths along a coastal highway in O 'ahu, Hawai'i) and The Eppley Foundation for Research (Blinded by the Light: Shearwater Deaths Along a Coastal Highway in O'ahu). The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript.

Comment 3: In your Data Availability statement, you have not specified where the minimal data set underlying the results described in your manuscript can be found. PLOS defines a study's minimal data set as the underlying data used to reach the conclusions drawn in the manuscript and any additional data required to replicate the reported study findings in their entirety. All PLOS journals require that the minimal data set be made fully available. For more information about our data policy, please see http://journals.plos.org/plosone/s/data-availability.

Upon re-submitting your revised manuscript, please upload your study’s minimal underlying data set as either Supporting Information files or to a stable, public repository and include the relevant URLs, DOIs, or accession numbers within your revised cover letter. For a list of acceptable repositories, please see http://journals.plos.org/plosone/s/data-availability#loc-recommended-repositories. Any potentially identifying patient information must be fully anonymized.

Important: If there are ethical or legal restrictions to sharing your data publicly, please explain these restrictions in detail. Please see our guidelines for more information on what we consider unacceptable restrictions to publicly sharing data: http://journals.plos.org/plosone/s/data-availability#loc-unacceptable-data-access-restrictions. Note that it is not acceptable for the authors to be the sole named individuals responsible for ensuring data access.

We will update your Data Availability statement to reflect the information you provide in your cover letter.

Response: We provided a data file for inclusion as supplementary table 2, which includes all road survey observations and environmental data, sampled every three days between Nov. 6 and Dec. 21 for eight years (2012 – 2019). These were the data used to develop the multi-year full model and the eight single-year yearly models.

Comment 4: We note that Figure 1 in your submission contain map images which may be copyrighted. All PLOS content is published under the Creative Commons Attribution License (CC BY 4.0), which means that the manuscript, images, and Supporting Information files will be freely available online, and any third party is permitted to access, download, copy, distribute, and use these materials in any way, even commercially, with proper attribution. For these reasons, we cannot publish previously copyrighted maps or satellite images created using proprietary data, such as Google software (Google Maps, Street View, and Earth). For more information, see our copyright guidelines: http://journals.plos.org/plosone/s/licenses-and-copyright.

We require you to either (1) present written permission from the copyright holder to publish these figures specifically under the CC BY 4.0 license, or (2) remove the figures from your submission:

a. You may seek permission from the original copyright holder of Figure 1 to publish the content specifically under the CC BY 4.0 license.

We recommend that you contact the original copyright holder with the Content Permission Form (http://journals.plos.org/plosone/s/file?id=7c09/content-permission-form.pdf) and the following text:

“I request permission for the open-access journal PLOS ONE to publish XXX under the Creative Commons Attribution License (CCAL) CC BY 4.0 (http://creativecommons.org/licenses/by/4.0/). Please be aware that this license allows unrestricted use and distribution, even commercially, by third parties. Please reply and provide explicit written permission to publish XXX under a CC BY license and complete the attached form.”

Please upload the completed Content Permission Form or other proof of granted permissions as an "Other" file with your submission.

In the figure caption of the copyrighted figure, please include the following text: “Reprinted from [ref] under a CC BY license, with permission from [name of publisher], original copyright [original copyright year].”

b. If you are unable to obtain permission from the original copyright holder to publish these figures under the CC BY 4.0 license or if the copyright holder’s requirements are incompatible with the CC BY 4.0 license, please either i) remove the figure or ii) supply a replacement figure that complies with the CC BY 4.0 license. Please check copyright information on all replacement figures and update the figure caption with source information. If applicable, please specify in the figure caption text when a figure is similar but not identical to the original image and is therefore for illustrative purposes only.

The following resources for replacing copyrighted map figures may be helpful:

USGS National Map Viewer (public domain): http://viewer.nationalmap.gov/viewer/

The Gateway to Astronaut Photography of Earth (public domain): http://eol.jsc.nasa.gov/sseop/clickmap/

Maps at the CIA (public domain): https://www.cia.gov/library/publications/the-world-factbook/index.html and https://www.cia.gov/library/publications/cia-maps-publications/index.html

NASA Earth Observatory (public domain): http://earthobservatory.nasa.gov/

Landsat: http://landsat.visibleearth.nasa.gov/

USGS EROS (Earth Resources Observatory and Science (EROS) Center) (public domain): http://eros.usgs.gov/#

Natural Earth (public domain): http://www.naturalearthdata.com/

Response: Figure 1 has not been previously published elsewhere but is an original created using Arc GIS Pro. This figure shows a series of geographic features overlaid on the ArcGIS Pro Software Version 2.5 topographic basemap (ESRI 2020). To this end, we added a citation to the references section:

ESRI (2020). ArcGIS Pro (Version 2.5). Esri Inc. https://www.esri.com/en-us/arcgis/products/arcgis-pro/overview.

Comment 5: Please include a copy of Table A1 which you refer to in your text on page 11.

Response: We apologize for this oversight. We added this table, relabeled as Table S1, in the Supplementary materials at end of manuscript.

Comment 6: Please upload a copy of Supporting Information Figure S1 and S2 which you refer to in your text on page 18 and 19.

Response: We apologize for this oversight. We have uploaded Figures S1 and S2, in the Supplementary materials at end of manuscript.

Comments from Reviewer 1

Comment 1: Firstly, I think that moving sections of full model before the yearly models could improve the focus of the study.

Response: Thank you for the suggestion. We reorganized the ms and now focus on the multi-year model, which addresses the shift from the HPS to the LED lights. Then, we discuss the year-to-year variability using the single-year models.

Comment 2: Furthermore, the authors could consider including the timing of moonrise, moonset, as well as cloud cover

Response: While we agree that developing a more comprehensive metric if lunar illumination that includes the timing of moonrise / moonset and the cloud cover could improve the fit of the model, we chose to not do so for two reasons: (1) complexity of weather conditions and (2) applicability of the model results. Thus, we respectfully would prefer to not modify the current environmental datasets and resulting models.

It is inherently difficult to obtain a cloud cover metric that accurately describes our study site, which encompasses a section of the windward and a section of the leeward coasts of our island. Because storms sweep across from the north or from the south, the weather on the windward and leeward shores of our island is often different. Thus, cloudiness and precipitation are often very different on either side (northwest or southeast) of Makapuu Point. Thus, developing a metric for the entire survey would require a weighing the average of the cloud cover on both shores, using the relative survey length on along shores. While this approach is doable, we fear that it would hinder the applicability of our model results, for resource managers and wildlife rescue organizations.

Thus, to ensure the wide use of our model results, we decided to used “simple” environmental variables available from internet datasets: percent moon illumination from the US Naval Observatory and hight-time (6 pm to 6 am) hourly wind speed from the Pacific Islands Ocean Observing System. Using these variables will allow managers and wildlife rescuers to develop predictive models using wind data predictions and the lunar almanac.

Comment 3: Finally, I include minor points in the attachment that could improve clarity and readability of the manuscript

Accepted most minor edits and made changes accordingly in manuscript.

Comments from Reviewer 2

Comment 1: I found the methods section too detailed and would suggest compressing some information and maybe place it in a supplement material annex

Response: We reorganized the methods section and moved some of the material to a “Fallout Modelling” section, to minimize redundancy and streamline the methodological descriptions. While we did not manage to shorten this section substantially, we feel this information is useful so readers can fully evaluate the data analysis. Thus, we would like to retain this material in the ms, rather than moving it to the supplementary materials section.

Comment 2: Both the results and the discussion highlight the environmental aspects of the analysis while the LED is nearly absent from the results and, albeit well discussed, does not have a prominent placement in the discussion. I would suggest rearranging the paper in a way that highlights the LED vs HPS change as this is a most crucial information for current light pollution mitigation efforts.

Response: Originally, we focused the ms on the environmental drivers of fallout, rather than on the influence of the changes in highway lighting, because the statistics highlighted the significant influences of moonlight and wind and did not find an overall difference before / after the changes in lighting regimes. Yet, given the reviewer’s suggestions we have refocused the ms to highlight the analysis of the lighting regimes in two ways: (i) we have modified the title of the ms, and (ii) we have rearranged the material in the methods / results / discussion sections, by placing the multi-year model before we discuss the yearly models.

Comment 3: I would also suggest that the authors increase the analysis of the data by including if possible the total fallout per year registered by the SLP which correspond to the survey area, thus increasing the scope of light regime change analysis. You provide the totals in Fig.2 it would be possible to add this data to the light regime comparison? I understand that by drawing a parallel between SLP fallout records and your survey dead birds you are showing that your data is potentially indicative of the overall fallout and I agree, however I feel the analysis and results feel flat and could be worked on to present a more robust evidence.

Response: While analyzing the SLP intake fallout records would be a valuable and insightful exercise, we feel that this analysis is not appropriate in our current ms for three reasons:

First, there is a spatial mis-match of the two datasets: our surveys focus on a fallout hotspot within a relatively small geographic area, and the SLP intake records include island-wide returns of grounded birds. Moreover, because most of the grounded birds do not include detailed location information, usually merely the city or highway where they were found, it would be impossible for us to subset these records to only analyze returns from our survey area. In fact, 20 – 40 of the records in any given year, do not include any location data. Moreover, while we know when the lighting regime changes for our survey area occurred between spring and summer of 2016, the shift from HPS to LEDs has been a lengthy process, with different areas of the island transitioning at different times over the span of a 2-year period (2015 – 2017). Thus, we would not be able to break up the SLP time series into a HPS vs an LED lighting regime. These uncertainties would greatly complicate the analysis of fallout trends.

Second, there is an environmental driver mismatch: due to the different spatial extent and the timing (daily vs every 3 days) of the two datasets, we would have to either perform the SLP analysis using daily environmental data or we would have to aggregate the SLP data into 3-days periods to match the timing of the road surveys. The former approach would complicate the interpretation of the two separate analyses, and the latter approach would miss the fine temporal resolution in the SLP data.

Finally, due to these inherent mis-matches, we are concerned that including both analyses in this ms would greatly blur the main messages of the ms, because we would be discussing and comparing four analyses: two multi-year models (SLP intake and road surveys) and two sets of yearly models (SLP intake and road surveys). If the model results agreed, it would still be difficult for the readers to digest all of these results. If the model did not agree, then the discussion section of the ms would have to discuss the reasons behind these discrepancies.

Thus, we would like to only include the highly-standardized and focused road surveys in this ms. Once we have shown that the change in highway lighting did not have an effect, future papers can investigate island-wide fallout patterns using a larger dataset (2012 – 2021).

Comment 4: Personally I would also suggest changing the title, moon and wind effects on fallout are known and not novel, their interaction is expected, on the other hand LEDs effects on fallout are unknown, understudied and needed!

Response: Following this reviewer’s suggestion, we have changed the title of the ms to reflect the influence of changing lighting conditions, rather than the environmental drivers (moon and wind). However, because two changes in lighting conditions took place simultaneously: light bulb (HPS to LED) and light fixture design (unshielded to shielded), we do not explicitly mention LEDs in the title. Thus, our revised title states: “Quantifying wedge-tailed shearwater (Ardenna pacifica) fallout after changes in highway lighting on Southeast Oʻahu, Hawaiʻi”

Comment 5: The authors collected data on utility poles (lighting systems) nearest the grounded bird. They could present an analysis of this data, i.e. was it possible to identify specific areas within the transects or fallout was widespread across it? Does this coincide with the previous research of Friswold et al 2020 (so data from 2002-2010) ? it could be interesting to discuss the implications of, after the change in lighting systems, the locations for fallout remain the same or change, especially in relation to the colonies you identify in line 138-142.

Response: We absolutely agree with this comment! In fact, after investigating the influence of the lighting regimes on overall fallout first, we performed a detailed spatial analysis of these fallout data. We chose to focus this first ms on the lighting regime and will publish the fine-scale spatial analyses on a second ms. We felt that publishing both analyses together would yield a rather long and unfocused ms.

Comment 6: The authors only used data of dead birds. Were live birds observed during the transects and if so were they added to the rescue center tally? I miss some discussion regarding the dead versus live birds in relation to the surveys. I understand that the authors provided a parallel between the two datasets by comparing proportions of live (rescue center data) and dead (this study), but if possible it would be interesting to include the live rescued birds in the full models. For instances the change to LED did not provide increased deaths however is it possible to evaluate its effect on the total fallout numbers provided by the rescue center? I understand that it might not be possible to confirm the location of fallout for all birds and that the rescue center possibly obtains birds from outside the survey area, however if it were possible to include the birds that have been rescued within the survey area it would greatly improve the results and further increase the impact of this work. Thus discussion not only the overall negative effect of the light pollution (fallout) and the ‘no-effect’ of the light change as well as the mortality associated with both.

Response: First of all, we would like to clarify that our surveys involve both live and dead birds. Whenever we found live birds along the road, we recorded them for our survey and brought them to Sea Life Park. Yet, the vast majority (99%) of the WTSH we encounter during our road surveys are dead, because by the time our surveys occur in the morning, most of the grounded shearwaters have either made their way off the road or have been hit by cars.

Regarding the suggestion to include the SLP intake records of rescued birds in the analysis, we would like to keep these two distinct datasets apart due to their inherently different spatial / temporal scales. Please see previous explanation in response to a similar suggestion by reviewer #1.

The reasons why we included the SLP intake records in this paper were to: (i) show that our surveys from Nov 6 to Dec 21 sampled the period of fallout throughout the island; and (ii) show that – despite the small geographic scale of our study, the year-to-year variability in fallout we documented was correlated to the pattern of island-wide fallout. Because there was a significant cross-correlation of both datasets (which shared 85% of their year-to-year variability), we contend that the same interannual drivers (e.g., ocean productivity, weather patterns) of island-wide fallout also influence the magnitude of fallout within the hotspot in the SE corner of the island.

Comment 7: Figure 4 is missing. If possible maybe combine the two time series graph (I understand this will generate three y scales but perhaps there is a way to illustrate the moon illumination as well).

Response: Apologies for erroneously leaving out Figure 4. I have added it back in. I had considered combining the two time series, and attempted it, but felt that it made the graph difficult to read and interpret, thus decided to keep them separate.

Comment 8: reduce the size of paragraph 5 in the introduction

Response: We removed some text and condensed the introduction.

Comment 9: methodology could be shortened and the information placed in a supplementary material, for example the passages pertaining to the specificity of Poisson distributions and the AIC (lines 192-195; 211-221)

Response: We removed some text and reorganized this section, which includes a general opening and then splits into two modelling sections (multi-year vs yearly models). This new organization removed some redundancy

Comment 10: If possible include 95% Confidence intervals in the results from your models. I fell it would provide a more cohesive interpretation of the results.

Response: While we see how adding 95% confidence intervals may enhance the interpretation of the results, we included merely the parameters and associated values for two reasons: 1) logistically, we wanted to keep the table small ensure that it was not overwhelming and 2) statistically, because the yearly models are based on small sample sizes (n=16), they provide an exploratory perspective. Thus, we want to de-emphasize the specific parameter values. Rather we only provide the parameter estimates to show the +/- signs and the p-values for reference. If we added the SE’s we would also want to add Z scores which would add two more columns for each explanatory variable. On the other hand, because the multi-year model has a much larger sample size (n=128) this analysis provides a more rigorous statistical analysis. Thus, when we report this model’s result, we provide the parameter values, +/- SEs, Z scores, and p-values so the reader can fully evaluate the significance of the model.

Minor Changes:

>I have added line numbers to the text version of the ms as they were missing.

Sorry for not including line numbers. I have added them in to my final draft.

> Careful with the use of abbreviations such as HDOT. PLOS ONE guidelines state ‘Do not use non-standard abbreviations unless they appear at least three times in the text.’

I removed the abbreviation and spelt out the acronym.

> in the CCT mentions throughout the text, remove the degree symbol before K (absolute temperature Kelvin is not used with the degree symbol)

Thank you for making me aware of this. I have removed the degree symbol.

> line 43. Remove comma after petrels: ‘petrels and puffins’

Done

> line 50. ‘conditions’ is repeated

Thank you for catching this mistake.

> line 60. Add reference regarding powerline collisions Travers et al. 2021 Avian Conservation Ecology

Thank you for sharing this reference. I have added it.

> line 79. I disagree that a CCT >2700K is high. I would substitute by ‘recommended’ for example. Organism friendly.

Thank you for this recommendation. I have changed the wording.

> line 292 and 293. Revise text. Two sentences are unconnected, ‘yet’ followed by ‘however’ is confusing. Thank you. Rephrased.

> Table 1. Correct to ‘mean ratio (VMR).

Not sure what is meant with this suggestion. The text states” variance to mean ratio”.

> Figure 3. Perhaps it would be appropriate to add a line plot with total of birds from the rescue centre.

We respectfully disagree because we have provided the yearly totals in S1 Table, and in Figure 2.

> Figure 5. I find this particularly useful for the wind and moon integrated evaluation. The smaller circles represent 0 fallout? I would substitute all 0 fallout records with a different symbol, for example an ‘x’ or a ‘+’. As it is its not possible to discern surveys with no fallout or surveys with low fallout. I wonder if using two colors to represent the two light regimes would also be usefull in this graph.

Thank you for this suggestion. We changed the zero values from an open circle to a small solid black dot, and edited the figure legend to explain the symbology.

Comments from Reviewer 3

Comment 1: Page 4, last paragraph: “wedge-tailed shearwaters” should be capitalized.

Response: Thanks for catching that! Capitalized.

Comment 2: Page 5, second line: USFWS should be spelled out the first time this acronym is used.

Response: Good point. Spelled out.

Comment 3: Page 6, second paragraph, last sentence: The prediction that northeasterly winds would lead to higher fallout rates goes against the observation that more birds were found stranded during anomalous southerly winds in 1994, as stated on page 5, second paragraph. Please clarify. Also, it would be useful to add an explanation of why >8 mph winds was chosen as “strong”. For example, in other parts of the world, winds under 13 mph would be classified as low, so it would be good for the reader to have a better understanding of what typical wind speeds are for this particular region. Also, for consistency, it would be better to present all wind speeds as knots, rather than a combination of mph and knots.

Response: We have reworded our expectations by removing the 8 knot threshold and by explaining the link between NE winds and shearwater deposition along the SE shore of the island, downwind from two nesting colonies on offshore islets. Now, the text states: “Because WTSH rely on wind to take flight and may become disoriented in the absence of moonlight, we predicted higher fallout during windy nights of low moon illumination. In particular, due to the location of our study area, southwest from two breeding colonies, we anticipated that strong northeasterly winds would drive the fledging birds towards shore”. We provide an explanation below and can include similar wording in the ms, if you deem that it would help clarify our hypothesis.

When relating WTSH fallout to wind direction, we expect different patterns when we compare the counts of grounded birds within our small-scale study area versus the island-wide results. Work and Rameyer (1999) analyzed island-wide patterns in 1992 – 1994, using Sea Life Park (SLP) intake records. They reported widespread fallout in 1994, with WTSH distributed throughout the windward (east), leeward (south and west) and north shores of the island (see enclosed figure).

Yet, while WTSH fallout was widely distributed in 1994, when southerly winds deposited the birds throughout the entire island, large numbers of grounded birds were found within our study area, in the vicinity of the breeding sites on offshore islets (in the SE corner of the island). This result underscores the correlation we documented between SLP intake records and our road surveys.

Thus, we hypothesize that strong winds lead to higher fallout. In particular, we feel like the tradewinds (NE winds) will deposit birds down-wind from the offshore islet colonies, leading to highly focused fallout in our study area, rather than widely-dispersed fallout throughout the island.

Comment 4: Table 1: the p-values for moon and date were cut-off from the table in the produced document.

Response: We apologize if the table was not readable in this format. The journal submission guidelines stated not to reformat the table if it did not fit within the page dimensions, because it will be turned sideways and displayed in landscape orientation.

Comment 5: While the Figure 4 caption is present in the text, the actual figure is missing from the document.

Response: Apologies for erroneously leaving out Figure 4. We added it back in.

Additional Editorial Comments:

• Update formatting and file name - upload figures as individual files (Fig1.tif)

Done

• Add supporting information at the end of manuscript after refs

Done

• 5. Please include a copy of Table A1 which you refer to in your text on page 11.

We removed Table A1, which originally referred to the table showing all tested model combos. Now, Table A1 refers to the fallout summary from the SLP Intake Records and the Road Surveys

• 6. Please upload a copy of Supporting Information Figure S1 and S2 which you refer to in your text on page 18 and 19

We added these figures to the supplementary materials

• Update the role of the funder

We updated the role of the funder

• Make data publicly available

We provide a copy of the data used for the fallout models as S Table 2

• Make sure copyright info is all good for figure 1 map - added line in caption indicating that map was made in ArcGIS Pro and added ArcGIS Pro reference in references section. I made this map, it was no taken from another source.

Submitted filename: PONE-D-21-31078_R1_ResponseReviewers_122821.docx

Click here for additional data file.

2 Feb 2022

6 Mar 2022

Reviewer 1

Comment #1 Line 40 – remove unnecessary bracket.

Response: Thank you for catching this mistake, I removed it.

Comment #2 Line 42 – USFWS should be spell out.

Response: Thank you for this suggestion, I have spelled it out.

Comment #3 Table 1 – spell out Wind Speed, as WS is not clear.

Response: Thank you for this suggestion. I have typed out wind speed in table #1. I left the abbreviation in table 2 as it is identified in the figure caption, and there is limited space in the column headings.

Comment #4 Line 275 – I am uncertain what do you mean by ‘larger degree angle’? Maybe try wind coming from land/offshore.

Response: Thank you for your comment. I agree that ‘larger degree angle’ is not very clear. I have changed it to say “wind blowing from the southwest” which I believe will be more intuitive to the readers.

Comment #5 Line 289 – can you specify ‘a single new moon period’

Response: Changed to “a new moon period occurred in the middle of the fledging season, leading to a single peak in fallout”. This is opposed to the previous line describing a year with two new moon periods, early and late in the season.

Comment #6 Lines 406-408 – It could be also a result of other moon cycle variables (timing, location of moon in respect to earth) and environmental conditions (cloud cover) that were not taken into account in the model.

Response: If the periodic moon cycle or other episodic environmental drivers (cloud cover and wind) were the drivers of fallout, we would expect that the 9-year (2002 – 2010) dataset of daily fallout records to yield a broad fallout distribution, caused by the overlapping of distinct annual peaks with different timing. Instead, we hypothesize that the consistent weekly fallout peak (Nov. 21 – 27) is caused by the timing of fledging, riven by the phenology of chick hatching and development. Nonetheless, I inserted more detail in this section:

“When we replicated this analysis for WTSH, the quadratic model was not significant (R2=0.038, F2,5=1.140, p=0.39), suggesting that annual fallout did not follow the same pattern with the timing of the full moon. Although, other variables such as the timing of moon rise, cloud cover, and topography blocking the moon were not taken into account and may play a role in the moon’s influence on fallout.”

Comment #7 Figure 4 – I do not see solid black dots indicating absence of fallout.

Response: Thank you for catching this. I changed it to “small solid dots”, they are grey not black. I have updated the figure caption.

Reviewer 2

Comment #1

Still unclear in the text regarding the state of collected birds: I understand from the authors that both live and dead birds were collected during the surveys, even if 99% of birds were dead. I maintain my recommendation to include this information in the ms, thus facilitate comparison with other studies and clarifying the work. For example: in methods explain that both live and dead birds were collected but since most were found dead (99%) only these were used for the analysis (unless all were used for the analysis, in which case please correct line 127 and other mentions of 'carcass' across the text); In the discussion, mention that even if this study only used dead birds, as you have found a good parallel between SLP intakes (island wide records of fallout, both live and dead) and this study, there is no evidence for a different effect of light change to the state of the birds, for example. These additions do not need to be lengthy, a short sentence will suffice.

Response: Thank you for these suggestions. I have added some clarifying language to distinguish that our surveys focused on dead birds, and we only used dead birds in our analysis.

“while visually searching for dead birds in each lane and along the shoulder. Since these surveys were conducted in the morning, likely a full 12 hours after fledging time, almost all the birds we observed were deceased. In 8 years of surveys, we observed 2 live birds, which were brought to Sea Life Park for rehabilitation and not counted in our analysis. All dead birds sighted”

“Our results are reassuring because they suggest that the shielded LED streetlights did not increase WTSH mortality due to fallout, as we hypothesized. Given the strong correlation between the dead birds observed in our road surveys and the live birds brought to SLP, there is no evidence suggesting that the shielded LED streetlights impacted the number of birds affected by fallout overall.”

Comment #2

Introduction: 4th paragraph (line 31-38) is a bit redundant, it could be shortened to be more concise and direct.

Response: Thank you for addressing this. I agree that this paragraph sounded misleading. I have rephrased this section to address the point that many newly implemented lights feature both shielding (good for birds) AND broad spectrum LEDs with high CCT values (potentially bad for birds). So the point I am trying to make is that we have a good change and a potentially bad change happening at the same time, and we don’t know how these changes coupled together are impacting birds.

“Mitigation measures often target light directionality, whereby streetlights are shielded through the use of a “full-cutoff” design, which inhibits light emission above the horizontal plane of the fixture. This approach, when applied to HPS lights, reduced Newell’s Shearwater (Puffinus newelli) fallout on Kauai (Hawaiʻi) [16]. Although mitigation is being addressed through shielding, the common use of optimized LEDs with broad spectra and Correlated Color Temperature (CCT) greater than the maximum recommended value for wildlife (2200 K) may be a cause for concern [17]. While modern LED lights possess the flexibility to give off a range of low to high CTT, short-wavelength light with high CCT is a common choice because of its efficiency [19]. The effectiveness of light shielding coupled with the use of broad spectrum, high CCT LEDs is unknown.”

Comment #3

Abstract: 'due to exhaustion or collision' (use singular for collision).

Response: Thank you for the suggestion, I have made this edit.

Comment #4

Line 14 - missing word: 'especially in the absence of "moonlight"...'

Response: Thank you for your thorough review and for catching this mistake! I have added in the word.

Comment #5

Line 20 - 'that can affect seabirds' or better yet, 'that can affect fallout'?

Response: Thanks for the suggestion, I have changed the phrasing to say “that can affect fallout”.

Comment #6

line 64: 3000-4000 K LED lights (missing LED)

Response: Thank you, I have added in “LED”

Comment #7

Line 64: I think (might be wrong) that is more accurate to say that CCT is the measure of how warm and cold a light appears, Kelvin is the unit.

Response: I think you are correct. I have rephrased this to say “where Kelvin (K) is a unit of measurement for CCT. Lower CCT indicates a warm yellow-orange appearance whereas higher CCT indicates cool blue light [18].”

Comment #8

Line 123: I would use full extent month names throughout the text (November 6 instead of Nov.). Or at least homogenize across the text.

Response: Thank you for pointing this out. I used full names in text and left abbreviations in tables only.

Comment #9

Line 285: instead of 'throughout this study' something like 'within the time frame of this study'

Response: Good suggestion, made change.

Comment #10

Line 324: distinguish between these two factors

Response: Thanks for the suggestion, I have made this edit.

Comment #11

Line 337: I agree that evidence indicates stronger winds will increase number of fledgling. But wind affects groundings two ways a) increase the number of fledglings and b) push these inland (especially when southwestern winds are about correct?). I would add this last part. you can reference 10 and 11 refs [Rodríguez et al 2014 and Syposz et al 2018]

Response: I agree that this is likely the case, but this section is specifically referencing our multi-year model, which identified the Wind Speed * Moon interaction as the single most important variable, but did not include wind direction. Thus, we would like to avoid referring to wind speed and the advection of chicks inland. We are developing a second ms analyzing the spatial pattern of fallout in relation to wind speed and direction, and this ms will analyze and discuss those patterns.

Reviewer 3

Comment #1

Reviewer #3: While the authors have addressed the previous round of comments, I have noticed some conflicting information in the introduction which I feel needs to be resolved as it impacts the logic of the main hypothesis. Line 34 states that light with high CCT (>2700K) is recommended for wildlife (note that a reference to support this statement is needed). Line 64 states that streetlights were changed from 2200K HPS lights to shielded 3000-4000 K LED lights (note that the word "LED" is missing in line 64 and should be added). Following the logic from the former statement (i.e., that light with >2700K is recommended for wildlife), the new LED light regime would be beneficial to shearwaters, however, the hypothesis states the opposite. Furthermore, the hypothesis is based on differences in wavelength rather than on CCT and K which are the light characteristics (related to temperature) discussed in detail in the introduction. Pulling all this together, it appears that the main hypothesis is based on the previous study of shearwaters which showed maximum light absorption by white LED lights emitting short wavelengths vs HPS lights which emit longer wavelengths, which presumably means that they are more attracted to white LED lights. As currently written, the introduction lacks a clear description of the relationship between the two light characteristics presented, namely: temperature and wavelength, and how these informed the predictions of the hypothesis.

Response: Thank you for these comments. I agree that this section sounded misleading. I have rephrased this section to address the point that many newly implemented lights feature both shielding (good for birds) AND broad spectrum LEDs with high CCT values (potentially bad for birds). The point I am trying to make is that we have a good change and a potentially bad change happening at the same time, and we don’t know how these changes coupled together impacts birds.

Also thank you for stating that a description of the relationship between wavelength and CCT is lacking, I have added this in the introduction:

“Mitigation measures often target light directionality, whereby streetlights are shielded through the use of a “full-cutoff” design, which inhibits light emission above the horizontal plane of the fixture. This approach, when applied to HPS lights, reduced Newell’s Shearwater (Puffinus newelli) fallout on Kauai (Hawaiʻi) [16]. Although[JU1] [JU2] mitigation is being addressed through shielding, the common use of optimized LEDs with broad spectra and Correlated Color Temperature (CCT) greater than the maximum recommended value for wildlife (2200 K) may be a cause for concern [17]. While modern LED lights possess the flexibility to give off a range of low to high CTT, short-wavelength light with high CCT is a common choice because of its efficiency [19]. The effectiveness of light shielding coupled with the use of broad spectrum, high CCT LEDs is unknown.”

Additionally I corrected this line and provided a proper reference:

Corrected this line - added reference: greater than the recommended value for wildlife (<2200 K) may be a cause for concern [17].

Other minor edits:

Comment #2

Line 20: "seabird" should be plural

Response: I changed this to say “fallout” instead, as per Reviewer 2 recommendation

Comment #3

The sentence in lines 65 and 66 seems out of place as it talks about CCT while the previous sentence talks about K; the latter sentence needs to be tied somehow to the previous one.

Response: Thank you - rephrased 3000 – 4000 K LED lights, where Kelvin (K) is a unit of measurement for CCT. Lower CCT indicates a warm yellow-orange appearance whereas higher CCT indicates cool blue light [18].

Submitted filename: PONE-D-21-31078_R2 Response to Reviewers_0306_2022.docx

Click here for additional data file.

9 Mar 2022

Quantifying wedge-tailed shearwater (Ardenna pacifica) fallout after changes in highway lighting on Southeast Oʻahu, Hawaiʻi.

PONE-D-21-31078R2

Dear Dr. Urmston,

We’re pleased to inform you that your manuscript has been judged scientifically suitable for publication and will be formally accepted for publication once it meets all outstanding technical requirements.

Within one week, you’ll receive an e-mail detailing the required amendments. When these have been addressed, you’ll receive a formal acceptance letter and your manuscript will be scheduled for publication.

An invoice for payment will follow shortly after the formal acceptance. To ensure an efficient process, please log into Editorial Manager at http://www.editorialmanager.com/pone/, click the 'Update My Information' link at the top of the page, and double check that your user information is up-to-date. If you have any billing related questions, please contact our Author Billing department directly at authorbilling@plos.org.

If your institution or institutions have a press office, please notify them about your upcoming paper to help maximize its impact. If they’ll be preparing press materials, please inform our press team as soon as possible -- no later than 48 hours after receiving the formal acceptance. Your manuscript will remain under strict press embargo until 2 pm Eastern Time on the date of publication. For more information, please contact onepress@plos.org.

Kind regards,

Vitor Hugo Rodrigues Paiva, Ph.D.

Academic Editor

PLOS ONE

Additional Editor Comments (optional):

Reviewers' comments:

15 Mar 2022

PONE-D-21-31078R2

Quantifying wedge-tailed shearwater (Ardenna pacifica) fallout after changes in highway lighting on Southeast Oʻahu, Hawaiʻi.

Dear Dr. Urmston:

I'm pleased to inform you that your manuscript has been deemed suitable for publication in PLOS ONE. Congratulations! Your manuscript is now with our production department.

If your institution or institutions have a press office, please let them know about your upcoming paper now to help maximize its impact. If they'll be preparing press materials, please inform our press team within the next 48 hours. Your manuscript will remain under strict press embargo until 2 pm Eastern Time on the date of publication. For more information please contact onepress@plos.org.

If we can help with anything else, please email us at plosone@plos.org.

Thank you for submitting your work to PLOS ONE and supporting open access.

Kind regards,

PLOS ONE Editorial Office Staff

on behalf of

Dr. Vitor Hugo Rodrigues Paiva

Academic Editor

PLOS ONE