MARINE ECOLOGY PROGRESS SERIESMar Ecol Prog Ser

Vol. 513: 225237, 2014doi: 10.3354/meps10991

Published October 22

INTRODUCTION

Following large oil spills at sea, seabirds are partic-ularly sensitive to both internal and external oil expo-sure (Leighton 1993), and their foraging habits,preening behavior, and resting requirements lead tofrequent contact with surface oil. Whereas proximateexposure, cause-of-death, and pathologies for indi-vidual birds can be directly examined (Balseiro et al.2005), population-level effects must be approximatedindirectly (Wilhelm et al. 2007).

Bird mortality resulting from oil spills has usuallybeen estimated by tallying recovered carcasses,

assessing the probability of recovery, and using anassumed expansion factor to statistically extrapolatethe carcass tally to an estimate of total mortality(Wiese & Robertson 2004, Haney et al. 2014, this volume). However, when mortality occurs far off-shore, winds and currents, scavenging by consumerspecies, and sinking from loss of buoyancy duringdecomposition combine to greatly diminish carcassrecoveries (Wiese 2003, Munilla et al. 2011).

The Deepwater Horizon blowout began on 20April 2010 and discharged liquid and gaseoushydrocarbons continuously into the Gulf of Mexicoover the next 86 d. Before the seafloor wellhead was

The authors 2014. Open Access under Creative Commons byAttribution Licence. Use, distribution and reproduction are un -restricted. Authors and original publication must be credited.

Publisher: Inter-Research www.int-res.com

*Corresponding author: chaney@defenders.org

Bird mortality from the Deepwater Horizon oil spill.I. Exposure probability in the offshore Gulf of

Mexico

J. Christopher Haney1,4,*, Harold J. Geiger2, Jeffrey W. Short3

1Terra Mar Applied Sciences LLC, 123 W. Nye Lane, Suite 129, Carson City, NV 89706, USA2St. Hubert Research Group, 222 Seward, Suite 205, Juneau, AK 99801, USA

3JWS Consulting LLC, 19315 Glacier Highway, Juneau, AK 99801, USA4Present address: Defenders of Wildlife, 1130 17th Street, NW, Washington DC 20036, USA

ABSTRACT: Following the 2010 Deepwater Horizon MC 252 blowout in the Gulf of Mexico, mostsurface oil remained more than 40 km offshore, precluding reliable estimation of offshore avianmortality based on shoreline counts. Using an exposure probability model as an alternativeapproach, we estimated that between 36 000 to 670 000 birds died in the offshore Gulf of Mexicoas result of exposure to oil from the Deepwater Horizon, with the most likely number near 200 000.Our exposure probability model is a technique for estimating this offshore component of avianmortality as the product of the oil slick area, the density of the birds above the oil slick, and theproportionate mortality of birds that could be exposed to oil during an assumed exposure period.The duration of the exposure period is treated as an estimated parameter to account for oil slickmovement, exposure of birds immigrating to the oil-contaminated area, and re-exposure of birdsthat survived prior vulnerability to exposure. Total avian mortality is determined as the sum ofmortalities from each exposure period. Exposure probability may be the only method available toestimate bird mortality from large, remote oil spills in the open ocean where carcasses are unlikelyto ever reach shore. In the case of the Deepwater Horizon, the uncertainty interval is quite largebecause several parameters could not be well estimated. Historically sparse survey coverageeffectively led to an under-appreciation of the effects of this spill on marine birds.

KEY WORDS: Avian mortality Exposure probability Oil spill Deepwater Horizon Gulf of Mexico Offshore habitat Marine birds

OPENPEN ACCESSCCESS

Mar Ecol Prog Ser 513: 225237, 2014

capped on 15 July 2010, the Deepwater Horizon haddischarged about 6.7 108 l of liquid petroleum intothe Gulf, of which about 3.3 108 l reached the sea surface (McNutt et al. 2012), equivalent to about 8Exxon Valdez spills. Extensive satellite and otherrecon naissance detected a massive surface oil slickin the north central Gulf of Mexico (Fig. 1) thatspread over the outer continental shelf, continentalslope, and abyssal plain. Frontal eddies of the LoopCurrent (Vukovich 1995) entrained some oil fromthe well towards southwestern Florida (Liu et al.2011), with smaller, isolated patches of surface oiladvected toward the Yucatan coast of Mexico(NOAA 2013).

Climatic and oceanographic conditions during theDeepwater Horizon blowout acted to suppress shore -line deposition of bird carcasses along the coastline(Haney et al. 2014). Outflow by the Mississippi Riverimpeded surface oil and other drifting objects fromreaching shorelines during the early phase of oil dis-charge (Kourafalou & Androulidakis 2013). Othersurface oil was advected by Loop Current frontaleddies even further away from the coast (Liu et al.2011). Wind-driven transport of oil toward the shore-

line was negligible except near the immediate coast(Huntley et al. 2011). Because more than 80% of thecumulative oil slick occurred 40 km or more offshore,an alternative to carcass recovery is needed to esti-mate avian mortality in most of the spill zone.

We applied different methods to estimate seabirdmortality from the blowout, de pending on distancefrom the shoreline, for several reasons. The offshoreseabird community consists almost entirely of aerialforagers, whereas the community closer to shore in -cludes surface foraging seabirds. Average aerial sea-bird densities are lower offshore (Tasker & Mustoe2003), but increase substantially approaching thecoast within 40 km (McFarlane & Lester 2005), so theaverage density of the offshore community may beconsidered nearly isotropic with respect to horizontaldirection, but the community within 40 km of theshore may not. Finally, in the northern Gulf of Mexico, seabird mortality within 40 km of the shoremay be estimated based on recoveries of oiled car-casses from shorelines, whereas seabirds killed morethan 40 km offshore have negligible probability ofreaching the coastline (Haney et al. 2014), and hencea different approach must be used.

226

90 W 85

30N

25

Fig. 1. Total duration of surface oil from the 2010 Deepwater Horizon MC 252 blowout in the northern Gulf of Mexico. Durationindicates the number of days that oil slicks were detectable by satellites. Duration does not necessarily imply continuous oilpresence at any location. Oil presence was analyzed with the Textural Classifier Neural Network Algorithm (TCNNA) for syn-thetic aperture radar (SAR; see also 'Data sources for surface oil' and 'Oil surface area measurement' in Supplement at

www.int-res. com/ articles/ suppl/ m513 p225_ supp. pdf) from NOAA (2013)

http://www.int-res.com/articles/suppl/m513p225_supp.pdf

Haney et al.: Bird mortality from Deepwater Horizon. I. Offshore

Here we develop an estimate of the offshore birdmortality caused by the blowout, along with the associated uncertainty using an ex posure probabil-ity model. This model is based on 3 parameters (oilslick size, bird density, and proportionate mortality)combined with a temporal expansion factor for theeffective duration of exposure for a cohort ofexposed birds. We estimate uncertainty by assigningproba bility distributions to the para meters to reflectalternative assumptions about the para meter values.We then use Monte Carlo simulation to assess theconsequences of these assumptions. The resultingframework provides an alternative ap proach forestimating seabird mortality at remote, open oceanoil spills.

METHODS

Study area and modeling domain

We considered the acute mortality phase to last103 d, from 20 April until 31 July 2010, to account forbird mortality from contact with lingering surface oilafter the wellhead was capped on 15 July (Aeppli etal. 2012). Oil slicks were detectable by satellite sen-sors beginning 22 April 2010 (Tables S1 & S2 in theSupplement at www.int-res.com/articles/suppl/m513p225_supp.pdf), but sea birds were exposed to oil assoon as it surfaced above the wellhead. Avian mor-tality clearly continued after the well was cappedbased on Wildlife Collection Reports that revealedan increasing ratio of dead-to-live bird recoveriesthrough late July 2010 (Belanger et al. 2010). In thepresent paper, we limit the scope of investigation towaters 40 km or more offshore (Fig. 1).

Exposure probability model

We assumed that the number of birds killed,denoted as N, can be approximated as the product of3 factors: oil slick surface area (A), seabird densityabove the slick (D), and a proportionate mortality (M):

NI = ADM (1)

where NI is the initial or conceptual approximation,prior to accounting for the temporal dynamics ofthe spilled oil. Eq. (1) is based on the fundamentalassumption that all birds over an oil slick during asufficient period of time (i.e. exposure period) wouldeventually be exposed to oil (Fifield et al. 2009), and

that birds flying above the uncontaminated sea sur-face in the same period of time will remain unaf-fected. Furthermore, we assume that some propor-tion M of the birds that would be exposed during thisperiod will die as a result of oil contact (Wilhelm et al.2007).

To estimate total mortality resulting from a pro-tracted oil discharge event such as the DeepwaterHorizon, we partitioned the assumed 103 d durationof the oil slick, denoted here as T, into a series of con-secutive exposure periods, denoted as P, and applyEq. (1) to each period. We treat the exposure period Pas an adjustable parameter to account for the time required to replenish birds killed by oil, either byslick movement to areas not previously oiled or notrecently oiled, or by birds that immigrate into oiledareas. Birds that immigrate into oiled areas mayinclude previously un-exposed birds and birds thatsurvived past vulnerability to exposure in immedi-ately adjacent areas currently un-oiled, or birds thatarrive to the region through longer-distance seasonalmigrations. We use the notation P to denote a vari-able describing an interval of time. We use P todenote a specific number of days for a particularexposure period. Symbolically there are T/P totalperiods and we index these with p = 1, 2, T/P.Parameters may vary by period, so that Ap denotesthe oil slick surface area during period p. Then, usingEq. (1) for each individual period, the second ap -proximation (denoted with the II subscript) of total number of birds killed by the oil spill can be foundwith summation:

(2)

We have assumed that the population exposedwithin an exposure period declines by proportion Mpby the end of the period, after which the population isrestored to a new density of birds, Dp +1, exposed dur-ing the succeeding period as a result of slick move-ment, bird immigration and survival. Varying theexposure period duration P then allows us to assessalternative assumptions regarding the time necessaryto replenish the density of birds above the oil slick.

The next simplifying assumptions we make arethat the exposure periods have equal durations P,that the proportionate mortality M is the same foreach period, and that the initial seabird density Daveraged over the oiled area for exposure period pis approximately constant for all exposure periods asa result of slick movement and bird immigration tooiled areas over the duration of the exposure period.If M and D are assumed constant, they may be fac-tored out of the summation in Eq. (2) to give

1

/

1

/N N A D MII p p p pp

T P

p

T P = = ==

227

http://www.int-res.com/articles/suppl/m513p225_supp.pdfhttp://www.int-res.com/articles/suppl/m513p225_supp.pdf

Mar Ecol Prog Ser 513: 225237, 2014

(3)

If we denote the average spatial extent of the oil slick as

(4)

then the number of bird deaths is found as the pro -duct of 3 parameters to describe the average deathsper period and the number of exposure periods:

(5)

Each parameter (A, D, M, and P) can be thought of as

an unknown random variable (i.e. in the Bayesiansense), and each random variable could be given aprobability distribution to represent alternative as -sump tions about values of the parameter. At othertimes we will view the problem in a sampling context(i.e. in the sampling sense the parameters are fixedand data realizations are thought of as random),denoting the parameters specific fixed numericalestimates as A, D, M, and P 1. In this latter case, wewill multiply the estimates together to de velop a sin-gle estimate of bird mortality. When we think of theparameters as random, we mean that we will assignprobability distributions to each parameter in orderto explore the alternative values of N that wouldhave come about as the result of alternative assump-tions about the values of the individual parameters(Silver 2012).

Parameter estimates and probability distributions

Several estimates of oil slick area over time areavailable for the Deepwater Horizon blowout (e.g. Huet al. 2011). Based on a synthesis of this information(Fig. 2; see also Tables S1 & S2 in the Supplement),the estimated average slick size, A, is 15 000 km2. Forthe purpose of assessing uncertainty, we assignedthis parameter a gamma probability distribution withmean 15 000 and SD 2300 km; 2 SD are approximately30% of the mean slick size, which reflects classifica-tion errors associated with remote sensing estimatesof oil ex tent (Garcia-Pineda et al. 2009). A gammadistribution was assigned because it has an inherentlower bound of zero but is otherwise similar to thenormal distribution with the parameter values chosen.

Because direct measurements of bird density abovethe oil slick were not made contemporaneously during the spill, we used estimates of bird densityreported from prior surveys of the northern Gulf ofMexico during spring and summer. While actual den-sity is highly variable on small spatial or temporalscales, the variability of averages across large spa-

tial and temporal scales would be considerablyless. Tasker & Mustoe (2003) covered a total surveydistance of 1903 km in the same season, along theouter continental shelf and continental slope near theDeepwater Horizon well, and observed a total of 912birds from 23 species (all aerial foragers). Multipliedby a transect width of 300 m (Tasker et al. 1984), spatial coverage was therefore 570.9 km2, resultingin an estimated density of 1.6 birds km2 (unadjustedfor detectability; e.g. Barbraud & Thiebot 2009). In adifferent Gulf survey in the oiled area from the 1990s,Hess & Ribic (2000) recorded 769 birds in Augustwithin a strip transect 300 m wide by 1592 km long(= 477.6 km2) for an identical value of 1.6 birds km2.The estimated density of birds, D, was thereforeassumed to be 1.6 birds km2. Assuming that the trueaverage density of birds must have been substan-tially lower than the observed average density incoastal areas (3.69.4 birds km2; McFarlane & Lester2005), and assuming that the true average densityshould have a very low probability of being morethan twice the 1.6 birds km2 reported by Tasker &Mustoe (2003) or Hess & Ribic (2000), this densityparameter was assigned a gamma distribution witha mean of 1.6 and SD of 0.61 birds km2.

In some previous studies of offshore spill mortalityto seabirds, all diving birds on the ocean surfaceinside the perimeter of an oil slick were assumed tohave died (Wilhelm et al. 2007), or if bird flight vec-tors intersected the slick within a 24 h period, all suchflying birds were also assumed to have died (Fifield

N ADM TP( )( )=

1

/A P

TApp

T P= =

1

/

1

/N DMA DM AIII pp

T Ppp

T P = == =

228

15-Apr 15-May 14-Jun 14-Jul

Sur

face

are

a (1

03 k

m2)

Date

0

10

20

30

40

50

60

70

80

90

Fig. 2. Solid line: extent (in km2) of the surface oil slick (wa-ters 40 km offshore) in the northern Gulf of Mexico oiledzone during the Deepwater Horizon MC 252 blowout (in-cludes days with no satellite coverage); dashed line: averageslick size used in the model; dotted lines: 2.5th and 97.5thpercentiles of a probability distribution for this average.Satellite measurements of slick extent were augmented withoil spill trajectory models and other ancillary data. See

Table S2 in the Supplement for details

Haney et al.: Bird mortality from Deepwater Horizon. I. Offshore

et al. 2009). Seabird vulnerability tosurface oil, however, can vary as afunction of the species time spent onthe sea surface (Williams et al. 1995).Although we assumed all birds overthe oil slick could have been exposed,we considered only some proportion ofthese birds, M, as having had contactsufficient to lead to death from oilexposure (whether via ingestion, res-piration, or plumage contamination).

Because many of the bird speciesaffected were storm-petrels, frigate-birds, tropicbirds, and pelagic terns,all of which spend a high proportionof time in the air (Table 1), we rea-soned that oiling mortality rates wouldbe lower for these birds than it wouldbe for species that spent more of theirtime resting on the sea surface. Pub-lished estimates of oiling mortalityrates for all seabird species, regard-less of foraging styles and from alllocations, range from 5 to 100%(Camphuysen & Heubeck 2001, Wil -liams et al. 1995). However, whenrestricted to those seabirds with aerialforaging styles, the range narrowsto 2289% (Camphuysen & Heu beck2001), including birds whose deathsin northern climates would be proxi-mately caused by thermal stress.Assuming that thermal stress wouldbe lower and that highly aerial sea-birds spend less time in surface con-tact in the Gulf of Mexico, we chose avalue of 0.33 as a reasonable estimateof proportionate mortality, M, for off-shore birds affected by the blowout.To represent un certainty, M was as -signed a beta distribution with a meanof 0.33 and SD of 0.15 so that M wouldrarely exceed 0.65 in simulation.

Based on oil slick movement, localbird movements, long-distance mi -gra tions, and weather changes (seethe Supplement: Migration and re -placement of seabirds over the spillzone), we estimated time to restoredensity following the loss of birds dueto mortality as likely between 1 and7 d. We further as sumed density ofbirds returned to the initial level of D

229

Species Common name Breeding Abundanceorigin H&R T&M

Calonectris diomedea Corys shearwater ONWA 10 1Puffinus gravis Great shearwatera ONWA 3 0Puffinus griseus Sooty shearwater ONWA 1 0Puffinus puffinus Manx shearwatera ONWA 5 0Puffinus lherminieri Audubons shearwatera WI/C/B 154 14Calonectris or Puffinus Shearwater speciesa WI/C/B 21 14Oceanites oceanicus Wilsons storm-petrel ONWA 10 50Oceanodroma leucorhoa Leachs storm-petrela NWA 1 4Oceanodroma castro Band-rumped storm- ONWA 250 189

petrelOceanodroma or Storm-petrel species ONWA 62 208Oceanites

Phaethon aethereus Red-billed tropicbird WI/C/B 2 0Phaethon spp. Tropicbird species WI/C/B 0 2Fregata magnificens Magnificent frigatebirda NGOM 4 13Fregata spp. Frigatebird speciesa NGOM 174 0Sula dactylatra Masked boobya SGOM 4 11Sula spp. Sulid species SGOM 0 3Pelecanus occidentalis Brown pelicana NGOM 0 2Phalaropus spp. Phalarope species ONWA 22 0Leucophaeus atricilla Laughing gulla NGOM 38 31Leucophaeus pipixcan Franklins gull ONWA 0 1Larus delawarensis Ring-billed gulla NWA 0 1Larus fuscus Lesser black-backed NWA 0 1

gulla

Larus spp. Gull speciesa NGOM 2 0Anous stolidus Brown noddy SGOM 0 16Onychoprion fuscatus Sooty terna SGOM 111 2Onychoprion anaethetus Bridled tern WI/C/B 70 75Sternula antillarum Least terna NGOM 1 0Chlidonias niger Black terna ONWA 1119 174Sterna dougallii Roseate tern SGOM 0 1Sterna hirundo Common terna NGOM 4 6Sterna paradisaea Arctic tern ONWA 2 0Sterna forsteri Forsters terna NGOM 0 1Thalasseus maximus Royal terna NGOM 19 14Thalasseus sandvicensis Sandwich terna NGOM 24 15Onychoprion and Tern speciesa NGOM 179 59Sterna spp.

Stercorarius pomarinus Pomarine jaeger ONWA 14 3Stercorarius parasiticus Parasitic jaeger ONWA 1 0Stercorarius longicaudus Long-tailed jaeger ONWA 2 0Stercorarius spp. Jaeger species ONWA 1 1

Total individuals 2310 912Percent aerial foragers 99.1 100Percent breeding outside northern Gulf of Mexico 81.5 84.5Percent breeding outside entire Gulf of Mexico 76.8 80.9aSpecies documented as visibly oiled during the Deepwater Horizon blowout

Table 1. Species composition, breeding origin, and relative abundances (total numbersobserved) of marine birds characteristic of May, June, and August in the offshore Gulfof Mexico within the oiled zone; H&R: Hess & Ribic (2000); T&M: Tasker & Mustoe(2003). Breeding originsONWA: outside the northwest Atlantic Ocean; NWA: north-west Atlantic Ocean; WI/C/B: West Indies, Caribbean Sea, or Bahamas; NGOM: north-ern Gulf of Mexico (in or near spill zone); SGOM: southern Gulf of Mexico (outside thespill zone). Totals in H&R included birds tallied both inside and outside the 300 m striptransect used to derive a density estimate. Aerial foraging behavior characterized birdsthat rely extensively on flight between exploited feeding patches (e.g. shearwaters) orthat use aerial feeding techniques (Nelson 1979), including surface plunging (e.g. boo-bies, some terns), aerial dipping (e.g. some terns), aerial pursuit (e.g. frigatebirds),skimming and hydroplaning (e.g. gulls, storm-petrels). Surface foraging behavior char-acterized birds that use prolonged pursuit diving (e.g. loons, sea ducks) and surfaceseizing (e.g. phalaropes). Taxonomy after American Orni thologists Union checklist of

North American birds (http://checklist.aou.org/taxa/; accessed 29 June 2014)

Mar Ecol Prog Ser 513: 225237, 2014

every 4 d, on average. These assumptions are sup-ported by Kinlan et al. (2012, p. 50), who reported onoffshore surveys of Atlantic birds and concluded that bird abundance de-correlates rapidly with time inthis region, supporting our assumption that repeatsurveys are approximately independent as long asthey are separated by a few days or longer. Un -certainty in P was then assessed by assigning thisparameter a uniform distribution on the interval 2 to6 d (SD approximately 1.15 d).

After assigning probability distri butions to theparameters, we drew 1 million random numbers fromthe assigned distributions (Table 2), and then werepeatedly used Eq. (5) to compute bird deaths. Thisgave a Monte Carlo probability dis tri bution for thenumber of bird deaths (rounded to no more than 2significant digits). From this distribution we con-structed what we call an uncertainty interval orMonte Carlo simulation interval. Our intervals arenot confidence intervals, because our simulations arenot an attempt to estimate a sampling distribution.We have chosen not to call them credible intervalsto stress that they were constructed from the priordistribution of the parametersand not from a pos-terior distribution.

To examine sensitivity of the overall uncertainty tocontributions to the variability of each individualparameter, each parameter in sequence was re -placed with its 2.5th and 97.5th percentile, while theother 3 parameters were fixed at their mean values.Although we call this a sensitivity analysis, the tradi-tional notion of sensitivity is intertwined with theprobability distribution for each parameter. The devi-ation, expressed as a range that we will call the devi-ation range, is the result. The cumulative deviationrange was computed as the sum of the individualdeviation ranges for each parameter. Sensitivity ofthe overall mortality estimate to each parameter was

expressed as a proportion of this cumulative devia-tion range, computed as the ratio of each parametersdeviation range and the cumulative deviation range.Taken together, these ratios indicate where newinformation could be most helpful in improving themortality estimate.

RESULTS

Mortality estimate with exposure probability

Regardless of the exposure period duration P chosen, the estimated average number of birds killedis given by the product of A, D, and M (= 15 000 km2

1.6 birds km2 0.33), or approximately 7900 birdsperiod1. For any period of length P days over the103 d duration of the slick, the estimated number ofbird deaths would be approximately 7900 103/P(= 820 000/P), such that N = 410 000 for P = 2, N =270 000 for P = 3, etc. Focusing on the value of P thatwe consider is most likely to represent the time toreplenish the bird density above the oil slicks to D(i.e. 4 d), the estimated number of bird deaths wouldbe approximately 200 000, after rounding to 2 sig -nificant figures to reflect the inherent overall im -precision of this estimate.

Using the named probability distributions to as -sess the effect of alternative parameter assumptions(Table 2), the central 95% of the probability was overthe interval 36 000 to 670 000 bird deaths, where asthe central 80% of the simulated probability coveredthe interval from 68 000 to 440 000 deaths. The prob-ability that bird deaths exceeded 97 000 is approxi-mately 80%, and the probability that this mortalityexceeded 120 000 is approximately 70%. The meanand median of the distribution were 230 000 and180 000, indicating a prob ability distribution skewedtoward larger values (Fig. 3).

Model sensitivity to parameters

The estimates of total bird deaths in the offshoreGulf of Mexico were most sensitive to the propor -tionate mortality, M (34.2% of cumulative deviationrange; Fig. 4). Similarly, the final mortality estimateswere sensitive to assumptions about the parametervalues for bird density (29.3% of cumulative devia-tion range) and exposure period (24.5%). The mor-tality estimate was least sensitive to the parameterthat was measured during the spill, the average oilslick area,

A (12.0%).

230

Parameter Assigned Mean SD 2.5th 97.5th distribution quantile quantile

A gamma 15000 2300 11000 20000D gamma 1.6 0.606 0.64 2.99M beta 0.33 0.15 0.09 0.65P uniform 4 1.15 2.1 5.9

Table 2. Probability distributions used to simulate the num-ber of offshore bird deaths caused by the Deepwater Horizon blowout.

A: average oil slick area (km2); D: average

bird density (birds km2); M: proportionate mortalityfor birds; P: exposure period needed for the bird density

over the oiled water to return to the initial value of D

Haney et al.: Bird mortality from Deepwater Horizon. I. Offshore

DISCUSSION

Based on our simulation, we found a very highprobability that between 36 000 and 670 000 birdsdied in the offshore waters of the Gulf of Mexico dueto the Deepwater Horizon blowout. A number of200 000 is the most reasonable single estimate for thisoffshore area. This estimate applies only to the acutedischarge phase of the blowout, which extended for103 d. Additional bird mortality not considered in ourestimates continued to be reported months after thewell was capped (Belanger et al. 2010).

Loss of 200 000 birds appears reasonable in com-parison with the number of birds in the region.The cumulative area of the oil slick more than40 km offshore was 100 000 km2 (NOAA 2013;see also Table S2 in the Supplement). With a den-sity of 1.6 birds km2, there would have been morethan 160 000 birds in total above the cumulative oilslick footprint, implying aggregated mortality ofessentially all of these birds plus others suppliedby migration. The upper bound of our uncertaintyinterval (670 000 birds) implies the death of allbirds above the cumulative oil slick area morethan 4 times over, requiring replacement of killedbirds through immigration at rates that ap proachimplausibility. Conversely, our lower bound (36 000birds) represents loss of less than 25% of birdsprojected over a large slick during a particularlylong-lasting discharge, which also approaches im -plausibility.

Given breeding population sizes at the speciesnearest origin, loss of 200 000 offshore birds (

Mar Ecol Prog Ser 513: 225237, 2014

Assumptions for exposure probability

Proportionate mortality of affected birds (M) con-tributed the most uncertainty to the overall mortalityestimate (Fig. 4). This parameter is also closely asso-ciated with the assumed period length (P) that corre-sponds to the time it takes for the density of birdsabove the oil slick to return to the assumed initialvalue through oil slick movement, local bird move-ment, and regional bird immigration. In both cases,these parameters were chosen without direct meas-urement from this oil spill, and had to be inferredfrom literature on bird migration, previous oil spills,and avian physiology.

We estimated bird density, D, with the value of1.6 birds km2 derived from observations across theoffshore Gulf of Mexico 7 yr or longer before theblowout. In the Terra Nova spill off southeasternCanada, bird density values collected at differenttimes or spatial scales did not greatly influence themortality estimates (cf. Fifield et al. 2009, Wilhelm etal. 2007). The value for D that we used is belowbird densities typically found in similar oligotrophicwaters elsewhere (Haney 1986). Except for localaggregation, bird density is characteristically low offthe southeastern USA. Additionally, our estimate wasderived without adjustments for de tection probabil-ity, a correction that when measured can raise appar-

ent density of birds surveyed by ~5 to 30% (Barbraud& Thiebot 2009).

The exposure probability method requires a rea -son able estimate of the average extent of the oilslick,

A. Yet measurements with satellite imagery

may have consistently under-estimated the surfaceextent of oil. Estimates of spatial extent depend onsensitivity of the particular satellite platform, influ-ence of cloud cover, and other factors related to thesensors spatial coverage. Estimates also varied withthe algorithms used to infer oil presence, thick ness, orconcentration (e.g. Leifer et al. 2012, Lindsley & Long2012). Dissolved hydrocarbons were invisible in sur-face waters (100 m) from thinner (to 0.1 m) oil sheens (Leiferet al. 2012). Underestimating oil slick size would leaddirectly to an underestimate of the number of birddeaths using Eq. (5).

By using assigned probability distributions foreach parameter, we were able to logically organize,weight, and present alternative assumptions (Figs. 3& 4). In each case, we considered a range of reason-able as sumptions about the parameters value. Sub-stantially different assumptions about parametervalues will lead to mortality estimates that are

232

Species Breeding Pop. size Notes Reference(s)origin (103 ind.)

Calonectris diomedea ONWA 6001200 Entire Atlantic Ocean population Brooke (2004), BirdLife Corys shearwater International (2014)

Puffinus gravis ONWA 15000 Entire Atlantic Ocean population Brooke (2004)Great shearwater

Puffinus lherminieri WI/C/B 60 Bahamas and West Indian popula- Mackin (2013)Audubons shearwater tions only

Oceanodroma leucorhoa NWA 400 Saint-Pierre and Miquelon, largest Lorme et al. (2012)Leachs storm-petrel North American breeding colony

Sula dactylatra SGOM 9 Campeche Bank, southern Gulf of Tunnell & Chapman (2000)Masked booby Mexico only

Leucophaeus atricilla NGOM 730 Estimated Gulf of Mexico popula- Haney et al. (2014)Laughing gull tion of breeders, non-breeding

adults, and sub-adults

Onychoprion fuscatus SGOM 96 Southwest Florida and Campeche Tunnell & Chapman (2000), Sooty tern Bank populations only Mackin & Lee (2014)

WI/C/B 576 Campeche Bank, USA, Bahamas, Tunnell & Chapman (2000), and West Indian populations Mackin & Lee (2014)

Chlidonias niger ONWA 100500 Total North American continental Heath et al. (2009)Black tern breeding population

Table 3. Breeding populations of offshore seabirds found affected by the 2010 Deepwater Horizon blowout. Total populationsizes are higher as a result of uncounted sub-adults, non-breeding adults, and adult breeders absent from the colony due toyear-round nesting (e.g. Tunnell & Chapman 2000). Unless stated otherwise, population size refers only to estimated breeders

from the source population(s) within or closest to the Gulf of Mexico (origin abbreviations as in Table 1)

Haney et al.: Bird mortality from Deepwater Horizon. I. Offshore

either so low or so high as to approach implausibility(cf. Fig. 3, Table 3).

The exposure probability approach is attractivebecause of its simplicity, which it achieves throughbroad assumptions. For example, we assumed thatbirds were neither attracted to nor repelled by sur-face oil. Birds are unlikely to avoid altogether an oilslick the size of the Deepwater Horizon (Fig. 1). Al -though some species may avoid small spills (Lorentsen& Anker-Nilssen 1993), observations elsewhere indi-cate no consistent spill avoidance behavior in sea-birds (French McCay & Rowe 2004). In 2 studies dur-ing the 2009 Montara blowout in the Timor Sea offnorthern Australia, seabirds were more abundantover oil slicks than in oil-free waters nearby (Mustoe2009, Watson et al. 2009). At smaller spatial scalesbirds could be attracted to surface slicks if buoyant,dark oil attracts seabird prey (see Watson et al. 2009),if weathered oil particles resemble small prey (Mus-toe 2009), or if thin oil sheens mimic convergenceslicks that attract dense foraging aggregations ofmarine birds.

Deepwater Horizon oil also presented 2 novelhazards to seabirds. Audubons shearwater and sev-eral other offshore Gulf seabirds are Sargassum-for-aging specialists (Haney 1986, Moser & Lee 2012).Therefore, risk to these seabirds increases at float-ing Sargassum when algal mats were coalesced bysurface processes that aggregate oil concurrently(cf. Carmichael et al. 2012, Powers et al. 2013). Also,flames from in situ oil burning (FISG 2010) in -creased mortality risk because storm-petrels andother nocturnal-feeding pelagic seabirds may beattracted to the light (Le Corre et al. 2002, Monte -vecchi 2006) and may then die in the flames, smoke,or residual oil.

We further assumed that proportionate mortalitywas independent of bird density. This assumptionmay not hold if seabirds are curious about oil-debilitated birds and approach them. In such cases,probability of oil exposure is greater than as sumed,implying that we underestimated offshore bird mor-tality. Because patterns of local seabird aggregation(e.g. Beauchamp 2011) as well as small-scale spatialprocesses that govern individual oil slicks (Chris-tensen & Terrile 2009) likely influenced the relation-ship between D and M in ways not captured by ourexposure probability model, we recommend futureresearch aimed at clarifying the key assumption ofindependence between D and M in exposure proba-bility models. This could be done by characterizingseabird behavior around individual oil patches dur-ing particularly large or long-lasting spills.

Probable fate of bird carcasses

An obvious question is, where did 200 000 carcassesgo if that many birds died? Carcasses were not sys-tematically surveyed in the offshore Gulf of Mexico;fewer than 30 carcasses were retrieved >40 km fromland, and all of these carcasses were recovered inci-dentally (Dead Bird Map for week of 14 December2010, Bird Recovery Map Archive, Bird Impact Data,www.fws.gov/home/dhoilspill/ collectionreports. html;accessed 25 Feb 2013).

Few carcasses could have persisted long in warmwaters at these latitudes because decomposition leadsto loss of carcass buoyancy and subsequent sinking.Decomposition in the Gulf of Mexico is likely to bemore rapid than the 5 to 10 d in cooler seas (Wood 1996,Wiese 2003). Carcasses also would have dis appearedquickly from opportunistic scavenging by other marineconsumers (Lowe et al. 1996) such as tiger sharks Ga-leocerdo cuvier (Kaufman 2012). Even partial scav-enging accelerates carcass sinking (Wiese 2003).

Carcasses were also destroyed by widespread insitu burning of surface oil from early May to mid July2010, which totaled 4.2 107 l, equivalent to oneExxon Valdez spill (FISG 2010). Most burns tookplace along current- and wind-driven convergencelines where buoyant oil was sufficiently concentratedfor ignition (FISG 2010). Such convergences arelikely to also aggregate and sink floating carcasses(Barstow 1983). Additional bird carcasses may havedisappeared from oil skimming operations.

It may seem unlikely that blowout responders wouldoverlook large numbers of bird carcasses, but birdswere not exposed simultaneously and so did not dieall at once or in one place. Aerially foraging birdstypical of the offshore Gulf (Table 1) likely dispersedfrom points of initial exposure. Deaths of birds fromcrude petroleum and weathered by-products can bedelayed (Nevins & Carter 2003). Following the ExxonValdez spill, >80% of dead birds were recoveredoutside the immediate spill zone (Piatt et al. 1990).

Even had all mortality occurred instantaneously,carcasses would be dispersed across an area greaterthan the surface area of Portugal at an average density of about 2 dead birds km2. Individual birdswould be exceedingly difficult to observe on theocean surface at appreciable distance (but see Hyren -bach et al. 2001). Furthermore, marine birds in thisoffshore region comprise species that are small,dark-plumaged, and very difficult to discern from a moving ship even when alive and active. Dark storm-petrels and black terns account for nearly 60% of survey observations for the offshore Gulf (Table 1).

233

Mar Ecol Prog Ser 513: 225237, 2014

Alternatives for estimating avian mortality at marine spills

In our companion article (Haney et al. 2014), weestimated bird mortality caused by the DeepwaterHorizon in the coastal waters of the northern Gulf ofMexico using both the exposure probability modeland a carcass deposition model. Results of these 2modeling approaches produced broadly overlappingestimates, with largely independent data sources.Furthermore, a ~60% decrease of the northern Gulfof Mexico population of the laughing gull Leuco -phaeus atricilla was evident in National AudubonSociety Christmas Bird Counts (NAS 2010), in broadagreement with the 36% decline of this species pro-jected on the basis of a carcass deposition model(Haney et al. 2014). This supports the exposure prob -ability approach we applied offshore.

Prior to the Deepwater Horizon blowout, an expo-sure probability approach had been applied to amuch shorter, smaller spill (~1.7 105 l), the TerraNova incident, covering 793 km2 over 6 d on theGrand Banks of eastern Canada (Wilhelm et al.2007); avian mortality estimated by exposure proba-bility gave similar results to an estimate interpolatedfrom a regression of bird mortality on spill volume.Moreover, this estimate was confirmed when some ofthe initial assumptions in the exposure probabilitymodel were validated by measuring seabird move-ments (Fifield et al. 2009).

In contrast to the exposure probability model,carcass sampling requires extensive field effort(Wiese 2003, Byrd et al. 2009). With less data-col-lection cost and effort, exposure probability modelscould provide estimates of comparable accuracyand precision to carcass sampling, but only if com-prehensive surface maps of oil slick area and sea-bird density estimates are available from surveycoverage of adequate resolution (e.g. Begg et al.1997).

Alternatively, measuring differences between pre-and post-spill colony attendance in coastal bird populations (Piatt & Ford 1996) can be used toinfer mortality after large marine oil spills. Thisapproach is not feasible for estimating avian mor-tality in offshore systems such as the Gulf of Mexico because virtually all bird species breed inhighly dispersed colonies far outside the spill zone(Tables 1 & 3; see also Fig. 3 in the Supplement).In summary, for large, offshore, deep-water oilspills, an exposure probability approach is the onlyfeasible method currently available for estimatingbird mortality.

Spill mortality in warm seas

With increases anticipated in offshore, deep-wateroil exploration (Pettingill & Weimer 2002), massive oildischarges such as the Deepwater Horizon spill arelikely to recur. The effects of offshore oil dischargeson ocean biota can easily go undetected (Williams etal. 2011) due to remoteness, unfavorable environmen-tal conditions, and logistical constraints on researchersimposed by spill response.

Except for the Gulf of Mexicos 1979 Ixtoc I (Jer -nelv & Lindn 1981) and the Australian 2009 Mon -tara blowouts (Watson et al. 2009), no other large oildischarges have affected warm-ocean assemblages ofmarine birds. Most post-spill assessments to date havebeen for avifauna in cool temperate seas (Balseiro etal. 2005, Oppel et al. 2009). Despite their aerial forag-ing habits, shearwaters, storm-petrels, boobies, andtropical terns (the same taxa found in the Gulf of Mex-ico; Table 1) were frequently ob served resting on andfeeding over weathered oil and sheen in the TimorSea after the Montara blowout (Mustoe 2009). Thesebirds and their prey were more closely associatedwith contaminated waters than with oil-free waters,with more than 80% of all encounters for some speciesoccurring over oil sheen (cf. Watson et al. 2009,OHara & Morandin 2010). Among birds exposed tooil and recovered, the proportionate mortality reached58 to 76% (Brassington & King 2010, Short 2011).

Despite protracted oil pollution in the Gulf of Mex-ico (Burgherr 2007), we know little about how eitherchronic or acute oiling affect the regions aerially for-aging seabird assemblage. Given the extent of thehydrocarbon industrys operations in the Gulf ofMexico (>4000 offshore oil production platforms;Dismukes 2010), the absence of region-wide, base-line seabird surveys prior to the Deepwater Horizonspill is of grave concern. Indeed, in no other region ofthe USA has the marine avifauna been so ineffectu-ally studied in conjunction with exploration anddevelopment of offshore energy.

Our ability to estimate bird mortality was limited bythe resolution of seabird surveys and a lack of directmeasurements of the mortality processes. These limi-tations are reflected in our large measures of uncer-tainty (Figs. 3 & 4). In order to reduce uncertainty andimprove assessment of avian mortality from futurespills, we strongly recommend (1) surveys that accu-rately and precisely measure seabird density and sea-sonal occurrence across the entire Gulf of Mexico,and (2) applied research on avian behavior at oilpatches, lethal dosages, and physiological responses(including effects of hydrocarbon inhalation).

234

Haney et al.: Bird mortality from Deepwater Horizon. I. Offshore

Acknowledgements. This study was funded jointly by TheMurray Firm and by Cossich, Sumich, Parsiola and TaylorLLC. Findings in this manuscript reflect those of the authorsonly; interpretations do not reflect positions that may beheld by any organization, entity, or other interest. All con-tent analyzed and reported here was available in the publicdomain. No data, information, documents, findings or anyother proprietary content protected by any confidentialityrestriction or agreement, including those pertaining to theNatural Resource Damage Assessment conducted for theDeepwater Horizon Mississippi Canyon 252 oil spill under61 Fed. Reg. 440, the Oil Pollution Act of 1990, were con-sulted or otherwise used in preparation of this manuscript.We thank Mark Tasker and Simon Mustoe for permission touse information from an unpublished report on marine birdssurveyed in the northern Gulf of Mexico. Steven C. Heinl,Charles H. Peterson, Steffen Oppel, Terrance J. Quinn II,Robert B. Spies, and an anonymous reviewer all made con-structive suggestions on earlier versions of this paper. DanThornhill reviewed and provided helpful suggestions on theanalyses. Anderson Shepherd, Jeff Lerner, and JenniferAllen assisted with illustrations.

LITERATURE CITED

Aeppli C, Carmichael CA, Nelson RK, Lemkau KL and oth-ers (2012) Oil weathering after the Deepwater Horizondisaster led to the formation of oxygenated residues.Environ Sci Technol 46: 87998807

Balseiro A, Esp A, Mrquez I, Prez V, Ferreras MC, GarcaMarn JF, Prieto JM (2005) Pathological features in mar-ine birds affected by the Prestiges oil spill in the north ofSpain. J Wildl Dis 41: 371378

Barbraud C, Thiebot JB (2009) On the importance of esti-mating detection probabilities from at-sea surveys of flying seabirds. J Avian Biol 40: 584590

Barstow SF (1983) The ecology of Langmuir circulation: areview. Mar Environ Res 9: 211236

Beauchamp G (2011) Fit of aggregation models to the distri-bution of group sizes in Northwest Atlantic seabirds. MarEcol Prog Ser 425: 261268

Begg GS, Reid JB, Tasker ML, Webb A (1997) Assessing thevulnerability of seabirds to oil pollution: sensitivity tospatial scale. Waterbirds 20: 339352

Belanger M, Tan L, Askin N, Wittnich C (2010) Chronologiceffects of the Deepwater Horizon Gulf of Mexico oil spillon regional seabird casualties. J Mar Anim Ecol 3: 1014

BirdLife International (2014) IUCN Red List for birds.(www.birdlife.org; accessed 28 Jan 2014)

Brassington GB, King B (2010) Ocean nowcasting and fore-casting for the Montara oil spill. Am Geophys Union,Ocean Sciences 2010, Portland, OR (www.cawcr. gov.au/staff/gbb/presentations/agu_ocean_sciences_2010_brassington.pdf; accessed 18 Sep 2013)

Brooke MDeL (2004) Albatrosses and petrels across theworld. Oxford University Press, Oxford

Burgherr P (2007) In-depth analysis of accidental oil spillsfrom tankers in the context of global spill trends from allsources. J Hazard Mater 140: 245256

Byrd GV, Reynolds JH, Flint PL (2009) Persistence rates anddetection probabilities of bird carcasses on beaches ofUnalaska Island, Alaska, following the wreck of the M/VSelendang Ayu. Mar Ornithol 37: 197204

Camphuysen CJ, Heubeck M (2001) Marine oil pollution

and beached bird surveys: the development of a sensi-tive monitoring instrument. Environ Pollut 112: 443461

Camphuysen CJ, Leopold MF (2004) The Tricolor oil spill: characteristics of seabirds found oiled in the Nether-lands. Atl Seabirds 6: 109128

Camphuysen CJ, Chardine J, Frederiksen M, Nues M(2005) Review of the impacts of recent major oil spills onseabirds. Rep WG Seabird Ecology. ICES CM 2005/C05

Carmichael CA, Arey JS, Graham WM, Linn LJ, LemkauKL, Nelson RK, Reddy CM (2012) Floating oil-covereddebris from Deepwater Horizon: identification and appli-cation. Environ Res Lett 7: 015301

Christensen KH, Terrile E (2009) Drift and deformation of oilslicks due to surface waves. J Fluid Mech 620: 313332

Dismukes DE (2010) Fact book: Offshore oil and gas indus-try support sectors. OCS Study BOEMRE 2010-042. USDept of the Interior, Bureau of Ocean Energy Manage-ment, Regulation and Enforcement, Gulf of Mexico OCSRegion, New Orleans, LA

Fifield DA, Baker KD, Byrne R, Robertson GJ and others(2009) Modelling seabird oil spill mortality using flightand swim behaviour. Environ Stud Res Fund Rep 186,Can Wildl Ser

FISG (Federal Interagency Solutions Group) (2010) Oilbudget calculator, Deepwater Horizon: a report to theNational Incident Command. Tech Doc, Oil Budget Sci-ence and Engineering Team (www. restorethegulf. gov/sites/default/files/documents/pdf/OilBudgetCalc_Full_HQ-Print_111110.pdf; accessed 17 Sep 2014)

French McCay D, Rowe JJ (2004) Evaluation of bird impactsin historical oil spill cases using the SIMAP oil spillmodel. In: Proc 27th Arctic Marine Oil Spill Program(AMOP) Tech Seminar. Emergency Sciences Division,Environment Canada, Ottawa, ON, p 421452

Garcia-Pineda O, Zimmer B, Howard M, Pichel W, Li X,MacDonald IR (2009) Using SAR images to delineateocean oil slicks with a texture-classifying neural networkalgorithm (TCNNA). Can J Remote Sens 35: 411421

Haney JC (1986) Seabird patchiness in tropical oceanic waters:the influence of Sargassum reefs. Auk 103: 141151

Haney JC, Geiger HJ, Short JW (2014) Bird mortality fromthe Deepwater Horizon oil spill. II. Carcass sampling andexposure probability in the coastal Gulf of Mexico. MarEcol Prog Ser 513: 239252

Heath SR, Dunn EH, Agro DJ (2009) Black tern (Chlidoniasniger). In: Poole A (ed) The birds of North Americaonline. Cornell Lab of Ornithology, Ithaca, NY (http://bna.birds.cornell.edu/bna/species/147; accessed 26 Feb2013

Hess NA, Ribic CA (2000) Seabird ecology In: Davis RW,Evans WE, Wrsig B (eds) Cetaceans, sea turtles andseabirds in the northern Gulf of Mexico: distribution,abundance and habitat associations, Vol II. Tech RepUSGS/BRD/CR-1999-0006 (OCS Study MMS 2000-003).US Dept of the Interior, Geological Survey, BiologicalResources Division, and Minerals Management Service,Gulf of Mexico OCS Region, New Orleans, LA, p 275316

Hu C, Weisberg RH, Liu Y, Zheng L and others (2011) Didthe northeastern Gulf of Mexico become greener afterthe Deepwater Horizon oil spill? Geophys Res Lett 38: L09601, doi: 10.1029/2011GL047184

Huntley HS, Lipphardt BL Jr, Kirwan AD Jr (2011) Surfacedrift predictions of the Deepwater Horizon spill: theLagrangian perspective. In: Liu Y, MacFadyen A, Ji ZG,Weisberg RH (eds) Monitoring and modeling the Deep-

235

http://dx.doi.org/10.1029/2011GL047184http://dx.doi.org/10.3354/meps10839http://dx.doi.org/10.5589/m09-035http://dx.doi.org/10.1017/S0022112008004606http://dx.doi.org/10.1088/1748-9326/7/1/015301http://dx.doi.org/10.1016/S0269-7491(00)00138-Xhttp://dx.doi.org/10.1016/j.jhazmat.2006.07.030http://dx.doi.org/10.2307/1521702http://dx.doi.org/10.3354/meps09022http://dx.doi.org/10.1016/0141-1136(83)90040-5http://dx.doi.org/10.1111/j.1600-048X.2009.04653.xhttp://dx.doi.org/10.7589/0090-3558-41.2.371http://dx.doi.org/10.1021/es3015138

Mar Ecol Prog Ser 513: 225237, 2014

water Horizon oil spill: a record-breaking enterprise.Geophys Monogr Ser 195. American Geophysical Union,Washington DC, p 179195

Hyrenbach K, Baduini CL, Hunt GL Jr (2001) Line transectestimates of short-tailed shearwater Puffinus tenuirostrismortality in the south-eastern Bering Sea, 19971999.Mar Ornithol 29: 1118

Jernelv A, Lindn O (1981) Ixtoc I: a case study of theworlds largest oil spill. Ambio 10: 299306

Kaufman R (2012) Sharks eating songbirds in Gulf of Mex-ico. (http: //news. nationalgeographic. com/ news/ 2012/ 01/120113-sharks-songbirds-gulf-of-mexico-animals-fish-science/; accessed 27 Feb 2013)

Kinlan BP, Zipkin EF, OConnell AF, Caldow C (2012) Statis-tical analyses to support guidelines for marine aviansampling: final report. OCS Study BOEM 2012-101,NOAA Tech Memo NOS NCCOS 158. US Departmentof the Interior, Bureau of Ocean Energy Management,Office of Renewable Energy Programs, Herndon, VA

Kourafalou VH, Androulidakis YS (2013) Influence of Mis-sissippi River induced circulation on the DeepwaterHorizon oil spill transport. J Geophys Res 118: 38233842

Le Corre M, Ollivier A, Ribes S, Jouventin P (2002) Light-induced mortality of petrels: a 4-year study from RunionIsland (Indian Ocean). Biol Conserv 105: 93102

Leifer I, Lehr WJ, Simecek-Beatty D, Bradley E and others(2012) State of the art satellite and airborne marine oilspill remote sensing: application to the BP DeepwaterHorizon oil spill. Remote Sens Environ 124: 185209

Leighton FA (1993) The toxicity of petroleum oils to birds.Environ Rev 1: 92103

Lindsley RD, Long DG (2012) Mapping surface oil extentfrom the Deepwater Horizon oil spill using ASCAT back -scatter. IEEE Trans Geosci Remote Sens 50: 25342541

Liu Y, Weisberg RH, Hu C, Kovach C, Reithmller R (2011)Evolution of the Loop Current system during the Deep-water Horizon oil spill event as observed with driftersand satellites. In: Liu Y, MacFadyen A, Ji ZG, WeisbergRH (eds) Monitoring and modeling the Deepwater Hori-zon oil spill: a record-breaking enterprise. GeophysMonogr Ser 195. American Geophysical Union, Wash-ington DC, p 91101

Liu Z, Liu J, Gardner WS, Shank GC, Nathaniel EO (2014)The impact of Deepwater Horizon oil spill on petroleumhydrocarbons in surface waters of the northern Gulfof Mexico. Deep-Sea Res II (in press) doi: 10.1016/ j. dsr 2.2014.01.013

Lorentsen SH, Anker-Nilssen T (1993) Behaviour and oilvulnerability of fulmars Fulmarus glacialis during an oilspill experiment in the Norwegian Sea. Mar Pollut Bull26: 144146

Lorme H, Delord K, Letrounel B, Barbraud C (2012) Popu-lation surveys of Leachs storm-petrels breeding atGrand Colombier Island, Saint-Pierre and Miquelonarchipelago. Wilson J Ornithol 124: 245252

Lowe CG, Wetherbee BM, Crow GL, Tester AL (1996) Onto-genetic dietary shifts and feeding behavior of the tigershark, Galeocerdo cuvier, in Hawaiian waters. EnvironBiol Fishes 47: 203211

Mackin WA (2013) Available breeding habitat and pop -ulation estimate for Audubons shearwater in the Caribbean. In: American Ornithologists Union, AnnualMeeting 2013, Chicago, IL, Poster no. 384 (http: //fieldmuseum.o rg/sites/default/files/2013AB_37.pdf; ac -cessed 5 Feb 2014)

Mackin WA, Lee DS (eds) (2014) West Indian Breeding Sea-bird Atlas (http://wicbirds.net/index.html; accessed 28Jan 2014)

McFarlane RW, Lester LJ (2005) Determination of nearshoreseabird density on the upper Texas coast. Use of Radarand Human Observation to Assess the Risk of Bird Mor-talities at Potential Wind Turbine Installations on theUpper Texas Coast, Final Report. State Energy Conser-vation Office, Austin, TX (www.seco.cpa.state. tx.us/ re/docs/re_wind_projects-bird.pdf; accessed 22 Feb 2013)

McNutt MK, Camilli R, Crone TJ, Guthrie GD and others(2012) Review of flow rate estimates of the DeepwaterHorizon oil spill. Proc Natl Acad Sci USA 109: 2026020267

Montevecchi WA (2006) Influences of artificial light on marine birds. In: Rich C, Longcore T (eds) Ecologicalconsequences of artificial night lighting. Island Press,Washington DC, p 94113

Moser ML, Lee DS (2012) Foraging over Sargassum by west-ern North Atlantic seabirds. Wilson J Ornithol 124: 6672

Munilla I, Arcos JM, Oro D, lvarez D, Leyenda PM,Velando A (2011) Mass mortality of seabirds in the after-math of the Prestige oil spill. Ecosphere 2: art83

Mustoe S (2009) Biodiversity survey of the Montara field oilleak. WWF Australia, Sydney, NSW (http://awsassets.wwf. org. au/ downloads/fs047.pdf; accessed 17 Sep 2014)

NAS (National Audubon Society) (2010) The Christmas BirdCount Historical Results [Online]. (http: //netapp. audubon.org/CBCObservation/Historical/ResultsBySpecies.aspx?1;accessed 12 Sep 2013)

Nelson JB (1979) Seabirds: their biology and ecology. A&WPublishers, New York, NY

Nevins HM, Carter HR (2003) Age and sex of common mur-res Uria aalge recovered during the 19971998 Point Reyestar ball incidents in central California. Mar Ornithol 31: 5158

NOAA (National Oceanic and Atmospheric Administration)(2013) ERMA Deepwater Gulf Response. NOAA, SilverSpring, MD (http: //response.restoration.noaa.gov/maps-and-spatial-data/environmental-response-management-application-erma/erma-gulf-response.html; accessed 10Feb 2013)

OHara PD, Morandin LA (2010) Effects of sheens associatedwith offshore oil and gas development on the feathermicrostructure of pelagic seabirds. Mar Pollut Bull 60: 672678

Oppel S, Dickson DL, Powell AN (2009) International impor-tance of the eastern Chukchi Sea as a staging area formigrating king eiders. Polar Biol 32: 775783

Peake DE (1996) Bird surveys. In: Davis RW, Fargion GS(eds) Distribution and abundance of cetaceans in thenorth-central and western Gulf of Mexico, Vol II. TechRep, OCS Study MMS 96-0027. US Dept Interior, NewOrleans, LA, p 271304

Peterson CH, Rice SD, Short JW, Esler D, Bodkin JL, Bal-lachey BE, Irons DB (2003) Long-term ecosystem responseto the Exxon Valdez oil spill. Science 302: 20822086

Pettingill HS, Weimer P (2002) Worldwide deepwater explo-ration and production: past, present, and future. LeadingEdge 21: 371376

Piatt JF, Ford RG (1996) How many seabirds were killed bythe Exxon Valdez oil spill? In: Rice SD, Spies RB, WolfeDA, Wright BA (eds) Proc Exxon Valdez oil spill sympo-sium, Anchorage, AK. Am Fish Soc Symp 18: 712719

Piatt JF, Lensink CJ, Butler W, Kendziorek M, NysewanderD (1990) Immediate impact of the Exxon Valdez oil spill

236

http://dx.doi.org/10.2307/4087623http://dx.doi.org/10.1190/1.1471600http://dx.doi.org/10.1126/science.1084282http://dx.doi.org/10.1007/s00300-008-0580-3http://dx.doi.org/10.1016/j.marpolbul.2009.12.008http://dx.doi.org/10.1890/ES11-00020.1http://dx.doi.org/10.1676/11-067.1http://dx.doi.org/10.1073/pnas.1112139108http://dx.doi.org/10.1073/pnas.1204729109http://dx.doi.org/10.1007/BF00005044http://dx.doi.org/10.1676/11-084.1http://dx.doi.org/10.1016/0025-326X(93)90124-3http://dx.doi.org/10.1016/j.dsr2.2014.01.013http://dx.doi.org/10.1109/TGRS.2011.2174369http://dx.doi.org/10.1139/a93-008http://dx.doi.org/10.1016/j.rse.2012.03.024http://dx.doi.org/10.1016/S0006-3207(01)00207-5http://dx.doi.org/10.1002/jgrc.20272

Haney et al.: Bird mortality from Deepwater Horizon. I. Offshore

on marine birds. Auk 107: 387397Powers SP, Hernandez FJ, Condon RH, Drymon JM, Free

CM (2013) Novel pathways for injury from offshore oilspills: Direct, sublethal and indirect effects of the Deep-water Horizon oil spill on pelagic Sargassum communi-ties. PLoS ONE 8: e74802

Short M (2011) Montara well head platform spill: Australiasfirst offshore oiled wildlife response. In: InternationalOil Spill Conference Proceedings. American PetroleumInstitute, Washington, DC (Abstract no. 208)

Silver N (2012) The signal and the noise: why so many pre-dictions failbut some dont. Penguin Press, New York,NY

Tasker ML, Mustoe S (2003) Bird observations from R/VGyre, 31 May21 June 2003, SWSS leg 1, Gulf of Mexico.Texas A&M University, Galveston, TX (http: //seawater.tamu.edu/swss/doc/bird_report.doc; accessed 21 Feb2013)

Tasker ML, Jones PH, Dixon T, Blake BF (1984) Countingseabirds at sea from ships: a review of methods em -ployed and a suggestion for a standardized approach.Auk 101: 567577

Tunnell JW, Chapman BR (2000) Seabirds of the CampecheBank islands, southeastern Gulf of Mexico. Atoll Res Bull482, Smithsonian Institution, Washington DC

Vukovich FM (1995) An updated evaluation of the LoopCurrents eddy-shedding frequency. J Geophys Res 100: 86558659

Watson JEM, Joseph LN, Watson AWT (2009) A rapidassessment of the impacts of the Montara field oil leak onbirds, cetaceans and marine reptiles. Dept Environment,Water, Heritage & Arts, Spatial Ecology Laboratory, Uni-versity of Queensland, Brisbane (www. environment. gov.au/ system/ files/ pages/ 1a63b46f-0a2a-4b4a-818c-319 ea8d7cd23/files/montara-rapid-survey.pdf; accessed 12 Sep2013)

Wiese FK (2003) Sinking rates of dead birds: improving esti-mates of seabird mortality due to oiling. Mar Ornithol 31: 6570

Wiese FK, Robertson GJ (2004) Assessing seabird mortalityfrom chronic oil discharges at sea. J Wildl Manag 68: 627638

Wilhelm SI, Robertson GJ, Ryan PC, Schneider DC (2007)Comparing an estimate of seabirds at risk to a mortalityestimate from the November 2004 Terra Nova FPSO oilspill. Mar Pollut Bull 54: 537544

Williams JM, Tasker ML, Carter IC, Webb A (1995) Amethod for assessing seabird vulnerability to surface pollutants. Ibis 137: S147S152

Williams R, Gero S, Bejder L, Calambokidis J and others(2011) Understanding the damage: interpreting cetaceancarcass recoveries in the context of the Deepwater Hori-zon/BP incident. Conserv Lett 4: 228233

Wood KA (1996) Experiments to determine the fate of deadseabirds off Wollongong, New South Wales. Corella 20: 3541

237

Editorial responsibility: Jacob Gonzlez-Sols, Barcelona, Spain

Submitted: October 16, 2013; Accepted: August 10, 2014Proofs received from author(s): September 30, 2014

http://dx.doi.org/10.1111/j.1755-263X.2011.00168.xhttp://dx.doi.org/10.1111/j.1474-919X.1995.tb08435.xhttp://dx.doi.org/10.1016/j.marpolbul.2006.12.019http://dx.doi.org/10.2193/0022-541X(2004)068[0627%3AASMFCO]2.0.CO%3B2http://dx.doi.org/10.1029/95JC00141http://dx.doi.org/10.1371/journal.pone.0074802

cite43: cite5: cite14: cite42: cite13: cite41: cite26: cite39: cite12: cite25: cite38: cite24: cite37: cite10: cite8: cite23: cite6: cite22: cite35: cite4: cite21: cite34: cite19: cite2: cite47: cite20: cite18: cite46: cite32: cite17: cite45: cite16: cite7: cite30: cite15: