Intra- and inter-annual variation in gray body condition on a foraging ground 1, 2 1 LEILA SOLEDADE LEMOS , JONATHAN D. BURNETT , TODD E. CHANDLER, 3 1 JAMES L. SUMICH , AND LEIGH G. TORRES

1Geospatial Ecology of Marine Megafauna Lab, Marine Institute, Fisheries and Wildlife Department, State University, Corvallis, Oregon 97331 USA 2Aerial Information Systems Laboratory, Forest Engineering, Resources and Management Department, Oregon State University, Corvallis, Oregon 97331 USA 3Fisheries and Wildlife Department, Oregon State University, Corvallis, Oregon 97331 USA

Citation: Soledade Lemos, L., J. D. Burnett, T. E. Chandler, J. L. Sumich, and L. G. Torres. 2020. Intra- and inter-annual variation in gray whale body condition on a foraging ground. Ecosphere 11(4):e03094. 10.1002/ecs2.3094

Abstract. store energy gained on foraging grounds to support reproduction and other metabolic needs while fasting for long periods during migration. Whale body condition can be used to moni- tor foraging success, and thus better understand and anticipate individual- and population-level trends in reproduction and survival. We assessed the body condition of eastern North Pacificgraywhales( robustus) on their foraging grounds along the , USA, from June to October of three consecutive years (2016–2018). We used drone photogrammetry and applied the body area index (BAI) to measure and compare whale body condition, which is a continuous, unitless metric similar to the body mass index in humans. A total of 289 drone flights were carried out over 106 photo-identified whales, which were grouped into demographic units by sex, maturity, and female reproductive status. Calves and pregnant females dis- played the highest BAIs, followed by resting females, mature males, and, finally, lactating females, reflecting the significant energetic demands on reproductive females. In all three years, gray whale body condition improved with the progression of feeding seasons, demonstrating the accumulation of body energy reserves on the foraging grounds. Yet, body condition was significantly better in 2016 than in 2017 and 2018 when overall body depletion was observed, indicating a difference in prey availability and/or quality between years. We analyzed local upwelling patterns between 2013 and 2018 as an oceanographic proxy for prey and determined significantly greater upwelling between 2013 and 2015 than low upwelling years between 2016 and 2018. We hypothesize that these upwelling patterns created ecosystem shifts in primary productivity and zooplankton prey of gray whales, causing carry-over effects between foraging success and body condi- tion in subsequent years. This study demonstrates the value of monitoring whale body condition to better understand temporal variation in foraging success, and potentially detect and describe the causes of anoma- lous changes in whale population health, such as the 2019 gray whale mortality event.

Key words: ; body area index; body condition; bottom-up trophic cascades; carry-over effects; drone; morphometrics; Oregon; USA; photogrammetry; unmanned aerial systems; upwelling.

Received 17 September 2019; revised 29 December 2019; accepted 7 January 2020; final version received 13 February 2020. Corresponding Editor: Hunter S. Lenihan. Copyright: © 2020 The Authors. This is an open access article under the terms of the Creative Commons Attribution License, which permits use, distribution and reproduction in any medium, provided the original work is properly cited. E-mail: [email protected]

INTRODUCTION viability of populations (Hunter 1978, Sinclair and Krebs 2002). In both terrestrial and Tight links exist between ecosystem productiv- marine ecosystems, grazer and predator popula- ity, trophic pathways, and the health and tions are dependent on bottom-up trophic

❖ www.esajournals.org 1 April 2020 ❖ Volume 11(4) ❖ Article e03094 SOLEDADE LEMOS ET AL. cascades based on primary productivity (Sinclair effects on individual whale body condition, and Krebs 2002, Benoit-Bird and McManus fecundity, reproduction, and survival (Lockyer 2012). Therefore, environmental perturbations 1986, Thomas 1990, Perryman et al. 2002, Har- that disrupt a predictable source of prey avail- rison et al. 2011, Braithwaite et al. 2015b). For ability could compromise a population’s viability instance, the resting intervals of female fin (Williams et al. 2013, Seyboth et al. 2016, Thomp- whales ( physalus) following birth son et al. 2018). While certain animal populations events may be longer if prey availability is insuf- may have evolved resiliency to periods of low ficient during the foraging season (Lockyer prey availability (Nattrass and Lusseau 2016), 1986). Additionally, a female gray whale cumulative prey shortages can cause poor body (Eschrichtius robustus) bioenergetics model condition, health implications, and ultimately described the energy requirements for a two-year population declines (Acevedo-Whitehouse and reproductive cycle (Villegas-Amtmann et al. Duffus 2009, Tollefson et al. 2010). Hence, popu- 2015) and determined that an annual energy loss lation management can benefit from understand- of (1) 4% would result in a pregnant female to ing variation in body condition of , as lose her calf; (2) 30–35% would prevent females related to normal life history phases such as to reproduce; (3) 37% would cause the lactating reproduction and growth, and in response to female to wean her calf at a lower mass; and (4) environmental conditions (National Academies 40–42% would prevent females to survive. Thus, of Sciences 2017). Monitoring individual- and foraging success impacts calving intervals and population-level body condition can illuminate rates, and calf and mother body condition and population health and associated factors, and the survival, all of which affect population growth limits of population resilience to food limitations. rates (Villegas-Amtmann et al. 2015, Braithwaite Marine ecosystems are particularly dynamic, et al. 2015a). both spatially and temporally, due to multiple Innovative methods have been applied to oceanographic forces that influence various study whale body condition and document vari- trophic pathways in the food web, from plankton ation relative to intrinsic (e.g., reproductive state; to whales (Cury et al. 2008, Heymans et al. 2014). Perryman and Lynn 2002, Christiansen et al. As large, long-lived, migratory marine animals, 2016), environmental (Bradford et al. 2012, Wil- baleen whales are capital breeders that rely on liams et al. 2013), and anthropogenic (e.g., entan- energy storage acquired on feeding grounds to glement in fishing gear; Pettis et al. 2017) factors. finance the costs of reproduction and fasting Previous studies of whale body condition used periods (Lockyer 1987b, Kasuya 1995). Thus, a data (Lockyer 1986, 1987a, Christiansen successful feeding season is critical to gain et al. 2013) or photogrammetry of aerial images energy stores to support their metabolic needs captured from airplanes (Best and Ruther 1992, and reproductive success. For instance, it is esti- Perryman and Lynn 2002, Miller et al. 2012) and, mated that female minke whales (Balaenoptera more recently, from unmanned aerial systems acutorostrata) investing in reproduction must (UAS; drone; Christiansen et al. 2016, Durban increase their body weight by 65% to sustain the et al. 2016, Burnett et al. 2018). While these stud- high energetic demands required by gestation ies did not monitor temporal trends in individual and lactation (Lockyer 1981). body condition due to logistical limits of repli- An unproductive foraging season resulting in cate data acquisition, other studies have exam- low individual fat reserves may affect future ined individual variation of body condition over reproductive performance and survival in subse- time on a breeding ground using drone pho- quent seasons, a phenomenon known as the togrammetry (Christiansen et al. 2018) and on a carry-over effect (Lockyer 1986, Harrison et al. foraging ground through qualitative categoriza- 2011). Short-term energetic balance disruptions tion of whale fat deposits in photographs (Pettis in whales with large reserves may lead to only et al. 2004, Bradford et al. 2012). However, quan- marginal effects, but persistent poor foraging titative aerial photogrammetry methods have not may profoundly affect individual fitness (New yet been applied to assess individual- and popu- et al. 2013). Cumulative poor foraging could trig- lation-level changes in whale body condition ger a downstream process that leads to adverse over a foraging season, which is critical to

❖ www.esajournals.org 2 April 2020 ❖ Volume 11(4) ❖ Article e03094 SOLEDADE LEMOS ET AL. understanding how whales respond to, and tol- whale body condition variation during three con- erate, changes in prey availability and quality. secutive foraging seasons; (2) describe how gray In this study, we use drones to repeatedly sam- whale body condition varies with respect to ple gray whale body condition while on their for- demographic units; and (3) explore possible asso- aging grounds along the Oregon coast, USA, ciations between gray whale body condition during three consecutive years. Through analy- variability and oceanographic conditions. ses of these replicate data, we describe body con- Our study is particularly relevant in light of dition variation relative to demographic unit, the 2019 unusual mortality event of ENP gray time, and environmental conditions, at both the whales with many emaciated whales found individual level and aggregated population level, stranded along the west coasts of , the adding to the understanding of gray whale United States, and Canada since January 2019 bioenergetics provided by previous studies (e.g., (NOAA 2019). Environmental anomalies and ele- Rice and Wolman 1971, Perryman and Lynn vated intraspecific competition due to limits on 2002, Bradford et al. 2012, Villegas-Amtmann carrying capacity are hypothesized causes of et al. 2015). these mortalities, with both mechanisms causing Eastern North Pacific (ENP) gray whales con- low prey availability and hence depleted whale duct one of the longest migrations of any mam- body condition. Our work establishes a baseline mal, as they reach up to 20,000 km round trip of gray whale body condition during the forag- every year between their breeding grounds in ing season and by demographic unit, which may Baja , Mexico, and feeding grounds in be used to identify trends, anomalies, and possi- the Bering and Chukchi seas, off the coast of ble causes of mortality events. Using the concept , United States (Jones and Swartz 2002, of carry-over effects, we examine possible links Sumich 2014). A subset of this population, between body condition and environmental vari- known as the Pacific Coast Feeding Group ability, which elucidates drivers of whale bioen- (PCFG), do not complete the full migration ergetics and can inform population management. (Calambokidis et al. 2002, Lang et al. 2014). This group of ~250 individuals feed during summer METHODS and fall along the coast from northern California through southeastern Alaska, including the Ore- Study area gon coast (Calambokidis et al. 2012, Weller et al. This study was performed on two gray whale 2013). Possible explanations for this truncated foraging areas along the coast of Oregon, USA: migration are energetic savings of a shorter tra- Port Orford (42°44059″ N, 124°29053″ W), in vel distance, leading to a longer foraging period, southern Oregon, and Newport (44°38013″ N, and efficient prey availability due to abundant 124°03008″ W), in central Oregon. The study sites food sources, such as mysid aggregations are ~180 km apart with photographically docu- (Holmesimysis sculpta; Newell and Cowles 2006). mented exchange of individual gray whales The development of drone-based whale pho- between areas within and between foraging sea- togrammetry methods facilitates easier, cheaper, sons (L. Torres, unpublished data). The two sites and safer collection of replicate aerial whale are environmentally similar, comprising rocky images in a noninvasive fashion (Christiansen nearshore environments, with the presence of et al. 2016, Smith et al. 2016, Burnett et al. 2018) patchy kelp habitats, and similar zooplankton than previously attainable. We implement the communities (L. Torres, unpublished data). quantitative and continuous metric of whale body area index (BAI) developed by Burnett Sampling methods et al. (2018) to assess PCFG gray whale body con- A small research vessel (a 5.4-m rigid-hulled dition. The BAI allows comparative assessment inflatable boat) was used to locate gray whales of whale body condition between time steps, and collect individual data during three feeding subgroups (e.g., by sex), and within individual to seasons between June and October of 2016, 2017, monitor and describe relative foraging success and 2018. Sampling was conducted in both study and health. The objectives of this study are to (1) sites in the first two years and only in Newport examine intra- and inter-annual PCFG gray in the last year, due to logistics. During whale

❖ www.esajournals.org 3 April 2020 ❖ Volume 11(4) ❖ Article e03094 SOLEDADE LEMOS ET AL. encounters, photographs of the left-hand side, Analysis methods right-hand side, and the fluke of each individual Photo-identification.—Gray whale photo-identi- were obtained for photo-identification by using fication was conducted using Adobe Bridge (ver- two cameras (Canon EOS 7D and Canon EOS sion 8.0.1.282; Adobe Systems, San Jose, CA, 70D) simultaneously. In addition, the number of USA) by visually assessing unique skin pigmen- sighted whales and calf presence were recorded. tation and marks (e.g., scars) on the left-hand Whale fecal samples were also collected oppor- side, right-hand side, and fluke that are stable tunistically using a 300-lm sampling net (subse- through time and identify distinct individuals quently placed on ice until stored in a 20°C (Hammond et al. 1990). Photograph quality was freezer), which enabled genetic sex determina- first assessed to ensure images were in-focus, tion. well lit, and not affected by glare, angle, or dis- When the whale behaved naturally during tance (Darling 1984, Hammond et al. 1990). observation (i.e., no behavioral changes due to Images of each identified whale were compared the presence of the research vessel) and weather with long-term (>30 yr) gray whale photo-id cat- conditions were appropriate, an UAS was alogs (held by the Institute at launched from the vessel to conduct whale over- Oregon State University, and by Cascadia flights and record videos during surfacing Research Collective, Olympia, , USA) events. Three different drones were used during in order to obtain sighting histories, which pro- this study. A DJI Phantom 3 Pro was used in vided information on the whale’s minimum age 2016, and a Phantom 4 was used in 2017. Both of estimate based on date of first sighting, and sex these aircrafts included a camera with 3.61-mm based on previous genetic analysis of collected focal length and pixel size (i.e., pitch) of approxi- tissue samples (Lang et al. 2014). When sex infor- mately 0.0015 mm (at 4000 9 3000 image resolu- mation was not available, genetic analyses were tion). In 2018, a DJI Phantom 4 Pro was used, performed on fecal samples collected from the with 8.8-mm focal length camera and pixel size individual whale, if available (see Appendix S1 of approximately 0.0033 mm (at 4000 9 3000 for fecal genetic analysis methods). image resolution). All aircrafts were remotely Assignment to demographic unit.—Each individ- controlled using an Apple iPad Mini 4 tablet with ual whale was assigned into a demographic unit the DJI Go 4 app. based on sex (male or female), maturity (imma- Collection of whale photogrammetry data ture, calves and juveniles; or mature, adults), and using the drone followed protocols described female reproductive status (resting, pregnant, by Burnett et al. (2018). In short, the drone lactating, and postweaning). It was assumed that hovered above the whale, at a minimum alti- (1) small whales (<8 m in length) swimming in tude of 25 m. Whales were recorded while close association with a large whale (>11 m in centered in the field of view and with the length) were calves (Perryman and Lynn 2002) camera pointing straight down at the whale and that the large whale was a lactating female (i.e., nadir). Video metadata at 1-s intervals mother, (2) a lactating female was a pregnant provided information on geolocation and alti- female in the previous year, and (3) a lactating tude relative to initialization height, which was female became a postweaning female when the consistent throughout the study (1.0 m) and calf was not associated with the mother in the compensated for in the photogrammetry analy- same year. Resting females represent mature sis. A calibration object of 1.0 m in length was females that were not pregnant or lactating in the centered in the frame of the drone camera dur- analysis year. Sexual maturity occurs at different ing the beginning of all flights and recorded ages and lengths in female and male gray whales from 10 to 40 m of altitude for correction of (Zimushko 1969, Zimushko and Ivashin 1980). In barometric altimeter errors during postprocess- females, it is attained between 8 and 12 yr, and ing. All overflown whales were simultaneously at approximately 12 m. In males, sexual maturity photographed from the boat to confirm indi- is reached at 6–8 yr and at around 11.5 m. To vidual identification. Whale behavior was also obtain confidence in individual maturity classifi- monitored to assess changes in response due cation, 8 and 12 yr of minimum age were applied to the drone presence. as maturity age cutoffs for males and females,

❖ www.esajournals.org 4 April 2020 ❖ Volume 11(4) ❖ Article e03094 SOLEDADE LEMOS ET AL. respectively. Individuals at a length range assessments were limited to between 20% and between 10 and 12 m of unknown sex or age 60% of the body length where body shape is were classified as undetermined. more likely to vary with energy reserves. Addi- Photogrammetry.—Drone videos were viewed tional morphometrics produced are the manual using the VLC Media Player software (version width (MW, widest width measured manually), 2.2.8; VideoLAN, Paris, France.). Full-resolution the optimized width (OW, point on parabola frames of the whale during surfacing events were nearest MW), and the surface area (SA) between extracted using the snapshot function. We aimed the two fitted parabolas. to extract five images of each individual per Morphometric attributes were measured in flight, but this was not always possible due to pixels and converted to meters via ground sam- non-ideal whale body positions and/or quality of pling distance (GSD, i.e., ground distance of one the image. Once images were extracted, image pixel), assuming that the aircraft altitude above quality was assessed and categorized as either sea level (ASL) was zero (Burnett et al. 2018). good or poor, according to attributes relative to Burnett et al. (2018) applied a GSD correction the whale body position (body straightness and model by calculating the empirical GSD (eGSD, body arching), camera focus, and environmental calibration object length in meters/calibration conditions (glare, water splash, and water turbid- object length in pixels) and regressing eGSD on ity; Appendix S2: Table S1, Fig. S1). If the image reported altitude. This model outputs a corrected was not in-focus, the whale’s body was bending GSD and accounts for systematic errors in or arched, or if any of the environmental factors altimeter observations (which propagate into masked >25% of one side of a body contour, the GSD). The program also incorporates lens distor- image was classified as poor and excluded from tion corrections for each of the three drone cam- the analysis. (Photogrammetry analysis methods eras used in this study. apply a best-fit parabola to analyst-defined Using the whale quantitative analysis program points along the body edge contour, allowing for (Burnett et al. 2018) in R (version 3.5.0; R Core some gaps in edge definition by the analyst; see Team 2019), measurement data of multiple Methods below and Burnett et al. 2018.) Five images of the same whale were grouped by date, images of the calibration object, at different alti- flight, and sighting to produce summarized mea- tudes, were also extracted from each drone surement data, including BAI values and error video. estimates based on the measurements (Table 1; Images of calibration objects and whales Fig. 1). To minimize the impact of measurement assessed as good quality were, respectively, mea- error, a threshold of 5% was applied for both WL sured using the whale calibration object mea- and BAI coefficients of variance (CV). Any image surement and whale measurements programs in MATLAB (version 9.3.0.7; MathWorks, Natick, Table 1. Morphometric attributes calculated for whale MA, USA) developed by Burnett et al. (2018). photogrammetry analysis. Using these programs, the analyst measured a Attribute Parameter Description single length value of the calibration object and a series of measurements of each whale image, 1 FW Fluke width, tip to tip fluke width which produced an output table containing 10 2 WL Whale length, rostrum to notch in tail whale morphometric attributes (Table 1, Fig. 1) 3 W20 Width at 20% of WL from rostrum 4 W30 Width at 30% of WL from rostrum for each image. Morphometric attributes 5 W40 Width at 40% of WL from rostrum included common morphometrics (whale length, 6 W50 Width at 50% of WL from rostrum WL, and width, MW; Perryman and Lynn 2002), 7 W60 Width at 60% of WL from rostrum and a series of width measurements at different 8 MW Manual width, manual measurement body length percentage points between 20% and of width at widest point – fi 9 OW Optimized width, width at point on 60% (W20 W60), to which the program ts parabola nearest MW parabolas along the body edges of these 20–60% 10 SA Surface area between 20% and 60% points. According to Koopman (1998), head, fins, of WL and fluke of cetaceans are not used as energy 11 BAI Body area index reservoirs. Therefore, body condition Note: Modified from Burnett et al. (2018).

❖ www.esajournals.org 5 April 2020 ❖ Volume 11(4) ❖ Article e03094 SOLEDADE LEMOS ET AL.

Fig. 1. Example surface image of a gray whale extracted from drone video and the morphometric measure- ments calculated during photogrammetry analysis using methods and programs developed by Burnett et al. (2018). (A) Extracted image of a whale at a surfacing event. (B) Measurements calculated: WL, whale length; FW, fluke width; OW, optimized width. 0.2–0.6: body proportion points, and parabolas (orange and purple lines) fit to the perimeter curves of the body based on location of 0.2–0.6 body percentage points. (C) Body surface area (SA) based on parabolas, from which the body area index (BAI) is calculated. causing a CV to be higher than 5% was excluded unitless and scale-invariant metrics of body con- from the final analysis. dition. BAI was developed based on the BMI for- BAI is similar to the body mass index (BMI) mula (mass (Kg)/height (m)2; Gallagher et al. used to evaluate human health because both are 1996) by using SA as a surrogate for body mass

❖ www.esajournals.org 6 April 2020 ❖ Volume 11(4) ❖ Article e03094 SOLEDADE LEMOS ET AL.

(Burnett et al. 2018) and calculated using the fol- and oceanographic conditions by extracting lowing equation: upwelling index data from 6-hourly composites of one-degree Fleet Numerical Meteorology and SA BAI ¼ 100: Oceanography Center sea level pressure at the ð0:4 WLÞ2 location closest to our main study site (Newport, OR, 45° N, 125° W). We sampled the data dur- Whale length (WL) is multiplied by 0.4 ing the study period (2016–2018) and the previ- because SA is only captured across 40% of WL ous three years (2013–2015) to evaluate possible (between 20% and 60% body length). The final carry-over effects on gray whale body condition. multiplication by 100 allows for a whole number The cumulative upwelling index sum (CUI) was BAI value. computed by first calculating the daily upwelling An advantage of the BAI assessment of whale index means from the six-hourly data, and then body condition is its scale invariance and inde- calculating the cumulative sum of daily upwel- pendence of body length, which enables compar- ling index means for each day based on the days isons over time of body condition within and since 1 January of that year. Hence, CUI provides among individuals regardless of different body an estimate of the net influence of upwelling lengths (i.e., calves:adults or males:females). In strength on the ecosystem as a proxy of variation addition, BAI is not influenced by scaling errors in productivity throughout the year (Bograd between pixels and metric units that may arise et al. 2009). during photogrammetry efforts if altitude is In order to compare upwelling conditions uncertain. For instance, we encountered inaccu- prior to the 2019 gray whale mortality event and rate altitude values in our 2016 data due to high a previous mortality event during 1999–2000 (Le flight speeds over the calibration board. (The Boeuf et al. 2000, Moore et al. 2001), we also con- flight speed was adjusted in the following years.) ducted the CUI analysis for the period of 1993– Consequently, absolute length and width values 1998. are not accurate from 2016, and only values from 2017 and 2018 are reported. However, since BAI Statistical methods is a unitless metric and is not affected by scaling The Shapiro-Wilk normality test verified the errors, BAI values remain accurate and compara- normal distribution of all variables assessed. All ble between all years, including 2016. To verify statistical tests were conducted in R (version the resilience of BAI values, we applied a linear 3.5.0; R Core Team 2019). regression to compare BAI values calculated To assess the influence of multiple factors on from raw pixel values and from scaled metric changes in BAI while also accounting for individ- values from the same whale images captured in ual replication, a linear mixed model (LMM) was 2017 and 2018. applied, using the lme4 package in R (Bates et al. Upwelling index.—In Oregon, coastal upwelling 2015). The influence of the following predictor occurs annually, from April to September, bring- variables on the response variable BAI was ing nutrient-rich water to the surface and fueling tested: demographic unit, day, month, year, and seasonal primary production (Mooers et al. 1976, study site (Newport or Port Orford). All models Harvey et al. 2018, Peterson et al. 2018). This included the whale identification as a random upwelling season overlaps with gray whale pres- effect to account for pseudoreplication. Model ence in the study area and may be a significant selection was based on Akaike’s information cri- driver of the availability of their zooplankton terion (AIC; Burnham et al. 2011). Nonsignificant prey. Although no study has directly addressed variables were excluded to use the most parsimo- this hypothesis, other studies in the California nious model. To evaluate the LMM fit, marginal fi 2 2 fi Current System (CCS) have found signi cant R (Rm, variance explained by xed effects) and 2 2 correlations between upwelling strength and conditional R (Rc , variance explained by both zooplankton biodiversity and abundance, includ- fixed and random effects, or the entire model) ing copepod and krill species (Thompson et al. were assessed using the MuMIn package (Naka- 2018). Thus, we investigated possible correlations gawa and Schielzeth 2013, Barto 2018). The P val- between population-level whale body condition ues and F statistics were obtained using the

❖ www.esajournals.org 7 April 2020 ❖ Volume 11(4) ❖ Article e03094 SOLEDADE LEMOS ET AL. lmerTest package (Kuznetsova et al. 2017). A Whale lengths and widths pairwise analysis of estimated marginal means For the combined 2017 and 2018 dataset, mean was conducted to compare significant fixed WL standard deviation (SD) and mean OW effects, using the emmeans package in R (Lenth SD for mature females (n = 29) were 12.43 m 2019). In addition, a generalized linear model 0.90 (range 10.23–14.12) and 2.13 m 0.23 (range (GLM) was conducted to determine the effects of 1.63–2.61), respectively, while mature male month and year on CUI, which followed a Gaus- (n = 29) values were 11.59 m 0.94 (range 9.81– sian distribution. For all tests, a P value of 0.05 13.88) and 1.99 m 0.19 (range 1.49–2.40). Imma- level of significance was considered. ture females (n = 8) displayed higher values for Linear regressions were also conducted using WL (10.42 m 0.94, range 8.44–11.56) and OW the lm function in R to describe (1) the trend in (1.78 m 0.17, range 1.42–1.92) when compared gray whale BAI change over the three seasons; to immature males (n = 9) WL (9.70 m 0.84, (2) the relationship between gray whale body range 8.30–10.75) and OW values (1.65 m 0.16, length (WL) and optimized width (OW); and (3) range 1.42–1.92). For calves (n = 2), the mean WL the resilience of BAI values calculated by raw was 7.61 m 0.48 (range 7.27–7.95) and mean pixel values relative to scaled metric values for OW was 1.39 m 0.10 (range 1.31–1.46; whales measured in 2017 and 2018. Appendix S2: Fig. S2). Photogrammetry analysis of good quality Absolute WL values in 2017 and 2018 and the images from 2016 was conducted by two analysts, strong correlation between WL and OW and a t-test was conducted to assess potential bias (R2 = 0.68, slope = 0.16, P < 0.001; Appendix S2: in calibration object and whale measurements Fig. S3) were comparable with data previously between analysts. As no statistically significant documented for gray whales (Perryman and difference between observer measurements was Lynn 2002), with calves representing the smaller found (t = 0.829, P = 0.41), only one observer whales (<8 m). Yet, it is the residuals around the (LSL) continued measuring the whales and cali- correlation line between WL and OW that bration boards in the following seasons. describes the variation of individual body condi- tion, which is captured through comparison of RESULTS the BAI metric that relates the whale’sSA between the 20% and 60% body length to WL. A total of 171 whales were photo-identified during our study period from 2016 to 2018, com- Model selection prising 40 females, 40 males, and 91 unsexed Ten linear mixed models were compared to individuals. Fifteen of the sexed individuals were assess the influence of multiple predictor vari- determined by fecal genetic analyses (eight ables on BAI (Table 3). Model 5 had the lowest females and seven males). Eleven of the unsexed AIC (AIC = 1306.06, df = 20), yet study site was individuals were calves sighted along with their not a significant variable (P > 0.05) in the model. mothers (2016, n = 9; 2017, n = 1; and 2018, Thus, model 4 was selected as the most parsimo- n = 1). nious model because it did not include study site A total of 289 drone flights (2016, n = 76; 2017, and only had a marginally lower AIC (LMM, = = = = 2 = 2 = n 113; and 2018, n 81) were performed over AIC 1307.74, df 19, Rm 0.24, Rc 0.37). In 62% of the identified whales (n = 106). Sixteen model 4, demographic units (F = 3.17, P < 0.01, unique gray whales were flown over in all the df = 10), month (F = 5.46, P < 0.001, df = 4), and three feeding seasons, 25 in two seasons and 65 year (F = 13.56, P < 0.001, df = 2) exhibited sig- only in one season. However, only 272 drone nificant effects on BAI, where year was the most observations of 104 individuals were analyzed significant factor. after image quality assessment, with varying The generalized linear model describing the sample sizes by year and demographic unit effects on CUI with the lowest AIC (AIC = (Table 2). 28960.22, df = 73, R2 = 0.55; Appendix S2: We noted no behavioral response of whales to Table S2) included the predictor variables month the drone overflights, either in the field or during (GLM, F = 153.90, P < 0.001, df = 11), year video review. (GLM, F = 41.26, P < 0.001, df = 5), and the

❖ www.esajournals.org 8 April 2020 ❖ Volume 11(4) ❖ Article e03094 SOLEDADE LEMOS ET AL.

fl Table 2. Unmanned aerial system (UAS) over ights of gray whales by number of individuals (Nind), number of

observations (Nobs), and respective mean and standard deviation (SD, in parentheses) of BAI per demographic unit per year.

2016 2017 2018 Demographic units Nind BAI Nobs BAI Nind BAI Nobs BAI Nind BAI Nobs BAI

Calf 3 44.61 (5.40) 3 44.61 (5.40) 1 41.33 1 41.33 1 41.89 1 41.9 Immature male 0 NA 0 NA 5 38.99 (2.68) 9 38.84 (2.33) 3 36.78 (1.87) 6 37.38 (1.82) Immature female 5 40.11 (2.41) 5 40.11 (2.41) 5 38.67 (2.79) 8 38.38 (2.58) 4 37.37 (3.42) 9 37.92 (2.57) Immature 0 NA 0 NA 5 38.15 (2.27) 6 38.49 (2.49) 4 40.50 (0.71) 7 40.47 (1.16) unknown sex Mature male 15 40.80 (3.39) 21 40.77 (3.63) 13 38.93 (2.89) 22 38.9 (2.67) 16 38.14 (2.31) 33 38.24 (2.74) Mature 2 40.86 (4.54) 2 40.86 (4.54) 2 37.65 (1.08) 2 37.65 (1.08) 4 38.05 (1.78) 8 39.09 (2.31) unknown sex Resting 14 41.19 (2.03) 23 41.21 (2.63) 12 38.23 (2.46) 28 38.51 (3.15) 13 38.21 (3.04) 26 38.4 (2.48) female Pregnant 1 43.87 1 43.87 1 43.45 1 43.45 1 42.42 1 42.42 female Lactating 3 35.75 (2.84) 4 36.28 (2.60) 1 34.61 1 34.61 1 38.07 2 38.07 (2.26) female Postweaning 1 38.32 2 38.32 (1.95) 1 36.37 2 36.38 (0.68) 1 38.11 2 38.11 (3.41) female Undetermined 11 40.20 (2.84) 16 40.21 (3.69) 9 37.83 (2.21) 13 37.7 (2.09) 6 39.64 (1.61) 7 39.85 (1.652) Notes: The number of observations is relative to the number of different flights conducted over these individuals. Individual whales may be represented in this table multiple times in case of (1) sightings in multiple years, and/or (2) change in reproduc- tive female cycle within a year. NA, not applicable. interaction of month 9 year (GLM, F = 12.24, Table 3. Linear mixed model selection parameters of P < 0.001, df = 55). gray whale body area index (BAI) relative to the pre- dictor variables demographic unit (DU), day of the Demographic unit comparisons year (DOY), month, year, and study site. Gray whale BAI varied significantly relative to demographic units (Fig. 2). Calves displayed the Models df AIC highest mean BAI (n = 5; 43.41 4.16 (1) BAI ~ DU + (1|ID) 13 1349.597 [mean SD]), followed by pregnant females (2) BAI ~ DU + month + (1|ID) 17 1329.607 ~ + + | (n = 3; 43.25 0.75), resting females (n = 26; (3) BAI DU year (1 ID) 15 1322.546 = (4) BAI ~ DU + month 19 1307.745 39.41 2.48), mature males (n 30; 39.16 2.51), + year + (1|ID) immatures of unknown sex (n = 8; 39.00 2.11), (5) BAI ~ DU + month 20 1306.065 matures of unknown sex (n = 6; 38.94 2.89), + year + study site + (1|ID) ~ + | immature males (n = 7; 38.85 2.62), immature (6) BAI study site (1 ID) 4 1364.056 = (7) BAI ~ DU + DOY + (1|ID) 14 1337.516 females (n 6; 38.29 2.68), postweaning females ~ + + + | = fi (8) BAI DU DOY month (1 ID) 18 1330.192 (n 3; 37.60 1.07), and, nally, lactating females (9) BAI ~ DU + DOY + month 20 1310.104 (n = 5; 35.99 2.37), which presented the most + year + (1|ID) depleted body condition. The high variation in (10) BAI ~ DU + DOY + month + year 21 1309.088 + | female body condition between reproductive study site (1 ID) phases was also detectable at the individual level Notes: All models used whale identification (ID) as a ran- fl dom effect. Model 4 (in bold) was selected as the best and where BAI values uctuated with repeated mea- most parsimonious model based on the Akaike information surements of the same individual during different criterion (AIC) and the insignificance of study site and DOY phases (e.g., Fig. 3). Furthermore, all (n = 4) but to the model. one lactating female examined in this study dis- played a worse body condition in the year they condition after weaning events (n = 3; lactating were lactating (36.41 2.15 [mean BAI SD]) females: 35.07 2.80; postweaning females: compared to years when they were in resting 37.60 1.06). (37.16 4.71)orpregnant(43.66 0.30) states. Significant BAI differences were found between Also, all lactating females exhibited improved body calves and immature females (P < 0.05), mature

❖ www.esajournals.org 9 April 2020 ❖ Volume 11(4) ❖ Article e03094 SOLEDADE LEMOS ET AL.

Fig. 2. The effect of demographic units on gray whale body area index (BAI) measured during June–October of 2016–2018 along the Oregon coast, USA, derived from linear mixed model results with whale identification as random effect (model 4). males (P < 0.05), resting females (P < 0.05), lactat- between 2016 and 2018 (P < 0.001), but not ing females (P = 0.001), postweaning females between 2017 and 2018 (Table 4). (P < 0.01), and undetermined (P < 0.05; Table 4). Gray whale body condition had significant inter-annual variation (Fig. 4B). The population Intra- and inter-annual body condition variation presented the best body condition in 2016 BAI values calculated from raw pixel values (n = 57; 40.82 2.91 [mean BAI SD]), fol- and from scaled metric values had a perfect lin- lowed by 2017 (n = 55; 38.67 2.45) and 2018 ear relationship (r = 1, P = 0), demonstrating (n = 54; 38.62 2.38), when considering a mean that 2016 BAI values are comparable to 2017 and per whale per demographic unit. 2018 measurements despite a lack of accurate altitude values in 2016 to facilitate scaled metric Upwelling index measurements. Significant annual variation in the CUI was Gray whales exhibited an overall intra-annual observed between the study years with the high- improvement in body condition across the forag- est CUI levels occurring in 2015 and 2013, fol- ing season months, indicating energetic gain at lowed by 2014 and 2018 with values near the the population level throughout the feeding sea- mean, and then, 2017 and 2016 had the lowest sons (Fig. 4A). However, the linear model results CUI levels, which were below the mean (Fig. 5). show a significant increase in body condition These data illustrate the relatively poor upwel- over the months in 2017 (rate of change = 0.03, ling conditions in the study area from 2016 = < 2 = F1,88 15.7, P 0.001, R 0.142) and 2018 (rate through 2018, particularly during 2016. All three = = < 2 = fi of change 0.01, F1,96 6.302, P 0.05, R years were signi cant predictors of CUI variation 0.01), while in 2016, improvement in body condi- (2016, F = 315.807; 2017, F = 422.073; and tion was nonsignificant (rate of change = 0.01, 2018, F = 582.113; P < 0.001). = > 2 = – F1,67 1.178, P 0.05, R 0.003). Assessment of the CUI for the 1993 1998 per- Significant BAI differences were found between iod that preceded the 1999–2000 gray whale mor- the months June and September (P < 0.01) and tality event indicates that three of the years October (P < 0.05), and between July and Septem- (1993, 1995, and 1998) were below the mean dur- ber (P < 0.01) and October (P < 0.05; Table 4). ing the foraging period (1 June–15 October; Significant BAI differences were also found Appendix S2: Fig. S4). In contrast to the observed between the years 2016 and 2017 (P < 0.001) and three-year period of relatively low CUI values

❖ www.esajournals.org 10 April 2020 ❖ Volume 11(4) ❖ Article e03094 SOLEDADE LEMOS ET AL.

22 June 2016 11 October 2016 7 September 2018 Lactating Postweaning Resting BAI = 32.52 BAI = 39.70 BAI = 37.49

Fig. 3. Body condition comparison of a female whale over the years and throughout different reproductive states using the body area index (BAI) metric. Differences in image quality are associated with the drone used for overflights (June 2016, DJI Phantom 3 Pro; October 2016, DJI Phantom 4; and September 2018, DJI Phantom 4 Pro). during the 2016–2018 period that preceded the body condition change over three consecutive 2019 gray whale mortality event, the three-year feeding seasons relative to time, demographic period of 1996–1998 that preceded the 1999–2000 unit, and environmental conditions. We docu- mortality event did not display a similar low ment significant differences in whale body condi- CUI trend. tion by month with general improvement as the foraging season progressed, by demographic DISCUSSION unit with lactating females having the least fat reserves, and by year with potential links to tem- This study builds upon foundational studies of porally lagged environmental conditions. These baleen whale body condition variation on forag- findings demonstrate the utility of the BAI metric ing grounds (Lockyer 1986, Bradford et al. 2012) to detect and quantify variation in whale body through application of quantitative drone-based condition at both the population level and indi- photogrammetry metrics to describe gray whale vidual level, which can be applied to monitor

❖ www.esajournals.org 11 April 2020 ❖ Volume 11(4) ❖ Article e03094 SOLEDADE LEMOS ET AL.

Table 4. Significant pairwise comparison of estimated Bradford et al. 2012), the two methods can be (1) marginal means (EMMs) of the most parsimonious used to validate each other and (2) integrated to linear mixed model (model 4) for gray whale body temporally extend assessment periods. area index (BAI) according to the variables demo- graphic unit, month, and year. Intra-annual body condition variation by demographic unit Contrast Estimate SE df T ratio P At the population level, gray whale body con- Calf, immature 4.88152 1.476 180.2 3.308 0.0436 dition improved significantly across months with female the progression of the feeding season in 2017 and Calf, mature 4.54966 1.275 223.3 3.569 0.0186 male 2018, indicating energy replenishment (Figs. 3, Calf, resting 4.47262 1.277 220.2 3.503 0.0231 4A). This result demonstrates that whales gain female mass during the feeding season, and aligns with Calf, lactating 6.91745 1.591 231.8 4.347 0.0010 fi female previous body condition assessments in n, and Calf, 6.75570 1.688 233.5 4.002 0.0040 western and eastern gray whales (Rice and Wol- postweaning man 1971, Lockyer 1986, 1987a, Perryman and female Lynn 2002, Bradford et al. 2012). Calf, 4.68251 1.309 224.5 3.577 0.0181 undetermined Interestingly, in 2016 when the whales were in June–September 1.693 0.485 255 3.492 0.0051 relatively good body condition following at least June–October 1.739 0.615 255 2.828 0.0401 three years of above-average upwelling condi- July–September 2.127 0.608 249 3.501 0.0049 tions, whale body condition did not show a sig- July–October 2.173 0.701 252 3.100 0.0182 fi – < ni cant trend over months. Perhaps when 2016 2017 2.0038 0.449 253 4.459 0.0001 whales are already robust and energetic require- 2016–2018 2.0296 0.432 254 4.693 <0.0001 ments are satisfied, they reduce their energy intake rate, as has been postulated by Machovsky-Capuska and Raubenheimer (2020), baleen whale population health relative to envi- gain mass more slowly, or have limited capacity ronmental and anthropogenic perturbations. to gain more mass. However, studies on this topic are scarce and a longer time series is The utility of BAI needed to assess these ideas. We documented a perfect fit of BAI values Gray whale body condition also varied accord- between scaled and unscaled photogrammetry ing to sex, age, and reproductive status, with images. Therefore, despite unscaled images in calves displaying the highest mean BAI, followed 2016, we demonstrate that the unitless and scale- by pregnant females, while lactating and post- invariant BAI metric can still be confidently weaning females had the lowest BAI values applied to assess whale body condition and (Fig. 2). These results are consistent with previ- allow comparisons across time periods when ous findings reported for western gray whales derived from measurements of low-distortion (Bradford et al. 2012) and right whales (Eubalaena nadir images over relatively flat water. Hence, australis and E. glacialis; Angell 2006), yet et al. the BAI can be applied to many datasets of aerial (2016) used drone-based photogrammetry to whale images captured in nadir that lack accu- document that (Megaptera rate altitude data (e.g., from helicopters, air- novaeangliae) calves on a breeding ground exhib- planes, or drones) for robust assessment and ited the worst body condition index (BCI), fol- comparison of body condition. We are confident lowed by immature, mature, and lactating that the BAI can add significant value to many females. This contrast in relative calf body condi- historical aerial image datasets to foster explo- tion between studies may be due to the method ration of pressing research questions regarding of body condition assessment (BAI vs. BCI), the impacts of environmental change on whale where BCI is an absolute measure of body condi- health. Additionally, while drone-based quantita- tion that does not account for the length of the tive photogrammetry assessments of whale body animals. Furthermore, et al. (2016) investigated condition, such as BAI, improve upon categorical how whale body condition declined on breeding determinations from oblique photographs (i.e., grounds over a two-month period, while the

❖ www.esajournals.org 12 April 2020 ❖ Volume 11(4) ❖ Article e03094 SOLEDADE LEMOS ET AL.

Fig. 4. Gray whale body area index (BAI), measured during June–October of 2016–2018 along the Oregon coast, USA, as a function of (A) month and (B) year, derived from linear mixed model results with whale identifi- cation as random effect (model 4). No sampling effort was performed in July 2016. present study investigated the improvement of represent the most depleted group, while preg- body condition on foraging grounds over five- nant females represent one of the most robust month periods in three different years. Although groups on the foraging grounds, aligning well our dataset contained no repeated photogram- with previous studies (Lockyer 1987a, Perryman metric data for any calf within a year, the high and Lynn 2002, Pettis et al. 2004, Miller et al. BAI values for all calf measures indicate that they 2011, Bradford et al. 2012). Baleen whales are were well nourished, reflecting the high milk fat iteroparous with lactation considered the most transfer from mother to calf (Rice and Wolman energetically demanding phase of the female 1971). reproductive cycle (Lockyer 1981, Gittleman and Our assessment of body condition by demo- Thompson 1988). Thus, it is crucial that pregnant graphic unit also indicates that lactating females females develop a significant energy storage

❖ www.esajournals.org 13 April 2020 ❖ Volume 11(4) ❖ Article e03094 SOLEDADE LEMOS ET AL.

Fig. 5. Cumulative upwelling index (CUI; m3/s/100 m coastline) computed from the daily upwelling index at the west coast of United States, offshore of Newport, Oregon (latitude 45° N, longitude 125° W), from 2013 to 2018. The black line represents the cumulative sum of the daily means over the six years. The black dashed verti- cal lines mark the study period (1 June–15 October). Daily upwelling index data obtained from http://upwell.pfe g.noaa.gov/erddap/. during the prior feeding season(s) to successfully foraging seasons to allow reproductive females gestate, birth, cover the high costs of lactation, to recover. Variations in timing and extent of and support her own metabolic needs (Lockyer body condition improvement of postweaning 2007, et al. 2013, 2016). Interestingly, we observed mothers may be due to individual foraging and nine mother–calf pairs in 2016, when the body movement patterns, energetic transfer to calves, condition was relatively good following multiple and environmental conditions that influence years of productive upwelling conditions. In con- prey availability. trast, we only observed one mother–calf pair in 2017 and one in 2018, when gray whale body Inter-annual body condition variation associated condition was relatively poor. These findings with the upwelling index indicate that prey resources in the previous years PCFG gray whales exhibited a significantly (2016 and 2017) may not have been sufficient for better body condition in 2016 when compared to females to store adequate energy reserves to the following years, suggesting a lack of ade- invest in reproduction in 2017 and 2018, which quate prey availability and energy reposition by aligns with previous findings (Perryman et al. whales preceding measurement during the 2017 2002). and 2018 seasons. This hypothesis is supported We also document improved body condition by our assessment of CUI from 2013 to 2018, of lactating females after weaning events, which which indicates the potential for carry-over is in line with previous findings (Miller et al. effects between environmental conditions (used 2011) and highlights the importance of successful as a proxy for prey availability) and whale body

❖ www.esajournals.org 14 April 2020 ❖ Volume 11(4) ❖ Article e03094 SOLEDADE LEMOS ET AL. condition in subsequent feeding seasons. We Disentangling the potential causes of the 2019 assume that favorable upwelling conditions mortality event is challenging, as both limits of result in increased prey for whales in Oregon, carrying capacity and unfavorable oceano- based on (1) the strong links between upwelling graphic conditions may be synergistic factors. in the CCS and increased primary production in While the qualitative correlation between CUI the spring (Schroeder et al. 2009), and (2) shifts in and gray whale body condition indicates the the copepod zooplankton species toward com- potential influence of upwelling strength in the munities lacking lipid content in Oregon during region, several other oceanographic events periods of reduced upwelling (Peterson et al. occurred in parallel to the weak upwelling from 2017). Although our analysis is limited to three 2016 to 2018, intensifying negative conditions years of gray whale body condition data relative (Appendix S2: Fig. S5). Starting in November to upwelling strength, an intriguing relationship 2013, anomalously warm conditions (event is highlighted that requires further data collec- known as The Blob) were observed in the entire tion and analysis to confirm or deny the correla- CCS, reaching its extreme in mid-2016 (Gente- tion. mann et al. 2017). In addition, the years 2015 and The one-year delay between the worst 2016 were characterized by the strongest El Nino~ observed upwelling year in 2016 and evidence of ever recorded, which intensified sea surface tem- poor gray whale body condition in 2017 is possi- peratures and caused a drastic decline in primary bly due to carry-over effects of low prey quality productivity (Peterson et al. 2002, Durski et al. or quantity conditions during the previous forag- 2015, Di Lorenzo and Mantua 2016). A positive ing season. Body condition potentially remained Pacific Decadal Oscillation (PDO) phase associ- poor in 2018 due to cumulative carry-over effects ated with reduced upwelling occurred during of another low upwelling year in 2017. Such tem- 2016 and 2017, peaking in April 2016 (Di Lorenzo poral lags between environmental conditions et al. 2008, Thompson et al. 2018), and a negative and whale biology have also been documented phase of the North Pacific Gyre Oscillation between the duration of the feeding season and (NPGO) was observed for almost the entire per- the number of gray whale calves in the following iod from 2013 to 2018 (Di Lorenzo et al. 2008, year (Perryman et al. 2002), and between water Thompson et al. 2018). There is also evidence temperature and southern (E. aus- that copepod and krill communities were tralis) reproductive success (Leaper et al. 2006). affected along the CCS during the same period The 2019 unusual mortality event of gray whales (Thompson et al. 2018). Therefore, it is possible along the west coasts of Mexico, the United that mysid prey of gray whales along the Oregon States, and Canada may also be related to the coast were also negatively affected. The cumula- cumulative impacts of oceanographic conditions tive effects of these oceanographic features that reduce prey availability. Three consecutive between 2013 and 2018 may have created pro- years of relatively poor upwelling may indicate longed non-nutritious feeding conditions in the the limit of gray whale resilience to prey short- CCS, causing depleted gray whale body condi- ages; gray whales may be resilient to one or two tion in the later years of our study (2017 and years of poor prey conditions (e.g., 2016 and 2018), and possibly contributed to the current 2017), and reach an energetic reserve threshold mortality event. after three years of reduced foraging success (e.g., 2016–2018), leading to high mortality rates CONCLUSIONS due to emaciation. Indeed, upwelling conditions prior to the 1999–2000 mortality event were This study demonstrates the utility of long- stronger than the recent period, although also term photogrammetric monitoring of baleen variable. Hence, factors other than environmen- whales on their foraging grounds. We document tal fluctuations that influence prey availability variability in gray whale body condition relative may have been responsible for the previous mor- to month of the foraging season, demographic tality event, such as thresholds of population car- unit, and year. We also present correlative evi- rying capacity, as hypothesized by Rugh et al. dence to support the hypothesis that upwelling (2005) and Coyle et al. (2007). conditions may have a carry-over effect on gray

❖ www.esajournals.org 15 April 2020 ❖ Volume 11(4) ❖ Article e03094 SOLEDADE LEMOS ET AL. whale body condition in the following years. research was conducted under NOAA/NMFS permits These findings help to fill knowledge gaps #16011 and #21678 issued to John Calambokidis. Publi- regarding how baleen whales recover from fast- cation of this paper was supported, in part, by the ing periods or reproductive phases (pregnancy; Henry Mastin Graduate Student Fund. lactation) and respond to environmental varia- tion. Additionally, we place our results within LITERATURE CITED the context of the 2019 gray whale unusual mor- tality event, illustrating the value of longitudinal Acevedo-Whitehouse, K., and A. L. J. Duffus. 2009. assessment of whale body condition for popula- Effects of environmental change on wildlife health. Philosophical Transactions of the Royal Society B: tion management. In addition, our methods can – be extended to the majority of the ENP gray Biological Sciences 364:3429 3438. Angell, C. M. 2006. Body fat condition of free-ranging whale population that feed in the Alaskan Arctic right whales, Eubalaena glacialis and Eubalaena aus- to increase sample size, allow comparison by tralis. Dissertation. Boston University, Boston, Mas- habitat and prey, and assess whale response to sachusetts, USA. larger scale events, such as . We Barto, K. 2018. MuMIn: multi-model inference. R pack- recommend continued monitoring of gray whale age version 1404. https://cran.r-project.org/packa body condition to better understand the causes ge=MuMIn of nutrition-limited mortality events, and the Bates, D., M. Maechler, B. Bolker, and S. Walker. 2015. subsequent consequences and recovery of this Fitting linear mixed-effects models using lme4. – population. Furthermore, we show that BAI is Journal of Statistical Software 67:1 48. robust to scaling errors and can be applied to aer- Benoit-Bird, K. J., and M. A. McManus. 2012. Bottom- ial whale image datasets that lack scale and alti- up regulation of a pelagic community through spa- tial aggregations. Biology Letters 8:813–816. tude data, such as our 2016 dataset or archived Best, P. B., and H. Ruther. 1992. Aerial photogramme- aerial image datasets from helicopters, enabling try of southern right whales, Eubalaena australis. valuable comparisons of whale health across Journal of Zoology 228:595–614. time periods and populations to inform conser- Bograd, S. J., I. Schroeder, N. Sarkar, X. Qiu, W. J. vation efforts. Sydeman, and F. B. Schwing. 2009. Phenology of coastal upwelling in the California Current. Geo- ACKNOWLEDGMENTS physical Research Letters 36:1–5. Bradford, A. L., D. W. Weller, A. E. Punt, Y. V. Ivash- This research was funded by the NOAA National chenko, A. M. Burdin, G. R. VanBlaricom, and R. L. Marine Fisheries Service Office of Science and Technol- Brownell. 2012. Leaner leviathans: body condition ogy Ocean Acoustics Program, Oregon Sea Grant Pro- variation in a critically endangered whale popula- gram Development funds, and the Oregon State tion. Journal of Mammalogy 93:251–266. University (OSU) Marine Mammal Institute. We are Braithwaite, J. E., J. J. Meeuwig, and M. R. Hipsey. thankful for the support of Brazil’s Science Without 2015a. Optimal migration energetics of humpback Borders Program, Brazil’s CNPq, the Harvard Laspau whales and the implications of disturbance. Con- Institute, and the Mamie Markham Research Award servation Physiology 3:1–15. (Hatfield Marine Science Center/OSU), which provided Braithwaite, J. E., J. J. Meeuwig, T. B. Letessier, K. C. S. financial aid and support for LSL. We are grateful to Jenner, and A. S. Brierley. 2015b. From sea ice to research assistants Danielle Marie Balderson and Tay- blubber: linking whale condition to krill abun- lor Mock for assistance with photogrammetry analysis, dance using historical whaling records. Polar Biol- Joseph D. Maurer (Statistics Department/OSU), Lucas ogy 38:1195–1202. Kopecky Bobadilla (OSU) and Mariana Bonfim (Tem- Burnett, J., L. S. Lemos, D. Barlow, M. Wing, T. Chan- ple University) for statistical support, Isaac Schroeder dler, and L. Torres. 2018. Estimating morphometric (Southwest Fisheries Science Center/NOAA) for advice attributes of baleen whales with photogrammetry on upwelling index analysis, John Calambokidis and from small UAS: a case study with blue and gray Alie Perez at the Cascadia Research Collective for assis- whales. Marine Mammal Science 35:108–139. tance with the photo-identification analysis, and Deb- Burnham, K. P., D. R. Anderson, and K. P. Huyvaert. bie Steel and Scott Baker (OSU) for assistance with 2011. AIC model selection and multimodel infer- genetic analyses. We are extremely grateful to Amanda ence in behavioral ecology: some background, Bradford and other anonymous reviewer for insightful observations, and comparisons. Behavioral Ecol- comments that improved this manuscript. This ogy and Sociobiology 65:23–35.

❖ www.esajournals.org 16 April 2020 ❖ Volume 11(4) ❖ Article e03094 SOLEDADE LEMOS ET AL.

Calambokidis, J., et al. 2002. Abundance, range and and the influence of the 2009–2010 El Nino.~ Ocean movements of a feeding aggregation of gray Dynamics 65:1643–1663. whales (Eschrichtius robustus) from California to Gallagher, D., M. Visser, D. Sepulveda, R. N. Pierson, southeastern Alaska in 1998. Journal of Cetacean T. Harris, and S. B. Heymsfield. 1996. How useful Research and Management 4:267–276. is body mass index for comparison of body fatness Calambokidis, J., J. L. Laake, and A. Klimek. 2012. across age, sex, and ethnic groups. American Jour- Updated analysis of abundance and population nal of Epidemiology 143:228–239. structure of seasonal gray whales in the Pacific Gentemann, C. L., M. R. Fewings, and M. Garcıa- Northwest, 1998–2010. International Whaling Reyes. 2017. Satellite sea surface temperatures Commission, Cambridge, UK. along the West Coast of the United States during Christiansen, F., A. M. Dujon, K. R. Sprogis, J. P. Y. the 2014–2016 northeast Pacific marine heat wave. Arnould, and L. Bejder. 2016. Noninvasive Geophysical Research Letters 44:312–319. unmanned aerial vehicle provides estimates of the Gittleman, J. L., and S. D. Thompson. 1988. Energy energetic cost of reproduction in humpback allocation in mammalian reproduction. American whales. Ecosphere 7:e01468. Zoologist 28:863–875. Christiansen, F., G. A. Vikingsson, M. H. Rasmussen, Hammond, P. S., S. A. Mizroch, and G. P. Donovan. and D. Lusseau. 2013. Minke whales maximise 1990. Individual recognition of cetaceans: use of energy storage on their feeding grounds. Journal of photo-identification and other techniques to esti- Experimental Biology 216:427–436. mate population parameters. International Whal- Christiansen, F., F. Vivier, C. Charlton, R. Ward, A. ing Commission, Cambridge, UK. Amerson, S. Burnell, and L. Bejder 2018. Maternal Harrison, X. A., J. D. Blount, R. Inger, D. R. Norris, body size and condition determine calf growth and S. Bearhop. 2011. Carry-over effects as drivers rates in southern right whales. Marine Ecology of fitness differences in animals. Journal of Animal Progress Series 592:267–281. Ecology 80:4–18. Coyle, K. O., B. Bluhm, B. Konar, A. Blanchard, and R. Harvey, C., et al. 2018. Ecosystem Status Report of the C. Highsmith. 2007. Amphipod prey of gray California Current for 2018: a Summary of Ecosys- whales in the northern : comparison of tem Indicators Compiled by the California Current biomass and distribution between the 1980s and Integrated Ecosystem Assessment Team (CCIEA). 2002–2003. Deep-Sea Research II 54:2906– NOAA Fisheries, Northwest Fisheries Science 2918. Center, Seattle, Washington, USA. Cury, P. M., Y.-J. Shin, B. Planque, J. M. Durant, J.-M. Heymans, J. J., M. Coll, S. Libralato, L. Morissette, and Fromentin, S. Kramer-Schadt, N. C. Stenseth, M. V. Christensen. 2014. Global Patterns in ecological Travers, and V. Grimm. 2008. Ecosystem oceanog- indicators of marine food webs: a modelling raphy for global change in fisheries. Trends in Ecol- approach. PLOS ONE 9:e95845. ogy and Evolution 23:338–346. Hunter, T. C. R. 1978. The importance of a relative short- Darling, J. D. 1984. Gray whales off , age of food in animal ecology. Oecologia 33:71–86. . Pages 267–287 in M. L. Jones, S. Jones, M. L., and S. L. Swartz. 2002. Gray Whale, L. Swartz, and S. Leatherwood, editors. The gray Eschrichtius robustus. Pages 524–536 In Encyclope- whale Eschrichtius robustus. Academic Press, dia of marine . Academic Press, San Orlando, Florida, USA. Diego, California, USA. Di Lorenzo, E., et al. 2008. North Pacific Gyre Oscilla- Kasuya, T. 1995. Overview of cetacean life histories: an tion links ocean climate and ecosystem change. essay in their evolution. Pages 481–498 in A. S. Blix, Geophysical Research Letters 35:1–6. L. Walløe and Ø. Ulltang, editors. Whales, seals, Di Lorenzo, E., and N. Mantua. 2016. Multi-year per- fish and man. Elsevier Science, Amsterdam, The sistence of the 2014/15 North Pacific marine heat- Netherlands. wave. Nature Climate Change 6:1042–1048. Koopman, H. N. 1998. Topographical distribution of Durban, J. W., M. J. Moore, G. Chiang, L. S. Hickmott, the blubber of harbor ( pho- A. Bocconcelli, G. Howes, P. A. Bahamonde, W. L. coena). Journal of Mammalogy 79:260–270. Perryman, and D. J. LeRoi. 2016. Photogrammetry Kuznetsova, A., P. B. Brockhoff, and R. H. B. Chris- of blue whales with an unmanned hexacopter. tensen 2017. LmerTest package: tests in linear Marine Mammal Science 32:1510–1515. mixed effects models. Journal of Statistical Soft- Durski, S. M., A. L. Kurapov, J. S. Allen, P. M. Kosro, ware 82:1–26. G. D. Egbert, K. Shearman, and J. A. Barth. 2015. Lang, A. R., J. Calambokidis, J. Scordino, V. L. Pease, A. Coastal ocean variability in the US Pacific North- Klimek,V.N.Burkanov,P.Gearin,D.I.Litovka,K.M. west region: seasonal patterns, winter circulation, Robertson,B.R.Mate,J.K.Jacobsen,andB.L.Taylor.

❖ www.esajournals.org 17 April 2020 ❖ Volume 11(4) ❖ Article e03094 SOLEDADE LEMOS ET AL.

2014. Assessment of genetic structure among east- Mooers, C. N. K., C. A. Collins, and R. L. Smith. 1976. ern North Pacific gray whales on their feeding The dynamic structure of the frontal zone in the grounds. Marine Mammal Science 30:1473–1493. coastal upwelling region off Oregon. Journal of Le Boeuf, B. J., H. Perez-Cortes, J. Urban, B. R. Mate, Physical Oceanography 6:3–21. and F. Ollervides. 2000. High gray whale mortality Moore, S. E., J. Urban, W. L. Perryman, F. Gulland, H. and low recruitment in 1999: potential causes and Perez-Cortes, P. R. Wade, L. Rojas-Bracho, and T. implications. Journal of Cetacean Research and Rowles. 2001. Are gray whales hitting "K" hard? Management 2:85–99. Marine Mammal Science 17:954–958. Leaper, R., J. Cooke, P. Trathan, K. Reid, V. Rowntree, Nakagawa, S., and H. Schielzeth. 2013. A general and and R. Payne. 2006. Global climate drives southern simple method for obtaining R2 from generalized right whale (Eubalaena australis) population linear mixed effects models. Methods in Ecology dynamics. Biology Letters 2:289–292. and Evolution 4:133–142. Lenth, R. 2019. emmeans: estimated Marginal Means, National Academies of Sciences, Engineering, and aka Least-Squares Means. R package version 1.1.3. Medicine. 2017. Approaches to understanding the https://cran.r-project.org/package=emmeans cumulative effects of stressors on marine mam- Lockyer, C. 1981. Estimation of the energy costs of mals. The National Academies Press, Washington, growth, maintenance and reproduction in the D.C., USA. female , (Balaenoptera acutorostrata), Nattrass, S., and D. Lusseau. 2016. Using resilience to from the Southern Hemisphere. Report of the Inter- predict the effects of disturbance. Scientific Reports national Whaling Commission 31:337–343. 6:25539. Lockyer, C. 1986. Body fat condition in northeast at New,L.F.,D.J.Moretti,S.K.Hooker,D.P.Costa,andS. antic whales, Balaenoptera physalus, and its rela- E. Simmons. 2013. Using energetic models to inves- tionship with reproduction and food resource. tigate the survival and reproduction of beaked Canadian Journal of Fisheries and Aquatic Sciences whales (family Ziphiidae). PLOS ONE 8:e68725. 43:142–147. Newell, C. L., and T. J. Cowles. 2006. Unusual gray Lockyer, C. 1987a. Evaluation of the role of fat reserves whale Eschrichtius robustus feeding in the summer in relation to the ecology of North Atlantic fin and of 2005 off the central Oregon Coast. Geophysical sei whales. Pages 183–203 in A. C. Huntley, D. P. Research Letters 33:1–5. Costa, G. A. J. Worthy and M. A. Castellini, editors. NOAA. 2019. 2019 Gray Whale Unusual Mortality Approaches to marine mammal energetics. Society Event along the West Coast. Marine life in distress. for Marine Mammalogy, Lawrence, Kansas, USA. https://www.fisheries.noaa.gov/national/marine- Lockyer, C. 1987b. The relationship between body fat, life-distress/2019-gray-whale-unusual-mortality- food resource and reproductive energy costs in event-along-west-coast north Atlantic fin whales (Balaenoptera physalus). Perryman, W. L., M. A. Donahue, P. C. Perkins, and S. Symposia of the Zoological Society of London B. Reilly. 2002. Gray whale calf production 1994– 57:343–361. 2000: Are observed fluctuations related to changes Lockyer, C. 2007. All creatures great and smaller: a in seasonal ice cover? Marine Mammal Science study in cetacean life history energetics. Journal of 18:121–144. the Marine Biological Association of the United Perryman, W. L., and M. S. Lynn. 2002. Evaluation of Kingdom 87:1035–1045. nutritive condition and reproductive status of Machovsky-Capuska, G. E., and D. Raubenheimer. 2020. migrating gray whales (Eschrichtius robustus) based The nutritional ecology of marine apex predators. on analysis of photogrammetric data. Journal of Annual Review of Marine Science 12:361–387. Cetacean Research and Management 4:155–164. Miller, C. A., P. B. Best, W. L. Perryman, M. F. Baum- Peterson, W. T., J. L. Fisher, C. A. Morgan, S. M. gartner, and M. J. Moore. 2012. Body shape Zeman, B. J. Burke, and K. C. Jacobson. 2018. changes associated with reproductive status, nutri- Ocean ecosystem indicators of salmon marine sur- tive condition and growth in right whales Eubal- vival in the northern California current. NOAA aena glacialis and E. australis. Marine Ecology Fisheries, Northwest Fisheries Science Center, Seat- Progress Series 459:135–156. tle, Washington, USA. Miller, C. A., D. Reeb, P. B. Best, A. R. Knowlton, M. Peterson, W. T., J. L. Fisher, P. T. Strub, X. Du, C. Risien, W. Brown, and M. J. Moore. 2011. Blubber thick- J. Peterson, and C. T. Shaw. 2017. The pelagic ness in right whales Eubalaena glacialis and Eubal- ecosystem in the Northern California current off aena australis related with reproduction, life history Oregon during the 2014–2016 warm anomalies status and prey abundance. Marine Ecology Pro- within the context of the past 20 years. Journal of gress Series 438:267–283. Geophysical Research: Oceans 122:7267–7290.

❖ www.esajournals.org 18 April 2020 ❖ Volume 11(4) ❖ Article e03094 SOLEDADE LEMOS ET AL.

Peterson, W. T., J. E. Keister, and L. R. Feinberg. 2002. The recommendations for researchers in the United effects of the 1997–99 El Nino/La~ Nina~ events on States. Journal of Unmanned Vehicle Systems 4:1– hydrography and zooplankton off the central Ore- 14. gon coast. Progress in Oceanography 54:381–398. Sumich, J. L. 2014. E. robustus: the biology and human Pettis, H. M., R. M. Rolland, P. K. Hamilton, S. Brault, history of gray whales. Whale Cove Marine Educa- A. R. Knowlton, and S. D. Kraus. 2004. Visual tion, Corvallis, Oregon, USA. health assessment of North Atlantic right whales Thomas, V. G. 1990. Control of reproduction in animal (Eubalaena glacialis) using photographs. Canadian species with high and low body fat reserves. Pages Journal of Zoology 82:8–19. 27–41 in R. E. Frisch, editor. Adipose tissue and Pettis, H. M., R. M. Rolland, P. K. Hamilton, A. R. reproduction (Progress in reproductive biology Knowlton, E. A. Burgess, and S. D. Kraus. 2017. and medicine). S. Karger, Basel, Switzerland. Body condition changes arising from natural fac- Thompson, A. R., et al. 2018. State of the California tors and fishing gear entanglements in North Current 2017–18: still not quite normal in the north Atlantic right whales Eubalaena glacialis. Endan- and getting interesting in the south. CalCOFI gered Species Research 32:237–249. Reports 59:1–66. R Core Team. 2019. R: a language and environment for Tollefson, T. N., L. A. Shipley, W. L. Myers, D. H. Keis- statistical computing. R Foundation for Statistical ler, and N. Dasgupta. 2010. Influence of summer Computing, Vienna, Austria. and autumn nutrition on body condition and Rice, D. W., and A. A. Wolman. 1971. The life history reproduction in lactating mule deer. Journal of and ecology of the gray whale (Eschrichtius robus- Wildlife Management 74:974–986. tus). American Society of Mammalogists Special Villegas-Amtmann, S., L. K. Schwarz, J. L. Sumich, Publications 3:1–142. and D. P. Costa. 2015. A bioenergetics model to Rugh, D. J., R. C. Hobbs, J. A. Lerczak, and J. M. Brei- evaluate demographic consequences of disturbance wick. 2005. Estimates of abundance of the eastern in marine mammals applied to gray whales. Eco- North Pacific stock of gray whales (Eschrichtius sphere 6:1–19. robustus) 1997–2002. Journal of Cetacean Research Weller, D. W., S. Bettridge, R. L. Jr Brownell, J. L. and Management 7:1–12. Laake, J. E. Moore, P. E. Rosel, B. L. Taylor, and P. Schroeder, I. D., W. J. Sydeman, N. Sarkar, S. A. R. Wade. 2013. Report of the National Marine Fish- Thompson, S. J. Bograd, and F. B. Schwing. 2009. eries Service gray whale stock identification work- Winter pre-conditioning of seabird phenology in shop. U.S. Department of Commerce, , the California current. Marine Ecology Progress California, USA. Series 393:211–233. Williams, R., G. A. Vikingsson, A. Gislason, C. Lock- Seyboth, E., K. R. Groch, L. D. Rosa, K. Reid, P. A. C. yer, L. New, L. Thomas, and P. S. Hammond. 2013. Flores, and E. R. Secchi. 2016. Evidence for density-dependent changes in body (Eubalaena australis) reproductive success is influ- condition and pregnancy rate of North Atlantic fin enced by krill (Euphausia superba) density and cli- whales over four decades of varying environmen- mate. Scientific Reports 6:28205. tal conditions. ICES Journal of Marine Science Sinclair, A. R. E., and C. J. Krebs. 2002. Complex 70:1273–1280. numerical responses to top-down and bottom-up Zimushko, V. V. 1969. Some data on the biology of processes in vertebrate populations. Philosophical gray whales. Pages 93–97 in V. A. Arseniev, B. A. Transactions of the Royal Society of London. Series Zenkovich, and K. K. Chapskii editors. Marine B: Biological Sciences 357:1221–1231. Mammals. Akademii Nauk, Moscow, Russia. Smith, C. E., S. T. Sykora-Bodie, B. Bloodworth, S. M. Zimushko, V.V., and M. V.Ivashin. 1980. Some results of Pack,T.R.Spradlin,andN.R.LeBoeuf.2016. U.S.S.R. investigations and whaling of gray whales Assessment of known impacts of unmanned aerial (Eschrichtius robustus, Lilljeborg 1861). Report of systems (UAS) on marine mammals: data gaps and the International Whaling Commission 30:237–246.

SUPPORTING INFORMATION

Additional Supporting Information may be found online at: http://onlinelibrary.wiley.com/doi/10.1002/ecs2. 3094/full

❖ www.esajournals.org 19 April 2020 ❖ Volume 11(4) ❖ Article e03094