European shag provisioning foraging dives

Foraging dives in birds and mammals involve complex physiological and behavioural adaptations to cope with the breaks in normal respiration. Optimal dive strategies should maximise the proportion of time spent under water actively foraging versus the time spent on the surface. Oxygen loading and carbon-dioxide dumping carried out on the surface could involve recovery from the consequences of the last dive and/or preparation in anticipation of the next dive depth and duration. However, few studies have properly explored the causal pattern of effects within such dive cycles, which is crucial prior to any assessment of optimal dive strategies. Using Time Depth Recorders and Global Positioning System loggers, we recorded over 42,000 dives by 39 pairs of male and female European shags (Phalacrocorax aristotelis). Dives either involved a straight descent and ascent, presumably reflecting an unsuccessful search for prey, or a descent followed by horizontal movement followed by an ascent, presumably reflecting active hunting pursuit of pelagic prey. Males were larger than females, but we were unable to distinguish between sex effects and the non-linear effects of body mass on dive behaviour. Path analysis showed that within-individual dive-to-dive variation in surface times can best be explained as recovery from the previous dive. As expected in a pelagic hunter with unpredictable dive durations, there was no evidence of anticipatory preparation of oxygen stores in pre-dive surface durations. Among-individual variation in dives showed that body mass directly affected descent durations, but individual variation in all other dive and surface durations was driven by variation in descent duration, suggesting a critical role for dive depth in overcoming body-mass dependent effects of hydrodynamic/wave drag and buoyancy. Our analyses tests for the first time certain critical assumptions for studies assessing optimal dive strategies in birds and mammals, thereby revealing new details and avenues for research concerning adaptive diving behaviour.

Methods
Study site:

The Sklinna archipelago, situated about 20km off the coast of Vikna in Trøndelag, Central-Norway (65°12’N 10°59’E), holds one of the largest shag colonies in Norway with ca. 2,000 breeding pairs in 2017.

Animal welfare note:

Capture and handling of birds were approved by the Norwegian Environment Agency (2013/2306, 2014/2179, 2015/3042, 2016/3366, 2017/4069, 2018/607) and the Norwegian Animal Research Authority (5148-2013/34672 (years 2013-2015), 7484-2015/55385 (years 2015-2017), 12163-2017/67495 (years 2017-2019). All handling of birds were done by Felasa C approved persons, or under supervision of such persons. 

Data Collection:

The fieldwork was conducted during June-July 2013-2018, including 78 birds (39 pairs) over 6 different breeding seasons. Chick rearing shags were chosen based on their nest accessibility and how ‘protective’ the pairs were, as those that aggressively stayed around the nest were easier to capture/recapture. Parental birds were fitted with loggers when nestlings were approximately 5-35 days old. Nestling age was determined using morphological criteria determined from control nests (from nesting areas in similar habitat within the Sklinna colony) checked every fifth day. The shags were captured and then recaptured at their nest by hand or using snares. Each individual was fitted with a GPS-logger (i-gotU GT-120, Mobile Action Technology, re-fitted in heat-shrink tubes) and Time Depth Recorders (TDR, G5, CEFAS Technology). TDR-loggers were attached to the GPS logger prior to instrumentation, and the loggers were attached to 3-4 middle tail feathers using TESA® tape. The maximum logger deployment weight was 30.6g, corresponding to 1.6% and 1.8% of mean body mass of males and females, respectively. The GPS loggers recorded location (± 10m) every 30 sec, and the TDR recorded water depth below (± 0.1m) every second. The loggers were removed during recapture after approximately 2-5 days. Deployment of loggers normally required less than 3 min of handling and retrieval less than 10 min, and no disturbance effects were noted in either adults or their chicks. In cases where there were signs of parental disturbance in the form of decreased nestling provisioning, then the second parent was not captured, and so these pairs were not included in the study.

The sex of adults was determined initially by body size features and ultimately via their vocalizations (Koffijberg and Van Eerden, 1995; Cramp and Simmons, 1977), because males and females made very distinct types of calls whilst defending the nest at our approach (Snow, 1960). At capture, body mass was obtained using a Pesola spring balance (accuracy ± 10g). Both adults in the pair were fitted with recording instruments during the same breeding season, although not overlapping in time, usually within only a few days of each other. At recapture, biometric measures were obtained (wing length (ruler ± 1mm), head and bill length (digital calliper ± 1mm) and body mass (see above)). Adult female average mass was 1610g (range 1370-1860g), whilst average adult male mass was 1920g (range 1660-2280g). Growth data (i.e. capture- recapture difference in chick weight) were collected for all nests during the time of recording and these measurements were compared to the control area within the same colony (see above) containing 50 nests where adults were not fitted with loggers. There was rarely indication of parents reducing their provisioning rates or changing any patterns of nestling feeding while fitted with loggers. There were no obvious differences in the number of surviving chicks in experimental versus neighbouring control nests, aside from impeded survival due to gull predation.

Data Handling:

Data handling and simulations was programmed in R 3.5.1 (R Core Team, 2018) and the TDR raw data was analysed with the package DiveMove (Luque, 2007). The total number of dives in this study was 46,103. The surface for dives was calibrated at ±1m, so that no dive movement less than 1m depth was counted as a real dive, which helped to remove possible non-foraging ‘cleaning’ dives (Christensen-Dalsgaard et al., 2017). The time submerged during foraging dives was divided into vertical descents and ascents involving <1m horizontal movement versus >1m horizontal movements. Such horizontal movement was calibrated with the package DiveMove’s Zero-Offset Corrected (ZOC) method (Luque, 2007), smoothed using ±4m depth filters, and registered as dive bottom duration. Dives were classified into two types according to the presence/absence of this horizontal dive bottom duration: U-shaped (with a horizontal dive bottom) versus V-shaped (with no horizontal dive bottom) dives (see Supplementary Materials A1). Pre- and post-dive durations at the surface longer than 360s were used to separate dive bouts (i.e. distinct sequences of successive dives at one location) whenever surface durations were too long to be explained by simple replenishment of O2 storages or momentary resting within a dive bout. GPS coordinates for each dive were assigned as the closest coordinates recorded within 30 secs before and/or after the dive (i.e. GPS locations could not be recorded during dives). GPS data were processed using R library ggmap (Kahle and Wickham, 2013). Merging, combining, and sorting of the data set was performed using the package dplyr (Wickham et al., 2018), and plots were generated by ggplot2 (Wickham, 2016). A total 24 ‘locations’ were identified as distinct places where most dives occurred (i.e. clusters of dives surrounded by areas with no dives), and distinguished as areas of uniform average depth and foraging conditions as determined from a topographical base map by Kystverket (https://kart.kystverket.no/). GPS coordinates were thus abbreviated to 2 decimal labels based on these dive locations, whilst the geographical size and number of observations varied between locations.

鸟类与哺乳类的觅食潜水，涉及应对正常呼吸中断的复杂生理与行为适应机制。最优潜水策略应最大化水下主动觅食时长与水面停留时长的占比。水面阶段的氧气加载与二氧化碳排出，既可能用于恢复前一次潜水带来的生理损耗，也可能为下一次潜水的深度与时长提前做好准备。然而，目前鲜有研究深入探究此类潜水周期内的效应因果模式，而这一内容正是评估最优潜水策略的核心前提。本研究借助时间深度记录仪（Time Depth Recorders, TDR）与全球定位系统（Global Positioning System, GPS）记录器，对39对雌雄欧洲粗颈鸬鹚（*Phalacrocorax aristotelis*）的超42000次潜水行为进行了采集。潜水行为可分为两类：一类为直下直上式，推测对应未成功搜寻猎物的潜水；另一类则为先下潜、再水平移动、最后上浮的模式，推测对应主动追捕远洋猎物的觅食潜水。雄性个体体型大于雌性，但本研究无法区分性别效应与体重对潜水行为的非线性影响。路径分析（Path analysis）结果显示，个体内单次潜水间的水面停留时长差异，最适宜解释为前一次潜水后的生理恢复过程。正如潜水时长不可预测的远洋猎手所预期的那样，本研究未发现潜水前水面停留时长存在氧气储备预准备的证据。个体间的潜水差异分析显示，体重直接影响下潜时长，但其余所有潜水与水面停留时长的个体差异均由下潜时长的差异驱动，这表明潜水深度在抵消体重依赖的流体动力/波浪阻力与浮力效应中发挥关键作用。本研究首次针对鸟类与哺乳类最优潜水策略评估中的若干关键假设进行了检验，从而为适应性潜水行为的相关研究揭示了新的细节与方向。

### 研究地点
斯克利纳群岛（Sklinna archipelago）位于挪威中部特伦德拉格郡维克纳海岸外约20公里处（北纬65°12′，东经10°59′），是挪威最大的粗颈鸬鹚繁殖群落之一，2017年时约有2000对繁殖个体。

### 动物伦理说明
鸟类的捕捉与处理流程已获得挪威环境署（审批号：2013/2306、2014/2179、2015/3042、2016/3366、2017/4069、2018/607）与挪威动物研究委员会（审批号：5148-2013/34672，适用2013-2015年；7484-2015/55385，适用2015-2017年；12163-2017/67495，适用2017-2019年）的批准。所有鸟类处理工作均由持有Felasa C资质的人员完成，或在该类人员的监督下进行。

### 数据采集
野外工作于2013-2018年的6-7月开展，共覆盖6个繁殖季的78只个体（39对）。研究选取育雏期的粗颈鸬鹚时，以巢穴可接近性与配对的“护巢性”为标准——会主动守卫巢穴的个体更易被捕捉与重捕。当雏鸟约5-35日龄时，为亲鸟佩戴记录器。雏鸟日龄通过形态学标准判定，该标准源自每5天巡查一次的对照巢（取自斯克利纳群落内相似生境的筑巢区域）。研究人员通过徒手或使用套索的方式，在巢穴处捕捉并重新捕获粗颈鸬鹚。为每只个体佩戴GPS记录器（型号i-gotU GT-120，Mobile Action Technology生产，外部加装热缩管）与时间深度记录仪（TDR，型号G5，CEFAS Technology生产）。TDR记录仪在安装前先固定于GPS记录器上，随后使用TESA®胶带将整套记录器固定于3-4根中部尾羽上。记录器的最大部署重量为30.6g，分别对应雄性与雌性平均体重的1.6%与1.8%。GPS记录器每30秒记录一次位置信息（精度±10m），TDR则每秒记录一次水下深度（精度±0.1m）。约2-5天后重捕时取下记录器。安装记录器的操作通常耗时不足3分钟，取回过程耗时不足10分钟，未观察到对成鸟或雏鸟造成干扰的迹象。若观察到亲鸟出现育雏量下降等受干扰迹象，则不会捕捉该配对中的另一只亲鸟，此类配对也不会被纳入本研究。

成鸟性别最初通过体型特征判定，最终通过鸣叫声确认（Koffijberg与Van Eerden, 1995；Cramp与Simmons, 1977）——当研究人员靠近巢穴时，雌雄个体的鸣叫声存在显著差异（Snow, 1960）。捕捉时使用Pesola弹簧秤称量体重（精度±10g）。同一繁殖季内，配对的两只成鸟均会佩戴记录器，但安装时间不重叠，通常间隔仅数日。重捕时采集生物测量数据：翅长（使用直尺测量，精度±1mm）、头喙长（使用数显卡尺测量，精度±1mm）与体重（同前）。成年雌性的平均体重为1610g（范围1370-1860g），成年雄性平均体重为1920g（范围1660-2280g）。本研究在记录期间为所有巢穴采集雏鸟生长数据（即两次捕捉间的雏鸟体重差值），并将其与同群落内的对照区域（见前文）进行对比，该对照区域包含50个未为成鸟佩戴记录器的巢穴。几乎未观察到佩戴记录器的亲鸟降低育雏频率或改变雏鸟喂食模式的情况。除了海鸥捕食导致的存活障碍外，实验组与邻近对照组巢穴的存活雏鸟数量无显著差异。

### 数据处理
数据处理与模拟程序基于R 3.5.1编写（R Core Team, 2018），TDR原始数据使用DiveMove软件包进行分析（Luque, 2007）。本研究共记录到46103次潜水。潜水的水面阶段校准阈值设为±1m，即水深不足1m的潜水活动不被计为有效潜水，这一设置有助于排除非觅食性的“清洁”潜水（Christensen-Dalsgaard等, 2017）。觅食潜水的水下时长被分为两类：水平移动距离<1m的垂直下潜与上浮阶段，以及水平移动距离>1m的阶段。此类水平移动通过DiveMove软件包的零偏移校正（Zero-Offset Corrected, ZOC）方法进行校准，辅以±4m的深度滤波进行平滑处理，并记为潜水底部时长。根据是否存在该水平潜水底部时长，潜水可分为两类：U型潜水（存在水平底部阶段）与V型潜水（无水平底部阶段）（详见补充材料A1）。当水面停留时长过长，无法仅用氧气储备补充或潜水序列内的短暂休息解释时，我们以时长超过360秒的潜水前、后水面停留时长来划分潜水序列（即同一地点连续潜水的独立集群）。每次潜水的GPS坐标被赋值为潜水前后30秒内记录的最接近坐标（即潜水过程中无法记录GPS位置）。GPS数据使用R语言库ggmap进行处理（Kahle与Wickham, 2013）。数据集的合并、整合与排序使用dplyr软件包完成（Wickham等, 2018），图表则通过ggplot2生成（Wickham, 2016）。本研究共识别出24个主要潜水发生地（即被无潜水区域环绕的潜水集群），并通过挪威海岸管理局（Kystverket）的地形底图（https://kart.kystverket.no/）判定，将其划分为平均水深与觅食条件均一的区域。基于这些潜水地点，GPS坐标被简化为两位小数标签，不同地点的地理范围与观测样本量存在差异。