Enhancing Marine Wildlife Observations: The Application of Tethered Balloon Systems and Advanced Imaging Sensors for Sustainable Marine Energy Development

Enhancing Marine Wildlife Observations: The Application of Tethered Balloon Systems and Advanced Imaging Sensors for Sustainable Marine Energy Development | Research Square window.SnipcartSettings = { analytics: { enabled: false } }; (function() { var accessVector = localStorage.getItem('access_vector') || ''; window.dataLayer = window.dataLayer || []; if (accessVector) { window.dataLayer.push({ user: { profile: { profileInfo: { snid: accessVector } } } }); } })(); (function(w,d,s,l,i){w[l]=w[l]||[];w[l].push({'gtm.start':new Date().getTime(),event:'gtm.js'});var f=d.getElementsByTagName(s)[0],j=d.createElement(s),dl=l!='dataLayer'?'&l='+l:'';j.async=true;j.src='https://www.googletagmanager.com/gtm.js?id='+i+dl;f.parentNode.insertBefore(j,f);})(window,document,'script','dataLayer','GTM-K279D39R'); Browse Preprints In Review Journals COVID-19 Preprints AJE Video Bytes Research Tools Research Promotion AJE Professional Editing AJE Rubriq About Preprint Platform In Review Editorial Policies Our Team Advisory Board Help Center Sign In Submit a Preprint Cite Share Download PDF Research Article Enhancing Marine Wildlife Observations: The Application of Tethered Balloon Systems and Advanced Imaging Sensors for Sustainable Marine Energy Development Alicia Amerson, Darielle Dexheimer This is a preprint; it has not been peer reviewed by a journal. https://doi.org/ 10.21203/rs.3.rs-5349011/v1 This work is licensed under a CC BY 4.0 License Status: Published Journal Publication published 03 Apr, 2025 Read the published version in Marine Biology → Version 1 posted 5 You are reading this latest preprint version Abstract This study investigates the capabilities of a tethered balloon system (TBS) for detecting and monitoring marine wildlife, primarily focusing on gray whales ( Eschrichtius robustus ) and various avian species. Over 55.7 h of aerial and surface footage were collected, yielding significant findings regarding the detection rates of marine mammals and seabirds. A total of 59 gray whale, 100 avian, and 6 indistinguishable marine mammal targets were identified by the airborne TBS, while surface-based observations recorded 1,409 gray whales, 1,342 avian targets, and several other marine mammals. When the airborne and surface cameras were operating simultaneously, 21% of airborne whale and 34% of airborne avian detections were captured with the airborne TBS camera and undetected with the surface-based camera. The TBS was most effective at altitudes between 50 to 200 m above ground, with variable-pitch scanning patterns providing superior detection of whale blows compared to fixed-pitch and loitering methods. Notably, instances of airborne detections not corroborated by surface observations underscore the benefits of combining aerial monitoring with traditional survey techniques. Additionally, the integration of machine-learning (ML) algorithms into video analysis enhances our capacity for processing large datasets, paving the way for real-time wildlife monitoring. Of the total number of blows detected by an ML algorithm, the percentage of blows identified by a human analyst was greater than that uniquely detected by the algorithm. Notably, more unique detections by the ML algorithm occurred during daylight, suggesting that sun artifacts may hinder human detection performance, thereby highlighting the added value of ML under these conditions. This research lays the groundwork for future studies in marine biodiversity monitoring, emphasizing the importance of innovative aerial surveillance technologies and advanced imaging methodologies in understanding species behavior and informing conservation strategies for sustainable marine energy, offshore wind development, and other marine resource management efforts. Tethered Balloon System Marine Wildlife Detection Machine Learning Infrared Sensors Human Observation Figures Figure 1 Figure 2 Figure 3 Figure 4 Figure 5 Figure 6 Figure 7 Figure 8 Figure 9 Figure 10 Figure 11 Figure 12 Figure 13 Figure 14 Figure 15 Figure 16 Introduction As various marine ecosystems increasingly face considerations for marine energy (ME) development, comprehensive environmental assessments have become necessary (Eaves et al. 2022). These assessments aim to evaluate the potential impacts of new technologies on energetically dynamic marine environments, particularly focusing on how marine wildlife interacts with ME devices. Disturbances from ME installations may lead to alterations in habitat use, behavioral changes, and shifts in population dynamics for key species, making it essential to provide field-tested recommendations for implementing environmental monitoring technologies and methodologies to understand these interactions (Amerson et al. 2022; Haxel et al. 2022; Hemery et al. 2022a; Hemery et al. 2022b; Reilly et al. 2022; Staines et al. 2022). There is a growing interest in utilizing cost-effective monitoring technologies that can also be implemented with minimal to no impact on wildlife (Gibbs et al. 1999; Thomas et al. 2011; Christie et al. 2016; Marvin et al. 2016; Stephenson 2020). While these technologies are more easily adapted for terrestrial wildlife, they also apply to observations of marine wildlife interactions with ME systems (Bicknell et al. 2016; Danovaro et al. 2016; Wang et al. 2019). However, gaps remain regarding the efficacy of aerial monitoring methods, particularly in varied marine and coastal conditions (Amerson et al. 2023). These gaps include scanning patterns, flight longevity, wind conditions, altitude variations, and comparisons between human observations and machine-learning (ML) detection. In a previous study, the research team conducted an initial flight trial of a tethered balloon system (TBS) and sensor package in La Porte, Texas (TX) (Amerson et al. 2023). During this study, no marine wildlife species were present. Therefore, there was a need to perform flights along a coastline with a known migratory path and a larger diversity of marine species. Furthermore, a consideration for the second deployment was to find an environment similar to areas of future ME development. An additional consideration was made to include flights during daylight and nighttime hours to evaluate the use of a TBS for ME environmental assessment for 24 h. Lastly, accumulating a large dataset from the effort in La Porte, TX presented a challenge associated with aerial monitoring: increased processing and analysis time by humans. The need for reliable ML applications may reduce this processing and analysis time, but these systems are currently under development and require reliable data libraries (Kellenberger et al. 2018; Corcoran et al. 2021; Aguilar-Lazcano et al. 2023; Clarfeld et al. 2023; Sharma et al. 2023). A reliable source of data for ML may be obtained from analysis that has been processed through human observations (Stewart et al. 2023; Barlow et al. 2024). This study aimed to address these gaps by integrating a TBS equipped with advanced imaging sensors to observe marine wildlife along the California coast, a critical migratory corridor for species such as gray whales and other marine mammals. Additionally, this study evaluated data collected by TBS sensors and human observations, reviewed various scan patterns and loitering altitudes, and leveraged ML programs to detect whale blows. By implementing ML, the goal was to compare the time and cost of data processing and analysis between humans and ML programs. The significance of this research lies in its potential to provide technological and methodological recommendations for regulatory decision-makers and to contribute to diverse environmental monitoring technology solutions for the future development of ME and offshore wind energy installations. This study aligns with the U.S. Department of Energy (DOE) Water Power Technologies Office’s (WPTO’s) commitment to advancing sustainable energy systems in U.S. waters, recognizing that ME involves generating energy from marine resources, such as waves, tides, and currents (Garson 2023). To this end, innovative monitoring approaches are essential for effective environmental management. It is hypothesized that continuous airborne thermal imagery from TBSs will enhance wildlife detection capabilities compared to traditional human observations or surface-based thermal imaging, particularly in low visibility and nighttime conditions. Prior studies suggest that aerial monitoring could increase the detection rates of large marine species (English et al. 2024; Farinelli et al. 2024; Panigada et al. 2024); however, the specific capabilities of TBSs in diverse marine environments remain fully unexplored. Preliminary tests were conducted to assess sensor performance in controlled fog simulations at Sandia National Laboratories (Sandia) to validate the methodology. This foundational work underscores the potential of TBS technology for monitoring marine wildlife under challenging conditions. Subsequently, a full TBS field operation was executed in Carmel, California, with the following objectives: (1) to detect live marine wildlife within the study area during both day and night, (2) to compare detection rates between the TBS and sensor packages at various altitudes against human observations from a land-based station, (3) to determine whether scanning or stationary imaging methodologies at altitude optimize wildlife detection, and (4) to evaluate the performance of ML algorithms in comparison to human post-collection analyses of TBS-collected imagery. By providing robust, scientifically grounded data, this research aims to contribute to existing knowledge regarding aerial technologies and methodologies for detecting and monitoring interactions between marine wildlife and ME systems. The subsequent sections will detail the methodology, results, and recommendations based on lessons learned and the future steps to advance TBSs and sensors, with an emphasis on aiding sustainable ME and offshore wind development. Methods TBS and Metocean Sensors Fifty-three hours of TBS flights were conducted by Pacific Northwest National Laboratory (PNNL) and Sandia at the Marine Pollution Studies Laboratory (MPSL) at Granite Canyon near Carmel, California, from January 25–29, 2024. The MPSL, which is jointly administered by the National Oceanographic and Atmospheric Administration (NOAA) and the University of California at Davis, is ideally located to monitor migrating gray whales on their southerly progression during the California winter at 36.44°N, 121.92°W and 21 m mean sea level (msl). TBS flights occurred between 0–300 m above ground level (agl; 21–321 m msl) during daylight and nighttime conditions, as summarized in Table 1. Table 1. TBS flights occurred between 0–300 m agl (21–321 m msl) during daylight and nighttime conditions. Altitude (m agl) 50 100 150 200 250 300 PST / UTC Hour 8 / 16 X 9 / 17 X 10 / 18 X X X 11 / 19 X X X X 12 / 20 X X X 13 / 21 X X X 14 / 22 X X X 15 / 23 X X 16 / 0 X X 17 / 1 X 18 / 2 X X 19 / 3 X X X 20 / 4 X X 21 / 5 X X 22 / 6 X X X X X 23 / 7 X X The TBS was composed of a 128 m 3 helium-filled aerostat powered by a 5 hp direct current (DC) permanent magnet motor controlled by a reversible regenerative driven variable-speed controller. The TBS operated airborne imaging sensor packages, as described in the next subsection (Figure 4a & b), day and night during varying atmospheric conditions and flight altitudes, as shown in Figure 1. The temperature, relative humidity, and altitude were measured with an InterMet iMet-4 RSB radiosonde on the TBS. Visibility was measured with a surface-based Campbell Scientific CS120A visibility sensor and typically decreased during daylight hours, as shown in Figure 2. Based on the results of a study conducted in Sandia’s Fog Tunnel in December 2023 (Dexheimer et al. 2024) the radiometric output of the TBS thermal imagers was expected to become increasingly inaccurate with decreasing visibility, and target detection to be impaired in reduced visibility because of the increased homogeneity within the radiometric image. Surface wind and wave properties were measured by the CODAR SeaSonde system at Granite Canyon, which is a high-frequency radar that measures surface currents from sea echo, in addition to deriving information on wind and wave properties from the sea echo. Imaging Sensors An ICI Mirage 640 P mid-wavelength infrared (MWIR) imager (Figure 4d) was used with 27 and 11 mm lenses, and an ICI 8640 long-wavelength infrared (LWIR) imager (Figure 4c) was used with a 50 mm lens. The Mirage 640 costs roughly 5 times more than the 8640 and uses a cooled chip with enhanced thermal imaging capabilities in colder temperatures. Both cameras were tested to determine if the Mirage provided increased detection capability. Multiple lenses were also tested to assess the comparative virtues of field of view (FOV) and resolution on the detection capability. At the start of each flight day, each thermal imager was calibrated at four pitch angles against a reference heated water bath with a stated temperature stability of ±0.07 °C, as shown in Figure 3. The emissivity value that allowed the radiometric temperature to match that of the calibration bath was recorded for each pitch and camera and lens combination to allow accurate radiometric output to be produced from the thermal images during post-processing. A Sony UMC-R10C camera (Figure 4d) was used to provide a visible reference during thermal imaging. Tallysman HC872 helical antennas and Hemisphere Vega 28 global navigation satellite system (GNSS) compass boards were used with the airborne Gremsy T7 camera gimbal to determine the distance to the imaging target. A full description of the imaging sensors and methodology is available in (Dexheimer et al. 2024). A RED Komodo 6K cinema camera (Figure 4e) was used with a Canon EF 100–400 mm L-series zoom lens to capture high-resolution images and video of marine wildlife. All imaging acquisition devices were time-synced daily TBS Operations Over 55 h of footage were collected during the study, as detailed in Table 2. Initially, the TBS carried out two opportunistic scanning patterns, which required the camera’s FOV to overlap with the position and timing of the present wildlife. Table 2. Throughout this study, 55.7 h of footage were collected from the surface and aloft. Location Camera and Lens Raw Processed Dates Duration (hours) Total Surface ICI 8640 and 50 mm 4.49 TB 292 GB January 27–30 38.7 Airborne ICI 8640 and 50 mm 133 GB 10.9 GB January 25–26 2.0 Airborne Mirage and 27 mm 687 GB 44.1 GB January 25–29 11.2 Airborne Mirage and 11 mm 234 GB 6.69 GB January 28 3.8 Total Airborne All 1.03 TB 61.7 GB January 25–29 17.0 Total Footage All 5.52 TB 353.7 GB January 25–29 55.7 A variable-pitch scan from shoreline to shoreline was performed using the Mirage camera and 27 mm lens; the pitch decreased from −3° to −75° below the horizon, with a −3° pitch equal to a 1.9 km distance to a target with the balloon 100 m agl. The balloon ascended in 50 m increments from 50 m to 300 m, with the pitch decreasing in increments corresponding to a change in the observed distance equal to half the vertical FOV. As the balloon ascended, the scan was initiated at steeper pitch angles, where a cutoff size of 4 pixels for an expected 7 m long target was reached based on the distance to the target. The camera operator maintained the camera at a fixed heading and pitch angle using in-flight data streaming from the differential GNSS antennas on the camera gimbal in addition to the real-time gimbal controller display. This scan pattern required 135 min to complete, with the scan at each altitude taking approximately 14 min. During the scan, a still image and 10 s video clip were continuously captured. The length of this scan pattern taxed the manual dexterity and visual endurance of the camera operators, so the scan pattern was revised to use a fixed pitch of −4° with the Mirage camera and either a 27 mm or 11 mm lens loitering at 50 m increments between 50 and 250 m agl. The −4° pitch radius was perceived to coincide with the region of most abundant marine wildlife based on camera operator observations during the study. This perception was later confirmed by the ML algorithm’s analysis of captured video determining that 75.8% of whale blows were detected between 1 and 2.5 km from the surface-based camera. The variable- and fixed-pitch scan patterns were conducted for over six and almost nine hours, respectively, during the TBS flight campaign. The TBS alternated the opportunistic fixed-pitch scan with an observer-driven loitering pattern, which stationed the Mirage camera with the 11 or 27 mm lens at a fixed altitude in 50 m increments between 50 and 250 m agl for 15–30 min with the camera pointed perpendicular to the shoreline on a 237° heading. The operator would look for any potential targets in the controller display within this period, while an additional visual observer simultaneously scanned for targets. If a target was identified by the observer, the camera operator would be verbally guided until the target was in frame; then, the target was tracked as a still image, and 10 s video clips were continuously captured. If no targets were identified, the still image and 10 s video clips were continuously captured throughout the scan. When the TBS was ascending or descending to a new flight level during all flight patterns, scanning would be suspended, and the airborne camera would be fixed on a 237° heading. Ascending or descending 50 m between flight levels generally occurred in 100 s. The loitering pattern was conducted for over 25 h during the campaign. Through the use of the scanning and loitering patterns, we intend to study the rates of comparative target detection between both operating methodologies. A surface-based ICI 8640 thermal imager was operated continuously from January 27 at 14:30 to January 30 at 03:00 on a 237° heading to provide a comparison of detection rates with the airborne thermal imagers. Target Detection and Visual Observations Airborne and surface thermal videos were imported into ICI’s IR Flash Pro software and exported as .mp4 files, which were then watched at a 3´ playback rate. Detected individuals were recorded with respect to species and time. NOAA visual observers independently conducted surface-based gray whale surveys at MPSL with binoculars from 07:30 to 16:30 on Monday through Friday from January 22 to February 1, 2024, with Thursday the 25 th and Friday the 26 th overlapping TBS flights. NOAA’s recorded sightings were compared with TBS thermal-imagery-based detections to determine if and when TBS-based observations may provide added value. RED camera video was encoded with RED’s proprietary RedcodeRAW codec to preserve image quality and was color-graded and converted to Rec709 .mp4 video files using Adobe Premiere Pro and Media Encoder software. RED camera video footage was evaluated to compare 2K, 4K, and 6K resolutions in terms of visual detail and clarity (Figure 7). The analysis presented in Figure 7 illustrates the distinct image quality and detail associated with each resolution. Higher resolutions, particularly 4K and 6K, provided enhanced depth of field, which may improve the detection and detail of whale observations. These findings highlight the role of a higher resolution in improving the detection of whale blows and other marine wildlife. Machine-Learning Detection Toyon Research Corporation (Toyon) was provided with converted 8-bit .mp4 files of surface and airborne camera footage. Infrared video was processed in Whale Spout Detector using both human-developed algorithms and artificial intelligence (AI) techniques. The human-designed algorithms served as a detector that identified possible locations of whale blows that were then fed to the AI model, which classified them as either whale blows, vessels, or other objects. The detector functioned by first building a background model of the scene to look for statistical anomalies using a single frame of data. Once an anomalous group of pixels was located, it triggered a tracking mechanism that followed the development of a candidate blow so that only objects that were similar in brightness and duration to a whale blow were passed along to the AI model. The AI model was a novel design developed at Toyon based on a convolutional 3D (C3D) architecture. Multiple 3D convolutions were performed so that both spatial and temporal features could be extracted. The model had been trained using thousands of samples of whale blows, vessels, and clutter. The trained model was embedded in the C++ software of Whale Spout Detector using the Open Neural Network Exchange (ONNX) format. The ONNX model allowed for seamless integration into C++ software, enabling real-time, efficient operation on various hardware platforms and allowing for much faster than real-time operation on the collected footage. Results Throughout this study, 55.7 h of footage were collected from the surface and aloft, as summarized in Table 2. From the airborne TBS, 59 gray whale, 100 avian target, and 6 indistinguishable marine mammal sightings, which were either sea otter or harbor seal ( Phoca vitulina ), were observed, while 1409 gray whales, 1342 avian species, 33 sea otters, 11 common dolphins ( Delphinus delphis ), and 19 indistinguishable mammals were observed by the more continuous surface-based thermal imager. Avian detection includes seabird and birds of prey species. As shown in Figure 8, most airborne whale sightings occurred in midmorning local time with a secondary peak in the early afternoon. Airborne avian sightings were distributed throughout the day and night, with peak observations occurring in the early afternoon. Harbor seals and sea otters were sighted in the morning. Most surface-based (non-TBS-derived) gray whale sightings occurred between sunset and midnight local time with a secondary peak in the afternoon. Surface avian observations peaked around midday and near sunset, and sea otter, common dolphin, and indistinguishable mammal sightings were most often observed during the day from midmorning to late afternoon. The only period of overlap between the airborne TBS and NOAA human observations occurred on January 25 and 26. The airborne TBS observations exhibit more diurnal variability than the human observer observations, but both methodologies indicate a similar magnitude of observations and decreased whale sightings in the late afternoon, likely attributed to changing environmental conditions. These conditions include increased surface glint from the sun setting, heightened wind speeds, and elevated Beaufort scale conditions. Toyon machine-processed and human-processed detections from the surface camera exhibited remarkably good diurnal agreement, lending confidence to both methods. In Figure 9, TBS flight altitudes were normalized by the total flight time and compared with the altitudes of whale detection normalized by total whale detections. A lower percentage of whale detections occur above 200 m in relation to the total flight time at or above 200 m. Based on real-time experience during the field campaign and post-processing, the reduced resolution at these higher flight altitudes resulted in difficulty in target identification. In contrast, a greater number of whale detections occurred at TBS altitudes of 50–100 m. This relatively low altitude indicates that marine mammal observations may not require aircraft and could occur from coastal instrumented towers or elevated structures, as well as offshore wind infrastructure. Of the 47 separate airborne captures of 59 total whales, 14 captures occurred when both the surface and airborne camera were operating simultaneously between January 27 and 29. Three of these fourteen capture events, or 21%, did not result in a surface whale observation within 3 min, which we interpret as an airborne detection/surface miss case. The median TBS altitude during the three-whale airborne detection/surface miss cases was 200 m, compared to a median TBS altitude of 153 m for all 14 simultaneous whale detection events, which suggests that additional observations may have been captured if the camera on the TBS had a larger FOV. The surface 8640 camera was expected to resolve a 7 m whale target into the minimum perceived detectable number of pixels, 4, at a 1.5 km distance to target. The airborne Mirage and 27 mm lens resolved the same 7 m target in 4 pixels at a 3.3 km distance to target. No whale blow observations were made with the Mirage and 11 mm lens, which resolved a 7 m target in 4 pixels at a 1.25 km distance to target. Figure 11a and b show an airborne detection/surface miss case observed with the TBS loitering at 200 m agl on January 28 at 23:47:11. The whale blow is observed at roughly half of the resolvable 3.3 km distance to target of the airborne Mirage camera (Figure 11a) and is beyond the 1.5 km distance expected to be resolved by the surface-based 8640 (Figure 11b). Of the 68 separate airborne avian captures, 62 captures occurred when both the surface and airborne cameras were operating simultaneously. Of these 62 capture events, 21, or 34%, were airborne detection/surface miss cases. The median TBS altitude during the 21 avian airborne detection/surface miss cases was 57 m, compared to a median TBS altitude of 102 m for all 62 simultaneous avian detection events. Because of the reduced target size of avian observations compared to whale observations, target detection at distance is limited, and it is likely that the increased FOV of the airborne Mirage camera resulted in observations that were not detected by the surface 8640 camera and 50 mm lens. Figure 12a and b show an airborne avian detection/surface miss case observed with the TBS loitering at 50 m agl on January 28 at 16:33:14, where the target was beyond the FOV of the surface-based 8640. At a 700 m distance to target, the surface-operating 8640 field dimensions were 134 m × 107 m, compared with 611 m × 489 m for the airborne Mirage and 11 mm lens. Of the 68 airborne avian captures, 15 were collected with the 11 mm lens on the Mirage and 53 were collected with the 27 mm lens, corresponding to similar respective detection rates of 4.0 and 4.7 avian detections per hour. The lens choice for avian detection should weigh the target size and the required resolution against the FOV. Although avian targets are small in comparison to whale blows, they are more easily identified because of their typically constant motion and trajectory. Images of a simultaneous airborne whale detection/surface detection case observed when the TBS was engaged in fixed-pitch scanning at 150 m agl on January 29 at 02:03:46 are shown in Figure 13a and b. A lens flare is visible in the lower left corner of the airborne image. An initial comparison of the scan patterns indicates that the variable-pitch scan pattern resulted in the highest rate of whale blow detections per hour. However, the variable-pitch scan pattern was only used on January 25 and 26 before it was replaced by alternating shorter periods of fixed-pitch scanning bookended by longer loitering periods. Because of the limited amount of potential testing time on site and the uncertainty related to the peak of the migratory rate, the variable-pitch scan pattern was replaced by the other two patterns because of the lengthy 135-minute period required to complete the scan. It should be noted that while the TBS conducted 53 h of flights between January 25–29, marine mammal imaging only occurred for approximately 40 h. The additional 13 h of flight time were typically spent troubleshooting or updating airborne instrumentation or occurred when it was difficult for the camera operators to see into the setting sun beginning about 90 min prior to sunset or when the solar elevation reached 15°. Table 3. Table of scan patterns: flight hours, whale blow detections, and avian detections comparing detection rates across scan patterns. Scan Pattern Flight Hours of Use Whale Blow Detections per Hour Avian Detections per Hour Variable Pitch 6.1 4.7 0.4 Fixed Pitch 8.7 1.0 2.2 Loitering 25.5 0.3 1.7 To determine if the variability in whale blow detection rate is related to the TBS scan pattern or migratory intensity, we reference the NOAA human observer data. Human-based whale surveys were generally conducted from 07:30–16:30 PST on weekdays from January 22 to February 2, which overlapped the TBS observations on January 25–26. On January 25, no human-based whale observations were made from 12:00–15:00 PST. A daily mean of 79 whale sightings were recorded in the human observations, with the mean on January 25 and 26 alone equaling 81 sightings per day. Since the mean number of whales observed on January 25 and 26 was consistent with the mean over the nine-day period, this indicates that the variable-pitch method has the greatest efficacy at identifying whale blows from the airborne TBS. The fixed-pitch and loitering patterns were alternated from January 27–29, so any variability in the migratory activity would be anticipated to impact the detections per hour for both patterns equally. While the least complex to execute or potentially automate in future iterations of this system, loitering exhibited the least efficacy for whale blow detection. For avian detection, the detection rate across the three scan patterns is relatively more uniform. Given that seabird populations tend to exhibit stable daily behavioral patterns rather than significant fluctuations due to migration or other factors, the fixed-pitch and loitering scan patterns might be more effective for detecting seabirds compared to the variable-pitch scan pattern. These scan patterns likely offer more reliable opportunities for detection under these stable conditions. Sandia’s student interns processed all 55.7 h of surface and airborne video footage at a rough cost of $65 per hour of footage. The advantages of human processing include the ability to process scanning or stationary footage and the ability to process the footage with respect to any identifiable animal species. The disadvantages are that human processing is tedious and time- and labor-intensive and requires significant data storage space that can be cumbersome to share between users. Toyon ML algorithms could not be run on the TBS airborne footage because the camera moved from one scene to another. The ML software builds a background model of the ocean to detect whale blows and will not operate if the camera is moved more often than every few minutes. This dependency may limit the future adaptability of this automated detection method for use with active scanning patterns or from nonstatic platforms. Currently, ML algorithms are significantly more expensive than human processing conducted by unspecialized labor and have a rough hourly cost that is 5 times greater than the human processing costs incurred during this study. Toyon Whale Spout Detector identified 1,281 whale blows in surface imagery collected between January 27 at 15:02 and January 29 at 23:55. Human processing resulted in 1,121 captures and 1,409 individual whale blows detected in surface imagery between January 27 at 14:35 and January 30 at 02:54. Almost 69% (879) of the Toyon blow detections occurred within 3 min of a human processing detection, indicating that the majority of blows detected by the Toyon detector were also detected by human analysts. In comparison, 591 or almost 53% of surface detections occurred within 3 min of a Toyon detection, indicating a greater number of detections occurred uniquely from human analysts in comparison to unique detections from the Toyon detector. An increased percentage of Toyon detections that were not detected by human analysts occurred during daylight, indicating that sun artifacts may have contributed to Toyon detections missed by human observers. This indicates a potential benefit of using ML algorithms on daylight imagery when human analysis may be impaired. Whale Spout Detector also estimates the range, coordinates, and bearing of a detected blow. Out of the total blows detected by the surface-based camera, 75.8% of blows were detected between 1 and 2.5 km, which informs the design criteria for coastal migratory whale imaging systems. Discussion/Recommendations In this study, 55.7 h of footage were collected using both surface-based and airborne thermal imaging systems to monitor gray whales, avian species, and other marine wildlife. The key findings demonstrate distinct temporal patterns in species detection, with gray whales being most frequently observed from the surface between sunset and midnight, while human-processed airborne detections peaked in the mid-morning and early afternoon. An intercomparison of the Toyon algorithm and human processing revealed that an increased percentage of Toyon detections that were not detected by human analysts occurred during daylight, indicating that sun artifacts may impair human image processing. The disparity between whale surface observations peaking from sunset to midnight and airborne human-processed observations peaking in mid-morning and early afternoon may also be attributed to increased imaging artifacts when the sun most impacted the airborne camera FOV near sunset. Avian species were observed both day and night, with peak sightings occurring around midday and in the early afternoon. Sea otters and harbor seals were primarily detected in the morning. Notably, whale detections were more successful at lower flight altitudes and higher camera resolutions, suggesting that future monitoring may benefit from coastal towers or offshore structures equipped with thermal imaging systems. Additionally, this study highlights the differences in detection efficiency between human observers and the TBS, with implications for optimizing future wildlife monitoring efforts. The comparison between surface-based and airborne observations revealed key insights into the efficacy of each method in detecting marine wildlife. Surface-based thermal imaging captured significantly more gray whale sightings in total than the airborne thermal imaging with the TBS because of the continuous nature of surface monitoring. However, the airborne TBS demonstrated unique advantages, particularly in detecting whales and avian species that were missed by surface cameras during simultaneous observations. For instance, 21% of airborne whale detections did not coincide with surface observations, indicating that airborne systems can identify animals in areas or at distances beyond the surface sensor’s FOV. These discrepancies suggest that a combination of both methodologies could enhance the overall detection capability, particularly in environments with challenging visibility or wide monitoring areas. Additionally, the choice of scan patterns, flight altitudes, and imaging equipment significantly influenced detection rates, with variable-pitch scanning proving more effective for whale blow detection. Further refinement of these operational parameters will be critical for improving the accuracy and efficiency of wildlife monitoring in future studies. To enhance the impact of this study, several recommendations for future research and practical applications emerge (Table 4). First, further investigations should focus on refining scanning methodologies, particularly the variable-pitch scan, which demonstrated the highest detection rates for whale blows. Automating this process could reduce operator fatigue and increase efficiency. Additionally, exploring the integration of ML algorithms for the real-time processing of thermal imagery could streamline data analysis, allowing for quicker decision-making in wildlife monitoring. Given the success of lower-altitude observations, future studies should consider deploying stationary monitoring platforms, such as coastal towers, to complement aerial efforts, especially in high biodiversity areas. Moreover, the findings underscore the importance of collaboration between human observations and automated systems, suggesting that integrating NOAA’s real-time data could optimize monitoring strategies. Ultimately, these advancements could inform conservation practices and regulatory frameworks, particularly in the context of emerging offshore developments, ensuring that wildlife protection remains a priority amid growing human activities in marine environments. Table 4. To enhance the operational efficiency and scientific output of the TBS and imaging sensors for detecting and tracking marine wildlife, the following recommendations aim to advance the use of the TBS and imaging sensor’s capabilities for wildlife observation in marine environments. Optimization of Scan Patterns Variable- vs. Fixed-Pitch Scans The variable-pitch scan exhibited a higher rate of whale blow detections, though it was more labor intensive and time consuming. Future efforts should explore automating the variable-pitch scan pattern to reduce operator fatigue and improve efficiency. Alternatively, reducing the number of scans to target the area of most frequent target detection could shorten the scan period from 135 min without significantly sacrificing detection rates. Automated Loitering Patterns Given the relatively low whale detection rate in the loitering pattern but its simplicity for automation, further research should be conducted on improving loitering pattern efficacy, particularly through the optimization of altitudes and camera angles. Machine Learning and Real-Time Processing Improved AI Models for Whale and Avian Detection While the Toyon ML algorithms were successful in detecting whale blows, continuous improvement in the AI models can be pursued by integrating more diverse datasets and enhancing the algorithms’ ability to differentiate species (i.e., marine mammals and avian species). Future studies should evaluate real-time AI performance and its integration into flight operations. Human vs. AI Processing The study revealed limitations in manual human processing due to time, cost, and labor constraints. Implementing real-time ML detection systems could reduce the need for post-flight human analysis, speed up data review, and improve real-time decision-making during TBS operations. To make this transition, real-time ML detection costs will need to be comparable to the cost of manual human processing. Environmental Conditions Impacting Detection Impact of Visibility and Atmospheric Conditions Reduced visibility impaired the radiometric performance of the TBS’s thermal imagers. Future studies should focus on developing or integrating sensors that can perform better under foggy or reduced visibility conditions, perhaps through multispectral or adaptive imaging technologies. Additionally, exploring atmospheric correction models to adjust imagery in real time could be valuable. Lower- vs. Higher-Altitude Observations As indicated by the higher detection rate at lower altitudes (50–100 m), future studies should consider deploying lower-altitude fixed monitoring platforms (e.g., coastal towers or offshore wind turbines) with thermal imaging systems. Comparative research on marine wildlife detection from both airborne and stationary systems would provide insights into the necessity of a TBS at certain altitudes. Imaging and Detection Equipment Lens and Camera Comparison The study showed that the 27 mm lens on the Mirage camera had better detection rates than the 11 mm lens. In future deployments, emphasis should be placed on using wider lenses like the 27 mm for wildlife detection. Further testing could explore a balance between resolution and FOV, especially to optimize the detection of different species. Enhancement of Thermal Imaging Systems The ICI Mirage 640 P MWIR camera is capable of increased detection performance in cold conditions compared to the 8640 LWIR camera, particularly for whale detection. Additional research into other camera systems or emerging technologies that enhance detection accuracy in diverse marine environments could greatly enhance marine mammal and avian surveys. Data Collection and Workflow Extended Operational Hours Given the diurnal variability in whale and seabird sightings, extending TBS flights to nighttime hours and early morning could help maximize the likelihood of detecting marine life. Using a combination of human visual observations and AI at night may also enhance detection. Collaboration Between Human and Machine-Learning Observations Comparing TBS detections with NOAA’s human visual surveys has proven effective. More research should focus on how both methodologies can be better integrated, for example, by incorporating real-time NOAA observations as feedback to the TBS, allowing more accurate and targeted camera adjustments. Long-Term Marine Wildlife Observation Programs Multiyear Campaigns Repeating observation campaigns over multiple years would enhance the understanding of seasonal migration and diurnal variations and identify the long-term effects of environmental changes on marine wildlife. This approach would allow for comparisons between human and AI-based detection systems and for AI-based systems to become reliable in the detection of various species and possibly behavioral changes, which would be beneficial data for renewable energy development and regulatory agencies. Strategic Siting Positioning the TBS in areas with high marine biodiversity, key wildlife corridors, or regions undergoing significant ecological changes (e.g., offshore wind installations and offshore oil platforms) will maximize data collection and impact. Multisensor Integration Future efforts should incorporate additional sensors like acoustic monitoring for whales, water quality sensors, and multispectral satellite data to create a more comprehensive observation system for marine ecosystems. Continuous Offshore Monitoring Deploying the TBS on offshore platforms (oil rigs or renewable energy developments, buoys) enables ongoing wildlife and environmental monitoring, supporting the long-term tracking of population trends and ecological changes. Offshore Wind and Marine Energy Monitoring The TBS can monitor wildlife interactions with offshore wind developments, providing critical data on the ecological impact throughout the construction and operation phases (Courbis et al. 2024). Sea Turtle Monitoring While not included in the current study because the study area does not encompassing sea turtles, the TBS is a valuable tool for monitoring them. It can assist in tracking migration routes, nesting sites, and interactions with offshore developments, thus contributing to conservation and regulatory efforts (Danovaro et al. 2024). Future studies in sea turtle habitats would be beneficial in indicating how TBS capabilities can inform conservation strategies, regulatory needs, and the installation of renewable energy sources with minimal or no impact on sea turtles. Conclusion This study highlights the capabilities of the TBS and advanced imaging sensors in effectively detecting and tracking marine wildlife, primarily gray whales and avian species. Over the course of 55 h of flights, we gathered extensive data on marine biodiversity, demonstrating the potential of thermal imaging technologies and innovative operational strategies. The findings indicate that TBS operations at altitudes between 50 to 200 m significantly enhance wildlife detection, with the variable-pitch scanning pattern emerging as the most effective method for identifying whale blows compared to fixed-pitch and loitering patterns. This suggests that adaptive scanning techniques can greatly improve our understanding of marine species’ distribution, movement, and potential behavior. The study revealed instances where airborne detections were not corroborated by surface observations, underscoring the complementary role of aerial monitoring to traditional survey methods. Furthermore, a comparative analysis of detection rates between the TBS and surface-based observations illustrates the strengths of aerial imaging technologies in identifying marine wildlife across various environmental conditions and ecosystems. The integration of ML algorithms into the workflow is set to enhance data processing efficiency, paving the way for real-time monitoring capabilities. The successful calibration and integration of diverse imaging sensors within the TBS framework further illustrate the potential for creating a comprehensive monitoring system that can adapt to environmental conditions and operational challenges. This research not only provides valuable insights into gray whale detections and population studies but also lays the groundwork for future studies in marine biodiversity monitoring, particularly in relation to conservation strategies and the sustainable development of ME and offshore wind resources. In conclusion, these findings underscore the critical importance of incorporating aerial surveillance technologies, advanced imaging sensors, and in situ methodologies in marine wildlife research. Such advancements facilitate a deeper understanding of species behavior while supporting effective conservation efforts. Future research should prioritize refining operational methodologies and exploring the applicability of TBSs in different ecological contexts, ultimately enhancing the capacity to observe marine ecosystems and wildlife, which is essential for the sustainable management of offshore wind developments and other ME initiatives. Declarations Conflicts of Interest The authors declare no conflict of interest. The funders had no role in the design of the study; in the collection, analyses, or interpretation of data; in the writing of the manuscript; or in the decision to publish the results. Funding This research was funded by the United States Department of Energy, Water Power Technologies Office, contract number DE-AC05-76RL01830. Ethical Compliance Statement This study did not involve the sampling of animals, and therefore, no permits were required. We affirm that all applicable international, national, and institutional guidelines for the care and use of organisms have been adhered to. As such, specific permissions related to animal sampling do not apply. We are prepared to provide any relevant documentation upon request. Regulatory Compliance Statement This study was conducted in full compliance with all applicable regulations, including obtaining the necessary permits from the Federal Aviation Administration (FAA) for aerial operations. We adhered to all FAA guidelines to ensure safe and responsible use of unmanned aerial systems during the research. Data Availability Statement Data will be made available under the license CC-Attribution 4.0 via the Portal and Repository for Information on Marine Renewable Energy (PRIMRE) on the Marine and Hydrokinetic Data Repository (MHKDR) [ https://mhkdr.openei.org/ , accessed on 31 December 2024]. Acknowledgment We want to express our gratitude to several individuals and organizations whose contributions made this research possible. We acknowledge Casey Longbottom, David Novick, Brent Peterson, Carlos Ruiz, and Gabrielle Whitson for their expert field operations of the tethered balloon system and development of the imaging sensors. Special thanks are extended to Dave Weller, Aimée Lange, and Trevor Joyce at NOAA Southwest Fisheries Science Center (SWFSC) for authorizing access to the Marine Pollution Studies Laboratory at Granite Canyon and for sharing valuable gray whale observation data. We are also grateful for the hospitality and support from UC Davis staff, which facilitated our site operations. The student staff led by Benjamin Hess undertook the effort to review all the footage from this effort, and we greatly appreciate the many hours spent at computers documenting whales and other wildlife. We would like to thank the CODAR Ocean Sensors, Ltd. team for providing wave height data from their continuous monitoring project at the Granite Canyon Laboratory Site. Additionally, we acknowledge Toyon Research Corporation for providing the machine learning detection system used in our study. We also extend our appreciation to the Triton Initiative leadership team for their invaluable support throughout this project. Lastly, we thank the reviewers for their valuable feedback, which greatly improved the quality of our manuscript. References Aguilar-Lazcano CA, Espinosa-Curiel IE, Ríos-Martínez JA, Madera-Ramírez FA, Pérez-Espinosa H (2023) Machine learning-based sensor data fusion for animal monitoring: Scoping review. Sensors 23:5732 Amerson A, Gonzalez-Hirshfeld I, Dexheimer D (2023) Validating a Tethered Balloon System and Optical Technologies for Marine Wildlife Detection and Tracking. Remote Sens 15:4709 Amerson AM, Harris TM, Michener SR, Gunn CM, Haxel JH (2022) A Summary of Environmental Monitoring Recommendations for Marine Energy Development That Considers Life Cycle Sustainability. J Mar Sci Eng 10:586 Barlow DR, Strong CS, Torres LG (2024) Three decades of nearshore surveys reveal long-term patterns in gray whale habitat use, distribution, and abundance in the Northern California Current. Sci Rep 14:9352 Bicknell AW, Godley BJ, Sheehan EV, Votier SC, Witt MJ (2016) Camera technology for monitoring marine biodiversity and human impact. Front Ecol Environ 14:424–432 Christie KS, Gilbert SL, Brown CL, Hatfield M, Hanson L (2016) Unmanned aircraft systems in wildlife research: current and future applications of a transformative technology. Front Ecol Environ 14:241–251 Clarfeld LA, Sirén AP, Mulhall BM, Wilson TL, Bernier E, Farrell J, Lunde G, Hardy N, Gieder KD, Abrams R (2023) Evaluating a tandem human-machine approach to labelling of wildlife in remote camera monitoring. Ecol Inf 77:102257 Corcoran E, Winsen M, Sudholz A, Hamilton G (2021) Automated detection of wildlife using drones: Synthesis, opportunities and constraints. Methods Ecol Evol 12:1103–1114 Courbis S, Williams K, Stepanuk J, Etter H, McManus M, Campoblanco F, Pacini A (2024) Technology Gaps for Monitoring Birds and Marine Mammals at Offshore Wind Facilities. Mar Technol Soc J 58:5–22 Danovaro R, Bianchelli S, Brambilla P, Brussa G, Corinaldesi C, Del Borghi A, Dell’Anno A, Fraschetti S, Greco S, Grosso M (2024) Making eco-sustainable floating offshore wind farms: Siting, mitigations, and compensations. Renew Sustain Energy Rev 197:114386 Danovaro R, Carugati L, Berzano M, Cahill AE, Carvalho S, Chenuil A, Corinaldesi C, Cristina S, David R, Dell'Anno A (2016) Implementing and innovating marine monitoring approaches for assessing marine environmental status. Front Mar Sci 3:213 Dexheimer D, Whitson G, Cheng Z, Sammon J, Gaustad K, Mei F, Longbottom C (2024) Tethered Balloon System (TBS) Instrument Handbook. DOE Office of Science Atmospheric Radiation Measurement (ARM) Program … Eaves SL, Staines G, Harker-Klimeš G, Pinza M, Geerlofs S (2022) Triton Field Trials: Promoting Consistent Environmental Monitoring Methodologies for Marine Energy Sites. J Mar Sci Eng 10:177 English G, Lawrence MJ, McKindsey CW, Lacoursière-Roussel A, Bergeron H, Gauthier S, Wringe BF, Trudel M (2024) A review of data collection methods used to monitor the associations of wild species with marine aquaculture sites. Reviews Aquaculture 16:1160–1185 Farinelli S, Keith-Diagne L, Garnica J, Keiman J, Luther D (2024) Quantifying minimum survey effort to reliably detect Amazonian manatees using an unoccupied aerial vehicle (UAV) at an ex situ soft-release site. Latin Am J Aquat Mamm 19:42–60 Garson J (2023) Water Power Technologies Office: 2022 Project Peer Review. EERE Publication and Product Library, Washington, DC. (United States) Gibbs JP, Snell HL, Causton CE (1999) Effective monitoring for adaptive wildlife management: lessons from the Galapagos Islands. J Wildl Manag : 1055–1065 Haxel J, Zang X, Martinez J, Polagye B, Staines G, Deng ZD, Wosnik M, O’Byrne P (2022) Underwater Noise Measurements around a Tidal Turbine in a Busy Port Setting. J Mar Sci Eng 10:632 Hemery LG, Mackereth KF, Gunn CM, Pablo EB (2022a) Use of a 360-degree underwater camera to characterize artificial reef and fish aggregating effects around marine energy devices. J Mar Sci Eng 10:555 Hemery LG, Mackereth KF, Tugade LG (2022b) What’s in my toolkit? A review of technologies for assessing changes in habitats caused by marine energy development. J Mar Sci Eng 10:92 Kellenberger B, Marcos D, Tuia D (2018) Detecting mammals in UAV images: Best practices to address a substantially imbalanced dataset with deep learning. Remote Sens Environ 216:139–153 Marvin DC, Koh LP, Lynam AJ, Wich S, Davies AB, Krishnamurthy R, Stokes E, Starkey R, Asner GP (2016) Integrating technologies for scalable ecology and conservation. Global Ecol Conserv 7:262–275 Panigada S, Pierantonio N, Araújo H, David L, Di-Méglio N, Dorémus G, Gonzalvo J, Holcer D, Laran S, Lauriano G (2024) The ACCOBAMS Survey Initiative: the first synoptic assessment of cetacean abundance in the Mediterranean Sea through aerial surveys. Front Mar Sci 10:1270513 Reilly CE, Larson J, Amerson AM, Staines GJ, Haxel JH, Pattison PM (2022) Minimizing Ecological Impacts of Marine Energy Lighting. J Mar Sci Eng 10:354 Sharma S, Sato K, Gautam BP (2023) A Methodological Literature Review of Acoustic Wildlife Monitoring Using Artificial Intelligence Tools and Techniques. Sustainability 15:7128 Staines GJ, Mueller RP, Seitz AC, Evans MD, O’Byrne PW, Wosnik M (2022) Capabilities of an acoustic camera to inform fish collision risk with current energy converter turbines. J Mar Sci Eng 10:483 Stephenson P (2020) Technological advances in biodiversity monitoring: Applicability, opportunities and challenges. Curr Opin Environ Sustain 45:36–41 Stewart JD, Joyce TW, Durban JW, Calambokidis J, Fauquier D, Fearnbach H, Grebmeier JM, Lynn M, Manizza M, Perryman WL (2023) Boom-bust cycles in gray whales associated with dynamic and changing Arctic conditions. Science 382:207–211 Thomas B, Holland JD, Minot EO (2011) Wildlife tracking technology options and cost considerations. Wildl Res 38:653–663 Wang ZA, Moustahfid H, Mueller AV, Michel AP, Mowlem M, Glazer BT, Mooney TA, Michaels W, McQuillan JS, Robidart JC (2019) Advancing observation of ocean biogeochemistry, biology, and ecosystems with cost-effective in situ sensing technologies. Front Mar Sci 6:519 Cite Share Download PDF Status: Published Journal Publication published 03 Apr, 2025 Read the published version in Marine Biology → Version 1 posted Editorial decision: Revise and Resubmit 22 Nov, 2024 Reviewers agreed at journal 03 Nov, 2024 Reviewers invited by journal 03 Nov, 2024 Editor assigned by journal 28 Oct, 2024 First submitted to journal 28 Oct, 2024 You are reading this latest preprint version Research Square lets you share your work early, gain feedback from the community, and start making changes to your manuscript prior to peer review in a journal. As a division of Research Square Company, we’re committed to making research communication faster, fairer, and more useful. We do this by developing innovative software and high quality services for the global research community. Our growing team is made up of researchers and industry professionals working together to solve the most critical problems facing scientific publishing. Also discoverable on Platform About Our Team In Review Editorial Policies Advisory Board Help Center Resources Author Services Accessibility API Access RSS feed Manage Cookie Preferences © Research Square 2026 | ISSN 2693-5015 (online) Privacy Policy Terms of Service Do Not Sell My Personal Information {"props":{"pageProps":{"initialData":{"identity":"rs-5349011","acceptedTermsAndConditions":true,"allowDirectSubmit":false,"archivedVersions":[],"articleType":"Research Article","associatedPublications":[],"authors":[{"id":373515765,"identity":"92ad95a7-cc02-4993-92cd-c604fcc2125b","order_by":0,"name":"Alicia Amerson","email":"data:image/png;base64,iVBORw0KGgoAAAANSUhEUgAAAZAAAAAyAQMAAABI0h/eAAAABlBMVEX///8AAABVwtN+AAAACXBIWXMAAA7EAAAOxAGVKw4bAAAA/ElEQVRIiWNgGAWjYDACdsYGhgQE1waImRvwa2FG1ZIGxIyEtKByDxPWwt/M3PjhQQ2DvG577+MPH3eczzOXbmxg/FJxGKcWicOMzRIJxxgMt505biY588ztYss5BxuYZc7g1sJwmLGNIYGNgXHbjTQ2Zt6224kbbiQ2MEu2peHUIQ/W8o/Bftv9Z8yfedvOEdZiANKSCETbbrAxSPO2HQBrYfzYZoNTiyHIL4l9EsnbzqSxSc5sSy42AGo5zHAGtxa54+0PP/74ZmO77fgx5g8f2+zyDG4kH3z4o0ICt/chAKEgAUQc5iGkARmAtTD+IEXLKBgFo2AUDHcAAFr5WUzEYS1AAAAAAElFTkSuQmCC","orcid":"https://orcid.org/0000-0002-0753-6735","institution":"Pacific Northwest National Laboratory Biological Sciences Division","correspondingAuthor":true,"prefix":"","firstName":"Alicia","middleName":"","lastName":"Amerson","suffix":""},{"id":373515766,"identity":"5fcd646e-7f4a-4fb1-852e-e438165dda7a","order_by":1,"name":"Darielle Dexheimer","email":"","orcid":"","institution":"Sandia National Laboratories","correspondingAuthor":false,"prefix":"","firstName":"Darielle","middleName":"","lastName":"Dexheimer","suffix":""}],"badges":[],"createdAt":"2024-10-28 17:40:59","currentVersionCode":1,"declarations":"","doi":"10.21203/rs.3.rs-5349011/v1","doiUrl":"https://doi.org/10.21203/rs.3.rs-5349011/v1","draftVersion":[],"editorialEvents":[{"content":"https://doi.org/10.1007/s00227-025-04618-3","type":"published","date":"2025-04-03T15:57:21+00:00"}],"editorialNote":"","failedWorkflow":false,"files":[{"id":70379613,"identity":"6feaf718-3a5a-4fdb-b75f-bc6710604734","added_by":"auto","created_at":"2024-12-02 15:50:10","extension":"jpeg","order_by":1,"title":"Figure 1","display":"","copyAsset":false,"role":"figure","size":97866,"visible":true,"origin":"","legend":"\u003cp\u003eThe TBS operated airborne imaging sensor packages day and night during varying atmospheric conditions and flight altitudes.\u003c/p\u003e","description":"","filename":"floatimage2.jpeg","url":"https://assets-eu.researchsquare.com/files/rs-5349011/v1/b80cc3c8cdf07961174f49e6.jpeg"},{"id":70381000,"identity":"d66a2eab-3ea3-406b-883c-02af5ad50929","added_by":"auto","created_at":"2024-12-02 15:58:10","extension":"png","order_by":2,"title":"Figure 2","display":"","copyAsset":false,"role":"figure","size":48090,"visible":true,"origin":"","legend":"\u003cp\u003eVisibility was measured with a surface-based Campbell Scientific CS120A visibility sensor and typically decreased during daylight hours.\u003c/p\u003e","description":"","filename":"floatimage3.png","url":"https://assets-eu.researchsquare.com/files/rs-5349011/v1/c4643127b9b26b38d83abf2a.png"},{"id":70378824,"identity":"aa4ddc21-71ac-4b5f-b8ac-0ba76c499373","added_by":"auto","created_at":"2024-12-02 15:42:10","extension":"jpeg","order_by":3,"title":"Figure 3","display":"","copyAsset":false,"role":"figure","size":338606,"visible":true,"origin":"","legend":"\u003cp\u003eAt the start of each flight day, each thermal imager was calibrated at four pitch angles against a reference heated water bath with a stated temperature stability of ±0.07 °C.\u003c/p\u003e","description":"","filename":"floatimage4.jpeg","url":"https://assets-eu.researchsquare.com/files/rs-5349011/v1/db1d55583dfe884cd750b571.jpeg"},{"id":70378816,"identity":"15d08eb5-094b-4cdd-8c0d-c3e60fec3f4a","added_by":"auto","created_at":"2024-12-02 15:42:10","extension":"jpeg","order_by":4,"title":"Figure 4","display":"","copyAsset":false,"role":"figure","size":1099785,"visible":true,"origin":"","legend":"\u003cp\u003e(a) The airborne camera boom launching on the TBS; (b) the camera boom, suspended gimbal, and GNSS differential antennas above the TBS winch; (c) the ICI 8640 camera continuously operated at the surface from January 27 to January 30; (d) the ICI Mirage and Sony R10C cameras on the Gremsy T7 gimbal with the differential GNSS antennas and Vega 28 positioning board; and (e) the RED camera.\u003c/p\u003e","description":"","filename":"floatimage5.jpeg","url":"https://assets-eu.researchsquare.com/files/rs-5349011/v1/6a19eb983a1a34acbeb71def.jpeg"},{"id":70379618,"identity":"1623fca7-e1bd-4196-89b5-2bee38c2404c","added_by":"auto","created_at":"2024-12-02 15:50:10","extension":"png","order_by":5,"title":"Figure 5","display":"","copyAsset":false,"role":"figure","size":217278,"visible":true,"origin":"","legend":"\u003cp\u003eTBS in-flight camera locations at Granite Canyon.\u003c/p\u003e","description":"","filename":"floatimage6.png","url":"https://assets-eu.researchsquare.com/files/rs-5349011/v1/c1f33d8fe248b0b4f17289fb.png"},{"id":70378828,"identity":"f012229d-641e-459c-9f4b-0a866d95e395","added_by":"auto","created_at":"2024-12-02 15:42:10","extension":"png","order_by":6,"title":"Figure 6","display":"","copyAsset":false,"role":"figure","size":1638443,"visible":true,"origin":"","legend":"\u003cp\u003eDepiction of a variable-pitch scan at a TBS altitude of 100 m agl. The direction of an arrow indicates the direction of a scan.\u003c/p\u003e","description":"","filename":"floatimage7.png","url":"https://assets-eu.researchsquare.com/files/rs-5349011/v1/c09329dc660576051f926775.png"},{"id":70381001,"identity":"e0aae829-4c0e-4c50-bb36-e32675978d56","added_by":"auto","created_at":"2024-12-02 15:58:10","extension":"jpeg","order_by":7,"title":"Figure 7","display":"","copyAsset":false,"role":"figure","size":532697,"visible":true,"origin":"","legend":"\u003cp\u003eRED camera video footage comparison at 2K, 4K, and 6K.\u003c/p\u003e","description":"","filename":"floatimage8.jpeg","url":"https://assets-eu.researchsquare.com/files/rs-5349011/v1/9acd7efc1c833decf712023f.jpeg"},{"id":70378819,"identity":"2e83a05e-487e-4e91-95ee-d16d5be65146","added_by":"auto","created_at":"2024-12-02 15:42:10","extension":"png","order_by":8,"title":"Figure 8","display":"","copyAsset":false,"role":"figure","size":63353,"visible":true,"origin":"","legend":"\u003cp\u003eTotal hourly human-detected airborne camera, human-observed, human-detected surface-based camera, and machine-learning-detected surface-based camera wildlife observations at MPSL from January 25–29, 2024.\u003c/p\u003e","description":"","filename":"floatimage9.png","url":"https://assets-eu.researchsquare.com/files/rs-5349011/v1/4629875ebbb669b851cdeaa5.png"},{"id":70378821,"identity":"5c171ac0-e112-4425-a7c1-b64ac3d27551","added_by":"auto","created_at":"2024-12-02 15:42:10","extension":"png","order_by":9,"title":"Figure 9","display":"","copyAsset":false,"role":"figure","size":18042,"visible":true,"origin":"","legend":"\u003cp\u003eTBS flight altitudes were normalized by the total flight time and compared with the altitudes of whale detection normalized by total whale detections.\u003c/p\u003e","description":"","filename":"floatimage10.png","url":"https://assets-eu.researchsquare.com/files/rs-5349011/v1/cadb4a21a3887ebb7b86cd26.png"},{"id":70378820,"identity":"52dff9b1-0229-4e74-bf33-aa6660b494c2","added_by":"auto","created_at":"2024-12-02 15:42:10","extension":"png","order_by":10,"title":"Figure 10","display":"","copyAsset":false,"role":"figure","size":36826,"visible":true,"origin":"","legend":"\u003cp\u003eComparison of simultaneous TBS airborne and surface observations to analyze detection success rates, highlighting instances where airborne cameras captured whales not observed at the surface despite successful airborne detection.\u003c/p\u003e","description":"","filename":"floatimage11.png","url":"https://assets-eu.researchsquare.com/files/rs-5349011/v1/6caf850485431190aee39799.png"},{"id":70378832,"identity":"3ef56cd0-4799-49be-a919-baccccfc0a4e","added_by":"auto","created_at":"2024-12-02 15:42:11","extension":"png","order_by":11,"title":"Figure 11","display":"","copyAsset":false,"role":"figure","size":916587,"visible":true,"origin":"","legend":"\u003cp\u003e(a) The airborne Mirage sensor detection of a whale blow. (b) The detection missed by the surface 8640 sensor.\u003c/p\u003e","description":"","filename":"floatimage12.png","url":"https://assets-eu.researchsquare.com/files/rs-5349011/v1/8618eaae8f3fc38d4c6161ec.png"},{"id":70378829,"identity":"745905b0-4f8d-4912-a7db-936cff68c91f","added_by":"auto","created_at":"2024-12-02 15:42:11","extension":"png","order_by":12,"title":"Figure 12","display":"","copyAsset":false,"role":"figure","size":845226,"visible":true,"origin":"","legend":"\u003cp\u003e(a) An airborne avian detection by the Mirage sensor. (b) A missed detection by the 8640 sensor because the target was beyond the field of view.\u003c/p\u003e","description":"","filename":"floatimage13.png","url":"https://assets-eu.researchsquare.com/files/rs-5349011/v1/5ed736fec663b129df4d7235.png"},{"id":70378831,"identity":"0ee59afc-f794-41f9-b724-3c982cd02926","added_by":"auto","created_at":"2024-12-02 15:42:11","extension":"png","order_by":13,"title":"Figure 13","display":"","copyAsset":false,"role":"figure","size":655123,"visible":true,"origin":"","legend":"\u003cp\u003eImages of an airborne whale detection/surface detection case observed when the TBS was engaged in fixed-pitch scanning at 150 m agl on January 29 at 02:03:46.\u003c/p\u003e","description":"","filename":"floatimage14.png","url":"https://assets-eu.researchsquare.com/files/rs-5349011/v1/69418d5165b807f898aa4890.png"},{"id":70381002,"identity":"49f16012-811c-40d9-8fd3-8f3f66938746","added_by":"auto","created_at":"2024-12-02 15:58:10","extension":"png","order_by":14,"title":"Figure 14","display":"","copyAsset":false,"role":"figure","size":22324,"visible":true,"origin":"","legend":"\u003cp\u003eDaily total NOAA human-observed whale sightings: Tracking the number of whale sightings recorded by NOAA observers over time.\u003c/p\u003e","description":"","filename":"floatimage15.png","url":"https://assets-eu.researchsquare.com/files/rs-5349011/v1/de78e1c44af47abf28cef691.png"},{"id":70379619,"identity":"7a66b46f-3ba5-4782-a379-e4206d3d24e1","added_by":"auto","created_at":"2024-12-02 15:50:11","extension":"png","order_by":15,"title":"Figure 15","display":"","copyAsset":false,"role":"figure","size":33379,"visible":true,"origin":"","legend":"\u003cp\u003eFraction of blow detections by hour from the surface 8640 camera that were not detected by the alternate detection method within 3 min.\u003c/p\u003e","description":"","filename":"floatimage16.png","url":"https://assets-eu.researchsquare.com/files/rs-5349011/v1/04332c36fcf2d56a950b66e0.png"},{"id":70379615,"identity":"c4030ee2-e515-4f09-a07f-4efc14bcf58f","added_by":"auto","created_at":"2024-12-02 15:50:10","extension":"png","order_by":16,"title":"Figure 16","display":"","copyAsset":false,"role":"figure","size":17010,"visible":true,"origin":"","legend":"\u003cp\u003eThe numbers of individual whale blows, detection events, and detections by the alternate processing method within 3 min for the human analyst and Toyon algorithm for surface 8640 camera video processing.\u003c/p\u003e","description":"","filename":"floatimage17.png","url":"https://assets-eu.researchsquare.com/files/rs-5349011/v1/c890943fb76d200771bf0e89.png"},{"id":80082269,"identity":"1583bb6d-b323-4ec0-a768-d7a485ba8981","added_by":"auto","created_at":"2025-04-07 16:08:12","extension":"pdf","order_by":0,"title":"","display":"","copyAsset":false,"role":"manuscript-pdf","size":7915657,"visible":true,"origin":"","legend":"","description":"","filename":"manuscript.pdf","url":"https://assets-eu.researchsquare.com/files/rs-5349011/v1/65948d7a-f76b-4c4e-8019-6b21d5054b29.pdf"}],"financialInterests":"","formattedTitle":"Enhancing Marine Wildlife Observations: The Application of Tethered Balloon Systems and Advanced Imaging Sensors for Sustainable Marine Energy Development","fulltext":[{"header":"Introduction","content":"\u003cp\u003eAs various marine ecosystems increasingly face considerations for marine energy (ME) development, comprehensive environmental assessments have become necessary (Eaves et al. 2022). These assessments aim to evaluate the potential impacts of new technologies on energetically dynamic marine environments, particularly focusing on how marine wildlife interacts with ME devices. Disturbances from ME installations may lead to alterations in habitat use, behavioral changes, and shifts in population dynamics for key species, making it essential to provide field-tested recommendations for implementing environmental monitoring technologies and methodologies to understand these interactions (Amerson et al. 2022; Haxel et al. 2022; Hemery et al. 2022a; Hemery et al. 2022b; Reilly et al. 2022; Staines et al. 2022).\u003c/p\u003e\n\n\u003cp\u003eThere is a growing interest in utilizing cost-effective monitoring technologies that can also be implemented with minimal to no impact on wildlife (Gibbs et al. 1999; Thomas et al. 2011; Christie et al. 2016; Marvin et al. 2016; Stephenson 2020). While these technologies are more easily adapted for terrestrial wildlife, they also apply to observations of marine wildlife interactions with ME systems (Bicknell et al. 2016; Danovaro et al. 2016; Wang et al. 2019). However, gaps remain regarding the efficacy of aerial monitoring methods, particularly in varied marine and coastal conditions (Amerson et al. 2023). These gaps include scanning patterns, flight longevity, wind conditions, altitude variations, and comparisons between human observations and machine-learning (ML) detection.\u003c/p\u003e\n\n\u003cp\u003eIn a previous study, the research team conducted an initial flight trial of a tethered balloon system (TBS) and sensor package in La Porte, Texas (TX) (Amerson et al. 2023). During this study, no marine wildlife species were present. Therefore, there was a need to perform flights along a coastline with a known migratory path and a larger diversity of marine species. Furthermore, a consideration for the second deployment was to find an environment similar to areas of future ME development. An additional consideration was made to include flights during daylight and nighttime hours to evaluate the use of a TBS for ME environmental assessment for 24 h. Lastly, accumulating a large dataset from the effort in La Porte, TX presented a challenge associated with aerial monitoring: increased processing and analysis time by humans. The need for reliable ML applications may reduce this processing and analysis time, but these systems are currently under development and require reliable data libraries (Kellenberger et al. 2018; Corcoran et al. 2021; Aguilar-Lazcano et al. 2023; Clarfeld et al. 2023; Sharma et al. 2023). A reliable source of data for ML may be obtained from analysis that has been processed through human observations (Stewart et al. 2023; Barlow et al. 2024).\u003c/p\u003e\n\n\u003cp\u003eThis study aimed to address these gaps by integrating a TBS equipped with advanced imaging sensors to observe marine wildlife along the California coast, a critical migratory corridor for species such as gray whales and other marine mammals. Additionally, this study evaluated data collected by TBS sensors and human observations, reviewed various scan patterns and loitering altitudes, and leveraged ML programs to detect whale blows. By implementing ML, the goal was to compare the time and cost of data processing and analysis between humans and ML programs.\u003c/p\u003e\n\n\u003cp\u003eThe significance of this research lies in its potential to provide technological and methodological recommendations for regulatory decision-makers and to contribute to diverse environmental monitoring technology solutions for the future development of ME and offshore wind energy installations. This study aligns with the U.S. Department of Energy (DOE) Water Power Technologies Office\u0026rsquo;s (WPTO\u0026rsquo;s) commitment to advancing sustainable energy systems in U.S. waters, recognizing that ME involves generating energy from marine resources, such as waves, tides, and currents (Garson 2023). To this end, innovative monitoring approaches are essential for effective environmental management. It is hypothesized that continuous airborne thermal imagery from TBSs will enhance wildlife detection capabilities compared to traditional human observations or surface-based thermal imaging, particularly in low visibility and nighttime conditions. Prior studies suggest that aerial monitoring could increase the detection rates of large marine species (English et al. 2024; Farinelli et al. 2024; Panigada et al. 2024); however, the specific capabilities of TBSs in diverse marine environments remain fully unexplored.\u003c/p\u003e\n\n\u003cp\u003ePreliminary tests were conducted to assess sensor performance in controlled fog simulations at Sandia National Laboratories (Sandia) to validate the methodology. This foundational work underscores the potential of TBS technology for monitoring marine wildlife under challenging conditions. Subsequently, a full TBS field operation was executed in Carmel, California, with the following objectives: (1) to detect live marine wildlife within the study area during both day and night, (2) to compare detection rates between the TBS and sensor packages at various altitudes against human observations from a land-based station, (3) to determine whether scanning or stationary imaging methodologies at altitude optimize wildlife detection, and (4) to evaluate the performance of ML algorithms in comparison to human post-collection analyses of TBS-collected imagery.\u003c/p\u003e\n\n\u003cp\u003eBy providing robust, scientifically grounded data, this research aims to contribute to existing knowledge regarding aerial technologies and methodologies for detecting and monitoring interactions between marine wildlife and ME systems. The subsequent sections will detail the methodology, results, and recommendations based on lessons learned and the future steps to advance TBSs and sensors, with an emphasis on aiding sustainable ME and offshore wind development.\u003c/p\u003e"},{"header":"Methods","content":"\u003ch2\u003eTBS and Metocean Sensors\u003c/h2\u003e\n\u003cp\u003eFifty-three hours of TBS flights were conducted by Pacific Northwest National Laboratory (PNNL) and Sandia at the Marine Pollution Studies Laboratory (MPSL) at Granite Canyon near Carmel, California, from January 25\u0026ndash;29, 2024. The MPSL, which is jointly administered by the National Oceanographic and Atmospheric Administration (NOAA) and the University of California at Davis, is ideally located to monitor migrating gray whales on their southerly progression during the California winter at 36.44\u0026deg;N, 121.92\u0026deg;W and 21 m mean sea level (msl). TBS flights occurred between 0\u0026ndash;300 m above ground level (agl; 21\u0026ndash;321 m msl) during daylight and nighttime conditions, as summarized in\u0026nbsp;Table 1.\u003c/p\u003e\n\u003cp\u003eTable\u0026nbsp;1. TBS flights occurred between 0\u0026ndash;300\u0026nbsp;m agl (21\u0026ndash;321\u0026nbsp;m msl) during daylight and nighttime conditions.\u003c/p\u003e\n\u003cdiv align=\"\"\u003e\n \u003ctable border=\"0\" cellspacing=\"0\" cellpadding=\"0\" width=\"337\"\u003e\n \u003cthead\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 107px;\"\u003e\n \u003cp\u003e\u003cstrong\u003e\u003cem\u003eAltitude (m agl)\u003c/em\u003e\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 38px;\"\u003e\n \u003cp\u003e\u003cstrong\u003e\u003cem\u003e50\u003c/em\u003e\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 38px;\"\u003e\n \u003cp\u003e\u003cstrong\u003e\u003cem\u003e100\u003c/em\u003e\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 38px;\"\u003e\n \u003cp\u003e\u003cstrong\u003e\u003cem\u003e150\u003c/em\u003e\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 38px;\"\u003e\n \u003cp\u003e\u003cstrong\u003e\u003cem\u003e200\u003c/em\u003e\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 38px;\"\u003e\n \u003cp\u003e\u003cstrong\u003e\u003cem\u003e250\u003c/em\u003e\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 38px;\"\u003e\n \u003cp\u003e\u003cstrong\u003e\u003cem\u003e300\u003c/em\u003e\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003c/thead\u003e\n \u003ctbody\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 31.9403%;\"\u003e\n \u003cp\u003e\u003cstrong\u003e\u003cem\u003ePST / UTC Hour\u003c/em\u003e\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 11.3433%;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 11.3433%;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 11.3433%;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 11.3433%;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 11.3433%;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 11.3433%;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 31.9403%;\"\u003e\n \u003cp\u003e\u003cstrong\u003e\u003cem\u003e8 / 16\u003c/em\u003e\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 11.3433%;\"\u003e\n \u003cp\u003e\u003cstrong\u003eX\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 11.3433%;\"\u003e\n \u003cp\u003e\u003cstrong\u003e\u0026nbsp;\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 11.3433%;\"\u003e\n \u003cp\u003e\u003cstrong\u003e\u0026nbsp;\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 11.3433%;\"\u003e\n \u003cp\u003e\u003cstrong\u003e\u0026nbsp;\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 11.3433%;\"\u003e\n \u003cp\u003e\u003cstrong\u003e\u0026nbsp;\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 11.3433%;\"\u003e\n \u003cp\u003e\u003cstrong\u003e\u0026nbsp;\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 31.9403%;\"\u003e\n \u003cp\u003e\u003cstrong\u003e\u003cem\u003e9 / 17\u003c/em\u003e\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 11.3433%;\"\u003e\n \u003cp\u003e\u003cstrong\u003eX\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 11.3433%;\"\u003e\n \u003cp\u003e\u003cstrong\u003e\u0026nbsp;\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 11.3433%;\"\u003e\n \u003cp\u003e\u003cstrong\u003e\u0026nbsp;\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 11.3433%;\"\u003e\n \u003cp\u003e\u003cstrong\u003e\u0026nbsp;\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 11.3433%;\"\u003e\n \u003cp\u003e\u003cstrong\u003e\u0026nbsp;\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 11.3433%;\"\u003e\n \u003cp\u003e\u003cstrong\u003e\u0026nbsp;\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 31.9403%;\"\u003e\n \u003cp\u003e\u003cstrong\u003e\u003cem\u003e10 / 18\u003c/em\u003e\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 11.3433%;\"\u003e\n \u003cp\u003e\u003cstrong\u003e\u0026nbsp;\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 11.3433%;\"\u003e\n \u003cp\u003e\u003cstrong\u003eX\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 11.3433%;\"\u003e\n \u003cp\u003e\u003cstrong\u003eX\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 11.3433%;\"\u003e\n \u003cp\u003e\u003cstrong\u003eX\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 11.3433%;\"\u003e\n \u003cp\u003e\u003cstrong\u003e\u0026nbsp;\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 11.3433%;\"\u003e\n \u003cp\u003e\u003cstrong\u003e\u0026nbsp;\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 31.9403%;\"\u003e\n \u003cp\u003e\u003cstrong\u003e\u003cem\u003e11 / 19\u003c/em\u003e\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 11.3433%;\"\u003e\n \u003cp\u003e\u003cstrong\u003eX\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 11.3433%;\"\u003e\n \u003cp\u003e\u003cstrong\u003eX\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 11.3433%;\"\u003e\n \u003cp\u003e\u003cstrong\u003e\u0026nbsp;\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 11.3433%;\"\u003e\n \u003cp\u003e\u003cstrong\u003eX\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 11.3433%;\"\u003e\n \u003cp\u003e\u003cstrong\u003eX\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 11.3433%;\"\u003e\n \u003cp\u003e\u003cstrong\u003e\u0026nbsp;\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 31.9403%;\"\u003e\n \u003cp\u003e\u003cstrong\u003e\u003cem\u003e12 / 20\u003c/em\u003e\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 11.3433%;\"\u003e\n \u003cp\u003e\u003cstrong\u003eX\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 11.3433%;\"\u003e\n \u003cp\u003e\u003cstrong\u003e\u0026nbsp;\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 11.3433%;\"\u003e\n \u003cp\u003e\u003cstrong\u003eX\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 11.3433%;\"\u003e\n \u003cp\u003e\u003cstrong\u003e\u0026nbsp;\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 11.3433%;\"\u003e\n \u003cp\u003e\u003cstrong\u003eX\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 11.3433%;\"\u003e\n \u003cp\u003e\u003cstrong\u003e\u0026nbsp;\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 31.9403%;\"\u003e\n \u003cp\u003e\u003cstrong\u003e\u003cem\u003e13 / 21\u003c/em\u003e\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 11.3433%;\"\u003e\n \u003cp\u003e\u003cstrong\u003eX\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 11.3433%;\"\u003e\n \u003cp\u003e\u003cstrong\u003e\u0026nbsp;\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 11.3433%;\"\u003e\n \u003cp\u003e\u003cstrong\u003e\u0026nbsp;\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 11.3433%;\"\u003e\n \u003cp\u003e\u003cstrong\u003eX\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 11.3433%;\"\u003e\n \u003cp\u003e\u003cstrong\u003eX\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 11.3433%;\"\u003e\n \u003cp\u003e\u003cstrong\u003e\u0026nbsp;\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 31.9403%;\"\u003e\n \u003cp\u003e\u003cstrong\u003e\u003cem\u003e14 / 22\u003c/em\u003e\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 11.3433%;\"\u003e\n \u003cp\u003e\u003cstrong\u003eX\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 11.3433%;\"\u003e\n \u003cp\u003e\u003cstrong\u003eX\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 11.3433%;\"\u003e\n \u003cp\u003e\u003cstrong\u003e\u0026nbsp;\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 11.3433%;\"\u003e\n \u003cp\u003e\u003cstrong\u003e\u0026nbsp;\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 11.3433%;\"\u003e\n \u003cp\u003e\u003cstrong\u003e\u0026nbsp;\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 11.3433%;\"\u003e\n \u003cp\u003e\u003cstrong\u003eX\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 31.9403%;\"\u003e\n \u003cp\u003e\u003cstrong\u003e\u003cem\u003e15 / 23\u003c/em\u003e\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 11.3433%;\"\u003e\n \u003cp\u003e\u003cstrong\u003e\u0026nbsp;\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 11.3433%;\"\u003e\n \u003cp\u003e\u003cstrong\u003eX\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 11.3433%;\"\u003e\n \u003cp\u003e\u003cstrong\u003eX\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 11.3433%;\"\u003e\n \u003cp\u003e\u003cstrong\u003e\u0026nbsp;\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 11.3433%;\"\u003e\n \u003cp\u003e\u003cstrong\u003e\u0026nbsp;\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 11.3433%;\"\u003e\n \u003cp\u003e\u003cstrong\u003e\u0026nbsp;\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 31.9403%;\"\u003e\n \u003cp\u003e\u003cstrong\u003e\u003cem\u003e16 / 0\u003c/em\u003e\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 11.3433%;\"\u003e\n \u003cp\u003e\u003cstrong\u003e\u0026nbsp;\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 11.3433%;\"\u003e\n \u003cp\u003e\u003cstrong\u003eX\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 11.3433%;\"\u003e\n \u003cp\u003e\u003cstrong\u003e\u0026nbsp;\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 11.3433%;\"\u003e\n \u003cp\u003e\u003cstrong\u003eX\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 11.3433%;\"\u003e\n \u003cp\u003e\u003cstrong\u003e\u0026nbsp;\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 11.3433%;\"\u003e\n \u003cp\u003e\u003cstrong\u003e\u0026nbsp;\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 31.9403%;\"\u003e\n \u003cp\u003e\u003cstrong\u003e\u003cem\u003e17 / 1\u003c/em\u003e\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 11.3433%;\"\u003e\n \u003cp\u003e\u003cstrong\u003e\u0026nbsp;\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 11.3433%;\"\u003e\n \u003cp\u003e\u003cstrong\u003e\u0026nbsp;\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 11.3433%;\"\u003e\n \u003cp\u003e\u003cstrong\u003e\u0026nbsp;\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 11.3433%;\"\u003e\n \u003cp\u003e\u003cstrong\u003e\u0026nbsp;\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 11.3433%;\"\u003e\n \u003cp\u003e\u003cstrong\u003eX\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 11.3433%;\"\u003e\n \u003cp\u003e\u003cstrong\u003e\u0026nbsp;\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 31.9403%;\"\u003e\n \u003cp\u003e\u003cstrong\u003e\u003cem\u003e18 / 2\u003c/em\u003e\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 11.3433%;\"\u003e\n \u003cp\u003e\u003cstrong\u003eX\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 11.3433%;\"\u003e\n \u003cp\u003e\u003cstrong\u003e\u0026nbsp;\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 11.3433%;\"\u003e\n \u003cp\u003e\u003cstrong\u003eX\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 11.3433%;\"\u003e\n \u003cp\u003e\u003cstrong\u003e\u0026nbsp;\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 11.3433%;\"\u003e\n \u003cp\u003e\u003cstrong\u003e\u0026nbsp;\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 11.3433%;\"\u003e\n \u003cp\u003e\u003cstrong\u003e\u0026nbsp;\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 31.9403%;\"\u003e\n \u003cp\u003e\u003cstrong\u003e\u003cem\u003e19 / 3\u003c/em\u003e\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 11.3433%;\"\u003e\n \u003cp\u003e\u003cstrong\u003e\u0026nbsp;\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 11.3433%;\"\u003e\n \u003cp\u003e\u003cstrong\u003eX\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 11.3433%;\"\u003e\n \u003cp\u003e\u003cstrong\u003eX\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 11.3433%;\"\u003e\n \u003cp\u003e\u003cstrong\u003eX\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 11.3433%;\"\u003e\n \u003cp\u003e\u003cstrong\u003e\u0026nbsp;\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 11.3433%;\"\u003e\n \u003cp\u003e\u003cstrong\u003e\u0026nbsp;\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 31.9403%;\"\u003e\n \u003cp\u003e\u003cstrong\u003e\u003cem\u003e20 / 4\u003c/em\u003e\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 11.3433%;\"\u003e\n \u003cp\u003e\u003cstrong\u003e\u0026nbsp;\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 11.3433%;\"\u003e\n \u003cp\u003e\u003cstrong\u003e\u0026nbsp;\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 11.3433%;\"\u003e\n \u003cp\u003e\u003cstrong\u003e\u0026nbsp;\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 11.3433%;\"\u003e\n \u003cp\u003e\u003cstrong\u003eX\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 11.3433%;\"\u003e\n \u003cp\u003e\u003cstrong\u003eX\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 11.3433%;\"\u003e\n \u003cp\u003e\u003cstrong\u003e\u0026nbsp;\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 31.9403%;\"\u003e\n \u003cp\u003e\u003cstrong\u003e\u003cem\u003e21 / 5\u003c/em\u003e\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 11.3433%;\"\u003e\n \u003cp\u003e\u003cstrong\u003e\u0026nbsp;\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 11.3433%;\"\u003e\n \u003cp\u003e\u003cstrong\u003eX\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 11.3433%;\"\u003e\n \u003cp\u003e\u003cstrong\u003eX\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 11.3433%;\"\u003e\n \u003cp\u003e\u003cstrong\u003e\u0026nbsp;\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 11.3433%;\"\u003e\n \u003cp\u003e\u003cstrong\u003e\u0026nbsp;\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 11.3433%;\"\u003e\n \u003cp\u003e\u003cstrong\u003e\u0026nbsp;\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 31.9403%;\"\u003e\n \u003cp\u003e\u003cstrong\u003e\u003cem\u003e22 / 6\u003c/em\u003e\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 11.3433%;\"\u003e\n \u003cp\u003e\u003cstrong\u003eX\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 11.3433%;\"\u003e\n \u003cp\u003e\u003cstrong\u003eX\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 11.3433%;\"\u003e\n \u003cp\u003e\u003cstrong\u003eX\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 11.3433%;\"\u003e\n \u003cp\u003e\u003cstrong\u003eX\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 11.3433%;\"\u003e\n \u003cp\u003e\u003cstrong\u003eX\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 11.3433%;\"\u003e\n \u003cp\u003e\u003cstrong\u003e\u0026nbsp;\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 31.9403%;\"\u003e\n \u003cp\u003e\u003cstrong\u003e\u003cem\u003e23 / 7\u003c/em\u003e\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 11.3433%;\"\u003e\n \u003cp\u003e\u003cstrong\u003eX\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 11.3433%;\"\u003e\n \u003cp\u003e\u003cstrong\u003eX\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 11.3433%;\"\u003e\n \u003cp\u003e\u003cstrong\u003e\u0026nbsp;\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 11.3433%;\"\u003e\n \u003cp\u003e\u003cstrong\u003e\u0026nbsp;\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 11.3433%;\"\u003e\n \u003cp\u003e\u003cstrong\u003e\u0026nbsp;\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 11.3433%;\"\u003e\n \u003cp\u003e\u003cstrong\u003e\u0026nbsp;\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003c/tbody\u003e\n \u003c/table\u003e\n\u003c/div\u003e\n\u003cp\u003e\u0026nbsp;The TBS was composed of a 128 m\u003csup\u003e3\u003c/sup\u003e helium-filled aerostat powered by a 5 hp direct current (DC) permanent magnet motor controlled by a reversible regenerative driven variable-speed controller. The TBS operated airborne imaging sensor packages, as described in the next subsection (Figure 4a \u0026amp; b), day and night during varying atmospheric conditions and flight altitudes, as shown in Figure 1. The temperature, relative humidity, and altitude were measured with an InterMet iMet-4 RSB radiosonde on the TBS. Visibility was measured with a surface-based Campbell Scientific CS120A visibility sensor and typically decreased during daylight hours, as shown in Figure 2. Based on the results of a study conducted in Sandia\u0026rsquo;s Fog Tunnel in December 2023 (Dexheimer et al. 2024) the radiometric output of the TBS thermal imagers was expected to become increasingly inaccurate with decreasing visibility, and target detection to be impaired in reduced visibility because of the increased homogeneity within the radiometric image. Surface wind and wave properties were measured by the CODAR SeaSonde system at Granite Canyon, which is a high-frequency radar that measures surface currents from sea echo, in addition to deriving information on wind and wave properties from the sea echo.\u003c/p\u003e\n\u003ch2\u003eImaging Sensors\u003c/h2\u003e\n\u003cp\u003eAn ICI Mirage 640 P mid-wavelength infrared (MWIR) imager (Figure 4d) was used with 27 and 11\u0026nbsp;mm lenses, and an ICI 8640 long-wavelength infrared (LWIR) imager (Figure 4c) was used with a 50 mm lens. The Mirage 640 costs roughly 5 times more than the 8640 and uses a cooled chip with enhanced thermal imaging capabilities in colder temperatures. Both cameras were tested to determine if the Mirage provided increased detection capability. Multiple lenses were also tested to assess the comparative virtues of field of view (FOV) and resolution on the detection capability. At the start of each flight day, each thermal imager was calibrated at four pitch angles against a reference heated water bath with a stated temperature stability of \u0026plusmn;0.07 \u0026deg;C, as shown in\u0026nbsp;Figure 3. The emissivity value that allowed the radiometric temperature to match that of the calibration bath was recorded for each pitch and camera and lens combination to allow accurate radiometric output to be produced from the thermal images during post-processing.\u003c/p\u003e\n\u003cp\u003eA Sony UMC-R10C camera (Figure 4d) was used to provide a visible reference during thermal imaging. Tallysman HC872 helical antennas and Hemisphere Vega 28 global navigation satellite system (GNSS) compass boards were used with the airborne Gremsy T7 camera gimbal to determine the distance to the imaging target. A full description of the imaging sensors and methodology is available in (Dexheimer et al. 2024). A RED Komodo 6K cinema camera (Figure 4e) was used with a Canon EF 100\u0026ndash;400 mm L-series zoom lens to capture high-resolution images and video of marine wildlife. All imaging acquisition devices were time-synced daily\u003c/p\u003e\n\u003cp\u003e\u003cbr\u003e\u003c/p\u003e\n\u003ch2\u003eTBS Operations\u003c/h2\u003e\n\u003cp\u003eOver 55 h of footage were collected during the study, as detailed in\u0026nbsp;Table 2. Initially, the TBS carried out two opportunistic scanning patterns, which required the camera\u0026rsquo;s FOV to overlap with the position and timing of the present wildlife.\u003c/p\u003e\n\u003cp\u003eTable\u0026nbsp;2. Throughout this study, 55.7\u0026nbsp;h of footage were collected from the surface and aloft.\u003c/p\u003e\n\u003ctable border=\"0\" cellspacing=\"0\" cellpadding=\"0\" width=\"624\"\u003e\n \u003ctbody\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 16.3462%;\"\u003e\n \u003cp\u003e\u003cstrong\u003eLocation\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 23.5577%;\"\u003e\n \u003cp\u003e\u003cstrong\u003eCamera and Lens\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 9.77564%;\"\u003e\n \u003cp\u003e\u003cstrong\u003eRaw\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 13.9423%;\"\u003e\n \u003cp\u003e\u003cstrong\u003eProcessed\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 17.7885%;\"\u003e\n \u003cp\u003e\u003cstrong\u003eDates\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 18.5897%;\"\u003e\n \u003cp\u003e\u003cstrong\u003eDuration (hours)\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 16.3462%;\"\u003e\n \u003cp\u003eTotal Surface\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 23.5577%;\"\u003e\n \u003cp\u003eICI 8640 and 50 mm\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 9.77564%;\"\u003e\n \u003cp\u003e4.49 TB\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 13.9423%;\"\u003e\n \u003cp\u003e292 GB\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 17.7885%;\"\u003e\n \u003cp\u003eJanuary 27\u0026ndash;30\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 18.5897%;\"\u003e\n \u003cp\u003e38.7\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 16.3462%;\"\u003e\n \u003cp\u003eAirborne\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 23.5577%;\"\u003e\n \u003cp\u003eICI 8640 and 50 mm\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 9.77564%;\"\u003e\n \u003cp\u003e133 GB\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 13.9423%;\"\u003e\n \u003cp\u003e10.9 GB\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 17.7885%;\"\u003e\n \u003cp\u003eJanuary 25\u0026ndash;26\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 18.5897%;\"\u003e\n \u003cp\u003e2.0\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 16.3462%;\"\u003e\n \u003cp\u003eAirborne\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 23.5577%;\"\u003e\n \u003cp\u003eMirage and 27 mm\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 9.77564%;\"\u003e\n \u003cp\u003e687 GB\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 13.9423%;\"\u003e\n \u003cp\u003e44.1 GB\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 17.7885%;\"\u003e\n \u003cp\u003eJanuary 25\u0026ndash;29\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 18.5897%;\"\u003e\n \u003cp\u003e11.2\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 16.3462%;\"\u003e\n \u003cp\u003eAirborne\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 23.5577%;\"\u003e\n \u003cp\u003eMirage and 11 mm\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 9.77564%;\"\u003e\n \u003cp\u003e234 GB\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 13.9423%;\"\u003e\n \u003cp\u003e6.69 GB\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 17.7885%;\"\u003e\n \u003cp\u003eJanuary 28\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 18.5897%;\"\u003e\n \u003cp\u003e3.8\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 16.3462%;\"\u003e\n \u003cp\u003eTotal Airborne\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 23.5577%;\"\u003e\n \u003cp\u003eAll\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 9.77564%;\"\u003e\n \u003cp\u003e1.03 TB\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 13.9423%;\"\u003e\n \u003cp\u003e61.7 GB\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 17.7885%;\"\u003e\n \u003cp\u003eJanuary 25\u0026ndash;29\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 18.5897%;\"\u003e\n \u003cp\u003e17.0\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 16.3462%;\"\u003e\n \u003cp\u003eTotal Footage\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 23.5577%;\"\u003e\n \u003cp\u003eAll\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 9.77564%;\"\u003e\n \u003cp\u003e5.52 TB\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 13.9423%;\"\u003e\n \u003cp\u003e353.7 GB\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 17.7885%;\"\u003e\n \u003cp\u003eJanuary 25\u0026ndash;29\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 18.5897%;\"\u003e\n \u003cp\u003e55.7\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003c/tbody\u003e\n\u003c/table\u003e\n\u003cp\u003eA variable-pitch scan from shoreline to shoreline was performed using the Mirage camera and 27 mm lens; the pitch decreased from\u0026nbsp;\u0026minus;3\u0026deg; to\u0026nbsp;\u0026minus;75\u0026deg; below the horizon, with a\u0026nbsp;\u0026minus;3\u0026deg; pitch equal to a 1.9 km distance to a target with the balloon 100 m agl. The balloon ascended in 50 m increments from 50 m to 300 m, with the pitch decreasing in increments corresponding to a change in the observed distance equal to half the vertical FOV. As the balloon ascended, the scan was initiated at steeper pitch angles, where a cutoff size of 4 pixels for an expected 7 m long target was reached based on the distance to the target. The camera operator maintained the camera at a fixed heading and pitch angle using in-flight data streaming from the differential GNSS antennas on the camera gimbal in addition to the real-time gimbal controller display. This scan pattern required 135 min to complete, with the scan at each altitude taking approximately 14 min. During the scan, a still image and 10 s video clip were continuously captured. The length of this scan pattern taxed the manual dexterity and visual endurance of the camera operators, so the scan pattern was revised to use a fixed pitch of\u0026nbsp;\u0026minus;4\u0026deg; with the Mirage camera and either a 27 mm or 11 mm lens loitering at 50 m increments between 50 and 250 m agl. The\u0026nbsp;\u0026minus;4\u0026deg; pitch radius was perceived to coincide with the region of most abundant marine wildlife based on camera operator observations during the study. This perception was later confirmed by the ML algorithm\u0026rsquo;s analysis of captured video determining that\u0026nbsp;75.8% of whale blows were detected between 1 and 2.5\u0026nbsp;km from the surface-based camera.\u0026nbsp;The variable- and fixed-pitch scan patterns were conducted for over six and almost nine hours, respectively, during the TBS flight campaign.\u003c/p\u003e\n\u003cp\u003eThe TBS alternated the opportunistic fixed-pitch scan with an observer-driven loitering pattern, which stationed the Mirage camera with the 11 or 27 mm lens at a fixed altitude in 50 m increments between 50 and 250 m agl for 15\u0026ndash;30 min with the camera pointed perpendicular to the shoreline on a 237\u0026deg; heading. The operator would look for any potential targets in the controller display within this period, while an additional visual observer simultaneously scanned for targets. If a target was identified by the observer, the camera operator would be verbally guided until the target was in frame; then, the target was tracked as a still image, and 10 s video clips were continuously captured. If no targets were identified, the still image and 10 s video clips were continuously captured throughout the scan. When the TBS was ascending or descending to a new flight level during all flight patterns, scanning would be suspended, and the airborne camera would be fixed on a 237\u0026deg; heading. Ascending or descending 50 m between flight levels generally occurred in 100 s. The loitering pattern was conducted for over 25 h during the campaign. Through the use of the scanning and loitering patterns, we intend to study the rates of comparative target detection between both operating methodologies. A surface-based ICI 8640 thermal imager was operated continuously from January 27 at 14:30 to January 30 at 03:00 on a 237\u0026deg; heading to provide a comparison of detection rates with the airborne thermal imagers.\u003c/p\u003e\n\u003ch2\u003eTarget Detection and Visual Observations\u003c/h2\u003e\n\u003cp\u003eAirborne and surface thermal videos were imported into ICI\u0026rsquo;s IR\u0026nbsp;Flash\u0026nbsp;Pro software and exported as .mp4 files, which were then watched at a 3\u0026acute;\u0026nbsp;playback rate. Detected individuals were recorded with respect to species and time. NOAA visual observers independently conducted surface-based gray whale surveys at MPSL with binoculars from 07:30 to 16:30 on Monday through Friday from January 22 to February 1, 2024, with Thursday the 25\u003csup\u003eth\u003c/sup\u003e and Friday the 26\u003csup\u003eth\u003c/sup\u003e overlapping TBS flights. NOAA\u0026rsquo;s recorded sightings were compared with TBS thermal-imagery-based detections to determine if and when TBS-based observations may provide added value. RED camera video was encoded with RED\u0026rsquo;s proprietary RedcodeRAW codec to preserve image quality and was color-graded and converted to Rec709 .mp4 video files using Adobe Premiere Pro and Media Encoder software. RED camera video footage was evaluated to compare 2K, 4K, and 6K resolutions in terms of visual detail and clarity (Figure 7). The analysis presented in Figure 7 illustrates the distinct image quality and detail associated with each resolution. Higher resolutions, particularly 4K and 6K, provided enhanced depth of field, which may improve the detection and detail of whale observations. These findings highlight the role of a higher resolution in improving the detection of whale blows and other marine wildlife.\u003c/p\u003e\n\u003ch2\u003eMachine-Learning Detection\u003c/h2\u003e\n\u003cp\u003eToyon Research Corporation (Toyon) was provided with converted 8-bit .mp4 files of surface and airborne camera footage. Infrared video was processed in Whale Spout Detector using both human-developed algorithms and artificial intelligence (AI) techniques. The human-designed algorithms served as a detector that identified possible locations of whale blows that were then fed to the AI model, which classified them as either whale blows, vessels, or other objects. The detector functioned by first building a background model of the scene to look for statistical anomalies using a single frame of data. Once an anomalous group of pixels was located, it triggered a tracking mechanism that followed the development of a candidate blow so that only objects that were similar in brightness and duration to a whale blow were passed along to the AI model. The AI model was a novel design developed at Toyon based on a convolutional 3D (C3D) architecture. Multiple 3D convolutions were performed so that both spatial and temporal features could be extracted. The model had been trained using thousands of samples of whale blows, vessels, and clutter. The trained model was embedded in the C++ software of Whale Spout Detector using the Open Neural Network Exchange (ONNX) format. The ONNX model allowed for seamless integration into C++ software, enabling real-time, efficient operation on various hardware platforms and allowing for much faster than real-time operation on the collected footage.\u003c/p\u003e"},{"header":"Results","content":"\u003cp\u003eThroughout this study, 55.7 h of footage were collected from the surface and aloft, as summarized in\u0026nbsp;Table 2. From the airborne TBS, 59 gray whale, 100 avian target, and 6 indistinguishable marine mammal sightings, which were either sea otter or harbor seal (\u003cem\u003ePhoca vitulina\u003c/em\u003e),\u0026nbsp;were observed, while 1409 gray whales, 1342 avian species, 33 sea otters, 11 common dolphins (\u003cem\u003eDelphinus delphis\u003c/em\u003e), and 19 indistinguishable mammals were observed by the more continuous surface-based thermal imager. Avian detection includes seabird and birds of prey species. As shown in\u0026nbsp;Figure 8, most airborne whale sightings occurred in midmorning local time with a secondary peak in the early afternoon. Airborne avian sightings were distributed throughout the day and night, with peak observations occurring in the early afternoon. Harbor seals and sea otters were sighted in the morning. Most surface-based (non-TBS-derived) gray whale sightings occurred between sunset and midnight local time with a secondary peak in the afternoon. Surface avian observations peaked around midday and near sunset, and sea otter, common dolphin, and indistinguishable mammal sightings were most often observed during the day from midmorning to late afternoon. The only period of overlap between the airborne TBS and NOAA human observations occurred on January 25 and 26. The airborne TBS observations exhibit more diurnal variability than the human observer observations, but both methodologies indicate a similar magnitude of observations and decreased whale sightings in the late afternoon, likely attributed to changing environmental conditions. These conditions include increased surface glint from the sun setting, heightened wind speeds, and elevated Beaufort scale conditions. Toyon machine-processed and human-processed detections from the surface camera exhibited remarkably good diurnal agreement, lending confidence to both methods.\u003c/p\u003e\n\u003cp\u003eIn Figure 9, TBS flight altitudes were normalized by the total flight time and compared with the altitudes of whale detection normalized by total whale detections. A lower percentage of whale detections occur above 200 m in relation to the total flight time at or above 200 m. Based on real-time experience during the field campaign and post-processing, the reduced resolution at these higher flight altitudes resulted in difficulty in target identification. In contrast, a greater number of whale detections occurred at TBS altitudes of 50\u0026ndash;100 m. This relatively low altitude indicates that marine mammal observations may not require aircraft and could occur from coastal instrumented towers or elevated structures, as well as offshore wind infrastructure.\u003c/p\u003e\n\u003cp\u003eOf the 47 separate airborne captures of 59 total whales, 14 captures occurred when both the surface and airborne camera were operating simultaneously between January 27 and 29. Three of these fourteen capture events, or 21%, did not result in a surface whale observation within 3 min, which we interpret as an airborne detection/surface miss case. The median TBS altitude during the three-whale airborne detection/surface miss cases was 200 m, compared to a median TBS altitude of 153 m for all 14 simultaneous whale detection events, which suggests that additional observations may have been captured if the camera on the TBS had a larger FOV.\u003c/p\u003e\n\u003cp\u003eThe surface 8640 camera was expected to resolve a 7 m whale target into the minimum perceived detectable number of pixels, 4, at a 1.5 km distance to target. The airborne Mirage and 27 mm lens resolved the same 7 m target in 4 pixels at a 3.3 km distance to target. No whale blow observations were made with the Mirage and 11 mm lens, which resolved a 7 m target in 4 pixels at a 1.25 km distance to target. Figure 11a and b show an airborne detection/surface miss case observed with the TBS loitering at 200 m agl on January 28 at 23:47:11. The whale blow is observed at roughly half of the resolvable 3.3 km distance to target of the airborne Mirage camera (Figure 11a) and is beyond the 1.5 km distance expected to be resolved by the surface-based 8640 (Figure 11b).\u003c/p\u003e\n\u003cp\u003eOf the 68 separate airborne avian captures, 62 captures occurred when both the surface and airborne cameras were operating simultaneously. Of these 62 capture events, 21, or 34%, were airborne detection/surface miss cases. The median TBS altitude during the 21 avian airborne detection/surface miss cases was 57 m, compared to a median TBS altitude of 102 m for all 62 simultaneous avian detection events. Because of the reduced target size of avian observations compared to whale observations, target detection at distance is limited, and it is likely that the increased FOV of the airborne Mirage camera resulted in observations that were not detected by the surface 8640 camera and 50 mm lens. Figure 12a and b show an airborne avian detection/surface miss case observed with the TBS loitering at 50 m agl on January 28 at 16:33:14, where the target was beyond the FOV of the surface-based 8640. At a 700 m distance to target, the surface-operating 8640 field dimensions were 134 m \u0026times; 107 m, compared with 611 m \u0026times; 489 m for the airborne Mirage and 11 mm lens.\u003c/p\u003e\n\u003cp\u003eOf the 68 airborne avian captures, 15 were collected with the 11 mm lens on the Mirage and 53 were collected with the 27 mm lens, corresponding to similar respective detection rates of 4.0 and 4.7 avian detections per hour. The lens choice for avian detection should weigh the target size and the required resolution against the FOV. Although avian targets are small in comparison to whale blows, they are more easily identified because of their typically constant motion and trajectory.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eImages of a simultaneous airborne whale detection/surface detection case observed when the TBS was engaged in fixed-pitch scanning at 150 m agl on January 29 at 02:03:46 are shown in\u0026nbsp;Figure\u0026nbsp;13a and b. A lens flare is visible in the lower left corner of the airborne image.\u003c/p\u003e\n\u003cp\u003eAn initial comparison of the scan patterns indicates that the variable-pitch scan pattern resulted in the highest rate of whale blow detections per hour. However, the variable-pitch scan pattern was only used on January 25 and 26 before it was replaced by alternating shorter periods of fixed-pitch scanning bookended by longer loitering periods. Because of the limited amount of potential testing time on site and the uncertainty related to the peak of the migratory rate, the variable-pitch scan pattern was replaced by the other two patterns because of the lengthy 135-minute period required to complete the scan. It should be noted that while the TBS conducted 53 h of flights between January 25\u0026ndash;29, marine mammal imaging only occurred for approximately 40 h. The additional 13 h of flight time were typically spent troubleshooting or updating airborne instrumentation or occurred when it was difficult for the camera operators to see into the setting sun beginning about 90 min prior to sunset or when the solar elevation reached 15\u0026deg;.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eTable\u0026nbsp;3. Table of scan patterns: flight hours, whale blow detections, and avian detections comparing detection rates across scan patterns.\u003c/p\u003e\n\u003ctable border=\"0\" cellspacing=\"0\" cellpadding=\"0\"\u003e\n \u003ctbody\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 15.0641%;\"\u003e\n \u003cp\u003e\u003cstrong\u003eScan Pattern\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 25.1603%;\"\u003e\n \u003cp\u003e\u003cstrong\u003eFlight Hours of Use\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 32.3718%;\"\u003e\n \u003cp\u003e\u003cstrong\u003eWhale Blow Detections per Hour\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 27.4038%;\"\u003e\n \u003cp\u003e\u003cstrong\u003eAvian Detections per Hour\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 15.0641%;\"\u003e\n \u003cp\u003eVariable Pitch\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 25.1603%;\"\u003e\n \u003cp\u003e6.1\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 32.3718%;\"\u003e\n \u003cp\u003e4.7\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 27.4038%;\"\u003e\n \u003cp\u003e0.4\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 15.0641%;\"\u003e\n \u003cp\u003eFixed Pitch\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 25.1603%;\"\u003e\n \u003cp\u003e8.7\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 32.3718%;\"\u003e\n \u003cp\u003e1.0\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 27.4038%;\"\u003e\n \u003cp\u003e2.2\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 15.0641%;\"\u003e\n \u003cp\u003eLoitering\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 25.1603%;\"\u003e\n \u003cp\u003e25.5\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 32.3718%;\"\u003e\n \u003cp\u003e0.3\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 27.4038%;\"\u003e\n \u003cp\u003e1.7\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003c/tbody\u003e\n\u003c/table\u003e\n\u003cp\u003eTo determine if the variability in whale blow detection rate is related to the TBS scan pattern or migratory intensity, we reference the NOAA human observer data. Human-based whale surveys were generally conducted from 07:30\u0026ndash;16:30 PST on weekdays from January 22 to February 2, which overlapped the TBS observations on January 25\u0026ndash;26. On January 25, no human-based whale observations were made from 12:00\u0026ndash;15:00 PST. A daily mean of 79 whale sightings were recorded in the human observations, with the mean on January 25 and 26 alone equaling 81 sightings per day. Since the mean number of whales observed on January 25 and 26 was consistent with the mean over the nine-day period, this indicates that the variable-pitch method has the greatest efficacy at identifying whale blows from the airborne TBS. The fixed-pitch and loitering patterns were alternated from January 27\u0026ndash;29, so any variability in the migratory activity would be anticipated to impact the detections per hour for both patterns equally. While the least complex to execute or potentially automate in future iterations of this system, loitering exhibited the least efficacy for whale blow detection. For avian detection, the detection rate across the three scan patterns is relatively more uniform. Given that seabird populations tend to exhibit stable daily behavioral patterns rather than significant fluctuations due to migration or other factors, the fixed-pitch and loitering scan patterns might be more effective for detecting seabirds compared to the variable-pitch scan pattern. These scan patterns likely offer more reliable opportunities for detection under these stable conditions.\u003c/p\u003e\n\u003cp\u003eSandia\u0026rsquo;s student interns processed all 55.7 h of surface and airborne video footage at a rough cost of $65 per hour of footage. The advantages of human processing include the ability to process scanning or stationary footage and the ability to process the footage with respect to any identifiable animal species. The disadvantages are that human processing is tedious and time- and labor-intensive and requires significant data storage space that can be cumbersome to share between users. Toyon ML algorithms could not be run on the TBS airborne footage because the camera moved from one scene to another. The ML software builds a background model of the ocean to detect whale blows and will not operate if the camera is moved more often than every few minutes. This dependency may limit the future adaptability of this automated detection method for use with active scanning patterns or from nonstatic platforms. Currently, ML algorithms are significantly more expensive than human processing conducted by unspecialized labor and have a rough hourly cost that is 5 times greater than the human processing costs incurred during this study.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eToyon Whale Spout Detector identified 1,281 whale blows in surface imagery collected between January 27 at 15:02 and January 29 at 23:55. Human processing resulted in 1,121 captures and 1,409 individual whale blows detected in surface imagery between January 27 at 14:35 and January 30 at 02:54. Almost 69% (879) of the Toyon blow detections occurred within 3 min of a human processing detection, indicating that the majority of blows detected by the Toyon detector were also detected by human analysts. In comparison, 591 or almost 53% of surface detections occurred within 3 min of a Toyon detection, indicating a greater number of detections occurred uniquely from human analysts in comparison to unique detections from the Toyon detector. An increased percentage of Toyon detections that were not detected by human analysts occurred during daylight, indicating that sun artifacts may have contributed to Toyon detections missed by human observers. This indicates a potential benefit of using ML algorithms on daylight imagery when human analysis may be impaired. Whale Spout Detector also estimates the range, coordinates, and bearing of a detected blow. Out of the total blows detected by the surface-based camera, 75.8% of blows were detected between 1 and 2.5 km, which informs the design criteria for coastal migratory whale imaging systems.\u003c/p\u003e"},{"header":"Discussion/Recommendations","content":"\u003cp\u003eIn this study, 55.7 h of footage were collected using both surface-based and airborne thermal imaging systems to monitor gray whales, avian species, and other marine wildlife. The key findings demonstrate distinct temporal patterns in species detection, with gray whales being most frequently observed from the surface between sunset and midnight, while human-processed airborne detections peaked in the mid-morning and early afternoon. An intercomparison of the Toyon algorithm and human processing revealed that an increased percentage of Toyon detections that were not detected by human analysts occurred during daylight, indicating that sun artifacts may impair human image processing. The disparity between whale surface observations peaking from sunset to midnight and airborne human-processed observations peaking in mid-morning and early afternoon may also be attributed to increased imaging artifacts when the sun most impacted the airborne camera FOV near sunset. Avian species were observed both day and night, with peak sightings occurring around midday and in the early afternoon. Sea otters and harbor seals were primarily detected in the morning. Notably, whale detections were more successful at lower flight altitudes and higher camera resolutions, suggesting that future monitoring may benefit from coastal towers or offshore structures equipped with thermal imaging systems. Additionally, this study highlights the differences in detection efficiency between human observers and the TBS, with implications for optimizing future wildlife monitoring efforts.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eThe comparison between surface-based and airborne observations revealed key insights into the efficacy of each method in detecting marine wildlife. Surface-based thermal imaging captured significantly more gray whale sightings in total than the airborne thermal imaging with the TBS because of the continuous nature of surface monitoring. However, the airborne TBS demonstrated unique advantages, particularly in detecting whales and avian species that were missed by surface cameras during simultaneous observations. For instance, 21% of airborne whale detections did not coincide with surface observations, indicating that airborne systems can identify animals in areas or at distances beyond the surface sensor\u0026rsquo;s FOV. These discrepancies suggest that a combination of both methodologies could enhance the overall detection capability, particularly in environments with challenging visibility or wide monitoring areas. Additionally, the choice of scan patterns, flight altitudes, and imaging equipment significantly influenced detection rates, with variable-pitch scanning proving more effective for whale blow detection. Further refinement of these operational parameters will be critical for improving the accuracy and efficiency of wildlife monitoring in future studies.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eTo enhance the impact of this study, several recommendations for future research and practical applications emerge (Table 4). First, further investigations should focus on refining scanning methodologies, particularly the variable-pitch scan, which demonstrated the highest detection rates for whale blows. Automating this process could reduce operator fatigue and increase efficiency. Additionally, exploring the integration of ML algorithms for the real-time processing of thermal imagery could streamline data analysis, allowing for quicker decision-making in wildlife monitoring. Given the success of lower-altitude observations, future studies should consider deploying stationary monitoring platforms, such as coastal towers, to complement aerial efforts, especially in high biodiversity areas. Moreover, the findings underscore the importance of collaboration between human observations and automated systems, suggesting that integrating NOAA\u0026rsquo;s real-time data could optimize monitoring strategies. Ultimately, these advancements could inform conservation practices and regulatory frameworks, particularly in the context of emerging offshore developments, ensuring that wildlife protection remains a priority amid growing human activities in marine environments.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eTable\u0026nbsp;4.\u0026nbsp;To enhance the operational efficiency and scientific output of the TBS and imaging sensors for detecting and tracking marine wildlife, the following recommendations aim to advance the use of the TBS and imaging sensor\u0026rsquo;s capabilities for wildlife observation in marine environments.\u003c/p\u003e\n\u003ctable border=\"1\" cellspacing=\"0\" cellpadding=\"0\" width=\"642\"\u003e\n \u003ctbody\u003e\n \u003ctr\u003e\n \u003ctd colspan=\"2\" valign=\"top\" style=\"width: 100%;\"\u003e\n \u003cp\u003eOptimization of Scan Patterns\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 32.3988%;\"\u003e\n \u003cp\u003e\u003cem\u003eVariable- vs. Fixed-Pitch Scans\u003c/em\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 67.6012%;\"\u003e\n \u003cp\u003eThe variable-pitch scan exhibited a higher rate of whale blow detections, though it was more labor intensive and time consuming. Future efforts should explore automating the variable-pitch scan pattern to reduce operator fatigue and improve efficiency. Alternatively, reducing the number of scans to target the area of most frequent target detection could shorten the scan period from 135\u0026nbsp;min without significantly sacrificing detection rates.\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 32.3988%;\"\u003e\n \u003cp\u003e\u003cem\u003eAutomated Loitering Patterns\u003c/em\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 67.6012%;\"\u003e\n \u003cp\u003eGiven the relatively low whale detection rate in the loitering pattern but its simplicity for automation, further research should be conducted on improving loitering pattern efficacy, particularly through the optimization of altitudes and camera angles.\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd colspan=\"2\" valign=\"top\" style=\"width: 100%;\"\u003e\n \u003cp\u003eMachine Learning and Real-Time Processing\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 32.3988%;\"\u003e\n \u003cp\u003e\u003cem\u003eImproved AI Models for Whale and Avian Detection\u003c/em\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 67.6012%;\"\u003e\n \u003cp\u003eWhile the Toyon ML algorithms were successful in detecting whale blows, continuous improvement in the AI models can be pursued by integrating more diverse datasets and enhancing the algorithms\u0026rsquo; ability to differentiate species (i.e., marine mammals and avian species). Future studies should evaluate real-time AI performance and its integration into flight operations.\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 32.3988%;\"\u003e\n \u003cp\u003e\u003cem\u003eHuman vs. AI Processing\u003c/em\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 67.6012%;\"\u003e\n \u003cp\u003eThe study revealed limitations in manual human processing due to time, cost, and labor constraints. Implementing real-time ML detection systems could reduce the need for post-flight human analysis, speed up data review, and improve real-time decision-making during TBS operations. To make this transition, real-time ML detection costs will need to be comparable to the cost of manual human processing.\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd colspan=\"2\" valign=\"top\" style=\"width: 100%;\"\u003e\n \u003cp\u003eEnvironmental Conditions Impacting Detection\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 32.3988%;\"\u003e\n \u003cp\u003e\u003cem\u003eImpact of Visibility and Atmospheric Conditions\u003c/em\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 67.6012%;\"\u003e\n \u003cp\u003eReduced visibility impaired the radiometric performance of the TBS\u0026rsquo;s thermal imagers. Future studies should focus on developing or integrating sensors that can perform better under foggy or reduced visibility conditions, perhaps through multispectral or adaptive imaging technologies. Additionally, exploring atmospheric correction models to adjust imagery in real time could be valuable.\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 32.3988%;\"\u003e\n \u003cp\u003e\u003cem\u003eLower- vs. Higher-Altitude Observations\u003c/em\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 67.6012%;\"\u003e\n \u003cp\u003eAs indicated by the higher detection rate at lower altitudes (50\u0026ndash;100\u0026nbsp;m), future studies should consider deploying lower-altitude fixed monitoring platforms (e.g., coastal towers or offshore wind turbines) with thermal imaging systems. Comparative research on marine wildlife detection from both airborne and stationary systems would provide insights into the necessity of a TBS at certain altitudes.\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd colspan=\"2\" valign=\"top\" style=\"width: 100%;\"\u003e\n \u003cp\u003eImaging and Detection Equipment\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 32.3988%;\"\u003e\n \u003cp\u003e\u003cem\u003eLens and Camera Comparison\u003c/em\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 67.6012%;\"\u003e\n \u003cp\u003eThe study showed that the 27\u0026nbsp;mm lens on the Mirage camera had better detection rates than the 11\u0026nbsp;mm lens. In future deployments, emphasis should be placed on using wider lenses like the 27\u0026nbsp;mm for wildlife detection. Further testing could explore a balance between resolution and FOV, especially to optimize the detection of different species.\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 32.3988%;\"\u003e\n \u003cp\u003e\u003cem\u003eEnhancement of Thermal Imaging Systems\u003c/em\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 67.6012%;\"\u003e\n \u003cp\u003eThe ICI Mirage 640 P MWIR camera is capable of increased detection performance in cold conditions compared to the 8640 LWIR camera, particularly for whale detection. Additional research into other camera systems or emerging technologies that enhance detection accuracy in diverse marine environments could greatly enhance marine mammal and avian surveys.\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd colspan=\"2\" valign=\"top\" style=\"width: 100%;\"\u003e\n \u003cp\u003eData Collection and Workflow\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 32.3988%;\"\u003e\n \u003cp\u003e\u003cem\u003eExtended Operational Hours\u003c/em\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 67.6012%;\"\u003e\n \u003cp\u003eGiven the diurnal variability in whale and seabird sightings, extending TBS flights to nighttime hours and early morning could help maximize the likelihood of detecting marine life. Using a combination of human visual observations and AI at night may also enhance detection.\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 32.3988%;\"\u003e\n \u003cp\u003e\u003cem\u003eCollaboration Between Human and Machine-Learning Observations\u003c/em\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 67.6012%;\"\u003e\n \u003cp\u003eComparing TBS detections with NOAA\u0026rsquo;s human visual surveys has proven effective. More research should focus on how both methodologies can be better integrated, for example, by incorporating real-time NOAA observations as feedback to the TBS, allowing more accurate and targeted camera adjustments.\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd colspan=\"2\" valign=\"top\" style=\"width: 100%;\"\u003e\n \u003cp\u003eLong-Term Marine Wildlife Observation Programs\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 32.3988%;\"\u003e\n \u003cp\u003e\u003cem\u003eMultiyear Campaigns\u003c/em\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 67.6012%;\"\u003e\n \u003cp\u003eRepeating observation campaigns over multiple years would enhance the understanding of seasonal migration and diurnal variations and identify the long-term effects of environmental changes on marine wildlife. This approach would allow for comparisons between human and AI-based detection systems and for AI-based systems to become reliable in the detection of various species and possibly behavioral changes, which would be beneficial data for renewable energy development and regulatory agencies.\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 32.3988%;\"\u003e\n \u003cp\u003e\u003cem\u003eStrategic Siting\u003c/em\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 67.6012%;\"\u003e\n \u003cp\u003ePositioning the TBS in areas with high marine biodiversity, key wildlife corridors, or regions undergoing significant ecological changes (e.g., offshore wind installations and offshore oil platforms) will maximize data collection and impact.\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 32.3988%;\"\u003e\n \u003cp\u003e\u003cem\u003eMultisensor Integration\u003c/em\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 67.6012%;\"\u003e\n \u003cp\u003eFuture efforts should incorporate additional sensors like acoustic monitoring for whales, water quality sensors, and multispectral satellite data to create a more comprehensive observation system for marine ecosystems.\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 32.3988%;\"\u003e\n \u003cp\u003e\u003cem\u003eContinuous Offshore Monitoring\u003c/em\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 67.6012%;\"\u003e\n \u003cp\u003eDeploying the TBS on offshore platforms (oil rigs or renewable energy developments, buoys) enables ongoing wildlife and environmental monitoring, supporting the long-term tracking of population trends and ecological changes.\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 32.3988%;\"\u003e\n \u003cp\u003e\u003cem\u003eOffshore Wind and Marine Energy Monitoring\u003c/em\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 67.6012%;\"\u003e\n \u003cp\u003eThe TBS can monitor wildlife interactions with offshore wind developments, providing critical data on the ecological impact throughout the construction and operation phases\u0026nbsp;(Courbis et al. 2024).\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 32.3988%;\"\u003e\n \u003cp\u003e\u003cem\u003eSea Turtle Monitoring\u003c/em\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 67.6012%;\"\u003e\n \u003cp\u003eWhile not included in the current study because the study area does not encompassing sea turtles, the TBS is a valuable tool for monitoring them. It can assist in tracking migration routes, nesting sites, and interactions with offshore developments, thus contributing to conservation and regulatory efforts (Danovaro et al. 2024). Future studies in sea turtle habitats would be beneficial in indicating how TBS capabilities can inform conservation strategies, regulatory needs, and the installation of renewable energy sources with minimal or no impact on sea turtles.\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003c/tbody\u003e\n\u003c/table\u003e"},{"header":"Conclusion","content":"\u003cp\u003eThis study highlights the capabilities of the TBS and advanced imaging sensors in effectively detecting and tracking marine wildlife, primarily gray whales and avian species. Over the course of 55 h of flights, we gathered extensive data on marine biodiversity, demonstrating the potential of thermal imaging technologies and innovative operational strategies. The findings indicate that TBS operations at altitudes between 50 to 200 m significantly enhance wildlife detection, with the variable-pitch scanning pattern emerging as the most effective method for identifying whale blows compared to fixed-pitch and loitering patterns. This suggests that adaptive scanning techniques can greatly improve our understanding of marine species\u0026rsquo; distribution, movement, and potential behavior.\u003c/p\u003e \u003cp\u003eThe study revealed instances where airborne detections were not corroborated by surface observations, underscoring the complementary role of aerial monitoring to traditional survey methods. Furthermore, a comparative analysis of detection rates between the TBS and surface-based observations illustrates the strengths of aerial imaging technologies in identifying marine wildlife across various environmental conditions and ecosystems. The integration of ML algorithms into the workflow is set to enhance data processing efficiency, paving the way for real-time monitoring capabilities.\u003c/p\u003e \u003cp\u003eThe successful calibration and integration of diverse imaging sensors within the TBS framework further illustrate the potential for creating a comprehensive monitoring system that can adapt to environmental conditions and operational challenges. This research not only provides valuable insights into gray whale detections and population studies but also lays the groundwork for future studies in marine biodiversity monitoring, particularly in relation to conservation strategies and the sustainable development of ME and offshore wind resources.\u003c/p\u003e \u003cp\u003eIn conclusion, these findings underscore the critical importance of incorporating aerial surveillance technologies, advanced imaging sensors, and in situ methodologies in marine wildlife research. Such advancements facilitate a deeper understanding of species behavior while supporting effective conservation efforts. Future research should prioritize refining operational methodologies and exploring the applicability of TBSs in different ecological contexts, ultimately enhancing the capacity to observe marine ecosystems and wildlife, which is essential for the sustainable management of offshore wind developments and other ME initiatives.\u003c/p\u003e "},{"header":"Declarations","content":"\u003ch2\u003eConflicts of Interest\u003c/h2\u003e\n\u003cp\u003eThe authors declare no conflict of interest. The funders had no role in the design of the study; in the collection, analyses, or interpretation of data; in the writing of the manuscript; or in the decision to publish the results.\u003c/p\u003e\n\u003ch2\u003eFunding\u003c/h2\u003e\n\u003cp\u003eThis research was funded by the United States Department of Energy, Water Power Technologies Office, contract number DE-AC05-76RL01830.\u003c/p\u003e\n\u003ch2\u003eEthical Compliance Statement\u003c/h2\u003e\n\u003cp\u003eThis study did not involve the sampling of animals, and therefore, no permits were required. We affirm that all applicable international, national, and institutional guidelines for the care and use of organisms have been adhered to. As such, specific permissions related to animal sampling do not apply. We are prepared to provide any relevant documentation upon request.\u003c/p\u003e\n\u003ch2\u003eRegulatory Compliance Statement\u003c/h2\u003e\n\u003cp\u003eThis study was conducted in full compliance with all applicable regulations, including obtaining the necessary permits from the Federal Aviation Administration (FAA) for aerial operations. We adhered to all FAA guidelines to ensure safe and responsible use of unmanned aerial systems during the research.\u003c/p\u003e\n\u003ch2\u003eData Availability Statement\u003c/h2\u003e \u003cp\u003eData will be made available under the license CC-Attribution 4.0 via the Portal and Repository for Information on Marine Renewable Energy (PRIMRE) on the Marine and Hydrokinetic Data Repository (MHKDR) [\u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://mhkdr.openei.org/\u003c/span\u003e\u003cspan address=\"https://mhkdr.openei.org/\" targettype=\"URL\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e, accessed on 31 December 2024].\u003c/p\u003e\n\u003ch2\u003eAcknowledgment\u003c/h2\u003e\n\u003cp\u003eWe want to express our gratitude to several individuals and organizations whose contributions made this research possible. We acknowledge Casey Longbottom, David Novick, Brent Peterson, Carlos Ruiz, and Gabrielle Whitson for their expert field operations of the tethered balloon system and development of the imaging sensors. Special thanks are extended to Dave Weller, Aim\u0026eacute;e Lange, and Trevor Joyce at NOAA Southwest Fisheries Science Center (SWFSC) for authorizing access to the Marine Pollution Studies Laboratory at Granite Canyon and for sharing valuable gray whale observation data. We are also grateful for the hospitality and support from UC Davis staff, which facilitated our site operations. The student staff led by Benjamin Hess undertook the effort to review all the footage from this effort, and we greatly appreciate the many hours spent at computers documenting whales and other wildlife. We would like to thank the CODAR Ocean Sensors, Ltd. team for providing wave height data from their continuous monitoring project at the Granite Canyon Laboratory Site. Additionally, we acknowledge Toyon Research Corporation for providing the machine learning detection system used in our study. We also extend our appreciation to the Triton Initiative leadership team for their invaluable support throughout this project. Lastly, we thank the reviewers for their valuable feedback, which greatly improved the quality of our manuscript.\u003c/p\u003e"},{"header":"References","content":"\u003col\u003e\u003cli\u003e\u003cspan\u003eAguilar-Lazcano CA, Espinosa-Curiel IE, R\u0026iacute;os-Mart\u0026iacute;nez JA, Madera-Ram\u0026iacute;rez FA, P\u0026eacute;rez-Espinosa H (2023) Machine learning-based sensor data fusion for animal monitoring: Scoping review. Sensors 23:5732\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eAmerson A, Gonzalez-Hirshfeld I, Dexheimer D (2023) Validating a Tethered Balloon System and Optical Technologies for Marine Wildlife Detection and Tracking. Remote Sens 15:4709\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eAmerson AM, Harris TM, Michener SR, Gunn CM, Haxel JH (2022) A Summary of Environmental Monitoring Recommendations for Marine Energy Development That Considers Life Cycle Sustainability. J Mar Sci Eng 10:586\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eBarlow DR, Strong CS, Torres LG (2024) Three decades of nearshore surveys reveal long-term patterns in gray whale habitat use, distribution, and abundance in the Northern California Current. Sci Rep 14:9352\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eBicknell AW, Godley BJ, Sheehan EV, Votier SC, Witt MJ (2016) Camera technology for monitoring marine biodiversity and human impact. Front Ecol Environ 14:424\u0026ndash;432\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eChristie KS, Gilbert SL, Brown CL, Hatfield M, Hanson L (2016) Unmanned aircraft systems in wildlife research: current and future applications of a transformative technology. Front Ecol Environ 14:241\u0026ndash;251\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eClarfeld LA, Sir\u0026eacute;n AP, Mulhall BM, Wilson TL, Bernier E, Farrell J, Lunde G, Hardy N, Gieder KD, Abrams R (2023) Evaluating a tandem human-machine approach to labelling of wildlife in remote camera monitoring. Ecol Inf 77:102257\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eCorcoran E, Winsen M, Sudholz A, Hamilton G (2021) Automated detection of wildlife using drones: Synthesis, opportunities and constraints. Methods Ecol Evol 12:1103\u0026ndash;1114\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eCourbis S, Williams K, Stepanuk J, Etter H, McManus M, Campoblanco F, Pacini A (2024) Technology Gaps for Monitoring Birds and Marine Mammals at Offshore Wind Facilities. Mar Technol Soc J 58:5\u0026ndash;22\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eDanovaro R, Bianchelli S, Brambilla P, Brussa G, Corinaldesi C, Del Borghi A, Dell\u0026rsquo;Anno A, Fraschetti S, Greco S, Grosso M (2024) Making eco-sustainable floating offshore wind farms: Siting, mitigations, and compensations. Renew Sustain Energy Rev 197:114386\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eDanovaro R, Carugati L, Berzano M, Cahill AE, Carvalho S, Chenuil A, Corinaldesi C, Cristina S, David R, Dell'Anno A (2016) Implementing and innovating marine monitoring approaches for assessing marine environmental status. Front Mar Sci 3:213\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eDexheimer D, Whitson G, Cheng Z, Sammon J, Gaustad K, Mei F, Longbottom C (2024) Tethered Balloon System (TBS) Instrument Handbook. DOE Office of Science Atmospheric Radiation Measurement (ARM) Program \u0026hellip;\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eEaves SL, Staines G, Harker-Klimeš G, Pinza M, Geerlofs S (2022) Triton Field Trials: Promoting Consistent Environmental Monitoring Methodologies for Marine Energy Sites. J Mar Sci Eng 10:177\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eEnglish G, Lawrence MJ, McKindsey CW, Lacoursi\u0026egrave;re-Roussel A, Bergeron H, Gauthier S, Wringe BF, Trudel M (2024) A review of data collection methods used to monitor the associations of wild species with marine aquaculture sites. Reviews Aquaculture 16:1160\u0026ndash;1185\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eFarinelli S, Keith-Diagne L, Garnica J, Keiman J, Luther D (2024) Quantifying minimum survey effort to reliably detect Amazonian manatees using an unoccupied aerial vehicle (UAV) at an ex situ soft-release site. Latin Am J Aquat Mamm 19:42\u0026ndash;60\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eGarson J (2023) Water Power Technologies Office: 2022 Project Peer Review. EERE Publication and Product Library, Washington, DC. (United States)\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eGibbs JP, Snell HL, Causton CE (1999) Effective monitoring for adaptive wildlife management: lessons from the Galapagos Islands. J Wildl Manag : 1055\u0026ndash;1065\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eHaxel J, Zang X, Martinez J, Polagye B, Staines G, Deng ZD, Wosnik M, O\u0026rsquo;Byrne P (2022) Underwater Noise Measurements around a Tidal Turbine in a Busy Port Setting. J Mar Sci Eng 10:632\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eHemery LG, Mackereth KF, Gunn CM, Pablo EB (2022a) Use of a 360-degree underwater camera to characterize artificial reef and fish aggregating effects around marine energy devices. J Mar Sci Eng 10:555\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eHemery LG, Mackereth KF, Tugade LG (2022b) What\u0026rsquo;s in my toolkit? A review of technologies for assessing changes in habitats caused by marine energy development. J Mar Sci Eng 10:92\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eKellenberger B, Marcos D, Tuia D (2018) Detecting mammals in UAV images: Best practices to address a substantially imbalanced dataset with deep learning. Remote Sens Environ 216:139\u0026ndash;153\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eMarvin DC, Koh LP, Lynam AJ, Wich S, Davies AB, Krishnamurthy R, Stokes E, Starkey R, Asner GP (2016) Integrating technologies for scalable ecology and conservation. Global Ecol Conserv 7:262\u0026ndash;275\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003ePanigada S, Pierantonio N, Ara\u0026uacute;jo H, David L, Di-M\u0026eacute;glio N, Dor\u0026eacute;mus G, Gonzalvo J, Holcer D, Laran S, Lauriano G (2024) The ACCOBAMS Survey Initiative: the first synoptic assessment of cetacean abundance in the Mediterranean Sea through aerial surveys. Front Mar Sci 10:1270513\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eReilly CE, Larson J, Amerson AM, Staines GJ, Haxel JH, Pattison PM (2022) Minimizing Ecological Impacts of Marine Energy Lighting. J Mar Sci Eng 10:354\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eSharma S, Sato K, Gautam BP (2023) A Methodological Literature Review of Acoustic Wildlife Monitoring Using Artificial Intelligence Tools and Techniques. Sustainability 15:7128\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eStaines GJ, Mueller RP, Seitz AC, Evans MD, O\u0026rsquo;Byrne PW, Wosnik M (2022) Capabilities of an acoustic camera to inform fish collision risk with current energy converter turbines. J Mar Sci Eng 10:483\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eStephenson P (2020) Technological advances in biodiversity monitoring: Applicability, opportunities and challenges. Curr Opin Environ Sustain 45:36\u0026ndash;41\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eStewart JD, Joyce TW, Durban JW, Calambokidis J, Fauquier D, Fearnbach H, Grebmeier JM, Lynn M, Manizza M, Perryman WL (2023) Boom-bust cycles in gray whales associated with dynamic and changing Arctic conditions. Science 382:207\u0026ndash;211\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eThomas B, Holland JD, Minot EO (2011) Wildlife tracking technology options and cost considerations. Wildl Res 38:653\u0026ndash;663\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eWang ZA, Moustahfid H, Mueller AV, Michel AP, Mowlem M, Glazer BT, Mooney TA, Michaels W, McQuillan JS, Robidart JC (2019) Advancing observation of ocean biogeochemistry, biology, and ecosystems with cost-effective in situ sensing technologies. Front Mar Sci 6:519\u003c/span\u003e\u003c/li\u003e\u003c/ol\u003e"}],"fulltextSource":"","fullText":"","funders":[],"hasAdminPriorityOnWorkflow":false,"hasManuscriptDocX":true,"hasOptedInToPreprint":true,"hasPassedJournalQc":"","hasAnyPriority":false,"hideJournal":false,"highlight":"","institution":"","isAcceptedByJournal":true,"isAuthorSuppliedPdf":false,"isDeskRejected":"","isHiddenFromSearch":false,"isInQc":false,"isInWorkflow":false,"isPdf":false,"isPdfUpToDate":true,"isWithdrawnOrRetracted":false,"journal":{"display":true,"email":"[email protected]","identity":"marine-biology","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":false,"externalIdentity":"mabi","sideBox":"Learn more about [Marine Biology](https://www.springer.com/journal/227)","snPcode":"227","submissionUrl":"https://submission.nature.com/new-submission/227/3","title":"Marine Biology","twitterHandle":"","acdcEnabled":true,"dfaEnabled":true,"editorialSystem":"em","reportingPortfolio":"Springer Hybrid","inReviewEnabled":true,"inReviewRevisionsEnabled":false},"keywords":"Tethered Balloon System, Marine Wildlife Detection, Machine Learning, Infrared Sensors, Human Observation","lastPublishedDoi":"10.21203/rs.3.rs-5349011/v1","lastPublishedDoiUrl":"https://doi.org/10.21203/rs.3.rs-5349011/v1","license":{"name":"CC BY 4.0","url":"https://creativecommons.org/licenses/by/4.0/"},"manuscriptAbstract":"\u003cp\u003eThis study investigates the capabilities of a tethered balloon system (TBS) for detecting and monitoring marine wildlife, primarily focusing on gray whales (\u003cem\u003eEschrichtius robustus\u003c/em\u003e) and various avian species. Over 55.7 h of aerial and surface footage were collected, yielding significant findings regarding the detection rates of marine mammals and seabirds. A total of 59 gray whale, 100 avian, and 6 indistinguishable marine mammal targets were identified by the airborne TBS, while surface-based observations recorded 1,409 gray whales, 1,342 avian targets, and several other marine mammals. When the airborne and surface cameras were operating simultaneously, 21% of airborne whale and 34% of airborne avian detections were captured with the airborne TBS camera and undetected with the surface-based camera. The TBS was most effective at altitudes between 50 to 200 m above ground, with variable-pitch scanning patterns providing superior detection of whale blows compared to fixed-pitch and loitering methods. Notably, instances of airborne detections not corroborated by surface observations underscore the benefits of combining aerial monitoring with traditional survey techniques. Additionally, the integration of machine-learning (ML) algorithms into video analysis enhances our capacity for processing large datasets, paving the way for real-time wildlife monitoring. Of the total number of blows detected by an ML algorithm, the percentage of blows identified by a human analyst was greater than that uniquely detected by the algorithm. Notably, more unique detections by the ML algorithm occurred during daylight, suggesting that sun artifacts may hinder human detection performance, thereby highlighting the added value of ML under these conditions. This research lays the groundwork for future studies in marine biodiversity monitoring, emphasizing the importance of innovative aerial surveillance technologies and advanced imaging methodologies in understanding species behavior and informing conservation strategies for sustainable marine energy, offshore wind development, and other marine resource management efforts.\u003c/p\u003e \u003cp\u003e \u003c/p\u003e","manuscriptTitle":"Enhancing Marine Wildlife Observations: The Application of Tethered Balloon Systems and Advanced Imaging Sensors for Sustainable Marine Energy Development","msid":"","msnumber":"","nonDraftVersions":[{"code":1,"date":"2024-12-02 15:42:05","doi":"10.21203/rs.3.rs-5349011/v1","editorialEvents":[{"type":"communityComments","content":0},{"type":"decision","content":"Revise and Resubmit","date":"2024-11-22T12:09:09+00:00","index":"","fulltext":""},{"type":"reviewerAgreed","content":"","date":"2024-11-04T00:11:54+00:00","index":0,"fulltext":""},{"type":"reviewersInvited","content":"","date":"2024-11-03T20:40:42+00:00","index":"","fulltext":""},{"type":"editorAssigned","content":"","date":"2024-10-28T18:37:44+00:00","index":"","fulltext":""},{"type":"submitted","content":"Marine Biology","date":"2024-10-28T13:38:47+00:00","index":"","fulltext":""}],"status":"published","journal":{"display":true,"email":"[email protected]","identity":"marine-biology","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":false,"externalIdentity":"mabi","sideBox":"Learn more about [Marine Biology](https://www.springer.com/journal/227)","snPcode":"227","submissionUrl":"https://submission.nature.com/new-submission/227/3","title":"Marine Biology","twitterHandle":"","acdcEnabled":true,"dfaEnabled":true,"editorialSystem":"em","reportingPortfolio":"Springer Hybrid","inReviewEnabled":true,"inReviewRevisionsEnabled":false}}],"origin":"","ownerIdentity":"55aed99a-647d-460f-9449-b1bf75533b4e","owner":[],"postedDate":"December 2nd, 2024","published":true,"recentEditorialEvents":[],"rejectedJournal":[],"revision":"","amendment":"","status":"published-in-journal","subjectAreas":[],"tags":[],"updatedAt":"2025-04-07T16:05:43+00:00","versionOfRecord":{"articleIdentity":"rs-5349011","link":"https://doi.org/10.1007/s00227-025-04618-3","journal":{"identity":"marine-biology","isVorOnly":false,"title":"Marine Biology"},"publishedOn":"2025-04-03 15:57:21","publishedOnDateReadable":"April 3rd, 2025"},"versionCreatedAt":"2024-12-02 15:42:05","video":"","vorDoi":"10.1007/s00227-025-04618-3","vorDoiUrl":"https://doi.org/10.1007/s00227-025-04618-3","workflowStages":[]},"version":"v1","identity":"rs-5349011","journalConfig":"researchsquare"},"__N_SSP":true},"page":"/article/[identity]/[[...version]]","query":{"redirect":"/article/rs-5349011","identity":"rs-5349011","version":["v1"]},"buildId":"qtupq5eGEP_6zYnWcrvyt","isFallback":false,"isExperimentalCompile":false,"dynamicIds":[84888],"gssp":true,"scriptLoader":[]}