Abstract
Wind energy is a sustainable source of electricity, and offshore winds are a particularly rich resource. To harness this energy with minimal impacts on the environment requires understanding the effects of offshore wind development on wildlife such as seabirds and bats. Quantifying these effects on seabird and bat populations is challenging due to the remoteness of and harsh conditions at offshore locations.
DOE’s Wind Energy Technology Office has funded the development of technology for understanding wildlife impacts to accelerate the development of offshore wind energy in the US. The ThermalTracker-3D (TT3D) technology was developed by the Pacific Northwest National Laboratory (PNNL) with DOE funding as a method for continuously monitoring bird and bat activity at remote locations, such as those offshore. Its detection and 3-D tracking capabilities were validated on land in 2019. To validate its performance offshore, in 2021 a marinized prototype TT3D was integrated with a Wind Sentinel™ buoy and deployed offshore. This report describes that initial offshore validation of the TT3D system for monitoring and quantifying bird and bat activity at offshore wind energy sites.
The buoy was deployed with the integrated ThermalTracker-3D at an area planned for wind energy development, 25 nautical miles off the coast of northern California. The buoy’s primary mission is to characterize the wind energy resource by measuring the wind speed and direction in the air column, up to 250 meters above the water surface. It provided a platform with power and a data link to shore for the TT3D. The TT3D camera assembly was mounted atop a camera stabilization system to hold the cameras relatively steady as the buoy was subjected to wave motion. The stabilizer stopped functioning early in the deployment; however, the TT3D continued to operate, and the situation provided an opportunity to study platform motion effects on the TT3D performance.
The prototype TT3D system operated continuously from May 4 through Aug 13, 2021 (14 weeks), at which time the buoy generator failed. Using data collected during the operational period, the technology was evaluated in terms of its reliability, output data quality, motion effects, hardware component performance, and platform integration.
The software operation and reliability largely exceeded expectations. The software ran autonomously without failure throughout its deployment and the associated scripts managing the disk space successfully maintained a healthy margin of free space, while those composing the status messages transmitted to shore through the buoy system ran reliably, providing continuous insight into the TT3D system’s health and status. Hardware – cameras, GPS, computer -- performed reliably and the platform integration worked well; the mechanical integration was secure and no failures of any connections or attachment points have occurred to date.
During the operational period, the TT3D recorded 2,440 valid 3D flight tracks, many of which were recorded during non-daylight hours. A review of a sample of the TT3D data revealed that the detection rate was lower than expected – 44% compared to 89% from previous testing. The settings that control the sensitivity of the detection algorithm were tuned remotely during the deployment, and the detection rate increased to 52%, indicating that further tuning could possibly have produced additional improvement. The platform motion may also have reduced the detection rate. A random sample of 205 detections was reviewed and 80% were recognizably birds. The other 20% appeared to be blurred by motion and were unrecognizable.
Camera motion was characterized by the rate of pitch, roll and yaw in degrees per second that occurred as an animal flew through the field of view. Most detections occurred with motion rates less than 15 deg/sec. The detections during motion greater than 5 deg/sec were more likely to be blurred and/or unrecognizable. The validity of the 3D tracks as determined by the geometry of the stereoscopic field of view was found to be more affected by the animal’s distance from the camera than by the camera motion. The distance effect is expected due to the nature of the stereo vision processing, which becomes less accurate for far away small objects. Extreme camera motion greater than 30 deg/sec did reduce the probability of a 3D track being valid.
This study successfully built an offshore prototype capable of autonomous long-term operation in an offshore environment, successfully integrated the prototype with a buoy, confirmed the reliability of the software, improved our understanding of the effects of motion, and collected seabird data for analysis and species identification, including data not previously available such as nocturnal activity and flight heights. Recommendations for future efforts and system improvements are as follows:
- Improve the detection rate by developing a workflow for tuning the detection sensitivity to the operating environment prior to deployment.
- Improve camera stabilization and add motion compensation in the software, as needed, to minimize motion effects from a floating platform.
- Optimize the computer and storage for constrained environments using the latest technology to increase the flexibility of the system.
- Develop automated taxonomic identification using data collected during this study combined with human observer data collected during a coincident survey of species in the buoy location.