Wind as Driver of Bird and Bat Abundance, Flight Direction, Altitude, and Speed on the North Atlantic Shelf

Technical Summary: Wind as Driver of Bird and Bat Abundance, Flight Direction, Altitude, and Speed on the North Atlantic Shelf

Problem Statement
The expansion of offshore wind energy along the U.S. northeast coast necessitates a robust understanding of flying animal (bird and bat) behavior to accurately quantify collision risks with turbines. Current collision risk models often rely on assumptions regarding animal behavior, such as avoidance rates and flight parameters, which are difficult to quantify. While factors like wind speed and direction are known to influence flight activity, previous studies have largely relied on historical climate data with limited temporal and spatial resolution (e.g., ERA5 reanalysis) or focused on European waters (North Sea). Furthermore, few studies have simultaneously analyzed animal abundance, flight height, and flight speed in relation to high-resolution, co-located wind measurements in U.S. waters. There is a critical need to understand how meteorological conditions drive animal presence and flight dynamics to improve collision risk assessments.

Methodology
This study presents a novel methodological framework integrating avian radar and wind lidar data collected from a research barge deployed on the Northeastern Shelf of the United States (40.9° N, 70.79° W) during a 5-week window in autumn 2024.

• Data Acquisition: An omnidirectional S-band radar (MERLIN True3D) captured bird and bat tracks, while two profiling lidars (WindCube v2.1 and Halo XR+) provided high-resolution wind speed and direction profiles.
• Data Processing: Radar tracks were filtered based on detection range (200–1100 m), flight speed (<35 m/s), and reflectivity (−25.5 to 10 dBsm) to exclude insects and low-quality data. Tracks were matched with concurrent wind data at flight altitude.
• Clustering Approach: To address the radar's inability to distinguish species, a hierarchical clustering methodology was developed using reflectivity as a proxy for animal size. The algorithm partitioned tracks into two groups ("small" and "big") based on flight behavior distributions (hourly abundance, flight height, direction, and speed). The optimal clustering split was determined at a reflectivity threshold of −8.54 dBsm.
• Statistical Analysis: Generalized Additive Models (GAMs) were employed to assess the drivers of hourly abundance for each cluster. Additionally, cumulative distribution functions were analyzed to investigate relationships between flight direction, height, and speed relative to wind conditions (tailwinds, crosswinds, headwinds).

Key Results
The analysis of 67,410 filtered tracks revealed distinct behavioral patterns driven by wind conditions and animal size:

• Abundance Drivers: For both size clusters, the abundance of animals in the previous hour was the strongest predictor of current abundance. Among environmental factors, wind speed and solar azimuth were most impactful. High wind speeds generally suppressed abundance, particularly for the "big" cluster at speeds exceeding 15 m/s.
• Size-Based Behavioral Differences:
  • Small Cluster (Lower Reflectivity): These animals (hypothesized to be smaller-bodied migrants) exhibited concentrated flight directions aligned with wind patterns. They flew at a variety of altitudes, often higher, and showed a strong propensity to fly with tailwinds (72.4% of tracks). Their abundance showed distinct temporal bursts consistent with migratory pulses.
  • Big Cluster (Higher Reflectivity): These animals (hypothesized to be larger-bodied residents) flew in a wide variety of directions but were concentrated at lower altitudes (82.3% below 100 m). They were more likely to fly in unfavorable wind conditions (crosswinds and headwinds) compared to the small cluster.
• Flight Dynamics:
  • Direction: Flight direction distributions shifted with wind direction. At higher wind speeds, both clusters showed a narrowing of flight direction distributions, with a higher percentage of animals flying with tailwinds.
  • Altitude: Flight height correlated with wind speed; more animals were detected at higher altitudes during periods of faster winds. Animals generally flew near the altitude of maximum wind assistance.
  • Speed: Ground speed increased with wind assistance, while air speed remained relatively constant (approx. 5 m/s for small, 8 m/s for big). This suggests larger animals maintain higher air speeds, potentially indicating higher energy input.

Significance and Claims
The authors claim this study provides the first insights into animal behavior at offshore wind sites using paired, high-resolution radar and lidar data. The primary contributions are:

1. Methodological Framework: The paper establishes a framework for integrating radar and lidar data to analyze flying animal behavior, offering a template for future offshore studies.
2. Behavioral Insights: The results demonstrate that wind is a primary driver of animal presence, flight direction, altitude, and speed. The study highlights that animal behavior is not uniform; distinct size groups (likely representing different ecological strategies such as migration vs. residency) respond differently to wind conditions.
3. Implications for Collision Risk: The findings suggest that collision risk models treating flight direction, height, and speed as static species-specific constants may be insufficient. The study indicates that:
   • The proportion of animals within the turbine rotor-swept zone (RSZ) decreases as wind speed increases.
   • Ground speed increases with wind assistance, affecting transit time through the rotor.
   • Animals actively adjust flight parameters (e.g., seeking tailwinds or specific altitudes) based on wind conditions.

The authors conclude that while the study is limited to a specific 5-week autumn window, the combined analysis of radar and lidar data reveals critical behavioral nuances that can refine collision risk assessments. They emphasize that future work should utilize longer-term datasets to develop predictive models for collision risk forecasting.