Making aerial wildlife surveys more reliable: Kayla Davis and Elise Zipkin research published in "Ecology and Evolution"

Wildlife policy and management decisions often rely on estimates of animal abundance, so inaccurate counts can have negative consequences, especially with rarer species. Biased information can seriously misinform policy decisions such as where to place energy farms or where to dig for oil and gas. Reliable analyses of aerial surveys can lead to better decisions for wildlife and the environment.

When Kayla Davis went to the Gulf of Mexico to collect data to help estimate the abundance and distribution of different marine birds, she ended up on a different quest. After witnessing surveyors’ difficulties counting individual birds from the air, she decided to change her focus and begin an investigation into aerial data collection methods.  “I couldn’t find a synthesis on the challenges of analyzing aerial survey data. So hopefully, this new research will be a useful tool for those using aerial survey data to understand wildlife patterns.” Kayla and her team worked in collaboration with the U.S. Fish and Wildlife Service and the U.S. Geological Survey.

“Aerial surveys are often the best approach for collecting data in vast or remote landscapes”, Elise Zipkin, senior author, said senior author Elise Zipkin. “However, analyses of aerial surveys can easily lead to biased inferences of wildlife abundance if not careful, because there are several confounding factors affecting the detection of species and individuals.”

The team used a three-pronged approach to tackle the problem. They analyzed waterbird data from the Gulf of Mexico Marine Assessment Program for Protected Species (GoMMAPPS), evaluated the literature on aerial surveys from the last 50 years, and created an online flock-counting quiz, which they administered to about 100 trained aerial observers.

Kayla and her colleagues identified three major detection errors with aerial surveys: non-detection, the failure to count animals that are present; counting error, when group size is counted inaccurately; and misidentification, incorrectly identifying species. Each of these issues can lead to biases in wildlife estimates. For example, when a survey is focused on multiple species, it often resulted in an over-count of one species and an undercount of another. In addition, the online survey revealed that observers tended to undercount flock sizes, even at small group sizes (less than 30 individuals), but that the margin of error increased as the group size increased, sometimes off by as much as 50 percent.

Kayla said, “A good first step when designing an aerial count survey, is to determine which issues are probable under given survey conditions. “f the target is a single, solitary species, non-detection may be the major source of bias. However, when similar-looking species aggregate such as in our GoMMAPPS case study, all three detection errors should be carefully considered.”

If researchers recognize conditions that are likely to present data issues, they can incorporate other resources to improve the survey quality. High-definition photographs and videos, pooling species records by family, and ordinal modeling using binned count data—a method of grouping data by certain levels or values—are ways to mitigate errors. In cases where those options are not possible, researchers may need to use additional data sources.

Elise said, “This work is the first to provide a comprehensive synthesis of the issues that affect accurate estimation of wildlife populations from aerial survey data, one of the most common approaches of collecting data. Our paper provides a roadmap of approaches to overcome these issues and, importantly, highlights future research directions.”