Solar power farms have many advantages in saving energy and using clean sources of electricity, but in the United States they are bringing a problem that affects the environment: the death of birds.
- Study associates exposure to gadgets' blue light with increased cancer risk
- USP adhesive sensor can track substances in sweat and detect new diseases
- Does it work? Company announces launch of anti-virus paint in Brazil
Companies have been finding dead bird carcasses on the ground for years, and since then, the question of what is causing this problem has haunted researchers. But in 2013, academics, environmental organizations and utility companies came together to form the Avian Solar Working Group, a group dedicated to developing strategies to mitigate bird deaths in solar facilities across the country.
Misti Sporer, a member of the group and chief environmental scientist at the North Carolina electricity company Duke Energy, says that initially no one could say what it meant to see a dead bird in these locations and that it was a challenge to obtain such data. According to a pioneering study on the subject in 2016, it was estimated that hundreds of large-scale solar farms in the United States could kill approximately 140,000 birds each year.
The number is low, representing less than a tenth of 1% of the estimated number of birds that are killed by fossil fuel plants, which can happen with electrocution, poisoning and collisions. However, the researchers believe that the number could triple as new solar farms start operating.
Even with the data in hand, it is not yet clear what the connection of the deaths with the panels is, but one of the main theories is that the glow emitted by the ceilings causes the birds to be confused with lakes. Then, when they dive into the fake lake, they end up in a deadly crash. The hypothesis, however, is based on a human perspective. "Do birds at least see the same way as humans? We need to collect more data to form a complete image", says the scientist.
Another interesting project also seeks to understand what happens to prevent more birds from dying every year. The United States Department of Energy funded, under a $ 1.3 million contract, researchers from Argonne National Laboratory, Illinois, to develop an artificial intelligence dedicated to studying the problem.
AI will analyze the behavior of birds in large-scale solar installations across the country, in the hope that the data collected will help ornithologists move forward with the discovery. Yuki Hamada, a biophysical scientist at Argonne and leader of the study, talks about the extreme importance of reducing this environmental impact. "These avian problems are a concern and something that the renewable energy industry wants to understand and mitigate," he says.
Challenges
Only a few regions in the US have regulatory standards that require reporting of bird deaths on the premises, but most large-scale solar farms are not too concerned about the problem. Those who are aware of what is happening, however, do not have sufficient resources for data collection and can only send surveyors to count dead birds once a month. According to the scientists, it is necessary to do the data verification in real time to have more conclusive results.
With the artificial intelligence created by Argonne, it will be possible to train the algorithm to recognize birds based on sizes, shapes and colors, identifying whether they are flying over the solar panel or just perched on it. Even before the idea of creating this AI, Adam Szymanski, software engineer at Argonne and project leader, was working on an initiative that aimed to automatically detect small drones in the air.
"The image learning research we are doing is a bit unique because we don't want to just classify an object into a single image. We need to classify a small, fast-moving object over time. So if a bird is flying, in some frames you will see a point and in others you will see their wings out, and we need to track objects as they move around the camera ", he explains.
Another problem to analyze this data is the fact that many of these solar farms are in remote places and do not have the necessary infrastructure for machine learning. "You must imagine that solar installations have energy because they generate. But they don't have electrical outlets connected to the panels," says Szymanski.
So, for the system to be able to capture this data, it will need extremely efficient hardware that works with its own small solar panels, or even with batteries. The system must also be able to store a large amount of data in real time.
Argonne then chose to use hardware developed by a company called Boulder AI, which monitors pedestrians and vehicle traffic. With a system that has a small camera, the current process involves collecting data from two solar facilities in Illinois, gradually expanding the program to dozens of other locations across the United States.
Initially, artificial intelligence will identify birds that enter without their field of vision, but only with more specific training will it be possible to differentiate their behavior with the panels. With the results in hand, researchers will be able to understand the possible causes of these deaths. "This technology allows us to envision a world that we do not normally see so that we can operate in a less impactful way for wildlife", adds Sporer.