One warm evening this fall a pair of U.S. Army UH-60M Blackhawk helicopters cruised low over New York City’s Staten Island, providing security for the United Nations General Assembly’s annual meeting in nearby Manhattan. Just after sunset a shoe box–size airborne object collided with one of the choppers, damaging its main rotor blade, window frame and transmission system. Inspection at a nearby airfield revealed evidence of something that had never happened before—a civilian drone had plowed into a crewed craft in U.S. airspace. That sent the Army, the National Transportation Safety Board and other government agencies scrambling to investigate how and why this had happened.

Those questions, although important, are less interesting to aviation researchers than determining just how much damage increasingly common drones—like the 1.4-kilogram quadcopter in the Staten Island incident—can potentially inflict on helicopters and small, low-flying aircraft. Data on the risk that lightweight drones pose to aviation safety is scarce, even as consumer and commercial drone sales take off. The U.S. Federal Aviation Administration (FAA) projects that almost seven million drones could be sold in the U.S. by 2020.

Pilots already take the potential for drone collisions very seriously, sometimes putting potentially lifesaving missions on hold for fear of such a strike. In August firefighters battling a large blaze on Montana’s Rice Ridge temporarily shut down helicopter operations when an unauthorized drone was spotted in the sky nearby. The following month government agencies issued terse warnings for civilians to keep their drones far away from low-flying aircraft being used to locate and evacuate Hurricane Harvey victims in Houston. It remains unclear, however, whether the risk of a drone collision is big enough to outweigh the need for those aircraft to carry out their missions.

Low-flying helicopters—such as those conducting search-and-rescue operations—likely face the most serious risk of collision because they often occupy the same airspace as multirotor consumer and commercial drones, says Ian Horsfall, professor of armor systems at Cranfield University and the Defense Academy of the United Kingdom. “Not only is there a threat of poorly controlled drones being in the wrong place, but also the nature of helicopter flight is such that [even] a properly licensed and piloted drone might be hit by a helicopter that strays into its airspace,” or vice versa, Horsfall says. The Army Blackhawk over Staten Island had been flying at an altitude of about 150 meters when it was hit. Hobbyist drones are barred from most parts of New York City but are permitted in some locations—including Staten Island’s LaTourette Park—as long as they stay below about 120 meters.

Fixed-wing aircraft may not spend as much time in quadcopter-accessible airspace as helicopters, but they could sustain severe damage from a collision. Barely a month after the Staten Island incident, a drone smashed into a twin-engine propeller plane near Quebec City. The Beechcraft King Air A100 aircraft was approaching a runway at Jean Lesage International Airport when its left wing collided with the drone at just above 450 meters. A ground inspection later revealed only minor harm: scratches on the wing’s top surface and scrape marks on a de-icing system. But Marc Garneau, Canada’s Minister of Transport, said during a press conference following the incident that the collision could have had “catastrophic” results if the drone had crashed through the cockpit window or damaged an engine.

Indeed, some of the most dangerous scenarios for low-flying aircraft involve drones smashing windshields or knocking out main engines. The hazards vary depending on the type of drone and piloted aircraft involved, according to Javid Bayandor, founder and director of the Crashworthiness for Aerospace Structures and Hybrids (CRASH) Lab at the University at Buffalo, The State University of New York. Previously based at Virginia Polytechnic Institute and State University, Bayandor and his colleagues have spent years putting together data on more than 150 drone models, and using computer simulations to study the potential threat to aircraft. “We're not only categorizing the drones and figuring out what their damage footprint would be on different locations on a number of aircraft types and engines,” Bayandor says. “We’re also looking into possible damage-mitigation methods."

A 2016 study by the U.K.’s Department for Transport and Military Aviation Authority concluded drone collisions could be far worse than the impact of soft-bodied birds. Tests conducted for the U.K. study involved shooting custom-built drones of different sizes at helicopter and large aircraft windshields, and found that a 1.5-kilogram drone could sometimes break through.

Consumer drones pose far less of a threat to high-flying commercial aircraft; a jet airliner is in drone airspace only during takeoff and landing, and is very fast-moving target for a quadcopter. The greatest risk would probably be a drone getting sucked into one of its turbines. Bayandor’s CRASH Lab once created a video showing a simulation of how a drone might get shredded into fragments inside a jet engine. “Those fragments can get shot around the engine, sometimes at very high velocities, and can continue to propagate the damage,” he says.

The overall chance of drone collisions with any type of aircraft actually remains relatively low, according to researchers. Without much data available about drone collisions or even near-misses, a March 2016 report by George Mason University’s Mercatus Center studied bird strikes as a proxy for quadcopters (pdf). The researchers found that although there are an estimated 10 billion birds in U.S. airspace, there are just hundreds of bird strikes per year. “Compared to the enormous population of birds, damaging bird strikes are rare,” according to the report, which concluded the risk to the airspace caused by drones weighing up to two kilograms is minimal.

“There weren’t any drone collisions until now, and when you consider how many flights there are, I think it confirms my suspicion that there is not this insane risk of planes going down regularly as sometimes the media had us fear would happen,” says Mykel Kochenderfer, an aeronautics engineer and director of the Stanford Intelligent Systems Laboratory at Stanford University. Still, he described the recent U.S. and Canada incidents as “a wake-up call to drone operators” about their potential impact on aviation safety.