Accidents Will Happen: A dataset of military drone crashes

Forensic experts investigate crash of US Predator near Incirlik Air Base, Turkey, Feb, 2016. (Depo photos)

Drone Wars is today publishing a dataset of just over 250 large military drone crashes that have taken place over the past decade (2009-2018). The full dataset is available online here.  This post is a brief summary of the data but there is a great deal more detail in our accompanying report which is available here.

Although there continues to be some disagreement about the classification of Unmanned Aerial Vehicles (UAVs), most adhere to the NATO system which divides them into three broad categories based on weight. Class I includes micro, mini and small drones, all under 150kg. Class II are drones that weigh over 150kg but less than 600kg, often referred to as ‘tactical’ drones, while Class III drones are those, including the US Predator and Reaper, that weigh over 600kg.

Click to open report

Our dataset details crashes of Class II and Class III UAVs. These are primarily operated by military forces, but crashes of such systems being operated by companies that develop them, or by civil security forces (such as US Department of Homeland Security), are included.  The information within our dataset comes from three primary sources: official air force accident investigations, responses to freedom of information requests and individual press reports. However, it is acknowledged from the outset that due to the secrecy surrounding the use of these systems, other crashes will have occurred which have not been publicised.

There are 254 drone crashes detailed in our dataset. Thirty-five (14%) are Class II with 219 (86%) being Class III drones. The vast number of US drone crashes in the database reflects the dominance of the US in using these systems.  Out of the 254 accidents in the database, 178 (70%) were being operated by various branches of the United States military. While a cursory glance at the data would appear to show that the number of US drone crashes have declined over the past three years, official aviation mishap statistics issued by the Pentagon show that in fact they remain relatively stable(see report for more details). What appears to be happening is that the public acknowledgement (and therefore reporting) of US crashes at the time they occur has declined while the release of official accident investigation reports take a significant amount of time.  At the time of writing, the most recent US accident investigation report of a USAF drone crash (released in February 2019) detailed a crash that occurred in August 2017.

The crash data also reflects the increasing proliferation of large military drones over the past decade.  In 2009 only two other countries besides the US had crashes of these types of drones. In 2018, nine countries had crashes of these class of drones alongside the US.  Overall 19 countries appear in the dataset as having had crashes of large military drones.  After the US, the UK has had the largest number of drone crashes (14) reflecting its long involvement in operating this type of system. Other states with a significant number of crashes include Israel, Turkey, India and Pakistan. In addition, we have included drone crashes when they were being operated by manufacturing companies, and two drones which crashed in 2014 while being operated by the UN peacekeepers.


US Reaper drone crashed in Nevada, Dec 2012
US Reaper crashed in Syria, July 2016


Alongside the fact of the crash, there are some details of the cause for just over half (56%) of the crashes. While these details primarily come from official US accident investigation reports and are therefore heavily skewed towards US drone crashes, the reasons are likely to be common across all drone crashes.  These include mechanical failure (such as tails shearing off or propellers snapping) , communications problems (known as ‘lost link’), engine failure (often due to oil or coolant loss), weather problems (including lighting strikes) and pilot error. While non-state groups regularly claim responsibility for crashed drones, we attribute between five and nine of the crashes in our dataset to being shot down.

Analysis of the data enables us to gain a good understanding of when, on average, drone crashes take place. 64% of the crashes took place while the drone was in mid-flight, while 20% occurred at the point of landing.  8% crashed during the take-off phase, with a small number of crashes (1%) taking place while the drone was taxiing along the runway.  For 7% of the recorded accidents, it is unknown at what stage the crash occurred.

While to some, particularly in the military, the use of remote controlled drones appears to be becoming normal, the reality is that the technology is far from mature and, as the data demonstrates, accidents occur frequently – around twice per month on average over the past decade.

There continues to be a great deal of debate about the safety and security implication of the rising use of small consumer drones.  However discussion in both the US and the UK about opening up the skies to large drones is largely taking place behind closed doors and between industry  lobbyists and government officials.  The fact that large military drones regularly fall from the sky must call this push to allow such systems to be flown in civilian airspace into questions. We hope that the release of this data will encourage a much wider debate on this issue within legislatures, the media and the public.


British Watchkeeper drone crashed near Salisbury, Nov 2015
Wing Loong, likely on demonstration flight, crashed in Pakistan, July 2016


See also: New military drone crash data undermines MoD case to fly Protector drone in UK

Leave a Reply