In charge that evening was 44-year-old Captain Alexander Cherepanov, with 9,229 flying hours under his wing, including 1,488 on this very type of aircraft. With him were four others: co-pilot Ermek Dzhangirov, a navigator, a flight engineer and a radio operator. ^[2]  

The Ilyushin was flying into a country whose own aviation industry had recently seen a rapid expansion. The open skies policy of a liberalising India introduced permits for private operators, meaning more planes and more fliers. As air traffic spiked, so did the need for enhanced safety measures. Just two months earlier, a panel had been appointed to probe air safety concerns. Its report was due.

In the 1990s, the English-speaking skills of pilots from the erstwhile Soviet Union were notably uneven. Maddeningly for the rest of the world, the former Soviet states also followed the metric system, mapping distances in metres and kilometres while most countries worked with feet and nautical miles. This complicated communications and opened up room for error so it was essential for air traffic controllers to speak slowly and enunciate clearly.

At Delhi’s Indira Gandhi International Airport (IGI), the ATCs had an additional challenge to manage. Ascending and descending passenger aircraft used the same stretch of stairway to and from the sky, largely since most of the airspace around the IGI was controlled by the Indian Air Force. Personnel had to manage comings and goings in the single corridor to perfection.

Separating aircraft in the single corridor was not new to VK Dutta, the ATC in-charge that evening, a man with 16 years of experience in aviation. He had to ensure that the planes stayed 1,000 feet apart, as per International Civil Aviation Organisation (ICAO) standards. In front of him was the old-fashioned primary radar where the “targets”— moving blips on the screen—gave him information on flight direction but, critically, not on altitude. For that, he would have to rely on the pilot’s word. Major Indian airports were in the process of installing secondary surveillance radar, indicating flight altitude and speed. But the technology was not yet at Dutta’s disposal. 

Flying from Islamabad to Delhi in a US Air Force plane, Captain Timothy Place was between 20,000 and 14,000 feet that evening when, fewer than 40 miles away, he saw an orange cloud burn up the fading sky. Was it a rocket launch, he wondered? The lurid colours seemed to intensify.

“That's not a missile, is it?” he asked Captain Rodney Marks on his right. Another plume appeared from the cloudy morass. Seconds later, it looked like both streams of fire were heading for the ground.

Stunned, but realising they had likely seen an accident, they informed Dutta, who immediately radioed both planes, desperately trying to establish some contact. It was too late.

At about 9pm, as Prannoy Roy finished reading for The News Tonight, the daily English bulletin on Doordarshan, wire services broke the story of a mid-air collision just outside Delhi. Most journalists had already left for the day, and the crowd at the studios of NDTV, a private news production company, was thinning. Writer and film-maker Natasha Badhwar, then a 25-year-old cameraperson two years out of Jamia Millia Islamia University, quickly volunteered herself for the story. Along with reporter Radhika Bordia and camera assistant Kanan Patra, she got into an office vehicle and headed towards Charkhi Dadri. As they approached the site, they saw a massive inferno raging in the distance, ringed by smaller fires.

They got out of the car and began to walk through the debris and darkness. When Badhwar struck something, the group stopped and switched on the camera light. A stiff body swam into view, dead, but not burnt or broken. The man looked in death as he might have in life. How do you film a dead body, Badhwar wondered. “There was a sense of horror I could not shut out,” she said. “There was a sense of incredulity: How could this happen? How could two planes collide?”

Through the night, journalists arrived to scenes of damage and disarray. Passports, bags of food, toys, slippers, wallets, wrenched-open suitcases and airplane seats were flung around. Disembodied human limbs lay scattered like confetti. The full horror would not be apparent until sunrise, but the acrid stench of burning fuselage and singed remains had already begun choking the air.

At one point, the grandson of a prominent politician waltzed in, ^[4] chaperoned by a group of sidekicks and handlers. He had no reason to be there but quickly became the centre of attention. When asked what he was doing there, he smarmily replied, “Plane curiosity.” 

Through the night, villagers tried to find survivors as emergency services arrived. By the morning, several bodies were moved into a school which had been converted to a makeshift morgue. Debris had scattered over a radius of several kilometres, complicating search and rescue operations.

KPS Nair was among the first team of investigators and diplomats to reach the scene as dawn broke. Then a Deputy Director with the Directorate General of Civil Aviation (DGCA), Nair had been in a meeting on air safety when news filtered in the previous evening. The irony was inescapable. “My first thought was, my god!” he recalled over the phone in June 2020. “Why do such things happen in spite of the efforts of human beings to make better flying machines and systems?”

As lead investigator, his first impulse was to locate the “black boxes”, vermilion-coloured crash-proof casings that held the cockpit voice recorder (CVR) and the flight data recorder (FDR). The CVR, which saved the last 30 minutes of cockpit conversation, could provide precious insight into what the crew had said to each other. The FDR would provide readings on altitude, speed, flight path and engine power to help build a picture of both planes’ vital parameters. “This,” Nair explained, “is the most important tool in the hands of an investigator from the accident site.”

Both black boxes were found before nightfall on the day following the accident. The cockpit of the Saudi aircraft was stuck deeper in the ground, irrevocably damaged, none of its instruments readable. All three of its altimeters were destroyed in the fire. 

The Kazakh plane had suffered less damage and yielded four altimeters. Two of them, curiously, showed different readings: 4,443m (14,576 feet) and 4,540m (14,895 feet). The other two were unreadable. 

Nair scoured the fields, analysing the debris scattered across two sites seven kilometres apart, searching for clues in the carcasses of the two planes. “We approach any accident with an open mind,” he said. “We ask, did everyone involved apply the necessary procedures? The contributing factors we look at include the ATC communications, pilot error, engineering and maintenance aspects.”

AK Chopra, then head of air safety for the North region at the DGCA, also reached the site with his team. Chopra sensed something puzzling, even counter-intuitive. The frontal structure of the Kazakh flight was mostly intact, which indicated that this hadn’t been a head-on collision. If the Kazakh plane was supposed to have been higher (15,000 feet), why did a primary assessment suggest that it had hit the Saudi plane from below?

By mid-afternoon, just 19 bodies had been claimed from the morgue. Rescue workers tried to pick through the debris, retrieving limbs and luggage. Badhwar even encountered the studio in-charge from Jamia’s Mass Communication Research Centre, mourning the loss of a younger brother who had been on the flight. 

For at least 15 hours after the crash, there was little sense of sanctity about the site, which was only cordoned off much later. ^[5] Some people rifled through the wreckage and corpses, trying to make off with the spoils. But for the most part, people were kind and helpful, bringing sheets from their homes for the bodies, volunteering tractors in search efforts, and offering tea to families.

Over time, the horror of losing a loved one morphed into the numbing pain of dealing with paperwork. Relatives scrambling to apply for compensation and obtain death certificates shuttled from one office to another. The Saudis announced compensation of up to GBP 12,000 for the families of passengers, and some compensation ^[6] for the villagers in whose fields the flights had crashed. 

In the end, 94 bodies were charred beyond recognition or totally mutilated. They lay piled up in the morgue, their limbs on ice. Many of the dead were from mofussil areas. Their families would hear about the tragedy only much later, particularly since some didn’t even know their relatives were on board. ^[7] Others, perhaps, never found out at all, given the underhand manner in which agents and touts recruited people for jobs abroad. ^[8]

After 15 days in the morgue, the unclaimed bodies were divided in proportion to the Hindus and Muslims on board according to the passenger manifest, a compromise between community leaders following an initial dispute. First, everyone went to the Muslim burial ground; later they went to the crematorium. Badhwar, still processing the disaster she had covered, attended the mass funerals and filmed them. “The smell of burning flesh still haunted me,” she said. “I felt I had some undone mourning to do.”

The media dredged up questions over Soviet accident records. Aeroflot, national carrier of the erstwhile Soviet Union, had been plagued with issues stemming from old planes, poor maintenance and corruption. ^[10] But the Kazakhs defended their dead countrymen and blamed Indian air traffic control. Saudi officials fumed about Indian media coverage of the crash. A human tragedy was quickly devolving into a blame game. 

Had either or both aircraft breached their allotted height? Had clouds hampered visibility? Had some as yet unknown mechanical failure occurred? RC Lahoti, a Delhi high court judge, was appointed to lead a court of inquiry to answer these questions. As the country where the crash took place, international accord deemed that India would lead the investigation. All the parties connected to the crash would get a hearing. A host of technical experts would weigh in.

Aviation professionals will tell you that plane crashes never happen for a single reason. There may be an identifiable  primary factor, but it’s usually a chain of events, an array of circumstances neatly piling up. What happened on 12 November 1996 was precipitated by one major factor—communication—but buttressed by stealthier causes. Unfortunately, all the holes lined up that evening.

An analysis of the wreckage pattern supported the suspicion that the collision had not been head-on. Rather, it appeared that the port wing ^[11] of the Kazakh plane had struck the port engines of the Saudi plane on first impact, and then sliced through its horizontal stabiliser, a component attached near the tail-end of an aircraft.

The devices found at the site offered little clarity at first. The Kazakh altimeters showed a 97m (319 feet) difference in readings. Since the Saudi altimeters could not be salvaged, it was hard to be precise about the altitude at which the collision had taken place.

But three months afterward, the centres at Moscow and Farnborough decoded both Flight Data Recorders. Both concluded that the collision had happened close to 14,000 feet, ^[12] a finding that all parties accepted before the court. This raised the next logical question: Why did the Kazakh aircraft leave its assigned height and descend to 14,000 feet?

The Kazakhs contended that their aircraft was not advised in time about the oncoming jet and sudden turbulence had forced the descent. ^[13] Even the flight data recorder seemed to suggest that the aircraft had quickly lost altitude, as much as 391 feet in half a second, or 47,000 feet per minute. At that rate, however, it should have disintegrated. Besides, there was no inkling of turbulence in the cockpit chatter.

Investigators eventually calculated that the descent had, in fact, been gradual. The solution to the puzzle, they believed, was that the sensor sending messages to the FDR was “sticky”—instead of constantly updating the flight parameters, the machine stuck periodically. The sensor’s moodiness yielded dramatic rather than graded changes in readings.

So that established that the descent was controlled. Why, then, had the Kazakh plane descended despite being told not to? Here is how the court’s accident report pieced together what happened.

At 6.36pm, SV 763 reports on the same radio channel: “Approaching FL 100.” 

Dutta allows them to climb to 14,000. 

At 6.38pm, SV 763 touches 14,000 and seeks permission to climb higher. 

Dutta tells them to stay put. 

The cockpit acknowledges this. This would have been audible to the Kazakh crew through the radio, and should have alerted them to another aircraft in the vicinity. 

Nine seconds later, still at 6.38pm, Repp tells Dutta they have reached 15,000 feet. But as the FDR later shows, Repp has got it wrong. At that moment, the Kazakh aircraft is at 16,348 feet. Repp is mistaken, perhaps because he doesn’t have his own instruments and hasn’t got a clear view of the panel. He may be misinterpreting the difference between “reached” and “approaching”.

Now, Dutta tells him to stay at 15,000 feet. There is traffic up ahead that will likely cross them in about five miles. Repp takes a while to absorb this and then acknowledges there will be oncoming aircraft: “Now looking.”

During this exchange, the pilot Cherepanov and co-pilot are chatting to each other about traffic. Are they aware they are supposed to be at 15,000? Here the report assumes the co-pilot continues to descend below that, despite being explicitly told there is a flight at 14,000 feet. Has he misunderstood that he has been cleared to 14,000? ^[16]

Cherepanov says nothing to his colleague. Aripbaev is also quiet.

A voice—most likely Repp—then calls: “Hold the level.”

6.39pm. Eighteen seconds before collision. Cherepanov belatedly asks: “What level we were (sic) given?” He doesn’t seem to have heard the ATC earlier.

The flight engineer shouts: “Maintain.” Confusion mounts. Maintain what? 14,000 feet?

Repp chimes in quickly, “Keep the 150th [15,000], don't descend!”

Has Repp realised they are in the wrong space? Suddenly, there is tumult. It’s 6.40pm. Eleven seconds from collision. The auto-pilot is switched off. Cherepanov asks the engineer to accelerate.

Four seconds to go. Repp, having probably seen the altimeter, yells desperately: “Get to 150 [15,000] because on the 140th [14,000] uh that one uh!”

Silence. The tape ends. The report presumes the Kazakh plane hit the Saudi plane while trying to climb back to 15,000 feet.

The report concluded it was a failure of the Kazakh crew. They had not only misunderstood the instructions, but also failed to coordinate and cross-check with each other. Cherepanov’s inept “crew resource management”, and “poor airmanship” came in for specific censure. Finally, the last hope—of the aircraft seeing each other—failed because clouds hindered clear visibility. 

The end, when it came, was swift. But in those final moments, both cockpits realised what was happening. ^[17] The Kazakhs were frantically trying to salvage the situation. In the Boeing, the Islamic prayer before death was chanted: “Astaghfir Allah, Asyhadu Unlaelaha Illallah.” God grant forgiveness, I witness no other God. 

As air traffic grew, ground facilities were unable to keep up. ATC teams were overworked and understaffed. In 1993, a US report advised flyers to avoid India and Colombia.

The accident and then Justice Lahoti’s report prompted several changes and sped up reforms. “It was done very fast,” said Captain Lumba, who flew through the 1990s. “It was a major cost to the country, but it improved safety tremendously. There was no dragging of feet.” Though the judge had been tasked with investigating this particular accident, the 15 recommendations looked at multiple aspects. ^[18]

Before Charkhi Dadri, controllers had to manage several movements per hour, often within the same air corridor, a rich arena for possible errors. One controller described the early 1990s as a “primitive” time. If a controller told a pilot to descend to 15,000 feet, were they actually doing so? With only primary radar, there was no way for the ATC to cross-check. They had to rely on the cockpit’s account, as Dutta had done that November evening.

Had Dutta had a secondary surveillance radar, which displays altitude, height and speed, he could have cross-checked the flight positions. The Lahoti report stressed that while secondary radar can’t “guarantee total safety by itself”, it would be useful at airports. In the months following the accident, media reports claimed that the secondary radar had been lying in storage at Delhi airport all along. That became a source of conjecture. Theories of theft, inadequate space and non-functional equipment were bandied about. The crash brought fresh attention to the radar issue, though according to news reports, it wasn’t until two years later that the technology became fully operational. ^[19]

The single air corridor had also played its role in the chain of events. Within nine months of the collision, incoming and outgoing flights from Delhi were allotted dedicated paths by carving out some of the airspace previously controlled by the Air Force. ^[20]

“If your report is spattered with blood, implementation becomes easy,” said AK Chopra, formerly of the DGCA. “We had been fighting for a secondary radar and separate corridors at the Delhi airport for months. We got them immediately after the accident.”

The other major change was that India made it mandatory for all airplanes flying in its airspace to be fitted with the Traffic Collision Avoidance System (TCAS). It became one of the first few countries in the world to impose this requirement. Five months after Charkhi Dadri, civil aviation minister CM Ibrahim announced in the Lok Sabha that TCAS would have to be installed in all 30-seater-plus flights after 31 December 1998. TCAS averts possible collisions by warning pilots of other aircraft intruding into their space. If the distance between flights falls below the mandated 1,000-foot vertical separation, the system triggers an alarm and advises evasive action.

“At the time, some airlines put up a strong opposition to mandatory TCAS because it was very costly,” said Chopra. “But we insisted on it, and it has paid off.” Though “near-misses” continue to occur (2018 saw as many as 64), the technology has more or less prevented catastrophic accidents.

Which brings us to the question: Could such a collision happen today? “It’s very unlikely,” said VP Agrawal, former Airports Authority of India chief. “The equipment, the automation gives you alerts ahead of time. The only problem is if there is a blackout of the automation system, if a computer freezes or there is a fire in the control tower. It is rare, but it can happen.” All things considered, Agarwal said, air travel remains the safest mode of transport. According to one estimate, for every 1 billion passenger miles travelled by car, 7.2 people die; by plane, it's 0.07 people.

India acted decisively to secure its skies but several other countries had been touched by the accident. The flights had Saudi and Kazakh crew; passengers were from India, Nepal, Kyrgyzstan, the United Kingdom; the aircraft had been manufactured in the USA and Russia.

“Corrective measures in India and around the world were more dramatic with this accident than many others,” said KPS Nair, formerly of the DGCA. “It was seen as a major accident not only because of the number of deaths but because mid-air collisions were so rare.” ^[21]

Miscommunication and language problems had plagued cockpits for decades, and routinely popped up in accident reports. ^[22] In 1993, Chinese pilots crashed into the ground partly because they could not comprehend why the ground warning system was telling them to “pull up”. In 1997, a Korean co-pilot, wary of upsetting the senior captain, allegedly used indirect language instead of spelling out the situation. The plane collided into a hill and 223 people died. ^[23]

In September-October 1998, India joined more than a 100 countries at the ICAO general assembly in Montreal with a proposal. It was simple, focused not on machines and technical specifications but the most primitive component of aviation operations: speech.

India claimed that ATC instructions were often disregarded or misinterpreted, resulting in situations where flight levels were mixed up, wrong run-ways were approached and wrong radials steered. The assembly accepted India’s working paper proposal to include language proficiency standards as part of licensing, and set in motion a long process to lay down the specifics.

A study group formed in 2000 spent nearly three years looking at the issue before releasing a document with a new six-level gradation of spoken English for crews and controllers. Personnel would have to achieve level four (out of six) levels to undertake international operations. Though no standardised test was designed, it was up to the countries to ensure proficiency as part of licensing requirements. They had until 2011 to comply.

Would the requirements have come in if the crash hadn’t happened? “No. It was India that took the initiative to bring it to the ICAO assembly,” said Captain Enrique Valdes on a Zoom call from the US. Valdes, a former American pilot and air crash investigator, was a member of the language proficiency study group. “India was the one that put it on the floor and was able to get the resolution to pass.” 

But language is never simply a question of speech and vocabulary. It is closely bound up with politics and power. Should English be the only language for aviation? This was discussed too, and eventually, the ICAO decided the English language proficiency requirements would apply only to international operations. ^[24]

In practice, it means a controller should speak English to an international aircraft but may switch to a local language with a domestic one. Since crews on the same radio frequency can hear conversations between the ATC and other aircraft in the vicinity, if the language isn’t English, it may not be understood by everyone. “Any time another language is used, pilots lose situational awareness (a sense of their own and other pilots’ positions ^[25] ) unless they speak that language. And pilots need situational awareness,” said Valdes.

The International Federation of Airline Pilots and the International Air Transport Association filed “a letter of difference” with ICAO, saying only English should be used for communications along international routes.

In 2019, scheduled services enabled 4.5 billion passenger journeys around the world—millions of individuals, steered safely by the magic of flight and the mundanity of clipped human chatter. But testing and licensing procedures can still be imperfect, and language always is. In 2017, the UK Civil Aviation Authority found that 267 Mandatory Occurrence Reports over 18 months were related to miscommunication. Problems included “the UK pilot misunderstanding the non-UK ATC (30%); and pilots (origins unclear) misunderstanding the UK ATC (18%)”.

More flights means more ground staff, more multicultural cockpits and more opportunities for misunderstandings. “To be honest, even now, in 2020, not everyone has reached full compliance with ICAO requirements,” said Roberts. “There is a gap not just in testing but also training. Of course, there have been a lot of changes since Charkhi Dadri, but it’s been very slow. I still don't think we are quite there yet.”

When 349 people die in such a gruesome manner right outside the national capital, how does it ebb so quickly from public memory? There was extensive news coverage of Charkhi Dadri in the immediate aftermath both in India and abroad, but it is barely remembered today outside aviation circles. There are no books about it. There is little institutional recognition of India’s role in the events that followed. Most people I spoke to below the age of 30 had no idea about the crash. 

Natasha Badhwar also covered the Uphaar Cinema fire in 1997 and the Kandahar hijacking in 1999. The coverage and national response to those tragedies had a different texture. Perhaps this was because they centred on identifiable people. Badhwar said she can recall specific families, stories and photographs. But with Charkhi Dadri, it was different. There were unclaimed bodies, a third of the passengers were foreigners, and a majority of the Indians on board came from rural areas. The families “didn't have albums to show you, for instance,” Badhwar said. “The kind of evidence we engage with as journalists, that was not there.”

Many passengers were also men and women headed to the Gulf to work in blue-collar jobs. Badhwar said, in her own imagination, working class people on flights “was an incongruity” at the time. “The human deaths in that plane crash did not evoke the same kind of horror or loss among the public,” she said. “No one important had died. They had not been people like us.”