sichuan 2008 earthquake case study

Benchmarks: May 12, 2008: Earthquake devastates western China

by Sam Lemonick Monday, August 31, 2015

Damage by earthquake-induced landslides in May 2008 was catastrophic and accounted for many casualties. Deadly debris flows hit the town of Qushan, which was destroyed by strong shaking and landslides. Credit: USGS, David Wald.

Just before 2:30 p.m. local time on May 12, 2008, a magnitude-7.9 earthquake shook Sichuan province in Western China. One eyewitness recalls seeing a mountain “blowing up” and boulders two-stories tall crashing into gorges. Another recalls thinking there had been a natural gas explosion, while a third described a hill split in half. It was the country’s largest earthquake in more than 50 years, and it left 18,000 people missing and presumed dead, nearly 375,000 injured and more than 69,000 confirmed fatalities.

Five years later, the wounds have begun to heal: Buildings have been rebuilt and lives are returning to normal. Despite that, deep grief and resentment remain about what locals see as the government’s failure to adequately protect people — especially children — and opinions differ on the success of rebuilding efforts.

The Great Sichuan Earthquake

A woman is rescued from the rubble following the Sichuan earthquake, which struck on May 12, 2008. Credit: courtesy of Miniwiki.org, Creative Commons Attribution-ShareAlike 3.0 Unported.

The Sichuan, or Wenchuan, quake struck on the eastern edge of the Tibetan Plateau, in an area with some of the most extreme topography on the planet. The rupture occurred on the Longmenshan Fault, a thrust fault associated with the collision between the Indo-Australian and Eurasian plates; the collision is uplifting Western China at a rate of roughly 5 millimeters a year. Walking across the fault, one would climb nearly 6,000 meters in 50 kilometers from the low-lying Sichuan basin to the tops of the tallest peaks.

The U.S. Geological Survey (USGS) located the epicenter of the quake about 80 kilometers west-northwest of Sichuan’s capital, Chengdu, and about 1,500 kilometers southwest of Beijing. From its hypocenter 19 kilometers below the surface, slip during the earthquake propagated northeast along the fault. According to both seismic data and survivor accounts, the earthquake lasted about two minutes. When the shaking stopped, the two sides of the fault had been offset by nine meters.

Reports of tremors came from thousands of kilometers away in Russia, Taiwan and Thailand, where office buildings reportedly swayed for several minutes. Workers evacuated office buildings in Beijing and in Shanghai, more than 1,600 kilometers away, about the distance from Boston to St. Louis.

Despite its proximity to the epicenter, downtown Chengdu — a city of 6.7 million people — suffered only minor damage. Reporters touring the city after the disaster saw cracks in the walls of buildings, although none had collapsed. In fact, all of the major cities in Sichuan suffered only moderate damage. Meanwhile, some towns in the surrounding countryside were destroyed.

Local and international engineers ascribe this dichotomy in part to China’s adoption of new building codes after a 1976 earthquake in northeast China that killed several hundred thousand people. New construction in China’s fast-growing cities was held to the stricter standards, while older buildings were often not upgraded.

Structures built with brick, mud and other materials before the codes were in place failed, often with tragic results. Among the most visible of these were the many schools that collapsed entirely. News reports worldwide featured heartbreaking pictures of parents who lost their children in devastated school buildings. A year after the quake, the education minister of Sichuan province reported that 5,335 schoolchildren had died in the earthquake.

Aftershocks and Landslides

This elevated highway was taken down by the Chinese Army because it was damaged beyond repair by the magnitude-7.9 quake. The road was the main route into the Beichuan area. Credit: USGS, Lynn Highland.

Numerous aftershocks — as many as 100 within the first three days — threatened survivors and delayed rescue efforts in the weeks following the main shock. As late as August, three months after the main shock, the area was hit by a magnitude-6 aftershock.

In addition to the danger of the aftershocks, numerous landslides in the mountainous region blocked rivers, forming 34 so-called quake lakes. The government evacuated hundreds of thousands of people downstream from some of these lakes in fear that the natural dams would burst, and the army airlifted in construction equipment to build sluices to release the water. Some feared that man-made dams might also give way. Despite damage to many, none failed.

Landslides that occurred both during and after the quake severely hampered the relief effort. Bridges and many vital roads were destroyed or covered in rubble. The Chinese military deployed tens of thousands of soldiers within a day of the quake, but they could not reach all of the remote areas most in need of help. One village reportedly spent five days without food and water until a rescue team could reach it.

Relief teams and supplies were flown in from neighboring countries and as far away as the U.S., while money came from around the world. Survivors in Chengdu also made extensive use of the Internet to get information and help coordinate relief and rescue operations.

The Agriculture Ministry reported that among the nonhuman victims of the earthquake were as many as 12.5 million farm animals, mostly chickens. Sichuan is also home to many of the world’s giant pandas, as well as a number of preserves and breeding centers. One panda in captivity died when the wall of its enclosure collapsed after a landslide at the Wolong Nature Reserve. Mud and debris flows also destroyed large swaths of bamboo forest, critical habitat for the wild panda population.

In the Aftermath, Seeking a Cause

An elderly woman in an emergency shelter after the quake. Credit: USGS.

As was apparent in Italy following the 2009 L’Aquila earthquake, many people expect geologists to be able to predict earthquakes, and the same was true in the aftermath of Sichuan. Many blamed the government for failing to take earthquake predictions seriously, pointing to a 2002 statistical analysis by Chinese seismologist Chen Xuezhong that suggested a strong possibility in the coming years of an earthquake in Sichuan greater than magnitude 7.

But the most contentious issue following the disaster was the charge that by overlooking corruption in the construction of schools, the government had failed to keep Sichuan’s children safe. Ultimately, the government’s tally of collapsed classrooms came to 6,898. Observers were quick to note that many schools collapsed while other nearby buildings suffered little damage.

The updated building codes announced after the 1976 earthquake held schools to an even stricter standard than residential buildings, and the government set out to renovate or rebuild schools that were out of code. After the 2008 quake, parents, journalists, activists and others accused school builders of ignoring the 1976 codes. Chinese critics use the colloquialism “tofu-dreg projects” to describe the shoddy construction of many schools.

An investigation by the Chinese magazine Caijing into why five schools in Sichuan had collapsed found multiple structural failures. Concrete walls, ceilings and floors in the schools had not been reinforced with steel rods called rebar, which gives concrete strength and flexibility, the investigation found. During the earthquake, the concrete simply cracked and crumbled without this internal steel structure.

As one headmaster interviewed by the magazine pointed out, the new schools replaced mud buildings that had leaked in the rain. Prior to the quake, parents and students had been happy to even have a new school, regardless of whether it met earthquake building standards. Bureaucracy, the magazine added, bogged down building inspections.

The Chinese government offered monetary compensation to parents who lost children and lifted its one-child policy for those who had lost an only child. In late May 2008, the government also promised a thorough investigation and severe punishments for construction companies and officials involved in building substandard schools. Despite several further announcements that year indicating the inquiry was in progress, no final report has been announced, at least in Western media. The government has, however, released new, stronger building codes.

Efforts to Rebuild

A residential building in Hanwang that was destroyed by the earthquake. Credit: USGS, Sara C. Behan.

Five years after the quake, the Chinese government and mainstream Chinese media have proclaimed “victory” in the rebuilding effort. The government committed $146 billion to relief, rescue and rebuilding. The World Bank reports that the Chinese government commissioned more than 41,000 reconstruction projects and completed 99 percent of them within two years.

In addition, China drew praise for its willingness to accept help from the rest of the world. In particular, the World Bank praised the government’s ability to marshal and direct the combined efforts of agencies, private organizations and individuals. One innovation was to pair counties in unaffected parts of the country with Sichuan counties to help rebuild.

The Chinese government also reformed building codes once again. City layouts were reorganized and damaged hospitals and sanitation plants were modernized. Schools were rebuilt.

Others, however, say that some parts of Sichuan still look like they did five years ago and some people are still living in temporary shelters. One Chinese writer who visited the area near the epicenter last summer reported that many roads remained unrepaired. He wrote that a 26-kilometer car ride took more than three hours because of damaged roads, and suggested the roads are ignored by officials focused on more visible “vanity” projects.

It may ultimately take years before the region fully recovers from the disaster.

In the meantime, the quake continues to be studied by seismologists both in China and around the world. After the quake, the China Geological Survey began a joint project with the USGS to study how to better predict landslide risks. A 2009 study in Science suggested that a dam in Sichuan completed in 2006 might have contributed to the earthquake. The authors proposed that the weight of the impounded water — 320 million metric tons by their estimate — could have altered the pressure on the fault and triggered the earthquake.

If there is a silver lining to the tragedy, it’s that researchers have studied the response to the quake to find better ways to save lives in the future. A study published last year looked at the medical response to the quake, from the initial search for survivors to epidemic outbreak programs in areas that lacked basic sanitation and medical care long after the quake. Experts have pointed to China’s coordinated response — between military and civilian leadership, between national, provincial and local governments, and between affected and unaffected parts of the country — as one of the most successful and important elements in the quake’s aftermath.

© 2008-2021. All rights reserved. Any copying, redistribution or retransmission of any of the contents of this service without the expressed written permission of the American Geosciences Institute is expressly prohibited. Click here for all copyright requests.

• Share full article

Advertisement

Supported by

Powerful Quake Ravages China, Killing Thousands

By Jake Hooker and Jim Yardley

• May 13, 2008

CHENGDU, China — A powerful earthquake struck Western China on Monday, toppling thousands of homes, factories and offices, trapping students in schools, and killing at least 10,000 people, the country’s worst natural disaster in three decades.

The quake, which was estimated preliminarily to have had a magnitude of 7.9, ravaged a mountainous region outside Chengdu, capital of Sichuan Province, just after lunchtime Monday, destroying 80 percent of structures in some of the towns and small cities near its epicenter, Chinese officials said. Its tremors were felt as far away as Vietnam and set off another, smaller quake in the outskirts of Beijing, 900 miles away.

Landslides, power failures and fallen mobile phone towers left much of the affected area cut off from the outside world and limited information about the damage. But snapshots of concentrated devastation suggested that the death toll that could rise significantly as rescue workers reached the most heavily damaged areas.

In the town of Juyuan, south of the epicenter in the city of Wenchuan, a school collapsed, trapping 900 students in the rubble and setting off a frantic search for survivors that stretched through the night. Two chemical factories in Shifang were destroyed, spilling 80 tons of toxic liquid ammonia, officials told Chinese state media.

The destruction of a single steam turbine factory in the city of Mianzhu buried “several thousand” people, the state-run Xinhua News Agency reported Tuesday morning.

The quake was already China’s biggest natural disaster since another earthquake leveled the city of Tangshan in eastern China in 1976, leaving 240,000 people dead and posing a severe challenge to the ruling Communist Party, which initially tried to cover up the catastrophe.

This time, officials quickly mobilized 50,000 soldiers to help with rescue efforts, state media said. Prime Minister Wen Jiabao flew to the scene and was shown coordinating disaster response teams from the cabin of his jet.

The prime minister was later shown on national television standing outside the damaged edifice of the Traditional Medicine Hospital in the city of Dujiangyan, shouting encouragement at people trapped in its ruins.

“Hang on a bit longer,” he said. “The troops are rescuing you. As long as there is the slightest hope, we will never relax our efforts.”

The quake was the latest in a series of events that have disrupted China’s planning for the Olympic Games in August, including widespread unrest among the country’s ethnic Tibetan population, which lives in large numbers in the same part of Sichuan Province where the earthquake struck.

The powerful initial quake struck at 2:28 p.m. local time, or 2:28 a.m. Eastern time, near Wenchuan County, according to China’s State Seismological Bureau. Most of the heavy damage appeared to be concentrated in nearby towns, which by Chinese standards are not heavily populated. Chengdu, the largest city in the area, with a population of about 10 million, is about 60 miles away and did not appear to have suffered major damage or heavy casualties.

But officials had yet to describe the impact in Wenchuan itself, which has a population of 112,000 and is home to the Wolong Nature Reserve, the largest panda reserve in China. The town of Beichuan, on the way from Chengdu to Wenchuan, suffered several thousand deaths, state media said.

China’s massive Three Gorges Dam, a few hundred miles east of the earthquake’s epicenter, reported no immediate problems.

At dawn on Tuesday morning in Chengdu, clusters of people were huddled outside, many saying they were too fearful of aftershocks to go indoors. Many wore plastic slickers to protect them from a steady drizzle. Wang Zihong, 35, a businessman from Gansu Province, had spent 12 hours outside his hotel, squatting with others on a street corner.

“It was a terrible shock,” he said. “I couldn’t stand up straight. We were on the second floor and we ran outside.”

Chengdu’s Huaxi Hospital, one of the largest in western China, started receiving patients from surrounding counties on Monday afternoon. By Tuesday morning, 180 patients had arrived from hard-hit surrounding counties.

“Seven thousand people have died in Beichuan, a single county, and we think Wenchuan will be similar, too, because it was the epicenter,” said Kang Zhilin, a spokesman for the hospital. He added: “The first patients who came had jumped from buildings because they were frightened.”

After the tremors shook Chengdu, roughly 4,000 frightened patients had been relocated from wards on the hospital’s upper floors to a courtyard outdoors. By Tuesday morning, the patients were sitting in the rain, covered in plastic.

A woman, Tang Hong, 50, sat beside her injured husband, Yan Chaozhong, in the hospital. They had arrived early in the morning from Dujiangyan County, one place that had suffered heavy damage. They had been inside their fourth-floor apartment when the quake hit. “It was violent,” she said. “Even when we crouched down, it flattened us.”

Ms. Tang said she and her husband had tried to escape down a stairwell, but a second tremor knocked her husband down the stairs, and he broke three ribs. She said four six-story buildings on her street had been flattened. She also wept as she described how a school for handicapped and deaf students collapsed while the children were in class. “It was horrible,” she said. “The entire school building collapsed.”

Minutes after the western temblor struck, a second, smaller quake struck Tongzhou, an outer district of Beijing. Thousands of office workers were evacuated in the capital city, but no damage was reported there.

“I suddenly felt very dizzy, as if I were heavily drunk,” said Zeng Hui, who works on the 22nd floor of an office tower in Beijing. “I thought I was seriously ill, then I looked around and saw my colleagues felt the same way.”

There were reports of fatalities in Chongqing Municipality, near Sichuan, where two primary schools were damaged. Four pupils died and more than 100 others were injured, state media reported. Xinhua devoted extensive coverage to the disaster, publishing regular updates on the situation, including latest death tolls, on its Chinese and English Web sites. The relatively vigorous flow of information and the fast response from top officials and rescue workers stood in stark contrast to the way China handled the Tangshan earthquake, or the way the military junta that rules neighboring Myanmar has managed the aftermath of a giant cyclone that killed nearly 32,000 people there this month, according to Burmese government estimates.

Efforts to reach people near the epicenter of the bigger quake in western China were hindered by damage to the telephone system. Some 2,300 towers used to transmit phone signals had fallen, the country’s main mobile phone company reported. The earthquake also disrupted air traffic control in western China, interfering with flights between Asia and Europe on Monday afternoon, although flight services were restored by the evening.

Cathay Pacific Airways announced that it had canceled two flights between Hong Kong and London — one in each direction — and had delayed the departure of a Monday afternoon flight from Hong Kong to London by 19 hours, to Tuesday morning.

While China Mobile acknowledged extensive damage to its cellphone towers, it is less clear how much damage occurred to the separate communications network that China’s authorities maintain for natural disasters and other contingencies.

Communications equipment vendors attending a police equipment exhibition in Beijing last month said that China maintained a separate network using different frequencies and other equipment from the main cellphone network. The separate network allows the police and other agencies to respond to emergencies even when the main landline and cellphone networks are overwhelmed with calls by residents.

Many Western countries also maintain separate communications systems for emergencies. China is still upgrading its emergency network by buying equipment from Motorola and other foreign companies, communications industry officials said at the exhibition.

Temporary disruption of the air traffic control system in western China strongly suggested that the authorities’ communications gear might also have been damaged at least temporarily. China has worked closely with the Federal Aviation Administration in the United States to improve air safety, and air traffic control operations in the United States have backup communications systems to avoid disruptions.

Jake Hooker reported from Chengdu, and Jim Yardley from Beijing. Ed Wong contributed reporting from Chengdu, and Keith Bradsher from Hong Kong. Zhang Jing contributed research..

Jake Hooker reported from Chengdu, and Jim Yardley from Beijing. Ed Wong contributed reporting from Chengdu, and Keith Bradsher from Hong Kong. Zhang Jing contributed research.

Become a member and discover where geography can take you.

• The geography ...
• Sichuan Earthq...

Sichuan Earthquake (2008): lessons learnt?

Key questions.

How do earthquakes happen?

What can be done to reduce the impact of earthquakes on people and places?

Earthquakes affect a huge number of people

The earth's crust is made up of huge tectonic plates which move slowly on the molten mantle beneath the earth's crust. The plates meet on conservative (plates slide past each other), destructive (plates collide into each other) and constructive (plates move away from each other) plate boundaries. Earthquakes occur along these plate boundaries or fault lines.

The impact of the 2008 Sichuan earthquake

The Sichuan earthquake occurred along a destructive plate margin, where the Indo-Australian Plate and the Eurasian Plate collide. The earthquake occurred along a mid-fracture (known as the Yingxiu-Beichuan fracture). The earthquake in Sichuan was a magnitude 7.9 on the Richter Scale, occurred at a depth of 19 meters and its epicentre was in Wenchuan, north-east in the Sichuan Province.

Dams - it is estimated that approximately 400 dams have been affected by the Sichuan earthquake (2008). The safety of these dams is a concern as many may not have been designed to withstand earthquakes. Some Chinese and US scientists believe that the quake could have been triggered by the huge Zipingpu dam. The dam is 50 stories tall and holds back several hundred million tons of water in the Zipingpu reservoir. It is situated just 500 meters from the Beichuan fault, only 5.5 kilometres from the epicentre of the earthquake. The Beichuan fault would have been under natural stress but this would have been multiplied by 25 times due to the weight of the water in the reservoir.

The Sichuan earthquake killed approximately 70,000 people, although other estimates put the total far higher. More than 10,000 children were killed when their schools collapsed or were buried beneath landslides. At least 4,727 children were orphaned as a result of the quake. The quake also left five million people homeless and 375,000 people injured. The impacts of this quake will be felt for generations to come especially as families tend to have only one child due to China's strict ‘One Child Policy'. Other impacts of the quake include: the collapse of buildings, particularly in the Beichuan County, the collapse of two chemical plants which led to the leakage of 80 tons of liquid ammonia which buried many people. Schools, homes and buildings were destroyed.

Reducing the impact of Earthquakes on people and places

The Sichuan earthquake killed many people due to unsafe building construction, particularly that of schools which were referred to by local people as ‘tofu dregs'. There is a great deal in modern construction techniques which can be done to reduce the impact of earthquakes on people and reduce the death toll. Within the USA, there are strict building codes for buildings within earthquake zones. These can help ensure minimum standards of building. In addition, instruments are installed in public buildings to measure the response of buildings to earthquakes. The response of buildings to earthquakes can be measured and then alterations to their construction can be made. It is also considered that metal buildings are far better at withstanding an earthquake than concrete ones because metal ones are more ductile (they can bend and flex without breaking). It is also important to consider the distribution of weight. A building which is top heavy is much more likely to fall than a building which is light weight on the top. Therefore, an earthquake building should be constructed of steel rebar but should be framed with lighter materials in the upper floors and have a heavily reinforced lower section. However, just because it is made of metal, does not mean it will not fall down during an earthquake.

In the USA, residential buildings are designed so that the roof falls directly in the middle of the room but stays up near the wall. People are then advised to take shelter in the doorways and away from the middle of the room. Scientists have considered the side to side movements during an earthquake in the design of earthquake proof buildings. However, they are now considering the up and down (vertical) movements also. To ensure that buildings are able to move in all directions, scientists suggest isolating the building from the ground in order to reduce the vibrations from an earthquake. Instead of having the foundation rest directly on soil or rock, use a material that will provide a cushioning effect and reduce the energy transfer from the ground to the building. Also, using liquid dampers. These are like putting a water tank at the top of the building. These are particularly useful where there are high winds.

Use the BBC's interactive earthquakes resource (online or print out) to explore how earthquakes happen. Note down key words from this resource and write a short paragraph summarising how earthquakes happen.

Main Activity

Comparing two earthquakes in China: Tangshan (1976) and Sichuan (2008)

Watch and read the PowerPoint China Earthquakes : Learning from the past and read the Sichuan quake factfile, referring particularly to the section "Poorly constructed schools".

Complete the card sort activity entitled Learning from the past: Tangshan and Sichuan compared. By doing this you start to think about how the two earthquake case studies are similar/different to each other.

Then produce a Venn-diagram on a large piece of paper to compare the two earthquakes. How were the earthquakes similar? How were they different? Once you have completed your diagram discuss and share your ideas with the class.

For the second part of this activity use the resource sheet Learning from the past: Changing the future. Write an email response to the Chinese Government with recommendations of how they could prepare better for another earthquake. Use this ScienceDaily link and search ‘Earthquake resistant building' in Google images to help.

Feed back your recommendations on how the Chinese government could reduce the impact of earthquakes in China to the class.

Extension activity

Design your own earthquake resistant building. Here are some key points to think about in planning its construction: How will you stop the building from falling down? If the building does not fall down, how will you prevent accidents both inside and outside as it moves? What about falling glass and rubble on the people outside?

File name Files

The Geography of Science Lesson 3 Teacher's Notes

The Geography of Science Lesson 3 Teacher's Notes (1)

The Geography of Science Lesson 3 China Earthquakes: Learning From The Past

The Geography of Science Lesson 3 Learning From The Past: Tangshan and Sichuan Compared

The Geography of Science Lesson 3 Learning From The Past: Tangshan and Sichuan Compared (1)

The Geography of Science Lesson 3 Learning From The Past: Changing The Future

The Geography of Science Lesson 3 Learning From The Past: Changing The Future (1)

The Geography of Science Lesson 3 Sichuan Earthquake Factfile

The Geography of Science Lesson 3 Sichuan Earthquake Factfile (1)

Back To Top

We use cookies. Read more about them in our Privacy Policy.

• Accept site cookies
• Reject site cookies
• Search Search
• Login Login
• Publications

Lessons from the Sichuan earthquake

• Issue 43 The role of affected states in disaster response
• 1 Aid and access in Sri Lanka
• 2 When the affected state causes the crisis: the case of Zimbabwe
• 3 Humanitarian governance in Ethiopia
• 4 The silver lining of the tsunami?: disaster management in Indonesia
• 5 Land and displacement in Timor-Leste
• 6 Lessons from the Sichuan earthquake
• 7 Britain and Afghanistan: policy and expectations
• 8 Are humanitarians fuelling conflicts? Evidence from eastern Chad and Darfur
• 9 Lessons from campaigning on Darfur
• 10 Supporting the capacity of beneficiaries, local staff and partners to face violence alone
• 11 Stuck in the 'recovery gap': the role of humanitarian aid in the Central African Republic
• 12 Out of site, out of mind? Reflections on responding to displacement in DRC
• 13 Making cash work: a case study from Kenya
• Twitter Twitter
• Facebook Facebook
• Share on Email Email

A t 2:28pm on 12 May 2008, a powerful earthquake struck China’s Sichuan Province. Some 87,500 people were killed, 45.5m affected and 14.4m displaced . Economic losses were estimated at $86 billion, with 21m buildings damaged. According to a recent DFID report, the earthquake drove an estimated 10m people below the poverty line, with overall poverty in badly affected areas increasing from 11% to 35% of the population . Despite the extent of the devastation, this earthquake was not China’s first experience with natural disaster. In fact, four of the ten most destructive earthquakes on record have occurred in the country, giving China extensive experience in coping with such emergencies – experience that was put to good use in the response to the Sichuan earthquake.

The role of the state in the earthquake response

The response to the earthquake was dominated by the Chinese government. Although the government invited international humanitarian assistance, few international NGOs engaged directly in emergency response, for a number of reasons. First, Sichuan was probably not a priority for organisations already involved in responding to concurrent disasters including Cyclone Nargis, which made landfall in neighbouring Myanmar just ten days before the earthquake. Second, lack of access and local experience may have prevented some INGOs from initiating operations in Sichuan. Third, in the context of an economic boom with over 10% annual growth since 2002, it is possible that international actors believed that the Chinese government had the capacity to respond. Although full recovery remains a distant reality for many, the Chinese state-led response to the Sichuan earthquake has generally been characterised as efficient and comprehensive. According to the government, despite the extent of the devastation, disease outbreaks were avoided, populations in danger from subsequent flooding or landslides were safely relocated, medical services were generally restored rapidly and a return to the baseline mortality rate was achieved relatively quickly. The direct provision of aid by the Chinese military was a key element in the emergency response phase. Officials reported that, within 14 minutes of the earthquake, the central government had dispatched the People’s Liberation Army (PLA) to the affected areas, and within days 113,000 soldiers and armed police had been mobilised. Of the nine government working groups set up for the relief effort, six were supported by the military.

One challenge to learning from the earthquake response is that data has not been made widely available by the government. For example, in the health sector Xinhua News reported that, as of late May, 45,000 medical workers were contributing to care following the earthquake, with 650 devoted to epidemic control . A Health Ministry representative also announced that the relief effort had eliminated the risk of a disease outbreak , and had even brought about a decline in infectious disease incidence in the worst-hit areas, compared to previous years. Unfortunately, evidence is not offered to support this statement, or similar statements in other sectors. Information from the few active organisations (including Médecins Sans Frontières, UNCEF, AmeriCares Foundation and Oxfam-Hong Kong) only capture the relatively small-scale activities of these agencies.

While working for a US-based organisation in two of the worst-affected counties in Sichuan directly following the earthquake, over the course of three months I encountered no other international NGO working on the ground. This is consistent with the general pattern of minimal INGO engagement in the emergency response. To accomplish our mission to re-establish referral care and provide urgently needed medical supplies, I coordinated all efforts in direct partnership with government agencies and the provincial and county-level Health Bureaus. In my day-to-day work, it became clear that the overall success of the government’s response was made possible by its authoritarian position, its experience of managing large population movements and natural disasters and the rapid deployment of the military. These three elements enabled the government to avoid or minimise many of the problems common to disaster response.

Mitigation strategies included an immediate emphasis on controlling infectious disease through widespread medical care and surveillance, the provision of tents for shelter (albeit insufficient in number at the outset and eventually upgraded to temporary, prefabricated structures), maintenance of security and the rule of law through substantial police and military deployments, traffic and supply-chain management at the regional and local level, as well as the triage of patients, the deployment of qualified volunteers and the efficient management of in-kind donations. For instance, as large quantities of unsolicited foreign medicines and supplies accumulated in airport warehouses (donated primarily by organisations without a physical presence in Sichuan), the provincial health bureau coordinated with the government body in charge of volunteers to assign pharmacology students on holiday to sort, translate and test these donations. Additionally, the movement of people was strictly regulated in the affected areas. For months, police and military roadblocks prevented non-essential personnel from entering the disaster zone (personnel also sprayed traffic passing through with disinfectant in the belief that this would reduce the risk of disease). These authoritarian measures largely succeeded in saving lives and reducing the secondary disasters of disease, flooding and damage from strong aftershocks; however, these results came at the expense of personal liberties, access to affected areas and, in some cases, the unquestioned acceptance of sub-standard living conditions.

Although the Chinese and foreign press have reported outrage among some parents who lost children in collapsed schools, the overall reaction of direct beneficiaries as regards the government aid they received was outwardly positive. Affected populations worked to reconstruct markets and establish a home in their government-issued tents, while awaiting further instructions from the local authorities. This differs from my experience in the North-West Frontier Province (NWFP) in Pakistan following the 2005 earthquake, where there was an elevated sense of anxiety, especially in remote rural areas. For example, in the Allai Valley of NWFP (population around 100,000), insufficient assistance saw virtually every family electing to migrate to camps at a lower elevation or moving to live with relatives elsewhere. Although some seasonal migration takes place annually in the region, post-earthquake migration occurred on a large scale, and was even encouraged by some NGOs. By contrast, the millions of people affected by the Sichuan earthquake, even those living in mountainous rural areas, stayed in close proximity to their destroyed homes.

In the days and months following the earthquake, many familiar disaster response tools and mechanisms were not utilised: there were no cluster meetings, and the Sphere Standards and other guidelines common in the humanitarian community were not in evidence. Instead, a coordinated response was achieved through the government’s hierarchical approach, and decisions followed the chain of command from national to provincial and down to the prefecture and county levels. In terms of coordination, after working side-by-side with my health bureau counterparts daily for nearly three months, I did not observe a single complaint about unwarranted time spent meeting donors or international aid groups (though there were complaints about the unaccompanied relief material pouring into the provincial airport and bonded warehouses). Unlike the direction eventually chosen by the government of Pakistan following the 2005 earthquake, the Chinese authorities did not immediately establish a parallel relief agency. Instead, relief activities were partitioned along the lines of the cluster approach , with the formation of working groups roughly corresponding with government agencies – an important approach for ongoing coherence in policy and practice.

Another partnership strategy used in the aftermath of the earthquake which may prove a model for long-term recovery was the ‘twinning’ of several badly affected counties and cities with other Chinese provinces and municipalities. These partnerships aimed to assist affected areas with resources, personnel and moral support for recovery. Teams of doctors, public health professionals and sanitation and disease control experts were immediately dispatched to the affected partner county; a reported 1–3% of the annual gross domestic product of sponsor provinces was pledged towards long-term recovery efforts in the affected county for at least three years. For example, Wenchuan County, the epicentre of the earthquake, was paired with wealthy Guangdong Province for long-term reconstruction assistance, including the provision of medical personnel to replace staff lost in the earthquake, and the training of Wenchuan-based staff in teaching hospitals in Guangdong.

The state-led response focused on efficiency in providing resources and services to the largest number of people possible. However, this came at a price; for instance, in order to get food to everyone who needed it nutritionally deficient instant noodles were provided for days on end in some locations. Shelter could not be manufactured quickly enough (despite temporary state seizure of suitable textile factories), resulting in up to 12 individuals sharing one family-size tent. The absence of water-borne diseases may actually be attributed to a culture of boiling water, rather than the government’s pervasive disinfection campaign. It is clear that action to protect against a secondary disaster did not come from abroad but from within China. Although the state deserves praise for its handling of the response, there are areas for improvement.

Lessons for the future

In the aftermath of every major recent natural disaster, from the Indian Ocean tsunami to the Pakistan earthquake and even the cyclone in Myanmar, a deluge of assistance from international non-governmental organisations has had a significant impact. This was not the case in China, where very little international assistance was provided and the response was very largely state-led – a vast relief effort launched by the Chinese government and carried out by hundreds of thousands of soldiers, civil servants and civilian volunteers. The government’s approach to the emergency response was effective in several respects; the setting of clear criteria and appropriate restrictions on unsolicited in-kind medical or other supplies, for instance, led to the more efficient use of resources and eased the supply-chain bottlenecks common in other disasters of similar magnitude, and overall the response was crucial in saving many lives. At the same time, however, greater efforts could have been made to enlist the support of specialised international agencies in specific areas, including emergency shelter, livelihoods and health. In the health sector, for instance, very little attention was paid to psychosocial and mental health programmes, especially among elderly people, who may well have benefited from specialised support from the humanitarian community. Finally, although the state deserves praise for its handling of the response, a lack of transparency in terms of specific data and details of the response have concealed many of these successes, as well as obscuring areas for improvement.

Brian Hoyer is an independent consultant. His email address is [email protected] .

Comments are available for logged in members only.

Issue 43 Contents

• Open access
• Published: 02 January 2024

A systematic scheme to develop dynamic earthquake rupture scenarios: a case study on the Wenchuan–Maoxian Fault in the Longmen Shan, China, thrust belt

• Rongjiang Tang   ORCID: orcid.org/0000-0001-9559-2712 1 &
• Ryosuke Ando 2  

Earth, Planets and Space volume  76 , Article number:  2 ( 2024 ) Cite this article

679 Accesses

2 Altmetric

Metrics details

The 2008 Wenchuan Mw 7.9 mainshock caused catastrophic destruction to cities along the northwestern margin of the Sichuan Basin. This earthquake did not activate the Wenchuan–Maoxian Fault (WMF) on the hinterland side and the conjugate buried Lixian Fault (LXF), but they could experience large earthquakes in the future. We propose a systematic scheme to develop scenario earthquakes for active fault systems with insufficient constrain of 3D fault geometries. We first performed stress tensor inversion to constrain the regional stress field. Then, we developed a new method to constrain fault geometries by inverting long-term slip rates under the given regional stress and applied it to the WMF. We conducted a set of 3D dynamic earthquake rupture simulations on the WMF and LXF to assess the scenarios of earthquake rupture processes. Several fault nucleation points, friction coefficients, and initial stress states are assessed, the general rupture patterns for these earthquake scenarios are evaluated, and finally, we find the scenarios that could fall into three groups. Depending on initial conditions, the dynamic rupture may start in the LXF, leading to magnitude-7.0 earthquakes, or start in the WMF, then cascade through the LXF, leading to magnitude-7.5 earthquakes, or both start and arrest in the WMF, leading to around magnitude-6.5 or -7.0 earthquakes. We find that the rupture starting on the reverse oblique-slip jumps to the strike-slip fault, but the reverse process is impeded.

Graphical Abstract

Introduction

The 2008 Mw7.9 Wenchuan earthquake caused devastating destruction to cities and counties along the active faults of the Longmen Shan thrust belt. Three parallel NW-dipping fault zones from the hinterland to the foreland [the Wenchuan–Maoxian Fault (WMF), the Beichuan Fault (BCF), and the Pengguan Fault (PGF)] constitute the crustal structure of the Longmen Shan Fault Zone (LMSFZ) (Fig.  1 ). The mainshock and aftershocks only ruptured the BCF and PGF (Xu et al. 2009 ; Feng et al. 2017 ; Liu-Zeng et al. 2009 ). Several counties, including Yingxiu, Beichuan, and Nanba, suffered a disaster in this event. One reason is that the fault almost passes through the center of these counties (Fig.  1 ). Similarly, the WMF also passes through the urban areas of Wenchuan, Maoxian, and several towns. Nearly 200,000 inhabitants live in the area along this fault. Should an earthquake in the future occur on the WMF, it would cause a significant hazard to densely populated towns. Assessing earthquake rupture scenarios in the future is essential to increase society’s preparedness.

Map showing the setting of major faults of the center Longmen Shan fault zone. The red bold lines indicate the field-confirmed rupture traces in the Wenchuan earthquake (Xu et al. 2009 ; Liu-Zeng et al. 2009 ). The blue solid lines represent the WMF and LXF, both of which did not slip in the mainshock but are very active in aftershocks. The green solid lines represent three active faults parallel to the LXF. Accurate fault trace of the WMF is derived from field investigation of Xie et al ( 2011 ), and the location of the LXF trace is deduced from the distribution of pure strike-slip aftershocks (Li et al. 2019 ). Historical earthquake catalog denoted by filled circles refers to Wang et al ( 2021 ) and China Earthquake Networks Center ( http://data.earthquake.cn ). Green hollow circles denote the historic strong earthquakes (M > 5) before 2008 Wenchuan event. The purple and gray coupled arrows depict the maximum and minimum horizontal compressional stress fields (Luna and Hetland 2013 ). The third primary stress (not shown) is vertical. The purple dashed regions, respectively, show the positions of Xuelongbao (north) and Pengguan (south) Massif (Shen et al. 2019 ). The blue box shows the seismic gap near Dayi town. WC—Wenchuan. MX—Maoxian.WMF—Wenchuan Maoxian Fault. LXF—Lixian Fault. BCF—Beichuan Fault. PGF—Pengguan Fault. XYDF—Xiaoyudong Fault. WCEQ—Wenchuan Earthquake. LSEQ—Lushan Earthquake

Dynamic rupture simulations intrinsically simulate physically self-consistent features of earthquake behavior and reproduce retrospectively large earthquake events (Aochi and Fukuyama 2002 ; Ando et al. 2017 ; Ando and Kaneko 2018 ; Ulrich et al. 2019 ). Since the dynamic rupture simulation is a forward problem to constrain the model based on the pre-seismic observations, efforts are made to use it to explore future scenario earthquakes under pre-seismic estimation of stress conditions and fault geometries (Oglesby and Mai 2012 ; Zhang et al. 2017 ; Aochi and Ulrich 2015 ; Ramos et al. 2021 ; Harris et al. 2021 ; Yao and Yang 2022 ).

Three-dimensional fault geometry is a dominant factor in earthquake rupture dynamics. Irregular fault geometry, such as bends, branches, intersections, and step-overs, have been evidenced to cause rupture initiation and termination (Oglesby 2005 ; Elliott et al. 2015 ; Yang et al. 2013 ; Ando and Kaneko 2018 ). Several studies (Shi et al. 1998 ; Oglesby et al. 2000 ; Ma and Beroza 2008 ; Tang et al. 2021a ) have shown that the dip angle, which leads to the asymmetric geometry of dip-slip or oblique faults intersecting the free surface, can affect the dynamics of earthquake rupture and ground motions considerably. In earthquake simulations, fault geometry is typically obtained from detailed relocation of aftershocks (Wu et al. 2009 ; Zhao et al. 2011 ; Yin et al. 2018 ), inversions using interferometric synthetic aperture radar (InSAR) measurements (Shen et al. 2009 ; Wan et al. 2016 ; Fukahata and Hashimoto 2016 ) and seismic reflection data (Hubbard et al. 2010 ; Jia et al. 2010 ; Lu et al. 2014 ). These methods are only available to investigate earthquakes that have already occurred. Seismic reflection and borehole surveys can estimate pre-seismic estimates of fault geometry, but they are costly in operations, and the data are not always available. Additionally, gravity and magnetic (Tian et al. 2017 ), seismic tomography (Tong et al. 2019 ), and magnetotelluric sounding (Zhao et al. 2012 ) cannot provide enough spatial resolution to describe the fault geometry, especially the dip angles. More straightforward and more effective methods have been awaited.

We propose a new method to infer the fault geometry by inverting the long-term slip rate based on the Wallace–Bott hypothesis (Wallace 1951 ; Bott 1959 ) to refine the fault geometry based on the existing inference of the WMF (Hubbard et al. 2010 ) and the surface fault trace (Xie et al. 2011 ). This hypothesis recalls that two bodies in contact slip toward the interfacial tangential traction and asserts that faults slip in the same manner (Fig.  2 ). This is also the base of the widely used stress tensor inversion method (Hardebeck and Michael 2006 ). Moreover, this hypothesis is confirmed based on a large natural earthquake’s seismically inverted slip distributions (Matsumoto et al. 2018 ). Recent detailed analyses of InSAR observations demonstrate that the slip direction at the ground surface is consistent with those of the overall characteristics at seismogenic depths (Fukahata and Hashimoto 2016 ). Therefore, in an ideal case, the fault plane and the principal stresses should uniquely determine the sense and direction of the long-term slip rate. Since geological and geomorphological observations can determine fault strikes, principal stress axes, and long-term slip rates, the Wallace–Bott hypothesis suggests that dip angles can be inferred based on these observations. In fact, this idea is a natural and quantitative extension of our commonly used experience to distinguish the near-vertical strike-slip faults from more tilted dip-slip faults. An advantage of this method is that we can provide observation-based and systematic inferences of the fault geometry even if more direct data such as seismic reflections survey and hypocenter locations are absent.

Wallace–Bott hypothesis. Two bodies contacted through the surface (fault) are subjected to the principal stresses S 1 , S 2 , and S 3 and the projection of them gives the on-fault tangential traction, which is oriented in the direction shown by open arrow. The slip occurs in the same direction (one-sided arrow indicates the motion of the lower plate)

In this study, we first calculate the distribution of heterogeneous stress fields near the WMF based on the focal mechanism solutions (FMS) of 2008 Wenchuan earthquake aftershocks (M ≥ 3.5). The heterogeneous on-fault initial stresses are resolved from the tectonic stress tensors. Then, we use the long-term slip rate together with the inferred stress to refine the WMF structure based on the Wallace–Bott hypothesis. We then implemented dynamic rupture simulations to investigate rupture propagations on the WMF and surrounding faults to assess potential rupture scenarios.

Seismic and tectonic condition of the Wenchuan–Maoxian fault and Lixian fault

The WMF has prominent fault orientation and geometry similarities to the BCF and PGF that caused the 2008 mainshock. The three faults are subparallel and horizontally spaced within 20 km (Fig.  1 ) and might root into the same main detachment with a depth of approximately 20 km. The main fault shapes display a classic ramp-flat geometry (Hubbard et al. 2010 ; Jia et al. 2010 ; Li et al. 2010 ). Tang et al.’s ( 2021b ) simulation revealed that the Wenchuan mainshock changed the stress on the WMF, generating a positive Coulomb-failure stress change ( \(\Delta CFS\) ) in the center portion of the WMF, suggesting that the 2008 mainshock might have clock-advanced the next earthquake on the WMF to some extent. The WMF has not experienced a strong earthquake for more than 300 years since the last M-6.5 earthquake in 1657 (National Earthquake Data Center, http://data.earthquake.cn ), and the current earthquake interval might provide sufficient accumulated strain for a moderate earthquake (5.5 < M < 6.5) (Additional file 1 : Fig. S1).

Field studies show that the WMF primarily exhibited dextral strike-slip faulting with a comparable reverse component (Tang 1991 ; Rongjun et al. 2007 ; Shen et al. 2019 ; Tian et al. 2013 ; Tan et al. 2017 ) since the Late Cenozoic, consistent with the coseismic slip sense and the long-term slip rate of BCF and PGF (Rongjun et al. 2007 ; Densmore et al. 2007 ; Xu et al. 2009 ; Liu-Zeng et al. 2009 ; Feng et al. 2017 ). Based on fission track and (U–Th)/He dating, Shen et al. ( 2019 ) obtained a 0.6 mm/yr long-term slip rate of WMF, which is as high as those along the frontal BCF (~ 0.54 mm/yr before the Wenchuan earthquake (Rongjun et al. 2007 ) and 0.88–0.91 mm/yr after the Wenchuan event (Ran et al. 2013 )].

In addition, according to observational seismic records before the Wenchuan earthquake (Additional file 1 : Fig. S1), four earthquakes occurred near the WMF out of 11 moderate earthquakes (5 < M < 6) in this area. Note that 12 Ms > 5.6 aftershocks occurred within 1 year after the 2008 Wenchuan earthquake, of which one Ms 6.1 event occurred on the southernmost end of the WMF and one Ms 6.0 event in the Lixian Fault (LXF) (Zhao et al. 2011 ).

The buried LXF is geometrically close to the WMF and bridges between the WMF and the BCF by its conjugate geometry (Fig.  1 ), which may result in cascading ruptures to form larger earthquakes. Geological studies are scant, and the activity is yet to be clarified for this fault due to the heavy vegetation cover and rugged terrain. The maximum altitude difference exceeds 3 km (Tan et al. 2017 ). The LXF appears to extend to the Miyaluo Fault, a well-defined geological structure (Fig.  1 ) displaying a sinistral sense of slip with unknown slip rates and thought to be active at least during the Pleistocene based on the fission track dating (Yang and Zhang 2010 ). Several earthquakes larger than M-7 were recorded on the Fubianhe and Songpinggou Faults, which are subparallel to the Miyaluo Fault with an NW strike orientation (Fig.  1 ). In the following section of the stress inversion analysis, the LXF exhibits the most favorable orientation in the regional stress field. Therefore, the LXF might display a comparatively high seismic potential, although none of the great earthquakes has been recorded to have occurred on it recently.

Regional stress inversion and modeling

The initial stress conditions are fundamental in controlling dynamic earthquake rupture processes (Harris 2004 ; Ramos et al. 2022 ). We inferred the present regional tectonic stress field by inverting earthquake FMSs (from Li et al. 2019 ) of 2008 Wenchuan earthquake aftershocks for computationally simulating earthquake scenarios in this region. We adopted a damped linear stress inversion method developed by Hardebeck and Michael ( 2006 ) to determine the stress field using FMSs. This method enables us to infer the directions of the principal stresses σ r ( r  = 1, 2, and 3 for the maximum, intermediate, and minimum principal stresses, respectively) and the ratio between the principal stresses σ r , defined as \(\varnothing =({\sigma }_{2}-{\sigma }_{3})/\left({\sigma }_{1}-{\sigma }_{2}\right)\) . The details of the method and results are described in Additional file 1 : Text S1.

However, in our dynamic rupture simulation, these observationally constrained parameters are insufficient to determine the absolute values of the principal stresses. Two unknowns are needed and obtained by imposing other prior information. For simplification, we first assume that one of the principal stress axes is vertical based on the observation near the WMF, as by Anderson ( 1905 ). Then, we constrain \({\sigma }_{H}/{\sigma }_{v}\) to 1.6, an approximate value from the in situ stress measurements along the BCF after the Wenchuan earthquake (Qin et al. 2015 ) and previous simulation works (Duan 2010 ; Zhang et al. 2019 ; Tang et al. 2021b ), where \({\sigma }_{H}\) and \({\sigma }_{v}\) denote the magnitudes of the maximum horizontal and vertical principal stresses. The \({\sigma }_{H}\) along the depth is constrained by assuming that the average stress drop satisfies the stress drop derived from empirical relations (Kanamori and Anderson 1975 ). We assume typical overburden stress and that the pore pressure increases with the depth, and the latter gradient gradually approaches the former (Rice 1992 ), so that the stress is tapered to the ambient value at a specific depth (here, we set 5 km) (Rice 1993 ). Figure  3 shows the resulting stress field model of the WMF. The along-strike rotation of the principal horizontal stresses reflects our inversion result. Since there are no independent constraints on the specific depth of the WMF, we discuss its influence in the section " Effect of saturation depth of stress ".

Along strike distribution of three-component principal stress ( \({\sigma }_{H}\) , \({\sigma }_{h}\) for the maximum and minimum horizontal principal stresses, and \({\sigma }_{v}\) for vertical principal stresses). The lower right figure shows the angle between the east and the maximum horizontal stress axis (negative denotes the clockwise), and black box denotes the projection of LXF

Fault geometry construction

Our dynamic model must match the observed oblique dextral slip WMF and the almost pure sinistral strike-slip LXF in the long-term slip rate. We set the LXF as vertical, with a 0.5 km buried depth and a 15 km basal depth inferred from the average seismogenic depth (Shen et al. 2009 ; Wan et al. 2016 ; Ramirez-Guzman and Hartzell 2020 ). In the simulation, we model only the southern portion of the LXF, and do not consider its possible extent further to the north, which could be hypothetically a link to the Miyaluo Fault (Yang and Zhang 2010 ).

The aftershocks of the M-7.9 Wenchuan earthquake and the 2013 M-7.0 Lushan earthquake delineated a prominent seismic gap approximately 50 km long near the town of Dayi (the blue box in Fig.  1 ). Several studies suggest that this gap can correspond to a possible unruptured fault segment (Wang et al. 2015 , 2018 ; Liu et al. 2018 ), and, in this study, we assume that the south part of the WMF terminates near this gap.

Although we could mimic this fault system with several planar rectangular fault segments, we developed a nonplanar fault model by considering existing data and geological inferences. We developed the WMF geometry model by considering (1) geological investigations and seismic reflection data (Xu et al. 2009 ; Hubbard et al. 2010 ; Jia et al. 2010 ), providing an approximate geometry of the WMF at a depth: a fault subparallel to the BCF, with a larger dip angle and listric geometry. (2) The surface fault trace from field investigations (Xie et al. 2011 ) provides along-strike constraints on the surface fault structure. (3) The early aftershock locations (Yin et al. 2018 ) help constrain the south section of the WMF geometry. Based on these constraints, we refine the geometry by developing a new method by inverting the long-term slip rate and the regional stress field based on the above consistency between the slip direction and the tangential traction (Wallace–Bott hypothesis) (Fig.  2 ). The direction of the on-fault tangential traction is fully determined from the stress tensor inversion analysis, including the orientations of the three principal stress axes and the ratio between the magnitudes of the principal stresses (stress ratio) (Michael 1984 ). The validity of this new method is observationally supported by Matsumoto et al ( 2018 ); in the case of the 2016 M7.0 Kumamoto, Japan, earthquake, they demonstrated the good coincidence of the directions between the coseismic slip constrained by the seismic inversion (Asano and Iwata 2016 ) and the tangential traction on the observationally constrained dipping fault surface where the tangential traction is resolved from the principal stresses obtained by the stress tensor inversion.

The readers should not confound our inferring dip angles with the internal or optimal frictional angles determined by the frictional coefficient. The dip angle inferred in our method is independent of the friction coefficient. In other words, our method does not hypothesize that faults occur in the most optimal orientation and has the flexibility to apply to misoriented faults under the recent tectonic stress conditions. Our approach can be valid if the regional stress fields are stable over time to have the observed long-term slip rates. While the temporal stability of tectonic stress regime (or the absolute stress level, not the stress change) considered here is an issue that needs more intensive investigations, at least, it appears to be insensitive to individual occurrence of large earthquakes (i.e., Imanishi et al. 2012 ).

Geometry inversion method

A proper choice of geometry models should be made depending on purposes by considering the available data and previous inferences (Xu et al. 2009 ; Hubbard et al. 2010 ; Jia et al. 2010 ). A model of listric fault geometry is considered here by following the above models. The general dip angles are allowed to vary continuously along the fault strike, and the surface fault traces are fixed, as given by the data. We use one inversion parameter \(\eta\) to describe the fault dip in the local area by assuming the listric fault geometry, as described by (Wan et al. 2016 )

where z is the depth of the fault plane, and y is the horizontal distance of the fault plane from its surface trace. The fault surface is described by the B-spline varying continuously along the strike, with its downdip curvature dictated by the parameter \(\eta\) (Fig.  4 ). \({h}_{0}\) is the fault depth to invert and is assumed to be the same for all fault segments. n B-spline curves are uniformly distributed along the fault strike with equal horizontal intervals X to create the fault surface. Therefore, we have n  + 1 nonlinear model parameters to determine (i.e., n \(\eta\) values plus \({h}_{0}\) )

Schematic to describe how the fault geometry is related to traction at measured points. We get the strike and dip angles for each collocation point using cubic spline interpolation (McKinley and Levine 1998 ), and the traction on each collocation point can be calculated through resolving the heterogeneous tectonic stress tensor along the fault strike and dip. Then, we get the averaged traction from all collocation points in the spherical computational region with radius R

The direction of the long-term slip rate is approximately collinear with the traction on the fault plane, which is equivalent to solving the problem of minimizing the difference between the two ratios. We solve the inverse problem in the sense of Tikhonov ( 1963 ), taking a regularized solution to be a model minimizing an objective function, \(\Psi\) , defined by

where T 1 and T 2 denote the averaged traction along the strike and dip within each spherical computational region with radius R, respectively, represented at the measured points. The averaged traction introduced here can be regarded as a comprehensive effect within the computational region R and depends on the regional stress and fault geometry (Fig.  4 ). The \({{\varvec{V}}}_{1}\) and \({{\varvec{V}}}_{2}\) denote the horizontal and vertical long-term slip rates at the measured points. We choose the damping matrix L to be a simple, first-difference operator (make sense for only \(\eta\) not \(h\) ). The regularization parameter \(\lambda\) is a positive number that balances the effects of data fitting and model smoothness, controlling the smoothness of model space.

However, the observed slip-rate ratios are not always obtained for each measured point, where either horizontal or vertical long-term slip rates can be given. We prefer to use an approximate form

where \(\alpha\) is a scaled value manually set to meet the best initial misfit. We construct the forward model to connect the fault geometry parameters to weighted average tractions \({{\varvec{T}}}_{1}\left({\varvec{m}}\right)\) and \({{\varvec{T}}}_{2}\left({\varvec{m}}\right)\) near the surface (Fig.  4 ). We used the conjugate gradient method to obtain the optimal solution of \({\varvec{m}}\) . The forward calculation and inversion procedures were further elaborated in detail in Additional file: 1 Appendix A of the supplementary material.

Application to the WMF

In this study, the model parameter \({\varvec{m}}\) include ten downdip curvature parameters \(\eta\) uniformly distributed along the fault strike, plus \({h}_{0}\) for all curves. Under the condition that the values of \(\eta\) can accurately describe the surface trace and the nonplanar fault surface, we expect as few parameters as possible to avoid the overfitting problem due to limited observed data. We collected the vertical and horizontal components of the long-term slip rate from a few studies (Ma et al. 2005 ; Rongjun et al. 2007 ; Ran et al. 2013 ; Shen et al. 2019 ) shown in Fig.  5 and implemented five measured points with seven slip rates along the WMF in the inversion. We choose the B-spline interval X = 16 km, radius R = 15 km, and the update strategy of the regularization parameter \(\lambda\) based on the additional tests (Additional file 1 : Text S2).

Map showing the measured points of long-term slip rate along the WMF. Each gray box contains the locations and long-term slip rates measured from different researchers. Dextral means dextral strike-slip rates and Up (NW) means vertical slip rates with northwestern block being hanging wall. The red and blue lines are the same with Fig.  1 . The red stars denote the measured points of long-term slip rate. Gray circles represent aftershocks of the 2008 earthquake (Yin et al. 2018 )

The long-term slip rate cannot be accurately measured due to the uncertainty of the geological age estimation of sediments (Rongjun et al. 2007 ; Shen et al. 2019 ). The inversion uncertainty is estimated using 100 bootstrap resamplings of the entire data set (seven slip rates) to obtain a more reliable result. The data from each measured point obey a Gaussian distribution. For each inversion, we set the same initial model as \({\varvec{m}}=[{\eta }_{0},\dots ,{\eta }_{0}, {h}_{0}]\) , where \({\eta }_{0}=5.0\) , h 0  = 18 km (Fig.  6 a). A few aftershocks are distributed at the south end of the WMF (Additional file 1 : Fig. S2), which is helpful as prior information to constrain one of the inversion parameters. We can fit the aftershock distributions using Eq. ( 1 ) with \({\eta }_{0}=5.0\) , h 0  = 18 km. Therefore, the parameter \(\eta\) at the southernmost position ( x  = 0 km) is enforced for little variation during the inversion.

Estimation of WMF geometry using 150 inversions. a Initial fault surface derived from the initial value of \({\eta }_{0}=5.0\) , h 0  = 18 km. b Final fault surface derived from average \(\eta\) of 150 inversions

An average of 100 inversions is derived to obtain the final fault model (Fig.  6 b), where the dip angles near the surface vary along the fault strike due to the constraint of the observed long-term slip rates (Fig.  7 a). The final inversion results are distributed in a narrow strand (Fig.  7 b). All objective functions steadily decrease versus the iterations (Fig.  7 c) and uncertainty of inverted dip angle is within 10° (Fig.  7 d), justifying our method. The predicted average fault dip angles at the intersection of WMF and free surface are approximately 60°, typically larger than that in the southern section of the BCF. The above results are consistent with the previous studies (Xu et al. 2009 ; Hubbard et al. 2010 ; Jia et al. 2010 ). Differences in the dip angle can lead to differences in rupture scenarios, including the strong ground motion and surface offset for the BCF and WMF.

Uncertainty of WMF geometry estimation. a Observed long-term slip rates and scaled traction. The blue and orange lines present weighted average shear stress along the dip and strike, respectively. The length of the bar indicates the standard deviation of slip rate. The vertical thin pink lines indicate the position of B-spline curves. b Distribution of inverted parameters τ, the red line is the average of 150 inversions (gray lines), and blue triangles indicate the position of interpolation curves. The average of inverted fault depth \({h}_{0}=16.11\) is not shown here. c Objective functions versus iterations from 100 inversions. d Standard variance distribution of fault dips derived from 150 inversions

While we believe that this method of fault geometry inversion is an effective physically based and observationally supported tool, this method could involve ambiguities due to the limitations of the available data and model constraints in inversion analyses. Furthermore, the Wallace–Bott hypothesis is not always satisfied, as the slip direction might deviate from the tangential traction orientation by considering several factors, including local variation in fault geometry (Lejri et al. 2015 ; Lisle 2013 ). Nevertheless, we can find the utility of our optimized search method when we wish to refine previous models (Hubbard et al. 2010 ) or decide on a better model based on prior information without more direct geophysical or geological information.

Dynamic rupture simulation

We computationally simulate spontaneous dynamic rupture propagation on the conjugate fault system of the WMF and LXF. The dynamic rupture problems are numerically solved by the Fast Domain Partitioning Boundary Integral Equation Method (FDP-BIEM) (Ando 2016 ; Ando et al. 2017 ; Ando and Kaneko 2018 ), which increases the efficiency of the elastodynamic boundary equation method (BIEM) without degradation in accuracy. We introduce additional ‘free surface elements’, on which a free surface boundary condition is satisfied (Ando and Okuyama 2010 ; Hok and Fukuyama 2011 ). This method could avoid further complicated theoretical computations for the integration kernels, and enable us to introduce rather complicated geometry of fault systems in a half-space medium. We consider a homogeneous elastic half-space, with Lamé parameters λ  =  μ  = 28 GPa and mass density ρ  = 2776 kg/m 3 ( v p  = 5500 m/s, v s  = 3175 m/s). These parameters are based on the first-order approximation of a three-dimensional velocity structure around the Longmen Shan Fault obtained from seismic tomography (Pei et al. 2010 ; Wang et al. 2021 ). The local variation of shallow seismic velocity structures or site effects amplifying ground motion is excluded from this model. Instead, we focus on the source effect.

The boundary condition for the fault surface is considered where the frictional strength is described by the linear slip-weakening friction law (Andrews 1976 )

where D c , \({\mu }_{s}\) , and \({\mu }_{d}\) denote the characteristic slip and the static and dynamic frictional coefficients, respectively. τ and σ n represent the shear traction and normal stress, respectively. Based on the results of high-speed friction experiments (Yao et al. 2013 ), we set the friction coefficients uniformly on the entire fault with μ s  = 0.53 and μ d  = 0.12. We choose D c  = 0.8 m, because its upper bound is estimated to be 1.0–2.0 m by Tang et al. ( 2021b ), who applied the approach proposed by Mikumo et al. ( 2003 ) to estimate the upper bound of D c of the 2008 Wenchuan earthquake, constrained by the slip rates at fault locations from the kinematic inversion (Zhang et al. 2014 ) of near-field waveforms. The neighboring parameter space is also investigated and discussed in the section " Discussion ", considering the uncertainties of these parameters. We first chose moderately low μ s , μ d , and D c to allow the rupture to propagate to the artificially assumed boundaries without being arrested to explore large earthquake scenarios.

As for the numerical model size, we typically employed the boundary elements of approximately 26,824. We control triangular elements to be 500 m on the fault and free surface in the vicinity of the fault, and then increase this size with a maximum rate of 7.5 times toward the outer boundary of the free surface until it reaches the maximum size of around 2000 m, such that higher calculation accuracy can be achieved near the fault plane, and computational expense can be greatly reduced due to the gradual distribution of mesh density (Additional file 1 : Fig. S3).

In the simulation, the ruptures are triggered by overstressing a small circle patch with a radius of 3 km \(,\) where the initial stress is uniform and slightly above the yield stress. Once a rupture is triggered, its subsequent development is controlled by the elastodynamic equation, the friction law, the initial stresses, as well as the fault geometry. Under our current state of knowledge, predicting a likely hypocenter on the fault system of the WMF and LXF is challenging. Therefore, we considered five nucleation locations at the center, the east and west ends of the WMF, and the north and south ends of the LXF. Considering stress changes caused by Wenchuan mainshock and aftershocks (Additional file 1 : Text S3), we set one of the potential epicenters near the Wenchuan County.

Figure  8 shows the total slip distribution on the WMF and LXF and the moment magnitude for each case. The rupture tends to jump from the oblique-slip fault (in the WMF) to the strike-slip fault (in the LXF), but the reverse is more difficult. For instance, a dynamic rupture might start in the central, eastern, or western parts of the WMF, and then cascade to the LXF, leading to three comparable-sized earthquakes with M-7.56, M-7.55, or M-7.57. However, if it starts in the south or north of the LXF, triggering a second rupture on the WMF will be challenging (Fig.  8 , video S1–S2), leading to a smaller earthquake with moment magnitudes 7.05 or 7.06.

Total slips and earthquake magnitudes from different nucleation points noted by gray arrows. The green arrow denotes the point in Fig.  13

The total slip on the LXF is smaller than that on the WMF. This might be surprising, because the stress drop on the LXF is significantly larger than that on the WMF (Additional file 1 : Fig. S4). Noting that the friction coefficients on the two faults are the same, and the faulting and geometry should cause the prominent difference between the coseismic slips. The WMF is a thrust fault with a larger extent in dip and strike than the LXF, which is a buried and smaller purely strike-slip fault. Under the same stress conditions, larger faults tend to slip more, because healing phases suppressing slip may arrive later, and the free-surface effect is reduced for buried faults. Numerical simulations show that thrust faults have larger coseismic slips than normal and strike-slip faults (Oglesby et al. 2000 ; Tang et al. 2021a ) named the hanging wall effect (Abrahamson and Somerville 1996 ), because the reflected waves from the free surface amplify the motions of the reverse fault near the free surface. Therefore, reverse faults tend to slip more than strike-slip faults, even in relatively lower stress conditions.

Rupture directivity effects caused by different nucleation positions lead to a prominent spatial difference, confirmed by surface displacements and PGV (Additional file 1 : Text S4). The site toward which the rupture front propagates will suffer stronger ground motions than the site where the rupture propagates away due to the cumulative effect of the seismic radiation called forward directivity (Somerville et al. 1997 ). We further evaluate the strong ground motion and surface offset in Additional file 1 : Text S4.

We explored several large earthquake scenarios. However, the dynamic rupture propagation results depend on several key factors, including the initial stress, fault geometry, frictional parameters, and hypocenters, which might lead to large uncertainty for different earthquake scenarios. For instance, different cases with a linear increase in stress from the surface down to some specified saturation depth might result in different ground motion distributions. We include this uncertainty by changing the specified depth but lowering the maximum stress, so that the same average stress drop can be maintained for the two cases (Fig.  9 ). Without loss of generality, we model 20 scenarios of dynamic rupture propagations (Table 1 ) to investigate the effects of neighboring parameter space, including μ d (0.12–0.15) (Yao et al. 2013 ), D c (0.8–1.2 m) (Tang et al. 2021b ), and the stress saturation depth.

Simulations (model E3 ~ C3 in Table 1 ) of the case 2 for depth-dependent vertical stress, the lower right sub-figure shows two different depth-dependent stress configurations with saturation depth 5 km and 10 km

Effect of saturation depth of stress

The average stress drops on the WMF and LXF are obtained from observations from the 2008 Wenchuan earthquake, but the distribution of stresses along the depth is poorly constrained. The magnitude of the principal vertical stress depends on the overburden pressure (density and depth), which might also be compensated for with increased pore pressure (Hardebeck and Okada  2018 ). Thus, we build two depth-dependent shear and normal stress regimes with the same average stress drop \({\Delta \sigma }_{w}\) = 2.9 MPa. The result of the first case has been discussed in the result section. However, the second case leads to different earthquake scenarios and distributions of the coseismic slip (Fig.  9 ). For the second case with deeper saturation depths, simulated ruptures starting in the east or center of the WMF propagate to 20 km beyond the nucleation area and arrest at the patch where the normal stress is stronger (Additional file 1 : Fig. S4). Furthermore, the rupture from the west part propagates over the WMF and cascades to the LXF but leads to a lower magnitude (Table 1 ) and coseismic slips than that of W1 (Fig.  8 ). However, unlike the scenarios on the oblique fault (the WMF), the coseismic slip and magnitudes on the sinistral fault (the LXF) are slightly larger than that in case 1. These phenomena suggest that depth-dependent stress patterns might affect dip-slip faults more than strike-slip faults, and a larger stress gradient near the free surface of an oblique-slip fault is more likely to trigger a stronger coseismic slip.

Patterns of ruptures in the WMF and LXF fault systems

The earthquake scenarios from all simulations in Table 1 can be summarized as six patterns (Fig.  10 a). The 20 cases evaluated do not cover the entire range of possibilities, and the assigned probabilities remain arbitrary, because the uncertainty of the geometry and initial stress is not fully considered. However, the proposed scenarios in this study are understandable from a mechanics standpoint; therefore, they appear possible. The spatial heterogeneity introduced in this study originates from fault geometry and inverted regional stress, which principally controls the macroscopic patterns. For instance, cascading ruptures jump from oblique to strike-slip faults, but the reverse process is less likely. The ruptures are arrested on the WMF, whereas they easily run through the entire LXF (Fig.  10 a). Furthermore, a rupture propagating eastward causes greater coseismic displacements than westward propagation.

Slip patterns from all models using simple schematic diagrams a and histogram of the magnitudes ( \({M}_{w}\) ) obtained in the 20 simulations b . Yellow highlighting indicates the path of propagating rupture, and solid red circles denote the nucleation sites

Figure  10 b represents a statistical analysis of the moment magnitude based on the deterministic simulation results. The earthquake scenarios fall into three groups: magnitudes of approximately 7.5, 7.0, or near 6.6. The first group includes earthquakes that start on the WMF and can propagate over all or most parts of the two faults. Earthquakes with magnitudes of approximately 7.0 typically come from the LXF or local ruptures on the WMF. The third group includes earthquakes that start on the WMF but quickly stop near the nucleation patch due to the unfavorable fault orientation because of the friction parameters and on-fault stress.

Incorporating all earthquake scenarios to further assess the potential seismic motion, we calculate the maximum surface displacements shown in Fig.  11 . Static surface deformation is crucial for near-source hazard analyses. Figure  11 shows the two prominent peaks at 40 km and 100 km along-strike distances. Wenchuan town is near the edge of the moderate slip, and Maoxian town is in the extent of the larger slip. The PGV distributions (Additional file 1 : Text S4) indicate moderately higher values of PGV around Maoxian than Wenchuan when a rupture nucleates at the center or eastern end of the WMF. The velocity structures, including site conditions, are not included in these simulations; therefore, the results for ground shaking from an actual earthquake, even if it were the same source as shown in these simulations, would differ significantly due to the structural effects.

The map for maximum surface displacements from 20 simulations. The white triangles present the two main cities (Wenchuan and Maoxian) and red stars denote the epicenters

Furthermore, the values of μ d  = 0.12–0.15 are slightly lower than those used in Wenchuan earthquake simulations ( μ d  = 0.18) (Zhang et al. 2019 ; Tang et al. 2021b ). Suppose the same friction parameters are used in this study, ruptures will be challenging to initiate in the designed nucleation zone, because relative strength parameter S (Bizzarri et al. 2011 ) will be larger. Besides, the steeper dip angle near the surface of the WMF (Fig.  6 ) tends to increase the normal stress, enhancing the fault clamping effect and making it more difficult to slip. Thus, if our model of the WMF with a steeper dip angle is correct, the WMF will be more stable than the BCF in the long term.

Explanation of rupture patterns

We simulated large earthquake scenarios (models E1–C1). Two interesting scientific issues are worthy of further discussion. First, why do cascading ruptures jump from oblique to strike-slip faults, but the reverse process is challenging? Second, why does the eastward rupture cause greater coseismic displacements than the westward propagation? Although we have mentioned that the rupture directivity can answer the second question, we hope to further explore the deeper reasons from a mechanics perspective.

We answer the first question by calculating the shear stress changes \(\Delta \tau\) , normal stress change \(\Delta \sigma\) (positive in compression), and the \(\Delta CFS\) on the two faults under the state when a rupture is approaching the fault intersection (Fig.  12 ). The \(\Delta CFS\) is calculated from \(\Delta CFS=\Delta \tau -{\mu }_{s}\Delta \sigma\) (King et al. 1994 ; Harris 1998 ), measuring whether the fault is brought closer or farther from failure after slip on an adjacent fault (Freed 2005 ; Parsons et al. 2008 ; Liu et al. 2018 ). Ruptures nucleating at the east or west ends of the WMF approach the fault intersection and induce a positive \(\Delta CFS\) on the LXF. The primary contribution of the increase in \(\Delta CFS\) comes from \(\Delta \tau\) rather than \(\Delta \sigma\) , although the normal stress change affects a larger area of the LXF. On the other hand, simulated ruptures nucleating at the north or south ends of the LXF approach the fault intersection and can induce a negative \(\Delta CFS\) on the WMF. The \(\Delta CFS\) in most areas near the intersection of the WMF is negative, although there are small-scale positive \(\Delta CFS\) locally.

Stress change near the fault intersection when rupture is approaching. Coulomb-failure stress ∆ CFS ( t ) = ∆τ( t ) −  \({\mu }_{s}\) ∆σ( t ), where ∆τ is shear stress change, ∆σ normal stress change, and \({\mu }_{s}\) is static frictional coefficient

The fault activity depends on its prestress state besides the \(\Delta CFS\) . However, it is more challenging to observationally determine the absolute prestress level or when in its seismic cycle than the principal stress orientation and the stress ratio. The prediction can be more complicated if we consider further stress heterogeneity than those in our modeling. With such a limitation, the LXF has higher initial shear stress than the WMF (Additional file 1 : Fig. S4); therefore, the LXF will be closer to failure than the WMF. Consequently, with the evolution of the \(\Delta CFS\) of the four cases, it is reasonable to expect the cascading ruptures to jump from oblique to strike-slip faults rather than the reverse.

We now answer the second question: Why does the eastward rupture cause greater coseismic slips than the westward propagation? We emphasize that all conditions in the two cases are the same except for the different nucleation positions or rupture propagating directivity. The increase in the coseismic slip is distributed over most of the fault rather than a local area, meaning that the stress evolution at a specific point (the point shown in Fig.  8 ) on the fault might help reveal the physical mechanism of this phenomenon (Fig.  13 ). The shear stress variations are similar in the two cases, and the dynamic stress drops are the same. However, the normal stress suddenly increases when the rupture propagates from east to west (Fig.  13 a). The seismic waves generated by the rupture superimpose on the edge of the rupture front and couple with the hanging wall moving in the opposite direction (oblique and dextral slip), forming a local compression. This transient increase in normal stress weakened the slip rate to reach a large peak value (Fig.  13 c), resulting in a lower coseismic slip than that of reverse propagating directivity. Our results correlate with the study by Tang et al. ( 2021a ).

Evolution of stress and slip for the cases with different nucleation positions. The location of fault point is indicated in Fig.  8

Conclusions

We developed a new algorithm based on the Wallace–Bott hypothesis for the inversion of fault geometry to infer nonplanar fault geometry in the Wenchuan region. Without sufficient geophysical or geological data to constrain the fault geometry, this is an effective scheme to refine fault geometry with long-term slip-rate data.

Combining deterministic simulations of earthquake dynamics, we assessed the rupture scenarios expected from the present stress field and given the fault geometry of the WMF and LXF. Areas of large ground motion largely depend on where the earthquake nucleates. Several fault nucleation points, friction coefficients, and initial stress states were evaluated, and the general rupture patterns for these earthquake scenarios could fall into three groups. Depending on the initial conditions, the dynamic rupture might start in the LXF, leading to magnitude-7.0 earthquakes, or start in the WMF, then cascade through the LXF, leading to magnitude-7.5 earthquakes, or both start and arrest in the WMF, leading to magnitude-6.5 or -7.0 earthquakes. We limited the extent of faults based on available surface fault data, and the maximum magnitudes of future earthquakes can be more significant if faults not in this model interact.

Furthermore, we find the general tendencies of a reverse oblique fault with a conjugate strike-slip fault. Due to the positive \(\Delta CFS\) near the fault intersection, a rupture starting on the reverse oblique-slip jumps to the strike-slip fault. However, the reverse process is challenging, because the negative \(\Delta CFS\) dominates near the intersection. Furthermore, the ground motion might be enhanced if the rupture propagation direction is consistent with the movement direction of the hanging wall, because a local decrease of normal stress enhances the coseismic slip.

While we applied our new systematic and physically based method to the case of WMF/LXF, this would apply to any other fault systems in the world where available data are limited to develop models. Therefore, we expect this method to be further tested in other cases and to be able to contribute to the assessment of future earthquake scenarios.

Availability of data and materials

Acknowledgement for the data support from China Earthquake Networks Center, National Earthquake Data Center ( https://data.earthquake.cn/ ). Videos in this paper can be downloaded from https://zenodo.org/record/7077841#.YyGRFnZBxD8 . The simulation data and code used in this study are available from the corresponding author upon request.

Abrahamson NA, Somerville PG (1996) Effects of the hanging wall and footwall on ground motions recorded during the Northridge earthquake. Bull Seismol Soc Am 86(1B):S93–S99

Article   Google Scholar  

Anderson EM (1905) The dynamics of faulting. Trans Edinburgh Geol Soc 8(3):387–402

Ando R (2016) Fast domain partitioning method for dynamic boundary integral equations applicable to non-planar faults dipping in 3-D elastic half-space. Geophys J Int 207(2):833–847

Ando R, Kaneko Y (2018) Dynamic rupture simulation reproduces spontaneous multifault rupture and arrest during the 2016 Mw 7.9 Kaikoura earthquake. Geophysical Res Lett. https://doi.org/10.1029/2018GL080550

Ando R, Okuyama S (2010) Deep roots of upper plate faults and earthquake generation illuminated by volcanism. Geophys Res Lett. https://doi.org/10.1029/2010GL042956

Ando R, Imanishi K, Panayotopoulos Y, Kobayashi T (2017) Dynamic rupture propagation on geometrically complex fault with along-strike variation of fault maturity: insights from the 2014 Northern Nagano earthquake. Earth Planets Space. https://doi.org/10.1186/s40623-017-0715-2

Andrews DJ (1976) Rupture velocity of plane strain shear cracks. J Geophys Res 81(32):5679–5687

Aochi H, Fukuyama E (2002) Three-dimensional nonplanar simulation of the 1992 Landers earthquake. J Geophy Res Solid Earth 107(B2):ESE-4

Aochi H, Ulrich T (2015) A probable earthquake scenario near Istanbul determined from dynamic simulations. Bull Seismol Soc Am 105(3):1468–1475

Asano K, Iwata T (2016) Source rupture processes of the foreshock and mainshock in the 2016 Kumamoto earthquake sequence estimated from the kinematic waveform inversion of strong motion data. Earth Planets Space 68(1):1–11

Bizzarri A (2011) On the deterministic description of earthquakes. Rev Geophys. https://doi.org/10.1029/2011RG000356

Bott MH (1959) The mechanics of oblique slip faulting. Geol Mag 96(2):109–117

Densmore AL, Ellis MA, Li Y, Zhou R, Hancock GS, Richardson N (2007) Active tectonics of the Beichuan and Pengguan faults at the eastern margin of the Tibetan Plateau. Tectonics. https://doi.org/10.1029/2006TC001987

Duan B (2010) Role of initial stress rotations in rupture dynamics and ground motion: a case study with implications for the Wenchuan earthquake. J Geophys Res Solid Earth. https://doi.org/10.1029/2009JB006750

Elliott AJ, Oskin ME, Liu-Zeng J, Shao Y (2015) Rupture termination at restraining bends: the last great earthquake on the Altyn Tagh Fault. Geophys Res Lett 42:2164–2170

Feng G, Jónsson S, Klinger Y (2017) Which fault segments ruptured in the 2008 Wenchuan earthquake and which did not? New evidence from near-fault 3D surface displacements derived from SAR image offsets. Bull Seismol Soc Am 107(3):1185–1200

Freed AM (2005) Earthquake triggering by static, dynamic, and postseismic stress transfer. Annu Rev Earth Planet Sci 33:335–367

Fukahata Y, Hashimoto M (2016) Simultaneous estimation of the dip angles and slip distribution on the faults of the 2016 Kumamoto earthquake through a weak nonlinear inversion of InSAR data. Earth Planets Space 68(1):1–10

Hardebeck JL, Michael AJ (2006) Damped regional-scale stress inversions: Methodology and examples for southern California and the Coalinga aftershock sequence. J Geophy Res Solid Earth 111(B11):B11310. https://doi.org/10.1029/2005JB004144

Hardebeck JL, Okada T (2018) Temporal stress changes caused by earthquakes: a review. J Geophys Research Solid Earth 123(2):1350–1365

Harris RA (1998) Introduction to special section: Stress triggers, stress shadows, and implications for seismic hazard. J Geophy Res Solid Earth 103:24347–24358

Harris RA (2004) Numerical simulations of large earthquakes: dynamic rupture propagation on heterogeneous faults. Pure Appl Geophys 161(11/12):2171–2181

Google Scholar  

Harris RA, Barall M, Lockner DA, Moore DE, Ponce DA et al (2021) A geology and geodesy based model of dynamic earthquake rupture on the Rodgers Creek-Hayward-Calaveras fault system, California. J Geophys Res Solid Earth 126(3):e2020JB020577

Hok S, Fukuyama E (2011) A new BIEM for rupture dynamics in half-space and its application to the 2008 Iwate-Miyagi Nairiku earthquake. Geophys J Int 184(1):301–324

Hubbard J, Shaw JH, Klinger Y (2010) Structural setting of the 2008 Mw 7.9 Wenchuan, China, earthquake structural setting of the 2008 Mw 7.9 Wenchuan, China, earthquake. Bull Seismol Soc Am 100(5B):2713–2735

Imanishi K, Ando R, Kuwahara Y (2012) Unusual shallow normal-faulting earthquake sequence in compressional northeast Japan activated after the 2011 off the Pacific coast of Tohoku earthquake. Geophysical Res Lett 39(9):100. https://doi.org/10.1029/2012GL051491

Jia D, Li Y, Lin A, Wang M, Chen W et al (2010) Structural model of 2008 Mw 7.9 Wenchuan earthquake in the rejuvenated Longmen Shan thrust belt, China. Tectonophysics 491(1–4):174–184

Kanamori H, Anderson DL (1975) Theoretical basis of some empirical relations in seismology. Bull Seismol Soc Am 65(5):1073–1095

King GC, Stein RS, Lin J (1994) Static stress changes and the triggering of earthquakes. Bull Seismol Soc Am 84(3):935–953

Lejri M, Maerten F, Maerten L, Soliva R (2015) Paleostress inversion: a multi-parametric geomechanical evaluation of the Wallace-Bott assumptions. Tectonophysics 657:129–143

Li Y, Jia D, Shaw JH, Hubbard J, Lin A et al (2010) Structural interpretation of the coseismic faults of the Wenchuan earthquake: three-dimensional modeling of the Longmen Shan fold-and-thrust belt. J Geophys Res Solid Earth. https://doi.org/10.1029/2009JB006824

Li X, Hergert T, Henk A, Wang D, Zeng Z (2019) Subsurface structure and spatial segmentation of the Longmen Shan fault zone at the eastern margin of Tibetan Plateau: evidence from focal mechanism solutions and stress field inversion. Tectonophysics 757:10–23

Lisle RJ (2013) A critical look at the Wallace-Bott hypothesis in fault-slip analysis. Bulletin De La Société Géologique De France 184(4–5):299–306

Liu Z, Liang C, Hua Q, Li Y, Yang Y et al (2018) The seismic potential in the seismic gap between the Wenchuan and Lushan earthquakes revealed by the joint inversion of receiver functions and ambient noise data. Tectonics 37:4226–4238

Liu-Zeng J, Zhang Z, Wen L, Tapponnier P, Sun J et al (2009) Coseismic ruptures of the 12 May 2008, Ms 80 Wenchuan earthquake, Sichuan East–west crustal shortening on oblique, parallel thrusts along the eastern edge of Tibet. Earth Planet Sci Lett 286(3–4):355–370

Lu R, He D, John S, Wu JE, Liu B, Chen Y (2014) Structural model of the central Longmen Shan thrusts using seismic reflection profiles: Implications for the sediments and deformations since the Mesozoic. Tectonophysics 630:43–53

Luna LM, Hetland EA (2013) Regional stresses inferred from coseismic slip models of the 2008 Mw 7.9 Wenchuan, China, earthquake. Tectonophysics 584:43–53

Ma S, Beroza GC (2008) Rupture dynamics on a bimaterial interface for dipping faults. Bull Seismol Soc Am 98(4):1642–1658

Ma B, Su G, Hou Z, Shu S (2005) Late quaternary slip rate in the central part of the Longmen Shan fault zone from terrace deformation along the Minjiang River. Seismology and Geology 27(2):234–242

Matsumoto S, Yamashita Y, Nakamoto M, Miyazaki M, Sakai S et al (2018) Prestate of stress and fault behavior during the 2016 Kumamoto earthquake (M7. 3). Geophys Res Lett 45(2):637–645

McKinley S, Levine M (1998) Cubic spline interpolation. Coll Redwoods 45(1):1049–1060

Michael AJ (1984) Determination of stress from slip data: faults and folds. J Geophys Res Solid Earth 89(B13):11517–11526

Mikumo T, Olsen KB, Fukuyama E, Yagi Y (2003) Stress-breakdown time and slip-weakening distance inferred from slip-velocity functions on earthquake faults. Bull Seismol Soc Am 93(1):264–282

Oglesby DD (2005) The dynamics of strike-slip step-overs with linking dip-slip faults. Bull Seismol Soc Am 95(5):1604–1622

Oglesby DD, Mai PM (2012) Fault geometry, rupture dynamics and ground motion from potential earthquakes on the North Anatolian Fault under the Sea of Marmara. Geophys J Int 188(3):1071–1087

Oglesby DD, Archuleta RJ, Nielsen SB (2000) The three-dimensional dynamics of dipping faults. Bull Seismol Soc Am 90(3):616–628

Parsons T, Ji C, Kirby E (2008) Stress changes from the 2008 Wenchuan earthquake and increased hazard in the Sichuan basin. Nature 454(7203):509–510

Pei S, Su J, Zhang H, Sun Y, Toksöz MN et al (2010) Three-dimensional seismic velocity structure across the 2008 Wenchuan Ms 8.0 earthquake, Sichuan, China. Tectonophysics 491(1–4):211–217

Qin X, Chen Q, Wu M, Tan C, Feng C, Meng W (2015) In-situ stress measurements along the Beichuan-Yingxiu fault after the Wenchuan earthquake. Eng Geol 194:114–122

Ramirez-Guzman L, Hartzell S (2020) 3-D joint geodetic and strong-motion finite fault inversion of the May 12, 2008, Wenchuan. China Earthquake Geophys J Int 222(2):1390–1404

Ramos MD, Huang Y, Ulrich T, Li D, Gabriel AA, Thomas AM (2021) Assessing margin-wide rupture behaviors along the Cascadia megathrust with 3-D dynamic rupture simulations. J Geophys Res Solid Earth 126(7):e2021JB022005

Ramos MD, Thakur P, Huang Y, Harris RA, Ryan K (2022) Working with dynamic earthquake rupture models: a practical guide. Seism Res Lett 93(4):2096–2110

Ran YK, Chen WS, Xu XW, Chen LC, Wang H, Yang CC, Dong SP (2013) Paleoseismic events and recurrence interval along the Beichuan-Yingxiu fault of Longmenshan fault zone, Yingxiu, Sichuan, China. Tectonophysics 584:81–90

Rice JR (1992) Fault stress states, pore pressure distributions, and the weakness of the San Andreas fault. Int Geophys 51:475–503

Rice JR (1993) Spatio-temporal complexity of slip on a fault. J Geophys Res 98(B6):9885–9907

Rongjun Z, Yong L, Densmore AL, Ellis MA, Yulin H, Yongzhao L, Xiaogang LI (2007) Active tectonics of the Longmen Shan region on the eastern margin of the Tibetan plateau. Acta Geologica Sinica-English Edition 81(4):593–604

Shen ZK, Sun J, Zhang P, Wan Y, Wang M, Bürgmann R et al (2009) Slip maxima at fault junctions and rupturing of barriers during the 2008Wenchuan earthquake. Nat Geosci 10:718–724

Shen X, Tian Y, Zhang G, Zhang S, Carter A, Kohn B et al (2019) Late Miocene hinterland crustal shortening in the Longmen Shan thrust belt, the eastern margin of the Tibetan Plateau. Journal of Geophysical Research: Solid Earth 124(11):11972–11991

Shi B, Anooshehpoor A, Brune JN, Zeng Y (1998) Dynamics of thrust faulting: 2D lattice model. Bull Seismol Soc Am 88(6):1484–1494

Somerville PG, Smith NF, Graves RW, Abrahamson NA (1997) Modification of empirical strong ground motion attenuation relations to include the amplitude and duration effects of rupture directivity. Seismol Res Lett 68(1):199–222

Tan XB, Xu XW, Lee YH, Lu RQ, Liu Y et al (2017) Late Cenozoic thrusting of major faults along the central segment of Longmen Shan, eastern Tibet: evidence from low-temperature thermochronology. Tectonophysics 712:145–155

Tang RC (1991) The Quaternary activity characteristics of several major active faults in the Songpan-Longmenshan region. Earthquake Res China 7:64–71

Tang R, Yuan J, Gan L (2021a) Free-surface-induced supershear transition in 3-D simulations of spontaneous dynamic rupture on oblique faults. Geophys Res Lett 48(3):e2020GL091621

Tang R, Zhu S, Gan L (2021b) Dynamic rupture simulations of the 2008 79 Wenchuan earthquake: Implication for heterogeneous initial stress and complex multifault geometry. J Geophys Res Solid Earth 126(12):e2021JB022457

Tian Y, Kohn BP, Gleadow AJ, Hu S (2013) Constructing the Longmen Shan eastern Tibetan Plateau margin: Insights from low-temperature thermochronology. Tectonics 32(3):576–592

Tian T, Zhang J, Liu T, Jiang W, Zhao Y (2017) Morphology, tectonic significance, and relationship to the Wenchuan earthquake of the Xiaoyudong Fault in Western China based on gravity and magnetic data. J Asian Earth Sci 138:672–684

Tikhonov AN (1963) On the solution of ill-posed problems and the method of regularization. Russian Academy of Sciences 151(3):501–504

Tong P, Yang D, Huang X (2019) Multiple-grid model parametrization for seismic tomography with application to the San Jacinto fault zone. Geophys J Int 218(1):200–223

Ulrich T, Gabriel AA, Ampuero JP, Xu W (2019) Dynamic viability of the 2016 Mw 7.8 Kaikōura earthquake cascade on weak crustal faults. Nature Commun 10(1):1–16

Wallace RE (1951) Geometry of shearing stress and relation to faulting. J Geol 59:118–130

Wan Y, Shen ZK, Bürgmann R, Sun J, Wang M (2016) Fault geometry and slip distribution of the 2008 Mw 7.9 Wenchuan, China earthquake, inferred from GPS and InSAR measurements. Geophys J Int 208(2):748–766

Wang H, Chen L, Ran Y, Lei S, Li X (2015) Paleoseismic investigation of the seismic gap between the seismogenic structures of the 2008 Wenchuan and 2013 Lushan earthquakes along the Longmen Shan fault zone at the eastern margin of the Tibetan Plateau. Lithosphere 7:14–20

Wang C, Liang C, Deng K, Huang Y, Zhou L (2018) Spatiotemporal distribution of microearthquakes and implications around the seismic gap between the Wenchuan and Lushan earthquakes. Tectonics 37:2695–2709

Wang Z, Wang J, Yang X (2021) The role of fluids in the 2008 Ms 8.0 Wenchuan earthquake. J Geophys Res Solid Earth 126:e2020JB019959

Wu JP, Huang Y, Zhang TZ, Ming YH, Fang LH (2009) Aftershock distribution of the MS8. 0 Wenchuan earthquake and 3D P-wave velocity structure in and around source region. Chinese J Geophys 52(1):102–111

Xie XS, Jiang WL, Feng XY (2011) Discussion appearance of surface fractures along the rear-range of Longmenshan mountain during 2008 Ms8.0 Wenchuan earthquake. Acta Seismol Sin 33(1):62–81

Xu X, Wen X, Yu G, Chen G, Klinger Y et al (2009) Coseismic reverse-and oblique-slip surface faulting generated by the 2008 Mw 79 Wenchuan earthquake, China. Geology 37(6):515–518

Yang N, Zhang YQ (2010) Tission-track dating for activity of the Longmenshan fault zone and uplifting of the western Sichuan Plateau. J Geomech 16(4):359–371

Yang H, Liu Y, Lin J (2013) Geometrical effects of a subducted seamount on stopping megathrust ruptures. Geophys Res Lett 40:1–6

Yao S, Yang H (2022) Hypocentral dependent shallow slip distribution and rupture extents along a strike-slip fault. Earth Planet Sci Lett 578:117296

Yao L, Ma S, Shimamoto T, Togo T (2013) Structures and high-velocity frictional properties of the Pingxi fault zone in the Longmenshan fault system, Sichuan, China, activated during the 2008 Wenchuan earthquake. Tectonophysics 599:135–156

Yin XZ, Chen JH, Peng Z, Meng X, Liu QY et al (2018) Evolution and distribution of the early aftershocks following the 2008 Mw 7.9 Wenchuan earthquake in Sichuan, China. J Geophys Res Solid Earth 123(9):7775–7790

Zhang Y, Wang R, Zschau J, Chen YT, Parolai S, Dahm T (2014) Automatic imaging of earthquake rupture processes by iterative deconvolution and stacking of high-rate GPS and strong motion seismograms. J Geophys Res Solid Earth 119(7):5633–5650

Zhang Z, Zhang W, Chen X, Li PE, Fu C (2017) Rupture dynamics and ground motion from potential earthquakes around Taiyuan, China. Bull Seismol Soc Am 107(3):1201–1212

Zhang Z, Zhang W, Chen X (2019) Dynamic rupture simulations of the 2008 Mw7.9 Wenchuan earthquake by the curved grid finite-difference method. J Geophys Res Solid Earth 124(10):10565–11058

Zhao B, Shi Y, Gao Y (2011) Relocation of aftershocks of the Wenchuan M S 8.0 earthquake and its implication to seismotectonics. Earthq Sci 24(1):107–113

Zhao G, Unsworth MJ, Zhan Y, Wang L, Chen X et al (2012) Crustal structure and rheology of the Longmenshan and Wenchuan Mw 7.9 earthquake epicentral area from magnetotelluric data. Geology 40(12):1139–1142

Download references

Acknowledgements

The comments and suggestions from reviewers Haiming Zhang and Thomas Ulrich are very appreciated. The authors would like to thank the editor Hideo Aochi, Junjie Ren, So Ozawa, Tawei Chang, and Ruth Harris for providing helpful suggestions. The authors would also like to thank Xinzhong Yin and Jiuhui Chen for providing aftershock data.

This work is supported in part by MEXT, Japan, under its Earthquake and Volcano Hazards Observation and Research Program, National Natural Science Foundation of China (4230040097), Basic Scientific Research Program from Yangtze Delta Region Institute (U032200109), as well as China Scholarship Council. Oakforest-PACS in the University of Tokyo was used for the numerical simulations under the “Joint Usage/Research Center for Interdisciplinary Large-scale Information Infrastructures” and “High Performance Computing Infrastructure” in Japan (Project ID: jh210023-NAH), and computational resources of Earth Simulator was provided by JAMSTEC through the HPCI System Research Project (Project ID: hp220105).

Author information

Authors and affiliations.

Yangtze Delta Region Institute of University of Electronic Science and Technology of China, Huzhou, Zhejiang, China

Rongjiang Tang

Department of Earth and Planetary Science, School of Science, University of Tokyo, Bunkyō, Tokyo, Japan

Ryosuke Ando

You can also search for this author in PubMed   Google Scholar

Contributions

RT conducted stress inversion, fault geometry inversion, and dynamic simulation, and drafted the manuscript. RA provided the FDP-BIEM code, editing, reviews, and improved methodology. All the authors read and approved the final manuscript.

Corresponding author

Correspondence to Rongjiang Tang .

Ethics declarations

Ethics approval and consent to participate.

Not applicable.

Consent for publication

Competing interests.

The authors declare that they have no competing interests.

Additional information

Publisher's note.

Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.

Supplementary Information

Additional file 1:.

Text S1. Stress inversion. Text S2. Test on parameters for fault geometry inversion. Text S3. The effect from Wenchuan mainshock . Appendix A. Conjugate gradient method for fault geometry inversion. Text S4 . Peak ground displacement and velocity. Figure S1 . Map showing the setting of major faults of the center Longmen Shan fault zone. The red bold black lines indicate the field-confirmed rupture traces in Wenchuan earthquake (Xu et al. 2009 ; Liu-Zeng et al. 2009 ). The blue solid and dotted lines represent the WMF and Lixian faults, both of which did not slip in the main shock but are very active in aftershocks. Accurate fault trace of the WMF is from the field investigation by Xie et al ( 2011 ), and the location of the LXF trace is from the distribution of aftershocks (Yin et al. 2018 ; Li et al. 2019 ). “Beach ball” represents the lower hemisphere projection of the focal mechanism (strike 225°/dip 39°/rake 120°) of the \({M}_{w} 7.9\) ( \({M}_{s} 8.0\) ) Wenchuan earthquake obtained from Zhang (2009). Pink circles—historic strong earthquakes (Before 2008) larger than M5. Yellow circles— historic strong earthquakes (Before 2008) with magnitude \(4\le M<5\) . The white arrow indicates block motion direction. CD—Chengdu; DY—Dayi; YX—Yingxiu; WC—Wenchuan; LX—Lixian; MX—Maoxian; BC—Beichuan; NB—Nanba. Figure S2 . Seismicity following the 2008 Wenchuan earthquake plotted in the along-strike distances ( a ) and cross-section along the purple box ( b ). Figure S3 . Boundary elements for each fault segment and free surface. We employed the boundary elements of approximately 26,824, and control triangular elements to be 500 m on the fault and free surface in the vicinity of the fault, and increase this size with a maximum rate of 7.5 times toward the outer boundary of the surface until it reaches the maximum size of around 2000 m. Figure S4. Initial on-fault shear stress and normal stress resolved by the tectonic stress, as well as Initial on-fault shear stress drop \(\tau_{drop} = \tau_{0} - \mu_{d} \sigma ,\) μ d  = 0.12. Figure S5. a The spatial distribution of 391 focal mechanism solutions (beachball, lower hemisphere stereographic projection) determined in this study from earthquakes between 2009 and 2016. The color and size of the beachballs indicate the type and seismic moment, respectively. Red, normal (NF) or normal oblique faulting (NS); green, strike-slip faulting (SS); blue, thrust (TF) or thrust oblique faulting (TS); black, mixed type. b Histogram illustrating the faulting type distribution of the earthquake focal mechanism data determined in this study. (From Li et al. 2019 ). Figure S6. Inferred regional stress field. a The trade-off between model length and data misfit, e = 1.6 is selected to be the optimal damping parameter. b An example to show the distribution of Closeness between best stress tensor and inverted stress tensors from 1000 times bootstraps in grid number 1. To estimate the stress uncertainty with 95% confidence, we pick 95% of them that have the highest scalar product with the best stress tensor. c Results of crustal stress field inversion from 440 fault plane solutions for the Longmen Shan region with a grid interval of 0.2° both in longitude and latitude (only the stresses in our calculation domain are shown). d The principal stress axes obtained in each block are denoted by red ( \({\sigma }_{1}\) ), blue ( \({\sigma }_{2}\) ) and green ( \({\sigma }_{3}\) ) in the lower hemisphere stereonet. Figure S7. Along strike distribution of scaled three-component principal stress. The scaled three-component principal stress are marked red ( \({\sigma }_{1}\) ), blue ( \({\sigma }_{2}\) ) and green ( \({\sigma }_{3}\) ) colors, which are derived using bootstrap resampling of the selected fault planes. The black circles denote the stress component closest to the vertical, and the dotted lines are the average of each stress component ( a ). The orange solid line and red dash line represent the trend and average of principal stress orientation ( b ). Figure S8. Inversions of fault geometry with different horizontal intervals of interpolated curves. From top to bottom: Observed long-term slip rates and scaled traction; Distribution of inverted parameters \(\eta\) ; Objective functions versus iterations; Final fault surface derived from inverted parameters \(\eta\) . Other details for each plot refer to Fig. 4 . Figure S9. Same as Additional file 1 : Fig. S7, but with different integral radius. Figure S10. Same as Additional file 1 : Fig. S7, but with regularization parameters. Figure S11. a Calculated Coulomb failure stress change and the movement of Lixian Fault, assuming \({\mu }_{s}=0.52\) (Modified from Tang et al. 2021a , b ). The spatial distribution of aftershocks is from Yin et al. ( 2018 ). b The Coulomb stress change of the WMF, assuming that the LXF ruptures at southern end. Figure S12. Three components and magnitude of surface slip, with eastern and western nucleation points. The white pentagrams denote the epicenter. The white arrow indicates the NE direction. See Fig. 5 for the definition of the x-y-z components. Figure S13. Magnitude of peak ground velocities from different nucleation points. The white pentagrams denote the epicenter. Figure A1 . Schematic to describe the calculation of traction from interpolated fault surface. Green circles are the control points at equal intervals on the dip curves, and orange circles denote the equally spaced points on the interpolation curves. The average interval of collocation points is around 3 km. The inverted triangles represent the locations of the measured points.

Rights and permissions

Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article's Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article's Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by/4.0/ .

Reprints and permissions

About this article

Cite this article.

Tang, R., Ando, R. A systematic scheme to develop dynamic earthquake rupture scenarios: a case study on the Wenchuan–Maoxian Fault in the Longmen Shan, China, thrust belt. Earth Planets Space 76 , 2 (2024). https://doi.org/10.1186/s40623-023-01932-2

Download citation

Received : 29 March 2023

Accepted : 06 November 2023

Published : 02 January 2024

DOI : https://doi.org/10.1186/s40623-023-01932-2

Share this article

Anyone you share the following link with will be able to read this content:

Sorry, a shareable link is not currently available for this article.

Provided by the Springer Nature SharedIt content-sharing initiative

• Wenchuan–Maoxian Fault
• Dynamic simulation
• Fault geometry inversion
• Strong ground motion

Guest Editorial to the Special Issue: Lessons Learned from post-2008 Wenchuan Earthquake Community Recovery

• Published: 19 October 2020
• Volume 104 , pages 1–3, ( 2020 )

Cite this article

• Robert Olshansky 1 ,
• Yu Xiao 2 &
• Daniel Abramson 3  

1615 Accesses

Explore all metrics

Avoid common mistakes on your manuscript.

It has now been 12 years since the Wenchuan Earthquake devastated Sichuan Province, China, in May 2008. By any measure, it was one of the largest natural disasters in modern times, severely affecting an area of 50,193 square miles (130,000 square kilometers), taking over 69,000 lives (with an additional 18,000 permanently missing), and temporarily displacing 15.1 million people. It is certainly the largest natural disaster to strike modern China, comparable only to the 1976 earthquake that killed 242,000 people and destroyed Tangshan, a city of 500,000 people (Zhang et al. 2014 ).

The rapid physical reconstruction, completed in most places in less than 3 years, was remarkable. Several publications at the time documented this rapid reconstruction process (e.g., Yong and Booth 2011 ; Dunford and Li 2011 ) or described the recovery planning and its administration (Ge et al. 2010 ).

But recovery is about more than reconstructing buildings and infrastructure. Successful recovery is about rebuilding economies, communities, and lives. Over the years, researchers have studied the effects of the earthquake and subsequent recovery on households (e.g., Yang et al. 2015 ; Wilczak 2017 ; Fayazi et al. 2019 ), but, with a few exceptions (e.g., Abramson and Qi, 2011 ; Dong, 2012 ; Jiang 2014 ; Liu et al. 2014 ; Chandrasekhar et al. 2014 ; Song et al. 2017 ), relatively little scholarship has reflected upon the quality of community-level recovery or community planning processes. Such studies would fill an important need in the scholarship of disaster recovery, especially because of both the large scale of this disaster—the Wenchuan Earthquake significantly affected one of the nation’s great economies—and because of the unique approach that China took to rapidly rebuild.

As the 10-year anniversary of the Wenchuan Earthquake approached, we sought to identify and highlight emerging research providing critical reflections, through various lenses, on the success of community-level recovery over time. With the generous opportunity provided by Natural Hazards to publish a Special Issue, we solicited papers for the annual conference of the International Association for China Planning, held at Harbin in June 2017. The response, which consisted of 13 papers spanning three conference sessions, confirmed that numerous scholars at multiple institutions throughout China were engaged in thoughtful research that asked important questions about the characteristics, progress, and quality of the recovery since 2008, such as: How has long-term community-scale recovery progressed, socially, economically, and physically? How was the innovative pair-assistance program implemented at the local level? How did recovery work in a range of communities, such as minority and indigenous ethnic communities? In what ways did NGOs facilitate recovery? What lessons does this recovery process provide regarding hazard mitigation, emergency management, preparedness and long-term community recovery practices?

We are now pleased to be able to present five of those papers plus three additional ones here in one volume. The first two papers, by Xiao et al. and by Tang, examine the overall administration of recovery after the 2008 earthquake. Xiao et al. review the planning, budgeting, and finance of the reconstruction, and particularly highlight the roles of pair assistance and land-based financing in paying some of the reconstruction costs. Tang demonstrates the value of taking a longer view of recovery, by describing how the initial 3 years of paired assistance provided a foundation for more deeply rooted collaborative networks that were essential to supporting ongoing recovery in the longer term.

Five of the papers provide us with glimpses into aspects of the community recovery experience at the local level. Shao et al. describe the local reconstruction processes in Weizhou and Yingxiu through the lens of resilience. In so doing, they provide insightful views into some of the local intergovernmental and political challenges regarding development and implementation of local recovery plans. Two of the contributions, by Zhang et al. and Fan et al., examine life in the famous town of New Beichuan, in which tens of thousands of earthquake-displaced inhabitants found new homes and jobs. Zhang et al., using observations and surveys, assess the social and economic life of the residents of New Beichuan. Notably, they contrast the experiences of post-disaster migrants with the indigenous peasants on whose land the new city was constructed. Fan et al. studied the evolution over time of small businesses that had restarted in New Beichuan. They document the unfortunate decline in the number of businesses, but are also able to identify several factors affecting the survival of business operations over time.

The cases of Weizhou, Yingxiu, and Beichuan are representative of urban areas affected by the earthquake, but widespread rural areas were also significantly affected. The paper by Di et al. examines the reconstruction and economic recovery of a rural mountain area with a strong tourism economy. Notably, the residents of Longmen Mountain Town demonstrated bottom-up adaptation to evolving post-earthquake conditions, changing their lifestyles, income sources, and construction methods to both prepare for future disasters and survive in the new post-disaster economy. Wide areas of Sichuan Province reconstructed by concentrating new homes and services into denser rural settlements closer to major transportation routes, infrastructure, and facilities such as schools and hospitals. Peng et al. evaluate the recovery success of three villages that reconstructed in this way. Given that the physical reconstruction and spatial concentration of rural settlements is the dominant form of rural development promoted by policy in China, even where no disaster requires rebuilding, this paper and others in the special issue have broad relevance beyond an assessment of the Wenchuan Earthquake recovery itself.

Finally, Lu et al. use the example of one NGO that actively provided assistance following both the 2008 Wenchuan Earthquake and the subsequent 2013 Lushan Earthquake, showing how it evolved from a provider of resources to the leading partner in a decentralized learning ecosystem. They propose this as a potential model for future NGO disaster response in China.

We believe this volume is the most substantial collection of scholarship accessible to international scholars in English on community recovery following the Wenchuan Earthquake. That said, it is far from a complete exploration of the evolution of communities following this disaster. We look forward to seeing continued scholarship published regarding this internationally and historically significant event.

Abramson D, Qi Y (2011) Urban-rural integration in the earthquake zone: sichuan’s post-disaster reconstruction and the expansion of the Chengdu metropole. Pac Aff 84(3):495–523

Article   Google Scholar  

Chandrasekhar D, Zhang Y, Xiao Y (2014) Nontraditional participation in disaster recovery planning: cases from China, India, and the United States. J Am Plann Assoc 80(4):373–384. https://doi.org/10.1080/01944363.2014.989399

Dong X (2012) Post-disaster recovery planning and sustainable development: a lesson from the Wenchuan earthquake, China 2008. Masters Thesis, University of Illinois at Urbana-Champaign

Dunford M, Li L (2011) Earthquake Reconstruction in Wenchuan: assessing the state overall plan and addressing the ‘forgotten phase’. Appl Geogr 31:998–1009

Fayazi M, Yeh ET, Li F (2019) Development and divergent post-disaster trajectories in a mountain village: temporal dynamics of differentiation after the 2008 Wenchuan earthquake. World Dev 124:10466. https://doi.org/10.1016/j.worlddev.2019.104663

Ge Y, Gu Y, Deng W (2010) Evaluating China’s National post-disaster plans: the 2008 Wenchuan earthquake’s recovery and reconstruction planning. Int J Disaster Risk Sci 1(2):17–27

Google Scholar  

Jiang W (2014) Towns undergoing changes: a case study on the recovery after the Wenchuan earthquake, China. Masters Thesis, University of Illinois at Urbana-Champaign

Liu L, Lin Y, Wang S (2014) Urban design for post-earthquake reconstruction: a case study of Wenchuan County, China. Habitat Int 41:290–299. https://doi.org/10.1016/j.habitatint.2013.09.001

Song Y, Li C, Olshansky R, Zhang Y, Xiao Y (2017) Are we planning for sustainable disaster recovery? Evaluating recovery plans after the Wenchuan earthquake. J Environ Plan Manag 60(1):1–25. https://doi.org/10.1080/09640568.2017.1282346

Wilczak J (2017) Reconstructing Rural Chengdu: urbanization as development in the post-quake context. Ph.D. Dissertation, University of Toronto. http://hdl.handle.net/1807/99022

Yang S, Du J, He S et al (2015) The emerging vulnerable population of the urbanisation resulting from post-disaster recovery of the Wenchuan earthquake. Nat Hazards 75:2103–2118. https://doi.org/10.1007/s11069-014-1413-z

Yong C, Booth DC (2011) The Wenchuan earthquake of 2008: anatomy of a disaster. Science Press and Springer, Beijing and Berlin

Zhang Y, Zhang C, Drake W, Olshansky R (2014) Planning and recovery following the great 1976 Tangshan earthquake. J Plann Hist 14(3):224–243. https://doi.org/10.1177/1538513214549435

Download references

Author information

Authors and affiliations.

Department of Urban and Regional Planning, University of Illinois at Urbana-Champaign, 611 Taft Drive, Champaign, IL, 61820, USA

Robert Olshansky

Toulan School of Urban Studies and Planning, College of Urban & Public Affairs, Portland State University, 506 SW Mill St, Portland, OR, 97201, USA

Department of Urban Design and Planning, University of Washington, 410 Gould Hall, Box 355740, Seattle, WA, 98195, USA

Daniel Abramson

You can also search for this author in PubMed   Google Scholar

Corresponding author

Correspondence to Robert Olshansky .

Additional information

Publisher's note.

Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.

Rights and permissions

Reprints and permissions

About this article

Olshansky, R., Xiao, Y. & Abramson, D. Guest Editorial to the Special Issue: Lessons Learned from post-2008 Wenchuan Earthquake Community Recovery. Nat Hazards 104 (Suppl 1), 1–3 (2020). https://doi.org/10.1007/s11069-020-04341-w

Download citation

Published : 19 October 2020

Issue Date : October 2020

DOI : https://doi.org/10.1007/s11069-020-04341-w

Share this article

Anyone you share the following link with will be able to read this content:

Sorry, a shareable link is not currently available for this article.

Provided by the Springer Nature SharedIt content-sharing initiative

• Find a journal
• Publish with us
• Track your research

Click through the PLOS taxonomy to find articles in your field.

For more information about PLOS Subject Areas, click here .

Loading metrics

Open Access

Peer-reviewed

Research Article

Rapid determination of seismic influence field based on mobile communication big data—A case study of the Luding Ms 6.8 earthquake in Sichuan, China

Roles Formal analysis, Software, Writing – original draft, Writing – review & editing

Affiliation Zhejiang Earthquake Agency, Hangzhou, China

Roles Data curation, Formal analysis, Methodology

* E-mail: [email protected]

Affiliation Beijing Earthquake Agency, Beijing, China

Roles Resources, Software, Supervision

Affiliation Zhejiang Development and Planning Institute, Hangzhou, China

Roles Data curation, Formal analysis

Roles Conceptualization

Roles Data curation

• Dongping Li, 
• Qingquan Tan, 
• Zhiyi Tong, 
• Jingfei Yin, 
• Min Li, 
• Huanyu Li, 
• Haiqing Sun

• Published: May 10, 2024
• https://doi.org/10.1371/journal.pone.0298236
• Peer Review
• Reader Comments

Smartphone location data provide the most direct field disaster distribution data with low cost and high coverage. The large-scale continuous sampling of mobile device location data provides a new way to estimate the distribution of disasters with high temporal–spatial resolution. On September 5, 2022, a magnitude 6.8 earthquake struck Luding County, Sichuan Province, China. We quantitatively analyzed the Ms 6.8 earthquake from both temporal and geographic dimensions by combining 1,806,100 smartphone location records and 4,856 spatial grid locations collected through communication big data with the smartphone data under 24-hour continuous positioning. In this study, the deviation of multidimensional mobile terminal location data is estimated, and a methodology to estimate the distribution of out-of-service communication base stations in the disaster area by excluding micro error data users is explored. Finally, the mathematical relationship between the seismic intensity and the corresponding out-of-service rate of communication base stations is established, which provides a new technical concept and means for the rapid assessment of post-earthquake disaster distribution.

Citation: Li D, Tan Q, Tong Z, Yin J, Li M, Li H, et al. (2024) Rapid determination of seismic influence field based on mobile communication big data—A case study of the Luding Ms 6.8 earthquake in Sichuan, China. PLoS ONE 19(5): e0298236. https://doi.org/10.1371/journal.pone.0298236

Editor: Rahul Priyadarshi, Siksha O Anusandhan, INDIA

Received: September 15, 2023; Accepted: January 19, 2024; Published: May 10, 2024

Copyright: © 2024 Li et al. This is an open access article distributed under the terms of the Creative Commons Attribution License , which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited.

Data Availability: All relevant data are within the paper and its Supporting Information files.

Funding: This study was supported by the Scientific Research Fund of Institute of Engineering Mechanics, China Earthquake Administration (2021D07), Project of Spark Program of Earthquake Sciences, China Earthquake Administration (XH23001B), Zhejiang Provincial Natural Science Foundation of China(LTGG24D040002).The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript.

Competing interests: The authors have declared that no competing interests exist.

Introduction

For a long time, most of the disaster distribution data related to post-earthquake rescue has been obtained by using expensive special equipment. However, due to the uncertainty of when and where an earthquake disaster may occur, the large-scale deployment and maintenance of professional equipment can incur considerable costs. It is also difficult to ensure the connectivity of this equipment and the full coverage of all affected people when an earthquake occurs [ 1 ]. With their ever-increasing popularity, smartphones, as the most widely used electronic devices, have been equipped with computing, communication, storage, and sensing capabilities. Even in disaster scenarios, the probability of people holding smartphones is still very high. Therefore, smartphones are capable of constructing direct field disaster distribution data with low cost and high coverage. The large-scale continuous sampling of mobile device location data provides a new way to estimate disaster distribution with higher temporal–spatial resolution [ 2 ].

Several earthquakes in recent years have shown that when the intensity of the epicenter reaches VIII, communication base stations will usually be out of service, which directly leads to a precipitous drop in the acquisition of smartphone location data after an earthquake. A large amount of smartphone location data disappears at a large scale after an earthquake, and the closer to the hardest hit area, the more obvious the data drop is. In addition, in the event of a nondestructive earthquake, there will be such phenomena as an increase in communication volume at the epicenter and location changes due to the flow of people avoiding the disaster. The change and distribution of mobile communication big data play an indicative role in estimating the extent of devastation in the first instances [ 3 ]. After the Wenchuan earthquake in 2008, communication facilities in the disaster area were severely damaged, and many mobile base stations stopped service, resulting in communication outages in the areas where these stations are located. We analyzed the affected areas and the extent of devastation by collecting the out-of-service data of mobile base stations and mapped the distribution range of the affected areas, which was highly consistent with the intensity distribution data obtained from the post-earthquake field survey ( Fig 1 ).

• PPT PowerPoint slide
• PNG larger image
• TIFF original image

https://doi.org/10.1371/journal.pone.0298236.g001

(We have collected information after the Wenchuan Ms 8.0 earthquake through 2 specialized BBS forums: https://www.txrjy.com/ ; https://club.mscbsc.com/ . Although the data we collected are incomplete and limited, we still have relatively and accurately demonstrated the distribution of the seismic influence field and the approximate area of the macroscopic epicenter through spatial interpolation).

With the prevalence of smartphones and the development of mobile Internet services, in combination with the popularization of global positioning technology, the technology of population data estimation based on precise geographic location has become increasingly mature [ 4 ]. When a user enters a certain geographic space, it is possible to obtain and verify the user’s location information, and the current population size of a geographic space can be inferred through a differentiated model. After an earthquake, we can use the changes in smartphone thermal data to infer the extent of damage to mobile communication infrastructure caused by the earthquake [ 5 , 6 ]. The data sources involved in this work are the various smartphone APP vendors. Because the data is location data of point groups reported in a certain area, personal privacy is not involved, and the data size covers billions of terminals.

Since 2012, telecom operators have successively applied location data analysis to mobile networks. Verizon, a U.S. operator, engages in business consulting by collecting information on the apps used and websites visited by its users, as well as their geographic locations [ 7 ]. The highway monitoring project of French telecom operator Orange and the smart footprint project of Spanish telecom company Telefonica are designed to provide location-derived information to users. Represented by the I-LOV project in Germany, many research institutions have participated in the construction of disaster emergency rescue systems based on smartphone signal searches [ 8 ]. At the World Internet Conference in Wuzhen in November 2014, the comprehensive analysis of China Mobile big data demonstrated a dynamic people flow big data analysis platform, and in December of the same year, the mobile location of big data provided decision-making information for the government in the emergency response to the stampede at the Bund in Shanghai [ 9 , 10 ]. "Location big data" is not merely the result of technological transformation in the computer industry. Cross-border thinking and big data thinking are also applicable to various other industries. Similar work has been conducted in the field of natural disaster research, where MyShake has been developed and built as a global smartphone seismic network that can detect and be triggered by P waves. With the constant downloading of MyShake, the scale of the seismic network has been rapidly developed, forming a global seismic network that utilizes personal smartphones to provide acceleration waveforms [ 11 ]. Since 2017, experts have started participating in the research on rapid acquisition of disaster situations and post-earthquake crowd flow analysis based on mobile Internet location data, and some scholars have begun using mobile location information data to study and track the distribution of earthquake-stricken areas and the responses of people after the earthquake [ 12 ]. Researchers have used smartphone data to analyze the crowd dynamics during the 2017 Jiuzhaigou earthquake. By analyzing the call and SMS data, they identified the densely populated areas and the migration trajectories during the evacuation process, which provided support for the development of effective urgent evacuation strategies [ 13 ]. The MIT Media Lab used smartphone data to analyze changes in human mobility and access to critical urban services in the aftermath of the 2015 Nepal earthquake, and the findings underscored the importance of considering post-disaster mobility dynamics in emergency response and recovery planning [ 14 ]. European-Mediterranean Seismological Centre used smartphone data to dynamically assess post-earthquake population displacement. By analyzing detailed call records [ 15 ], EMSC identified changes in human mobility patterns in the affected areas and quantified the displacement level, which provided information for post-disaster recovery and planning work. Taking an earthquake-stricken area as the main research object, Gao Na compared the demographic data obtained by smartphone location and found the role of annual difference data in the field of sudden disaster emergency relief [ 16 ]. The Nie group from the Institute of Geology, China Earthquake Administration used smartphone location data to analyze the indicators related to earthquake disasters and proposed to regard seismic intensity as a sensitive indicator for smartphone location data [ 17 ]. Zhang used the pre-earthquake and post-earthquake Internet smartphone location data in an earthquake area and adopted the standard deviation ellipse model in spatial econometric analysis to analyze the spatial distribution characteristics of the disappeared smartphone location data in the earthquake area and the oriented direction of its discrete point sets, which determined the direction of the seismic influence field, and further provided technical support for post-disaster situation assessment and emergency response services [ 18 ]. In the process of seismic data processing, relevant experts can address earthquake emergency response, rescue guidance, and other studies by integrating and mining a large amount of complex and multisource data [ 19 ]. In conclusion, the application of "mobile location big data" during earthquake emergency responses can improve the scientificity and accuracy of the decision-making processes during an earthquake and enhance the capabilities of earthquake early warnings and emergency responses [ 20 ].

Materials and methods

Principles of smartphone location big data.

The popularity of smartphones and the development of the Internet (according to the report of "Worldwide Quarterly Smartphone Tracker" of the International Data Corporation (IDC), as of the first quarter of 2022, the global smartphone penetration rate is approximately 82%) has ushered in the era of mobile Internet. In addition, with the development and wide use of the global positioning system (GPS), the generation and development of location-based services have become inevitable trends. Location-based services not only offer convenience but also provide new data sources and possibilities for business intelligence analysis, public affairs management, academic research, and other efforts. A large number of users around the world generate numerous information for sharing every day, and the shared information can be accessed via application interfaces [ 21 ]. The geographic data generated through smartphone location sharing services has brought a new revolution to GIS. The most fundamental issue involved in research using the shared data in mobile networks as a data source is the acquisition of data and the supporting platforms for the analysis and computation of acquired data. Population data estimation applies geo-fencing technology to push notifications. When a smartphone enters a geographical spatial scope, the location information of the smartphone can be acquired and logged. Since the size of a population is highly correlated to the number of smartphones, the current population size can be inferred by means of model simulation at different times and in different areas [ 22 ]. The volume of shared information in mobile networks is huge, which poses a great challenge in storing and analyzing this massive amount of data. Therefore, most of the research on mobile terminal location information focus on studying and predicting individual mobility patterns. Such analysis of individual mobility patterns, combined with the information from users’ social network applications, can be suitably applied to public affairs management, such as user profiling, service recommendations, and market predictions [ 23 ]. Communication data providers can also upload smartphone location data to cloud servers for real-time analysis and storage. Cloud servers can process large-scale location data, and the results of these cluster location data analyses can facilitate the development of more targeted emergency strategies in response to natural disasters. With the support of a large amount of data, the study of human activities on a larger scale will be of greater significance for urban and rural planning, population distribution, socio-economic indicators, and other aspects [ 24 , 25 ].

Data preprocessing

A magnitude 6.8 earthquake struck Luding County, Sichuan Province (at 29.59 degrees north latitude and 102.08 degrees east longitude) at 12:52 p.m. Beijing time on September 5, 2022, with a focal depth of 16 kilometers. The earthquake caused heavy casualties, with 46 deaths, and severe damage to water, electricity, transportation, and communication facilities and other infrastructure. The region was highly deformed due to crustal movements, and there have been other violent earthquakes. Since 1900, 21 earthquakes with a magnitude of 6.0 or greater have occurred within 200 km of the epicenter, and a 6.2-magnitude earthquake occurred 27 km from the epicenter in 1975. At the epicenter, we collected a total of 1,806,100 smartphone location records and 4,856 spatial grid locations within a range of 300 km from east to west and 220 km from north to south, and the collection scope covers areas of VI degree and above.

The analysis in this study used GIS data is obtained from the open-source data Open Street Map (OSM), which is available for free and can be downloaded from the portal website of Digital Crete ( https://www.openstreetmap.org/ ) The OSM data contains a series of data layers such as highways, railways, water systems, buildings, transportation facilities, etc. In this study, we only used data from residential points and areas, The DEM data adopts Copernicus DEM, which is a global open-source DEM data released by the European Space Agency(ESA) and can be downloaded from the ESA portal website ( https://panda.copernicus.eu/panda ). The DEM data of ESA has a 10 meter (EA-10) resolution for the European part and a 30 meter resolution for the global range. In this study, we used a 30 meter resolution in a raster format (Tif).The above data does not require authorization. The smartphone location records data obtained was authorized by the telecommunications company to the Zhejiang Earthquake Agency and provided to the author for use. The smartphone location records data used in this article uses Geohash encoding with an accuracy of Geohash7 (approximately 120m * 150m). This data is used to count the number of mobile devices in each Geohash grid within the earthquake zone range per minute.The data is group smartphone location data and does not involve the personal privacy of individual mobile phone users, so there is no concern about personal privacy leakage. All maps in this article are created using ARCGIS 10.6, The coordinates of the map are WGS84,The maps are oriented with North as up, and at this scale all maps in this article have an extent of 300km × 200km.The GIS data is obtained from open-source data websites and has been verified against the place name data of the epicenter area. All data does not involve copyright or legal disputes.

The epicenter of this earthquake was a mountainous area. The population density near the epicenter was not high, with only a few settlements and scenic areas for tourists. After the earthquake, the communication facilities in the disaster area were damaged, which led to a substantial reduction in mobile terminal connections. The disaster avoidance behavior of people and the rapid repair of mobile communication facilities also caused changes in the number of mobile terminals in the disaster area. In addition, there was also a huge quantitative difference between different time periods and areas. Therefore, we extracted the population information in different time periods. Specifically, starting from 10:00 a.m. on the 9th day, the data coverage was extracted every 1 hour, with 24 time periods in total. In this way, the dynamic damage condition of communication facilities in the earthquake area could be reflected in a relatively comprehensive way. In this work, due to the vast data size, accuracy and efficiency contradicted each other. Although the adoption of high precision could contribute to a more detailed representation of population distribution, the problem of high computational complexity would occur. Moreover, the terrain of the disaster area was complicated, and the shadowing effect of the mountains had a certain impact on the accurate positioning of smartphone locations [ 26 ]. Therefore, an excessive pursuit of accuracy would result in a certain amount of repeated calculation points and affect the calculation efficiency. In order to analyze the population in the earthquake area in a faster way, while balancing accuracy and efficiency, we selected a 150 m grid size to analyze terminal location distribution.

Population distribution simulation based on density analysis

Principle of kernel density analysis..

Within the 150 m grid, the population is not perfectly uniformly distributed. Thus, a mathematical approach is needed to simulate population density. We regard the center of the grid as a point. The value of the point is the size of population in the grid, and the distribution of population density is represented by calculating point density [ 27 ]. There are three commonly used methods for calculating point density: the quadrat density method, the kernel density method, and the Voronoi diagram density method. The quadrat density method randomly selects a number of quadrats in the space of the area being simulated and calculates the density of each quadrat by counting the number of individuals in each quadrat, with the average of the density of all quadrats as the density of the large area. However, random sampling is characterized by a certain degree of subjectivity, so the simulation results are relatively larger. This method is applicable to the sampling survey of a static population, but it has a poor effect in simulating a population with strong mobility and high density. The Voronoi diagram density method calculates a distance-based plane partition in geometric space by using data points as generators of a Voronoi diagram. There are n non-coincident seed points in the plane, and the plane is divided into n regions in such a way that the distance from a point in each region to the seed point in the region in which the point is located is closer than the distance from it to any seed point in any other region, and each region is called a Voronoi seed point region. Due to the abrupt density changes at cell junctions and the neglect of continuity in the occurrence of spatial phenomena, Voronoi diagrams also have certain limitations in population distribution estimation [ 28 , 29 ].

However, the above problems can be solved by the kernel density method. The value of kernel density gradually decreases with increasing center radiation distance, with consideration to the distance attenuation effect of the center point on its surrounding locations [ 30 ]. Conceptually, each point is covered with a smooth curved surface, and the surface value is highest at the location of that point. As the distance increases, the surface value decreases until the value turns to zero at a distance equal to the search radius. Each pixel value of the output raster is the sum of all surface values superimposed on the pixel. Thus, the kernel density estimation method can transform a point set into a surface that exhibits continuous density variation. It also is possible to transform a discrete set of points into a smooth density variation diagram, thus demonstrating their spatial distribution pattern. The higher the density value, the greater the aggregation extent of the point is. Kernel density analysis has obvious advantages in the simulation of population distribution, as population distribution is featured with clustering, and the farther away from the center, the less dense the population distribution is [ 31 ].

Spatial calculation method of kernel density.

https://doi.org/10.1371/journal.pone.0298236.g002

Population distribution simulation based on kernel density analysis.

We adopted the kernel density method and used the collected smartphone point data to simulate the population distribution within a range of 300 km from east to west and 220 km from north to south. Figs 3 and 4 show the simulated population distribution of Luding based on smartphone data. As shown in Fig 3 , the area near the epicenter is sparsely populated, with only a certain amount of population distributed in Moxi Town, Detou Town, and Dewei Township. In particular, the population density is very low in the west of the epicenter, with few large settlements located in the area. The population is concentrated in the eastern plain areas, which are far from the epicenter. Among these areas, Luding County is 40 km away from the epicenter, while Shimian County is 47 km away from the epicenter. The rest of the population is sporadically distributed along the bottom of the terrain ditches and the traffic lines. Fig 5 shows a heat map of population near the epicenter at 12:00 p.m. on September 5. It can be seen in Fig 5 that at 12:00 p.m., the area near the epicenter, where the government of Moxi Town is located, was densely populated, with only a few people sporadically distributed in the surrounding area. There were also few people in Detuo Town, and its neighboring township Dewei had a certain amount of population.

https://doi.org/10.1371/journal.pone.0298236.g003

https://doi.org/10.1371/journal.pone.0298236.g004

https://doi.org/10.1371/journal.pone.0298236.g005

To reflect the population change before and after the earthquake, we compared the population density at 13:00 pm after the earthquake on September 5 ( Fig 6 ) with the data at 12:00 pm before the earthquake ( Fig 5 ) and calculated the density difference between the 2 time periods so that the communication outage and the movement of people caused by the earthquake can be reflected more objectively. Figs 5 and 6 show the changes in smartphone density before and after the earthquake in the earthquake area and the epicentral area, respectively. As shown in Fig 6 , after the earthquake, the population density of highly seismic regions near the epicenter in the IX degree zone in the southern part of Luding County, such as Moxi Town, decreased significantly; as shown in Fig 7 , the population density of areas farther away from the epicenter (such as Hanyuan County) did not change much, and the overall density decrease and increase of Xingjing County in the VI degree zone were relatively in balance. Yucheng District in Ya’an City, the most densely populated area, in the V degree zone, experienced an increase in the number of smartphones turned on. In the earthquake area of Luding County, the population density in Moxi Town, Detou Town, Yanzigou Town, Dewei Town and its surrounding area, as well as Wanggangping Township, Caoke Township and its surrounding area in Shimian County dropped sharply, from which it could be inferred that a significant number of out-of-service base stations and power outages occurred in these areas. Along National Highway 318 from Luding County to Tianquan County, there was a certain increase in population density, and the population on the periphery of Luding County also increased to a certain extent, indicating that tourists had already begun to evacuate out of the scenic areas half an hour after the earthquake.

https://doi.org/10.1371/journal.pone.0298236.g006

https://doi.org/10.1371/journal.pone.0298236.g007

Results and discussion

Analysis of population change in the segmented zones of the disaster area.

After the earthquake, the population distribution in the disaster area displayed a dynamic change. Starting from the size of population affected by out-of-service base stations after the earthquake, to the repair of communication and electric lines, and then to the subsequent evacuation to the outside along the traffic lines, the number and location of the population had been changing.

We selected several typical settlements in the Ⅸ and VII degree zones for the time series analysis of communication terminal volume. Moxi Town, Caoke Township, and Detou Town in the Ⅸ degree zone were chosen, of which Moxi Town was closer to the epicenter and was also one of the settlements heavily struck by the earthquake. Fig 8 shows an analysis of the population data by area and time period within 24 hours since 12:00 pm on September 5. At 1:00 pm, communication and power systems were severely damaged after the earthquake, resulting in a sharp drop in communication terminal volume. However, the communication repair was made very quickly. After the emergency communication vehicles entered the earthquake area, communication was partially restored. As of 7:00 pm that night, part of the communication and power systems had been restored. The communication terminal volume began to grow slowly. At 06:00 am on September 6, the population evacuated from the scenic areas and the rescue forces entering the disaster area were superimposed. As shown in Fig 8 , the crowd flow in the three settlements showed a sharp increase; the situation in Caoke Township and Detou Township was basically similar, and the time series curves of the communication terminal volume in the two places were highly consistent.

https://doi.org/10.1371/journal.pone.0298236.g008

We also examined statistics on changes in Shimian County, Luding County, and Yidong Town in the VII degree zone ( Fig 9 ) within 24 hours. Because of the extremely uneven population distribution in this zone, in which Shimian Town and Luding County Town were relatively densely populated and the damage to communication facilities after the earthquake was less serious than that in the Ⅸ degree zone, the decline in communication terminal volume was relatively slower than that in the other zones. At 11:00 pm, the communication terminal volume showed a natural decline due to equipment shutdowns. Similarly, the population change in the three typical settlements in the VII degree zone is highly consistent. What is different from the Ⅸ degree zone is that there was a continuous outflow of people after the partial restoration of communication.

https://doi.org/10.1371/journal.pone.0298236.g009

Simulation of out-of-service base station situations in the disaster area

https://doi.org/10.1371/journal.pone.0298236.g010

https://doi.org/10.1371/journal.pone.0298236.g011

After the intensity map was officially released, we also calculated the changes in simulated population based on smartphones in different intensity zones before and after the earthquake, and in order to exclude the impact of smartphone shutdowns for lunch breaks on the out-of-service situation, we also calculated the changes in the smartphone-based population in the same time period and the same area on September 4. In the calculation of the final result, as the base, the data on September 4 was subtracted so that we could obtain the real out-of-service situation on September 5. Table 1 shows the statistics of the out-of-service rate based on smartphone population data. In Table 2 , it can be seen that the out-of-service rate in the VI degree zone is -1.62%, while the out-of-service rate on September 4 is 1.32%. The out-of-service rate decreased after the earthquake, which means that after the earthquake, some people in this zone were affected and kept learning about the disaster through their mobile terminals. By using the same method and after deduction of the base, we finally calculated that the out-of-service rate in the VII degree zone at 19.42%, which is contrary to the traditional view that service outages at communication base stations will occur only in the VIII degree zone; thus, it is inferred that there exists a certain out-of-service rate in the VII+ degree zone. In the VIII degree and IX degree zones, we believe that the number of people who were strongly affected by the earthquake and turned off their smartphones at 1:00 pm was small. Thus, the final results of the out-of-service rate in VIII degree and IX degree zones are 48.40% and 78.11%. The reduction in the number of smartphones after the earthquake may be affected by multiple factors, but an out-of-service base station is the most important one. Therefore, the above results can objectively reflect the base station out-of-service rate in different intensity zones.

https://doi.org/10.1371/journal.pone.0298236.t001

https://doi.org/10.1371/journal.pone.0298236.t002

With the same method, we tracked back to the population heat map data of earthquakes with a magnitude of 6.0 or above in mainland China since 2017 in the database, analyzed and processed the communication location big data in the zones of VII degree or above during multiple post-earthquake time periods in the course of earthquake emergency response, used the spatial change of terminal data to infer the damage of communication base stations and the response of people, and summarized the regression modeling so that we can spatially infer the distribution range of the hardest hit areas. Based on the change trend of terminal location data, the research proceeded with the extraction of seismic damage information to correct the empirical isoseismal line so as to make up for the lack of first-hand field data of the existing rapid assessment system. Finally, a rapid assessment method of the seismic influence field based on communication big data is established, and an operable software system is formed, which can provide technical support for earthquake emergency rescues. At present, the related research results have been popularized and applied in many regions.

https://doi.org/10.1371/journal.pone.0298236.g012

Conclusions

With good timeliness and high data accuracy, smartphone population heat map data can be applied to the actual practice of earthquake emergency response. Taking the Luding Ms 6.8 earthquake as an example, we have obtained the simulation data of the population distribution in the earthquake area in real time with the support of smartphone location data. Having been tested by dozens of large and small earthquakes, the data model has become more mature. Based on the cross-corroboration of multiple validation channels, we find that the data obtained at this time is consistent with the actual situation. In the hours after an earthquake, smartphone location data can provide strong support for the government’s disaster relief work when the actual disaster situation is still uncertain. The application of this method in the emergency management field is also widely recognized. Compared with traditional modes of disaster information acquisition, smartphone location data has obvious advantages in terms of cost, accuracy, efficiency, and other aspects. With high timeliness and good continuity of data, as well as the absence of additional investment in hardware equipment and organization of large-scale field investigations, this method can realize the rapid acquisition of the location change rule of communication terminal population in a disaster area after an earthquake. By analyzing the characteristics of spatial–temporal changes in smartphone location big data before and after the earthquake, we can infer the strength and distribution range of earthquake intensity. In addition, unlike the “black box period” in previous earthquakes, this work enables us to provide the government with a sufficient and reliable information basis for disaster relief within one hour after the earthquake, even when a more accurate and complete picture of the disaster situation is not available. This approach can greatly complement the shortcomings of the existing rapid seismic assessment systems and enables the rapid obtainment of highly credible disaster information in a timely manner when a sudden and destructive earthquake occurs ( Table 3 ).

https://doi.org/10.1371/journal.pone.0298236.t003

An out-of-service base station will cause a large amount of mobile terminal location data to disappear after an earthquake. During this earthquake, the magnitude reached 6.8, the maximum intensity of the epicenter was IX degree, and the hardest hit area experienced power interruptions and out-of-service base stations, which directly led to a cliff-like drop in the mobile terminal location data we obtained after the earthquake. The closer to the epicenter, the more obvious this phenomenon was. After the communication and power lines were repaired, the acquisition of smartphone location data began to recover gradually. Determining the base station out-of-service rate plays an important indicative role in estimating the seismic intensity of the hardest hit area in the first moments. This further helps us to obtain the seismic intensity, isoseismal line, and other critical information in the affected areas. With the support of communication big data, we can correct the empirical isoseismal lines based on the seismic damage information extracted from the model. In addition, we can also verify the locations that cannot be accurately assessed and dynamically modify the isoseismal lines to gradually improve the assessment accuracy. By combining empirical models and automatic computer processing technologies, we can obtain the earthquake disaster information in a faster manner. In the meantime, we still need to discover the characteristics of seismic damage in different regions and of different magnitudes, and on this basis, we can establish the basis of determining seismic intensity according to the performance of smartphone location data in different situations. This will ultimately further improve the method of obtaining the seismic intensity influence field based on mobile Internet data and enhance our earthquake response capability.

In practice, we have also found some shortcomings in using mobile terminals to judge disaster situations. Even in modern society, the utilization rate of smartphones is still closely related to factors such as age, geographic location and economic status. Large-intensity earthquakes usually occur in mountainous and sparsely populated areas. In the disaster area of this Luding earthquake, there was also an extreme situation where the population in the mountainous area in the northwest of the disaster area was extremely small and there was a clear gap in population composition, as most of the population were aged over 60 or children, with low smartphone usage rates, which may have led to underestimated location data. In contrast, Moxi Town and Detou Town near the epicenter were relatively densely populated. As Moxi Town is a scenic location, most of the population here were tourists, with a high smartphone usage rate, which may have led to overestimating the pre-earthquake data to a certain extent if a unified model was used. Therefore, in order to estimate and judge disaster situations more accurately, it is necessary to customize a suitable regional population model to take these differential factors into account. In the Luding Ms 6.8 earthquake, the research team applied different population models to adapt to the special conditions in Moxi Town and the northwestern mountainous areas. This has fully demonstrated the importance of flexibility and customization strategies. This approach can not only make up for the data bias caused by uneven smartphone usage rate but also help improve the accuracy of disaster assessment.

Smartphone-based location big data has great application potential in future earthquake emergency management. This technology can not only realize the real-time estimation of population distribution during an earthquake but also achieve in-depth analysis based on its rich attribute information, such as tracking the places of origin of the population, the population movement vector, traffic jams, etc., which could not be accomplished in past earthquake emergency responses. The mining of smartphone data can derive richer applications to better serve earthquake emergency responses. The application of mobile location big data sources can replace the traditional field investigation of seismic damages. We can calculate the spatial distribution of a disaster based on the disaster information reflected in big data so as to obtain the simulated results of disaster distribution, which will greatly improve the efficiency of seismic damage information acquisition.

Supporting information

S1 file. calculation on out-of-service rate based on mobile terminal location data in this earthquakes..

https://doi.org/10.1371/journal.pone.0298236.s001

• View Article
• Google Scholar
• PubMed/NCBI