The report that came to the attention of the United States military in April 2017 relayed devastating news from Iraq: More than 30 people, among them women and children, had been killed when aircraft from the American-led coalition fighting the Islamic State in Mosul struck a neighborhood known as Siha.
A civilian casualty cell of the U.S. military, which was charged with assessing reports of civilians killed or wounded in coalition operations, learned of the claim in a Facebook post published on April 11 by the news outlet, the Iraqi Spring Media Center.
The Pentagon began an inquiry, but only a week later its assessment officers couldn’t confirm whether coalition aircraft had targeted that location, and they dismissed the claim, saying Siha was not among “known districts of West Mosul.” There would be no further review.
But Siha wasn’t hard to find.
Reporters from The New York Times were able to locate the west Mosul neighborhood using just Google Maps. The name appeared slightly different, as “Sihah” instead of “Siha,” a spelling variation that is common when Arabic words are written in English.
Additionally, a simple Google search revealed several news reports published before April 2017, verifying the existence of Siha and its approximate location.
An analysis of confidential Pentagon documents by The Times’s Visual Investigations unit found that a number of allegations of civilian casualties had been dismissed as “noncredible” based on flawed reviews of evidence — oversights that Times reporters were able to detect using resources widely available to the public. That included websites like Google Maps and Wikimapia, a crowdsourced mapping platform. Typically, U.S. military assessors have access to far more robust resources, such as strike logs and video feeds of airstrikes.
“I’ll tell you what it is: That’s negligence,” said Marc Garlasco, a former Pentagon senior intelligence analyst. “That is plain and simple. It is the most basic level of investigation that they should be doing, and not to do it is completely negligent.”
The Times obtained more than 1,300 confidential Pentagon assessments of allegations of civilian casualties in the American-led air war in the Middle East, between September 2014 and January 2018, during the height of the war against the Islamic State. Based on those documents, The Times recently reported patterns of failed intelligence, decision-making and execution behind deadly airstrikes.
These documents detail the criteria and rationales for how the Pentagon chose to classify civilian casualty allegations as either credible or noncredible.
A vast majority of assessments — more than 1,100 — were deemed noncredible. In some cases, there was not enough information for reviewers to search for airstrikes that might coincide with allegations or to conclude that civilian casualties occurred as a result of a coalition strike. However, The Times had found that many allegations of civilian casualties were erroneously dismissed for reasons ranging from insufficient quality and quantity of video to the inability to determine which of many strikes in an area was the subject of a claim.
This investigation focuses on reviewers’ inability to establish details about the locations of strikes. In reviewing 80 assessments, including those with high numbers of reported civilian casualties, The Times repeatedly found what appeared to be simple mistakes. In a dozen instances, Pentagon assessors said that a location could not be identified, even though it was easily found on the internet, or they seemed to have just looked in the wrong place.
Following recent revelations in The Times about botched strikes by U.S. forces, the Pentagon has said that it is committed to investigating its mistakes. But this examination raises further questions about the capability, or willingness, of the U.S. military to accurately count civilian casualties from its air war.
“The entire effort was really about responding to reports of civilian casualties in public and getting ahead of the narrative,” said Daniel Mahanty, one of the lead authors of a 2020 report on how the United States assesses civilian casualty claims, and a former State Department official. “It was certainly not about doing anything differently to prevent harm as the operation proceeded.
Captain Bill Urban, spokesman for the U.S. Central Command, said the military “applies a policy of reviewing and assessing all reports of civilian harm, irrespective of their source, and where the information available warrants, conducting investigations, applying critical lessons learned, and acknowledging the civilian harm caused by our actions.” He did not comment on the specific errors identified by The Times’s investigation.
While the Pentagon reviews all allegations of civilian casualties, interviews with experts and current and former military personnel revealed systemic problems, including a lack of training, inaccurate airstrike logs and an overworked, rotating assessment team of usually only a few people.
There were significant inconsistencies in the quality of the assessments, pointing to a process whose success relied more on the skills and commitment of individual officers than on cohesive standards and methods.
The assessment process
A review into civilian casualty allegations can be prompted in several ways, including reports from the local news media and social media posts monitored by American military personnel. Most incidents are flagged by Airwars, a nonprofit watchdog group based in Britain that collects reports and provides summaries from local sources.
The allegations are typically sent to the civilian casualty cell, whose members receive no standardized training to become assessors.
“Each brings to the task his/her own unique skills, and each are being constantly required to hone those skills over time,” said Captain Urban, the military spokesman.
One former assessment officer, who asked to remain anonymous out of concern for retribution by the U.S. military, said that he had never heard of the civilian casualty cell until receiving the assignment and that many of the skills required to review casualty claims had to be learned on the job.
The Pentagon’s assessment process leaves a paper trail of confidential documents, including “initial assessment forms,” filled out by the civilian casualty cell to determine if a claim warrants further investigation. Such inquiries are often conducted by the command that carried out the strike.
Assessors answer a series of yes or no questions to establish details about an allegation, including location and timing.
In assessing the 2017 claim from Siha in Mosul, reviewers did not appear to have consulted basic resources like Google Maps. In other assessments, they made significant efforts to review claims, using a wide variety of tools to review information.
When one allegation stated that a swimming pool had been targeted in Syria, assessment experts used satellite imagery to identify “all pools within Raqqa” to search for nearby strikes, showing a thoroughness that wasn’t applied across the board.
If enough information is gathered, the final step of the initial assessment process is to check military records for airstrikes that may have hit the approximate location noted in an allegation. But the former assessment officer said those records could be inaccurate, making it difficult to pinpoint strikes.
Those concerns were confirmed by The Times’s own ground reporting, which found many instances in which the logged coordinates for airstrikes were more than 500 yards from the actual site of impact. One was as far as five miles away.
Some officers noted these inaccuracies in the assessments, with one writing that the logs “shouldn’t be used to identify strikes.” Instead, the officer recommended searching reports by aircrews after missions — a cumbersome and rare practice.
Captain Urban said that strike logs had become more accurate, but he did not provide specific details about the improvements or whether previously dismissed allegations had been reassessed using improved logs.
Sometimes, Pentagon assessors simply misread the allegations, leading to the dismissal of a report.
In an assessment involving Hit, Iraq, Airwars and local sources said that an airstrike had killed two people and wounded three in “Al-Bab Al-Gharbi,” which translates to “the Western Gate” and describes what the area is: the western entrance to the historical center of the city.
While the sources clearly referred to “Al-Bab Al-Gharbi” as a single neighborhood, the Pentagon reviewers incorrectly looked for “Al-Bab and Al-Gharbi,” as if they were two distinct areas.
The assessment document shows that the military did carry out strikes in Hit that day, but the claim was dismissed because assessors were unable to find the location.
Captain Urban said the Pentagon could not provide any insight into how this allegation — or any of the others reviewed by The Times — had been evaluated because the assessors “have moved on to new assignments.”
Inadequate Arabic skills
Documents show that even when precise information about the location of a reported strike was available, reviewers sometimes missed it because of a lack of Arabic skills.
In one assessment, the Pentagon deemed as noncredible an allegation that eight people had been killed, including four children, in part because it could not locate the Jerri neighborhood, also in Hit, Iraq.
The Jerri neighborhood can easily be found on Wikimapia — but only if searched for in Arabic.
Although assessors conducted some searches in Arabic, they did not do so routinely. Multiple people who worked on or with the civilian casualty cell told The Times that speaking or reading Arabic was not a requirement.
While Captain Urban said interpreters were available to assessors “where language skills are needed,” the review of documents suggested there were still oversights when it came to Arabic comprehension.
Mixing up locations
In some assessments, the Pentagon simply confused towns with the same or similar names and dismissed the claims, the documents show, as happened with a reported airstrike on a Syrian town in March 2017.
Several social media posts said that the strike had hit a neighborhood in Maskana, part of Aleppo Province in Syria, killing at least eight people. An internal Pentagon team flagged the claim for further review.
The documents show that assessors zeroed in on Maskana, but it was the wrong one. There is a town with the same name in Homs, a different province of Syria. The reviewers were unable to find correlating airstrikes, and the allegation was dismissed.
A few weeks later, military personnel dismissed another claim because they appeared to have mixed up two towns. According to a tweet from a Syrian news outlet, the U.S.-led coalition bombed the village of Sabha in Deir al-Zour Province, killing or wounding about 50 people. Again, an internal group at the Pentagon alerted the assessment team.
Analysts reviewing the allegation looked for a village called Sabha in Deir al-Zour Province. They found one, and stopped there.
But there is another town with the same name close by. That town matches the nearby location of the reported strike described by a local resident in a news story.
In its dismissal of the allegation, the Pentagon said that “the nearest strikes […] were 17 km away” from the Sabha the reviewers had focused on.
The possible strike location identified by Times reporters was almost exactly 17 kilometers away.
The Pentagon’s 2018 procedures for assessing civilian harm prompts analysts to “narrow the date/time/location of the allegation using photo/video evidence.” But in multiple civilian casualty assessments, this wasn’t done, a shortcoming that resulted in assessment officers’ missing important pieces of evidence.
That’s what happened in the assessment of the Sanjari family in January 2017. Friends and relatives had gathered at the family’s home in Mosul to mourn the passing of Aziz Ahmed Aziz Sanjari, a retired Iraqi Army colonel. An explosion ripped through the gathering, killing civilians, according to initial social media and news reports.
Airwars sent the claim to the Pentagon and said that the attack had taken place at a funeral. It included a link to a video from the Amaq News Agency, a news outlet linked to the Islamic State.
Pentagon assessors appear to have wrongly assumed the strike happened at a cemetery in an area about a half-mile from the Sanjari house. In their dismissal, they wrote that “no strikes were found within 100 m of the cemetery boundaries.”
The analysts also reported that they were “unable to access” the video link that Airwars had included in the claim. Whether the link was accessible at the time is unclear, but the video was posted elsewhere online the day of the strike and was easy to find when the Pentagon did its assessment. And it’s still online to this day — a search that took five minutes on Twitter, using the term “Mosul” in Arabic and the date of the strike.
In not reviewing the Amaq video, the Pentagon missed a key piece of evidence showing that the airstrike had hit just outside a civilian home, not at a cemetery.
Interviews with the family in June 2021 and death certificates confirmed that 11 people had been killed, including an unidentified woman and a girl who were walking near the home.
Ridhwan Ahmed Aziz Sanjari, who lost two of his brothers and his cousin in the airstrike, told The Times, “I just wanted to know why.”
Jeff Parrott, Hiba Yazbek, Abbie Cheeseman and Leila Barghouty contributed research. Momen Muhanned contributed translation. Drew Jordan and Michael Beswetherick contributed production.