The data has been collected through open sources and cross-referenced using a wide range of information. These include local and international news agencies and media reports; social media accounts, including Twitter, Facebook, YouTube, and other video footage, and WhatsApp; reports from international and national NGOs; official records from local authorities; and reports by international human rights groups. Where independent reporting is not available, the data has been cross-referenced with sources from opposing sides to the conflict as to ensure the reporting is as accurate and impartial as possible.
The dataset lists target category and subcategory for each incident, apart from incidents where no information on the target is available. The target category chosen for each incident refers to the original use of the target, e.g. a school hit by an airstrike is referred to as a school building, with no further assessment on its use at the time of the airstrike or the circumstances that led to the airstrike.
As the data is collected using open sources and as there is a general lack of transparency from the parties to the conflict, as well as a lack of independent reporting on the ground, the data presented has been verified and cross-referenced to the extent possible. It is important to note, however, that due to the context of the current situation in Yemen there are significant challenges in accessing independent sources for verification of incidents. The data presents our best current understanding of the incidents included and are reported in good faith.
The data has been collected with the purpose of creating a statistical overview of the impact of the aerial bombardments in Yemen, especially in regard to geographical spread and its targeting.
In YDP's data, an air raid refers to one incident. One air raid incident includes all air strikes on a single location within approximately one hour and therefore may comprise multiple airstrikes. Air strikes per air raid can vary greatly from a couple to several dozen per air raid. Due to the variations in quality of information on air strike counts and the challenges of verifying each individual strike on a single target or location, an air raid in YDP data, therefore, presents the most conservative estimate.
The required minimum for a recorded air raid incident is one airstrike. The maximum is the unverified number of air strikes, which is also noted in the details of the data. This unverified number of air strikes ranges from minimum airstrikes: 19,511 to maximum airstrikes: 45,563 for the first four years of the air campaign.
Casualty figures are a best estimate of numbers and are presented in good faith. YDP recognizes the limitations in the accuracy of the figures. Open source reporting of casualty numbers is often scarce or biased. Locations of bombings can be in very remote areas, or with limited access to multiple or independent sources that results in open source reporting of significantly different casualty numbers for the same incident. The process of the civilian casualty data collation includes in-depth research into incidents with civilian casualties, reviewing video material, published lists of victim names and cross-referencing with human rights groups and other entities collecting data on civilian casualties.
The research process is as meticulous as possible, and to remove bias or inflation in casualty numbers, the lowest reported casualty number is recorded in the data, unless verified numbers from human rights groups on the ground are available, in which case they are used. This means that the civilian casualty data we present is the least civilian casualties reported from airstrikes.
As part of our work to enhance conflict monitoring and data sharing, YDP is collaborating with the Armed Conflict Location and Event Data Project on collecting and collating countrywide political violence data that includes the wider conflict and ground war. ACLED’s methodology for political violence and protest events data collection differs from the process used for YDP’s separate air raids data collection.
For more details on the methodology used for political violence, please see ACLED’s detailed breakdown of methodology and a background paper explaining how YDP and ACLED collaborates on data collection .