Social Media Can Help Natural Disaster Reporting
By: Ubah Moallim and River Terrell
As we see a stark uptick in the number of natural disasters around the world, the grim data on worsening climate change impacts are ever concerning. Around the globe, forest fires, droughts, heat waves, cold fronts and hurricanes are shaking the bedrock of communities – sometimes irreparably. Damage to homes, roads and food production systems can spell doom for businesses and families picking up the pieces long after the initial catastrophe. Among the primary issues during environmental disturbances is the lack of accurate information coverage that stems from inadequate or incomplete reporting. Images especially are hard to come by, and local journalists often risk their personal safety to capture a newsworthy shot.
Without detailed information and comprehensive images of affected areas, engineers and other on-site specialists have a tough time gauging what’s been damaged or even where. Losses are exponentially compounded as information gathering is slow and inconsistent, tainted by fear and panic, delaying preventative procedures in the face of uncertainty. Overall, an unfortunate cycle of information deficits is cemented as the disaster unfolds: there are gaps in what is known; panicked and often incomplete information is provided, leaving more questions about the validity and usefulness of the fragmented information shared online. In the chaos of the disaster, this cycle becomes increasingly difficult to overcome.
Luckily, there is a silver lining: an emerging program offers potentially better outcomes, collecting and corroborating data to aid in disaster response. In a project funded by the National Science Foundation, researchers at the Computational Media Lab at UT Austin, led by Dr. Dhiraj Murthy, have devised a plausible mechanism that could vastly improve natural disaster responses and reduce the long-term consequences of future severe weather using data collected by social media. This program leverages reporting done during natural disasters by both known professional journalists and citizens. The CML demonstrated the potential of this new system by using Twitter posts and images about Hurricane Florence, which made landfall in North Carolina on September 14, 2018.
The hurricane is an apt example of the kind of destruction to infrastructure that can occur, as residents saw over 10,000 homes affected and $24 billion racked up in damage. During the storm, there was increased activity across various social media platforms as citizens and journalists alike tried to provide important details to reconnaissance specialists. However, the well-intentioned collective efforts of community action can sometimes be more of a headache than a help, adding more voices to the shouts of panic rather than providing clear direction. The CML’s program focuses on providing real-time and operational data that will employ the skills of qualified local journalists and citizens to observe and report on the complexities of natural disasters.
How it works:
The data collection, interpretation, and analyses start off as soon as the presence of the natural disaster is confirmed to proactively set up the framework for how coverage will then unfold and develop. The data itself is taken from the social media platform Twitter. An algorithm that was developed using image-based machine learning automates a way to digitally process damage. The system sifts through the highlighted and relevant mentions of the disaster, classifying images and sources. Keywords and hashtags are used to expedite the process. To ensure that the information is free of any potential corruption or inaccuracies, a preliminary process of elimination takes place. During this procedure, duplicate images are removed from the bank of data and a manual visual inspection of all the sources and images on Twitter is conducted. Once this is completed, the extracted images are used to create the first model of images. Additionally, a verification step happens where the biographies of accounts posting selected tweets are scrutinized, both for chosen keywords that identify the source as likely reputable and for their location. The location aspect ensures that any further information is posted within close proximity, so that updates are not stale by the time reconnaissance and rescue groups make it to the scene. The approved Twitter contributions then go through by a manual confirmation, and the critical information is distributed to local journalists who are actively reporting on the disaster. Ultimately, this program provides a real-time means of gathering accurate information during a disaster, making it easier for those on the ground – journalists, emergency relief workers, and citizen volunteers – to make decisions that will save lives.
The sheer influx of comments, posts, images, alerts, and the general background noise of social media (especially during a crisis) makes it hard to determine what information is reliable. Until now, first responders and engineers had no choice but to live with the reality that precious information (that could lead to a home recovered or a life saved) is buried under the clatter of the social and digital media ecosphere. The CML’s new method combines Twitter, journalist content, and computational tools to quickly gather and analyze relevant information and prioritize and disperse the most appropriate resources where most needed using artificial intelligence to accurately identify and label relevant pictures that are uploaded with location information along with contextual facts from the ground can help first responders in real time and later engineers who rebuild the infrastructure. This study forges a new path (combining the power of computer and human networks) to effectively cut through the noise of conflicting sources in real or near real time to preserve human life, property, and direct resources is not only possible, but effective and cost-efficient.
Collected data would subsequently undergo prioritization processes based on factors that are related to perceived relevance and then appropriately delivered to the key personnel on the ground. In addition, the voices deemed reliable by verification will be elevated and ordinary citizens will be empowered to act as journalistic authorities who are crucial to the recovery procedures. This type of work is called “citizen science”, and it refers to the kind of crowdsourcing that civically minded people engage in to improve turbulent conditions. Communities rally to provide real, quantifiable assistance to professionals and to supplement their knowledge with information that proves itself to be invaluable. Wide-reaching resource networks are more efficient, considering the prevalence of smartphones. As we’ve seen in the last few years, anyone with a device and a connection can record, videotape and project to a global audience, and sometimes, these citizen-captured footage is the only available content on disasters. Wide resource networks will help organizations and reconnaissance teams that focus on initial relief, helping overcome the expensive tasks of recruiting mobile workers and locating the real-time geographical locations when extreme weather occurs. These groups must make high-stakes judgment calls and decide which area is worth trying to recover, a decision made more complex by the high amounts of uncertainty surrounding it. When these one-or-the-other calls are made, some people suffer more than others, and particularly those in remote or rural areas are forced to fend for themselves. The rural nature of the impacted regions in North Carolina already meant a lack of infrastructure capable of withstanding a hurricane. Couple this with the generally low socioeconomic status of the local residents and there is a bleak prognosis. This is further exacerbated by the fact that the agricultural systems are usually constructed with less safety standards than more densely populated areas – according to a 2017 paper by the American Society of Civil Engineers on the associated criteria for buildings and their minimum design loads.
In an attempt to combat the costly endeavor of reconnaissance, the easily accessible collection of social-media information will make first responders out of those closest to a disaster. This system creates a map of knowledge that would otherwise not be available or usable, bringing eyewitness accounts and first-hand experiences documenting damage to the forefront of emergency response. This way, the engineers and relief workers in the thick of the disaster can properly pinpoint the kind of work needed to be done and what areas are in the most need, making them better able to restore neighborhoods and commercial areas back to safety and functionality. It paves the way for scores of individuals to sidestep the devastating realities that often accompany having to relocate with little to no personal belongings or financial resources to their name since people can return to their residences more quickly the issue of finding suitable shelter is eliminated. Community spaces can then be used more efficiently as stations for distributing essential services like water and food. Furthermore, the social media platforms can act as useful back channels that allow individuals to communicate personally or to a larger audience when phone lines and the internet have been felled by natural disasters.
An often-underappreciated perspective also looks at the social and mental wellbeing of those who are struggling to cope with these catastrophes. Studies suggest that the overall improvement of an individual’s levels of anxiety and that of the impacted society as a whole is very attainable. As posts are shared the knowledge bubble increases and informed citizens can relieve the anxiety of uncertainty in chaos. Also, the tendency of social media to relieve stress and provide humor like memes in the face and aftermath of severe loss can demonstrably lend itself to the recuperation and mental as well as physical fortitude of countless people. Even when online content pertains to the disaster, having readily available images that provide an accurate framework for residents to perceive the scale and seriousness of the situation at hand is helpful. Communities can empower themselves through non-traditional means, using information that might not have been otherwise utilized to care for and rebuild their communities.
Professional journalists can aid in these processes by encouraging the active participation of those affected by disasters. Extending the wide net of their expertise and credibility might even help in establishing reputations of transparency. The other – though less tangible – benefit is that alleviating the disinformation and misinformation storm that is brought about without available reliable information, is alleviating the lack of trust in existing reporting structures. This tool could potentially transform the nature of disaster reporting to be more proactive than reactive, collecting and presenting real-time data from those in the midst of a disaster to aid workers and professional journalists. Disaster preparedness researchers and developers can use relevant information to formulate more timely, cost-effective, and efficient methods of saving lives and helping communities.
As climate change worsens and unpredictable weather patterns persist, we must find new ways to protect our communities from the consequences of natural disasters. Though just a start, this program can begin the process of developing new tools, taking advantage of citizen journalism to respond to the natural disasters that we will surely continue to face.