How AI and drones fight wildfires around the world

The next wildfires will be fought with the aid of technology, as artificial intelligence and internet-of-things devices help prevent, detect, fight fires, to ultimately assist in keeping people safe.

Already, Natural Resources Canada shares vital satellite imagery with provincial teams to support firefighting efforts. San Diego Gas & Electric puts drones and AI to work inspecting their power lines for fire hazards. Australian company, Exci, uses ground-based cameras and satellite imagery to automatically detect bushfires. BlackBerry AtHoc, meanwhile, helps coordinate disaster response, providing real-time communication, coordinated information to first responders.

All of these systems have something in common, they run on the Amazon Web Services (AWS) cloud.

Firefighters need access to high-quality, up-to-date information to assess wildfires and determine how to get them under control. With a landmass as vast as Canada, this is no easy task.

Natural Resources Canada (NRCan) assists by collecting earth observational satellite imagery from the Sentinel-2 satellite, and creating the Canadian Wildland Fire Information System. The satellite dataset is massive and would be difficult to transfer to the provinces without the help of the AWS Open Data Registry, allows NRCan to store and make the wildfire dataset publicly available at no cost. This way, provinces, like the British Columbia government, can easily and quickly pull down the satellite imagery at no cost and create a daily mosaic map to support fire teams.

San Diego Gas & Electric serves 3.6 million customers in Southern California, an area prone to wildfires due to its dry climate, strong winds, and abundant vegetation. The utility maintains about 240,000 power poles that carry electricity across its service area. Faulty or damaged equipment can spark fires if not repaired in time.

In 2019, the utility began flying drones over 75,000 power poles at highest risk of wildfire. The technology would need to not just take pictures, but identify high-priority problems requiring immediate repair.

They built an elegant solution with a team from Accenture (an AWS Partner) that combines the considerable experience of the utility’s qualified electrical workers, with the bird’s-eye view captured by the drones, and powerful machine-learning models running in the AWS cloud.

These machine-learning models are trained on millions of images labeled by the utility’s experts. They analyze images taken by the drones, and flag anomalies or defects that could pose a fire hazard. The human experts review the output and provide feedback to improve its accuracy.

The models use Amazon SageMaker, a service that makes it easy to build, train, and deploy machine-learning models at scale.

The drone program has helped San Diego Gas & Electric reduce inspection time by 40%, increase inspection coverage by 55%, and improve inspection quality by 15%. It allows the utility to prioritize repairs based on risk, optimizing its maintenance resources.

Most importantly, it has helped the utility enhance its wildfire prevention efforts and protect customers and communities.

Exci, an Australian company, uses AI and machine learning to detect bushfires automatically within minutes of ignition. Bushfires are a common occurrence in Australia, especially during the summer months. They can spread rapidly and unpredictably, threatening lives and property. Early detection is crucial for effective fire suppression and prevention.

The Exci system acquires imagery from ground-based cameras and sensor data from satellites. A sophisticated AI processes the data, searching for signatures of fires such as smoke and heat. If it detects one, it reports it to first responders or property owners.

The system automatically detects even small fires within minutes with a near-zero rate of false positives.

Exci runs its system on AWS using services such as Amazon EC2 for compute power, Amazon S3 for data storage, Amazon Kinesis for data streaming, Amazon Rekognition for image analysis, and Amazon Comprehend for natural language processing. These resources scale up or down to meet demand.

Exci has deployed its system across 30 million acres of Australia and in the United States so far.

Emergency responders now face data flowing in from sensors, drones, and mobile devices. Cloud-based technology helps them make sense of it, revealing big-picture trends and safety risks. And they assist in managing resources by streamlining communication and collaboration.

One example is BlackBerry AtHoc, a critical event management platform solution that helps first responders and citizens strategize and communicate during emergencies.

Operating in the cloud, this system can be deployed locally in a single day to provide timely, accurate, and confirmed information that can save lives during fast-moving events.

With a single alert, a command center quickly and easily notifies responders and partners of the exact location, the type of incident, hazards present, access to the scene, the number and severity of casualties, and which emergency services are on the scene. Recipients can confirm they received the message with a single click.

It can instantly alert the public within a geo-fenced area to shelter in place or evacuate immediately. This addresses the problem of so many channels of information — news sites, social media, different communications equipment by responders from different jurisdictions — to avoid silos and bottlenecks.

BlackBerry AtHoc runs on AWS as it provides reliability, scalability and agility to adapt to whatever the situation in front of them.

California Department of Forestry and Fire Protection used another such emergency management system also running on the AWS cloud, this one from ArcGIS, to help manage the 2018 Camp Fire, the deadliest and most destructive wildfire in the state’s history.

On top of offering the computing firepower required to analyze the massive datastream from firefighting drones and mobile-equipped first responders, the AWS cloud lowers carbon footprints by nearly 80% versus servers housed separately in private customer’s Enterprise Data-Centers.

That will grow to 96% by 2025, as we close in on our target for Amazon to be powered 100% by renewable energy. And that helps address climate change, the long-term cause of more and more wildfires.