Last December, roughly 190 nations took an unprecedented step – they approved a number of United Nations measures aimed at trying to halt the loss of our world’s biodiversity. Among the concrete actions officials adopted was the pledge to protect nearly a third of Earth’s land and ocean by 2030. To hit this 30x30 target as it is prominently known as, nations need the tools and mechanisms required to effectively and efficiently enforce these wildlife refuges. Despite the urgent need to protect these critical areas namely the ocean, the global community has a long way to go. According to The Marine Protection Atlas, currently 10% of the ocean is under protection, but only 3% of that number is strongly protected.
There are many levers governments and non-governmental organizations are currently utilizing to realize these ambitious targets. Policy shifts, new legal frameworks, and more resources earmarked for enforcement operations are a few, not least technology. This is where Skylight is playing a unique role. As more data becomes available, the platform is advancing AI approaches to accelerate the quality and shorten the time to deliver actionable analysis for enforcement and deterrence – a key in addressing illegal, unreported, and unregulated (IUU) fishing
The latest publicly available satellite data Skylight is making available globally, daily, and with minimal delay for all users is Night Lights. Night Lights data from the Visible Infrared Imaging Radiometer Suite (VIIRS) “Day Night Band” can detect light emissions – weather permitting – from vessels that use light to attract catch like tuna or squid at nighttime. This is a particularly important development in the fight against IUU fishing as a significant proportion of the vessels plaguing our seas are doing it at night.
Night Lights have several benefits that will further empower Skylight users. Most notably, every location is imaged at least once a day. This is a big advantage for areas especially many of the small island developing nations that aren’t regularly covered by other satellites. Night Lights collect in the middle of the night – between 1:00 a.m. and 4:00 a.m. – filling an important gap when other remote sensing methods are not gathering data. Combining these data with environmental data can be additionally helpful for tracking fishing activities. For example, maritime analysts may utilize Night Lights data in Skylight and compare it against the lunar phases to better understand the fishing happening in their waters and plan enforcement activities, if they are needed.
The Colorado School of Mines has made vessel detections from Night Lights data available to communities for many years. Organizations like Global Fishing Watch have also enabled publicizing this data for the global maritime and conservation community. Skylight’s focus is on making Night Lights more relevant to operationally minded users who need information as quickly as possible to take action. This includes processing the Night Lights data for vessels, checking if they are transmitting automatic identification systems (AIS), and delivering the data – as quickly as within two hours.
As promising as the data coming from Night Lights are, there were challenges. Firstly, Night Lights detections come from weather satellites that weren’t designed to detect lights on ships. The VIIRS sensor is mainly geared toward providing data that organizations like NOAA and the National Weather Service could use to pinpoint hurricanes, snow cover, and wildfires, data from Night Lights by itself, is insufficient to classify vessels. The images collected are very low resolution and look like a series of pixels making it hard for analysts to determine vessel details and lack location precision. And simply detecting collections of these bright pixels on a dark background would lead to a number of false positives. That’s because there are many sources of lights at sea causing interference. These range from those naturally occurring such as moonlit clouds and lightning to others artificial in nature like oil platforms and light pollution from coastal cities.
While these were obstacles, Skylight’s engineers have been able to operationalize Night Lights for those working to protect and monitor the ocean. To do so, they built seven different computer vision models that all work in parallel and feed into one model. The result is the Night Lights vessel detection model that’s able to filter out most of the “noise,” or false positive detections.
There are certainly limitations to the type of information Night Lights can provide for those tackling IUU fishing. But when Night Lights are combined with the platform’s existing AI applications, imagery sources, and types, the level of quality and assurance in Skylight’s information greatly increases. For example, where Night Lights struggles, chiefly moonlit clouds, the platform’s computer vision for satellite-based Synthetic Aperture Radar (SAR) imagery – a source of data that can penetrate clouds day or night – can close the information void. This sort of data quality assurance is precisely what maritime analysts and marine protected area (MPA) managers need to help make data-driven decisions that are critical to delivering on the promise of MPAs as a way for restoring the health of our ocean.
For more technical information on Night Lights, read more about it on Skylight's Knowledge Base here.