Deep Dive: Impact AI on Marine Ecology. Part 1.

Over the years, AI and oceanography have created a powerful synthesis that can shed light on many important issues related to protecting and exploring the earth’s water surfaces. The emerging challenges require more efficient, accurate, and scalable solutions for marine ecology.

Machine learning and AI offer new insights through which we can explore the oceans, reduce water pollution, monitor marine biodiversity, or predict tides. All of this remains promising and offers many new and effective solutions.?

According to the study, ocean industries such as fishing, shipping, and energy generation generate at least $1.5 trillion in economic activity each year and support 31 million jobs.

Problems faced by traditional methods

Traditional methods of researching the components of the ocean are gradually exhausting themselves, due to which they currently face a considerable number of problems, such as:

The Danger of the Sea Depths For the Human Body

Oceanographers and marine biologists regularly face many problems, most of which are related to the danger to the human body. The deep oceans are inaccessible, devoid of light, and subject to extreme pressure, directly affecting research and making it risky and resource-intensive. The tasks faced by traditional methods require huge resources and time, and they are dangerous for people. Also, many changes and phenomena are invisible to the human eye. According to the National Oceanic and Atmospheric Administration (NOAA), this capability is key because more than 80% of Earth’s oceans remain uncharted, uncharted, and unexplored

A Large Amount of Data

Artificial Intelligence becomes crucial in processing a large amount of data, recognizing patterns, or tasks that require large resources. Data captured by tag-and-track devices, underwater cameras, and remote sensing technologies can be easily analyzed using Artificial Intelligence. This answers many questions such as migration patterns, predator-prey relationships and responses to environmental stressors.“A big ocean needs big data. Researchers are collecting large quantities of visual data to observe life in the ocean. How can we possibly process all this information without automation? Machine learning provides an exciting pathway forward – MBARI Principal Engineer Kakani Katija?

The Scale of the Oceans and the Variability of Their Ecosystems

The scale of the world’s oceans, and their variability, require a complex interpretation of complex biological, chemical, and physical interactions. Because of this, with traditional processing, the data is often incomplete and ambiguous.

Spatial and Temporal Resolution of the Data

Traditional methods such as diver surveys and boat sampling require considerable costs, time and resources but are still limited to small local areas. Therefore, a comprehensive understanding of marine ecosystems requires large-scale models that can be provided by Artificial Intelligence. In addition, the rapid pace of changes in the environment can make it difficult to interpret data manually.

How can AI help tackle plastic in the ocean and its Impact on Marine Ecology

Plastic waste in the ocean is an issue that is becoming more and more acute every year. Negative effects on the reproductive system of animals and fish, the destruction of ocean fauna, the disappearance of rivers due to pollution, the emission of chemicals that penetrate every particle of our planet – this is far from a complete list, this is only its beginning.

The world produces more than 430 million tons of plastic per year. Two-thirds become waste after the first use. According to the Ocean Conservatory, eleven million metric tons of plastic enter the ocean each year, on top of the approximately 200 million metric tons already flowing through our marine environment.

Nicola Simpson, head of the United Nations Development Programme’s Barbados and Eastern Caribbean Blue Economy Promotion Lab, says that at these production rates, there will be more plastic than fish in the ocean by mid-century.

Therefore, this is the #1 question facing our generation. We need to take all the necessary measures and direct all efforts to save the ocean, correct our mistakes and leave to our descendants the unique nature that we have been given.

In this matter, the technologies of Artificial Intelligence and Machine Learning have considerable potential. And now, we’ll tell you how it can help tackle the problem of ocean plastic pollution.

Automatic Analysis Using Satellite Images

Researchers from Wageningen University and EPFL published a study in Cell iScience in 2023 in which they developed an AI-based detector that estimates the probability of marine debris shown in satellite images. This technology can help get rid of plastic waste by removing it with the help of ships. The European Space Agency’s Sentinel-2 satellite records the accumulation of waste in the seaThe images are freely available and require analysis by Artificial Intelligence and deep neural networks, as they contain terabytes of data.

The founders of the study are Davis Tuya, associate professor at EPFL and director of the Computational Science for Environmental and Earth Observation (ECEO) Laboratory in Zion, Mark Rooswurm, associate professor at Wageningen University and former researcher at EPFL, and Sushen Jilla Venkatesa Environmental Computational Science and Earth Observation (ECEO) Laboratory.

Mark Rooswurm says about this technology,“These models are trained on examples provided by oceanographers and remote sensing specialists who have visually identified several thousand marine debris in satellite images around the world. In this way, they “taught” the model to recognize plastic waste.

Every 2-5 days, Sentinel-2 captures images of coastal areas around the world. The detector, based on Artificial Intelligence, evaluates each pixel in satellite images to detect debris. It learns according to the principles of artificial intelligence. In this way, it works on data to make the best use of limited training data. The detector uses Computer Vision algorithms to recognize objects visible on the screen and provide this data to oceanographers and remote sensing experts.

The detection model can also spot debris in daily PlanetScope images taken from cubic nanosatellites.

“Combining weekly Sentinel-2 with daily PlanetScope collections can close the gap in continuous daily monitoring,” Russwurm says. “In addition, PlanetScope and Sentinel-2 sometimes record the same area of marine debris on the same day within minutes of each other. This double image of the same object in two locations shows the direction of drift due to wind and oceanic thin studies.”

Research in these areas will be continued by Mark Rooswurm from the University of Wageningen, together with Professor Tim van Emmerik, an expert on river plastics, and partners in the Netherlands, such as Ocean Cleanup. The company collects plastic in the open ocean on special ships. This is a collaboration for the AI for Detection of Plastics with Tracking (ADOPT) joint project in collaboration with the Swiss Data Science Center (a joint venture between ETH Zurich and EPFL) and EPFL Prof. Davies Tuia, Dr. Emmanuel Dalsasso and Prof. Marc. Rooswurm.

The Ocean Cleanup

In 2021, the NGO The Ocean Cleanup offered its own AI-powered monitoring and mapping tool.

Using Machine Learning technologies, Ocean Cleanup detects plastic pollution and models its movement in the ocean. This helps passive cleaning systems remove plastic.

The model was trained on a huge number of source images. A neural network detects objects on them. It consists of a series of mathematical equations with different settings called weights. Thanks to this, the neural network learned to detect the object based on the training examples it received by passing the example images.

The training images also needed a lot of work. To obtain them, Ocean Cleanup tagged approximately 4,000 examples of objects in photographs from their Aerial Expedition (2016) mission and the 2018 System 00.1 (“Wilson”) voyage. The transformed images were fed to the model for training, after which a dataset of 18,589 images was usable.

In addition, the Ocean Cleanup team collected a dataset from aboard the Maersk Transporter during their System 001/B mission in 2019. In doing so, they captured more than 100,000 time-lapse photos from the port and starboard sides of the vessel. The photos were geotagged, which helped to obtain unique GPS coordinates.

Knowing the location of each object, the Ocean Cleanup team grouped the GPS coordinates of the photos into different sectors. From this, they obtained an estimated number density per transect by grouping the detected objects in the transect and dividing this by its total area. In this way, it allows the creation of a map of the concentration of plastic in the ocean.

By 2024, Ocean Cleanup plans to reduce ocean plastic by 90% by combining the two systems and scaling them up.

Razer and Clearbot

In 2021 Razer, in partnership with ClearBot, created an automated robot that works based on Machine Learning and Artificial Intelligence and detects marine plastic at a distance of two meters in turbulent water. The robot runs on solar energy and can collect up to 250 kg of plastic in just one cycle.

ClearBot’s CEO says, “The Razer team’s action-oriented approach to tackling marine litter has been extremely impressive. We are grateful to the team that volunteered their time for this project. With the new model, we are confident that we will be able to expand our global reach to protect marine waters, starting with partners including seaport operators in Asia and non-governmental organizations who have already expressed interest. Together with Razer, we look forward to making positive changes in the world.”

That year, for World Oceans Day, ClearBot called for pictures of marine plastic debris to be uploaded to its website. This is usually found in open water. AI models were trained to detect waste based on these photos.

Gringgo Indonesia Foundation

For Indonesia, the problem of plastic waste is very acute. Much of the trash that is located along the coastline can end up in the ocean. That is why the Gringgo Indonesia Foundation decided to solve this problem with the help of technology and also with the support of Google.

In 2019, Gringgo was named one of 20 grantees of the Google AI Impact Challenge, giving them resources to get started.

Gringgo Indonesia Foundation aimed to create an image recognition tool that could classify materials and give them a monetary value. This would help reduce ocean pollution, improve public awareness, and help create an economic model for waste management.

In 2017, they launched several apps that allowed recycling workers to track the amount and type of waste they collect. This saved time by offering a more organized route and profit estimation. Using Image Recognition technology, waste management workers can take pictures of trash and identify its value.

Thanks to this, employees of waste processing services can estimate the market value of materials and optimize their activities.

Febriadi Pratama, CTO & co-founder at Gringgo says: “Within a year of launching the apps, we were able to increase recycling rates by 35 percent in our first pilot village, Sanur Kaja in Bali. We also launched a public app that connects people with services collection of waste in their homes”.

With the support of Google, the Gringgo Indonesia Foundation is working with the Indonesian startup Datanest. Together, they are working to create an image recognition tool using Google’s TensorFlow machine learning platform. The goal of the project is to enable waste workers to better analyze and classify waste, as well as quantify its value.

Open Ocean Engineering

Hong Kong startup Open Ocean Engineering aims to fight pollution with zero-emission automated boats called Clearbots. They can scoop trash out of the water and bring it ashore for recycling.

In 2022, Open Ocean Engineering won the Alibaba Group Jumpstarter 2022 Global Pitch Competition for startups.

Clearbot Neo is an automated boat that sails through harbors, canals or rivers to collect trash that would otherwise end up in the ocean. Siddhant Gupta and Utkarsh Goel, the creators of Clearbot, were inspired to create it after a trip to Bali. There, local workers removed garbage from the sea in small boats. That is why they decided to automate this process, and in conclusion, Clearbot was born.

The robot works from an electric motor on solar batteries. It picks up the debris in the water and collects it. One such robot can collect a metric ton of garbage per day. It can also clean up localized oil or fuel spills if equipped with an individual boom. In this case, it can collect up to 15 liters of pollutants per day. Clearbot Neo has a dual-camera detection system. This allows it to monitor the surface of the water look for debris or any obstructions and scoop the debris onto an onboard conveyor between the twin hulls. At the same time, another camera photographs the garbage that is already entering the conveyor. The image and location are received by the corresponding system hosted on the Microsoft Azure platform.

In the spring of 2020, Open Ocean Engineering received an AI for Earth grant from Microsoft. This helped them create their model, based on Artificial Intelligence on the Azure platform.

“We’re figuring out how the trash gets into the water in the first place,” Gupta said. “This adds significant transparency to the sea cleanup process. We’re collecting data about what’s in the water, what’s the composition of the things that are in there, how much of it can be recycled, and what materials we should focus on.”

Read more here - https://amazinum.com/insights/deep-dive-ais-impact-on-marine-ecology/