Artificial intelligence, machine learning chip in to fight climate change, protect environment3 min read . Updated: 30 Jan 2020, 09:31 PM IST
- Did you know you could contribute idle computing power to help environment research projects?
- Google’s flood forecasting system uses ML to give a more accurate picture of areas that will be flooded so that people can be alerted in time
NEW DELHI : Global warming-induced climate change has started taking its toll on animals, plants and even humans. Until 2019, around 14,000-25,000 pairs of breeding emperor penguins would gather every year at Halley Bay colony in Antarctica, making it the second-largest penguin colony in the world after Coulman Island, also in Antarctica. Today, the colony at Halley Bay has almost disappeared, as per a 2019 study by the British Antarctic Survey. The deteriorating air quality in large cities like Delhi, particularly during winter, has serious health implications. To be sure, as per the Centre for Science and Environment, 30% of premature deaths in India are caused by air pollution. The recent bushfire in Australia killed an estimated one billion animals and has endangered more than 100 different species.
While cutting down global emissions and switching to renewable sources of energy will help in the long run, emerging technologies such as artificial intelligence (AI) and its various branches, along with internet of things (IoT), can help in the conservation of plants, animals and birds by providing more accurate climate predictions. They are also helping authorities detect acts of deforestation and poaching while some are creating awareness among masses.
Take the machine learning (ML)-based detector model by Princeton, US-based data science company Gramener, which can estimate the number of penguins from the images captured by multiple camera traps located across Antarctica. Gramener used an image dataset of penguin colonies in Antarctica that included images from more than 40 locations.
To clean up the penguin image dataset, they used multi-column convolutional neural networks (CNN). They then trained the deep learning model on Microsoft’s DSVM (data science virtual machine) platform and used Intel’s Xeon scalable processors to repurpose and benchmark it.
The model uses a density-based counting approach, which not only estimates the numbers faster but can also give more accurate results when compared with the manual counting of penguins from the images. Gramener has also used AI and data analytics to estimate population of 12 salmon species.
Another AI-driven initiative empowering environmentalists with better insights is the Elephant Listening Project, which uses acoustic sensors to listen to forest elephants in Africa. Their population has dwindled due to poaching. The brainchild of California-based Conservation Metrics, the project uses ML to more accurately identify the low-frequency rumbling sounds made by elephants to communicate with each other amid sounds made by other animals.
By using Microsoft’s Azure cloud computing services, researchers at Conservation Metrics have been able to process months of sound data in a few weeks. Both projects were supported by Microsoft’s AI for Earth programme.
Using ML to identify unique sound patterns and then using sensors to capture that data can also be very effective against illegal deforestation and poaching. San Francisco-based Rainforest Connection is making the most of it with an acoustic alert system that listens for indicators of deforestation and poaching such as sound of chainsaws, vehicles or guns that can be used to immediately alert authorities.
“Global warming has changed the way climate modelling is done. Using AI/ML is very important as it will make things happen faster. All this will require lots of computing power and, going forward, quantum computers might play an important role," said Jaspreet Bindra, digital transformation expert and author of The Tech Whisperer. IBM’s new AI-driven global high resolution atmosphere forecast (GRAF) can detect high-impact events such as tropical cyclones and thunderstorms. Predictions powered by GRAF can also be accessed on IBM’s The Weather Channel app.
Solving many of the scientific and mathematical problems related to the environment require massive computing power. UC Berkeley’s BOINC (Berkley Open Infrastructure for Network Computing) is helping researchers, including climatologists, by crowdsourcing the processing resources of multiple PCs and smartphones from around the world.
Users who want to contribute can install the BOINC software on their devices. When their system is idle, the available CPU and GPU resources will be automatically channelled to power computing projects. The project has more than 310,000 participants, with the computing power of 800,000 devices at their disposal.
Among Indian companies, Gurugram-based Blue Sky Analytics is using AI-powered solutions to analyse high volumes of satellite data, ground-level measurements from sensors and ancillary public datasets to deliver real-time high resolution data on the environment. One of their solutions, Zorro, is helping with environment regulation by monitoring industrial emissions with satellite data, another called BreeZo provides information on air quality while Zuri allows compliance authorities to monitor farm fires.
AI and IoT are changing the world’s approach to problem solving. Tasks that took months earlier take a few minutes now. Environmentalists need to embrace them on a larger scale to boost their efforts.