There’s no doubt that Climate Change is an urgent concern for humanity. India's Monsoon Floods, Canada's wildfires, and Brazil's rain-induced landslides are just some of the devastating consequences of this environmental doom. We need fast, efficient, and innovative solutions to manage our natural resources in this critical moment. Could Machine Learning (ML), one of the fastest-growing technologies, unleash its transformative power in tackling climate change?
Machine Learning in a Nutshell
From voice search technology, like Siri and Alexa, to automated translation, it's almost impossible to live your everyday life without coming into contact with at least one ML-driven technology. Essentially, Machine Learning is a branch of Artificial Intelligence where computers learn how to make decisions by themselves through processing large amounts of data. One of its most important divisions when talking about climate change is Artificial Neural Networks (ANN), which, as you can probably tell by its name, was modeled after the human brain. It consists of three main layers:
- Input layer- introduces initial information.
- Hidden layers- computes, does calculations based on input, and transfers results to output.
- Output layers- obtains final results.
As climate complexities arise, ANN excels in tackling interconnected factors like temperature, humidity, wind patterns, and greenhouse gasses (Kumari et al., 2023) .Priya Donti, the co-founder of Climate Change AI (CCAI), reveals ML's prowess in climate change–from revolutionizing our approach to agriculture to improving our climate models–ushering in a sustainable future.
Improving agriculture approaches
Although it might seem harmless, producing food is one of the main contributors to large-scale Greenhouse Gas emissions, particularly through crop production. Industrial farming involves more than just growing crops: clearing the native vegetation releases stored carbon, and tilling exposes soil, releasing even more carbon dioxide. While widely practiced for many years, this traditional agricultural method is unsustainable in our current world.
Precision agriculture represents a smarter way of farming, whereby embracing the natural diversity of farmland and crops, it maximizes yields and minimizes waste (Rolnick et al., 2022). Recurrent Neural Networks and Convolutional Neural Networks are two very important techniques for the existence of precision agriculture. These neural networks work with complex data, like information and images gathered over a long period of time, allowing researchers to make better predictions and decisions. Further, a study conducted by National Institute of Technology Durgapur researcher, Manas Mohanty, proved the potential of these techniques. The proposed model improves localization accuracy, helping to effectively locate small devices that gather and share information from their surrounding area called sensor nodes in agricultural fields. Using the predicted locations of these sensor nodes, a fertigation (simultaneous irrigation and fertilization) management system is proposed to optimize water and nutrient application (Mohanty et al., 2023). Innovations like these are very important for reducing the waste of resources. They can also significantly increase crop yields and minimize the toxicity of our fruits and vegetables, ultimately aiding in the fight against climate change.
This new farming method holds a lot of promise, but there are some valid concerns about its use. Over-reliance on ANN-based systems may lead to data privacy breaches, and excessive dependence on technology may disconnect farmers from traditional knowledge. A balanced approach—namely, integrating neural networks with existing methods and encouraging farmers to retain their expertise for optimized benefits—would be crucial for leveraging the best of both worlds.
Advancing Sea Level Monitoring and Climate Models
It's evident in the events of the past few years that extreme alterations are happening in the oceans. Hurricanes are becoming stronger than ever, the constant occurrence of floods is devastating whole communities, and according to NASA, the Arctic Ocean will become ice-free before mid-century (NASA, 2023) But what is causing these changes, anyway? Well, simply put, global warming is making ice sheets melt, which causes the ocean water to warm up and expand.
One of the main ways technology has helped us keep track of this fluctuation is through advanced satellite data. Within this data, ANN can help recognize intricate patterns allowing experts to capture subtle changes and trends necessary for accurate predictions within the dynamics of the ocean. This empowers professionals to safeguard vulnerable ecosystems and coastal areas that billions call home (Kumari et al., 2023). Neural networks are favored over traditional methods mainly because these variations happening in the sea are not linear, meaning that while ANN can work around gaps in data or lack of measurements, other statistical analysis procedures cannot (Liu et al., 2010).
ANN's capacity also successfully decodes weather patterns through abundant data. Unlike climate models designed for yearly predictions, neural networks can revolutionize forecasts on medium time scales (like weeks or months) with its unmatched pattern recognition and feature extraction techniques. This newfound potential could be the game-changer in enhancing our preparedness against climate disasters' escalating frequency and intensity (Hwang et al., 2019).
However, there is a possible implication of underestimation bias in the data (Blanzeisky et al., 2021). While these evolutions in sea level monitoring and climate forecasting have been much help, the neural network's ability to recognize intricate patterns may wrongly lead to underestimating the severity of specific changes. The continuous update and diversification of the training data to include extreme events and incorporating physical models (a representation of the attributes of the natural system) and additional data sources for cross-validation is essential.
Will ML make an impact in the fight against climate change?
There is no doubt that Machine Learning can have a transformative effect on our fight against climate change. But, unfortunately, it's prone to fallibilities that can cause a lot of issues. Continuing this journey, many questions come up. How can we reduce bias in ML models? Is it possible to overcome extreme data breaches? As researchers work on answering these questions, it is invaluable for us to keep on hoping.
References
Kumari, N., & Pandey, S. (2023). Application of artificial intelligence in environmental sustainability and climate change. Visualization Techniques for Climate Change with Machine Learning and Artificial Intelligence, 293–316. doi:10.1016/b978-0-323-99714-0.00018-2
Rolnick, D., Donti, P. L., Kaack, L. H., Kochanski, K., Lacoste, A., Sankaran, K., … Bengio, Y. (2022). Tackling Climate Change with Machine Learning. ACM Computing Surveys, 55(2), 1–96. doi:10.1145/3485128
Mohanty, M. K., Thakurta, P. K. G., & Kar, S. (2023, February 8). Efficient sensor node localization in precision agriculture: An Ann based framework - OPSEARCH. SpringerLink. https://link.springer.com/article/10.1007/s12597-023-00625-4
Climate Change, N. G. (2023, February 1). Sea Level | NASA Global Climate Change. Climate Change: Vital Signs of the Planet. https://climate.nasa.gov/vital-signs/sea-level
Liu, Z., Peng, C., Xiang, W., Tian, D., Deng, X., & Zhao, M. (2010). Application of artificial neural networks in global climate change and Ecological Research: An overview. Chinese Science Bulletin, 55(34), 3853–3863. doi:10.1007/s11434-010-4183-3
Hwang, J., Orenstein, P., Cohen, J., Pfeiffer, K., & Mackey, L. (2019). Improving subseasonal forecasting in the western U.S. with Machine Learning. Proceedings of the 25th ACM SIGKDD International Conference on Knowledge Discovery & Data Mining. doi:10.1145/3292500.3330674
Blanzeisky, W., & Cunningham, P. (2021a). Algorithmic factors influencing bias in machine learning. Communications in Computer and Information Science, 559–574. doi:10.1007/978-3-030-93736-2_41