Data Science and Sustainability: Understanding Human Impact on the Planet

Sustainability is not only an application field for data science and Artificial Intelligence (AI) but also a matter concerning their own development. This was the central theme of the presentation given by Antonietta Mira, Professor at the Faculty of Economics of Università della Svizzera italiana, during VisionScience, a teacher training event held at ETH Zurich.

In her contribution, Antonietta Mira addressed sustainability as a multidimensional challenge intertwining environmental, social, and economic aspects, highlighting the central role of data science in measuring, understanding, and governing the impact of human activities on the planet. This theme was introduced through the concept of the AI divide: today, only a limited number of countries host advanced infrastructure such as AI data centres, with direct consequences for who can innovate, control, and benefit from these technologies. This imbalance raises crucial questions regarding equity, governance, and sustainable development.

The presentation showed how data analysis allows us to reconstruct profound transformations of the Earth system. Among the examples cited was a study published in Nature Communications that reconstructs the evolution of global mammal biomass from 1850 to the present. While in the mid-19th century the biomass of wild mammals was comparable to the combined biomass of humans and livestock, today the situation has radically reversed. Human and livestock biomass has increased approximately fivefold, while wild mammal biomass has been halved, primarily due to habitat loss and unsustainable hunting.

A second emblematic result concerns so-called anthropogenic mass. A study published in Nature shows that around 2020, the total mass of human-made objects surpassed the total biomass of all living organisms on Earth. This sustained growth, which accelerated after World War II, reflects the intensity of the pressure exerted by human activities on ecosystems.

The presentation then highlighted how integrating satellite data, field measurements, and statistical models has revolutionised the estimation of natural resources. This is particularly evident in the case of forests: recent studies estimate the existence of approximately three trillion trees on the planet, but the number of trees per person is steadily decreasing, offering a concise and immediate indicator of human pressure on terrestrial ecosystems.

A further example concerned marine ecosystems. In research conducted by Antonietta Mira in collaboration with the Queensland University of Technology, the use of machine learning algorithms, georeferenced imagery, and citizen science contributions allowed for the observation of an increasing homogenisation of the benthic composition of the Great Barrier Reef following cyclones and bleaching events—an early signal of biodiversity loss linked to climate change.

The final part of the intervention focused on the sustainability of data science and AI themselves. Although some applications, such as Large Language Models, require more energy per interaction compared to traditional digital services, the data shows a more balanced picture. According to the World Energy Outlook 2024 by the International Energy Agency, data centres currently consume less than 2% of global electricity, and their contribution to energy demand growth will remain limited in the coming years, thanks in part to significant efficiency improvements. At the same time, AI can become an active tool for climate mitigation, contributing to the optimisation of energy systems, the development of renewables, and the discovery of new materials.