Data & Sustainability: The Impact of Data Analytics on a Sustainable Future
Can Data Really Solve all our Problems?
Most would agree that to an extent, having so much data at our fingertips has drastically changed our lives. Advances in technology have changed how we interact with the world around us, and greatly increased our understanding of modern human life. In recent years, pressure from both government agencies and consumers have placed a growing demand on businesses to be more accountable for their environmental impacts, and create products and services that are more sustainable. With the growing concerns about the global impacts of carbon emissions, material waste, energy consumption, and the availability of limited resources, many businesses are trying to find new ways to use data to create more sustainable and resilient products, and make their businesses more environmentally friendly. However, not everything is without fault. The energy consumption it takes to build the infrastructures for collecting, sorting, and using data needed to assess their sustainability efforts have their own negative impact on the environment. Could the very tool we use to try and improve our environment contribute to its worsening condition more than it can help, or is it outweighed by the potential good it can do? Is there a way to ensure the benefits of using data outweigh any negative impacts it might have?
Companies create more data than you might realise. The types of data one business can generate can include information on energy consumption of offices and warehouses; CO2 emissions from manufacturing and supply chains; methods for obtaining raw materials and scarce resources; processing of raw materials into usable products; and the use and/or disposal of products by consumers. All of these aspects of a business have measurable impacts on the environment. Until recently, most businesses did not fully understand the impact of their own operations on the environment, as information was spread out across different departments or operations. Something as simple as energy consumption might have been difficult to measure, with data spread across different documents and formats throughout an organisation. The use of Cloud systems, and more streamlined generation of data, allow for better access — and thus understanding — of their own data. Understanding the data generated can help companies make informed changes to lessen their global impact. Data can be used to notice trends, optimise supply chains or processing methods, or even predict trends in wastage or energy use.
The United Nations (UN) suggests that similar data analysis methodologies used in the private sector — such as consumer profiling and predictive analytics — can be used alongside new sources of data, such as satellite data, to more accurately identify vulnerable populations and people’s wellbeing. Many governments, especially poorer and more marginalised nations, do not have access to adequate population-level data that is needed in order to identify issues and make societal changes. Most of the useful data needed to accomplish real change is collected by private sectors. One area where partnerships between private and public sectors would be beneficial is for measuring progress of Sustainable Development Goals (SDGs) — a blueprint for addressing various global challenges such as poverty, inequality, climate change, environmental degradation, peace, and justice. Though the use of data can help measure SDG progress in fairer and more inclusive ways, a UN report states that there isn’t enough data for 68% of the environment-related SDGs to track their progress.
With the partnering of digital technology, big data, and the right governance structures and business models, organisations can achieve greater outcomes and reduce environmental impacts at scale, even now. The collaboration between IBM and YARA — a global leader in crop nutrition and digital farming solutions — is one such example. By working together, they were able to co-create the Open Farm & Field Data Exchange, a platform that helps farmers optimise and reduce the environmental footprint of their food production in real-time. The platform is compatible with any cloud system, and farmers are able to use data on weather, crops, and soil, to measure and reduce their footprint.
The Kellogg Company — one of the largest food manufacturing companies globally — have also demonstrated how the use of data helped them cut emissions, while also saving millions of dollars annually. In 2005, they committed to reducing greenhouse gas emissions, starting with the use of a data management platform. This platform was able to track and manage energy-use data at their company headquarters, and found that some of their HVAC systems were working against each other. Kellogg’s was able to use this information to fix HVAC systems across many of their plants, leading to a carbon savings of almost 400,000 metric tons over a 7-year period. This carbon savings is equivalent to removing 114,000 cars from the road every year.
Data and analytics can also be used to improve performance of many power companies to stabilise energy grids to meet customer demands. In the Netherlands, IBM assisted TenneT (a European electricity transmission system) in launching Equigy, a blockchain platform designed to stabilise how energy is fed into the power grid. For example, wind and solar power can experience unpredictable fluctuations, such as when too much wind leads to a surplus of energy. The Equigy blockchain platform can help regulate how energy is fed into the power grid to keep it stable. In North America, one of the largest low-cost clean power generators is Ontario Power Generator (OPG). By using data and predictive analytics, OPG was able to build a dashboard that recognised patterns of grid functionality or problems, and create corresponding notifications over the cloud. Such analytics can predict problems in the grid before they get worse, and allow for quick action to be taken.
While data and analytics can help improve sustainability efforts, we can’t forget that there is still a price to using this technology. Digital technology — from streaming, gaming, to online banking and trading — is often overlooked as a primary carbon producer, having its own environmental impact. Between 2013 and 2020, the energy consumption of digital technology increased by nearly 70%. In the US alone, data centres are responsible for 2% of the country’s total electricity use, and while many countries claim to use greener energy sources, many major companies still depend, at least partly, on fossil fuels to power their data centres.
IT and data infrastructures are a significant contributor to carbon and other greenhouse emissions from the creation and use of these infrastructures and devices. The carbon footprint of the Information and Communication Technology (ICT) sector — including the construction and use of user devices, networks and data centres — accounts for about 1.4% of total global greenhouse gas emissions, and approximately 3.6% of global electricity consumption. To put this into perspective, the UK could reduce its carbon emissions by over 16,000 tons if each adult sends one less email per day. If such a simple change can lead to such a large decrease in carbon emissions, you can probably imagine just how much these infrastructures really impact the global environment.
Some are concerned over potentially higher electricity consumption with newer technologies all the time. The machine learning algorithms mentioned above, while beneficial, still use a lot of energy to compile and process vast amounts of data. Though the infrastructures pertaining to data and analytics can have a negative impact on the environment, it does positively contribute to solutions for industries and individuals alike to establish more large-scale sustainable practises. Like anything in life, balance is key. We need to be mindful of how often and how much we use data and technology for the benefit of the planet, and that it outweighs any negative environmental impacts the use of this technology might have.
But, people are developing new ways to determine the energy efficiency of machine learning algorithms, and believe it should be an important component for evaluating the accuracy of an algorithm in the near future. Researchers at the University of Copenhagen have created a free, open-source program called Carbontracker. This program can track and predict the approximate carbon footprint of machine learning models. Their hope is simple — that once someone is aware of the carbon footprint of a machine learning algorithm, they can take steps to reduce their footprint. Even though this might be a long way off from being widely used, these researchers have already suggested changes that can be made now to reduce an algorithm’s carbon footprint.
Running machine learning algorithms in parts of the world, and certain times of day, where energy is less carbon-intensive, are just some ways to reduce digital carbon footprints. For example, training a model in Sweden produces much less of a carbon footprint than in Estonia. Also, the time of day might also affect the carbon footprint of running a model — choosing hours of the day where energy is less carbon-intense can cut emissions. Choosing low-carbon-intensity hours can cut the carbon footprint by three-quarters in Denmark, and by half in the UK. Algorithms should also aim to be as efficient as possible. Efficiency means you can minimise computing power, and thus carbon emissions, required for training a machine learning model. The researchers involved with creating Carbontracker feel that these models should not just be evaluated based on their accuracy, but also their energy and carbon footprints, to help make the field of data analytics more and machine learning more environmentally friendly.
Overall, the carbon footprint of ICT and data is relatively small compared to its potentially negative impacts, but as the world becomes increasingly digital at an alarming rate, the opportunity for a larger digital carbon footprint is apparent. If steps are taken early on to make data and analytics more energy efficient, the potential benefits for other sectors — from forming to manufacturing — greatly outweigh any negative energy impact. Currently, existing ICT solutions have the potential to reduce global carbon emissions by up to 15%.
