27th July 2021

How can software help fight climate change?

climate change

Climate change is the greatest trial that humanity has ever faced. We have the numbers, and we know the timeline. If we do not cut global emissions in half by 2030 and eliminate them by 2050, the cascading effects of global warming will be irreversible. Slowly but surely, the Earth will be ravaged by increasingly frequent catastrophes and extreme weather events like wildfires, heatwaves, hurricanes, and floods, a trend we are already beginning to witness.

We have to utilize every tool and resource at our disposal to combat the impending crisis, and experts in the field suggest that software technologies like Digital Twins, Predictive Analytics, Machine Learning and Artificial Intelligence could offer highly promising solutions to our problem. To clarify, since the terms are often used interchangeably – Artificial Intelligence is the broad term for machines performing tasks in a way that would be considered ‘intelligent,’ whereas Machine Learning is an application of AI that involves feeding machines relevant data to learn to perform tasks without being explicitly programmed a certain way.

Here are a few areas where software can help us in our fight against global warming:

📡Modelling of Earth Systems using satellite and sensor data

As per the climate scientist Michael E. Mann, “looking down at the Earth from above is our most critical space mission”. AI can be combined with a new spectrum of satellite and imaging data to generate new insights into regional weather patterns and pollution levels.

NOAA-20, a highly advanced polar-orbiting satellite, operated by the U.S. National Oceanic and Atmospheric Administration (NOAA), is one of the new environmental satellites used to collect data about the Earth. NOAA-20’s sensors can image pollution levels and ice-sheet recession, as well as calculate the net heating of the Earth’s surface.

The classification of observational data has always been a tedious manual task, but Machine Learning classification solutions show promise in automating this process.

As satellite databases grow, the number of possible applications for ML is likely to grow as well. An increase in reliable data and instantaneous classification of this data would mean that AI-based climate models and forecasting programs would become highly accurate and efficient. Consequently, precise forecasting of extreme weather events could save thousands of lives in the years to come.

🧪AI-Assisted Material Invention

Cement and steel production together make up over a tenth of global greenhouse gas emissions. Researchers have begun to combine Machine Learning with generative design to create structural materials and products that require less raw material to produce.

Advances in materials science could allow ML researchers to access databases like the UCI Machine Learning Repository to invent new, climate-friendly materials. Researchers have already used semi-supervised generative models to create novel low-emission concrete formulas that could satisfy the required structural characteristics but only create a fraction of the emissions during production.

An experiment at Northwestern University used AI to figure out how to create new metal-glass hybrids hundreds of times faster than they would have using conventional trial-and-error methods.

Another group of researchers used databases to pinpoint 3D materials that can be peeled into 2D layers with desirable properties that they do not possess in their 3D form. An example of this is graphene – essentially a single sheet of graphite – a ‘miracle material’ that the scientific community has awaited for years.

📜Informing policy and climate investment 

The fight against climate change requires writing, assessing and passing new policies at all levels. Machine Learning programs can use a wide variety of datasets to improve policy analysis – evaluating the effects of past policies and the potential impacts of future alternatives.

We already know that Machine Learning can help gather relevant data about potential policy targets – sources of emissions, analyzing traffic patterns, etc. Following this, ML applications show potential in assessing policy options and evaluating the effects of policy decisions post-deployment.

Machine Learning can also be applied to climate analytics and investment – predicting the financial repercussions of climate change and optimizing investment portfolios based on companies’ carbon footprints to help strike a balance between profitability and sustainability.

⚡Electrical Grid & Utility-Scale Energy Storage Systems 

The transition to low-carbon energy sources is essential to mitigate emissions. There are two types of low-carbon sources – variable and controllable. Variable sources refer to renewable energy sources that depend on fluctuating external factors like wind and sun. Controllable sources refer to those that can be turned on and off, such as nuclear and geothermal power plants. 

Since renewable energy sources are often inconsistent, it is challenging to ensure demand and supply equilibrium – a mandatory requirement for the modern power grid. Utility-scale energy storage systems help stabilize the grid during periods of excess energy demand or supply, reducing the need to rely on gas-fired plants which pollute the environment, while enabling the addition of a higher renewable energy component to the energy mix. 

For the grid to become truly sustainable, it will require additional infrastructure to be built for long-duration energy storage. Given the capital intensive nature of these energy storage systems, technologies like digital twins and predictive analytics are going to be key in ensuring optimum utilization of these assets.

By building digital models of these assets, the software can enable front-of-the-meter storage operators and owners to make data-driven decisions to improve asset performance. Predictive analytics can alert potential failures weeks in advance, improving visibility and plan for timely replacements to avoid downtime.

Additionally, machine learning can help energy producers and operators better map demand and improve dispatch scheduling. Machine Learning deployment to forecast electricity demand and schedule dispatch is not just a concept but an existing technique that is continuously improving. The further optimization of supply-demand balance will depend on the inclusion of external data in the predictive models. An algorithm that simultaneously accounts for internal and external factors such as – demand, weather, emissions, cost – could be the key to a seamless, safe, green power system.


The transportation sector accounts for about 24% of global CO2 emissions. Passenger and freight transportation each accounts for almost half of transportation-related greenhouse gas emissions.

We are already seeing rapid advances being made to make mobility cleaner and sustainable whether it’s through electrification, shared mobility, vehicle connectivity or autonomous mobility. Talking about electrification according to Bloomberg NEF, there are 12 million passenger EVs, 1 million commercial EVs, and over 260 million electric two-and-three wheelers on the road globally today.

Despite these numbers, it’s not enough. A lot more needs to be done to increase EV adoption and make it comparable to ICE vehicles in terms of cost and efficiency. And software here can accelerate this transition if we are to achieve our common goals of making mobility zero-carbon by 2050.

Commercial Vehicles represent less than 1% of today’s EV fleet, however, thanks to such factors as falling costs, widening availability, and support from policymakers, EV fleets’ total cost of ownership is 15-25% less than ICE vehicles, encouraging fleet operators to increase the share of EVs in their fleet. McKinsey projects that by 2030, the size of the US commercial and passenger EV fleets would be eight million EVs (compared with fewer than 5,000 in 2018), which would amount to between 10 and 15 per cent of all fleet vehicles.

All this investment in EVs will require additional infrastructure in terms of charging. It’s estimated that the United States alone would require 13 million chargers by 2030 to service all of the country’s EVs. Such mass deployment of EV charging stations will require intelligent management and decision making. ML algorithms can enable fleet operators to maximize their fleet’s runtime while minimizing the cost of charging by simulating different charging patterns based on forecasted demand. Algorithms that model transport demands and routes – both passenger and freight – can significantly increase efficiency and cut down the trip frequency, transport time and fuel consumption. The number of buses that should run at a particular time of day, which goods should be shipped together, which pathways the ships carrying the goods should take; all of these complex decisions can be simplified using ML.

Lastly, Machine Learning will be essential for developing autonomous vehicles, which promise a reduction in energy consumption by reducing congestion and increasing efficiency. Imagine a grid of cars all talking to each other, constantly optimizing travel routes.

💻Software technologies have to make significant contributions to realize our vision for a zero-carbon future

Whether it’s through pinpointing pollution and climate modelling, assisting material invention and structural design, grid transformation or transport optimization, the potential applications of advanced software continue to grow with growth in the climate-relevant datasets. 

Combining powerful software tools like Digital Twins, Machine Learning and AI with fields both within and outside computer science can lead to powerful innovation, and this innovation has the potential to turn the tables in the fight for humanity’s future.

Recent Posts