• Businesses have not paid enough attention to the increasing role software will have in environmental regulation and environmentalism;
  • Machine learning can help environmental regulators analyse enormous data streams, helping them uncover new information and work more efficiently;
  • Software can help build new ways to monitor carbon emissions and evaluate carbon offset efforts, spotting problems and identifying opportunities.

In 2011, Marc Andreessen, co-founder of Netscape and venture capital firm Andreessen Horowitz, wrote an op-ed in the Wall Street Journal titled “Why software is eating the world”. His contention was simple: software can now transform industries globally owing to the rapid decline in cost of microprocessors and the transition to the modern internet which connects commercial networks and enterprises.

Throughout the 1990s and early 2000s, the role of software in businesses was deceptively subtle. At first, the only companies that used software products were software product companies. Then software crept into the administration of enterprises as companies sought to increase efficiencies in their businesses. Owing to the internet and more recently smartphones, customer behaviour has changed so dramatically that software is being implemented in all sectors. Companies are now expected to reimagine their business to ensure they are not the next taxi company caught off-guard by their Uber-equivalent.

Yet, while businesses wishing to avoid becoming the next Kodak have paid considerable attention to the “disruption” that software start-ups or software-converted competitors can cause, little focus has been devoted to what impact the increasing role software will have in environmental regulation and, more broadly, environmentalism.

The preponderance of sensors attached to smartphones, drones or satellites, creates enormous data streams which we can plug into machine learning models to tell us things that previously were not obvious or available to environmental regulators. Through software, this information could not only save regulators time and money but also increase their effectiveness.

For example, our team's analysis highlights how software can be used to help Chinese regulators realize their ambitious plan to be carbon neutral before 2060. Our report used satellite imagery and machine learning to estimate production from coal power plants in China, which represent 30% of carbon emissions. These models currently achieve a mean average error (MAE) of 14% at the plant level based on data from the EU, the US and Australia; and 11% at the provincial level based on disaggregated data from the Chinese government.

The breakdown of global CO2 emissions by region
The breakdown of global CO2 emissions by region
Image: Our World in Data

These carbon emissions estimates could be used to reduce monitoring, reporting and verification (MRV) costs associated with China’s recently launched emissions trading system (ETS). With emissions monitoring, Chinese regulators could develop an anomaly detection system to prompt site audits based on the probability of data falsification. Instead of auditing the 2,225 electricity generators currently regulated by China’s ETS, regulators could identify data falsification in near-real-time.

China already has a continuous monitoring system for air pollution which, with the help of the Institute of Public & Environmental Affairs’s Blue Map database, has identified and rectified numerous non-compliance events. As China looks to expand its ETS to cover heavy industry sectors, including steel, cement and chemicals, regulators have a unique opportunity to expand their continuous monitoring system for air pollution to cover carbon emissions as well. This data would shine a spotlight on industries that account for a substantial portion of global carbon emissions.

Illustration of an anomaly detection system to determine MRV audits for companies regulated by China’s ETS
Illustration of an anomaly detection system to determine MRV audits for companies regulated by China’s ETS
Image: Turning the Supertanker: Powering China’s coal to clean transition with actionable analytics, TransitionZero

Continuous monitoring systems are already being deployed to deal with methane leaks from upstream oil and gas production. Until recently, little was known about these leaks. Research by the Environmental Defence Fund found that the US oil and gas industry was emitting at least 13 million metric tonnes of methane a year – about 60% more than the Environmental Protection Agency estimated at the time. Start-ups LongPath Technologies and Quanta3 have developed technologies and software to monitor leaks from US drilling rigs. These technologies offer the potential to curb a potent contributor to global warming while saving the US oil industry billions of dollars in lost gas.

Monitoring systems are not only limited to sources of emissions but also sinks for the booming offset market. For example, carbonplan, a US non-profit, recently identified systematic over-crediting of forest offsets in California’s forest carbon offsets programme – the largest such programme in existence, worth more than $2 billion. To bring this discussion into the open, carbonplan created a digitized project database to allow regulators to act to rectify these shortcomings.

Over the coming decades, humanity will need to deliver economic growth without putting intolerable pressure on the world’s scarce resources. As sensor technologies, cloud computing and predictive analytics continue to course through every sector of the economy, regulators have a unique opportunity to use software to identify and deal with environmental problems at a fraction of the cost and time of traditional regulatory approaches.