The importance of transparency in relation to company carbon emissions has grown significantly in recent years due to the growing awareness of the impact of greenhouse gas emissions on the environment and the urgent need to address climate change. Companies, governments, and other organizations are increasingly being held accountable for their carbon emissions, and are expected to publicly disclose their emissions data and progress in reducing their carbon footprint. This increased transparency helps stakeholders to make informed decisions, encourages organizations to reduce their emissions, and holds them accountable for their actions. It is also seen as a key component of efforts to achieve a more sustainable and low-carbon future.
Gaps in company carbon emissions data can pose significant problems in the effort to accurately assess and reduce greenhouse gas emissions. These gaps can arise for a number of reasons, including a lack of reporting requirements, inaccurate reporting as well as a lack of standardization.
These gaps in company carbon emissions data can undermine efforts to effectively address climate change by making it difficult to accurately assess emissions and track progress. Addressing these gaps requires a concerted effort to improve reporting requirements, increase accuracy and transparency, and standardize methodologies for measuring and reporting emissions.
What drives the need for corporate emissions predictions?
Accurate company carbon emissions reporting: Why does it matter?
Accurate corporate emissions predictions are crucial for several reasons:
Climate policy and regulation: Accurate emissions predictions are essential for the development and implementation of effective climate policies and regulations. These policies and regulations aim to reduce greenhouse gas emissions and mitigate the impacts of climate change.
Corporate responsibility: Companies are becoming more aware of their responsibility to reduce their carbon footprint and operate in a more sustainable manner. Accurate emissions predictions enable companies to set targets for reducing their emissions, track their progress, and demonstrate their commitment to sustainability.
Investment decisions: Investors are increasingly looking to invest in companies that are reducing their carbon footprint and operating in a sustainable manner. Accurate emissions predictions help investors make informed decisions and assess the sustainability of potential investments.
Climate risk management: Climate change and its impacts pose a growing risk to businesses and the global economy. Accurate emissions predictions are essential for companies to understand their exposure to these risks and take appropriate action to mitigate them.
Overall, accurate corporate emissions predictions are crucial for the development of effective climate policies, for companies to take responsibility for their emissions, for informed investment decisions, and for effective climate risk management.
What are Scope 1-3 emissions?
Scope 1-3 emissions refer to three categories of greenhouse gas (GHG) emissions that are used to quantify and track a company’s or an organization’s carbon footprint.
Scope 1 emissions: These are direct emissions from sources that are owned or controlled by the company, such as fuel combustion in boilers, vehicles, and other on-site sources.
Scope 2 emissions: These are indirect emissions from the generation of electricity, heating, and cooling that the company consumes but does not produce itself.
Scope 3 emissions: These are all other indirect emissions that are not included in Scope 2 and are a result of the company’s activities, but occur from sources outside of the company’s direct control. This includes emissions from the production of purchased goods, transportation and waste disposal.
Scope 1-3 emissions are used to provide a comprehensive picture of a company’s carbon footprint and are often reported in accordance with the GHG Protocol, a widely recognized accounting standard for GHG emissions. Understanding and tracking these emissions is important for companies to set targets for reducing their carbon footprint, track their progress, and demonstrate their commitment to sustainability.
Using machine learning to improve the accuracy of company carbon emission estimates
To solve this, Permutable has been applying it’s cutting-edge machine learning technology and is pleased to announce positive initial results from its ongoing carbon emissions data modelling programme in partnership with Innovate UK.
Permutable’s goal is to fundamentally improve and add more detailed insights into how carbon emissions can be better estimated through the use of artificial intelligence, particularly for companies whose carbon emissions remain unreported.
Stastical models vs machine learning
Previously, emissions would typically be estimated using statistical models – e.g. mean, weighted average (by # of employees) for a certain industry/country pair. These approaches can be very rigid compared to machine learning models. That is because they focus on their ability to explain results, starting from a hypothesis and then modelling predictions based on that. By contrast, machine learning approaches are more flexible because they focus on what best fits the data they are trained for.
Company carbon emission reporting: Increased accuracy of up to 97%
The initial results reported here confirm that, using machine learning techniques, the accuracy of carbon emissions modeling for the different emission scopes is 68 – 99% better than using country averages, using smaller amounts of data[i].
- 96% improvement over industry mean across >900 companies
- 74% improvement over industry mean across >800 companies
- 97% improvement over industry mean across >500 companies
There were 11 industry groups considered in total including Infrastructure, Extractives and Minerals Processing, Financial, Technology & Communications, Resource Transformation, Transportation, Health Care, Services, Consumer Goods, Food & Beverage and Renewable and Alternative Energy.
Charts above depict:
Chart 1: Representation of the scope 1 predictions against expectations and typical statistical estimates (the mean across US / industry clusters). The predictions were made using both financial & energy features, where multicollinear features (p>0.8) were removed. The current model shows a 68% decrease in MSE (mean squared error) compared to the industry average and a 76% decrease in MSE compared to the full cluster average.
Chart 2: Representation of the scope 2 predictions against expectations and typical statistical estimates (the mean across US / industry clusters). The predictions were made using both financial & energy features. The current model shows a 95% decrease in MSE (mean squared error) compared to the industry average and a 90% decrease in MSE compared to the full cluster average.
Chart 3: Representation of the scope 3 predictions against expectations and typical statistical estimates (the mean across US / industry clusters). The predictions were made using only the first most important financial features, as determined by the feature importance on the baseline random forest training set. The current model shows a 99% decrease in MSE (mean squared error) compared to both the industry average and the full cluster average.
Overall, the above findings suggests great potential for more accurately predicting corporate emissions across multiple countries and industries, and is expected to increase overall transparency, reliability and therefore confidence in carbon emissions predictions.
Wilson Chan, Founder and CEO of Permutable commented, “We are very pleased with the initial results from the carbon emissions modelling project as part of our work with Innovate UK. The initial results are encouraging and we look forward to building on these results in phase two by incorporating higher quality data from more industries and countries.”