Updated 2/8/16 at 1:32 pm and 2/10/16 at 2:50 pm
Many companies are making their carbon emissions public, to show they are doing their part to fight climate change.
But new research by Lux Research indicates most companies in the U.S. are either underestimating or overestimating their emissions.
Ory Zik is Vice President of Analytics for Lux Research. He says estimating one's own carbon emissions is very difficult. That's because electricity moves from region to region on grids.
"If you look at the plug in your home, or wherever, it's very hard to know whether the electricity flows from a nearby coal-fired plant or a remote hydro plant or a solar power plant," says Zik. "So it's a hard problem but it can be mathematically solved if you collect the data."
So, that's what Lux Research did.
The group says most companies use the U.S. Environmental Protection Agency (EPA) Emissions & Generation Resource Integrated Database (eGRID) to estimate their emissions. However, eGRID divides the U.S. electricity grid into just 24 broad regions, and is updated only infrequently – the most recent information available is from 2012.
Data scientists at Lux Research divided the grid into 134 regions, instead of just 24, providing more granular insight, and used U.S. Energy Information Administration (EIA) data that is updated monthly, as opposed to three-year-old annual data.
The researchers say its new mathematical model improves the accuracy of carbon reporting by a factor of 80. Using the new model, researchers say they discovered that Google underestimated its carbon emissions at four out of seven of its U.S. data centers.
At its South Carolina data center, the company underestimated its carbon emissions by a full 30%.
The research shows Amazon is off, too. Its 23 Virginia-based cloud services data centers use about 43% electricity from coal – not 35% as inferred using eGRID.
Zik says they could have chosen any number of companies to call out, because virtually all are relying on the less accurate eGRID information.
They picked Google and Amazon as examples because both clearly have the analytical expertise to do a good job when reporting their emissions - but don't.
Zik says the study shows the need to be skeptical of companies' claims and to hold them to a higher standard.
"If a large company makes the claim that they built a carbon-neutral data center or a data center that is 50% carbon neutral, when they don't even bother to calculate what they use today - then we need to be very careful about these claims," he says.
Zik says as inaccurate as U.S. emissions calculations are, it's even worse overseas.
Spain has 3000 power plants, he says, yet the government lumps the data from all of them together, which makes it almost impossible to figure out a company's individual emissions.
"And if you have an operation in China, the level of data transparency is is so low that it's even hard to know where the power plants are placed."
Zik says Lux Research has plans to duplicate its U.S. research for other countries, so that companies that report their global emissions can do so more accurately.
Google declined to comment on the record. Amazon did not respond to a request for comment.
The EPA responded thus:
eGRID is a comprehensive database of the environmental characteristics of all grid connected electric power generated in the United States, providing researchers, policy makers and the public with a powerful set of information about electricity generation. For over 17 years, eGRID data have been widely used by EPA and others to quantify the air pollution from power plants as well as the benefits from reduced electricity demand. In addition, eGRID’s data on carbon dioxide emission rates are commonly used to estimate the indirect emissions from electricity purchases in most carbon footprint calculators and greenhouse gas inventories.
Re: timing of eGRID updates
eGRID has been collecting and synthesizing power sector data for over 17 years and the long-term and widespread use of eGRID ensures accurate benchmarking and meaningful comparisons over time.
eGRID gathers data from all available sources - including annual data that is directly reported from power plants to the Department of Energy’s Energy Information Administration and EPA - and synthesizes these into a usable format. eGRID uses annual data to reduce the seasonal variability in generation and emissions. This data then undergoes extensive analysis, quality control, and peer review - analytical steps that take time but which result in a comprehensive, accurate and robust data set. EPA has begun implementing a series of data handling improvements that will result in quicker releases of future eGRID updates.
Re: number of subregions
eGRID provides emission factors at many levels (i.e., plant, State, and Power Control Area (PCA)). Given the interconnectedness of the electrical grid, however, EPA recommends using the emission rates from the 26 eGRID subregions. This level appropriately addresses power imports and exports and better represents the actual mix of generation types (e.g., renewable, natural gas, coal) used by individuals within the eGRID subregion.