Not long ago in the southwest of England, a local community set out to replace a 1960s-vintage school with a new building using triple-pane windows and superinsulated walls to achieve the highest possible energy efficiency. The new school proudly opened on the same site as the old one, with the same number of students, and the same head person — and was soon burning more energy in a month than the old building had in a year.
The underfloor heating system in the new building was so badly designed that the windows automatically opened to dump heat several times a day even in winter. A camera in the parking lot somehow got wired as if it were a thermal sensor, and put out a call for energy any time anything passed in front of the lens. It was “a catalogue of disasters,” according to David Coley, a University of Bath specialist who came in to investigate.
Many of the disasters were traceable to the building energy model, a software simulation of energy use that is a critical step in designing any building intended to be green. Among other errors, the designers had extrapolated their plan from a simplified model of an isolated classroom set in a flat landscape, with full sun for much of the day. That dictated window tinting and shading to reduce solar gain. Nobody seems to have noticed that the new school actually stood in a valley surrounded by shade trees and needed all the solar gain it could get. The classrooms were so dark the lights had to be on all day.
Faulty energy modeling
It was an extreme case. But it was also a good example, according to Coley, of how overly optimistic energy modeling helps cause the “energy performance gap,” a problem that has become frustratingly familiar in green building projects.
The performance gap refers to the failure of energy improvements, often undertaken at great expense, to deliver some (or occasionally all) of the promised savings. A study last year of refurbished apartment buildings in Germany, for instance, found that they missed the predicted energy savings by anywhere from 5% to 28%. In Britain, an evaluation of 50 “leading-edge modern buildings,” from supermarkets to health care centers, reported that they “were routinely using up to 3.5 times more energy than their design had allowed for” — and producing on average 3.8 times the predicted carbon emissions.
The performance gap is “a vast, terrible, enormous problem,” in the words of one building technology specialist, and that’s not an exaggeration. Though much of the public concern about energy consumption and climate change focuses on automotive miles-per-gallon, the entire transport sector — including trains, planes, ships, trucks, and cars — accounts for just 26% of U.S. climate change emissions. Buildings come in at 40%, and they are the fastest growing source of emissions, according to the U.S. Green Building Council.
Eliminating the performance gap matters particularly for European Union nations, which have a legally binding commitment to reduce emissions by 80% to 95% below 1990 levels by mid-century. But knowing with confidence what savings will result matters for anybody trying to figure out how much to invest in a particular energy improvement.
Builders and occupants are not to blame
Researchers have generally blamed the performance gap on careless work by builders, overly complicated energy-saving technology, or the bad behaviors of the eventual occupants of a building. But in a new study, Coley and his co-authors put much of the blame on inept energy modeling.
The title of the study asks the provocative question, “Are Modelers Literate?” Even more provocatively, a press release from the University of Bath likens the misleading claims about building energy performance to the Volkswagen emissions scandal, in which actual emissions from diesel engine cars were up to 40 times higher than “the performance promised by the car manufacturer.”
For their study, Coley and his co-authors surveyed 108 building industry professionals — architects, engineers, and energy consultants — who routinely use energy performance models. To keep the problem simple, the researchers asked participants to look at a typical British semi-detached home recently updated to meet current building codes. Then they asked test subjects to rank which improvements made the most difference to energy performance.
Their answers had little correlation with objective reality, as determined by a study monitoring the actual energy performance of that home hour-by-hour over the course of a year. A quarter of the test subjects made judgments “that appeared worse than a person responding at random,” according to the study, which concluded that the sample of modelers, “and by implication the population of building modelers, cannot be considered modeling literate.”
The “garbage in, garbage out” problem
Predictably, that conclusion raised hackles. “The sample seems odd to me,” said Evan Mills, a building technology specialist at Lawrence Berkeley National Laboratory, “to include so many people who are junior in the practice, and then to be criticizing the industry at large.” He noted that almost two-thirds of the 108 test subjects had five years or less experience in construction. But Coley and his co-authors found that even test subjects with “higher-level qualifications, or having many years of experience in modeling,” were no more accurate than their juniors.
In any case, Mills acknowledged, “the performance gap is real, and we must be aware of models not properly capturing things. We have cases where modelers will come up with a savings measure that is more than the energy use of the house, because they are just working with the model,” and not paying attention to the real house.
That sort of problem — energy models showing unreasonable results — also turns up at the preliminary stage on 50% of projects going through the LEED certification process, said Gail Hampsmire of the U.S. Green Building Council. Designers have a tendency to take a “black box” approach, providing whatever inputs a particular energy model requires and then accepting the outputs “without evaluating the reasonability of those results,” she said. “You always have the issue of garbage in/garbage out, and the capability of the modeler to identify whether they are getting garbage out is critical.”
Finding a cure
So what’s the fix? The current accreditation requirements for energy modelers are “very gentle,” said Coley, but “when you’re trying to get something off the ground relatively quickly, you can’t send everybody back to college for three years.” In any case, the problem isn’t really education in the formal sense.
“It has to do with feedback,” he said, or the lack of it. The culture of building construction says it’s perfectly reasonable for architects — but not energy modelers —to travel hundreds of miles to see how the actual building compares with what they designed. For energy modelers, there’s not even an expectation that they’ll get on the phone with the building manager at year one and ask how energy usage compares with the original model. As a result, said Coley, energy modeling can become like theoretical physics: “You can very easily create a whole web of theories, and then you find yourself studying the physics of your theories, not the physics of the real world.”
The answer, he suggested, is a regulatory requirement that modelers follow up on their work by routinely checking their predictions against a building’s actual energy consumption. A system of modest inducements could also make that feedback more broadly available — for instance, by promising to take three weeks off the planning permissions process for developers who commit to posting actual energy usage to an online database. The Green Building Council has begun to require that sort of reporting for projects seeking LEED certification, said Hampsmire, with an online platform now in development “for building owners to track their own performance and compare it with other buildings.”
A second problem, according to Coley, is the tendency of government agencies to require simplified energy models at the start of the design process. The requirements often include certain uniform assumptions about energy use, making it easier to compare one building with another.
“Because you have to do that at the start, it becomes the default, and this sets up a kind of ‘Alice in Wonderland’ world, and it’s not surprising that modelers model this artificial world.” But at least in the United States that has become less of a problem in recent years, according to Hampsmire. Current building code requirements are “fairly good,” she said. “They don’t say, ‘Model energy use for a building occupied eight hours a day,’” or some other arbitrary standard. Instead, “they specifically state that all energy use has to be modeled as anticipated.”
Builders need realistic modeling
The takeaway from all this isn’t to discredit energy modeling but to improve it. Builders increasingly need realistic modeling, said Coley, by people with a deep knowledge of building physics and at least as much experience with real buildings as with energy models. Without that, the result will be even more $500-million office blocks with too much glass on the southern exposure, causing everybody inside to bake on a hot summer afternoon. Without smart energy modeling, the result will be a world spinning even faster into out-of-control climate change.
“This isn’t rocket science,” said the Berkeley Laboratory’s Mills. But then he added, “It’s harder than rocket science.”
Richard Conniff is a National Magazine Award winning writer whose latest book is House of Lost Worlds: Dinosaurs, Dynasties, and the Story of Life on Earth. This post originally appeared at the website Yale e360.