Tuesday, October 27, 2009

The Challenge of Building Energy Consumption Benchmarking

For a long time now I have been advocating that building energy consumption benchmarking is not realistic without a statistically representative database. The fundamental problem with existing databases used for benchmarking (such as DOE's CBECS database) is that building use categorization is much too general and in each building use category there are an insufficient number of buildings to provide statistically meaningful comparisons. Most researchers understand the need for statistically representative data. Unfortunately, the typical response today for relying of any benchmarking database with known shortcomings is "but it's the best we have." I have always believed that it is better not to benchmark than to benchmark against a database that has not been statistically proven to be representative.
Let's look at some of the variables that certainly must be factored into any building energy consumption benchmarking analysis. Of course, the more variables there are, the larger the database that must be built to provide meaningful benchmarking data at a reasonable confidence level, and it starts right with building use. To classify building use into the four major uses, i.e., office, retail, industrial and lodging, would be to grossly oversimplify the picture as there are literally dozens of major building use categories. Each one of these major building use categories can then be further subdivided to reflect buildings with common characteristics. For example, take the office building category which is probably the simplist of building use categories to characterize. Office buildings may be constructed tall or wide, with significant square footage or relatively small square footage. There are different classes (at least three). They may be attached to other buildings or stand-alone. Buildings may have been constructed a hundred years ago or more recently (perhaps under tighter energy efficiency building codes). The buildings may be located in a warm climate or a cold climate or anything in-between. Some office buildings include retail. Some office buildings have high energy consumption facilities such as data centers. There may or may not be standby emergency generators that must be tested periodically. And all this is to say nothing about how the building is operated! For example, what are the building's hours of operation? How does the building handle maintainence and repairs? What amount of space is occupied and what amount is vacant? I can go on and on, and this is just for the "simple" office building category. While lodging has just about as many variables as office buildings, these pale against the many more variables associated with retail and industrial use!
As I said, there literally are dozens of significantly different major building uses. The entire CBECS database (which identifies only 17 different "principal" building uses) includes slightly more than 5,200 buildings to represent the entire country. In view of the wide standard deviation around energy consumption data for each building use, the number of buildings that should be included in the benchmarking database to achieve a 90% or 95% confidence level for each building use category/sub-category is significantly more than currently in the CBECS database.
What should we as an industry do about this problem? I believe that two things are a must. First, every effort must be made to develop a single building energy consumption database that will include all the major building uses (properly categorized with common characteristics) and be statistically representative for benchmarking purposes. This should be a priority of both DOE (with its CBECS database) and EPA (with its Energy Star database). Second, we must refrain from using any benchmarking database, e.g., CBECS, just because "it is the best that we currently have." By using a benchmarking database that has significant deficiencies, the results can (and will) be questioned and may do more harm than good. We as an industry must come together now on this critically important problem and deal with it once and for all.

Wednesday, October 7, 2009

How much energy do green buildings really save?

LEED continues to receive considerable attention in the media since it is the country's leading green building certification program. However, questions about the program's true effectiveness are beginning to surface. A recent New York Times article (August 31, 2009, "Some Buildings Not Living Up to Green Label") reported that a LEED-certified Federal Building in downtown Youngstown, Ohio was "hardly a model of energy efficiency." In fact, according to an environmental assessment conducted last year, the building reportedly did not even score high enough to qualify for the Energy Star label. This is not really "news" as the U.S. Green Building Council (USGBC) that administers the LEED program in a study last year of 121 new buildings certified through 2006 itself found that more than half (53%) did not qualify for the Energy Star label and 15% scored below 30, meaning that they used more energy per square foot than at least 70% of comparable buildings. One of the problems appears to be where the LEED points come from. The General Services Administration, which owns the Youngstown building, indicated in the Times article that points were "racked up for things like native landscaping rather than structural energy-saving features." The inferrence clearly is that perhaps the LEED criteria need re-visiting.

Interestingly, the USGBC also found from its own research that a quarter of the new buildings that have been certified do not save as much energy as their designs predicted, probably because the energy models used to predict how much energy a planned building will consume are inexact. Clearly, much still needs to be done to develop a comprehensive database of modelled versus actual building energy use information. Such a database would unquestionably improve building energy performance models.

LEED assigns credits before a building has been operated, but the only real way to know how a building is performing is to collect energy use data over time. Fortunately, as of this year the LEED program will require all newly constructed buildings to provide energy and water bills for the first five years of operation as a condition for certification.

The Canadian National Research Council also recently published an interesting study by Guy Newsham, Sandra Mancini and Benjamin Birt in August (Report NRCC-51142) that added more fuel to the fire. Their analysis of 100 LEED-certified buildings found that while LEED buildings used 18-39% less energy per square foot than their conventional counterparts in the CBECS database, 28-35% of LEED buildings used more energy. In addition, the measured energy performance of LEED buildings had little correlation with the certification level of the building or the number of energy credits achieved by the building at the time of design. This clearly presents a problem for building owners and operators who are not realizing the energy performance they were expecting.

These problems and issues have the potential to raise questions about the credibility of green building rating systems. This would be unfortunate as it could jeopardize the overall societal benefits that still accompany these rating systems. The fundamental problem in my view is that we are not giving sufficient recognition to the fact that there is a large variability between buildings (even between the same types of buildings in the same geographic location) and to do true comparisons at a statistically meaningful confidence level requires a much larger dataset than currently exists. As such, many of the comparisons (particularly against the relatively small CBECS dataset) leave much to be desired, and may in fact do more harm than good as they can be misleading. Unfortunately, the media, as expected, will often pick up the newsworthy conclusions of a just-published study, but fail to note any of the limiting conditions upon which the conclusions were drawn. As such, a top priority of our industry should be to build a much larger and more complete (and transparent) building energy performance dataset against which statistically significant energy consumption benchmarking can be done.