Show off your holiday lights and you could win an iPad! Enter your photo by December 13. Winner will be selected by popular vote.
Monday, July 29, 2013
In the power generation industry, engineers frequently use simulation tools to predict the behavior of various pieces of power generation equipment.
For electrical problems, we use circuit simulators based on fundamental laws such as conservation of charge and conservation of energy. They are exceptionally accurate.
For cooling problems, things are more difficult. Heat transfer simulations include exact laws, such as conservation of energy, but also include inexact parameters that approximate imperfectly understood phenomena, such as airspeed variations between heat sink fins.
For problems where air is being driven steadily by a fan or a blower, we can usually predict temperature rises within 5 percent, but for problems where airflows are dominated by the buoyancy effect of heated air, we’re lucky if we can predict temperature rises within 30 percent. And if erroneous modeling assumptions are included, discrepancies can become arbitrarily large.
The climate simulations used to study anthropogenic global warming are vastly more complicated than our heat transfer simulations. In addition to airflows being completely driven by buoyancy effects, climate simulations include hundreds of parameters that attempt to approximate the effects of water phase changes (precipitation and evaporation); time-varying heating caused by a moving sun and varying cloud cover; chemical reactions occurring within plant life, the ocean surface, ground-level smog, the ionosphere and more. The models are stunningly complex.
Facing such complexity, can we trust the results of the various climate simulations out there? If we had the ability to run multiple planetary-level experiments on multiple atmospheres, we could quickly determine their accuracy. Alas, we cannot. We have but one Earth with one atmosphere with no way to run experiments.
Another way to judge simulation quality is to see how closely different models from different teams match. For electrical problems, I’ve seen spreadsheet models and sophisticated time-domain circuit simulators match exactly. Results can be trusted. With climate models, though, we’re not so lucky.
The climate simulations used by the Intergovernmental Panel on Climate Change yielded temperature rise estimates that covered a wide range, with the coolest predicting 2.7 degrees Fahrenheit by the year 2100, and the hottest predicting 10 degrees. Climate scientists elsewhere have created simulations that predict rises as small as 0.4 degrees and as large as 20 degrees. The range of predictions covers a 50-to-1 ratio. And for sea-level rise estimates, some simulations have predicted rises approaching 20 feet, while others predict less than an inch: a 300‑to-1 ratio. Simply put, results are all over the map
What are we supposed to do in the face of such uncertainty? One common trick is to treat the various simulations as statistical samples, but this isn’t appropriate. There is no guarantee that the simulation results will be centered around the true answer. (This is important to note: There is a true answer. All simulations that fail to match it are erroneous to some degree.) If the simulations all share a common modeling error, they will all err in the same direction. And the “spread” seen in this data set is due to different simulations containing different modeling errors, as opposed to true variations seen in the climate over time.
Because of this extreme spread in predictions, we’re reduced to lumping the various results into three basic categories:
n The first are “cool” results, where expected warming is so mild that no harm will come. If the true answer falls into this range, we can safely ignore the problem.
n At the other end are “super hot” results, where the planet has already been pushed past a “tipping point” and is in feedback-driven thermal runaway, beyond the aid of CO2 mitigation efforts. If the true answer falls into this range, restricting carbon emissions is pointless, and we should instead stockpile food before the upcoming biosphere collapse. Seriously.
n And between those two are “medium” results, where CO2 reduction could actually make a difference.
Into which category does the true answer fall? I do not know. I’m not a climate scientist. Just an engineer who occasionally correlates simulation with experiment. However, based on the extremely wide range of results coming out of the climate modeling community, I strongly suspect that climate scientists don’t know either.
Their inability to provide a consistent answer clearly indicates an immature science. Before embarking on costly carbon restriction strategies that disproportionately hurt America’s poor, we should understand the science better. The proposed Environmental Protection Agency crackdown on power plant CO2 emissions is premature.
Weather JournalEarly mix, then ice storm Sunday