by Gordy Slack
Nineteenth century naturalist and writer John Muir called the Sierra Nevada Mountains the ‘Range of Light’ because of the stunning luminosity of its peaks viewed from California’s Great Central Valley. Today, Muir would be heartbroken to see his favorite mountains obscured by the thick soup of pollution that so often fills the valley these days.
The particulate constituents of that smog, which blows east from the Pacific coast’s urban areas, are generated by highway traffic and agricultural activity in the valley itself, and come increasingly from other valley-based industries as the region’s economy grows. Air pollution is effectively trapped in the valley, which is capped by an inversion above and by mountain ranges that define its north, south, and eastern edges. In winter, the inversion layer drops and concentrates the trapped and dirty air, leading to health crises for asthmatics and others with allergies and lung conditions. And it is just plain unhealthy for everyone else. According to the US EPA, five of America’s ten most air-polluted cities are in California’s Central Valley.
As he drives to work each day, Shawn Newsam, Assistant Professor of Electrical Engineering and Computer Science at UC Merced, gazes toward the Sierras twenty miles to the east. Sometimes the mountains look clear and close enough to touch, he says, and other times he cannot see them at all. If he can see the peaks clearly, he worries less about his two children playing outdoors. But if that dark haze obscures the mountains, he may recommend keeping them inside.
It is not a very scientific approach to measuring particulates in the air, acknowledges Newsam, but right now, there is really no better working alternative. With only a handful of point measurements of particulate matter for the entire Central Valley, air quality warnings just are not local enough to be of much use. “Air quality in the valley is a quickly changeable and highly localized affair,” Newsam says. A reading from Fresno, say, may not reflect the air conditions in nearby Modesto or Clovis.
Supported partly with a grant from CITRIS, Newsam is developing a network of several dozen cameras that can collect data and possibly analyze air particulates around the Central Valley. The project could provide a quick, easily accessible way to evaluate local air quality in real time.
This spring, Newsam will aim digital cameras at the horizon and analyze the resulting atmospheric images in hope of finding meaningful associations between them and the particulates in the air they depict. Broadly speaking, he will be taking two approaches, one from above the data, the other from beneath.
First, from above, he will develop models of the different light dispersal effects of different size particles. “We may then be able to automate the application of such analyses to the images for lots of real time data about both the concentration of particulates in the air and the size of the particles themselves,” he says.
The size of airborne particles matters as much as their composition, says Anthony Wexler, UC Davis Professor of Mechanical and Aeronautical Engineering and a specialist on air pollution. “Larger particles affect the health of the lung, whereas smaller ones can penetrate into the bloodstream, circulate, and affect other organs,” says Wexler, who is consulting with Newsam on the project.
But there may be a simpler way to establish the correlation between what Newsam’s camera’s record and the air particles' size and concentration. Simple data mining, Newsam says, may be a shortcut to just as potent a result. By capturing and analyzing numerous images and then comparing them to direct particulate measurements (which are much more expensive) simultaneously taken and analyzed by the San Joaquin Valley Air Pollution Control District (Valley Air District), Newsam may be able to find correlations between the two.
To accomplish this, Newsam will apply different data mining techniques, including Bayesian classifiers, decision tree classifiers, neural network classifiers, and support vector machines.
Newsam will use both linear and nonlinear regression to model the relationship between the features of the images he gathers and the particulate data he gets from the Valley Air District. Quantifiable image texture features will be used to examine the amount of detail visible on distant objects, such as mountains, as well as analyze the spectral signatures of the light scattered by the particles, possibly employing a hyperspectral camera that can take pictures at narrow light wavelength bandwidths. And he will also compare images taken using polarization filters at different angles in case there is a detectable and interpretable polarization effect on light scattered by different size particles.
One way or the other, Newsam expects his research to bear fruit. As his research is, as he calls it “investigatory, very experimental,” he does acknowledge the possibility that neither the modeling nor the data mining will result in reliable ways to draw useful particulate information from his photographs.
“The Holy Grail would be finding a quick way to draw conclusions about particulates from images,” Newsam says. “But even if I do not get that, we will still have a far more complete track record of whether visibility is getting worse or better and by how much.”
Such a record will become more valuable as the human population in the valley grows and, in all likelihood, as air quality worsens, Newsam says. And with the aid of projective computer technologies, graphic what-if scenarios could be accurately illustrated. “If a politician or bureaucrat wanted to know what the Sierra Nevada viewed from the Merced Campus would look like ten years from now given different economic and emission scenarios, we could show them,” Newsam says.
The data collected by his widely distributed cameras may serve another purpose as well. As photons become an increasingly important source of energy in the valley, the need to make accurate, short-term predictions about how much energy the sun will contribute to the grid at any given time becomes increasingly important. Newsam’s images could provide up-to-the minute projections of solar irradiance over large geographic areas and thus make inexpensive forecasts of the potential output of photovoltaic systems in the area. Newsam has approached PG&E with his research plan, and the energy company has shown interest in developing a real-time map of solar irradiance. In this way, by aiding solar energy’s reliability, reading air pollution could have the side effect of reducing it. That would please John Muir as well as Newsam’s kids, who just want to play outside.