The Berkeley Science Review: Articles

Watching the Watershed
Wireless sensor networks uncover the real water cycle (view PDF)
by Tim De Chant

"Water, water, everywhere, / Nor any drop to drink." These words, written by Samuel Taylor Coleridge in the 18th century, may soon describe the 21st. In fact, access to fresh water is already an issue of global concern. The genocide in Darfur may have brought forth underlying social tensions, but the conflict is largely the result of a drought that forced the nomadic and Islamic Baggara people south onto non-Arab farming land. Further east, the Himalayan glaciers that feed many of Asia's largest river systems could disappear in the next 20 years, leaving 2.4 billion people on the brink of disaster. Worldwide, the United Nations estimates that over one billion people already do not have access to fresh water, a number predicted to increase over the next 50 years.

Population growth is a large part of the problem, but climate change is the compounding factor that can turn water shortages into disasters. Understanding climate change is the first step in developing strategies to deal with impending crises that could threaten global supplies for drinking, sanitation, and irrigation. At the same time, even our understanding of the seemingly simple process of the water cycle is still mired in uncertainty. Scientists involved in UC Berkeley's Keck HydroWatch Center, established in 2006, are hoping to correct our sketchy comprehension of the water cycle by delving into a study so detailed that the water will have nowhere to hide.

Water world
To fully appreciate the mechanisms underlying climate change, we must first understand the water cycle and its relationship to the climate. To do this, we must be able to accurately and precisely track water's detailed movements within a given area. Years of scientific research have provided a somewhat coarse understanding of the water cycle. Evaporation from oceans, lakes, and streams, and the release of water from plants (known as transpiration) are the primary drivers of atmospheric water content, commonly referred to as humidity. As humidity builds, clouds begin to form and, when fully saturated, water precipitates out and hits the earth as rain or snow. This precipitation either runs off into streams and rivers or seeps directly into the soil. The water's eventual re-evaporation completes the cycle. This basic model is well-known, but many finer details, such as the daily water use of a single tree or how much soil moisture influences atmospheric water content, are still murky. Vital questions remain unanswered because current methods for studying the hydrologic cycle are lacking. As Jim Kirchner, professor of earth and planetary science and HydroWatch principal investigator, puts it, "It's like trying to understand a Beethoven symphony when we can only hear a note every minute or two."

Previous studies of the hydrologic cycle have fallen short by providing us with only a small sampling of the symphony of variation in the underlying processes. Published research has covered a variety of spatial scales, but with sampling at relatively low frequencies. The Keck HydroWatch project is different as it seeks to understand the big unknowns of the water cycle by sampling at very fine spatial and temporal resolutions. The project centers on two California watersheds, one in the Angelo Coast Range Reserve and another in the Sagehen Creek Field Station. Their local climate regimes are both driven by pulses of precipitation in the winter and dry summers. The coastal Angelo experiences winter rainfall followed by cool summers while the mountainous Sagehen is buried in winter snow and endures hot summers. These precipitation cycles, combined with their seasonal temperature differences, position the watersheds at the extremes of climatic variation in California. But it's the unprecedented levels of detail at which these two watersheds will be studied that place this project at the cutting edge of climatological and hydrological research. David Culler, professor in the Department of Electrical Engineering and Computer Science and another of the project's principal investigators, draws a parallel between the project and Antonie van Leeuwenhoek's discovery of single-celled organisms: "We didn't know about bacteria until we had a microscope."

Scaling the motes
Wireless sensor networks are the microscopes through which the HydroWatch project researchers hope to more closely examine the water cycle. While not a new invention, wireless sensor networks (also called sensor webs when used for environmental monitoring) have only recently become advanced enough to be used in such a demanding project. The wireless sensors are tiny, specialized computers, about as big as a stack of six half-dollar coins. Their small size and energy-efficient operation allow them to collect data at exact locations around the clock for months or even years, giving scientists unprecedented amounts of information. And because they communicate with each other wirelessly, the footprint of their installation and thus their impact on the system they monitor is reduced. The need for miniaturization of the wireless sensors' computing components, combined with the complexity of their network, would have made a successful deployment a decade ago as likely as being dealt a royal flush. Fortunately, Culler's years of experience has rigged the deck.

Culler began his work on the hardware and software necessary to power environmental sensor networks in the late 1990s. "At the time, I was working on huge complexes of machines," Culler says. It was around this same time, though, that computing devices were becoming smaller and more pervasive. "The idea was that computing is going to be everywhere," he recalls. Working with that ideal in mind, he used what parts were available and created the prototype for the first wireless environmental sensor, or mote. Each mote consists of four sensors, one each for total solar radiation, photosynthetically active radiation (the part of the spectrum to which the light absorbing plant pigment chlorophyll is sensitive), temperature, and humidity. These sensors work in concert to measure the biophysical processes that drive the water cycle. The radiation sensors monitor the amount of sunlight that works its way into various parts of the forest while the temperature and humidity sensors keep track of those changing conditions. Humidity is a complex climatic factor and of great importance in understanding the water cycle. It is driven by a combination of factors, including temperature, surface evaporation, and transpiration. Yet humidity is not simply driven by these variables—it also influences them. For example, as humidity in the low levels of the forest rises, the temperature lowers, stemming further evaporative loss and stabilizing the humidity. The straightforward question, "How humid is it today?" is thus underpinned by a host of complex processes and interactions.

The sensors' mechanics, on the other hand, are surprisingly simple. "Strangely, almost all electrical components are [technically] sensors," says Culler. The light sensors are not that different from standard solar panels found in calculators. The resistor, a small component that steps voltages up or down within electronics, sees its electrical resistance fluctuate in response to changes in temperature. Common resistors react in relatively unpredictable ways, but their design can be modified to react predictably to changes in temperature, turning a simple resistor into a highly useful environmental sensor. The humidity and soil moisture sensors similarly employ changes in electrical current, but with a twist. Rather than measuring the change in resistance through a resistor, these sensors measure the changes in electrical current as it travels across the air or soil between two terminals. Higher moisture allows for greater electrical conductance across the gap.

The relative simplicity of the four basic environmental sensors belies the difficulty in designing them for the real world. The challenges that Culler and his students faced in creating the wireless sensors were staggering. They were a completely new paradigm of computing. Their small size limited their computing power, data storage, and power availability. "That austere set of requirements is actually very stimulating," says Culler, "because you have to tackle some hard problems. Loss of data and devices are common, and you always have to deal with uncertainty and noise." Those hurdles, however, were surmounted in relatively short order. Within a few years, Culler's lab had a working hardware prototype and a new operating system, TinyOS, both of which were made open-source to allow anyone to help smooth out the kinks. Now, years later, parts for motes are commercially available and Culler is the consumer, taking both parts and software off the shelf to build the HydroWatch sensors. Each mote operates on an industry standard wireless communication protocol similar to Wi-Fi. Central towers on site collect the motes' chatter and funnel it to the Internet. A standard database then records the changes in solar radiation, temperature, and humidity at each node. Think of the collected data like the New York Stock Exchange for microclimatic conditions. Values for each measured variable constantly change––like stocks in active trading. Hundreds of thousands of data points are generated each day. The total amount of data is simply staggering. And collecting it is only half the story.

If a tree transpires in the forest...
Once the data have been recorded, making sense of the millions of data points requires careful analysis. Fortunately for the HydroWatch Center, Culler and another HydroWatch principal investigator have already experienced this data overload firsthand. Five years ago, he and Todd Dawson, a professor in the Department of Integrative Biology, collaborated on a similar project, albeit smaller in scale. Their goal was to describe the atmosphere immediately surrounding a single redwood tree, a seemingly simple goal that turned out to be far more complex than either of them imagined.

Dawson's research is focused on the question of how plants interact with and influence their climatic surroundings. One of the plants he studies is the coast redwood, Sequoia sempervirens. As the tallest tree in the world, redwoods are both slaves to and masters over the microclimate that surrounds them. They constantly lose massive amounts of water, possibly dozens of gallons a day, through transpiration; this loss, in turn, creates a negative feedback loop that raises the surrounding humidity, lowers the temperature, and slows further loss. This fine level of control is fantastically dynamic, making Sequoia sempervirens a perfect test for environmental monitoring via wireless sensor networks.

Dawson and Culler deployed 33 wireless sensors in a redwood grove for 44 days. They predicted revolutionary results. What they hadn't predicted was the staggering amount of data those 33 motes could provide in one month. "It took us eight months to distill the data, troubleshoot, pull out outliers, and figure out where we had bad, uncalibrated sensors," says Dawson. This lesson has not been wasted on Dawson and Culler as they begin the deployment phase of the Keck HydroWatch project in earnest. "There's been a real push to get the statistical community involved to help us so this doesn't happen again."

The astonishing amount of data for the microclimate around a handful of redwoods hints at the complexity of the global climate, something that Ron Cohen, associate professor of chemistry and HydroWatch principal investigator, is keen to investigate. Cohen wants to better understand the behavior of water on a global scale. For Cohen, a single watershed is an excellent starting point. "My interest in Keck starts from a technical standpoint of how we can change the way we make atmospheric observations," he says. Cohen wants to move beyond detailed measurements at a single point to extrapolate out to ever-increasing areas. HydroWatch is the first step, moving from one point to dozens, perhaps hundreds of points when the project is running at full steam.

Cohen hopes that as a result of the tools developed by HydroWatch, he can help build more accurate atmospheric models that can "predict outside the envelope of past experience and into new conditions." One global phenomenon that Cohen is investigating is cloud formation. "If you look at the Earth from space," says Cohen, "a lot of it is white because of the clouds. That's all directly reflected sunlight!" The percentage of the Earth covered by clouds is hugely important to atmospheric modelers because it dictates how much sunlight eventually reaches and heats the Earth's surface. Understanding the origin of clouds, then, is incredibly important for future climate predictions. Previous efforts have turned a relative paucity of data into system-wide extrapolations of potentially lower quality, Cohen concedes. HydroWatch promises to upend the status quo, narrowing the gap between known data and assumptions and providing scientists with more accurate atmospheric models.

Vapor action
"Water is important for life; that's not something that is new," says Inez Fung, professor in the Department of Earth and Planetary Science and another Keck principal investigator. "But being able to predict it is something else." Her work forecasting the drivers of climate change is steeped in uncertainty. Climate models suffer from uncertainties with respect to atmospheric carbon dioxide and plant growth, but they absolutely fall apart when soil moisture is added to the equation. The Keck HydroWatch project already has a startling preliminary finding from their few months in the field that may change how the models incorporate vegetation and soil moisture into predictions of water vapor.

Current climate models assume all plants in an ecosystem of the same functional type (for example, all coniferous trees) contribute equally to atmospheric water vapor. Fung, however, says plants near the canyon bottom of the Angelo site are contributing little to the total atmospheric water vapor. A drastic temperature inversion near the stream has stratified the air column, keeping the water transpired by nearby plants from escaping the canyon. "It's those poor sods uphill in the water-stressed areas," says Fung, "who have to dig their roots down to the fractured soils to get water." When they do reach water, the higher temperatures at their exposed positions cause the water they do obtain to evaporate quickly. Understanding how those plants obtain their water, from where it is obtained, and the rate at which they lose it could be a central piece of enhanced climate models.

The HydroWatch team hopes their discoveries will not only flesh out the murky details of the water cycle, but also create a set of tools by which policy makers can better manage their water resources. Fifty years ago, water resource management consisted of building dams and removing vegetation to prevent it from using up water. "As everything gets scarcer, we as a society that sets rules for managing these lands needs to decide what our priorities for different vegetation types are," says Bill Stewart, cooperative extension specialist in the Department of Environmental Science, Policy, and Management. While removing vegetation may still produce more water, it can also eliminate critical habitats. Is a slight increase in water worth three more species on the endangered species list? And what about the new plants that replace those that were removed? Stewart points out that many of the plant species that repopulate disturbed lands are excellent at sucking up water, potentially nullifying the initial increase in water production. Finally, few scientists fully understand the influence of the changing composition of the atmosphere on plants. Many fossil fuel industry groups would have us believe that higher carbon dioxide levels will make our future a greener place. Trees may benefit to a small degree, but the more likely outcome favors more water-thirsty weeds that could further diminish our already strained water resources.

A warning in the water
For accurate models to become a reality, scientists will need more than a basic understanding of climate dynamics. As Fung puts it, "I'm a scientist working on the climate problem, so that means I need to predict the carbon dioxide in the atmosphere. And I cannot predict the carbon dioxide until I know the biosphere. And I really can't know the biosphere until I know the soil moisture." All of these major components—the carbon dioxide, the biosphere, and the soil moisture—are tightly bound to one another. A small variation in one part of the system can cause drastic changes in others. The water cycle ties in to each part of the problem, so understanding its intricacies is vital to producing new climate models that allow us to react to new developments in real time. Most research to date has been retrospective and has not allowed us to fix the problem before the damage has been done. "What I hope comes out of this," says Dawson, "is a tool for monitoring environmental change in real time so that we might even be able to use it as a warning."

Tim De Chant is a graduate student in environmental science, policy, and management.

Want to know more? Check out:
bie.berkeley.edu/keck

Comments on this article? Drop us a line at with 'letter to the editor' in the subject!