TwitterTwitterLinked InContact

NATURAL RESOURCE SCIENCE AND MANAGEMENT IN THE WEST

Supercomputer-Powered Model Improves Water Planning

Dec 23, 2014

A Hi-Resolution Hydrologic Model Peers into the Future of Western Water

Inside the University of Wyoming’s 3-D visualization cave, winter is coming. Through special glasses, a viewer watches winter snow pile up in the Wind River Mountains. The simulation shifts. Spring comes, and the snowpack begins to melt, waning gradually with warmth. Underfoot, the ground becomes translucent. Water accumulates and moves around, resurfacing as it feeds tributaries of the Green River, at times evaporating and returning further down the mountains as rainfall.

Supercomputer-Powered Model Improves Water Planning

Wearing special goggles, a man interacts with a model in UW’s 3-D visualization cave.

This holodeck-like experience of a year in water is just a glimpse into an incredibly detailed hydrologic model created by University of Wyoming engineering and environment and natural resources professor Fred Ogden. Called ADHydro, the model harnesses the power of supercomputing to create high-resolution, physics-based simulations of how water moves through very large watersheds. The project takes advantage of computing power at the new Cheyenne-based supercomputer facility the University of Wyoming shares with the National Center for Atmospheric Research.

“Right now we are simulating the Upper Green River basin in Wyoming,” said Ogden, who has been working on the project for the last two years.

As climate change alters the way water falls on and flows through western landscapes, Ogden hopes his model can help water managers better understand how those changes play out, despite the uncertainties the future holds.

 

Looking to the past to estimate the future

The importance of the Colorado River is often summed up with the following facts: The river’s basin touches seven Western states before reaching Mexico. It supplies life-giving water to 40 million people. It can generate 4,200 megawatts of hydroelectric power, and its water slakes the thirst of 5.5 million agricultural acres.

The Colorado is probably one of the most oft-studied and worried-over rivers in the country. Yet despite decades of efforts measuring and tracking the river’s flows, every year when mountain snowpack begins to roll off rocky slopes, water managers still wrestle with many unknowns: How fast will the snow will melt? How much will get used up by trees? How much will evaporate before it gets into a stream? Until fairly recently, managers looked to weather from the past hundred years to try and find analogs that would help them plan for how the snow would behave.

“We really relied pretty dominantly on the historical records,” said Jim Prairie, a hydrologic engineer with the Bureau of Reclamation’s Upper Colorado Regional Office.

In the 2000s, though, as drought spread across the West, scientists and water managers realized they might need to look further back in time to understand was happening to the river. They turned to tree rings, which tell a story of bigger floods and longer droughts in centuries past.

“That allowed us to get a much broader picture of how hydrology may be changing in the future,” said Prairie. “What we have seen in the last 1,200 years is we could have higher and lower flows [than in the past 100].”

This knowledge gave water managers a better sense of what the snowpack, and the river, was capable of. Climate change, though, adds another wrinkle. The challenge, said Ken Nowak, the Bureau of Reclamation hydrologic engineer for the river’s lower half, is that climate change means the next 100 years may not look like the last 100, or even the last 1,200.

“The past is no longer indicative of the future,” said Nowak.

That’s where models come in. To start, water managers get results from a set of climate model projections called the Coupled Model Intercomparison Project. These coupled models combine representations of the atmosphere and the oceans, as well as the ice sheets, sea ice, and the land surface, painting a fuller picture of the Earth’s climate. For the purposes of predicting the future of Western water supplies, such models produce two key pieces of information: projected future temperatures and precipitation.

Of the two, changes in future temperatures are more predictable. The physics underlying what happens when carbon dioxide concentrations increase, as they have been since the Industrial Age, are pretty simple: add more CO2, the atmosphere heats up. While the amount of warming won’t be the same everywhere, the models show that the Rocky Mountains of the future will be warmer than they are now.

Supercomputer-Powered Model Improves Water Planning

Inside the National Center for Atmospheric Research (NCAR) Wyoming
Supercomputing Center.

Precipitation is more fickle. Scientists know a warmer earth will hold more moisture in its atmosphere. But since there is not a direct relationship between temperature, CO2, and precipitation, the question of where that moisture will go is more difficult to answer. Most researchers think wet areas, like the tropics, will tend to get wetter, and dry areas, like the desert Southwest, drier. The future of many places in between, like the Rocky Mountain region, is basically a toss-up at this point.

The coupled models run under a range of scenarios. One keeps the world on a high-emissions path, where carbon dioxide levels continue to increase rapidly. Others take a middle road. There is also a low-emissions option, to include the possibility of global action to drastically reduce emissions or capture carbon dioxide from the atmosphere. Each scenario results in different futures, some a degree or two warmer, others many degrees warmer, with corresponding changes to rain and snowfall regimes.

To get to the localized impacts of these climate changes on water, researchers plug the results of the climate models into hydrologic models representing all kinds of natural physical processes: snowpack, how that water gets taken up by trees, whether it evaporates or soaks into soils, how fast it makes its way downstream.

 

Hi-resolution hydro

Ogden’s ADHydro effort represents a significant improvement over existing hydrologic models. Many past models, including one of Ogden’s now used by the Army Corps of Engineers, use what’s called a square mesh to represent a watershed. Essentially, the landscape and processes are divided up into small squares representing a given area. This gives you a standard resolution across the entire watershed.

That works well for smaller watersheds, said Ogden, but he’s trying to represent something much larger—first, the Green River watershed, and later the entire Colorado River basin. So ADHydro uses a different approach.

“We are using an unstructured mesh, triangles, and they can change size,” he said.

Since a whole lot of computing power is needed to run such models, this approach is more efficient. In the mountains, where a lot is going on in a small space, the model has tiny triangles, 50 meters on a side. It can represent processes like snowmelt, soil moisture, and groundwater recharge on a very detailed scale, where even subtle changes can have big effects downstream.

When a river gets to the plains, where not a lot is changing, that high resolution becomes superfluous. The ADHydro model then can bump its triangle mesh size up to 500 meters on a side, saving valuable computing resources for the areas that truly need it. Even with a supercomputer running the model, such high resolution uses a lot of processing power, so Ogden has tried to make it as efficient as possible. The model also represents processes in three dimensions, but saves resources by being “quasi-3-D,” said Ogden.

“The models that we are trying to improve upon … tend to run at lower resolution. And because of that they don’t really simulate feedbacks properly, particularly between groundwater and surface water,” he said.

In contrast, the ADHydro model includes real world processes such as the effect of wind on sublimation, or snow evaporation. Other models can’t represent this, so they do something called parameterization, “a fancy word for fudging it,” said Ogden. And, in addition to simulating water, air, and soil, the new model includes layers like tree cover. Since trees use a lot of water, whether they are alive or dead, growing or not, can have a big impact on water availability.

Kristi Hansen, a water economist at the University of Wyoming, has been working with Ogden to create a layer for his model that will represent decisions water managers make about storing and releasing water. Management scenarios could include information about a city’s rights to Colorado River diversions and its projections of future metropolitan growth. Such inputs allow researchers and policymakers to play around with different configurations of water use in the future. Say Las Vegas cuts its water use still further, but demand from a rapidly growing Phoenix jumps.

Hansen is excited about the model’s improved resolution over existing ones, saying when there are shortfalls on the Colorado River, “the finer scale has the potential to tell us more about who is affected and by how much…a better, more detailed understanding of the hydrology can help us to make good decisions in the future.”

Ogden imagines a case in Wyoming where the state might need to make a decision on a new pipeline to take water from the Green River to Colorado’s Front Range. The State Engineer’s Office could use his model to find out what granting approval for the withdrawal would mean.

“They [the engineers] could pull open the Web browser, say we are going to divert this amount under these rules, and then select a future climate scenario and maybe a future land use scenario, and then click ‘run.’ And then sometime later they would get an e-mail back with a report telling them what the effect of that was on, say, the water level of Lake Powell in the future.”

By summer 2015, the model will be running simulations on the National Center for Atmospheric Research supercomputer, using output from coupled global climate models to model water behavior in large western watersheds, said Ogden.

 

From models to action

No matter how good your model, moving from improvements in research models into improvements in the actual operation of a complex system of dams, diversions, and cities in the arid West is a monumental task. As a hydrologic model gets layered on different climate model outputs, each with its own assumptions, the view of the future actually becomes fuzzier, said Jeff Lukas, a senior research associate with the Western Water Assessment. That’s because whenever multiple modeling steps are used, the range of possible outcomes broadens.

“The uncertainty increases because you have made these choices along the way,” said Lukas.

Supercomputer-Powered Model Improves Water Planning

Lukas points to local and regional efforts by utilities like Denver Water and others. With information from climate and hydrologic models, they are improving their flexibility to react to various circumstances, preparing for whatever the future brings.

Laurna Kaatz, the climate policy analyst for Denver Water, which serves a quarter of the state’s population, thinks a lot about the future. She said climate models and hydrologic models give managers insights into how the region might change.

“They help us play out ‘if’ scenarios. If this happens, in the future, then this is what it could mean to the water system.”

Even though models have uncertainty and do not predict the future, it’s still “really valuable information,” she said. “If we didn’t have these than we would just be making assumptions about what the future could be.”

Lukas agrees this is where models come in handy.

“Really, the important thing is not to be surprised,” he said. “To have at least some ability to prepare for an event you haven’t seen before.”

Ogden acknowledges that exactly predicting future flows “is almost impossible.” But he says ADHydro is an improvement over current options. He notes a 2012 effort from the Bureau of Reclamation looking at the future of the Colorado River that painted a “pretty dire picture” of future Colorado River water availability. ADHydro could be used in similar sorts of research.

“What we hope to do would to be able to improve on those because we are including more feedback processes because of our higher resolution,” Ogden said.

Denver Water’s Kaatz agrees.

“Advancements in modeling are really important. Because it helps us better understand the system that we are trying to manage.”

As snow coats mountaintops this winter, water users and water managers have begun their anxious watch, tallying inches and snow water equivalent, checking weather stations and SNOtel sites. No one knows what this winter, or the next, or the one 30 years from now, will bring. With models like ADHydro though, we may at least be better prepared for that uncertain future.

By Stephanie Paige Ogburn

Stephanie Paige Ogburn reports on science and environment in the West from Denver, Colorado. Find more of her work at stephaniepaigeogburn.com.