Geologist Andrew Elmore and his colleagues tromped up mountain stream channels to record data about the locations of their headwaters. They merged that information with other data to build a new, detailed model of small and buried streams in Maryland west of the Chesapeake Bay. Photograph, UMCES/Cheryl Nemazie

TO PROTECT STREAMS FLOWING DOWN towards the Chesapeake Bay, you sometimes have to journey up to the mountaintops.

That's what Andrew Elmore and his colleagues did. Again and again, in dozens of Maryland forests, the scientists scrambled uphill, tracking the course of small streams. It was part of a labor-intensive effort to build a novel, detailed map showing Maryland streams not recorded on other maps. By mapping the headwaters of small streams near the tops of forest ridges and hills, the researchers worked to create a computer model that predicts the locations of small streams across all of Maryland west of the Chesapeake Bay. The model offers a tool to protect streams from development and to improve the region's water quality.

These small headwater streams are easy enough to ignore — many are small enough to step over as you walk through a forest. But knowing their locations is important for several reasons. First, they are important pockets of biodiversity. Biologists have found that relative to larger streams, the smaller ones support a more diverse array of aquatic species, like fish and insects. The mix of species can be different from stream to stream — and this diversity can be easily lost when construction of new homes and roads fills in or buries a stream.

Small streams are also important for water quality. Compared to large streams, more of the water in small streams makes contact with the soils in the channel. And more leaves are washed into small streams relative to the amount of water there. These conditions help to promote a biological process, denitrification, that removes nitrogen from the water. Mile for mile, small headwater streams are the most efficient at this among all streams. However, small streams are also the most likely to be filled in or buried by construction and development.

Elmore, a geologist, became interested in mapping streams after he came to work in 2006 at the Appalachian Laboratory, part of the University of Maryland Center for Environmental Science (UMCES), in Frostburg.

While developing a map of buried streams in Baltimore, he and a colleague, Sujay Kaushal of UMCES, noticed that many of them were not shown on the National Hydrography Dataset, an existing, widely used, nationwide database about surface waters. That database was created in the 1990s after streams in areas like Baltimore were already buried. They also noticed that small streams in non-urban areas weren't included on the NHD map, either.

Seeing an opportunity to create a more detailed map, Elmore used funding from Maryland Sea Grant to begin mapping streams across Maryland west of the Chesapeake. It was, Elmore says, "the most ambitious attempt yet to model stream networks over a large region."

For this project, his scientific collaborators were Steven Guinn and Matthew Fitzpatrick of the Appalachian Laboratory and Jason Julian, now at Texas State University. Their approach was based on a fundamental idea: to know where buried streams might be located today, you have to know what the Maryland landscape must have looked like centuries ago, a landscape crisscrossed by streams, before suburbs and houses spread across the state. Parts of Maryland provide clues about that seemingly pristine landscape: the state's remaining forests.

The new model of stream locations, created with funding from Maryland Sea Grant, shows previously unrecorded Maryland streams (blue), adding detail to stream maps previously created as part of the National Hydrography Dataset (black). In this map, forested areas are colored green, agricultural lands are yellow. Map, UMCES Appalachian Laboratory

The researchers figured that if they devised a computer model predicting where streams flow today within those Maryland forests, they could use the same model to accurately predict the location of stream channels elsewhere, including in non-forested suburban and urban locales where houses and roads now stand. They could create a road map to find small and buried streams.

To create that computer model, Elmore says, the scientists had to collect several types of data. They searched existing sources of information about terrain elevation and slope and soil characteristics, features that can indicate the presence of streams.

But they needed other information that was missing: the locations of a sample group of "channel heads" where stream headwaters begin. To map those headwaters, they had to drive and hike to the tops of mountains, like Dan's Mountain Wildlife Management Area in Alleghany County. And to the tops of hills, like those in Prince William Forest park in Virginia. Elmore describes how the researchers literally followed the evidence.

"At each forested watershed, we would work in groups of two to three people to map," Elmore says. "We would start at a rather large stream and walk upstream to the first confluence. One of our group would start following the smaller stream, still walking uphill, and the rest of us would keep walking up the larger stream until we found another confluence and another small stream to walk up. The walking continued until we reached the channel head — channel heads are the most uphill evidence of a stream channel. When we found this location, we recorded the GPS coordinates."

The trio ended up often having to make their way through dense vegetation when there were no established trails to follow, Elmore recalls. "To make the bushwhacking easier, we only mapped streams in the spring, winter, and autumn, when undergrowth vegetation was sparse." In all, the scientists walked about 85 miles of stream length, and they found more than 250 channel heads. The new data came at some cost: they often emerged from the forest with many small cuts on their legs.

When the scientists put all of this data together, their computer model predicted the paths of streams as they flowed downhill from upland, forested areas. Elmore and his colleagues then extended the model to predict where streams would probably flow today across all of Maryland west of the Bay, including in non-forested areas. "The map is really a map of what the stream network would look like if the entire landscape had the same land use, and land use history, as our forests," Elmore says. In all, the map covers 23,000 square miles including the Potomac River watershed and five smaller watersheds.

To check the model's accuracy, the scientists used existing field data about the actual presence or absence of streams in more than 10,000 locations in Maryland. Eighty-four percent of the model's predicted stream locations correctly identified an actual stream, an improvement over the existing, nationwide map. Only 55 percent of stream locations shown in the National Hydrography Dataset (NHD) were correct.

Elmore's model filled in blank spots on the NHD map with many new, thin squiggly lines representing the probable location of streams. In some portions of Maryland, the "stream density" in Elmore's model (measured as kilometers of stream length per square kilometer) is 2.5 times the density shown in maps of the same area based on NHD data.

The differences between the two maps reflect several differences in how they were created, Elmore says. The U.S. Geological Survey (USGS) produced the NHD in the 1990s using aerial photographs, he notes. What's more, for the purpose of creating stream maps, the USGS defines a stream as a body of flowing water that contains water most of the time. However, some of the smaller streams shown in Elmore's model may flow only intermittently — during spring rains, for example.

Eventually, the national NHD dataset will evolve to include a higher level of detail similar to that in Elmore's model, says Jeffrey Simley, a USGS cartographer. "The only thing holding us back is the lack of funding for such development and the need for such detail in many parts of the country," he says.

Elmore and his colleagues published a description of the model in 2013 in the journal PLoS One.

Maryland has a lot of freshwater, non-tidal streams and rivers — more than 19,000 miles. And many streams in Maryland, and throughout the Chesapeake watershed, are in bad shape. The Maryland Department of the Environment issues a biennial report about the state's waterways and their compliance with federal water-quality regulations. In 2014 the department estimated that about half of the state's stream and river miles violated at least one of the water-quality standards — for example, to support healthy populations of fish and other aquatic life.

The model is "an important tool for improving our understanding of how to keep the Bay clean," says Christine Conn, director of the integrated policy and review unit of the Maryland Department of Natural Resources. "If we don't know where these streams are, we have difficulty managing the resource, both for conservation and restoration."

The department has begun using the model to review impacts from proposed construction projects and to identify small streams that are habitats for brook trout, which the agency manages. Conn adds that her agency may incorporate Elmore's data into its next update of the maps used to create its GreenPrint tool, a statewide map that identifies lands and watersheds with high ecological value as priorities for conservation.

Elmore says that the model could help inform decisions about where to locate conservation projects, such as artificial wetlands, to maximize benefits. "If you're looking at a broad region, you don't want to cluster all your restoration projects in one area, you want to distribute them on the landscape," he says.

Officials in several Maryland counties have contacted Elmore about using the model to help them comply with new rules, called TMDLs (Total Maximum Daily Loads), intended to improve the Bay's water quality. The model could help inform where to plant stream-side buffers of trees to help remove nutrients and sediments from runoff. The model "opens up the amount of land where we could potentially plant buffers to meet those TMDLs," he says.

And, Elmore jokes, "If you're in the business of water-proofing people's basements, I can give you a great map of who to send fliers to." The researchers posted the stream map on a website (http://streammapper.al.umces.edu) that shows local streets superimposed. Zoom in at the block level, and you might notice a buried stream running near or under your house.