In 2011, regional planners in Virginia found that the way the Bay Program’s computer models assigned land use often didn’t reflect what they saw when they looked out the window.

For example, the model said that there were confined animal feeding operations in the resort city of Virginia Beach.

“It’s true, there are some hog operations down along the North Carolina border, but at Lynnhaven Inlet where the land use was called a CAFO, the closest thing we could find was Captain George’s Seafood Buffet,” said James Davis-Martin, Chesapeake Bay coordinator for the Virginia Department of Environmental Quality.

“This may sound funny, but it shows how ‘off’ the model could be in describing local land use,” Davis-Martin said.

This was not an isolated example. There were many other places in Virginia — as well as in other Bay states — where the model said one thing about how land was being used when local planners knew otherwise.

And it was important, because the model assigns each land use — such as a residential or commercial development, pasture, forest or wetland — an estimated runoff rate for nitrogen, phosphorous and sediment into local rivers and streams. These are the key pollutants that have degraded Bay water quality.

The disconnect between the modeled world and local knowledge has consequences under the Chesapeake Bay Total Maximum Daily Load, or pollution diet. Each state has developed watershed implementation plans describing how it would reduce nutrient and sediment pollution from each land use to meet TMDL goals. Those plans are tied to real-world investments in wastewater treatment plants, stormwater management and programs to improve agricultural practices and septic systems.

“As Maryland was completing its Phase II watershed implementation plan, we heard from our local jurisdictions that one of their biggest concerns was the level of accuracy in the land use data,” said Lee Currey, the Maryland Department of Environment director of science services. “They told us, ‘our information is different from what’s in the Bay model; how can we trust the results?’ ”

Local governments wanted confidence that reduction measures they were planning were adequate, but not overly stringent — and therefore unnecessarily expensive.

To answer these concerns and to take advantage of technology and new data, the state-federal Bay Program partnership is building a new, more accurate picture of the watershed’s widely varying land uses. It will include more precise data and more local information obtained from cities, counties and townships across the watershed.

And, to prevent things like CAFOs that turn up in cities, the Bay Program is also asking local governments to review the updated data before it is used in new modeling. Bay Program officials hope that this two-way process will build confidence in the model, which is used to measure cleanup progress, as well as provide useful information to local governments.

This confidence, and trust in the model and the validity of the data it uses, “is critically important,” said Rich Batiuk, associate director for science with the EPA’s Bay Program Office, “because the work that needs to be done to restore the Chesapeake Bay is now more than ever going to require collaboration at the local level where most land use, conservation and restoration decisions are made.”

In the past, the Bay Program always tried to use data sources that were consistent throughout the Bay watershed, said Peter Claggett, U.S. Geological Survey geographer and coordinator of the Bay Program’s Land Use Workgroup.

Because of that, he said, the Bay Program was unable to use locally derived data. “But, there’s been a major shift in the thinking — now we want to use the best data we can get.”

And who has the best land use data? Usually, Claggett said, it’s the localities.

Evolution of Bay models

Implementing TMDLs under the Clean Water Act involves the use of models to estimate the amount of a pollutant that a waterbody can assimilate and still meet water quality standards agreed upon by the states and the EPA. This amount is called the Total Maximum Daily Load, or TMDL. Developing a TMDL involves defining a numeric limit on the amount of pollution that can reach a water body, like the Bay.

In the case of the Chesapeake Bay TMDL — where the “impaired” water body receives water and discharges from a 64,000-square-mile drainage basin — the only way to determine necessary reductions of nitrogen, phosphorous and sediment was to create a suite of interconnected models and datasets that describe the Bay, its tributary rivers and the lands they drain.

There is one model that describes nitrogen deposition from the atmosphere; another that describes the circulation patterns of the Bay. Other models describe river flow and erosional processes, and another predicts how land use will change. The models are coupled with real-world data such as precipitation rates and land types and uses, and the nutrient and sediment output is matched against water-quality monitoring data from selected points around the Bay and its watershed to help ensure model results match actual observations.

The TMDL created in 2010 for the Bay was based on a Chesapeake Bay model (as most refer to the collective suite of models and data sets) that many agree is one of the most sophisticated in the world. But while the models do well at assigning pollution rates from major tributaries, they lose accuracy when trying to describe what is happening at the county or city scale.

“Land use and precipitation patterns are probably the two most important inputs to the suite of Chesapeake Bay models,” Batiuk said, so getting the best data is really important.

For consistency across the seven watershed jurisdictions, the Bay Program has historically relied on nationally available imagery to develop its land cover data.

The version of the model used to develop the Bay TMDL utilized land cover data derived from satellite imagery accurate to a 30-meter resolution.

The land cover — which refers to what actually covers the ground, such as impervious surfaces, tree canopy and herbaceous cover — is digitized so that it can be measured and counted. These land cover types fall into broad categories such as agricultural, urban and natural, which are then more finely classified into ways the land is actually used — the “land use” categories. The model used for TMDL planning had 31 land uses, such as “harvested forest,” “low-intensity developed pervious,” “hay without nutrient (applications),” and “nursery.” Each land use has its own assigned rates for nutrient and sediment pollution. Each use also has its distinct cleanup opportunities.

But the 30-meter resolution can’t accurately detect urban tree canopy, lawns in small yards, narrow roads and small buildings — and modelers had “to make broad inferences about the presence of these features based on U.S. Census Bureau and other sampled data,” according to Bay program officials. As a result, estimates fell short when put to the test by the states — and their local government partners — trying to develop local cleanup plans.

“There was a lot of criticism, and much of it was valid,” said Karl Berger, principal environmental planner for the Metropolitan Washington Council of Governments.

“Land use is an easy thing for people to understand. You can stand in a place and know what the land use is, so it’s easy for people to focus on,” Berger said. “As a result, a lot of dissatisfaction about the whole Bay TMDL cleanup process coalesced around the accuracy of the model.”

The dissatisfaction was understandable, said Jenny Tribo, senior water resources planner with the Hampton Roads Planning District, which includes local governments that operate their stormwater systems under federal permits. “When it became clear that the Bay TMDL was going to be used to establish allocations under stormwater permits, the permitees were just not comfortable with these discrepancies.”

The Bay Program took the criticism to heart. In 2012, it convened a workgroup consisting of local and state planners, modelers and others whose charge was to create accurate land use data across the watershed for all jurisdictions. The workgroup was to find and use the best available data, ensuring that “the land use data used in the watershed model is perceived as relevant at the local government scale.”

Higher-resolution data

One way to make the model more relevant is to use the same data that local governments use for planning or for implementing programs like stormwater utility fees. Even in 2010, larger cities and conservation organizations used land cover data at resolutions finer than 30 meters — but these data weren’t available watershedwide. And it was expensive to acquire, maintain and replicate.

But spurred by the potential impacts of inaccurate data, in 2012 Virginia tasked its own Virginia Geographic Information Network (VGIN)— which maintains statewide geographic data — with obtaining better land cover information consistent with the Bay Program model. Virginia set out to use aerial imagery acquired by planes at altitudes considerably closer to the ground than satellites.

They also used LiDAR [Light Detection and Ranging] imagery that can measure the height of objects at accuracies of up to 30 centimeters. Together, these high-resolution data can help determine whether an impervious surface is a house or a road — and accuracies of 1-meter or less can be achieved.

Virginia was not alone in developing land cover data for localities. The Chesapeake Conservancy, a nonprofit organization based in Annapolis, was developing the methods to efficiently transform high-resolution aerial imagery and LiDAR into the same land cover classifications used by the Bay model.

In 2014, the Bay Program accepted their proposal to cover the whole watershed, and the conservancy staff is poised to deliver the data — except Virginia, which is providing its own — starting in January 2016.

But to make land cover data useable by the Bay modeling team, it has to be classified into land uses. That final link is made possible by integrating the high-resolution land cover data with other information, like U.S. Department of Agricultural census data, which can help the land use analysts determine whether, say, undeveloped grassland is pasture or suburban lawn. If it is pasture, modelers assume a certain amount of runoff with a certain amount of nutrients and sediment — and how much pollution reduction a specific management practice can achieve. Likewise for lawns.

It is at this phase — land use classification — where local information can be most useful. A county’s own land use, zoning and planning maps can help classifiers distinguish between pasture and lawn. “We’re asking for things like sewer service areas, parcel data and the county’s own land use maps,” Claggett said. These provide

the context necessary for the Bay Program to more accurately classify land use.

And to make sure the land use classifications are as accurate as possible, the Bay Program officials are asking localities to review the way they’ve classified land so that by the time the modelers start working on the updated suite of Bay models in late 2016, they are using the best available information — information that local governments have endorsed.

Based on the output of those models, the Bay Program will refine nutrient and sediment goals at the end of 2017.

“We’re providing several opportunities along the way for localities to review the land use data,” Batiuk said. “What we want to end up with is locally reviewed and approved data.”

Call for local data

The Bay Program and its state partners have started asking local governments to provide the most updated land cover and land use information they have — and local officials will be asked to check whether the program’s interpretation of land use is accurate. Though the Bay Program’s work team led by Claggett is coordinating the outreach, each state has its own system for contacting local governments. Maryland’s Department of the Environment, for example, is taking advantage of already established lines of communication between the state and the 23 county-based watershed implementation teams.

Pennsylvania is working through the County Commissioners Association of Pennsylvania and other state-level organizations to request local data and alert staff that a “data review” will soon be asked of them. In Virginia, VGIN is using an existing network between the state agency and local governments.

These efforts are in stark contrast to earlier Bay Program efforts to obtain and use local data as the watershed implementation plans were being developed in 2011, said Mary Gattis, coordinator of the Bay Program’s Local Government Advisory Committee. Counties sometimes viewed data requests from the Bay Program with suspicion, unsure whether sharing the data might have unintended consequences, because the model process was not well understood. Now, said Gattis, “we’re trying to help local governments understand that they have a role in ensuring that the model is as accurate as it can be.”

The land cover maps being developed by the Chesapeake Conservancy (and by Virginia) are going to be free to local governments and the public, as will be the land use classification developed by the Bay Program. These data would normally cost a county in the range of $30,000 to $70,000, Claggett said.

Some counties, especially the less urbanized ones, don’t have the resources to obtain the kind of land use maps needed for planning watershed improvements. VGIN’s Dan Widner said that there are a number of governments “chomping at the bit” to have the improved Virginia land cover map, so that they can use it in their local planning. “They’ve never had this level of detail — down to 1-meter resolution — and are excited to get it.”

Malcolm Derk, commissioner of Snyder County, PA said that his planning department consists of all of one person. He expects that the new land use data will be helpful for the county to target places where buffer restoration is needed and to encourage more voluntary agricultural practices.

Tex Weaver, GIS Manager for Albemarle County, a locality that has had high-resolution data since 2009, is also looking toward the future. “Our hope is that this new data will help us evaluate changes since the last time we classified land use in the county — changes like deforestation and development.”

To to submit local land use data and review the Bay program’s land use classification, contact Quentin Stubbs, 410-267-9853, or see