In North Carolina, heavy rainfall from Hurricane Floyd in 1999 revealed limitations in the state’s flood hazard data and maps. The majority of the state’s FIRMs were at least 10 years old, with many maps compiled from approximate studies in the 1970s without detailed H&H (hydrologic and hydraulic) analyses. With FEMA’s limited budget, North Carolina received on average only one updated flood insurance study for one county per year. After Hurricane Floyd, it was clear that most of North Carolina needed to be re-mapped digitally, consistent with FEMA’s Map Modernization Plan, using improved H&H modeling and analyses that define current flood risks with greater accuracy. Over 50 counties needed to be remapped as soon as possible.
In 1968, the U.S. Congress established the National Flood Insurance Program (NFIP) in response to escalating costs to taxpayers for flood disaster relief. The goal of the NFIP, which is administered by the Federal Emergency Management Agency (FEMA), is to minimize future flood damages. Compliance with NFIP regulations saves an estimated $770 million annually in building and contents damage. Accurate FIRMs are key to the success of this program.
FIRMs are used for floodplain management, for regulation of new construction and for determination of flood insurance requirements. Since the beginning of the NFIP, FEMA has produced FIRMs for over 18,000 flood-prone communities. However, most of these maps were produced with manual cartography and often lacked the rigorous H&H analyses needed for detailed studies. Furthermore, minimal funds have been available over the years for updating FIRMs as natural and man-made changes in the floodplain alter flooding conditions.
As part of FEMA’s Cooperating Technical Partner (CTP) initiative, North Carolina entered into a Cooperating Technical State agreement whereby the state will assume primary ownership of, and responsibility for, the NFIP maps for all North Carolina communities, to include resurveying the state, conducting flood hazard analyses, and producing updated, digital FIRMs (DFIRMs) as shown on page 24. DFIRMs allow the latest GIS technology to be used, facilitate updating in the future, and allow on-line distribution. Furthermore, the new DFIRMs include flood hazard data overlaid on digital orthophoto quarter-quads (DOQs) that allow homeowners to clearly recognize their homes relative to floodplain boundaries. The state provided the majority of funding for the project. The North Carolina Geodetic Survey (NCGS) serves as the state’s technical lead for the surveying and mapping phases of the total initiative, and Dewberry & Davis LLC serves as FEMA’s Map Coordination Contractor (MCC) for this project.
Digital Elevation Data for Hydrologic and Hydraulic (H&H) ModelingHydrologic models predict peak discharges for standard flood events (10-, 50-, 100- and 500-year floods) over entire watersheds. These discharges are predicted from rainfall, flood routing and storage and watershed characteristics; e.g., land cover, soils and terrain slope. When the rain comes, some water will soak into the ground and some will run off, depending on soil type, soil moisture content (saturation), vegetation cover, terrain slope, etc. Whenever humans cut down trees, construct buildings, and pave streets or parking lots, more water will run off in a given time period than in the past; therefore, peak discharges typically increase from such human activity, even if rainfall amounts are the same. The routing of runoff depends on terrain slope and aspect of all streams in a watershed. For such calculations, the resolution and accuracy of uniformly spaced Digital Elevation Models (DEMs) or Triangulated Irregular Networks (TINs) do not need to be high. The final product of a hydrologic model is peak flood discharges expressed in terms of a certain number of cubic feet per second at key points along streams in the watershed.
For the second half of the H&H process, hydraulic models compute flow velocities and elevations predicted from hydrologic model discharges and channel and floodplain cross-section characteristics (e.g., area, slope, vegetation roughness), and information on bridges, culverts, dams and levees. Water-surface elevations are used to develop flood profiles and flood boundaries for standard flood events (10-, 50-, 100- and 500-year floods). For hydraulic modeling, accurate TINs are needed to include breaklines where there are distinct changes in slope, such as at the tops and bottoms of stream banks.
When North Carolina first advertised for firms to perform the new Flood Insurance Studies, no technologies were specified. The focus was on high-resolution and high-accuracy digital elevation data suitable for semi-automated H&H modeling. All firms proposed using LIDAR (Light Detection and Ranging) data instead of photogrammetry to generate the TINs and DEMs, although some proposed photogrammetry to generate the breaklines, which are difficult to determine from LIDAR data alone. The two winning teams were headed by Watershed Concepts (Greensboro, N.C.) and Greenhorne & O’Mara (Raleigh, N.C.). Watershed Concepts used Earth Data International (High Point, N.C.) to perform LIDAR/photogrammetric operations, and G&O used 3Di EagleScan (Wilmington, N.C.). Watershed Concepts is using ESP Associates (Charlotte, N.C.) as the ground surveying subcontractor, and G&O is using its own surveyors as well as McKim & Creed (Wilmington, N.C.), and Hobbs, Upchurch & Associates (Southern Pines, N.C.).
Both LIDAR firms were tasked to generate TINs with a vertical root mean square error (RMSE) of 20 cm in coastal areas and 25 cm inland. This is approximately equivalent to elevation data accuracy suitable for contours with intervals of 2.16 and 2.70', respectively. Of course, photogrammetry could be used to generate such contours or equivalent TINs, but the (low altitude) data acquisition and photogrammetric compilation would have been lengthy and very expensive.
The 20 to 25 cm RMSE requirement was a compromise from FEMA’s 15 cm LIDAR standard, partly because no LIDAR data acquisition to date has been able to achieve a 15 cm RMSE vertical accuracy in dense vegetation, and partly because FEMA has a four-foot contour interval standard when the topography is mapped conventionally. The 20 to 25 cm RMSE requirement may also prove to be unachievable in the forested areas of North Carolina, but this requirement was established with the intent to obtain the highest accuracy realistically achievable, considering the large project area, aggressive time schedule and costs.
The LIDAR ApplicationLIDAR uses thousands of laser pulses per second to accurately measure distances to ground targets from the aircraft. Each LIDAR pulse is a laser beam of light, perhaps 1 to 3' in diameter, that is reflected back (referred to as a “return”) each time it hits an object. When a laser pulse first hits a treetop, for example, the “first return” would provide the 3-D coordinates of the treetop. However, part of that same laser beam proceeds through the tree and hits intermediate targets such as tree branches. The “last return” is hopefully the ground where the beam can proceed no further. Some LIDAR sensors record either first or last returns, some record both, and some record first, last and intermediate returns.
In addition to the LIDAR sensor itself depicted at right, two other enabling technologies are critical. Airborne GPS enables the continuous high-accuracy 3-D positioning of the LIDAR sensor in the air, and Inertial Measuring Unit (IMU) technology enables the roll, pitch and heading of the sensor to be continuously measured with high accuracy. These technologies enable the 3-D position and orientation of the origin of each laser pulse to be known, and the timing of return signals enables the 3-D coordinates of first, last and intermediate return targets to be computed for ground surfaces.
The last LIDAR returns are not necessarily at the bare earth; sometimes laser pulses hit rooftops or dense vegetation and never penetrate to the ground. Post-processing (e.g., to remove points that fall on buildings and dense vegetation) to produce bare-earth DEMs/TINS comprises an estimated 60 percent to 80 percent of the total cost of a LIDAR project.
LIDAR has four major advantages compared with photogrammetry:
1. LIDAR can acquire accurate elevation data while flying from a higher altitude than can aerial photography. The accuracy of photogrammetric data is inversely proportional to the flying height, whereas LIDAR accuracy degrades less significantly with increased flying heights.
2. LIDAR needs only a single near-vertical laser pulse to penetrate between the trees (or through the trees) to measure the ground elevation, whereas photogrammetry requires two different lines of sight to both see the same points on the ground from two different perspectives. This means that LIDAR will have far fewer areas where the terrain is obscured by trees that block the lines of sight.
3. LIDAR can collect both first- and last-returns, with 5,000 to 33,000 pulses per second, so as to semi-automatically map the elevation of the tree canopy as well as ground elevations with high-density, high-accuracy data (but subject to post-processing and QC steps). Some LIDARs also collect intermediate returns between the first and last. Photogrammetry can generate high-density elevation points also, but only by expensive manual compilation of individual points, or by automated image correlation that surveys many points per second, but normally of treetops and rooftops instead of bare-earth terrain.
4. LIDAR data can be acquired both day and night, whereas aerial photography is acquired only during limited daylight hours when the sun angle is optimum.
However, LIDAR also has four major disadvantages compared with photogrammetry:
1. LIDAR pulses are often absorbed by water, and water returns are unreliable. It is difficult to determine the edge of lakes and rivers from LIDAR data alone. Normally, digital orthophotos are used to determine the limits of water boundaries, and LIDAR returns within those boundaries are discarded.
2. LIDAR data are ill-suited for determination of breaklines. If LIDAR pulses have a nominal point spacing of 5 meters, for example, it is hard to determine the location of breaklines at the tops and bottoms of stream banks that fall somewhere between the elevation points 5 meters apart, especially when data in the stream itself are also unreliable. Thus, breaklines estimated from digital orthophotos or photogrammetrically compiled breaklines are used to augment the LIDAR data as needed for hydraulic modeling.
3. LIDAR is a relatively new technology, and standard procedures have not yet been developed to yield data with predictable accuracy comparable to that from photogrammetry where flying height, focal length and photogrammetric procedures consistently yield predictable results.