FINAL REPORT Evaluation of Alternatives in Obtaining Structural Elevation Data Part I — Assessment of Elevation Strategies Part II — Providing Structural Elevation Data Prepared by Dewberry & Davis LLC for Federal Emergency Management Agency Contract EMW-2002-CO-0267 January 31, 2005 PURPOSE Insurance agents and write-your-own (WYO) companies have long affirmed that the requirement for Elevation Certificates (ECs) is a major impediment in selling flood insurance. In 2000, FEMA held a forum for parties interested in developing an eRating system for flood insurance policies. The purpose of the meeting was to exchange ideas on the best strategy to achieve FEMA’s goals for an eRating system and to discuss alternatives for overcoming the difficulties and high cost of implementing such a system. FEMA had not developed a requirement for an eRating system but simply sought input from industry, other government agencies, and academia on the best strategy to develop such a system. After careful consideration of the issues raised at the eRating forum, FEMA decided not to pursue developing an eRating system for flood insurance but rather, concluded that it must (1) examine whether it is appropriate, feasible, and legally possible for the government to provide elevation information on individual structures for use in rating structures, and (2) determine if it is technically feasible and cost effective to do so. The purpose of this study is to determine if it is appropriate, feasible, and legally possible for FEMA to obtain the elevation data on individual structures and to make this elevation information available to properly rate the structures for flood risks and flood insurance premiums so that ECs costing hundreds of dollars each would not be needed in most cases for insurance rating. The study examines the legal issues involved in collecting and making elevation information available and assesses five approaches for obtaining structure elevation information. The cost effectiveness of these various approaches is evaluated and a recommendation is provided for implementation of an elevation registry. This study will allow FEMA to determine if providing individual structure elevation data best serves the National Flood Insurance Program’s (NFIP) needs. Initially, the Dewberry team submitted a report on legal issues, prepared by FEMA Law Associates and the EOP Foundation, which summarized research and analysis on legal issues relevant to a determination by the Federal Emergency Management Agency (FEMA) of whether it can develop an elevation registry ("registry") of structural elevation data for National Flood Insurance Program (NFIP) purposes. The Dewberry team sought to identify and evaluate the significance of potential legal obstacles to developing this registry primarily in three areas: (1) the Privacy Act of 1974 and other privacy issues; (2) potential exposure to liability for inaccurate elevation information; and (3) potential ownership rights that third parties may have to elevation data. As indicated in APPENDIX A of this report, we identified no legal issues that would preclude FEMA from establishing and maintaining a registry and making it available to insurance companies and agents writing NFIP policies, or even to the general public. Creation of the proposed registry is an activity well within the authority granted by the National Flood Insurance Act. An elevation registry as described in the Statement of Work, and in subsequent meetings with FEMA, would not violate federal or state privacy law or policy or significantly expand the liability exposure of participants in the NFIP. Part I of this Final Report summarizes Dewberry's research and analysis of five (5) elevation strategies/alternatives considered for acquiring data for an elevation registry, explains how data were gathered and evaluated for utility in populating this registry, and assesses whether it is technically feasible and appropriate to utilize any or all of the five elevation strategies proposed by Dewberry to develop an elevation registry of structures. The five elevation strategies evaluated in Part I of this Final Report are as follows: ? Strategy A: Maximize use of existing Elevation Certificates ? Strategy B: Maximize use of airborne remote sensing (photogrammetry, LIDAR and IFSAR) ? Strategy C: Evaluate use of mobile photogrammetric vans ? Strategy D: Maximize cost-effectiveness of future Elevation Certificates ? Strategy E: Leverage alternative data sources for an elevation The purpose of Part I of this Final Report is to document the accuracy standards and the evaluation of each strategy, and to develop conclusions and recommendations regarding the utility of these strategies for populating an elevation registry. The purpose of Part II of this Final Report is to document the process for arriving at the recommended strategy or strategies for obtaining structure elevation data. TABLE OF CONTENTS PURPOSE ii PART I — ASSESSMENT OF ELEVATION STRATEGIES BACKGROUND 1 ACCURACY OF ELEVATION CERTIFICATES 4 LESSONS LEARNED IN NATIONWIDE SURVEYS 7 STRATEGY ASSESSMENTS 12 Strategy A — Maximize use of existing Elevation Certificates 14 Sources of Elevation Certificates 15 A.1 Insurance Services Office (ISO) 15 A.2 LOMA 2000 Database 16 A.3 Dewberry and URS 18 A.4 FEMA Regions and State NFIP Coordinators 21 A.5 Local Communities 22 A.6 U.S. Army Corps of Engineers 23 A.7 Policies in Force Database 24 A.8 Limitations of existing Elevation Certificates 25 A.9 Tools for Georeferencing existing Elevation Certificates 27 A.10 Conversion of paper Elevation Certificates to digital format 31 Strategy B — Maximize use of Airborne Remote Sensing 32 B.1 Photogrammetry 32 Conventional (Vertical) Photogrammetry 32 Conventional Photogrammetry with Measured Offsets 33 Oblique Photogrammetry — Pictometry 34 Photogrammetry Conclusions 38 B.2 Light Detection and Ranging (LIDAR) 39 LIDAR Data of Charlotte/Mecklenburg County, NC 42 LIDAR Data of Prince George's County, MD 44 LIDAR Data of Harris County, TX 44 LIDAR Data of Beaufort County, SC 45 LIDAR Conclusions 46 B.3 Interferometric Synthetic Aperture Radar (IFSAR) 47 IFSAR Data of Jefferson County, CO 49 IFSAR Conclusions 49 Strategy C — Evaluate use of Mobile Photogrammetric Vans 50 C.1 VISAT™ Photogrammetric Van 50 C.2 SideSwipe™ Vehicle Mounted Side Scan LIDAR 54 C.3 Mobile Remote Sensing Van Conclusions 55 Strategy D — Maximize Cost-Effectiveness of Future Elevation Certificates 56 D.1 Future Web-Based Elevation Certificates 56 D.2 Legal Considerations for Web-Based Elevation Certificates 56 Strategy E — Leverage Alternative Data Sources for an Elevation Registry 57 E.1 U.S. Census Bureau 57 E.2 U.S. Postal Service 58 E.3 U.S. Army Corps of Engineers 58 E.4 Community GIS Data 58 E.5 National Parcelmap Data Portal 59 E.6 CitySets 59 E.7 Home Owner Support 60 E.8 Flood Zone Determination Companies 61 E.9 Insurance Industry 63 E.10 NEMIS Database 63 Summary of Technology Capabilities 64 COST-EFFECTIVENESS (CE) ANALYSES 66 Methods for Populating an Elevation Registry 66 Base Scenario 73 CE Model Sensitivity to EC Total Value 81 CE Model Sensitivity to Elevation Registry Cost Parameters 82 CE Model Sensitivity to Accuracy/Digitization Cost of Existing ECs 83 CE Model Sensitivity to Accuracy of Future Ground-Surveyed ECs 85 CE Model Sensitivity to Accuracy/Cost of Photogrammetric ECs 86 CE Model Sensitivity to Accuracy/Cost of Pictometry ECs 88 CE Model Sensitivity to Accuracy/Cost of LIDAR ECs 90 CE Model Sensitivity to Accuracy/Cost of IFSAR ECs 93 CE Model Sensitivity to Accuracy/Cost of Photogrammetric Van ECs 94 STRATEGY SUMMARIES 96 Strategy A — Existing Elevation Certificates 99 Strategy B — Airborne Remote Sensing 100 Photogrammetry 100 Pictometry 101 LIDAR 102 IFSAR 103 Strategy C — Vehicular Remote Sensing 104 Strategy D — Future Web-Based Elevation Certificates 105 Strategy E — Leverage Alternative Data Sources 106 STRATEGY RECOMMENDATIONS 108 Strategy A — Maximizing the use of existing ECs 108 Strategy B — Maximizing the use of existing LIDAR or photogrammetric data 111 Strategy C — Utilizing data from mobile photogrammetric vans 113 Strategy D — Web entry of future ECs 114 Strategy E — Leveraging alternative data sources 114 PART II — PROVIDING STRUCTURAL ELEVATION DATA Purpose 117 Summary of Part I Requirements for eRating 117 Summary of Part I Legal Findings 117 Summary of Part I Technical Findings 119 Elevation Registry 120 Web Services 122 Populating the Registry 123 Registry Maintenance and Updates 124 Registry Use by Insurance Agents 127 Judgment Ratings 127 Registry Use by Others 130 FEMA Implementation Costs 132 Cost Recovery 133 Community Implementation Costs 134 Incentives 134 Advantages and Disadvantages 135 Summary 135 LIST OF TABLES Table 1 — Elevation Certificate Information 14 Table 2 — Summary of ISO Elevation Certificate Holdings 16 Table 3 — Relevant Data from LOMA 2000 Database 17 Table 4 — Dewberry's Elevation Certificates 19 Table 5 — URS' Elevation Certificates 20 Table 6 — Example File from Policies in Force Database 24 Table 7 — Same House with Four Different Elevation Certificates 26 Table 8 — Comparison of Geocoding Services 30 Table 9 — Pictometry Accuracy Comparisons 38 Table 10 — Mecklenburg County LIDAR Accuracy Comparison 43 Table 11 — Prince George's County LIDAR Accuracy Comparison 44 Table 12 — Beaufort County LIDAR Accuracy Comparison 45 Table 13 — Jefferson County IFSAR Accuracy Comparison 49 Table 14 — DataQuick Data Availability 62 Table 15 — Technology Suitability Matrix 64 Table 16 — Summary of Twenty Alternative Methods 66 Table 17 — Cost-Effectiveness Model Base Spreadsheet 76 Table 18 — CE Model Sensitivity to EC Total Value 81 Table 19 — CE Model Sensitivity to Elevation Registry Cost Parameters 82 Table 20 — CE Model Sensitivity to Accuracy/Digitization Cost of Existing ECs 83 Table 21 — CE Model Sensitivity to Accuracy of Future Ground-Surveyed ECs 85 Table 22 — CE Model Sensitivity to Accuracy/Cost of Photogrammetry ECs 86 Table 23 — CE Model Sensitivity to Accuracy/Cost of Pictometry ECs 89 Table 24 — CE Model Sensitivity Accuracy/Cost of LIDAR ECs 90 Table 25 — CE Model Sensitivity to Accuracy/Cost of IFSAR ECs 93 Table 26 — CE Model Sensitivity to Accuracy/Cost of Photogrammetric Van ECs 94 Table 27 — Summary of Elevation Alternatives 96 Table 28 — Methods Ranked by Overall Value of EC Records 97 Table 29 — Methods Ranked by Overall Cost of EC Records 98 Table 30 — Theoretical Flood Insurance Premium Increases 129 LIST OF FIGURES Figure 1 — Example of Poor Absolute Accuracy 11 Figure 2 — Photogrammetric Spot Heights 34 Figure 3 — Sample Pictometry Oblique Photo 36 Figure 4 — Sample 4-View Pictometry Images 37 Figure 5 — LIDAR Surfaces Before and After Post-Processing 40 Figure 6 — Examples of Building Footprints and Buffer, Centroids, and Parcel Polygons 41 Figure 7 — Comparison of ORI and Orthophoto Images 48 Figure 8 — VISAT Van 50 Figure 9 — VISAT Camera Array Configuration 51 Figure 10 — VISAT Navigation Route and Display 52 Figure 11 — VISAT Van Camera Comparison 53 APPENDICES APPENDIX A — Report of Legal Findings 137 APPENDIX B — Geospatial Accuracy Standards 234 APPENDIX C — Elevation Surveys 240 APPENDIX D — ISO Elevation Certificate Data 244 APPENDIX E — Elevation Certificate holdings of Dewberry and URS 275 APPENDIX F — Comparison of Commercial Geocoding Services 283 APPENDIX G — Photogrammetry Accuracy Analyses 287 APPENDIX H — Pictometry Explanation 289 APPENDIX I — Pictometry Accuracy Analyses 292 APPENDIX J — LIDAR Automated Data Extraction Report 297 APPENDIX K — LIDAR Accuracy Analyses 380 APPENDIX L — IFSAR Accuracy Analyses 404 APPENDIX M — VISAT Accuracy Analyses 406 APPENDIX N — SideSwipe Vehicle Mounted Side Scan LIDAR 408 APPENDIX O — Legal Comments on Web-Based Elevation Certificates 414 APPENDIX P — Elevation Registry System Description 418 APPENDIX Q — CE Spreadsheets for Base Scenario 421 APPENDIX R — Proposed Data Dictionary 424 APPENDIX S — Web-Based Registry 427 APPENDIX T — Community Rating System 437 PART I — ASSESSMENT OF ELEVATION STRATEGIES BACKGROUND The National Flood Insurance Act of 1968 created the National Flood Insurance Program (NFIP). The NFIP is a cooperative venture involving the federal government, state and local governments and the private insurance industry. The federal government sets insurance rates, provides the necessary risk studies to communities, and establishes floodplain management criteria guiding construction in the floodplain. Communities must adopt and enforce floodplain management standards for new and substantially improved structures. Flood insurance is only available in those communities that enact and enforce these measures. Private insurance companies, under an arrangement known as the Write Your Own (WYO) program, sell and service federal flood insurance policies in their own name and withhold part of the premium for their efforts. The government also sells flood insurance directly through its servicing contractor and retains the risk for all flood insurance policies. Within the Department of Homeland Security (DHS), the Mitigation Division, a component of FEMA, administers the NFIP. The regulations governing the NFIP appear in Title 44 of the Code of Federal Regulations and in manuals, procedures and other documents. The provision of insurance, the regulation of the floodplain, and the enforcement of the mandatory purchase requirements depend on three things: ? Flood hazard identification and risk assessment -- certain key information about the nature and extent of the flood risk in a given area ? Floodplain management -- the elevation of the structure ? Insurance rating -- structural characteristics such as the number of floors and occupancy type Flood Hazard Identification and Risk Assessment: FEMA provides flood-zone information in the form of a Flood Insurance Study (FIS) and Flood Insurance Rate Map (FIRM). The FIRM shows the flood risk in a given jurisdiction and serves as the guiding document for communities in the regulation of floodplain construction and for lenders in enforcing the mandatory purchase requirements. The primary flood risk characteristics shown on the FIRMs are the Special Flood Hazard Areas (SFHAs) or areas inundated by a one-percent annual probability flood and the elevation relative to the mean sea level to which the floodwaters will rise. (A discussion of accuracy standards is in APPENDIX B, and a discussion of mean sea level and vertical datums is contained in APPENDIX C). Whereas Base Flood Elevations (BFEs) are also shown on FIRMs, communities and surveyors usually get this information from a community's flood insurance study profile. The FIRM also serves insurance companies and agents as the source of needed risk information for writing and rating applications for flood insurance under the NFIP. Agents must have the flood zone and BFE to rate a flood insurance policy. For most Post-FIRM structures within the SFHA, the agent gets the flood zone and BFE from the Elevation Certificate (EC). In other cases, the agent must get the flood-zone and BFE from the community's FIRM and FIS. Locating a property on the paper copy of the FIRM has historically been a problem for agents, a problem that has not diminished substantially over the years. Flood zone determination companies and some WYO companies have digitized much of the information on the FIRMs and now provide this information to some agents. Zone information is far from universally available from the WYO companies, and agents may not want to pay the fee that flood zone determination companies charge for the service. The primary clients of the flood zone determination companies are federally regulated lenders who need the information to comply with the mandatory flood insurance purchase requirements of the National Flood Insurance Reform Act of 1994. Lenders are able to pass along the fee for the service to borrowers as part of a mortgage loan's closing costs. However, insurance agents may hesitate to do the same with their customers because charging for this service could jeopardize their competitive position. In 2002, FEMA made all effective maps available in raster scan version through the Map Service Center. Digital map files, flood insurance studies and flood profiles are now available on FEMA's website and on CD. This greatly improves accessibility for agents by eliminating the need to maintain paper copies of the maps and providing data which an agent or others can use to calculate the BFE. Floodplain Management: FEMA cannot provide flood insurance unless a community adopts and enforces a floodplain management ordinance that meets or exceeds the minimum requirements of the NFIP. Elevation information is needed to guide floodplain construction and to rate insurance applications. The community must ensure that the lowest floor elevation of a new structure, built in the SFHA after the date of the current effective FIRM, is at or above the BFE shown on the FIRM. To encourage community participation in the NFIP and the purchase of flood insurance, Congress subsidized the insurance premiums for buildings constructed before the issuance of a FIRM or before 1975, whichever is later. The NFIP does not require an EC to rate these buildings though, in the case of better-situated Pre-FIRM properties, actuarial rates may be lower. The NFIP, as adopted and enforced by each participating community, requires the community to obtain the elevation of the lowest floor, which includes the basement, of all new and substantially improved buildings, and to maintain a record of such information [44 CFR 60.3(b)(5) and 60.3(e)(2)]. The local floodplain administrator must determine which level is the "as built" lowest floor to verify whether the building complies with the community's floodplain management regulations. "As built" means that construction of the building is complete and the building is ready for occupancy. For new construction and substantially improved building in the floodplain, there are a series of surveys and inspections to verify that construction takes place according to plan and that it meets floodplain management requirements. FEMA's EC is an important tool that communities can use to document the level of flood protection for various building components and the "as built" lowest floor elevation. All communities must obtain and retain elevation information but they are not required to document that information on a FEMA EC. Communities participating in the NFIP's Community Rating System (CRS), which account for 66 percent of the policies, must obtain and retain this information on a FEMA EC (currently FEMA Form 81-31, July 2000). Communities often archive the elevation data they maintain and hence it is not readily accessible. The passage of time may have compromised the reliability of older elevation information. Insurance Rating: All applicants for flood insurance on Post-FIRM structures within the SFHA must provide the lowest floor elevation on a property in the form of an EC completed by a licensed engineer or surveyor. For these buildings, flood insurance rates take into account a number of different factors including the flood risk zone shown on the FIRM, the elevation of the lowest floor above or below the BFE, the type of building, the number of floors, and the existence of a basement or enclosure. The NFIP uses elevation information to determine rates for flood insurance coverage. The EC shows the structure's elevation relative to the mean sea level. Insurance agents writing a flood insurance policy use this information to determine a structure's lowest floor elevation and calculate the difference between the BFE and the lowest floor elevation to determine the proper rate for insurance coverage. Approximately 40 percent of the policies sold require an EC. The cost for the certificate is usually more than $300. The insurance agent obtains the relevant structural characteristics from the insured. For example, the property owner can supply information about the number of floors, occupancy type, date of construction, etc. to the agent. The NFIP's Flood Insurance Manual provides detailed guidance for the agent's use in rating flood insurance policies, which the agent uses to determine the type of building. Building types include those with no basement, an unfinished basement, a finished basement, mobile homes on foundations, and elevated buildings. The agent also classifies the building's occupancy type as single- family dwelling, two to four family dwelling, other residential building, or a non- residential building. The agent determines if the structure is used for commercial or residential purposes, records the value of the structure, its owners and, as appropriate, the elevation ACCURACY OF ELEVATION CERTIFICATES This section explains why it is important for structural elevation data to be accurate, especially for structures in or near the SFHA. Flood insurance premiums, for a post-FIRM structure (built after publication of a FIRM) and showing the structure's location to be within a SFHA, are largely based on the difference in elevations between the BFE and the top of the bottom floor of the structure. If the structure's top of bottom floor elevation is above the BFE, insurance premiums are much lower than when the top of bottom floor is below the BFE. For example, using NFIP flood insurance premiums as of May 1, 2004, for a post-FIRM building in the SFHA of a non-CRS community, annual premiums shown below are for $150,000 in building coverage and $75,000 in contents coverage for a one-story building with no basement and a $500 deductible: ? When the top of bottom floor is 2 ft above the BFE: $418 ? When the top of bottom floor is 1 ft above the BFE: $595 ? When the top of bottom floor equals the BFE: $892 ? When the top of bottom floor is 1 ft below the BFE: $3,201 ? When the top of bottom floor is 2 ft below the BFE: $4,040 From this example, it is obvious that elevation errors could have a major effect on actuarial rates charged for post-FIRM buildings, whereas the "subsidized" rate charged for pre-FIRM buildings (constructed prior to publication of a FIRM) would be $1,471, regardless of the top of bottom floor elevation. If the building is outside the SFHA in areas of low or moderate flood risk (shown as B, C, or X zones on a FIRM) and with no significant history of flooding, its premium for a Preferred Risk Policy (PRP) would be $264 (with $60,000 in contents coverage) regardless of the top of bottom floor elevation. Higher accuracy top of bottom floor elevation data also reduces risks in actuarial ratings. For this study FEMA assumes that ground-surveyed ECs are accurate to 0.5 ft at the 95% confidence level. For example, if an EC surveyor certifies a top of bottom floor elevation to be 1 ft higher than the BFE, FEMA can assume (with 95% confidence) that the actual top of bottom floor elevation is between 0.5 ft and 1.5 ft above the BFE, the structure has a relatively low risk of flooding and insurance premiums would be at a relatively low rate ($595 in the above example). However, if a less-accurate "approximate" aerial survey process were used to determine the top of bottom floor elevation, accurate to 2 ft at the 95% confidence level for example, then a top of bottom floor elevation determined to be 1 ft higher than the BFE might actually be between 1 ft below the BFE and 3 ft above the BFE at the 95% confidence level. Then, FEMA would have less confidence that the structure has a low risk of flooding and would need to charge a higher "judgment rating" premium (somewhere between $595 and $3,201) to account for increased risk of flooding, even though the top of bottom floor elevation is probably on the "safer side" of the BFE. Executive Order 12906 requires all Federal agencies collecting or producing geospatial data to comply with standards adopted through the Federal Geographic Data Committee (FGDC) which requires accuracy to be reported in ground distances at the 95% confidence level. As stated in FGDC-STD-007.1- 1998, Geospatial Positioning Accuracy Standards, Part 1: Reporting Methodology, the Federal Geographic Data Committee (FGDC) defines geospatial positioning accuracy in terms of local accuracy and network accuracy, described in APPENDIX B. To understand local accuracy, one needs to understand the concept of relative accuracy, as applied to a 95% confidence level. The most common ground surveys are referenced to local survey monuments or benchmarks, including temporary benchmarks such as elevation reference marks (ERMs), often selected for their proximity or convenient location rather than for their accuracy or stability. To understand network accuracy, one needs to understand the concept of absolute accuracy, as applied to a 95% confidence level. Such control surveys are referenced to a rigorous geodetic control network of survey monuments that are both accurate and stable. FGDC-STD-007.1-1998 states: "Geodetic control surveys are usually performed to establish a basic control network (framework) from which supplemental surveying and mapping work, covered in other parts of this document, are performed. Geodetic network surveys are distinguished by use of redundant, interconnected, permanently monumented control points that comprise the framework for the National Spatial Reference System (NSRS) or are often incorporated into the NSRS. These surveys must be performed to far more rigorous accuracy and quality assurance standards than control surveys for general engineering, construction, or topographic mapping. Geodetic network surveys included in the NSRS must be performed to meet automated data recording, submission, project review, and least squares adjustment requirements established by the National Geodetic Survey (NGS). The lead agency is the Department of Commerce, National Oceanic and Atmospheric Administration, National Ocean Service, NGS; the responsible FGDC unit is the Federal Geodetic Control Subcommittee (FGCS)." In addition to understanding distinctions between the various accuracy standards explained in APPENDIX B, it is also important to understand distinctions between ellipsoid heights (from GPS surveys that follow the rules of geometry) and orthometric heights (from differential leveling surveys that follow the rules of gravity). The term "elevation" is normally meant to infer orthometric heights. These distinctions, and other factors necessary in the generation of "accurate" elevation data, are discussed in APPENDIX C. As referenced in Table B.4 of APPENDIX B, accuracies in this study will refer to vertical errors at the 95% confidence level and equivalent contour interval (CI), as follows: ? ECs, whether surveyed by GPS or conventional means, are assumed to be accurate to 0.5 ft at the 95% confidence level. This means that 5% of the ECs will have errors larger than 0.5 ft when compared against a standard of higher accuracy such as geodetic surveys that satisfy NGS 2- cm or 5-cm standards. ? If alternative ECs are equivalent to 1' contours, as with the highest accuracy LIDAR surveys, this means that 95% of EC elevations checked should be accurate within 0.6 ft when compared against a standard of higher accuracy. ? If alternative ECs are equivalent to 2' contours (as with the most common LIDAR or high accuracy photogrammetric surveys), this means that 95% of EC elevations checked should be accurate within 1.2 ft when compared against a standard of higher accuracy. ? If alternative ECs are equivalent to 5' contours (as with mid-accuracy photogrammetric surveys), this means that 95% of EC elevations checked should be accurate within 3.0 ft when compared against a standard of higher accuracy. ? If alternative ECs are equivalent to 10' contours (as with IFSAR or USGS DEMs produced from 10' contours), this means that 95% of EC elevations checked should be accurate within 6.0 ft when compared against a standard of higher accuracy. A "standard of higher accuracy" is assumed to be at least three times more accurate than the product being evaluated, e.g., geodetic surveys accurate to 5- cm (?2") at the 95% confidence level are suitable for checking the accuracy of another product to determine if it is accurate to 6" at the 95% confidence level. As explained later in this report, FEMA considers elevations to have zero value for an elevation registry when elevation errors are 4 ft or worse at the 95% confidence level. LESSONS LEARNED IN NATIONWIDE SURVEYS This section describes early GPS survey projects that provided lessons learned for subsequent surveys performed by Dewberry, URS, and G&O. Occasionally, Dewberry has been asked to review ECs produced by other firms that were unaware of these lessons learned. Regardless, the major problem remains today for most EC surveys nationwide — that local surveys are still performed relative to the most convenient and accessible benchmarks, regardless of accuracy and stability, rather then using more rigorous (and expensive) procedures to guarantee some reasonable level of network accuracy. For decades, Dewberry surveyed ECs with conventional survey procedures; it has been using combinations of GPS and conventional survey procedures since 1993. Dewberry has also been hired to determine the most likely reasons for errors in elevation surveys performed by others, once a client determined that errors had occurred. This section summarizes lessons learned during the past decade. The following studies provided valuable input into shaping Dewberry’s current GPS elevation survey procedures. ? In 1993-1994, Dewberry performed GPS surveys of thousands of homes flooded in Georgia, Alabama, Florida, and Texas. For those surveys, FEMA only needed GPS to determine the latitude and longitude of the flooded homes. At that time, GPS elevation survey procedures had not yet been published. For elevations, FEMA asked only for "windshield survey estimates" of the depth of interior flooding as well as the area of each building's footprint, plus costs to repair (estimated by a Certified Flood Adjustor looking only at external conditions without leaving the car). ? In early 1995, Dewberry was tasked by FEMA to survey 1,300 structures in Louisville and Jefferson County, KY, to demonstrate the viability of generating low cost and expedient ECs and to provide homeowners with credible, personalized flood risk information on which to determine their need for flood insurance. This project, which became known as the "GPS shootout," compared the capabilities of a relatively unsophisticated GPS Backpack solution with those of a highly sophisticated GPS TruckMAP solution, both using "stand-off" survey procedures that did not require surveyors to walk on private property. ? In 1995, Dewberry, with support of ISO/CRS specialists, performed No- Cert GPS surveys of 1,468 structures. Dewberry developed Standards and Specifications for GPS "No Cert" Reference Level Surveys to be conducted in eight states (NJ, NC, SC, FL, LA, TX, CA, and WA), using procedures validated by the National Geodetic Survey (NGS). Demonstrations were conducted to validate the accuracy of the procedures to be used. ? In 1996, Charlotte and Mecklenburg County, NC, concluded that the mass production of accurate elevation surveys was the key to proactive floodplain management in their community. They hired Dewberry to survey approximately 2,190 floodprone buildings throughout the county and to develop a GIS database designed for proactive floodplain management. This database has since grown to over 3,000 structures, and all structural elevation data is freely available to the public on-line. ? ECs and databases were similarly prepared for Boone, NC; Roanoke, VA; and Prince George's County, MD. ? In 1998, as part of a Price-Waterhouse study of the economic effects of actuarially based premiums, Dewberry surveyed 7,628 pre-FIRM houses in 23 communities nationwide. By then, GPS elevation survey procedures were established and published by NGS for achieving 5-cm vertical accuracy. Local and temporary bench marks are not always accurate. When asking the Director of Public Works (DPW) in Albany, Georgia to recommend a survey monument to be used as the GPS base station, he asked, "Do you want a high one or a low one?" Dewberry replied that we wanted the most accurate one. The DPW replied that he didn't know which monument was actually the most accurate, but he knew that if one (high) monument was used as the reference station, elevations would be about one foot higher than if a different (low) monument was used. Because Dewberry didn't need accurate elevations for this particular project, this discrepancy did not need to be resolved. This was Dewberry's introduction to the fact that local surveyors often know which monuments are high or low, but may not know which one is more accurate; furthermore, it could be perfectly legal for a surveyor to choose a high monument, when surveying an EC, and this could artificially cause a house to appear to be less floodprone. In Jefferson County, KY, the County Surveyor provided a list of temporary benchmarks (TBMs) in the vicinity of the Southern Ditch which posed a threat of flooding to the majority of the 1,300 houses surveyed. These TBM descriptions were typical of FEMA's Elevation Reference Marks (ERMs), e.g., railroad spikes in power poles and trees (that grow), and chisel marks on bridge abutments, for example. While recognizing that none of these were suitable for accurate vertical surveys, Dewberry checked two telephone poles with railroad spikes for which elevations were provided. One had two, and the other had three railroad spikes at different elevations on the referenced power poles, visible from different directions. Depending on the surveyor's care and angle of approach, he/she could have selected a railroad spike several feet higher or lower than the one intended. Also, none of the benchmarks on this list referenced the vertical datum used. Unfortunately, this is typical of local benchmarks used for surveys of many ECs, where surveyors traditionally select the most convenient benchmarks, rather than those of higher accuracy at greater distances which are more expensive to survey. Surveyors may not be able to detect the presence of basements from the street. Of the 62 most difficult houses surveyed in Jefferson County, KY, the lowest floor elevations surveyed independently by two different methods all agreed within 1 inch, except for one house (3140 Sunny Lane) where two significantly different survey reference points were selected in large part because of the desire to minimize intrusion on private property. A neighbor had indicated to one survey team that the house next door had no basement under the northern half of the split-level home; this team classified the house as building diagram number 3 and surveyed the ground level entrance door to the southern half of the house. The other team detected the presence of basement windows in the northern section, classified the house as building diagram number 4, and surveyed the bottom of siding of the northern section with an offset of 8 feet to the basement floor. The elevation difference was over 5 feet between these two split levels. Failure to detect the presence of basements could cause lowest floor elevations to be in error by a full story (9 feet when including floor joists and flooring materials). Because of the high visibility failure at 3140 Sunny Lane in Jefferson County, KY, all surveys conducted by Dewberry since 1995 have included brief intrusions onto private property, if only to detect the presence or absence of basement windows and/or walk-out basement doors. (Uncertainty in the height of floor joists is the main reason why surveyors prefer to survey the top of the foundation and then subtract 8 feet -- the standard height of construction forms used to pour concrete foundations -- to determine the elevation of basement floors.) ERMs shown on FIRMs may not be reliable or may not be recoverable. To achieve high absolute accuracy from GPS surveys, it is important to first validate the accuracy of all monuments to be used as GPS base stations. Dewberry recommends that four of the best monuments surrounding a project area be checked relative to each other and validated, prior to actual surveys of structures. If the relative elevations of these four monuments are consistent within 1 inch, then any of them could be used as a GPS base station without causing significant errors in surveys derived therefrom. Dewberry attempted to survey the elevation of ERM 45, on Jefferson County's FIRM panel 170, which is nearest to the majority of the houses surveyed during this project, but this ERM could not be located after 2 hours of searching. ERMs from FIRM panels were found to be unreliable in all 8 states included in the No-Cert GPS Survey. They often were obsolete, having been destroyed because of construction or buried under concrete years ago. Most ERMs could not be recovered, and when they were recovered, they were often found to be inaccurate. For example, RM61 in Carteret County, NC, was documented as 10.41 ft on the FIRM, when its elevation was actually 5.78 ft per the NGS Data Sheet and confirmed by Dewberry's surveys as having an elevation of 5.78 ft. This was a significant error of nearly 5 feet. Other ERMs typically had errors of 6 to 18 inches. Note that FEMA no longer shows ERMs on its FIRMs but rather, on newer FIRMs, includes only NGS benchmarks of First or Second Order Vertical and a stability classification ranking of A, B, or C as defined by NGS. Local vertical monuments also may be shown on the FIRM with the appropriate designations. Local monuments shall be placed on the FIRM only if the community has requested that they be included, and if the monuments meet the NGS inclusion criteria. Additional information on qualifying criteria is as follows: ? They must be surveyed per NGS-58 guidelines for Secondary Base 5- centimeter monuments relative to existing NSRS monuments. ? They must have stability classifications of A, B, or C. ? Global Positioning System (GPS) files and station descriptions must have been previously submitted and accepted by the NGS for inclusion in the NSRS. Survey monuments must be verified before beginning a survey project. Before starting the surveys in Louisville and Jefferson County, KY, the elevations of four control points, recommended for use by the County Surveyor, were checked for accuracy. Three of the four were consistent within 1 inch, but control point BF26- 01 was found to be in error by 30 feet; its published elevation was 494.97 ft. but its correct elevation was 464.97 ft. During the No-Cert GPS Survey study, Dewberry's GPS teams invariably arrived in a community and had to "start from scratch" using NGS monuments to determine survey control. Local benchmarks were generally found to be inaccurate, also with errors of 6 to 18 inches, even in areas where subsidence was not a problem. In Charlotte-Mecklenburg County, NC, Dewberry's surveyors spent the first two weeks attempting to sort out the discrepancies found between the various survey monuments and benchmarks required as GPS base stations. Many discrepancies were over 12 inches, whereas Dewberry's standard was one inch. Dewberry required all surveys to be performed relative to accurate, reliable and stable benchmarks documented in the National Spatial Reference System and internally consistent within one inch; but it took two weeks to resolve control discrepancies throughout the county. To avoid multi-path errors, GPS elevations must be validated. To prevent potential GPS multipath errors, NGS-58 states that single elevation points should be surveyed twice, on successive days with distinctly different satellite geometry. Alternatively, pairs of inter-visible points can be surveyed the same day, using conventional survey procedures to survey a "backsight" and validate the elevation differences between the two points; if they agree within a few centimeters, then no multipath errors occurred. Subsidence can be a problem. In Louisiana and Texas, subsidence was a problem in several communities. In some cases, several days were spent by 2- person survey crews trying to resolve 12 inch discrepancies in elevations of control points that should have been accurate to a fraction of an inch, but which were apparently sinking at different rates as a result of subsidence. Accurately located structures must be compared to accurately located floodplain boundaries to make in/out determinations. The left side of Figure 1 shows a segment of FIRM panel 25 of 50 in Omaha. The SFHA to the east includes two tree-lined streets that closely parallel the creek. The pre-FIRM houses on these two streets are clearly in the SFHA. However, the right side of the figure shows that an automated determination would plot the houses outside the SFHA -- not because they were actually outside the SFHA, but because the entire paper FIRM, from which the Q3 Flood Data was produced, lacks absolute horizontal accuracy. The GPS points with absolute accuracy (network accuracy) could not be accurately registered to the less accurate base map that has relative accuracy (local accuracy) only. Similar problems occur with elevation surveys where ECs may have good local accuracy relative to the nearest benchmarks, but lack good network accuracy relative to the geodetic datum which should form the vertical basis for all Flood Insurance Studies. Figure 1 — Example of Poor Absolute Accuracy STRATEGY ASSESSMENTS FEMA’s intent in creating the elevation registry is to expedite and simplify the rating and issuance of flood insurance policies by insurance agents, WYO companies, and the FEMA contractors issuing FEMA flood insurance policies directly and, when possible, avoid the need for new ECs to be surveyed. If FEMA decides to establish an elevation registry, it would probably be a subset of the NextGen data warehouse, with firewalls to prevent Privacy Act violations. The data will be available to WYO companies and agents in a format capable of linking to their existing computer systems. Further, for purposes of rating and writing policies, FEMA intends that agents and companies be able to rely on elevation data in the registry, and policies properly written and rated consistent with elevation data in the registry will be deemed correct until the registry information is changed. The registry, at minimum, will provide to insurance agents and companies improved and simplified access to a key element of evaluating flood risk: elevation of the structure as compared to the BFE as determined in that area. As noted in Dewberry's Report on Legal Issues at APPENDIX A, Registry data will likely also be available and accessible to homeowners, potential homeowners, communities, lenders, and any private companies requesting access to this data. While the registry is not designed for this purpose, homeowners or prospective homeowners might seek to use the data to evaluate flood risk of their homes, or of properties prior to purchase. Communities might use this data in studies of flood prone areas or as part of a building permit process. For communities that maintain their own ECs online, the registry might simply provide a link to the community web site. FEMA should consider granting CRS credits for providing this information to the public. From the Dewberry team's legal analysis at APPENDIX A, we understand that elevation information required for use in determining premiums for an actuarially sound flood insurance program need not be as accurate as information required for evaluating the true flood risk of individual structures. An actuarially sound program can average out modest positive and negative errors in elevations of individual buildings, whereas those same errors could hide true flood risk for the owner of a particular structure. Whereas FEMA can accept some uncertainty in approximate, uncertified elevations of existing structures for insurance rating purposes and use judgment rating procedures to increase flood insurance premiums accordingly, communities and home owners need elevation data with absolute accuracy, providing certified assurance that new construction does not result in floodprone structures intended to be built at higher elevations. Communities would be advised not to rely on an elevation registry of approximate top of bottom floor elevations for evaluating a permit or for determining compliance of an existing structure that is being substantially improved or was substantially damaged. A new EC would have to be obtained to determine the "as built" information on the structure. In the case of new construction, an EC would not even exist. Elevation information used for floodplain management purposes must be as accurate as possible for any proposed construction in the floodplain. This elevation information includes the BFE, any topographic information, and the proposed building elevations of all new and substantially improved structures that are provided to the community as part of the application for a development permit. It also includes “as built” elevation information the community must obtain once the structure is completed before it can issue a certificate of occupancy or compliance. Information in a registry cannot properly be used as a substitute for “as built” information because it is generally not available at the time the building is completed and may not be of the required level of accuracy. To ensure that potential users of the registry are aware of its limitations, the registry should include a prominent notice stating that it may be used in lieu of ECs in rating or writing flood insurance policies but that the approximate elevation information may not be sufficiently accurate for other purposes, particularly in determining whether to purchase a structure in the flood plain or to permit new construction or renovation in the floodplain. Relying on elevation data that is costly to obtain led FEMA to examine whether it is legally possible, appropriate and feasible to obtain and make available the elevation information necessary to rate a flood insurance policy. FEMA's goal is to make elevation information more accessible to foster the development of an eRating system that supports the actuarial rating of a flood insurance policy. For this study, FEMA originally identified two strategies to obtain elevation information to eRate flood insurance policies, which FEMA wanted thoroughly examined. (1) The first strategy called for a means to efficiently gather into a single, accessible database all available ECs for structures in the floodplain and to continually update this database as additional or better structure elevation information becomes available; this strategy is designed to capture the elevation data needed to rate a flood insurance policy in a single database. (2) The second strategy called for exploring new mapping technologies and approaches, combined with other property data, to gather elevation data. For example, Light Detection and Ranging (LIDAR) and Interferometric Synthetic Aperture Radar (IFSAR) can provide information on the lowest adjacent grade near a structure from which it is possible to determine the ground elevation and estimate the structure's lowest floor elevation, using foundation types or some other parameter, measured from that ground elevation. From FEMA's two strategies, Dewberry proposed five strategies to be evaluated for populating an elevation registry once it was determined that there were no major legal impediments for doing so. These strategies are evaluated in the following sections. Strategy A — Maximize use of existing Elevation Certificates Strategy A is based on gathering all available ECs for flood prone structures and capturing the elevation information, needed to rate a flood insurance policy, into an accessible elevation registry that would be maintained and updated as new information becomes available. Owner names would be deleted from the elevation registry because of Privacy Act considerations. Since ECs are required to rate a policy when the structure meets certain conditions (e.g., date of construction, flood hazard zone, etc.) and to obtain a LOMA or LOMR-F, EC information is vital for the elevation registry. Strategy A will be applicable to structures for which ECs have been developed and where they are most readily available, preferably in electronic format. Strategy A alone will not result in a structural elevation database for all structures in and near the SFHA. In most communities, ECs have only been produced where required for selected structures. Several previous studies have identified the potential need to evaluate existing ECs for completeness and accuracy. The other strategies assessed in this study will shed light on this issue through the comparison of ECs with other data sources. ECs include structure-specific information about the elevation of various features of that structure. Depending on the type of structure, as depicted in standard building diagrams, the information in Table 1 is currently required by FEMA and by insurance agents to rate policies for flood risks. Table 1 — Elevation Certificate Information General Information Elevation Information Address Base Flood Elevation (BFE) Flood zone Lowest adjacent grade (LAG) Building use* Highest adjacent grade (HAG)* Building diagram number Top of bottom floor (TBF) Latitude/longitude (optional)* Top of next higher floor (TNHF)* Source of latitude/longitude* Bottom of lowest horizontal structural member (LHSM) Horizontal datum* Top of slab of attached garage* Source of elevation information Lowest elevation of machinery and/or equipment servicing the building* Vertical datum * Note: Older ECs may not contain these items. ECs also typically include other information such as FIRM panel number and date. ECs that are already in a computer database format will be the least costly to convert to an elevation registry. Ideally, all pertinent elevation information has been captured from the ECs and the data would be able to be imported directly into the registry. Sources of Elevation Certificates In order to populate an elevation registry, the logical starting point is to determine where existing elevation data exists, to determine the data format(s), to assess the data suitability to the needs of the registry, and to assess the ease with which the data could be obtained and imported. Seven potential data sources were evaluated: (1) databases of the Insurance Services Office, Inc. (ISO), (2) FEMA's LOMA 2000 database, (3) ECs available to Dewberry and URS Corporation, (4) ECs available to FEMA Regions and state NFIP coordinators, (5) ECs available at selected local communities where large numbers of digital or hardcopy ECs are available, (6) ECs available from the U.S. Army Corps of Engineers, and (7) FEMA's Policies in Force database. In order to document the existence of ECs that might be available from various sources, Dewberry developed an Existing Data Review Form to use as an aid when contacting agencies that have ECs. The form was designed to be used to document the findings of the available ECs and/or elevation databases in the various FEMA regions and states. With few exceptions, Dewberry learned that such elevation information is only available at the community level, except for data already being collected by ISO. For the most part, Dewberry did not contact individual communities as part of this study. The exception was that we contacted communities whose ECs we needed to augment with additional data to determine if ancillary data could be made available for geocoding addresses. A.1 Insurance Services Office, Inc. (ISO) Under this contract ISO was tasked to inventory the ECs and/or elevation databases that they have access to. This includes the ECs submitted annually by Community Rating System (CRS) communities. All CRS communities are required to maintain ECs for all buildings built in the SFHA after the date of their application to the CRS. The community must make copies of the ECs available to all inquirers, and FEMA publishes a listing of the phone number of the point of contact for the ECs for each CRS community. Additional CRS credits are given for maintenance of ECs for older buildings and for maintenance of ECs in an electronic format. FEMA estimates that 63% of the flood insurance policies are in CRS communities. Approximately 25% of the CRS communities receive credits for maintaining their ECs in a computer format. They submit these data annually to ISO which collects them on behalf of FEMA. Additionally, ISO participated in a 1995 study on the retrieval and conversion of ECs to a database format. At the time, significant difficulties were encountered in creating a database from information documented on more than 7 different FEMA EC forms and provided in hardcopy format of varying quality, completeness, and legibility. However, there remain a significant number of ECs in this database. ISO reported the EC holdings in Table 2. Table 2 — Summary of ISO Elevation Certificate Holdings Source Format Number of CRS communities Number of Elevation Certificates 1995 conversion project Access database 315 33,865 CRS program Diskettes 90 17,751 (Non-CRS) 50 1,367 Total 404 52,983 Dewberry subsequently asked ISO to provide existing ECs and/or elevation databases for Pinellas County, FL; Beaufort County, SC; Jefferson County, CO; and Harris County, TX to support Strategies B and C. Several issues were found with the data that are noted later in this chapter. A complete listing of ISO’s data sets can be found in APPENDIX D. A.2 LOMA 2000 Database LOMA 2000 is a software application used by all of FEMA’s Mapping Coordination Contractors (MCCs). LOMA 2000 automates the writing of Letters of Map Amendment (LOMAs) and Letters of Map Revision based on Fill (LOMR- Fs) and their attachments. Approximately 20,000 LOMAs and LOMR-Fs are processed annually by the MCCs. For most, if not all, LOMAs and LOMR-Fs, an EC and additional information, e.g., lowest elevation on the parcel is required. Owners typically request that their house be administratively removed from the SFHA because the lowest grade adjacent to the structure is higher than the BFE on the FIRM. LOMA 2000 has been in use since 1999 and contains approximately 163,000 records with EC information. Historic Letters of Map Change (LOMCs) have been entered into LOMA 2000; however, pertinent information is missing in the database for these older records. The LOMA 2000 data dictionary includes the elements, listed in Table 3, that could be relevant to an elevation registry, as well as many other elements that are non-relevant. It should be noted that the lowest floor elevation (LFE) often is not the same as the top of bottom floor elevation because the LFE may include the lowest insurable elevation, to include crawl space, floor of attached garage, or lowest elevation of machinery. Table 3 — Relevant Data from LOMA 2000 Database General Information Horizontal Location Data Elevation Data Community code Latitude BFE State code Longitude BFE source Street address Lat/Long source LAG (elevation) City Lat/Long datum LAG source Zip code Old zone Lowest floor (elevation) County New zone Lowest floor source Lot Elevation datum Block Section Panel Panel date LOMA LOMR_F CLOMA CLOMR_F Dewberry alone currently has approximately 130,000 addresses in the LOMA 2000 database. Depending on the age of the ECs, they may or may not include latitude and longitude. Approximately 80,000 records in Dewberry's LOMA 2000 database include latitude and longitude, Baker has approximately 12,000 files, with 4,200 geocoded. PBS&J has approximately 21,000 records, with 7,400 geocoded. All of the MCCs use commercial geocoding packages to estimate the latitude and longitude of LOMA 2000 addresses if they have not been provided on the EC. Only 124 entries in LOMA 2000 list latitude and longitude as having been derived by GPS survey. The elevation data in the LOMA 2000 database was obtained from many different documents including various editions of the FEMA Elevation Certificate. Therefore, the data will only be as good as the knowledge of the persons completing and interpreting the form before entering it in the database. For insurance rating, FEMA considers data in LOMA 2000 to be less reliable than data on an EC submitted with a LOMA application. Whereas the ECs submitted with a LOMA application may be highly accurate, the elevation data in the LOMA 2000 database is believed to be less accurate. Moreover, the latitude and longitude errors that may exist within LOMA 2000 from the use of commercial geocoding packages may be several hundred feet in any direction, as documented in A.9 below. Of note, the FEMA DFIRM Database design team considered adding point locations of LOMCs as a layer in the DFIRM Database, using LOMA 2000 as the source. However, because of the known geocoding variances, it was decided that a relational table listing LOMC cases by panel was a more prudent option than including approximate structure locations. The LAG elevations in the LOMA 2000 database do have value for the elevation registry. A.3 Dewberry and URS Dewberry and URS each searched their archives to determine the availability of existing EC data that they had produced. The complete results are at APPENDIX E. A summary is provided below. Dewberry Dewberry has produced thousands of GPS ECs as part of the 1995 No Cert study; the 1999 Study of the Economic Effects of Charging Actuarially Based Premium Rates for Pre-FIRM Structures; for pro-active communities such as Charlotte-Mecklenburg, NC; and as part of post-disaster surveys of damaged structures. As indicated in Table 4, Dewberry was able to assemble a combined Access database of 16,381 GPS ECs. The Dewberry ECs included a mix of residential, commercial, and public structures; were collected by GPS survey; include latitude/longitude, LAG, and BFE; include structure details such as building diagram number or building description; and most contain elevations for top of bottom floor or top of reference floor. Some of the earliest surveys, including the No Cert surveys in 1995 and post- flood surveys in 1993-94, were not retrievable in digital format and/or lacked suitable information for an elevation registry. The hardcopy deliverables were provided to FEMA. The 1993-94 post-flood surveys in Georgia, Alabama, Florida, and Texas did not include any actual elevations, but depths of interior flooding to the nearest whole foot, based on "windshield surveys" from the car. Dewberry also obtained structure information from the City of Austin, TX, to be used for a demonstration by FIA at the 2000 National Flood Conference of the feasibility of automating flood policy writing. The data provided by the City of Austin included addresses, structure type, latitude/longitude and first floor elevation for 863 structures in and near the floodplain of Waller Creek. The city has collected this type of information for an EC database. Additionally, the city maintains an address centroid database of 272,127 addresses for the entire city (not including first floor elevation). Table 4 — Dewberry's Elevation Certificates Source Geographic Area Date Structure Category Format Number of Elevation Certificates HMTAP West Virginia 1996 Damaged structures Database 1,129 Actuarial Study Throughout U.S. 1997 Pre-FIRM structures in SFHA Database 8,083 HMTAP North Topsail, NC 1997 Damaged structures Database 2,046 Charlotte- Mecklenburg County, NC Mecklenburg County, NC 1997 All structures in SFHA Database 2,197 (Add’l ECs done by county since) HMTAP Horry County, SC 2000 Damaged structures Database 207 HMTAP Coastal counties of Maryland 2000 Damaged structures Database 84 Boone, NC Boone, NC 2001 Selected structures in SFHA Database 378 Project Impact Roanoke Valley Alleghany Regional Commission Roanoke, Roanoke County, Vinton, & Salem, VA 2001 Selected structures in SFHA Database 1,495 (Add’l ECs done by PDC since) Prince George’s County, MD Prince George’s County, MD 2002 & 2003 Selected structures in SFHA Database 762 Subtotal Database 16, 381 No-Cert GPS Survey Study Florida, New Jersey, North Carolina, & South Carolina 1995 Post-FIRM structures in SFHA Hardcopy 1,368 Post-Flood Surveys Georgia, Alabama, Florida, and Texas 1993- 1994 Damaged structures Hardcopy (not true ECs) 7,963 No absolute elevations Total 25,712 URS Under various HMTAP task orders managed by URS, contractors such as Dewberry, Greenhorne & O'Mara (G&O), GRI (now Baker), and SKW have surveyed thousands of additional ECs, as summarized in Table 5. The data in these ECs were collected using conventional and GPS surveys. All are available in hardcopy format and some may be available in digital format. These surveys normally include the pertinent EC items. Table 5 — URS' Elevation Certificates Source Geographic Area Date Structure Category Format Number of Elevation Certificates HMTAP TO012 Sonoma County, CA Russian River 1995 Damaged structures Hardcopy 450 HMTAP TO032 Lexington, VA & surrounding counties 1995 Damaged structures Hardcopy 750 HMTAP TO048 Wyoming, Bedford, & Lycoming Counties, PA 1996 Damaged structures Hardcopy 1,050 HMTAP TO079 Hoisington, KS 2001 Damaged structures Hardcopy 20 HMTAP TO081 Wyoming County, WV 1996 Damaged structures Hardcopy (see Dewberry’s listing for database) 175 HMTAP TO082 West Virginia 1996 Damaged structures Hardcopy (see Dewberry’s listing for database) 1,060 HMTAP TO113 North Topsail, NC 1996 Damaged structures Hardcopy (see Dewberry’s listing for database) 1,000 HMTAP TO122 Shenandoah County, VA 1996 Damaged structures Hardcopy 46 HMTAP TO126 Danville & South Boston, VA 1996 Damaged structures Hardcopy 14 HMTAP TO129 Barbour & Harrison Counties, WV 1996 Damaged structures Hardcopy 172 HMTAP TO130 West Virginia 1996 Damaged structures Hardcopy 240 HMTAP TO131 Hampshire County, WV 1996 Damaged structures Hardcopy 80 HMTAP TO139 Page & Warren Counties, WV 1997 Damaged structures Hardcopy 17 HMTAP TO142 West Virginia 1997 Damaged structures Hardcopy 330 HMTAP TO144 Shenandoah & Rockingham Counties, VA 1997 Damaged structures Hardcopy 10 HMTAP TO373 Horry County, SC 2000 Damaged structures Hardcopy (see Dewberry’s listing for database) 220 TOTAL 5,634 Additional information about the URS data sets can be found in APPENDIX E. A.4 FEMA Regions and State NFIP Coordinators For this study, Regional engineers and State NFIP Coordinators nationwide were telephoned by Dewberry, URS and G&O to determine the availability of ECs at the regions and states. Without exception, the regions and state NFIP coordinators indicated that individual community NFIP coordinators would need to be queried to determine what was available at community level, because the regions and state coordinators do not maintain ECs. It was beyond the scope of this study to contact all individual community NFIP coordinators, but some state coordinators provided information about communities known to have large numbers of ECs. A few of these communities were contacted as noted below. A.5 Local Communities Several local communities believed to have large holdings of ECs were contacted regarding available data. Some of them have ECs online for public outreach; to avoid duplication of effort, FEMA's elevation registry should (as a minimum) provide a link to such sites. Several communities with larger numbers of electronic ECs are noted below. Additionally, a few local communities provided other GIS data or support that enabled Dewberry to make use of the ECs for use in evaluating Strategies B through D. These are also noted below. Monterey County, CA. Monterey County maintains a database containing the information from Sections A-F of the new (2000) EC form. The database is sent to FEMA during every verification cycle for the Community Rating System (contained on CD). The builder collects most of the data; however, some data are verified by the County. Currently the database contains 383 records. This is consistent with ISO’s EC database records for this community (374). The community is also receiving full CRS credit for maintaining ECs in a computer format. Sacramento County, CA. Sacramento County maintains a database that contains all the EC data for structures built in the floodplain within the last few years. The database is directly linked to an EC template that can be printed. Approximately 90% of the data was collected by county surveyors specifically for this purpose. Additional data is kept for local flooding regulatory elevations such as high-water marks, etc. Currently the database contains approximately 200 records out of approximately 3000 structures. ISO reports that this community is not receiving full credit for maintaining ECs in a computer format and reports 0 database records for this community. Santa Barbara County, CA. Santa Barbara County keeps hardcopies of ECs filed by parcel. In addition, the County maintains an internally developed Access database that duplicates all of the EC information. All structures (pre- and post- FIRM) are kept in the database, which numbers approximately 1000 entries. It is required that all elevation data be obtained by a licensed land surveyor, and are tied into USGS benchmarks. ISO reports that this community is not receiving any CRS credit for maintaining ECs in a computer format and reports 0 database records for this community. Maricopa County, AZ. Maricopa County currently maintains a database of all EC data. This database is linked to their geographic information system (GIS) using ArcView. A database query can be used to complete and printout ECs for any parcel. They are in the process of linking this information to their internet site so that it will be available to the public. All of their ECs are in the database. The number of structures currently numbers 662. ISO reports 932 database records for this community. The community is also receiving full CRS credit for maintaining ECs in a computer format. Simi Valley, CA. The City of Simi Valley maintains EC data for all residential structures in the form of a database that contains all of the same information. Elevation data for larger structures (commercial, industrial, etc.) are kept on file in hardcopy form. ISO reports that this community is receiving full CRS credit for maintaining ECs in a computer format and reports 13 database records for this community. Charlotte and Mecklenburg County, NC. The Charlotte-Mecklenburg (NC) Storm Water Services provided over 3,000 ECs used by Dewberry for evaluation of low- resolution LIDAR data produced of Mecklenburg County as part of the North Carolina Floodplain Mapping Program. A GIS database with building footprints, plus the raw LIDAR dataset, was also provided by the community. Note that Dewberry also lists 2,197 EC records for this community; the remaining ECs were surveyed by other firms. This is one of those communities that maintain ECs online for public information and outreach. Prince George's County, MD. Prince George's County, MD provided ECs used by Dewberry for evaluation of mid-resolution LIDAR data and oblique Pictometry images. Note that Dewberry also lists 136 EC records for this community. Beaufort County, SC. Beaufort County, SC provided GIS data used by Dewberry for evaluation of high-resolution LIDAR data also provided by the county. Jefferson County, CO. The Jefferson County Planning and Zoning Department provided Dewberry with 25 ECs as well as accurate geographic coordinates for each structure. Pinellas County, FL. The GIS Coordinator for Pinellas County, FL provided a GIS file (Pinellas_co_parcels_roads.dxf) used by Dewberry to georeference ECs. A.6 U.S. Army Corps of Engineers (USACE) The Philadelphia District of USACE used contractors to survey thousands of floodprone houses for the Susquehanna River Flood Warning and Response System (FWRS) in Pennsylvania. They used a quasi-photogrammetric method whereby photogrammetric spot heights were established of the terrain surrounding the corners of each house visible in stereo, and then surveyors were hired to measure the vertical offset up or down from these spot heights in order to determine the top of bottom floor elevation, top of next higher floor elevation, and lowest adjacent grade elevation of each house. The specifications for photogrammetric spot heights called for vertical accuracy of 0.5 ft (6 inches) at the 90% confidence level, with spot height accuracies equivalent to 2' contours. They have approximately 1,200 structures with street addresses and elevation data, plus an additional 1,400 structures with elevation data and only partial or incomplete addresses. A. 7 Policies in Force Database FEMA’s Policies in Force database contains some 3 million records, 80,000 of which have elevation data. The Policy database includes the following information as shown in the example from CSC's BureauNet (sometimes called FIANet) in Table 6 below. Note that latitude and longitude are not available for many records. Table 6 — Example File from Policies in Force Database Company No: 23779 Pol Nbr: 5400518089 Pol Status: Expired more than 94 days Pol Eff Dt: 08/12/2001 Pol Exp Dt: 08/12/2002 Org Nb Dt : 08/12/2000 End Eff Dt: 08/12/2001 Org Con Dt: 01/01/1996 As of Date: 01/31/2003 Community : 370246 CRS Class : 0 Probation : 0 First Name: RICHARD Last Name : EDDINS Address 1 : Address 2 : 204 DULCIMER LN City : ZEBULON State: NC Zip Code : 27597 2876 Addr Key : NC2145LN2043425 WYO Rate Data Program : Regular Rate Meth : Manual Rollover : New Policy Exp Const : 0 Condo Ind : Non-Condo Condo Unit: 1 Prem Pay I: Bldg Basic: Occupancy : Single Family Building : One Floor Bsmt/Encl : None Bldg Addtl: Post Firm : Y Flood Zone: AE Loc Cont : Cont Basic: Crse Const: N State Own : N Dis Assist: 0 Cont Addtl: Pol Term : 1 Small Bus : N Ins To Val: ICC Prem : 0 Comm Prob : 0 Premium : 209 Pol Fee : 30 NFIP Expc : 50 Deduct Pct: Bldg Covg : 900 Cont Covg : 50 Rep ICCCov: 200 NFIP ICC $: 6 Bldg Deduc: 500 Cont Deduc: 500 Base Flood: 232.5 Low Floor : 244.4 Elev Diff : 12 Diagram # : 8 Low Adj Gr: 235.3 Fld Proof : N Obstruct : 10 Elev Cert : 3 Post V Crt: N Longitude : .000000 Latitude : .000000 GEO Result: N GEO Census: A.8 Limitations of Existing Elevation Certificates As noted above, ISO provided Dewberry with a spreadsheet of EC data for CRS communities within four counties: Pinellas County, FL; Beaufort County, SC; Jefferson County, CO; and Harris County, TX. In working with these data and through evaluation of other available ECs, several limitations have been identified. These are issues that will make the creation of a consistent, up to date, and accurate elevation registry challenging. Data are not centralized. Most ECs are maintained at the local level. This means that obtaining the information that would be needed for an elevation registry will require significant effort to identify and obtain. Additionally, most communities that maintain their ECs in a hard copy format reported that substantial effort would be required to collect and submit ECs. Most ECs are not digital. Most communities do not maintain their ECs in an electronic format. Those databases that do exist tend to contain only newer structures and/or newer LOMAs and LOMR-Fs. The older records that are stored on paper in scattered locations may be much more difficult and potentially cost-prohibitive to retrieve. A 1995 study by ISO on the retrieval and conversion of ECs to a database format resulted in over 30,000 records that were provided to FEMA and to the CRS communities that originally supplied the information. At the time, significant difficulties were encountered in creating a database from information documented on seven different FEMA EC forms and provided in hardcopy format of varying quality, completeness, and legibility. These issues would still exist with ECs not currently in a digital format. Many EC database records are missing information or appear to contain questionable information. Related to the issue of older EC forms noted above is the fact that certain pertinent pieces of information for an elevation registry may not have been included in older forms. Highest Adjacent Grade (HAG) is an example of an item currently required but not included on older EC forms. Many of the ECs that were retrieved for use in evaluating Strategies B and C lacked relevant elevation data. Pinellas County, Florida (1,524 records) Of the 1,524 Pinellas County records, 1,361 (89.3%) had lowest floor elevations in A-zones, 27 (1.8%) had lowest floor elevations in V-zones, 136 (8.9%) had no lowest floor elevations, and 581 (38.1%) had no LAG information. Of the 1,524 records, 1,306 had elevations less than 25 feet (most were between 5 and 15 feet); but 55 had elevations over 100 feet, with two over 500 feet and one over 800 feet. Thus, 55 (4.2%) of the 1,306 elevations were probably in error, especially since there were no elevations between 25 feet and 100 feet. Of these 55 erroneous elevations, 17 listed NGVD as the vertical datum, whereas the remaining 38 had the datum field blank in the database. Thus, of the 1,524 records in Pinellas County, 191 records (12.5%) either had no lowest floor elevation data or the elevations were grossly in error, and 581 (38.1%) had no LAG elevations. There were about 400 records that were duplicates for fewer than 200 homes, normally resurveyed on different dates, with different elevations. In one interesting example (see Table 7), the lowest floor elevations for the same house vary between 1.4 and 11 feet, LAG elevations vary between 6.5 and 10.4 feet, and BFEs vary between 10 and 12 feet. The lowest floor elevation change from 11 feet to 1.4 feet may also be caused by illegal construction below an elevated structure. Table 7 — Same House with Four Different Elevation Certificates House Number Street Prefix Street Name Street Suffix EC Date Zone BFE LAG Lowest Floor A-zone Lowest Floor V- zone 4200 S 54th Ave 7/11/91 A12 10 blank 10 blank 4200 S 54th Ave 2/19/92 A12 11 10.4 11 blank 4200 S 54th Ave 11/25/92 A12 10 6.5 blank blank 4200 S 54th Ave 3/05/93 V15 12 blank blank 1.4 Based on prior experience, it is common to survey LAG and lowest floor elevations that differ by a foot or more when surveyors base their surveys on different elevation reference marks (ERMs) that are unstable, inaccurate, and not accurately surveyed with GPS relative to the National Spatial Reference System (NSRS) maintained by the National Geodetic Survey (NGS). Furthermore, as demonstrated in the above example, elevations are sometimes in error by 8-9 feet because of confusion by surveyors as to which floor is the lowest, a reason why changes were made to FEMA's new EC Form 81-31 in 2000. Beaufort County, South Carolina (448 records) Of the 448 records of Beaufort County, 122 (27.2%) had no lowest floor elevation, and 126 (28.1%) had no LAG elevation. Only 46 of 448 (10.3%) of the records had street addresses and lowest floor elevations for the same records. Others had street names that duplicated other records, but no house numbers to distinguish one building from the other; but they did have lot numbers. Jefferson County, Colorado (10 records) Of the ten records of Jefferson County, only three (30%) had lowest floor elevations, and none had LAGs. Harris County, Texas (22 records). Of the 22 Harris County records, 13 had LAG elevations and 21 had lowest floor elevations, but one of these elevations was (erroneously) 5,429 ft, whereas all other elevations in the county's dataset were less than 25 feet. ISO, the source of the EC database information cited above, noted that "gaps" in the data exist and "vary greatly by community because in most cases the community is not only the source of the data but also the checker of data quality." ISO also provided additional rationale: ? “Communities transfer the EC information from hard copy to a data set, not ISO. If information is missing it is because the community failed to enter it. ? "ISO only randomly quality checks hard copy ECs. We do not compare the hardcopy and the data sets, as this is the communities’ responsibility. ? "No EC information is typically found for properties in the un-numbered A zones or AO zones, C or X zones because the FIRMs do not show elevation data. Gaps in EC data can subsequently result. ? "EC data quality also varies greatly by state." No latitude/longitude. As expected, none of the street addresses found in the ISO database were georeferenced, i.e., none had latitude/longitude, UTM coordinates or State Plane northings/eastings. This means that an alternative means for determining geospatial coordinates would need to be identified so that the EC data could be used and maintained in an elevation registry. See section A.8 below for a further discussion of geocoding options. Non-standard addresses. Of the 1,524 Pinellas County EC records, 371 (24.3%) had no street addresses; most of these had some form of lot number, but some listed only a name, e.g., "Tooke, O.J. UNREC" in the address column. Most of the 1,153 addressed records did have different columns for the property's house number, house suffix, street prefix, street name, street suffix, and apartment number, but many of the records had the street suffix merged with the street name. Pinellas County’s database was designed in a way that made it very difficult to link the county's street addresses with those from the ECs. The Pinellas County database had all address information merged in a single column, and many of the addresses were very complex and not suited for "normal" address matching. After reviewing the EC data from these four counties, Dewberry concluded that quality control review changes would be needed in the way community data are entered into a database such as that developed for the CRS program before it could be reliably used to populate an elevation registry. A.9 Tools for Georeferencing existing Elevation Certificates Latitude and longitude are required to georeference/geocode a structure in a GIS, but such geographic coordinates are an optional entry on FEMA Form 81- 31, July, 2000. As a result, a very small percentage of ECs include this optional entry, except when Global Positioning System (GPS) procedures are used for the survey. Even when GPS procedures are used for the surveys, the latitude and longitude are not always provided since the entry is optional. Without geographic coordinates, or comparable UTM or State Plane coordinates, street addresses alone are not adequate to accurately determine the location of a structure in a GIS. For this study, without accurate geocoding, existing ECs could not be used as "ground truth" to validate the accuracy of the alternative elevation strategies discussed below. For a potential elevation registry, the lack of accurate geocoding would impact the accuracy of revisions to the elevation registry when new DFIRMs or other changes would normally dictate the need for maintenance and updating of registry records. GPS Surveys. The most accurate and direct way to establish latitude/longitude for an EC is to use GPS procedures which automatically yield geographic coordinates of all points surveyed. When using differential GPS procedures with survey-grade receivers, GPS is capable of producing centimeter-level accuracy; however, when using a single mapping-grade receiver, GPS produces positions with error on the order of 10 meters horizontally and 20 meters vertically. In surveying buildings, GPS antennas cannot be placed immediately adjacent to a building because the building itself would block many of the GPS satellite signals, and visible satellites would suffer from multipath errors -- causing errors in x/y/z coordinates so surveyed. For these reasons, differential GPS procedures are used to survey temporary benchmarks in front of each building to be surveyed (often PK nails driven into the street pavement), followed by conventional surveys from the PK nails to the survey reference points being surveyed on or near the building, e.g., bottom of front door, top of foundation, LAG or HAG point. Such high-accuracy GPS/conventional survey procedures were used for the ECs used as "ground truth" in Charlotte-Mecklenburg County, NC and Prince George's County, MD. For these two counties, each street address was directly linked to the surveyed latitude/longitude of the front door of each building. Digital Orthophotos. An accurate but indirect way to establish latitude/longitude for an EC is to utilize digital orthophotos, combined with some other means for identifying which rooftop image on the orthophoto goes with each street address. In Harris County, TX, georeferencing of existing ECs was performed by the Harris County Flood Control District (HCFCD) which had a GIS database that linked street addresses to a vector polygon bounding each parcel/lot to which a street address is referenced. For each street address for which an EC was provided in Houston, Mike Walters (713-684-4173) at HCFCD established the parcel/lot polygon from the HCFCD's GIS database; overlaid each parcel/lot polygon on top of the city's digital orthophotos, manually selected the rooftop within that parcel/lot, and then selected the latitude/longitude of the rooftop centroid. When performed correctly, this is a perfectly acceptable way to geocode the centroid of a building from its street address. As demonstrated in a Dewberry study for FEMA Region 5 in 1998 and this current study, documented below, it is relatively easy to determine the latitude and longitude of buildings from various forms of digital orthophotos commonly used and available nationwide, but it is very difficult to identify the correct street addresses for those buildings from commonly-used geocoding software programs. GIS Polygons. A slightly less accurate way to georeference an EC, based on its street address, is to utilize tax parcel polygons from the community's GIS, but without refinement by digital orthophotos. The tax assessor and city planner are among the officials who typically utilize such a GIS that digitizes the parcel/lot perimeter boundary lines, as with the HCFCD database above. Such polygons are typically digitized by using Computer Aided Design and Drafting (CADD) coordinate geometry (COGO) procedures to enter the boundary survey information (line distances and angles) for each boundary line segment bounding a tax lot or parcel. However, rather than overlaying the parcel/lot polygon over a digital orthophoto to manually select the location of the building centroid, the GIS itself is used to automatically place a centroid at the center of the lot or some alternative means to estimate the location of the main building on the lot. Resulting geocoding errors are insignificant on small lots, but could be significant on large lots. Several states, notably Maryland (http://www.op.state.md.us/data/mdview.htm) and New York (http://www.nysgis.state.ny.us/inventories/orps.htm) provide statewide parcel data to the public (for a fee), as do numerous local and county entities. However, if New York is representative, some entities may begin restricting access to this type of information due to increased security concerns. Commercial Geocoding Services. A number of commercial companies offer georeferencing software and/or services to perform either or both of the following: (a) geocoding -- providing latitude/longitude values for a known list of street addresses, but excluding P.O. box addresses or rural route addresses, and (b) reverse geocoding -- providing street addresses for a known list of geographic coordinates. For establishment of an elevation registry, FEMA has need for both geocoding and reverse geocoding. As discussed above, the geocoding of ECs with known street addresses would be required for records derived from existing ECs. However, an elevation registry could also be populated by any of the aerial remote sensing techniques described in this study, for which geographic coordinates (latitude/longitude) are known but street addresses are unknown when surveyed from the air; this would require reverse geocoding. Commercial geocoding solutions typically rely on street centerlines with address ranges or zip-code points that serve known address ranges. Linear interpolation of the target address between the low and high address ranges is used to identify where along the block and on which side of the block (odd or even) the address falls. If the address ranges of the street centerlines are larger than the real addresses, as is most often the case, a “clustering” effect can happen, with all of the addresses landing at the “low” address end of the block. Standard offsets or setbacks of the house from the street are usually included and can sometimes be varied. If an address cannot be found, a default location at a zip-code centroid is sometimes returned. Accuracy of geocoding relies on the spatial accuracy and currency of the street centerlines and the accuracy and completeness of the street names and address ranges used. Rural route addresses, post office boxes, and lot and block numbers are addressing systems that do not lend themselves to geocoding. Four leading commercial georeferencing services were evaluated by Dewberry to test the geocoding of a small sample set of 53 ECs in the City of Houston, TX. One of these four services readily admitted that procedures were approximate and could not distinguish between neighboring houses or houses across the street from each other. Table 8 summarizes some of the differences between the other three geocoding services for the same addresses; the names of these services will remain anonymous as results may vary widely in different communities, and none was clearly superior to the others. The complete results are at APPENDIX F with geocoded coordinates compared with "ground truth" coordinates provided by Harris County derived from parcels and rooftops identified on digital orthophotos (described above). All of the geocoding services delivered approximate positions, but yielded positioning errors of several hundred feet. Thus, none of the commercial services evaluated can be relied upon to distinguish between neighboring houses or houses across the street. Still, as with the LOMA 2000 database, such crude geopositioning is better than no geopositioning. Table 8 — Comparison of Geocoding Services Geocoding Service A B C Address matching 50 of 53 52 of 53 53 of 53 ?N average (Northing) 129.80 ft 152.97 ft 178.79 ft ?E average (Easting) 362.97 ft 206.37 ft 212.68 ft ?N maximum (Northing) 1055.66 ft 1127.10 ft 894.98 ft ?E maximum (Easting) 7684.08 ft 1469.53 ft 767.13 ft ?N 95th percentile 360.13 ft 509.85 ft 428.63 ft ?E 95th percentile 449.52 ft 600.65 ft 533.35 ft Horizontal errors at the 95% confidence level * 575.99 ft 787.86 ft 684.24 ft * Because of systematic formulas that interpolate street addresses, there is no reason to assume that geocoding errors follow a normal distribution, therefore, the 95th percentile method is warranted and the horizontal (radial) error at the 95% confidence level is assumed to equal the square root of [(?N 95th percentile)2 + (?E 95th percentile)2]. A.10 Conversion of paper Elevation Certificates to digital format As noted previously, most communities do not maintain their ECs in an electronic format. These records tend to be stored in scattered locations and may be cost- prohibitive to retrieve. A 1995 study by ISO on the retrieval and conversion of ECs to a database format resulted in over 30,000 records that were provided to FEMA and to the CRS communities that originally supplied the information. At the time, significant difficulties were encountered in creating a database from information documented on more than 7 different FEMA EC forms and provided in hardcopy format of varying quality, completeness, and legibility. Scanning and Optical Character Recognition (OCR) software was tested by ISO on the conversion project and abandoned due to the poor quality of the scans. Dewberry contacted S.A.I.D. Inc. regarding the feasibility and cost of scanning paper ECs and digitizing the pertinent data into a database. S.A.I.D. Inc. was contacted because they provide similar scanning services for FEMA’s Engineering Study Data Packages and have provided data entry for other services. S.A.I.D. estimated $5 per EC for double-entry digitization (whereby two different personnel enter the data which is then compared for quality control purposes to detect differences/errors) and 300 dpi PDF files for each EC form. The assumptions used for this cost estimate are as follows: ? The ECs would be 2-sided forms or 2 pages. ? There would be a minimum of 50,000 ECs to be digitized. ? Within those 50,000 EC forms, there will be up to 7 types of EC forms. These forms will have varying amounts of data, similar in nature to the current FEMA Form 81-31. ? Up to 50% of the EC forms may be handwritten. ? The output will be a PDF multi-page image (2 pages) and an ASCII data file, including the image file name, to satisfy the format of the data dictionary for the elevation registry. ? The size of the 300 dpi PDF file for each EC form would be approximately 138 Kb. Strategy B — Maximize use of Airborne Remote Sensing B.1 Photogrammetry Conventional (Vertical) Photogrammetry. Photogrammetry is that branch of surveying that deduces the physical three-dimensional measurements of objects from measurements on stereo photographs that photograph an area from two or more different perspectives. The 3rd dimension (elevation) is normally mapped as contours of equal elevation, or as spot heights for which the z-value (elevation) of each point is carefully measured. Spot heights are normally mapped at tops of mountains, bottoms of depressions, centers of road intersections, tops of dams or dikes, or other locations where there is a need for an accurate elevation value; but spot heights can also be mapped at LAG or HAG points (if these points are visible on both of the stereo photographs) or at the four corners of a building, for example. Normally, vertical stereo photography is flown of entire communities with numerous adjacent/parallel flight lines. The area imaged with each photograph overlaps the adjoining photo (before and after) in the same flight line by about 60% and has 10-20% sidelap with photos from adjoining flight lines. With 60% forward overlap, all of the terrain area can be seen on at least two successive photos, and up to 30% of the terrain area can be seen on three successive photos. The camera's focal length and the aircraft's flying height dictate the accuracy of elevation data surveyed photogrammetrically. When elevation data are acquired, mapping cameras with a standard 6" focal length are normally used, and flying heights are varied to satisfy requirements for a specified contour interval to be mapped. Subsequently, when mapped to National Map Accuracy Standards, spot heights measured from this stereo photography will have 90% of the elevations accurate to one-fourth of the contour interval or less, with no spot height elevation errors larger than one-half the contour interval. For example, to produce a map with 2 ft contours, it is common to acquire the aerial photography from an altitude of 4,000 ft above the mean elevation of the terrain being mapped; then, at least 90% of the spot height elevations should be accurate to 0.5 ft, and the remaining 10% of the spot height elevations should be accurate to 1.0 ft. Whereas the National Map Accuracy Standard expresses accuracies at the 90% confidence level, the new National Standard for Spatial Data Accuracy requires accuracies to be expressed at the 95% confidence level, as used throughout this report. The three-dimensional (3-D) coordinates (latitude, longitude, and elevation) of any point can be surveyed photogrammetrically if the point can be seen on two or more stereo photos. Some ground points cannot be seen in stereo when tall trees or buildings block the view to the ground from one or more perspectives. For the purpose of this study, vertical aerial photography (aimed straight down) can accurately survey rooftops and many points on the ground including LAGs and HAGs, but basement windows cannot normally be seen in stereo because one photo might see a basement window, but the second photo will look straight down on the house, and the third photo will see the opposite side of the house -- thus no stereo images of the same basement window feature. Conventional Photogrammetry with Measured Offsets. As described at www.nap.usace.army.mil/GIS/fwrs.htm, the Philadelphia District of the U.S. Army Corps of Engineers (USACE) executed the structure inventory portion of the Susquehanna River Flood Warning and Response System (FWRS) in 2000. A part of the FWRS involved the Corps using a multi-technology method to survey the top of bottom floor elevations of thousands of floodprone houses along the Susquehanna River in Pennsylvania. The District hired a photogrammetric firm (BAE/ADR) to establish photogrammetric spot heights on the ground adjacent to the corners of each house (as many corners as could be seen in stereo). The Corps also hired a survey firm to measure the offset distances up/down from one of the surveyed spot heights (per structure) to indirectly compute the top of bottom floor elevation and other elevations relative to the spot heights. The specifications for the photogrammetry were that the spot heights should satisfy National Map Accuracy Standard for 2 ft photogrammetric contours. Figure 2 (left) illustrates how spot heights (shown here as red dots) might initially appear when photogrammetric x/y/z coordinates are provided as spot heights. At this point, there is no basis for reference. Normally only two or three spot heights can be established on the ground adjacent to the corners of any building because, from one or two directions, the building itself blocks the view of one of the photographs needed to make stereo measurements on the ground. With only georeferenced points, it is difficult for the land surveyor to determine which house, and which corner of which house those coordinates (dots) pertain to. Figure 2 (right) illustrates how those same spot heights might be plotted on top of a digital orthophoto or other base map so that the surveyor can determine the spot height, for each building, most appropriate for use for measurement of vertical offsets. Not all surveyors have GIS tools for overlaying such "dots" on top of orthophotos or other base maps. Similarly, for home owners, they too would not be able to easily determine which "dots" pertained to their house. However, most communities have GIS specialists who could easily perform such tasks on a community-wide basis. Furthermore, home owners may actually be able to recognize the pattern of dots relative to streets and homes in the area. Figure 2 — Photogrammetric Spot Heights To verify the accuracy of this method, Dewberry hired a survey firm to use GPS and conventional survey procedures to directly survey the top of bottom floor elevations of 40 of the same houses previously mapped by the Corps of Engineers. The comparisons of the expedient photogrammetric method vs. ground survey method are at APPENDIX G for the 40 houses selected at random. Assuming the GPS surveys were correct with all errors attributed to the photogrammetric surveys and/or offset measurements, the vertical accuracy of the top of bottom floor elevations by this method was 1.19 ft at the 95% confidence level. This exactly satisfied the mapping standard for 2 ft contour interval accuracy and proves that this is a viable method for obtaining accurate structural elevation data. A total of 31 of the 40 top of bottom floor elevations (77.5%) were accurate within 0.5 ft. Oblique Photogrammetry - Pictometry. Because Pictometry's oblique imagery will be new to most readers of this report, the company-provided information is included at APPENDIX H to explain the products and how they are acquired and used. Figure 3 shows a sample Pictometry oblique image, photographed from an elevation of approximately 2,000 feet above the mean terrain. Figure 4 provides samples of Pictometry imagery, zoomed-in from four different perspectives. When features are not in the shadows, the oblique views enable the GIS analyst to see and measure the bottom of doors, tops of foundations, presence of walk-out basements, or basement windows, for example. In fact, this is the only airborne remote sensing technology that is able to "see" such features needed to determine the elevation of the lowest floor. All other airborne remote sensing technologies are able to survey the rooftop and the LAG and HAG, for example, but only estimates or infers the lowest floor or top of bottom floor elevation. However, as can be seen from the four views at Figure 4, because of shadows and shrubbery, it is nearly impossible on this particular residence to see basement windows or flood vents from any of the four views. Dewberry learned during this study that Pictometry's elevations are actually relative rather than absolute. For example, Pictometry can accurately measure the distance up from the ground (e.g., LAG point) to the top of foundation, and then subtract 8 feet to compute the top of bottom floor elevation of the basement, but if it doesn't know the absolute elevation of the ground, then the absolute elevation of the lowest floor will also be in error by the error in the ground elevation. The next three subsections describe Pictometry projects with different techniques for obtaining digital elevation data from which relative height differences were determined and tested by Dewberry. 2-View Pictometry with USGS DEM. Pictometry had flown over 100 counties/communities by this method, but none where Dewberry had georeferenced ECs to serve as ground truth. However, Pictometry had flown over Prince George's County, MD, while flying Montgomery County, MD, Washington D.C., and Arlington County, VA and had acquired imagery of Prince George's County from two directions instead of four. Pictometry provided measurements of 29 buildings, using 2-view images, where Dewberry had georeferenced ECs to serve as ground truth. Of these 29, three houses could not be accurately surveyed for various reasons and their top of bottom floor elevations were left blank; they were obscured by trees or neighboring buildings on the two sides of the houses where photos were available and they did not have photos of the front and rear of these houses. Several other houses were measured with questionable accuracy because the analyst could not clearly see whether or not these houses had basement windows. Furthermore, Pictometry used USGS DEMs to determine the elevation of the ground, from which offset measurements were made to determine top of bottom floor elevations. See APPENDIX I. Of the houses measured, the average absolute error in the top of bottom floor elevation was 2.61 ft; the error at the 95% confidence level was 6.34 ft; and the largest error was -10.75 ft, caused by identifying a basement where none actually existed. While unimpressive in and of itself, Dewberry still considered this potentially encouraging for three reasons: (1) the 2.61 ft average elevation error was strongly influenced by several houses where the wrong assessment had been made regarding the presence or absence of basement windows, and the Pictometry analyst can annotate the database with a confidence level indicator, rating houses where he/she is highly confident that the house has a basement or doesn't have a basement, or various degrees of diminished confidence; (2) confidence levels should be increased if the normal 4-view images were available to view each building from all four sides rather than from only two sides as in PG County; and (3) because USGS DEMs (with potentially large elevation errors) had been used as the reference elevation for each building. Figure 3 — Sample Pictometry Oblique Photo Figure 4 — Sample 4-View Pictometry Images 2-View Pictometry with LIDAR DEM. Dewberry subsequently provided Pictometry with a LIDAR dataset of Prince George's County, MD, and these same 29 buildings were re-measured. This time, the analyst provided top of bottom floor elevations for all 29 homes. The average (absolute) error in the top of bottom floor elevation was 2.53 ft; the top of bottom floor error at the 95% confidence level was 4.66 ft; and the largest top of bottom floor error was 5.63 ft, again influenced by the same factors (1) and (2) above, but without major errors from factor (3). See APPENDIX I 4-View Pictometry with Spot Heights Finally, to eliminate factors (1) and (2), Dewberry decided to survey additional houses in Arlington County, VA for comparison with the 4-view Pictometry images previously available. Since LIDAR data was not available for this county, Dewberry provided Pictometry with the surveyed spot height elevations at three corners of 27 houses. This is essentially the most accurate DEM that could be provided, allowing Dewberry to isolate errors from Pictometry's measurement process from errors in the DEM. Each of these houses had dense tree cover on at least two sides. For these 27 houses, the average error in the top of bottom floor elevation was 1.59 ft; the error at the 95% confidence level was 5.01 ft; and the maximum error was 5.85 ft. See APPENDIX I This time, many houses again had misidentified basements. Pictometry Accuracy Summary Table 9 summarizes the accuracies achieved in three different evaluations of Pictometry datasets compared with ECs. Table 9 — Pictometry Accuracy Comparisons Pictometry Dataset Evaluated LAG errors 95% Conf. Average absolute LAG errors HAG errors 95% Conf. Average absolute HAG errors TBF errors 95% Conf. Average absolute TBF errors Prince George's County, MD w/USGS DEM 3.96 ft 1.65 ft 3.61 ft 1.54 ft 6.34 ft 2.61 ft * Prince George's County, MD w/ LIDAR 3.99 ft 1.62 ft 3.83 ft 1.70 ft 4.66 ft 2.53 ft ** Arlington County, VA, w/surveyed spot heights N/A N/A N/A N/A 5.01 ft 1.59 ft * Initially, only 26 of 29 homes were surveyed for top of bottom floor (TBF) elevations; three were not surveyed because existence of basements could not be determined ** Subsequently, all 29 homes were surveyed for top of bottom floor elevations. The same Pictometry images were used. LIDAR results would have been better if the analyst had not guessed on the questionable basements. Photogrammetry Conclusions The major photogrammetry conclusions are as follows: ? Conventional vertical photography can survey LAG and HAG elevations as well as spot heights of the terrain at multiple corners of a structure; but this technology cannot directly survey the lowest floor elevations because the bottom of front door or other survey "target points" are not normally visible on aerial photographs looking straight down at rooftops. ? When conventional photogrammetric spot heights are combined with on- site tape measurements from the ground to bottom of front door or top of foundation, for example, then the top of bottom floor elevations could be computed with errors comparable to the elevations interpolated from topographic contours, i.e., 90% of top of bottom floor elevations accurate within ˝ the contour interval and the remaining 10% accurate within the full contour interval. However, errors in measuring the offset distances could also be a factor. For the 40 Susquehanna structures, the vertical error at the 95% confidence level was 1.19 ft, equivalent to 2' contours, and 31 of 40 lowest floor elevations were accurate within 6 inches. However, it must be recognized that discrepancies were not necessarily due to errors in the photogrammetry but partly due to two different surveyors measuring vertical offsets by different methods, or selecting different points on the houses on which to base the lowest floor measurements. ? Oblique aerial photography, from Pictometry for example, cannot directly survey LAG and HAG elevations or spot heights, but can indirectly survey top of bottom floor elevations relative to elevations of surrounding terrain. However, in all three tests performed, there difficulties in detection of basements, causing top of bottom floor elevations to have average errors between 1.59 and 2.61 ft, and top of bottom floor errors at the 95% confidence level between 4.66 ft and 6.34 ft. For these reasons, Dewberry concludes that Pictometry imagery can not be reliably used to determine top of bottom floor elevations but, instead, has its best value for other applications, such as providing a "birds' eye" view of the property so an insurance agent or others can see the house to be insured, and/or to provide a means to check for unauthorized construction. The exception is in some areas where there are no basements, such as in many Florida counties; then Pictometry can provide more information without fear that a wrong assessment is made regarding the presence or absence of basements. B.2 Light Detection and Ranging (LIDAR) LIDAR collects thousands of spot heights every second of flight, currently with up to 100,000 laser pulses per second. LIDAR is most commonly flown of entire counties or communities to establish the elevation layer of their GIS. Each LIDAR pulse can receive multiple returns, yielding elevations (actually x/y/z coordinates) of features mapped. The first return for each pulse provides the elevation of the first thing hit by the pulse, to include treetops and rooftops. Some of the light from each laser pulse penetrates through or between the trees and hopefully hits the ground for use in establishing a bare-earth digital terrain model (DTM). With dense vegetation, LIDAR pulses might never penetrate the vegetation to reach the ground. Vegetated features are "soft" where there is a difference between the elevation of the first and last return. Other features, including bare earth, sand, concrete, rock, short grass, and building rooftops are "hard" where the elevations from the first and last returns are the same, i.e., where there is no LIDAR penetration of the feature. To generate a bare-earth DTM, the LIDAR data is post-processed by computer algorithms to "remove" buildings and vegetation. Figure 5 provides an example of LIDAR processing. The left image shows LIDAR last-return elevations prior to post-processing. The center image shows the bare-earth DTM after post- processing for removal of buildings and vegetation and interpolation to fill in the missing spaces where elevation points were deliberately removed. The right image shows where there is no longer elevation data. There is no data in the black areas either because there were no LIDAR returns in the first place (in water) or because those returns were deliberately deleted during post-processing because they mapped trees, rooftops or other elevated features above the bare- earth DTM that was needed. Note the bridge decks that were deliberately "cut out" for hydro-enforcement of the streams and canal for hydraulic modeling purposes. The reverse of the right image would show where LIDAR elevation points remain after removal of trees and buildings. As with the right image, a surveyor or home owner could probably navigate to specific streets just by recognizing the LIDAR dot pattern for the streets and buildings when using the LIDAR's bare-earth point file. Figure 5 — LIDAR Surfaces Before and After Post-Processing When LIDAR data is used to automatically determine the LAG and HAG of structures, or to estimate their lowest floor elevations, it is preferred to have a GIS file of structure footprints, as maintained by many communities. Such footprints are most commonly mapped photogrammetrically or with rooflines digitized from digital orthophotos. Figure 6 shows an example of building footprints (top left), and those same footprints with a surrounding buffer (top right), shown in red, that can be automatically generated by a GIS with any buffer width desired. Some community GIS files have these footprints georeferenced so that the street addresses are known for each building. Other community GIS files do not link their footprints to street addresses. For analyses of the various technology sub-options for both LIDAR and IFSAR, the base scenario assumption is that footprints are linked to street addresses; but separate calculations are also performed with the assumption that street addresses are not linked to footprints. With bare-earth DTMs, there is no need for a buffer zone because each footprint can directly "cut" the DTM to establish the LAG and HAG. Figure 6 — Examples of Building Footprints and Buffer, Centroids, and Parcel Polygons Some communities do not have footprints, but they do have building centroids linked to street addresses. The center left image at Figure 6 shows examples of such centroid points, and the center right image shows those centroids superimposed on top of digital orthophotos. Still other communities may have neither footprints nor centroids but instead have parcel polygons that are linked to street addresses as shown at the bottom left image at Figure 6 and superimposed on top of digital orthophotos at the bottom right image. The superimposition of footprints, centroids, or parcel polygons on top of digital orthophotos provides graphic orientation but not street addresses. Such addresses must be established in the GIS database. Of all these options, footprints are preferred because they can be overlaid on top of a bare-earth LIDAR TIN to "cookie cut" the LIDAR data to determine the LAG and HAG. In this report, the "with footprint" process will be abbreviated as the "w/FP" method. With either centroids or parcel polygons, there is less certainty in the computation of LAG and HAG elevations from LIDAR data because the location of LAG and HAG points must be estimated; however, when either centroids or parcel polygons are overlaid on top of digital orthophotos, it is a simple GIS task to generate footprints around the perimeters of visible rooftops. When footprint files are not available, Dewberry uses Computational Consulting Service, Inc. (CCS) which has developed sophisticated algorithms to process the raw LIDAR "point cloud" dataset in order to detect building locations and search for the LAG and HAG. This "no footprint" process will be abbreviated as the "NoFP" method. When LIDAR data is widely spaced (e.g., 4-5 meter post spacing), CCS computer algorithms have a much harder time detecting building locations than when the post spacing is narrower. With 4-5 meter point spacing, it is possible to have only one LIDAR pulse hit a rooftop, making it impossible to determine the shape of the roof for estimating the shapes of the building footprints and their buffers. With wide post spacing, it is even possible that no LIDAR pulse hits a rooftop. One purpose of this research project is to determine how well narrower post spacings perform in helping these computer algorithms to estimate building footprints so that buffers can be accurately established. As described below, CCS processed four different LIDAR datasets and prepared the LIDAR Automated Data Extraction Report at APPENDIX J that describes procedures for extraction of buildings from LIDAR data, determination of main parameters of buildings from LIDAR data, and determination of additional parameters of buildings using 3D models created of those buildings. LIDAR data of Charlotte/Mecklenburg County, NC. Raw LIDAR data of Mecklenburg County was provided by the Charlotte-Mecklenburg Storm Water Services. The LIDAR data was flown by EarthData in 2003 and had nominal post spacing of approximately 16 feet. This is considered to be a low resolution dataset because each house might have only one LIDAR pulse hit an entire rooftop, and some houses might even have no pulse hit an entire rooftop if the width of the house, for example, is less than 5 meters (16 feet). Building footprint files were also provided by the County for some of the buildings. Dewberry chose a test area that included 2617 buildings with footprints. CCS' "NoFP" processing was performed which automatically identified 2270 of those 2617 buildings for which one-to-one GIS relationships were identified, missing 347 (13.3%) of the buildings; Dewberry considered this a success because of the low resolution dataset that might have no LIDAR pulse, or perhaps only one or two pulses hit many of the rooftops. Some of the one-to- one "misses" were actually cases where there were one-to-many or many-to-one relationships because of rows of townhouses, for example, with different rooftop elevations for different units, but for which multiple units may have only a single footprint. Dewberry then determined that 217 of these 2160 buildings also had ECs for use as "ground truth" elevations, so these 217 buildings became the basis for comparison of CCS' automated "NoFP" method to be used when there are no footprints, and Dewberry's automated "w/FP" method to be used when there are footprints. Of these 217 ECs, 215 had LAG elevations, and 108 had HAG elevations (HAG elevations were not required on earlier versions of FEMA form 81-31). The spreadsheet that computes the overall statistics for LAG and HAG elevations, comparing CCS' "NoFP" method (without footprints) with Dewberry's "w/FP" method (with footprints) is at APPENDIX K — LIDAR Accuracy Analysis (Mecklenburg County, NC). The results are summarized in Table 10. Table 10 — Mecklenburg County LIDAR Accuracy Comparison Mecklenburg County, NC, LAG/HAG from LIDAR with 16 ft nominal post spacing "NoFP" LAG "w/FP" LAG "NoFP" HAG "w/FP" HAG LIDAR average post spacing 16 ft Number of Houses 215 108 Standard Deviation 1.91 ft 1.13 ft 1.84 ft 1.15 ft Average (absolute) Error 1.22 ft 0.87 ft 1.11 ft 0.71 ft Minimum Elevation Error -9.39 ft -4.64 ft -5.75 ft -5.75 ft Maximum Elevation Error 9.47 ft 3.63 ft 9.44 ft 4.05 ft 95th Percentile Error 3.79 ft 2.82 ft 3.57 ft 2.09 ft 90th Percentile Error 2.55 ft 1.98 ft 2.28 ft 1.61 ft 85th Percentile Error 1.95 ft 1.43 ft 1.71 ft 1.16 ft ? For CCS' "NoFP" method (no footprints), the average LAG elevation error was 1.22 ft, but the LAG elevation error at the 95% confidence level was 3.79 ft. Similarly, the average HAG elevation error was 1.11 ft, but the HAG elevation error at the 95% confidence level was 3.57 ft. The spreadsheet at APPENDIX K shows some larger outlier errors that indicate potential systematic errors when the two methods both yield poor results. ? For Dewberry's "w/FP" method (with footprints), the average LAG elevation error was 0.87 ft, but the LAG elevation error at the 95% confidence level was 2.82 ft. Similarly, the average HAG elevation error was 0.71 ft, but the HAG elevation error at the 95% confidence level was 2.09 ft. ? For systematic errors, it is possible that the ECs or footprints include errors in horizontal position, that the LAG/HAG elevations on the ECs may include errors, or that Dewberry's attempts to establish one-to-one GIS relationships between ECs, building footprints, and LIDAR data failed for some records — all potentially causing the wrong elevations to be compared. ? It is interesting to note that errors at the 95% confidence level are nearly twice as large as errors at the 85% confidence level. LIDAR data of Prince George's County, MD. LIDAR data of Prince George's County was acquired by Waggoner Engineering, Inc. in 2000 and had nominal post spacing of approximately 8 feet. This is considered to be a medium resolution dataset because each house should have several LIDAR pulses hit individual rooftops. See APPENDIX K — LIDAR Accuracy Analysis (Prince George's County, MD). The results, summarized at Table 11, are considerably better than Table 10, demonstrating the benefits of narrower post spacing for this purpose. By using 8 ft spacing instead of 16 ft, the LAG elevation errors at the 95% confidence level were reduced from 3.79 ft to 1.68 ft when using CCS' "NoFP" method (no footprints), and they were reduced from 2.82 ft to 2.02 ft when using Dewberry's "w/FP" method (with footprints). Table 11 — Prince George's County LIDAR Accuracy Comparison Pr. George's County, MD, LAG/HAG from LIDAR with 8 ft nominal post spacing "NoFP" LAG "w/FP" LAG "NoFP" HAG "w/FP" HAG LIDAR average post spacing 8 ft 8 ft 8 ft 8 ft Number of Houses 579 579 579 579 Standard Deviation 0.57 ft 0.61 ft 1.71 ft 0.60 ft Average (absolute) Error 0.53 ft 0.80 ft 0.77 ft 0.65 ft Minimum Elevation Error -5.15 ft -4.28 ft -1.76 ft -3.11 ft Maximum Elevation Error 1.91 ft 2.51 ft 29.69 ft 4.91 ft 95th Percentile Error 1.68 ft 2.02 ft 2.40 ft 1.82 ft 90th Percentile Error 1.11 ft 1.54 ft 1.34 ft 1.44 ft 85th Percentile Error 0.82 ft 1.19 ft 0.91 ft 0.95 ft LIDAR data of Harris County, TX. Raw LIDAR data of Harris County was provided by TerraPoint and had nominal post spacing of approximately 5 feet. This is a high resolution dataset that became available for evaluation during the progress of the study. However, this dataset was flown with an older sensor not optimized for foliage penetration, and there were considerable difficulties with the old ECs that lacked geographic coordinates. When the ECs were geocoded, they appeared to be far out of registration with the LIDAR data, causing CCS and Dewberry to be unsure of the validity of comparing the LIDAR data with EC data that appeared to be questionable at best and erroneous at worst. Furthermore, compounding this issue is the fact that Houston suffers from severe subsidence, and there was a distinct possibility that the land subsided significantly between the time when the ECs were surveyed (up to 20 years ago) and recent years when the LIDAR was flown. For these reasons, Dewberry abandoned any attempts to evaluate the Harris County LIDAR dataset. Fortunately, an alternative high resolution LIDAR dataset of Beaufort County, SC was already available. LIDAR data of Beaufort County, SC. LIDAR data of Beaufort County was provided by the county's GIS coordinator. The LIDAR data was flown by Laser Mapping Specialists, Inc. (LMSI) and had nominal post spacing of approximately 4 feet. This is the highest resolution dataset evaluated in this study; each house should have many LIDAR pulses hit individual rooftops. See APPENDIX K — LIDAR Accuracy Analysis (Beaufort County, SC). Table 12 — Beaufort County LIDAR Accuracy Comparison Beaufort County, SC, LAG/HAG/TBF from LIDAR with 4 ft nominal post spacing "NoFP" LAG Elevations "w/FP" LAG Elevations "NoFP" HAG Elevations "w/FP" HAG Elevations "NoFP" TBF Elevations Average post spacing 4 ft 4 ft 4 ft 4 ft 4 ft Number of Houses 27 38 27 38 27 Standard Deviation 0.43 ft 0.28 ft 0.39 ft 0.39 ft 0.78 ft Average (abs) Error 0.42 ft 0.28 ft 0.37 ft 0.95 ft 2.93 ft Minimum Error -1.37 ft -0.65 ft -0.25 ft 0.23 ft -0.19 ft Maximum Error 0.27 ft -0.55 ft 1.19 ft 1.73 ft 3.63 ft 95th Percentile Error 1.09 ft 0.59 ft 0.97 ft 1.60 ft 3.58 ft 90th Percentile Error 0.91 ft 0.54 ft 0.84 ft 1.50 ft 3.50 ft 85th Percentile Error 0.77 ft 0.51 ft 0.77 ft 1.40 ft 3.44 ft The results summarized at Table 12 are considerably better than Table 11 for all of the statistics shown in these two tables, again demonstrating the benefits of narrower post spacing. By using 4 ft spacing instead of 8 ft, the LAG elevation errors at the 95% confidence level decrease from 1.68 ft to 1.09 ft when using the "NoFP" method and from 2.02 ft to 0.59 ft when using the "w/FP" method. Similarly, the HAG elevation errors at the 95% confidence level decrease from 2.40 ft to 0.97 ft when using the "NoFP" method, and decrease from 1.82 ft to 1.60 ft when using the "w/FP" method. CCS's estimation of top of bottom floor elevations, using "NoFP" methodology, yielded errors of 3.58 ft at the 95% confidence level. This was the only dataset that yielded top of bottom floor elevations that could even be considered for the registry, and these results are this good in large part because the test houses in Beaufort County had no basements. If these houses had basements, the "NoFP" top of bottom floor elevation accuracies would probably have been poorer. Overall, for estimation of LAG elevations, CCS' "NoFP" method yielded errors of approximately 1.09 ft at the 95% confidence level. Similarly, Dewberry's "w/FP" method yielded LAG elevation errors of 0.59 ft at the 95% confidence level. This equals the accuracy expected of data equivalent to 1 ft contours, although most LIDAR datasets are compiled to meet 2 ft contour interval standards. Throughout the remainder of this study, LIDAR data will be evaluated as though equivalent to 2 ft contours — in spite of the fact that this particular LIDAR dataset in Beaufort County is considerably more accurate than 2 ft. LIDAR Conclusions. At the 2004 International LIDAR Mapping Forum (ILMF), several presentations pointed out the fact that LIDAR firms in Europe and Japan routinely collect much higher resolution LIDAR data than in North America, with some countries now collecting data at extremely high resolution, i.e., up to 28 points per square meter, whereas in the U.S. there is normally one LIDAR point for several square meters. The reason for this difference is that LIDAR data is most commonly used for engineering design applications in Europe and Japan and is collected with helicopter-based sensors, whereas LIDAR data is most commonly used for mapping applications in North America and is collected with fixed wing aircraft designed for flying longer distances at higher altitudes. Regardless of the type of aircraft used, LIDAR systems now being sold in the U.S. have extremely high pulse repetition rates, now up to 100,000 pulses per second. This alone will cause high resolution LIDAR datasets to become the norm rather than the exception in North America. The results achieved in Beaufort County, SC make it reasonable for Dewberry to proceed with this study assuming that LAG and HAG elevations, when footprints are available to "cookie cut" the LIDAR TIN surface, can be derived with accuracies comparable to 2 foot contours, i.e., accurate to 1.20 ft or less at the 95% confidence level. When proceeding on this assumption, LIDAR datasets should have independent confirmation of the overall accuracy of the data. In all cases, the availability of building footprints makes it possible to obtain the most accurate LAG/HAG elevations from existing LIDAR datasets. B.3 Interferometric Synthetic Aperture Radar (IFSAR) For a quick reference to Interferometric Synthetic Aperture Radar (IFSAR), the reader is referred to Intermap's web site at www.intermaptechnologies.com, Product Handbook. IFSAR products have traditionally consisted of ortho-rectified radar images (ORI) and digital surface models (DSMs). An ORI is a grayscale image of the earth's surface that has been corrected to remove geometrical distortions that are a normal part of the imaging process. Although they are similar to black and white aerial photographs, ORIs differ because, instead of being made of visible light, the radar pulses the ground with "flashes" of radio waves which then return from imaged features to the antennas to give distance and intensity measurements. The key feature of ORI imagery is that it provides a means of viewing the earth's surface in a way that accentuates features far more than is possible with aerial photography. The radar looks to the side of the aircraft and casts "shadows" that enable the user to visually perceive the elevation information in the image. See Figure 7 for comparison of ORI imagery with traditional digital orthophotos. IFSAR DSMs are derived from the return signals received by the two radar antennas on the aircraft. The signals bounce off the first surface they strike, making the DSM a representation of any object large enough to be resolved. This includes buildings, vegetation and roads, as well as natural terrain features. DSMs ma