Transportation and Climate Change Clearinghouse
2 - Study Methodology
Eight of the nine scenarios of sea level rise used in this study are based on the Third Assessment Report, because this study was begun before the release of the IPCC's Fourth Assessment Report; a ninth scenario from the Fourth Assessment Report was recently added to reflect the full range of results. All of the scenarios used in this study are in line with the results of the Fourth Assessment Report.
The first 8 SLR scenarios examined (6 cm to 48.5 cm) are based on the range of increases in global eustatic5 sea level rise by 2100 referred to as the range of averages of the Atmosphere-Ocean General Circulation Models (AOGCMs) for all 35 SRES (Special Report on Emission Scenarios), reported in figure 11.126 of the IPCC's Third Assessment Report (2001).7 The range of averages is narrower than the range of results for the complete set of models and scenarios, 9 to 88 cm from 1990 to 2100, which includes uncertainties in land-ice changes, permafrost changes and sediment deposition. The 8 SLR scenarios represent points along the high and low lines that bracket the range of averages, spaced in 25 year increments. The 9th SLR scenario considered in the study, 59 cm, corresponds to the high end of the six illustrative scenarios considered in the Fourth Assessment.8 See Appendix 3 for a list of maps and statistics created as part of this report.
While methods for estimating changes have significantly improved, the overall picture of the predicted changes relevant to this study remains relatively unchanged. The results of the two IPCC reports are in fact not all that different, if differences in the way the analyses were conducted are considered. The IPCC notes that if two differences in the analysis are taken into account, the Third Assessment Report model means would be within 10% of the central estimates of the Fourth Assessment Report results. These two differences are: 1) while the Third Assessment Report gives projections for 2100, the Fourth Assessment Report gives projections for 2090-2099, and 2) the Third Assessment Report analysis includes some small constant additional contributions, which are not included in the Fourth Assessment Report analysis. Furthermore, the IPCC notes that the ranges in the Third and Fourth Assessment Reports would have been similar if uncertainties had been treated the same.9
It is also noteworthy to consider that this study, like the Third Assessment Report and the Fourth Assessment Report, does not include the effects of full melting of either the Greenland or West Antarctic Ice Shelf. Combined or individually, melting of these ice features would add significant additional water to the global oceans and raise the level beyond the scenarios considered in this study.
For each scenario two areas of concern were established. These are:
- regularly inundated, for areas that would be permanently under water under the given SLR scenarios
- at-risk, for areas that could be temporarily flooded due to storm surge under the given SLR scenarios
The regularly inundated areas are described as all the areas falling between NOAA's mean higher high water (MHHW)10, the study definition of sea level, in 2000 and the projected sea level under each SLR scenario (MHHW in 2000 plus each of nine sea level rise scenario increments in Phase I and 5 in Phase II up to 59 cm).
The at-risk areas are the areas falling between the adjusted MHHW, and NOAA's highest observed water level (HOWL) plus the of sea level rise projected for the particular scenario (HOWL in 2000 plus each sea level rise increment up to 59 cm). This is the study definition of storm surge. Note that any potential changes in storm intensity and resultant surge due to climate change are not considered by this study.
Figure 2-1 Provides a description of how regularly inundated and at-risk areas are defined for each SLR scenario. The projected sea level is based on the range of averages (the dark shaded areas) of the IPCC's Third Assessment Report, and the 59 cm level is from the Fourth Assessment Report.
Figure 2-1: Global average sea level rise 1990 to 2100 for the SRES scenarios. ICF used the upper and lower limits of the dark shaded area in this study as the basis for the changes in sea level for eight of the nine sea level rise scenarios. These figures are based on the range of averages of the Atmosphere-Ocean General Circulation Models (AOGCMs) for all 35 SRES Scenarios as reported in figure 11.12 from the IPCC's third assessment report (2001)11.
Given that sea level is not a flat and easily defined surface, a surface model that suits the study needs was required. NOAA's National Ocean Service (NOS)12 maintains numerous tidal stations along the coast of the United States that are used to measure the daily variances of sea level. These tidal station data are maintained as a matter of public record13 mainly as a service to ensure commercial and private maritime safety. While it is important for sea going vessels to understand how low the low tides may be, so they do not run aground, they also need to know how high the high tides (Mean Higher High Water) are expected to be so that they do not collide with the underside of structures such as bridges. This latter measurement is useful to this study to determine areas that are regularly inundated and is therefore the basis for our current (or base year 2000) sea level model. This area defines the highest areas that are wet on a regular basis and would therefore be of concern to those who plan and maintain transportation infrastructure.
Figure 2-2: An example of the tidal station data collected from the NOS showing the location of the facility, and all of the National Tidal Datum Epoch (NTDE) data for the tidal epoch of 1983-2001 are shown above. The NOS defines a tidal epoch as "the specific 19-year period adopted by the National Ocean Service as the official time segment over which tide observations are taken and reduced to obtain mean values (e.g., mean lower low water, etc.) for tidal data. It is necessary for standardization because of periodic and long term trends in sea level. The present NTDE is 1983 through 2001 and is actively considered for revision every 20-25 years. Tidal data in certain regions with anomalous sea level changes (Alaska, Gulf of Mexico) are calculated on a Modified 5-Year Epoch."14
Figure 2-3: an exaggerated 3D view of the MHHW sea level surface in the Chesapeake Bay area.
There are 632 tidal stations from New York to the Atlantic coast of Florida. Of those, 410 include the data needed (MHHW and NAVD15) to produce a surface model of the sea.16 To use these measurements across the broad area of the Atlantic coast, a surface was needed to approximate the elevation of the ocean at MHHW. Given the sparse population of discreet data from the tidal stations, this interpolation does not account for all local variations in the real world environment. This sparseness also introduces some value uncertainty. However, for the prescribed broad usage of this study, it does provide enough information to estimate the shape of the surface of sea level. In order to model this, the actual ground elevation (MHHW less NAVD) of the MHHW from the tidal stations was entered into a Geographic Information System (GIS) and a Triangulated Irregular Network (TIN) surface was interpolated. In the table above from the Washington, Potomac River tidal station, MHHW is 0.965 meters above MLLW and MLLW is 0.425 meters below NAVD, the benchmark ground elevation. By subtracting the NAVD from the MHHW the actual ground elevation of the MHHW can be found, in this case 0.965 (MHHW) - 0.425 (NAVD) = 0.54 meters. This process was performed on each tidal station and the TIN was interpolated from these points. The TIN created by this process was used to represent base year (2000) sea level. An example of the surfaces created by this process is found in Figure 2-3.
Working with the base year MHHW data from the tidal stations, additional TINs were created for each scenario by adding that scenario's estimated increase in sea level to the base year tidal station data. For example, in the table above from the Washington, Potomac River tidal station, the actual ground elevation of MHHW is 0.54 meters (see section 2.1 for further explanation of the process) and one of the scenarios for the increase in sea level for regular inundation is 48.5 cm (0.485 m). The addition of the estimated increase (48.5 cm) to the base MHHW level for this station (54 cm) equals 1.024 meters. This process was repeated for each tidal station and sea level rise scenario and a new surface model TIN created.
The Highest Observed Water Level (HOWL) data was extracted from the same tidal station data source (NOAA's National Ocean Service) used to create the current sea level models. The HOWL represents the highest recorded water level at that station and the date on which that observation was made. Therefore the HOWL data is completely dependent upon the length of time that the tidal station has been in existence. The oldest HOWL was recorded in 1898.
This data was used to model the base year (2000) surface representing areas at-risk of periodic inundation (storm surge). Of the 632 Atlantic coast tidal stations with full tidal data, 208 maintain data on HOWL, resulting in some value uncertainty in the base year surface.
The same process was used to create the HOWL surface as was used in creating future sea level surface models. For example, in the table above from the Washington, Potomac River tidal station, the actual ground elevation of HOWL is 2.943 meters (see section 2.1 for further explanation of process) and one of the scenarios for the increase in sea level is 48.5 cm (0.485 m). The addition of the estimated increase to the base year provides a sum of 3.428 meters. This process was repeated for each tidal station and sea level rise scenario and a new surface model TIN created for a total of 9 HOWL surface models.
The areas of concern are the areas that will be regularly inundated, the areas falling between the current MHHW and the projected sea level under each sea level rise scenario; and that are at-risk of periodic inundation, areas that fall between projected sea level and the projected HOWL under each SLR scenario.
These areas were produced by using a 3D geographic information system tool that compared the surfaces created in the previous steps to Digital Elevation Models (DEMs) produced by US Geological Survey (USGS) for the National Elevation Dataset (NED). These have a horizontal grid size of 30 meters. Please see section 2-5 for more about the DEM data used in this study.
These DEMs were then resampled to a 5 meter resolution using a bilinear interpolation to prevent "terracing" that occurs at such coarse scales as the 30 meter resolution.17 This function smoothes out the DEM and provides interpolated elevation data between the cells.
Figure 2-4: Areas of Concern
The surface models for all scenarios were then compared to the DEMs to determine where the surface models were above the elevation of the DEMs. For each sea level rise scenario, this comparison identified areas that would be regularly inundated or at-risk of periodic inundation due to storm surge. The results are created as polygon features.
The Digital Elevation Models (DEMs) used in this study are a product of the USGS and are know as the National Elevation Dataset (NED). The USGS DEMs were, at the initiation of this study, "the highest-resolution, best quality elevation data available across the United States."18 Additionally, the NED was at the time of this study the only consistent elevation dataset publicly available for the entire study area. The DEMs were downloaded from the US Department of Agriculture's (USDA) Geospatial Data Gateway.19 North Carolina, Virginia, Maryland, Washington DC, Delaware, Pennsylvania, New Jersey, and New York DEMs are in UTM (Universal Transverse Mercator) zone 18, while Georgia, South Carolina, and the Atlantic Coast of Florida were all in UTM zone 17. The NED has full coverage of the United States at 30 meter resolution, and partial coverage of the United States at 10 and 3 meter resolution.20 The 30 meter resolution was used in this study because it was the only dataset that covered the entire study area. A need for consistency in methodology overrode the potential for higher detail in certain areas.
Although this was the most appropriate data set for this study, there were some issues identified. These were primarily either due to the process used to create the DEM, or the quality of the "best quality elevation data" source data that went into making the DEM. Bordering DEMs can have vastly different quality, depending on how or when the DEM or source data was updated. See Figure 2-5 for an example of the edge variations. USGS DEMs have been produced using several very different methods, the accuracy of which has progressively improved over time. Outdated techniques for creating large scale DEMs, which are no longer used, include digitizing topographic maps and then processing them into a grid format, creating orthophotos and manually profiling them (which were notorious for creating artificial ridges and valleys throughout the DEM, making it look as if it has stripes), or using stereo plotters to create contours from stereo photographs. Currently, DEMs are created from interpolating digital line graph data reprocessed into a raster format. In some cases the edges of two adjacent DEMs are inconsistent. Often this has to do with temporal variations in the data collection. In one example in southern Florida, an artificial cliff runs the entire length of the two DEMs. This was caused by a significant difference in source data ages and collection and processing methodology. Another issue encountered involved DEMs that appear "speckled". This may be a result of down sampling a DEM. This would be done in order to make the higher resolution DEM match the pixel size and resolution of the larger dataset.21This was the most prominent issue in the Virginia portion of our study area - an example of this can be found in Figure 2-5: Examples of issues identified with the USGS DEMs. Both images illustrate how differences in source data used to create a single DEM can produce different quality results. The top image shows the differences in quality of the source data between adjacent DEMs in Virginia, while the bottom image shows the differences in quality between adjacent DEMs in Florida. There is a distinct difference in quality among the four tiles on the top (VA) image. The upper right tile is the most detailed and the bottom two are of lower quality and show signs of significant ‘speckling'. On the bottom image (FL) there is a visual difference between the quality of the left and the right sides.
While these issues are important to understand, they actually affect only a small portion of the overall data. In fact, these issues were a factor in only 36 out of the 1098 total DEMs in the study area - or about 3% of the DEMs. And since the areas of inundation do not always cover the entire DEM area, the area affected by these issues is actually less than 3%.
Figure 2-5: Examples of issues identified with the USGS DEMs. Both images illustrate how differences in source data used to create a single DEM can produce different quality results. The top image shows the differences in quality of the source data between adjacent DEMs in Virginia, while the bottom image shows the differences in quality between adjacent DEMs in Florida. There is a distinct difference in quality among the four tiles on the top (VA) image. The upper right tile is the most detailed and the bottom two are of lower quality and show signs of significant 'speckling'. On the bottom image (FL) there is a visual difference between the quality of the left and the right sides.
Once the areas of concern polygons were created, they were overlaid upon the transportation network data to identify potentially affected transportation infrastructure.
The data used in this analysis include:
1:100K scale Road data from the National Highway Planning Network (NHPN)22 including. The study included:
- Interstate Highways
- Non-Interstate Principal Arterials (hereafter referred to as Principal Arterials)
- Minor Arterials (Including all Rural Minor Arterials, but not the Urban Minor Arterials)
- National Highway System (NHS)23
1:100K scale Rail data from the Federal Railroad Administration
1:100K scale Airport boundaries and runway areas from TeleAtlas24
1:100K scale Port boundaries digitized from DOQQs25 for the land boundaries and the MHW line for the water boundaries. Ports included are part of the following port authorities:
- The Port Authority of New York & New Jersey
- The South Jersey Port Corporation, NJ
- The Philadelphia Regional Port Authority, PA
- The Delaware River Port Authority
- The Port of Baltimore, MD
- The Virginia Port Authority
- The North Carolina State Ports Authority
- The South Carolina State Ports Authority
- The Georgia Ports Authority
- The Florida Ports Council (Atlantic coast ports only)
The roads and rails were overlaid with the areas of concern to identify the linear distance in kilometers affected within each scenario. The airports, runways and port areas were also intersected with the areas of concern to identify the area in acres affected under each scenario. A portion of the bus public transit system impacts could be reflected in the results for roads, and the commuter rail system results are reflected in the results for rail. While heavy rail and light rail public transportation systems such as subways and metros were not assessed, systems in areas that are regularly inundated or at risk to storm surge could also be affected.
Since the elevations from the DEMs represent the actual ground elevation, this study did not account for situations where infrastructure is artificially elevated. However, the results in this study are still relevant in those areas. For example, a highway with a high bed is indicated as inundated in this study. While the road itself may not be underwater, the bed, which is inundated, was not likely designed to be permanently underwater and thus must still be considered for mitigation.
From the analysis in the previous steps, statistics at the county and state level were created for each scenario. For each scenario the statistics include:
- Kilometers of Interstate Highways potentially impacted
- Kilometers of Non-Interstate Principal Arterial roads potentially impacted
- Kilometers of Minor Arterial roads potentially impacted
- Kilometers of National Highway System facilities potentially impacted
- Kilometers of Railroads potentially impacted
- Total acres of Land potentially impacted
- Acres of Airport Property potentially impacted
- Acres of Airport Runways potentially impacted
- Acres of Port Property potentially impacted
The statistics tables include both regularly inundated and at-risk land areas. These are mutually exclusive, meaning the areas at-risk do not also include regularly inundated areas. The sum of these two fields equals the total land area potentially impacted by the effects of SLR and storm surge under the 59 cm SLR scenario. For example, in the table below, the total area for the 59 cm scenario is the sum of the regularly inundated or permanently flooded area, 236,581 acres, and the area at-risk to temporary flooding due to storm surge, 237,971 acres, for the total 474,552 acres impacted by either regular inundation or potentially storm surge.
Figure 2-6: An example of the output statistics for the state of Maryland showing the 59 cm scenario.
To visualize the data created in the previous steps, maps were created. For each state an overview map for each sea level rise scenario was created. Similarly, for each county that was affected a map for each scenario was created. The maps contain both regular inundation and at-risk areas for each scenario for a total of nine maps per county for Maryland, Virginia, and North Carolina. For New York, New Jersey, Pennsylvania, Delaware, South Carolina, Georgia, and Florida a total of five maps per county were created. Note that since Washington D.C. is not a state, it's "State" and "County" maps are one and the same. In Figure 2-7 below, the map depicts the city of Virginia Beach, VA and is representative of the other county level maps created under this study. For Florida, the eastern half of the State only was included.
Figure 2-7: a representative output map from this study showing regular inundation and at-risk areas at the 59 cm scenario.
5 Eustatic sea level rise refers to a uniform change in sea level created by any volumetric increase in the oceans worldwide, primarily due to thermal expansion (caused by higher temperatures) and ice melt.
6 IPCC3, WG1, c.11, page 671. http://www.grida.no/climate/ipcc_tar/wg1/pdf/TAR-11.PDF
7 IPCC3, 2001, WG1, c.11, pp. 671-72. http://www.grida.no/climate/ipcc_tar/wg1/pdf/TAR-11.PDF
8 IPCC4, 2007, WG1, summary for policy makers, p. 13.
9 IPCC4, WG1, c.10, pp. 820-822. http://ipcc-wg1.ucar.edu/wg1/Report/AR4WG1_Pub_Ch10.pdf
10 NOAA's mean higher high water (MHHW) level approximates the average shoreline at the daily highest tide computed over an epoch (19 year period). See Figure 3-2 for more on this subject.
11 IPCC3, 2001, WG1, c.11, pp. 671. http://www.grida.no/climate/ipcc_tar/wg1/pdf/TAR-11.PDF
14 See http://tidesandcurrents.noaa.gov/datum_options.html for definitions
15 The North American Vertical Datum of 1988 (NAVD 88) is the vertical control datum established for vertical control surveying in the United States of America. NAVD is a benchmark for ground elevation.
16 This model estimates all elevations by using the North American Vertical Datum of 1988 (NAVD)
17 The term "terracing" refers to the effect produced when a continuous surface (land elevation in this case) is represented by discrete data at large intervals. In this case, the DEMs used take an elevation reading every 30 meters and assign that elevation to the entire grid cell, thus making unnatural cliffs and flat areas where cells converge.
20 30 meter resolution means that 1 elevation point represents an averaged elevation of an area of 30 square meters.
22 The NHPN is a product of the U.S. Department of Transportation's Federal Highway Administration.
23 There are other roads identified on the lower functional systems to include the remainder of the National Highway System (NHS). There may be other roads identified which are Non-NHS/Non Arterial, but these systems are not complete in the NHPN.
24 This data was extracted from ESRI's StreetMap Pro dataset which uses TeleAtlas North America data.
25 A digital orthophoto quarter quadrangle (DOQQ) is a computer-generated image of an aerial photograph in which image displacement caused by terrain relief and camera tilts has been removed. For more information see: http://www.usgsquads.com/prod_doqq.htm.