Mapping Sea-Level Rise
Data
Accurate mapping of potential inundation due to sea level rise depends on high resolution terrain data. There are two main types of elevation data that are commonly available: USGS digital elevation models (DEMs) and LIDAR data. The USGS 7.5-minute DEM is the highest resolution elevation data set that covers the entire nation. However, with its root mean squared error (RMSE) of 7 m, which easily surpasses most estimates of projected sea-level rise over the next 100 years, it may not be valid to use this DEM to assess impacts of sea-level rise on coastal areas. LIDAR data are much more accurate, but their coverage is limited to a thin strip of area of 300 to 500 m, along selected stretches of coastline.
Although the USGS 7.5-minute DEM has an RMSE of 7 m, this error is not evenly distributed. Because of the interpolation methods used in deriving the DEM from its 7.5- minute topographical maps, it is reasonable to assume that such error will be small in relatively flat terrain and large in rapidly changing terrain. Since most coastal areas are relatively flat, DEMs covering these areas likely have smaller error than average. Therefore, a comparison of DEM data with the LIDAR data is conducted in order to determine the level of accuracy of the DEMs in coastal areas and the associated error in the inundation estimates.
Five sites were chosen in CARA region. Each site covers a stretch of coastal area of 0.5-1 km wide and 3-5 km long. These sites were chosen to be representative of major coastal states in CARA, and also based on data availability. For each site (except Maryland), both 30-m and 10-m DEMs (DEM30 and DEM10) were used. For the Maryland site, only the 30-m DEM is available. Because one of the primary purposes of LIDAR data is to monitor the dynamics of coastal changes, many coastal states collect LIDAR data repeatedly over important stretches of the coast. For each site, all available LIDAR measurements were used, ranging from 1 (for New Jersey) to 4 (for Virginia and Maryland) time periods. The original LIDAR data, in the form of x,y,z points, were interpolated to a 5-m grid. In total, 23 data layers for 5 sites were collected. All layers are converted to Universal Transverse Mercator (UTM) projections. Horizontal datum is North American Datum 1983 (NAD83), and vertical datum is North America Vertical Datum 1988 (NAVD88).
All DEM layers were resampled to 5-m grids using the nearest neighbor method. For each site, we compared elevation values of the DEMs with those of the mean of all LIDAR layers at each grid point. The mean, RMS and maximum difference in elevation were calculated. The internal variations among the LIDAR layers were also calculated. Results for all sites were pooled together by calculating the area-weighted mean and RMS differences to summarize the average discrepancies between DEM and LIDAR over all sites (Table 1).
|
Mean Difference |
RMS Difference |
Maximum Difference |
DEM30 and DEM10 |
0.023 |
0.344 |
8.159 |
DEM30 and LIDAR mean |
0.192 |
1.243 |
27.528 |
DEM10 and LIDAR mean |
0.125 |
1.161 |
19.047 |
Among LIDAR layers |
0.017 |
1.121 |
62.557 |
Table 1: Differences between DEM and LIDAR data sets (m)
We also calculated areas that will be inundated under different sea-level rise scenarios, using all DEM and LIDAR layers. The sea-level rise scenarios used were 0.38, 0.66 and 1 m, which were based on low, medium and high sea-level rise projections made by Najjar et al. (2000). The inundated areas derived from different data sets are compared for each site, and the results were also pooled to determine the average differences in results derived from DEM and average LIDAR across all sites (Table 2).
|
Inundated area (km2) |
Difference with LIDAR(%) |
|||
Sea-level Rise |
DEM30 |
DEM10 |
LIDAR mean |
DEM30 |
DEM10 |
Below 0.38 m |
8.72 |
8.70 |
8.09 |
7.80 |
7.47 |
Below 0.66 m |
8.87 |
8.88 |
8.54 |
3.95 |
4.04 |
Below 1 m |
10.29 |
10.32 |
9.47 |
8.66 |
9.03 |
Table 2: Areas inundated by 0.38, 0.66 and 1 meter sea-level rise using DEM and LIDAR
In conclusion, the average difference between the DEMs and LIDAR is less than 0.2 m, and the RMS difference is 1.24 m for the five coastal sites we evaluated. The average discrepancies between the DEMs and LIDAR in calculating areas inundated by projected sea-level rise are between 4 to 9% under different scenarios. DEM10 is very consistent with DEM30, and so, in summary, we conclude that DEM30 is appropriate for making inundation estimates due to future sea-level rise.
Sea-Level Rise Scenarios
According to Mid-Atlantic Regional Assessment (Fisher et. al., 2001), sea-level is most likely to rise 0.6 meter (2 ft) by 2100 for the mid-Atlantic coast, ranging from 0.4 to 1m. The average normal tidal range for this region is about 2.3 m (7 ft), ranging between 1.3 (VA, MD) to 3.2 (MA) meters (4-10 ft). Spring tides are typically 20% greater than normal tides, adding another 0.5m in tidal range. The DEMs we used in this study is registered to North American Vertical Datum 88, i.e. the mean sea-level of 1988 for the region, and its vertical resolution is 1m. Therefore, we chose 1m as the range for future sea-level rise to reflect the probably higher end of the model projection, given the data resolution. We also decided to include 3m to include land that will be affected by tides with elevated sea-level, which can be as much as 1.5 m higher than future mean sea-level during spring high tides (i.e. 2.5m higher than present mean sea level).
Finally, this assessment of risk is made based on elevation only. These maps do not depict future shorelines, since there are a variety of other factors that will affect the shape of future shorelines such as coastal erosion, wetland accretion, the existance of man-made shoreline protection structures/projects, to name but a few. These maps represent a first-order assessment, based on which more detailed analysis can be conducted at a finer scale.
For more information, contact Shuang-Ye Wu or Ray Najjar.