EPA's Bush-era sea level rise study
Part 1: Coastal Elevation Mapping
James G. Titus and Dave Cacela. 2008. Uncertainty Ranges Associated with EPA's Estimates of the Area of Land Close to Sea Level. In Background Documents Supporting Climate Change Science Program Synthesis and Assessment Product 4.1. Edited by James G. Titus and Elizabeth M. Strange. 2008. Washington, D.C.: U.S. Environmental Protection Agency. pp 68-144. EPA 430-R-07-004.
The abstract is on this page. You can also download a printer quality version of this uncertainty analysis of coastal elevations (pdf MB, 77 pp).
Previous assessments of the land vulnerable to sea level rise mostly failed to provide an uncertainty range. The few studies that did provide an uncertainty range id so without incorporating the uncertainty associated with topographic information on which the studies were based.
This study uses the elevation data from Titus and Wang (2008), as formatted by Jones and Wang (2008), to quantify an uncertainty range for the amount of land below various elevations for each county along the U.S. Mid-Atlantic Coast. That data set estimated the elevations of all land above spring high water. That is, it estimated elevations for dry land and nontidal wetlands, but did not estimate elevations for tidal wetlands. (Knowing that land is tidal wetland tells us that the land elevation is below spring high water and above mean low water, which provides a narrower uncertainty range about the elevation than if we know only that the land is below, for example, the 10-ft contour on a topographic map.) Because they obtained data for the entirety of the study area, there is no sampling error. The source of error stems entirely from the limitations in precision of the Section 1.1 results.
The overall approach is to make an assumption about the potential vertical error of the elevation data and the extent to which that error is random versus systematic. The magnitude of the error varies by data source: because we assume that error is a function of contour interval, which in turn varies by topographic quad, we calculate error separately for each topographic quad. Let us first explain our basis for focusing on vertical error of the elevation data, and then explain how low and high estimates for areas were calculated where the input data were USGS contour maps and other data with relatively coarse contour intervals (1 meter or worse), as well as our procedure for when the data had higher quality (2 feet or better).