Item talk:Q228022

{

 "@context": "http://schema.org/",
 "@type": "WebPage",
 "additionalType": "Activity",
 "url": "https://www.usgs.gov/special-topics/significant-topographic-changes-in-the-united-states/science/accuracy-assessment",
 "headline": "Accuracy Assessment of Elevation Data",
 "datePublished": "October 14, 2019",
 "author": [
   {
     "@type": "Person",
     "name": "Dean Gesch, Ph.D.",
     "url": "https://www.usgs.gov/staff-profiles/dean-gesch",
     "identifier": {
       "@type": "PropertyValue",
       "propertyID": "orcid",
       "value": "0000-0002-8992-4933"
     }
   }
 ],
 "description": [
   {
     "@type": "TextObject",
     "text": "The differences were calculated by subtracting the GPS point elevation from the NED and SRTM elevations. This calculation results in statistics that are easy to interpret, with a positive error meaning that the NED or SRTM was too high at that point, and a negative error meaning that the NED or SRTM was too low."
   },
   {
     "@type": "TextObject",
     "text": "Comparison with the 2001 version of NLCD in this area confirms that the elevated areas that are very apparent in the SRTM shaded relief image are groups of trees, as they are included in the NLCD canopy density dataset."
   },
   {
     "@type": "TextObject",
     "text": "The tables below list the summary statistics of the errors for NED and SRTM, respectively, segmented by land cover class."
   },
   {
     "@type": "TextObject",
     "text": "An important aspect of the accuracy assessment was the calculation of vertical accuracy by land cover class, recognizing that SRTM is a first return system; the elevation measured is that of the first reflective surface that the radar signal encounters. Because of the relatively short wavelength (5.6 cm) of the C-band radar used by SRTM, the reflective surface in most vegetated areas is located within the canopy. In areas where buildings are prevalent, the measured SRTM elevations represent the combined effects of the rooftops and other structures within the resolution cell. This phenomenon of SRTM recording non-bare earth elevations where vertical features are present on the surface has been well documented, and it is being exploited by several researchers to produce maps of canopy height and biomass."
   },
   {
     "@type": "TextObject",
     "text": "The reported accuracies for SRTM fall well within the mission specifications, which agree with previous assessments. The effects of land cover on the uncertainty of SRTM elevation measurements and the overall better performance by NED in terms of vertical accuracy have also been documented in previously published results."
   },
   {
     "@type": "TextObject",
     "text": "Thus, the derived SRTM elevation measurement is for a surface located significantly above the elevation of the ground surface as represented in the reference geodetic points. This effect is supported by the measurement of the overall positive elevation bias for evergreen forest that approaches 2 meters. In other words, the SRTM elevations for evergreen forests are, on average, nearly 2 meters above the ground surface elevations. Interestingly, the bias for deciduous forests is much lower at less than one-half meter. A partial explanation for this observation may be that SRTM data were collected in February, which represents leaf-off conditions for deciduous forests in the conterminous United States, and more of the radar signals probably penetrated the relatively open canopy and reflected from the ground surface."
   },
   {
     "@type": "TextObject",
     "text": "The root mean square error (RMSE) was calculated, which is a commonly used method to express vertical accuracy of elevation datasets. The overall absolute vertical accuracy calculated for the NED is 2.44 meters (RMSE), whereas the assessment of the SRTM showed an accuracy of 3.53 meters (RMSE). The assessment included examination of the accuracy as a function of specific terrain conditions. Land surface characteristics, including slope, aspect, and local relief, were derived directly from the NED and were associated with each control point. Local relief was calculated within a 1,500 x 1,500-meter area (about 1 mi2) centered on the point location."
   },
   {
     "@type": "TextObject",
     "text": "The thresholding approach (described below) used to distinguish significant differences from the raw difference grids requires a measure of the absolute vertical accuracy of each input elevation dataset. The NED and SRTM data were compared to an independent reference geodetic control point dataset from NGS. These points have centimeter-level accuracy in their horizontal and vertical coordinates, as they are produced by high precision GPS observations on established survey bench marks. Because of their broad distribution across the conterminous United States and their very high accuracy, the points served as an excellent reference dataset for assessing the absolute vertical accuracy of the NED and SRTM."
   },
   {
     "@type": "TextObject",
     "text": "The first return nature of SRTM data is a critical characteristic to consider when generating SRTM NED differences grids. In vegetated and built-up areas, many of the vertical differences are not indicative of topographic surface changes but reflect the presence of radar scatterers above the ground surface. This leads to uses of the difference grids beyond just detecting geomorphic changes. Indeed, that has already been the case with the difference grids generated from this project, which have been distributed to researchers for use in biomass and carbon modeling, production of NLCD 2001 data over Kentucky, and riparian vegetation mapping in Texas for air and water quality monitoring."
   },
   {
     "@type": "TextObject",
     "text": "The next two figures show the error for NED and SRTM, respectively, at each GPS control point location plotted against elevation, slope, aspect, and local relief. The NED and SRTM errors appear to be truly random; there is no discernible correlation or relationship with any of the terrain parameters. This randomness is evident in the distribution of data points in the scatterplots; in each case, the values are uniformly distributed around the zero error axis."
   },
   {
     "@type": "TextObject",
     "text": "The graph below compares the accuracies for NED and SRTM for each NLCD class. For all classes except one (quarries/strip mines/gravel pits), the uncertainty, as measured by the RMSE, of the SRTM exceeds that of the NED. This is not unexpected given the nature of the SRTM system that measures the height of the first reflective surface. For the quarries/strip mines/gravel pits class, the areas would generally be devoid of vegetation and buildings, so the SRTM measurement more likely represents ground level and thus is closer to NED elevation."
   },
   {
     "@type": "TextObject",
     "text": "The graphs below compare the accuracies for six land cover classes for which the SRTM data are more likely to be affected by the reflective surface above ground level, the urban/built-up and forest classes. The largest uncertainty in SRTM data occurs in the evergreen forest class (5.16 m RMSE), which is most likely due to the SRTM radar signal penetrating very little into the dense conifer canopy before being reflected."
   },
   {
     "@type": "TextObject",
     "text": "The reference control points were in the same coordinate system as the input elevation data (decimal degrees of latitude/longitude in NAD83 for horizontal; decimal meters in NAVD88 for vertical), which facilitated comparison with the NED and SRTM data. The control point dataset was intersected with the NED and SRTM data, as well as with the NLCD to label each point with the land cover at each location. The control points were located in 849 of the 934 1x1-degree tiles used as processing units. The elevations from NED and SRTM were derived for each control point location through bilinear interpolation. Differences were calculated between the reference point elevation and the corresponding NED and SRTM elevations, and summary statistics were accumulated."
   },
   {
     "@type": "TextObject",
     "text": "The illustration below shows the effects of the presence of a tree canopy on SRTM data and the derived SRTM NED difference grid. In this illustration for an area in southern Michigan, the large vertical differences are due solely to the presence of trees."
   }
 ],
 "funder": {
   "@type": "Organization",
   "name": "Significant Topographic Changes in the United States",
   "url": "https://www.usgs.gov/special-topics/significant-topographic-changes-in-the-united-states"
 },
 "about": [
   {
     "@type": "Thing",
     "name": "Geology"
   },
   {
     "@type": "Thing",
     "name": "Environmental Health"
   },
   {
     "@type": "Thing",
     "name": "Accuracy Assessment of Elevation Data"
   },
   {
     "@type": "Thing",
     "name": "earth resources observation and science center"
   },
   {
     "@type": "Thing",
     "name": "Methods"
   },
   {
     "@type": "Thing",
     "name": "Topographic Change"
   },
   {
     "@type": "Thing",
     "name": "Integrated Land Change Monitoring"
   },
   {
     "@type": "Thing",
     "name": "Energy"
   },
   {
     "@type": "Thing",
     "name": "Methods and Analysis"
   },
   {
     "@type": "Thing",
     "name": "Climate"
   },
   {
     "@type": "Thing",
     "name": "Information Systems"
   },
   {
     "@type": "Thing",
     "name": "topographic"
   },
   {
     "@type": "Thing",
     "name": "Water"
   },
   {
     "@type": "Thing",
     "name": "Science Technology"
   },
   {
     "@type": "Thing",
     "name": "Topochange"
   },
   {
     "@type": "Thing",
     "name": "Coastal Changes and Impacts"
   }
 ]

}