5.A.iv. DEM for Italy

---

Primary Developers:		Servizio Geologico Nazionale (SGN) and National Geophysical Data Center (NGDC) (now NCEI)
Title:		30" DEM for Italy
Publication Date:		1994
Bibliographic Citations:		Carrozzo, M.T., D. Luzio, C. Margiotta, T. Quarta, F. Zuanni, A. Chirenti, and A.M. Tundo, 1985. Data base of mean height values for the whole Italian landmass and surrounding areas: determining and statistical analysis. Bollettino di Geodesia e Science Affini, No. 1, 1985, pp. 38-56.
	*	Servizio Geological Nazionale and Row, L.W., 1994. 30 Arc-second Digital Elevation Model for Italy, in Row and Hastings, 1994 (in GLOBE Task Team and others, 1999).
Post-processing:		None. This DEM was originally created for TerrainBase.
Source/Lineage Category:		10

---

* Primary reference citation for all data from this source

The DEM for Italy was derived at NGDC (now NCEI), from a 7.5x10-second DEM developed by the Servizio Geologico Nazionale (SGN) of Italy. This section summarizes documentation prepared by Row and others (1995, pp. 4-91 through 4-95) for TerrainBase (Row and Hastings, 1994), and is also described by Carrozzo and others (1985).

The 7.5x10-second DEM developed by SGN was developed for use in gravity terrain corrections. Organizations involved with either developing, analyzing, or financing the data set included AGIP (a major petroleum company in Italy), the National Research Council, National Group of Solid Earth Geophysicists (GNGTS), Ministry of Education (MPI), and the University of Lecce. Source materials included several topographic map sources cited in Row and others (1995), at scales of 1:25,000 to 1:50,000 for land, plus hydrographic maps at 1:100,000.

Mean height values were manually extracted from the topographic maps at regular grid intervals. The size of the grid was selected to maximize the amount of reliable detail obtainable from the source maps, while keeping the resolution from becoming so fine as to overly burden the developers. Grid spacings of 7.5x10 seconds were used for 1:25,000 scale maps, 15x20 seconds for 1:50,000 maps, and 30x40 seconds for the 1:100,000 scale maps. A map of sources is presented in Carrozzo and others (1985), and reproduced in Row and others (1995, p. 4-94).

Grid lines were printed on transparent acrylic sheets in 20-minute latitude sections, and overlain on the maps to guide data collection. For each grid cell, a mean height was manually estimated from the contours and point heights that appeared within and adjacent to each cell.

The 6,210,000 elevation values were derived from 3450 maps, 1,878,000 heights in coastal regions from 130 coastal maps, and 3,456,000 depth values from 120 maps. The coarser grids were then interpolated to 7.5x10-second gridding by an undocumented technique. SGN then tiled the data into 20x30-minute areas.

Data were analyzed for errors by SGN using the following procedures:

1. SGN checked that data were expressed in proper metric terms, and that they were intrinsically valid. SGN verified that the data fell within a predetermined range of values for the area under consideration.
2. SGN checked the slope at each cell by testing whether the difference between each cell and its neighbors did not exceed a threshold value as determined by the type of local terrain.
3. SGN spot-checked data by independently collecting data from various cells twice. SGN then verified that the two values did not differ beyond predetermined values, based on the local terrain.
4. SGN generated contour maps from the gridded data in each area, and overlaid these on the source maps for comparison.
5. SGN generated contour maps from several adjacent areas, and overlaid these on 1:100,000-scale source maps, to confirm proper edge matching between processed areas.

SGN and NGDC jointly developed a method for compiling these proprietary data into an unrestricted 30" grid. According to this method, NGDC computed the mean value of all 7.5x10-second grid values contained in each 30" output grid cell. Areas containing significant errors were removed from the model (these were filled by NIMA DTED Level 0 discrete values during the patching process making GLOBE Version 1.0).

Cell-centered registration, mean of source 7.5" x 11.25" grid cells in output 30" cell. Graphic describing georeferencing and sampling for the DEM for Italy.

Plate 16. Plate 17. Click on each image to view larger size.

Plates 16 and 17 are histograms for the DEM of Italy, on linear and semi-logarithmic scales, respectively. There is a trough in the histogram for elevations below 80m in the SGN DEM, a broad peak at about 70m to 200m (the elevations of the lakes in the Italian Alps), and modest spikes at about 400m and 440m. Otherwise, the histogram is relatively smooth. The semi-logarithmic plot shown is not quite linear between 100m and 2000m elevation. Then a modest bulge occurs at about 2500m, with a rapid diminution of elevation values above about 2800m. This DEM is not quite so smooth as the GSI DEM for Japan. Nevertheless, it is unusually free of spikes, considering its cartographic source.