SRTM Data Processing Methodology

Introduction

The first release of Shuttle Radar Topography Mission (SRTM) data was provided in 1-degree digital elevation model (DEM) tiles from the USGS ftp server (ftp://e0srp01u.ecs.nasa.gov/srtm/) in 2003. The data was released continent by continent, as and when the data was processed by NASA and the USGS. For the United States, data was made available at 1-arc second resolution (approximately 30m at the equator), but for the rest of the world the 1-arc second product is degraded to 3-arc seconds (approximately 90m at the equator). SRTM elevation data has now been released for the entire terrestrial surface, and a “Finished” product has now been released (ftp://e0srp01u.ecs.nasa.gov/srtm/version2/SRTM3/).

In this web site, the Consortium for Spatial Information (CGIAR-CSI) of the Consultative Group for International Agricultural Research (CGIAR) is offering post-processed 3-arc second DEM data for the globe. The original SRTM data has been subjected to a number of processing steps to provide seamless and complete elevational surfaces for the globe. In its original release, SRTM data contained regions of no-data, specifically over water bodies (lakes and rivers), and in areas where insufficient textural detail was available in the original radar images to produce three-dimensional elevational data. There are a total of 3,436,585 voids accounting for 796,217 km2, and in extreme cases, such as Nepal they constitute 9.6% of the country area with some 32,688 voids totalling an area of 13,740 km2.  No-data regions due to insufficient textural detail were especially found in mountainous regions (Himalayas and Andes, for example), or desertic regions (e.g. Sahara). The existence of no-data regions in a DEM cause significant problems in using SRTM DEMs, especially in the application of hydrological models which require continuous flow surfaces. For the CGIAR-CSI SRTM data product we apply a hole-filling algorithm to provide continuous elevational surfaces.

The data is projected in a Geographic (Lat/Long) projection, with the WGS84 horizontal datum and the EGM96 vertical datum.

Methodology

We follow the method described by Reuter et al. (2007). The first processing stage involves importing and merging the 1-degree tiles into continuous elevational surfaces in ArcGRID format. The second process fills small holes iteratively, and the cleaning of the surface to reduce pits and peaks. The third stage then interpolates through the holes using a range of methods. The method used is based on the size of the hole, and the landform that surrounds it. The processing is made using Arc/Info AML model. Specifically:

1. The original SRTM DEM (finished grade data downloaded from ftp://e0srp01u.ecs.nasa.gov/srtm/version2/SRTM3/ is used to produce contours or points (depending on the interpolation methodology to be used for the void). Processing was made on a void by void basis.
2. In cases when a higher resolution auxiliary DEM was available, a point coverage is produced of the elevation values at the centre of each cell of the auxiliary DEM within void areas. When no high resolution auxiliary DEM is available, the 30 second SRTM30 DEM is used as an auxiliary for large voids.
3. For areas with a high resolution auxiliary DEM: The contours and points surrounding the hole and inside the hole are interpolated to produce a hydrologically sound DEM using the TOPOGRID algorithm in Arc/Info. TOPOGRID is based upon the established algorithms of Hutchinson (1988; 1989), designed to use contour data (and stream and point data if available) to produce hydrologically sound DEMs. This process interpolates through the no-data holes, producing a smooth elevational surface where no data was originally found. Drainage enforcement is activated, and the tolerances set at 5 for “tolerance 1”, representing the density and accuracy of input topographic data, and a horizontal standard error of 1m and vertical standard error of 0m.
4. For areas without a high resolution auxiliary DEM: The most appropriate interpolation technique is selected based on void size and landform typology, and applied on the data immediately surrounding the hole, using SRTM30 derived points inside the hole should it be of a certain size or greater. The best interpolations methods can be generalised as: Kriging or Inverse Distance Weighting interpolation for small and medium size voids in relatively flat low-lying areas; Spline interpolation for small and medium sized voids in high altitude and dissected terrain; Triangular Irregular Network or Inverse Distance Weighting interpolation for large voids in very flat areas, and an advanced Spline Method (ANUDEM) for large voids in other terrains.
5. The interpolated DEM for the no-data regions is then merged with the original DEM to provide continuous elevational surfaces without no-data regions. This entire process is performed for tiles with large overlap with neighbouring tiles, thus ensuring seamless and smooth transitions in topography in large void areas.
6. The resultant seamless dataset is then clipped along coastlines using the Shorelines and Water Bodies Database (SWBD). This dataset is very detailed along shorelines, and contains all small islands. More information about this dataset is available in USGS (2006c). .

Auxiliary DEMs were available from the following sources:

• USA – NED 3-arc second digital elevation model for mainland USA, Alaska, Hawaii and Puerto Rico.  Available from http://seamless.usgs.gov/  
• Mexico – 90m DEM available from http://www.inegi.gob.mx/geo/default.asp  
• Canada – Canadian Digital Elevation Data Level 1derived from 1:50,000 and 1:250,000 topographic maps, available from http://www.geobase.ca/geobase/en/data/cded1.html
• New Zealand – 100m DEM made kindly available by Geographx (http://www.geographx.co.nz/downloads.html)
• Australia – GEODATA TOPO 100k contour data, interpolated to produce a 90m DEM available form http://www.ga.gov.au/products
• Mountainous areas in Central Asia, China, Europe, Caucasus, Northern Andes and Southern Andes based on data from Jonathan de Ferranti’s webpage:  http://www.viewfinderpanoramas.org/dem3.html
• Costa Rica – 50m DEM derived from digitized topographic maps made available to CIAT by Antonio Trabucco.
• Ecuador – a 90m DEM derived from digitized topographic maps available from http://rslultra.star.ait.ac.th/~souris/ecuador.htm#DEM30
• Global – Where other auxiliary DEMs were not available, the SRTM30 1km product was used as an auxiliary DEM (USGS, 2006d).

This method produces a smooth elevational surface of no-data regions. Whilst micro-scale topographic variation is not captured using this method, most macro-scale features are captured in small-intermediate sized holes. Jarvis et al. (2004) (available here) make a detailed analysis of the accuracy of the interpolated elevational data in a region in Colombia with 43% of the region containing no-data in the original SRTM release. They find an average vertical error of just 5m in interpolated regions when compared with a DEM derived from cartographic maps, though the maximum error stretches to 257m in a region with approximately 1500m elevation. When hydrological models are applied to the interpolated DEM and the cartographic DEM, little difference is found in hydrological response in terms of overland flow and discharge.

The method presented here for filling in the no-data holes in the original SRTM release is by no means the only method available. For a complete review of methods for hole-filling in SRTM data, readers are referred to an article produced by the Alpine Mapping Guild, Gamache (2004). Martin Gamache has since produced some detailed analysis of the data offered here by the CSI, concluding that the hole-filling algorithm is quite successful in representing broad scale patterns in topography in data holes.

A detailed evaluation of the hole-filling methodology is available at:
http://www.terrainmap.com/downloads/Gamache_final_web.pdf

Changes from Version 1 to Version 2

• Version 2 includes DEM data for Australasia and small islands in the Atlantic, Indian and Pacific Oceans.
• Version 2 has the shorelines clipped.
• Version 2 have no “cliffs” on tile joins, brought about by insufficient overlap in interpolation in Version 1.
  

Change from Version 2 to Version 3

• Version 3 includes Finished grade SRTM data
• Version 3 uses the SWBD database to clip the coastlines
• Version 3 uses auxiliary DEMs to fill the voids
• Version 3 differs from Version 2 with a ½ grid pixel shift

Change from Version 3 to Version 4

• Version 4 uses a number of interpolations techniques, described by Reuter et al. (2007)
• Version 4 uses extra auxiliary DEMs to fill the voids and SRTM30 for large voids
• Version 4 differs from Version 3 with a ½ grid pixel shift which definitively solves this confusion.

Known issues and future improvements

We plan to continue improving the data as and when high resolution auxiliary datasets become available.  Updates are planned that will use high resolution ASTER DEMs for filling holes in particularly troublesome areas (Sahara, for example).

Data Use and Distribution

This data has been generated by not-for-profit institutions with the objective of supplying accessible and useful information to developing country organizations. We actively encourage use of these products for scientific purposes.
This is not however the case for commercial purposes. The entire dataset is available for commercial use at a modest cost, but permission must be sought. Commercial sectors interested in using this data should contact Dr. Andy Jarvis (a.jarvis@cgiar.org).

This dataset should be cited as follows:
Jarvis, A., H.I. Reuter, A. Nelson, E. Guevara, 2008, Hole-filled SRTM for the globe Version 4, available from the CGIAR-CSI SRTM 90m Database: http://srtm.csi.cgiar.org.

References

• Gamache, M. (2004). Free and Low Cost Datasets for International Mountain Cartography, http://www.icc.es/workshop/abstracts/ica_paper_web3.pdf.
  
• Hutchinson, M. (1988). Calculation of hydrologically sound digital elevation models. Third International Symposium on Spatial Data Handling, Columbus, Ohio, International Geographical Union.
  
• Hutchinson, M. (1989). "A new procedure for gridding elevation and stream line data with automatic removal of spurious pits." Journal of Hydrology 106: 211-232.
  
• Jarvis, A., J. Rubiano, A. Nelson, A. Farrow and M. Mulligan (2004). Practical use of SRTM data in the tropics: Comparisons with digital elevation models generated from cartographic data. Working Document no. 198. Cali, International Centre for Tropical Agriculture (CIAT): 32.
• Reuter H.I, A. Nelson, A. Jarvis, 2007, An evaluation of void filling interpolation methods for SRTM data, International Journal of Geographic Information Science, 21:9, 983-1008.
  
• USGS, 2006a, Shuttle Radar Topography Mission (SRTM) "Finished" 3-arc second SRTM Format Documentation, Available online at: http://edc.usgs.gov/products/elevation/srtmbil.html (accessed 01/08/2006).
  
• USGS, 2006b, Shuttle Radar Topography Mission DTED® Level 1 (3-arc second) documentation, Available online at: http://edc.usgs.gov/products/elevation/srtmdted.html (accessed 01/08/2006).
  
• USGS, 2006c, Shuttle Radar Topography Mission Water Body Dataset, Available online at: http://edc.usgs.gov/products/elevation/swbd.html (accessed 01/08/2006).
  
• USGS, 2006d, SRTM30 Documentation, Available online at: ftp://e0srp01u.ecs.nasa.gov/srtm/version2/SRTM30  (accessed 01/08/2006).
  
• Wessel, P., and W. H. F. Smith, A Global Self-consistent, Hierarchical, High-resolution Shoreline Database, J. Geophys. Res., 101, #B4, pp. 8741-8743, 1996.

Back to top