We are only as good as our source

The creative commons elevation data sets available from Geoscience Australia offer an improved product compared to the data sensed by their source provider, NASA.

The 1 second (30 meter resolution) SRTM derived datasets (which can be downloaded from the National Elevation Data Framework portal) are those used in the Blue Mountains Project being documented on this blog. The origin of the dataset is from a shuttle mission flown by NASA in 2000 to scan the surface terrain of the earth at 30m resolution.

However, the nature of scanning the Earth’s terrain at an angle of 45 degrees results in some problems, most notably where there are steep cliffs. The raw SRTM data have had a number of algorithmic techniques applied to provide an enhanced product for Australia and this is well documented in the excellent user guide produced by Geoscience Australia (Gallant, J.C., Dowling, T.I., Read, A.M., Wilson, N., Tickle, P., Inskeep, C. (2011) 1 second SRTM Derived Digital Elevation Models User Guide. Geoscience Australia) [11.6 MB PDF]

Whilst exploring certain key areas of the Blue Mountain data in the project visualization platform it can be seen that some of these corrective techniques have introduced some peculiarly high elevations over the area of the cliff face along the Grose Valley near Blackheath. This is also documented in Gallant et. al. (2011) p. 37 [11.6 MB PDF].

Data Errors in Grose Valley NSW. Data: (c) Commonwealth of Australia (Geoscience Australia) 2012; Image (c) John Whelan (2013)
Data Errors in Grose Valley NSW. Data: (c) Commonwealth of Australia (Geoscience Australia) 2012; Image (c) John Whelan (2013)

 

Data Errors in Grose Valley NSW. Data: (c) Commonwealth of Australia (Geoscience Australia) 2012; Image (c) John Whelan (2013)
Data Errors in Grose Valley NSW. Data: (c) Commonwealth of Australia (Geoscience Australia) 2012; Image (c) John Whelan (2013)

This level of error will clearly effect any surface dependent visualization and analysis. Areas of the valley floor have also been inserted from data derived from a lower resolution source, so, how to improve this area remains a question.

When sensing and measuring we are only as good as our measuring tools and when auto-rectifying and processing we are only as good as our algorithm design. When a trained human is in the loop, in the case of manual contour interpolation, a number of problem points are still present, but a high-level of intelligent adaption and cartographic insight is also there. This semantic understanding applied to automatic generalisation and visualization is a still a challenge in digital cartography.