Comparing the Quality – CGIAR DEM vs MERIT DEM
I would like to illustrate some of the differences in quality between two global DEM rasters – the CGIAR SRTM 90m DEM and the MERIT DEM. Both rasters have 3 arc-second horizontal resolution and provide near-global coverage. Both rasters use Shuttle Radar Topography Mission (SRTM) data as the primary data source.
You can download the MERIT DEM as tiles in either ESRI FLT or GeoTiff format, or as a single raster in MRR format, from here.
You can download the CGIAR SRTM 90m DEM as tiles in ESRI ASCII format from here. At present, I am unable to provide a download link to this raster in MRR format but I will update the article if I can make that available.
The two rasters share the same horizontal resolution – 3 arc-seconds or approximately 90 metres. CGIAR has a vertical resolution of 1 metre whereas MERIT has a vertical resolution of 1 centimetre (as an MRR and even smaller in tiled formats). However, the vertical resolution does not say anything about vertical data accuracy. In both cases, the vertical accuracy is considerably worse than 1 metre. My view is that the CGIAR vertical resolution of 1 metre is more appropriate than the MERIT vertical resolution of 1 centimetre. If the rasters are in MRR format, then using an appropriate vertical resolution serves to reduce the file size. The CGIAR raster is an 11.5 GB MRR file whereas the MERIT raster is a 47.1 GB MRR file.
In the images in this article, drag the slider to the right to reveal the CGIAR raster and to the left to reveal the MERIT raster. I have used the same rendering algorithm for both rasters – the same color table, the same color mapping, and the same shadow intensity factors. They present a perfect like-for-like comparison. The images were rendered in MapInfo Pro (https://www.precisely.com/product/precisely-mapinfo/mapinfo-pro) using rendering algorithms generated by ProRaster (https://robertsgeospatial.com.au/). Both rasters provide near-global coverage, however, this does not include Antarctica in either case. MERIT includes the northern latitudes north of 60 degrees, which is a considerable improvement and advantage. The CGIAR raster does not include Alaska, Norway, or most of Russia.
Global Coverage Comparison
The rasters take a different approach to coastal clipping. I cannot pick a winner here because the effect changes from one locality to the other. The images below show the mouth of the Amazon River in Brazil and the mouth of the Hawkesbury River in New South Wales, Australia.
Coastal Clipping Comparison – Amazon River
Coastal Clipping Comparison – Hawkesbury River
Both rasters claim to be a DEM rather than a DTM. This is because the SRTM radar instrument measures the top of the radar reflective surface and this might include trees, buildings, and other cultural noise. MERIT has been processed to reduce this cultural noise where possible and the effect is easily visible. The examples below are from Florida in the USA and the Murray River in Australia. MERIT is smoother, better reflecting the true nature of the terrain surface in both examples. In the Murray River example, farm boundaries, roads, and rivers are easily visible due to the trees that are often found along these boundaries. In the MERIT raster, this effect is reduced considerably.
Cultural Noise Comparison – Florida
Cultural Noise Comparison – Murray River
The example below is Mt Taranaki on the North Island of New Zealand. The volcano is a reserve and is surrounded by farmland, so there is a considerable circular height discontinuity running around the mountain (One Ring to Rule Them All). In the MERIT raster, this discontinuity is less pronounced.
Treeline Comparison – Mt Taranaki
Another source of noise in the SRTM are orbital artefacts and other processing artefacts. The image below shows orbital artefacts in Namibia, Africa. In the MERIT raster, these are not visible.
Orbital Artefacts Comparison – Namibia
The SRTM data suffers from dropouts where no valid data was returned by the orbital instrument. In both cases, efforts have been made to fill these regions with data from other sources. You can see from the example below, of the Western Sahara Desert, that MERIT contains considerably higher quality data fill.
Dropouts Comparison – Sahara Desert
Dropouts are particularly troublesome in mountainous regions and in steep valleys and gorges. One unfortunate result from this is to disrupt and compromise the hydrology information and to place “dams” into river systems where they do not exist. The first example below is the Three Gorges Dam in China where you can see MERIT has considerably improved detail in the steep sided gorges and improved the continuity of the river valley. The second example is of the Colo River in the Blue Mountains in Australia. The narrow and steep gorge systems in the Wollemi National Park are represented considerably better by the MERIT DEM.
Dropouts Hydrology Comparison – Three Gorges Dam China
Dropouts Hydrology Comparison – Colo River Australia
Finally, there are many areas where MERIT contains higher detail than CGIAR. The following example is the Grand Canyon in the USA where you can see the detail in the canyon walls is considerably improved. The second example is centred on the peak of Chomolungma (Mt Everest) in Tibet. In the CGIAR raster the mountain is smooth, whereas in the MERIT raster you can, I think, even pick out the Hillary Step.
Detailed Fill Comparison – Grand Canyon USA
Detailed Fill Comparison – Mt Everest
In summary, the MERIT DEM is a considerably improved global scale terrain DEM. It provides more coverage, more detail, and more accuracy. Even though the file is 4X bigger, MERIT is a clear winner in this comparison, and I recommend it!
You can download a PDF of this article here.