CleanTOPO2
Edited SRTM30 Plus World Elevation Data

Tom Patterson, US National Park Service

CleanTOPO2 is a touched up and generalized version of SRTM30 Plus, a public domain dataset that combines sea floor and land elevation data of the entire world. The intent of these changes is to create an elevation dataset more applicable to the graphical needs of cartographers, such as for making shaded relief and 3D panoramas. SRTM30 Plus and an earlier related dataset, ETOPO2, feature bathymetry data released by Smith and Sandwell in 1997. As remarkable as this dataset is, it nevertheless contains numerous artifacts that mar map presentations. In CleanTOPO2, manual editing to the elevation data itself has removed many of the bathymetry artifacts (Figure 1). Until the day arrives that improved bathymetric data are released by the scientific community, CleanTOPO2 offers a stopgap solution for those creating maps and related graphics.


Figure 1. (left) Shaded relief rendered from ETOPO2 reveals linear artifacts on the Hatteras Abyssal Plain southwest of Bermuda—the triangular shape is only a coincidence. (right) Artifacts are less visible in the shaded relief rendered from CleanTOPO2.


Bathymetry edits

Terrestrial edits
Data
Comments and tips




Bathymetry edits

Shaded relief generated from SRTM30 Plus and ETOPO2 data invariably impresses readers with its sheer amount of detail and the unusual topographic forms found on the ocean floor. Upon closer inspection of the shaded relief, however, readers see elements that are decidedly unnatural—sharp incisions and rows of small bumps that cut straight through seamounts, abyssal plains, and trenches. These lines are in fact artifacts imbedded in the data that trace the routes taken by oceanographic survey ships (Figure 2). Because the artifacts create graphical noise and can mislead readers about the true character of the sea floor, I removed them wherever possible.

Figure 2. The striations on the Atlantic sea floor radiating out from Cape Town, South Africa, correspond to the routes of survey ships. They are not natural.


Procedure

I used Adobe Photoshop CS2 to smooth out the offending artifacts by importing the bathymetry elevation data as a 16-bit grayscale image—think of it as a type of raster Digital Elevation Model (DEM). Manual edits to the grayscale DEM were not possible, however, because it was dark and all but impossible to interpret. Instead, I edited a shaded relief rendered from the DEM that served as proxy image—seeing the topography clearly was essential to success. Drawing on a layer mask allowed me to transfer the edits to the DEM later.

In the first step, using Natural Scene Designer Pro 4.0, I rendered a detailed shaded relief of SRTM30 Plus from a DEM downsampled to 2-arc-minute resolution (10,800 x 5,400-pixels wide). Both the DEM and the shaded relief were the same size in pixels. I then opened the shaded relief in Photoshop, converted it to grayscale mode, and duplicated the layer (Figure 3, top). On the bottom layer applying Gaussian blur (a value 6 was used) smoothed the topography, including the artifacts. Next, I added a Layer Mask to the top shaded relief that did not receive Gaussian blur.

Having set up the file, I then began editing on the layer mask. Using a pressure sensitive Wacom stylus and tablet, I painted with a soft brush set at 50 percent opacity until the artifacts disappeared or were diminished. If an artifact was bold, I pressed harder with the stylus to apply denser tone to the mask until it melted away from view; removing faint artifacts required only light pressure.

In this manner I systematically canvassed our planet painting out the artifacts while at the same time taking pains not to alter actual bathymetry data. By replacing the original data with a softened version of itself, the potential for doing real harm was slight. When in doubt about whether a feature was natural or unnatural, I tended to leave it alone. In some cases I chose an intermediate solution and suppressed the feature to make it less visible. Manual edits took six hours to complete.

To complete the procedure I set up a two-layer Photoshop file similar to shaded relief file described above, but using the 16-bit grayscale DEM instead. As before, the bottom layer received Gaussian blur and the top layer a layer mask. I then copied and pasted the layer mask contents from the shaded relief file to the DEM file (Figure 3, bottom). Finally, I flattened the DEM file and saved it as 16-bit grayscale TIF image.

Figure 3. (Top) The shaded relief file in Photoshop used to make manual edits on a layer mask. (Bottom) The DEM file to which the edits were transferred.


Deciding what to edit

A large number of bathymetry artifacts exist in SRTM30 Plus data and my editing did not remove them all. I concentrated on deep ocean basins where the artifacts were most noticeable and the bottom flat, making the edits easier to accomplish. By contrast, the fractured topography of mid ocean ridges received minimal editing. Here the artifacts largely become lost amidst the complexity and editing is prone to degrading the data.

Some oceans required more editing than others (Figure 4). Problem areas included the North Pacific south of the Aleutians, a region marred by parallel north-south artifacts; the Atlantic off the US east coast; and, radiating linear artifacts found near Cape Town, Honolulu, and Tokyo. At 72 degrees North and South latitude the Smith and Sandwell bathymetry merges abruptly with polar datasets derived from other sources. I erased the seam lines.

The Indian Ocean and central Pacific required little editing, perhaps because of the relative absence of data collected by survey ships in these areas. Not that these areas were completely clean. Much of the Smith and Sandwell bathymetry derives from satellite altimetry, a collection process that may give the ocean floor a slightly dimpled texture. This is much in evidence in deeper portions of the Gulf of Mexico in the older ETOPO2 data. However, the same area is smooth in SRTM30 Plus, which uses upgraded data from the NGDC Coastal Relief Model. Other than the southern Gulf of Mexico adjacent to smoothed data mentioned above, the dimpled texture remained unedited.

Besides the linear artifacts, I also removed or diminished other suspicious data. This included isolated pits and bumps on the continental shelves that appeared unusually deep or high, suggesting erroneous depth values.

Figure 4. The black marks show the extent of bathymetry edits.




Terrestrial edits

The terrestrial data in SRTM30 Plus, and by extension CleanTOPO2, contain significant improvements compared to the coarse data found in parts of ETOPO2 (Figure 5). As the name suggests, SRTM30 Plus uses NASA's SRTM 30 elevation dataset for areas between 60 degrees north and 56 degrees south latitude, the poleward limits of this coverage. At higher latitudes venerable GTOPO30 is used.

Figure 5. Guiana highlands, South America. CleanTOPO2 uses downsampled SRTM30 data to more accurately represent terrestrial topography.


If SRTM30 Plus has a fault, it is too much detail, or, more specifically, unbalanced levels of detail in land and ocean areas. Compared to the bathymetry based on a 3.7-kilometer grid, the terrestrial elevations contain samples taken every 900 meters on the Earth’s surface. The result is a merged dataset that, well, looks like a merged dataset. This dataset creates legibility problems in cartographic presentations. Shaded relief rendered from SRTM30 Plus depicts land areas with an excessive amount of detail that clashes with the generalized bathymetry.

Solving this problem involved discarding terrestrial data—less is often more when making maps. Downsampling SRTM30 Plus to 3.7-kilometer resolution was the first step. This step was not quite enough, however.  In many areas the bathymetry data appear coarser than 3.7-kilometer resolution, the native resolution of 2-arc-minute data. By downsampling SRTM30 Plus terrestrial data to 4.9-kilometer resolution (the grid width decreased from 10,800 to 8,100 height samples) and then upsampling the data back to the original 10,800 width, I finally created a merged terrestrial/bathymetric dataset that looked visually balanced when rendered as a shaded relief (Figure 6).

Figure 6. (left) A shaded relief created with 3.7-kilometer resolution terrestrial data. (right) A relief created from with downsampled (4.9-kilometer resolution) terrestrial data. Bathymety is unchanged in the illustrations.




Data — hosted by Florida State University, FREAC

CleanTOPO2 comes in three versions tailored to meet different design and production needs. All versions are 16-bit grayscale TIF files measuring 10,800 x 5,400-pixels and include a World (.tfw) file and "Read me" document with projection and datum information. The Geographic projection is used. All data available here are in the public domain.

CleanTOPO2 (68.8MB) - Full resolution terrestrial and bathymetric elevation data. Note: the terrestrial data in this version may be too detailed for some small-scale mapping.

CleanTOPO2 - 2D version (68.8MB) - Slightly downsampled terrestrial and full resolution bathymetric elevation data. This version is optimized for creating small-scale 2D shaded relief.

CleanTOPO2 - 3D version (68.5MB) - Downsampled terrestrial and bathymetric elevation data. This version also uses resolution bumping to control tall solitary peaks (including seamounts and guyots) from spiking upwards with the application of vertical exaggeration in a 3D scene (Figure 7).

Figure 7. CleanTOPO2 - 3D version rendered as an oblique view and draped with Natural Earth. Click the image for a larger version.


More data...

Natural Earth plus bathymetry (89.6MB) - This image was used as a drape in the 3D scene shown in Figure 7. It features Natural Earth landcover and bathymetric tints without shaded relief. It is as a zipped GeoTIF measuring 15,000 x 7,500-pixels. Click here to see a preview image.




Comments and tips

• Elevations in CleanTOPO2 range from 8,248 to -10,701 meters, a range less than the top of Mt. Everest (8,850 m) to the Challenger Deep (-10,920 m), the highest and lowest points on Earth. Apparently the 3.7-kilometer data grid does not align exactly with these extreme points.

• When CleanTOPO2 is opened in an application that reads DEM data, specifying an elevation range of 8,248 to -10,701 meters places the zero elevation at sea level, more or less. Along shallow coasts the land/water boundary in CleanTOPO2 varies considerably from surveyed coastlines.

• CleanTOPO2 includes terrestrial areas below sea level, such as Death Valley and the Dead Sea rift valley. Be aware in your software that including an ocean level set to zero will inundate these dry areas with water.

.• CleanTOPO2 bathymetry does not include the Caspian Sea, Aral Sea, and all freshwater lakes, including the Great Lakes of North America.

• Despite my attempt to cleanse CleanTOPO2 of artifacts, many remain. If you see something that should not be in the data, consider editing it out using the same Photoshop technique described above. You can apply editing done in this manner multiple times to the same data.

• If I removed a bathymetry feature from CleanTOPO2 that I shouldn’t have, you can put it back in using the original SRTM30 Plus data and applying the Photoshop editing technique in reverse. The key to doing this is making sure that both grayscale DEMs are the same size in pixels and contain exactly the same range of 16-bit grayscale values representing elevation.

• The easiest way to touch up shaded relief and panoramas rendered from CleanTOPO2 are with post-rendering edits in Photoshop. This is a one-time solution, however, that does not fix the underlying problem in the data.

• If shaded relief rendered from CleanTOPO2 looks too noisy and harsh, render it with less vertical scaling or print it lightly to disguise the flaws. The same idea applies to 3D scenes; there is a limit to the amount of vertical exaggeration that you can apply to a scene without it becoming visually displeasing. Use your judgment.

• Shaded relief generated from CleanTOPO2 is not ideal for extreme enlargement beyond its native resolution—you can push the 10,800 x 5,400 height samples only so far. As much as you might want to show bathymetric shading on a big map, the wiser choice is often leaving it off. However, if you must push the size limits of CleanTOPO2, I suggest emphasizing hypsometric tints (elevation colors) and deemphasizing the shaded relief, which is more revealing of imperfections in the data.

• When making small-format presentations (such as illustrations for the Web) from CleanTOPO2, the results usually read better when generalized data are used. Most raster mapping applications have downsampling functionality. To downsample CleanTOPO2 in Photoshop, go to Image/Image Size in the drop menu and specify smaller pixel dimensions. You will need to experiment. I usually start by discarding one half of the data and evaluate the new results after rendering it.


Return to shadedrelief.com