Thoughts - Generating Reliable Contours from SRTM Data

Isla la Plata - Other Precision Map (1-20,000)

Here is a comment and question concerning my use of SRTM data...

I am working on a similar project, in some aspects, about generating contours, and was wondering when the contours do not line up with your coastline (having 25 m intersect the shoreline of your polygon) do you alter the polygon or the raster data before generating the contours? Anyways, I think you work is beautiful, keep it up.

Here is my response...

Thank you for your kind words. I love making maps of islands and to be able to add meaningful contours is a great benefit. However, my experience with the SRTM 30-meter and 90-meter data is as follows. After loading both a georectified image and an SRTM file into Global Mapper, I am able to quickly generate contours at an interval I deem appropriate for the scale I am working with. In nearly every instance the lower level contours do not conform to the actual imaged shorelines. I end up manually realigning the contours to best match the imaged shoreline. As a result my lower level contours are reasonable, but suspect.

When I am dealing with a very large void that cannot be handled by combining SRTM 30-meter with SRTM 90-meter data or be filled with existing void-filling software (I use SRTMFill), I end up guesstimating alignments of contours. Consequently, all of my maps containing contours are as good as the underlying SRTM data - sometimes excellent (upper elevations of Isla la Plata map) and less reliable (10-meter to 30-meter contours and northern contours due to a void).

I am not practiced enough to correct the underlying SRTM ASCII data, so I rely upon working with software (SRTMFill), imagery to contour reconciliation, or when cloud-obscured, an educated guess using secondary imagery.

Hopefully, this answers your questions. Perhaps you have techniques or know of SRTM void-filling software that might correct many of my issues concerning misaligned or voided elevation data. I do believe that SRTM 30-meter and Landsat ETM+ at 14.5-meter resolution can work reasonably well together. However, combining SRTM 30-meter with 1-meter resolution imagery most often creates many slight, but consistent contour misalignments.

SRTM elevation data, as stated above, gives us a reasonable idea of the topography of the underlying land. I would use it to go hiking, but I sure would not bid a construction job based on it.

Enjoy!

Pulau Sematan ID - Kidney Bean Shaped Island

Pulau Sematan ID - Offshore Island Near Pulau Sulawesi

Pulau Sematan ID - Image From Landsat ETM+ N-50-00_2000 (1:62,380)

Located off of the northern coast of Pulau Sulawesi, is Pulau Sematan. After much research I have learned it is a dive destination and home to a few hardy fishermen. It is covered with lush vegetation and based on it's proximity along the equator, it is most likely, hot, humid and wet.

Pulau Sematan ID - EVS Precision Map With 50-Meter Contour Intervals (1:62,380)

The map itself was fast to do. From start to finish it took about 3-hours. Nothing particularly challenging, just a nice EVS precision map of a small island.

Enjoy!

Musandam Peninsula OM - 50-Meters Contours Using Plans One Through Four

Musandam Peninsula - EVS Precision Map (1-175,000)

I posted the above map of the Musandam Peninsula a few weeks ago. The geography is fascinating. However, I knew more work had to be done to complete my map. Specifically, I wanted to create contours to overlay onto the area.

My first attempt, Plan One, was straightforward - download the SRTM HGT file, load it into Global Mapper, generate contours and I would be finished . . . Whoops! Not so. It turns out that the SRTM HGT file for this area is full of voids, those nasty little holes of nothing. With voids my generated contours were grossly inaccurate.

I've been at this mapping business long enough to know that if one source doesn't work, there is most probably another, like Plan Two - the CGIAR-CSI SRTMs. These are void filled versions of the HGTs. I downloaded the appropriate SRTM tile, loaded it, generated contours and . . . Whoops! Not quite right.

Still not to be deterred, I went to Plan Three - using a void-filling piece of software called Void Killer 2 I tacked the contour generating problem once more. I am aware of a number of freeware SRTM void fillers, but I wanted to try the trial version of VK. I downloaded the trial copy, unzipped it and loaded the requisite SRTMs . . . Whoops! Won't let me make a copy and I didn't want to spend $43 to test it's claim as the best void filling software.

Musandam Peninsula - EVS Precision Map with 50-Meter Contour Intervals (1-175,000)

Plan Four to the rescue. In order to use Void Killer 2, it was necessary to convert the CGIAR-CIS SRTM tile from TIFF into DEM format. I loaded this DEM tile on top of the void peppered HGT file and generated contours . . . voila! It worked. Somehow the combination of the two sandwiched together allowed GM to generate contours without voids. Some of the 50-meter contours spill across water versus land, but that can be cleaned up. The HGT voids were filled with reasonable contours. The combination of SRTM HGT and SRTM DEM did the trick.

I would caution the state of Oman - Gentlemen, don't embark upon a multi-million petro-dollar project based upon these contours. However, I am confident that GM's contour generating algorithm is sound for the intrepid Omani backpacker, who could rely on these contours with a fair level of confidence.

Enjoy!

How To Realign Misaligned Contours Using the "Eye-Ball" Method

Rapa Iti - NE Coast GeoTIFF Image False Color (1:10,000)

I have been working on a redo of Rapa Iti Island, using an exquisite, cloud-free GeoTIFF image, for the past month, an "on-again, off-again" project. I have completed the island's revised shorelines taking them from EVS precision to EEVS precision. That is from Enhanced Vector Shorelines using Landsat ETM+ base imagery to Extremely Enhanced Vector Shorelines using 1-meter resolution base imagery. The "zooming" capabilities of EEVS are pretty amazing. If you map it, one could zoom in to view your backyard which is in need of mowing.

Anyway, the revised Rapa Iti shoreline is finished. I still have a variety of vegetation layers to digitize, along with human ground signatures (dwellings, roads, paths, harbor features, etc). Before I tackled these challenging layers I thought I would quickly generate topographic contours. Using Global Mapper's "Generate Contours" routine, one merely needs to import the appropriate SRTM image for your area of interest, select a few options and one has contours generated in seconds. These contours, when SRTM is accurate and complete (no voids), accurately reflect the elevations of your target area. But, SRTM tiles often have little voids, which can be quickly filled using software tools (my personal favorite is SRTMFill) and in certain areas, big voids, which are difficult to reliably fill. After much thought, research and counting my money (I hate to spend money on pricey software), I decided to use a tried-and-true method "eye-ball-it". Any cartographer worth their salt cringes at this method. Too often, unscrupulous map-makers foist their finished product off as a rigorously crafted map following standard cartograph protocols, when, in fact, the map is more fiction than fact - none of which speaks kindly of the "eye-balling" method. However, since I do not have the means to obtain surveyed contour lines or modified SRTM data of my target island, I'll just "eye-ball" this portion of the project. This involves looking at the underlying land features and comparing them to accurate, reliable contours, guess-timating possible alignment of the various misaligned contour tracks (i.e., 20-meters, 40-meters, etc) and finally realigning them.

Rapa Iti - NE Coast Misaligned Contours Realigned (1:10,000)

Let's begin with Maputu Pt or the NE Coast of Rapa. Because the SRTM data was incomplete, containing a large void (no data or incomplete data), this point had contours that were spilling into the ocean.

Rapa Iti - NE Coast SRTM Base Image Misaligned Contours Realigned (1:10,000)

The above image shows the SRTM limits as a darker shade of green, when, in fact, the actual shoreline is west of the SRTM shoreline. Using the GeoTIFF image, I was able to realign the misaligned contours (see the blue contours). The method I used is "eye-balling". I looked at the shape of the easterm side of the point and beginning with the highest elevation, I realigned each contour level. The eastern side of Maputu Point appears to be quite steep, based on the GeoTIFF image and my contour realignments.

Rapa Iti - NE Coast EEVS Map with Realigned Contours Draped over SRTM Image using GM's 3D Effect (1:10,000)

Using Global Mapper's 3D Effect, I was able to see my work in a different, more revealing way. I think my "eye-balling" of the misaligned contours works quite well.

Rapa Iti - NE Coast GeoTIFF Image False Color (1:10,000)

From my GeoTIFF image to . . .

Rapa Iti - NE Coast EEVS Map Misaligned Contours Realigned (1:10,000)

Rapa Island, Maputu Point using EEVS precision (hi-res imagery). The green contours are generated from SRTM data. The blue contours are my realigned contours. They track the elevations more accurately then the SRTM based contours.

Rapa Iti - S Coast GeoTIFF Image False Color (1:10,000)

Rapa Iti - S Coast Misaligned Contours (1:10,000)

Along with the misaligned contours in the NE, another area of misaligned contours is along the southern shore. Looking at the GeoTIFF image without contours, one sees a coastline fronted by some pretty steep south facing slopes. Looking at the contours generated from the SRTM data, it is very easy to locate the problem contours. The steep slope is either poorly defined or nondefined by contours. Based on SRTM data, it appears the immediate shoreline is a very steep cliff fronting the shore. The contours do nothing toward resolving the remainder of the slope. Also the point that extends out is not defined.

Rapa Iti - S Coast EEVS Map Misaligned Contours Draped over SRTM Image with 3D Effect (1:10,000)

Looking at the contours draped over the SRTM image the reason for the misalignment is obvious. Where there should be elevation data there is none.

So I will carefully "eye-ball" the slope and assigned contours where I guess-timate they belong. You know and I know these realigned contours are nothing more than educated guesses. In the case of this map of Rapa and in consideration of the quality of the base GeoTIFF imagery, the educated guesses will ultimately lead to a good map that is useful to gather an understanding of the topography of this island. If I were to have actual contours that could be applied to this map, I could produce a high quality product useful for a wide variety of purposes. As it is, my map containing "eye-balled" contours will be a quality product useful for a variety of purposes.

Stay tuned. This project should be finished within the next few weeks. Then, you be the judge.

Enjoy!