[TowerTalk] HFTA Question

[jimlux]

On 9/24/20 4:43 PM, Bill via TowerTalk wrote:
> I've used N6BV's HFTA for over 15 years, love it and found it very useful.
> Recently I downloaded the terrain from K6TU for a new planned station in V31 and found some strange numbers.  All latitude and longitude numbers were double checked so that is not an issue.  From being in the area where the station will be located and from the Google Earth terrain numbers, things don't agree.  I'm talking about 50 foot hills that don't exist.  From observation, I know there is a steady slow slope to the ocean from an elevation of 45 feet to ocean level 2500 ft away.  No hills, and Google Earth confirms this, but the terrain numbers from K6TU claims there is one
> I manually downloaded the numbers for several different directions and compared the output between the numbers from K6TU and the ones from Google and there are differences.  No gross changes, but enough to change the output by 3 or more dB in many cases.  I'm not badmouthing  K6TU and appreciate his work, but this is quite puzzling.  I've never had this happen before.  In low flat sugar cane areas, Google gives numbers that conform, but the numbers I got from K6TU  show rises of almost 40 ft.
> My guess is that the numbers outside the US are not quite as accurate and perhaps  thick jungle gives an artificial height.  Any thoughts?
>

My experience is from looking at DTEDs and DEMs both pre-SRTM and post-SRTM.

Yes, there will sometimes be unusual anomalies, and anyone who's doing 
"production" coding needs to check for this.  Sometimes, it was things 
like a flipped bit (What is that -10,000 meter that's only 50 meters 
wide) somewhere in the processing chain that generated the model files.

This is especially true when the file you are using has been 
interpolated from some raw measurement file.  Yes, there are data 
validity checks and stuff, but stuff gets through.

For "outside the US" there are issues with
a) governments not wanting too much accuracy in published descriptions, 
so the database generators comply with the request
b) difficulty in getting ground control points to calibrate the 
measurements.

For older (pre-SRTM) data, the raw files were typically digitized from 
topographic maps.  Someone would take digital line graph (DLG) data 
which in turn was hand digitized by someone off a topo map, or 
transformed from some image processing.  As you can imagine, you can 
misplace a dot, or skip a contour in this without too much trouble.  Big 
area databases were often digitized from 1:250,000 scale maps, which are 
1 by 2 degrees.  Sometimes, there is digitized data of areas of "special 
interest" where someone funded scanning/digitizing/processing maps at 
the 1:24,000 or 1:25,000 scale (the familiar 7.5 minute quads).  For 
areas outside the US, the source maps are of varying scale and quality.

A bigger issue is that the maps are of different source age and 
geometric transformations - so you're trying to blend a map with 
elevation data from 1937, done by a crew in the field with a plane table 
and theodolite, with elevation data on a map from 1960 reduced 
semi-automatically from aerial photographs, with maps from 2010 with 
data from SRTM or Airborne LIDAR.

  But a 40 foot or 20 foot elevation error wouldn't surprise me in 
moderately hilly terrain - that's one contour line.  National Map 
Accuracy standards are 1/50" inch on the map. On a 1:24,000 topo that's 
40 feet, horizontally.

SRTM data is better, sometimes.  There's a lot of processing to try and 
not measure the top of the forest canopy, but it's not perfect.  LIDAR 
data, depending on the age, is great at measuring the tops, but it also 
gets the ground return, if the foliage isn't too dense (you get the 
return from the scattered light as well as the "first return") - what 
they do is take the LIDAR data and merge it with SRTM data to try and 
figure out what the true surface profile is.  Sometimes it works, 
sometimes it doesn't.   SRTM's beam width is fairly large, so it sort of 
"averages" (actually, you get a distribution back).  LIDAR is tiny point 
measurements, but lots of them.  Consider a 30 degree hillside that is 
mostly grass and bushes with some 50 foot high 30 foot diameter oak 
trees around.

SRTM's basic vertical accuracy is about 11 meters vertically (mostly due 
to the phase noise, (8m), and the baseline angle (7m)

https://www2.jpl.nasa.gov/srtm/SRTM_paper.pdf

Describes the system design, how they make DEMs, etc. and a lot of 
discussion of the errors.