Mail
NEW

Comparing programs that calculate the skyline profile

Computer programmes

There are, to my knowledge, three programmes, that are based on Shuttle Radar Topography Mission 90m (SRTM 3") and 30m (SRTM 1") resolution datasets. And these programmes calculate the skyline profile as seen from a certain location:

HeyWhatsThat by Michael Kosowsky (accessed on May 10th, 2015)
A program that is based on web services and is relatively easy to use. One can chose a location on a map and then calculate the skyline profile seen from that location. The default observer height is 2m (but can be changed in steps of 1m). One can also make a Stellarium landscape file and one can download a csv file so one can include it in a spreadsheet. It has a Webservices interface (and for instance incorporate in skyscapeR). The profile can also be included in Google Earth. The min. azimuth resolution is 0.125 degrees.
It is based on the SRTM 3" data and uses 2D nearest-neighbor interpolation. I helped testing some aspects of this service.
SyntH by Fernando Patat (acquired on May 8th, 2015)
The programme is not available for running on the web or on one's own computer. The default observer height is 1.5m (but can be changed). It can provide a csv file so one can include it in a spreadsheet or Stellarium.
The program supports SRTM 3" and SRTM 1"and uses 2D nearest-neighbor interpolation. I helped testing some aspects of this program.
Horizon (0.13c) by Andrew Smith
It is a standalone progamme that can generate the skyline profiles. The default observer height is 1.5m (but can be changed). One can also make a Stellarium landscape file in which is horizon.txt, and this txt file can be included in a spreadsheet.
It supports SRTM 3", SRTM 1" and many other datasets (like LIDAR). It uses linear interpolation.
StandingStones by David Hoyle
A web based service with a very nice user interface, it has a pre-calculated set of skylines. The observer height is 2m. It has a Webservices interface. One can also generate a Stellarium landscape file.
It supports SRTM 3", SRTM 1" and many other datasets (like LIDAR). It uses linear interpolation.
r.horizon in GRASS
This is a standalone GIS progamme that can generate viewsheds and skyline profiles.
It supports SRTM 3", SRTM 1" and many other datasets. It uses 2D nearest-neighbor interpolation.

Panorama photos and theodolite measurements

Beside the above skyline profiles programmes, Mike Wilson manages the Mega-What website which has skyline profiles from several Irish Axial Stone Circles (ASC) based on real panorama photos and theodolite measurements. A great database! The height of the recording equipment is around 1.25m.

So the results of the above four computer programmes have been compared with Mega-What theodolite measurements and panorama photos.

Comparison

The comparing results of the four programmes and theodolite measurements of the panorama photo can be seen in the below Fig. 1.

Fig. 1 Comparing skyline profiles from
computer programmes: HeyWhatsThat (red [SRTM 3"]), SyntH (purple [SRTM 1"]),
Horizon (orange [SRTM 1"]) and r.horizon (green [STRM 1"])
with theodolite/photmetric measurements from Mega-What (black triangles/background)

In Fig 2 there is a comparison between SyntH, Horizon and StandingStones-linear.

Fig. 2 Comparing skyline profiles from
computer programmes: SyntH (purple [SRTM 1"]),
Horizon (orange [SRTM 1"]) and StandingStones-linear (background [STRM 1" linear])
with theodolite measurements from Mega-What (black triangles)

SyntH (SRTM 1") and r-horizon (STRM 1") is quite close to the theodolite measurements, while HeyWhatsThat (based SRTM 3") a little less.
The computer skyline profiles don't map the Mega-What theodolite measurements sometimes, like: the small evaluation around 310 azimuth (this is a small nearby hill); and the peaks on the right side of the hill. If a lower eye height (say 0.5m) would be chosen, this small hill will emerge in the skyline (like in r.horizon).
The skyline profile of Horizon programme and StandingStones-lineair webpage have the same outline (as tested for STRM 1"), but both have somewhat different apparent altitudes around azimuth between 120 and 270 deg compared to the others. The difference is due to parallax: the nearby horizon (200m to 600m) and the low resolution of STRM 1" (see next section).

In an earlier panorama of Mega-What there was a shift of the small hill (at azimuth of ~40 degrees), this was due to misalignment when stitching the panorama photos. This was improved with the help of the above computer programmes.

Sensitivity analysis of position of observer and nearby horizons

A sensitivity analysis was done for three Irish ASC sites (Kealkil, Lettergorman and Cashelkeelty West) using Horizon.

Site	Latitude [^o]	Longitude [^o]
Kealkil	51.745204	-9.370549
Lettergorman	51.658575	-9.067067
Cashelkeelty West	51.757194	-9.814921

In Horizon the longitude, latitude and eye height of the observer were varied until Horizon's results were as close as possible (matching) to the theodolite/photometric results of Mega-What.
A square of q*q m (q being 10, 20 or 30m in NS-EW direction) was defined centred on the longitude & latitude of the site. At the corners of this square, the observer was positioned and checked if Horizon matched most Mega-What. When finding the most matching corner, the eye height (1.5, 2, 2.5 or 3.5m) was varied to see if another eye height matched better (in all cases no improvement from 1.5m). For each site between 5 to 13 postions we tested. The matching was effectively done for nearby skylines, as the far away skylines were not really influenced by these small changes in longitude, latitude and eye height.
This provided the most matching observer position in Horizon compared with theodolite/photometric of Mega-What. Remember that the most matching observer postion should have been at the monument, but for some programs this is not the case (eg due to low resolution of STRM1" for nearby horizons).

Here are the most matching observer position:

Site	Square q*q [m]	Most matching corner	Most matching eye height [m]	Figure
Kealkil	20*20	SW	1.5	3
Lettergorman	20*20	NW	1.5	4
Cashelkeelty West	20*20	SE	1.5	5

So you see that an Eye height of 1,5m in Horizon looks most matching for these three sites. This is a nice consistent result. And the most matching longitude and latitude of the observer in Horizon was some 15m (10*√2) out from the site location. Furthermore not always the same corner of the square was most matching, so it is not likely that there are STRM coordinate reference issues. And this off-set of some 15m is not that large compared to the STRM 1" pixel size of 30*30m (and its related accuracy): this results in parallax due to the nearby horizon (200m to 600m) and the low resolution of STRM 1"
The interesting thing is that SyntH (STRM 1") results are, by default, very close to Mega-What's and thus the most matching Horizon's results. But that is very likely a coincidence.

Here are the pictures with Mega-What, SyntH and the most matching observer position using Horizon.

Fig 3 Kealkil
<the difference around 130 and 200 deg is due to paralax: low resolution of STRM 1" pixels and a 180m nearby horizon>

Lettergorrman: Horizon SyntH Mega-What

Fig 4 Lettergorman
<difference between SRTMs and theodolite around 250deg azimuth, could be due to forestation when STRMs were measured>

CashelkeeltyW: Horizon SyntH Mega-What

Fig 5 Cashelkeelty West

Summary

Careful stitching of panorama photos (like in Mega-What) is essential to overcome photo distortions and rotation errors. That is why theodolite (reference) measurements or computer programmes comparisons are important.

Of course a photo (or theodolite measurements) can be most of the time better than a computer programme. But for photos one needs to have the opportunity to visit the site at clear weather conditions, which is not always possible. So computer programmes can help solve issues with far away skylines, due to atmospheric conditions they are sometimes not visible on photos/theodolite.

For skylines further than 5km, all programmes, datasets and the photo give similar results.

The above comparison shows that the SRTM 1" dataset (30m resolution) is better (due to a higher resolution) than the SRTM 3" dataset. And one can assume that LIDAR would even be better (as it has a resolution of around 0.5m or better). The artifacts seen in STRM1" though would be very similar with LIDAR; except the nearby horizon would need to be 60 times nearer by. So it is important, regradless of dataset, to have the correct rendering methodology.

Both HeyWhatsThat, SyntH and r.horizon follow more or less the skyline shown in the Mega-What panorama photo. The SRTM 1" resolution (min. 30m) is too low to depict nearby features (say closer than 1.6 km) accurate enough (say better than a forth of the Sun/Moon's diameter) in the skyline.

One has to remember that low level vegetation will have influence on STRM (not so much on LIDAR and perhaps theodolite). And for nearby hoirzons this can also bring up differences

The height plateaus in HeyWhatsThat, SyntH and r.horizon could perhaps be made more natural(?) by using a proper smoothing/interpolation method, such as linear interpolation.

Horizon and StandingStones-linear are very similar and have a slightly different skyline seen for a Mega-What location, and this can be expected due to parallax of nearby horizon and the low resolution of STRM 1".

A combination of all methods (visits, photos, theodolite and computer programmes) is really important to appreciate reality.

References

Patat, Fernando. 2011. 'Horizon synthesis for archaeo-astronomical purposes', Astronomische Nachrichten, Vol 332: pp. 743-49.

Acknowledgments

I would like to thank the following people for their help and constructive feedback: Brian Doyle, David Hoyle, Michael Kosowsky, Andrew Smith, Fernando Patat, Mike Wilson, and all other unmentioned people. Any remaining errors in methodology or results are my responsibility of course!!! If you want to provide constructive feedback, let me know.

Disclaimer and Copyright

Home

Mail

Major content related changes: May 24, 2015