Skrifennow

My blog, imported from Blogger and converted using Jekyll.

The Bayesian classifier function in the dissertation results - Part 1: acquiring and preprocessing data

Jan 4, 2016

I have mentioned before a little about having a Bayesian classifier for the Martian terrain segments, which measures the similarity to the Martian terrain.

This is detailed in section 3.4.2 on pages 36-43 of the dissertation, or 43-52 of the tablet-optimised version I will try to recap it here.

The first step was to gather the data:
Beginning with the tabulated Souness et al. 2012 objects, I identified and downloaded the relevant High Resolution and Stereo Camera DTM tiles, reprojected to a common equicylindrical cordinate system with a standard latitude of 40°, and then create derived topographic variables.
I originally did this manually, inputting Souness GLF coorindinates manually into the web interface at PDS Geosciences Node Mars Orbital Data Explorer (ODE). 
I retrieved both the nadir image files and the areoid elevation.

It is also possible to download shapefiles of coverage footprints of various Mars satellite data and identify coverage of Souness GLF footprints in a more automated fashion.


The actual 6 band feature vector was
  1. Nadir image
  2. DTM elevation
  3. Slope (degrees)
  4. Aspect (in absolute degrees from north)
  5. Cross-sectional curvature
  6. Longitudinal curvature

The software I used for this was Jo Wood's LandSerf. In his PhD thesis there is a good explanation of the various different types of topographic curvature. Unfortunately this is no longer easily accessible, except via the Internet Archive however the various types of curvature are also explained in this website forming part of an online textbook on Geospatial Analysis.

After processing in LandSerf I used gdal at the command-line to generate a layerstacked .kea file for each tile.

The longitudinal and cross-sectional curvatures are defined in the slope's own plane, i.e. the plane containing the slope vector with normal pointing in the aspect direction.

The longitudinal curvature is a measure of convexity/concavity of the slope, and cross-sectional curvature the rate of change of the slope vector, when considered in the slope's own reference frame (as opposed to plan curvature which considers the same from vertically above).

The 10 band layerstacks, which I didn't process further due to the processing time also included Mean curvature, plan curvature, profile curvature and Landserf feature classification.

Later on, in reprocessing the results after the dissertation I used the original aspect rather than the absolute aspect from N.

TuiView screenshot of the h0037 HRSC DTM tile. I have coded red:slope, green:nadir image and blue:elevation.





More zoomable images - on Mars this time

Dec 31, 2015

I previously shared my index to some of my posts about Martian glaciers.

I have produced zoomable images of the various parts of the Martian surface, with imagery from HRSC nadir images along with elevation.

Unfortunately, I have not been able to show the full resolution of these, because the method of tiling the 16384x16384 images I use as the basis of the work, using gdal2tiles.py and displaying using leaflet.js, involved a very large total number of files, and it seems my current web host at neocities.org has a maximum.

Nevertheless, the results can be seen at taklowkernewek.neocities.org/mars/hemispheres.html and taklowkernewek.neocities.org/mars/regions.html.

Sample screenshot from my website









Update - it looks to me like even with only 4 zoomlevels, I have still managed to max out my number of files count. I am also in process of adding links to filtergraph.com where I have a number of scatter plots showing the Souness GLFs by region.



Numbers of Souness GLFs by region.

An example of using Filtergraph to plot the various Souness GLFs in the Olympus Mons area.

Using Leaflet.js for making my maps zoomable

Dec 30, 2015

I have come across the Javascript library Leaflet.JS which I have begun applying to my maps of Cornwall.

It may be used, in conjuction with Mapbox, to render OpenStreetMap in a number of styles, for example:  link to taklowkernewek.neocities.org.

I have also made one of the versions of my cycling maps of Cornwall zoomable. This is simply done by rendering a 16384x16384px image of the map in QGIS, and then simply using Leaflet.js to make it zoomable, with the assistance of gdal2tiles.py to make the various 256px square tiles needed.

Link: taklowkernewek.neocities.org/mappys/leafletJStest/westcornwall.html

Follow link for zoomable version of this map, and also further maps of mid and east Cornwall.

Assessing possible typographic errors in Souness catalog - southern hemisphere

Dec 28, 2015

Detailing the Souness GLFs which have a large offset between the quoted centre, or are duplicates, or have been visually identified as potentially having a typographic error, I now continue with the southern hemisphere.

Argyre

see also previous post.
S732,  S809
Souness 732 in the wall of crater Arkhangelsky at 24°W, 41°S has an off-centre 'centre' position, but does not have HRSC DTM coverage.

Souness 809 at 49°S in tile h2508 has a long and sinuous shape where the 'centre' falls outside of the channel.
Showing broad context of this area, to the west of Argyre Planitia, a system of valleys is seen, along with dunes.

It is possible that Souness 809 is an example of a glacier forming in a pre-existing older fluvial valley system.

West Hellas

see also previous post
S1124, S1145, S1175, S1183, S1188, S1189, S1197, S1237, S1253/S1309:
Souness 1124 in an unnamed crater not covered by a HRSC DTM.

Souness 1145 also falls outside of HRSC DTM coverage.

Souness 1175 is among many GLFs around the walls of a crater.

Several Souness GLFs are shown in anaglyph ESP_014335_1450.
Souness 1197 is found on the north wall of a crater where the floor is covered by a dune system.

Souness 1237 on the southern wall of crater 'Russell' which contains a system of dunes.

Crater Russel at ~ 55°S, 13°E.
S1253/S1309 are duplicates in this crater containing many GLFs at 38°S, 16°E.

East Hellas

see also previous post.
S950/S1271, S951/S1272, S978/S1274 , S999/S1278, S1000/S1277, S1016, S1073/S1294, S1085/S1292, S1098/S1297, S1107/S1298, S1117/S1301, S1275, S1300.

Souness 950 is a duplicate with Souness 1271 and S951 with S1272 in the north wall of Greg crater.

A wider view, also showing the duplicates S978/S1274

Souness 978 and 1274 are duplicates and another two Souness GLFs are found around a mesa to the NW of Greg crater at ~ 111°E, 37°S.

Souness 999 and 1000 are duplicates with S1278 and S1277 in an interesting region covered previously.

Broader context for S999/1278, S1000/1277 near Reull Vallis which is at the NW of this image. Souness 1275 is seen at the S of this image.

S1016 is out of HRSC DTM coverage on a piece of raised terrain with Souness GLFs on both north and south facing sides at 41.5°S, 105°E.

Broader context for S1016.

S1073 and S1085 are duplicates with S1294 and S1292.

Broader context for Souness 1073/1294 and Souness 1085/1292. Souness 1098/1297 is also seen.
Souness 1098 is a duplicate with S1297.

Souness 1107/1298 is seen in a shadowed area on the margin of Reull Vallis itself. It is covered by an anaglyph PSP_001978_1445.

Souness 1107/1298 and S1299 are visible in this anaglyph.
Souness 1117 is a duplicate with S1301 and the malformed S1300 (typo for the 'head' location?) also overlaps.

Souness 1275. The centre is displaced out of the outline of the extent due to the glacier flowing around the topography.

Southeastern Highlands

see also previous post.
S1270
Souness 1270 within a crater at 169°E 46°S.

Zooming into the crater wall.

Conclusion

For the purposes of the Bayesian classifier, I believe that only 2 of the 1309 GLFs would have been particularly problematic, S359 and S1300. The naive assumption of constant channel width does add in some of the wall terrain anyway, even if the 'centre' is still within the 'extent'. These two would be the only ones to be worth manually correcting for my purposes.

I think it could be beneficial is to retrain the classifier with the 'centre' areas only, i.e. use only the terrain elements that overlap with the centre of each GLF. Theoretically, the 'head' areas should be the terrain that governs accumulation of glaciers, and this I have already done.

An extended investigation could be to take (some?, all?) of the GLFs and try to draw a profile of the mid-line using best available imagery and then try to draw conclusions based on this.

Assessing possible typographic errors in Souness catalog

Dec 26, 2015

In the Souness et al. 2012 paper, a catalog is presented with the various glacier-like-forms identified.

In my analysis, I assume a constant channel width, which is not always a good assumption, and in a few cases, there are duplicates, or what appears to be a possible typographic error. There are only a few out of the 1309 which are likely to be affected.

I have identified some of these in the spreadsheet at this link.

I show the count of Souness GLFs by the regions I used earlier:

Filtergraph link: filtergraph.com/4134840
There are only relatively few that look like they have a typographical error, for example Souness 359, and Souness 1300:

Red dot shows 'head', white 'centre' and dark blue 'terminus'

S1117 and S1301 are duplicates and S1300 appears to contain a typo

I decided to go through systematically, and look at Souness objects that either looked odd in shape visually, or had a distance between the 'centre' recorded by Souness and the middle of the channel that was more than the channel width.


I below detail the individual Souness objects where there is some kind of query, either being a duplicate or centre out of the mid-channel. In Google Earth, I show heads in blue, termini in yellow, the 'centre' with a snowflake, and left and right mid-channel in magenta and green.

Overall, I don't think this on its own will have made a lot of difference to the classifier, since any DEM segments that overlap the extent would be chosen for the input, and a little more conservative 'extent' for those Souness GLFs that currently include non-glacial terrain would not likely make a decisive difference.

Northern Hemisphere

North of Olympus Mons

See also previous post.

In this area, Souness objects 87, 87, 96 and 108 stand out as odd-shaped.
Location, with the four Souness objects noted highlighted in yellow.

Souness 87 and 88 overlap significantly near the edge of h0037. Souness 91 and 92 are also visible.
Oblique view in Google Earth showing S87, 88, 91 and 92.

Souness 96 starts outside the HRSC tile h0037's area of coverage and joins what looks like a larger valley fill glacial form. The crater also appears to have glacial features infilling it, that are not catalogued by Souness.
Souness 96 looking west with CTX imagery in Google Earth.

Looking down on S96 with CTX imagery. There are other glacial like forms not catalogued by Souness since his intention was to catalogue analogues of valley-glaciers on Earth. In this area of Mars, recent (last few Myr) glacial activity is believed to be superimposed on an older landscape of the larger topographic forms.

S108 is outside any HRSC DTM tile.
The same area, in Google Earth. Note the slope liniae in the crater at lower-right. These may be the result of current episodic brine flows.

Zooming in on S108, a slope line is seen on the slope constraining it topographically.

Mareotis Fossae area

See also previous post.
Souness 12 in context within HRSC DTM tile h5401.

Souness 12 looking NW in an oblique CTX image mosaic. It is an interesting question why exactly Souness 12 is formed where it is when to the right gullies cover the slope instead.
Souness 47 (left) and Souness 31 (right). 46°N, 82-85°W.

Souness 44 at upper-left, with S43 and S45 nearer the centre.
Souness 47.
Souness 38 amid a glaciated landscape.

Souness 31 and 32 (north to south).
Souness 11, 12, and 14 within context showing Bayesian classifier result.

Souness 47 (left) and Souness 31 (right) with S43, 44, 45, 38 and 32 also shown. See also same area in previous post. 46°N, 82-85°W.



Deuteronilus Mensae

See also previous post.

In this area S6, 590, 617, 618, 634, 693, 708, 720, and 724 either have a markedly off-centre 'centre' or are duplicates.

Souness 6. 30°N, 11°W in HRSC tile h1498.

Souness 6 CTX images.
The duplicated Souness objects S567/578 and S566/577 are shown to the right above.

Souness 590 among several objects not covered by any HRSC DTM.
The same GLFs as the previous image, showing CTX images via Google Earth. Perhaps there has been several glacial epochs in the 'recent' times of the last few Myr that have been superimposed on the longer-term formed topographic channel?

Souness 617 and 618 among other GLFs along a topographic boundary.

S617/S618
Souness 634 at a topographic boundary where it appears there is other glacially reshaped terrain not classified as Souness objects.

S634

S634 in broader context.
Souness 693 among other GLFs and at least one more uncatalogued example near the top of this field.

S693 in context, looking west with CTX imagery.
Souness 720 and 693 highlighted.

S720 in upper-centre of image.

S720.
Souenss 708 and 707.

Souness 724.
S724 oblique view.

Protonilus Mensae

See also previous post.


Souness 337 and 338, along with some further Souness objects. ~ 36N, 57°E.

Souness 351 shows an odd shape, and is here depicted with other Souness GLFs. S348 perhaps has a position error, if it is in fact a little to the SW of where it is plotted. ~ 37°N, 56°E.

Visualising the Bayesian classifier Ln(K) for extents (blue) and 'head' areas (red).

Souness 380. 53°E, 40.5°N.

Souness 380, 389 and 390 in context along with a number of others.

Souness 389, 390.
Anaglyph PSP__009020_2210_PSP_008519.

Souness 400, 401, 402, 403, 404. ~ 52°E, 40°N.

Souness 445 shows an off-centre centre, where two other Souness GLFs are found on a mesa at 49°E, 46°N.

Souness 472, 473 and 474.

Souness 491.

 

Nilosyrtis Mensae

see also previous post.
Souness 204 and 205.
Souness 204 in the anaglyph ESP_027823_2100.

Souness 230 and 231 merging, along with 2 other Souness GLFs pictured.

Souness 245 at the low latitude of 28°N, where most of the surrounding valleys are non-glaciated, or perhaps where glaciers have now ablated.

Souness 247 is off-centre, shown here with some surrounding GLFs.

Souness 260.

Phlegra Montes

see also previous post.
Tiles h1423 and h1412, showing S150 at top, the duplicates S154 and S158 near top, and S156 near the bottom.
Souness 156.

Duplicates S154 and S158.
S156 in an oblique view looking W showing CTX imagery in Google Earth.

A higher view showing S156 and the duplicates S154/158.


S156, and S159 amid a complex glacial landscape.

S156 and S159
The southern hemisphere fields are shown in a post to follow.

All Posts