PriorityFlow : slopes' spatial distribution is faulty

[mvhulten]

In 6cb8289 slopes are calculated.

Slopes in x- and y-direction appear okay, except that the right half of both variables are sort of a ghost mirror image of the left half. Figure out at which step this happens.

[mvhulten]

Strange output.

Comes from modTopo.py.

Output from burnShape2Topo.py is fine and uses SLOTH routines to create netCDF files, so do the same in modTopo.py.

[mvhulten]

modTopo.py doesn't seem to do anything, inspect at later, now use burnt and not further modified HSURFBurned.nc.

Transposing solved the main issue:

[mvhulten]

@kgoergen, visual comparison of slopex:

Left: from original $\hspace{7cm}$ Right: generated with PriorityFlow

[mvhulten]

No pitt filling on the right!
Still, that would not explain the large differences…

edit: Pitt-filling results in different slopes at some places, but visually looks the same as the right figure (and thus still very different from the original).
Then flooding depressions and resolving flats results in yet slightly more different output.
166a6d4

[mvhulten]

After pit-filling, flooding and resolving flats, with PriorityFlow::SlopeCalcUP(), we get:

Same range and scale as above.
The sparseness of the previous plot is gone.
It still doesn't look like what we had.

[kgoergen]

@mvhulten thanks. We may discuss during Z04 meeting and maybe afterwards. Seems to be going into the right direction, lowermost plot has very pronounced burned rivers it seems. Need to check the river burning perhaps.

[mvhulten]

Left figure is from what we have and want to roughly reproduce, the right two are created by PriorityFlow with and without preprocessing (burning and such) of the HSURF.nc input.

original/TSMP1 $\hspace{4.5cm}$ not burnt $\hspace{4.5cm}$ with burning, pit filling, flooding, inflating

Further, I checked that the strong slopes are due to the burning alone (not pixel adjustment, filling, flooding or inflating).
I wonder if burning is done at all in the original slope files.

edit: fixed the plots (no landmask/Problem 2 and no interpolation)

edit: Same but reduced scale: burning now also visible in original (left figure).

[kgoergen]

the middle figure seems to be hinting to the right direction it seems.

[mvhulten]

Problem 1.

Burning makes it less similar to original/TSMP1. Specifically, features from Donau iron door and the Netherlands are not there. It looks like burning has not been done originally.

Problem 2.

Larger part of Norway and Sweden have no hills in new version (independent of the burning). Our DEM—generated in the preceding mklandmask step—shows mountains in Norway, so I think it's some computational artifact, happening probably during slope calculation.
edit: not just Scandinavia have zero values for $x$ and $y$ slopes, also many part of southern Europe is in fact patchy!

I've checked the HydroRIVERS shapefiles that are used for burning: they show rivers everwhere, including in the zero-slope regions from my figures.

Topography (HSURF from generated EUR-11_TSMP_FZJ-IBG3_464x452_EXTPAR.nc) looks fine as well, including high mountains e.g. in Norway and Marocco.

[mvhulten]

Re: Problem 2., it appears the sea mask is mirrored from top to bottom and vice versa! Weird but probably easy to figure out where the problem comes from.

[mvhulten]

Skipping pit-filling doesn't change much. In the non-burnt case, pit-filling has a generally small effect and everywhere less than 1 cm. In the burnt case, filling has a bit larger effect of up to 2 cm compared to no filling.

not burnt $\hspace{8cm}$ burnt

Top figures are without filling and bottom figure with filling (pysheds).

---

Terminology here:

• "burning" is burning and further pixel adjustments (pysheds)
• "pit filling" is filling, flooding depressions and solving flats (original/Nicolas)