Generate LUT for terrain_shadow

davidoesch commented 2 months ago

GOAL

have assets for each DOY /or / solar azimut and solar zenit for the time covered by S2. ( bin mask: 1 shadow, 0 is no shadow)
add switch to code above to use assets instead fo caluclating terrain_shadow
use https://github.com/ChristianSteger/HORAYZON approach / library

out of scope

Deliverables

[ ] for each DAY OF YEAR one COGTIFF in EPSG 32632, for each time resoltion one band with 0 for no shadow and 1 for shadow, ending up in a total of 365 files

STEPS

[x] Define start end hours of S2 orbits : assuming that INGESTTIME according to https://code.earthengine.google.com/197d0cb28197c8cef1646976468818aa 10:10 UTC and 10:48 UTC for all 19k+ tiles
[x] Define temporal daily resolution : 10 min, 15min, 30 min: by running for one day in spring different resolutions to detect if at different altitudes / exposition the shadow moves more than 10m in the timesspan given
[x] merge swisssurface/swissalti to fill gaps
[x] max value of merged product of each rastercell stored GRI grid 10m
[ ] upload to collection TERRAINSHADOW_SWISS , asset: asset name is TERRAINSHADOW_DOY<doy[3]> , set band name to time in "08:03:31 UTC" or "080331Z"
[ ] adapt JS Code to use by default the assets: define DOY , define the BAND to use, etc
[ ] adapt PY Code to use by default the assets: define DOY , define the BAND to use, etc

davidoesch commented 1 month ago

@Tschoun please confirm Define start end hours of S2 orbits : assuming that INGESTTIME according to https://code.earthengine.google.com/197d0cb28197c8cef1646976468818aa 10:10 UTC and 10:48 UTC for all 19k+ tiles

davidoesch commented 1 month ago

use swissalti or swissurface as input: for this ,the "load dem " function needs to be adapted lon, lat, elevation = hray.load_dem.srtm(file_dem, domain_outer, engine="gdal") to use swissalti or swissurface as input. Inspiration can be found here https://github.com/ChristianSteger/HORAYZON/blob/4ed462c3dc00869ab1d665c16c468af2ab1a3f57/horayzon/load_dem.py which does contains hints
make a mask with 0 and 1 ( currently : 0: illuminated, 1: self-shaded, 2: terrain-shaded, 3: not consider)
set band names to minute of the day , so it can be filtered for it. (currently mmhh, easier for inspection)

davidoesch commented 1 month ago

result SHADOW in RED 10:10, BLUE 10:15 , YELLOW 10:20

if we zoom in , we see that within 5min in April we have 10m changes in shadow:

Key findings

[x] r.sunmask.datetime fits the purpose in combination with swissurface3d
[x] 5min steps do make sense
[ ] computational time for whole of switzerland
[x] python script to generate all DOYS qgis_grass_DOY_SHADOW.py.zip is ready generating this file 001_shadow_SWISSSURFACE.tif
001_shadow_SWISSSURFACE.tif.zip

next steps

merge swisssurface and swissalti to fill misisng data of swisssurface in "Meiringen"
resample merged result to S2 GRI 10x10m (might use the hightest cell of swissalti in the GRI cell)

davidoesch commented 3 weeks ago

Test_Swisssurface with max Values in 10meters in 32632 test_small.zip

davidoesch commented 2 weeks ago

Bild unten ROT: Swissalti gemäss satromo 1.0.0 , WEISS gemäss test_swisssurface

swisstopo / topo-satromo