[GRASS-user] High tension with r.resamp.rst

[Ghislain Vieilledent | CIRAD]

Helena Mitasova sent me an answer yestarday on the topic:

I haven’t used (or looked at) r.resamp.rst for a long time, but the 
tension may be OK becuase it depends on scale
- if you are working in lat/long, the 30sec resolution in decimal 
degrees is a very small number (compare to e.g. 10m resolution DEM),
so the large tension compensates for that. We normalize the data to 
mitigate the scale dependency but for latlong
it is still an issue.
Have ou tried some of the other resampling modules, just to get two 
independent results ? They should be pretty close.

Helena

_______________________________________________________

Ghislain VIEILLEDENT

CIRAD - UMR AMAP
botAnique et Modélisation de l'Architecture des Plantes et des végétations
TA A51/PS2
F-34398 Montpellier cedex 05
FRANCE

Phone: +33.6.24.62.65.07
Skype: ghislain.vieilledent
E-mail : ghislain.vieilledent at cirad.fr

Research unit web site: http://amap.cirad.fr
Personal web site: https://ecology.ghislainv.fr
BioSceneMada project: https://bioscenemada.cirad.fr
_______________________________________________________

On 06/02/2019 15:24, Ghislain Vieilledent wrote:
> Ciao Nikos,
> The pdf is available at the web link provided in the email. Thanks for 
> giving a look.
> Cheers,
> Ghislain
> 
> Le 6 février 2019 15:19:22 GMT+01:00, Nikos Alexandris 
> <nik at nikosalexandris.net> a écrit :
> 
>     * Ghislain Vieilledent | CIRAD <ghislain.vieilledent at cirad.fr> [2019-02-06 14:11:04 +0100]:
> 
>         Dear GRASS GIS user,
> 
>         I am using r.resamp.rst to interpolate climatic data at 30sec
>         resolution over Africa.
> 
>         Grass script, data, and results are available here:
>         https://nextcloud.fraisedesbois.net/index.php/s/CKzsZCSCqMTY9rq
> 
>         #===================
> 
>         # Import Africa shapefile
>         v.in.ogr --overwrite input=gisdata/vectors/Africa layer=Africa
>         output=Africa
> 
>         # Rasterize Africa at 30 arc-sec resolution
>         g.region -ap vector=Africa res=0:00:30
>         v.to.rast input=Africa type=area output=Africa use=val value=1
> 
>         # Import climate
>         r.in.gdal --overwrite input=gisdata/rasters/test1.1_044.tif
>         output=test1_1
> 
>         # Resample with RST
>         g.region -ap raster=test1_1
>         # Note: use maskmap=Africa to save computation time
>         r.resamp.rst --overwrite input=test1_1 elevation=test1_1_rst \\
>         ew_res=0.00833 ns_res=0.00833 maskmap=Africa
>         r.info test1_1_rst
> 
>         # Resample to exact resolution 0:00:30
>         g.region -ap res=0:00:30
>         r.resample --overwrite input=test1_1_rst output=test1_1_30s
> 
>         # Export
>         r.out.gdal --overwrite input=test1_1_30s \\
>         output=output/test1_1_30s.tif type=Float32 \\
>         createopt='compress=lzw,predictor=2'
> 
>         #===================
> 
>         When I use r.info on the interpolated raster, it indicates a
>         tension
>         of 7278 in the comments which seems very high to me as the default
>         value is 40.
> 
>         | Comments:
>         | tension=7278.552524
>         | dnorm=5.495598, zmult=1.000000
>         | KMAX=50, KMIN=35, errtotal=0.000535
>         | zmin_data=0.056488, zmax_data=3.756866
>         | zmin_int=-0.066779, zmax_int=3.214336
> 
>         On the other side, results do not look so bad (see test.pdf figure).
> 
> 
>     I am following this thread out of curiosity/interest on resampling.  The
>     .pdf file is not attached.  How big is it?
> 
>     Nikos
> 
> 
>         Would you have any comment on this high value for the tension
>         and do
>         you think interpolation has been performed correctly ?
> 
>         Best regards,
> 
>         Ghislain
> 
> 
> -- 
> Envoyé de mon appareil Android avec Courriel K-9 Mail. Veuillez excuser 
> ma brièveté.
> 
> _______________________________________________
> grass-user mailing list
> grass-user at lists.osgeo.org
> https://lists.osgeo.org/mailman/listinfo/grass-user
>