[GRASS-user] raster boxplots wrong assignment

Sat Jun 8 07:05:37 PDT 2024

Dear Micha

Thank you very much.

That’s really strange, using the same code, I got another boxplot with the wrong assignment.

Did you re-order or postprocess “Untersuchungsquadrate_2020-2024”?

New mask before v.extract? For example 511150 is >52000 but does not appear in your  boxplot. However, the assignment in your boxplot is perfect.

v.extract Untersuchungsquadrate_2020_2024 out=quadrat_3 where="ID_Quadrat>520000 and ID_Quadrat<600000" --o

v.to.rast quadrat_3 out=quadrat_3 use=attr attribute_column=ID_Quadrat --o

r.boxplot -o input=al_pres_apple_spring_merged zones=quadrat_3 output="al_pres_boxplot_3.png" raster_statistics=median –o

Kind regards

Sibylle

From: Micha Silver <tsvibar at gmail.com> 
Sent: Saturday, June 8, 2024 11:50 AM
To: sibylle.stoeckli at gmx.ch; grass-user at lists.osgeo.org
Subject: Re: [GRASS-user] raster boxplots wrong assignment

Here it is:

$ v.extract quadrat out=quadrat_3 where="ID_Quadrat>520000 and ID_Quadrat<600000" --o
$ v.to.rast quadrat_3 out=quadrat_3 use=attr attribute_column=ID_Quadrat --o
$ r.boxplot -o input=al_pres zones=quadrat_3 output="al_pres_boxplot_3.png" raster_statistics=median --o

HTH

On 07/06/2024 16:56, sibylle.stoeckli at gmx.ch <mailto:sibylle.stoeckli at gmx.ch>  wrote:

Dear Micha

Many many thanks for your valuable GRASS GIS support.

It would be very interesting to see your output using v.extract with three boxplots

v.extract Untersuchungsquadrate_2020_2024 out=quadrat_hi where="ID_Quadrat>520000 and ID_Quadrat<600000" 

Kind regards

Sibylle

From: sibylle.stoeckli at gmx.ch <mailto:sibylle.stoeckli at gmx.ch>   <mailto:sibylle.stoeckli at gmx.ch> <sibylle.stoeckli at gmx.ch> 
Sent: Wednesday, June 5, 2024 2:53 PM
To: 'Micha Silver'  <mailto:tsvibar at gmail.com> <tsvibar at gmail.com>
Subject: RE: [GRASS-user] raster boxplots wrong assignment

Dear Micha

Hmm it doesn’t make any difference. Strange.

Thanks a lot

Sibylle 

From: Micha Silver <tsvibar at gmail.com <mailto:tsvibar at gmail.com> > 
Sent: Wednesday, June 5, 2024 1:19 PM
To: sibylle.stoeckli at gmx.ch <mailto:sibylle.stoeckli at gmx.ch> ; grass-user at lists.osgeo.org <mailto:grass-user at lists.osgeo.org> ; p.vanbreugel at gmail.com <mailto:p.vanbreugel at gmail.com> 
Subject: Re: [GRASS-user] raster boxplots wrong assignment

On 05/06/2024 11:33, sibylle.stoeckli at gmx.ch <mailto:sibylle.stoeckli at gmx.ch>  wrote:

Dear Micha

Thank you very much. Perfect I was able to reproduce your code with v.extract.

Yes it makes sense. The values in the raster file are the output of a pollination model. Before investing a lot of time and to compare the output of different land-use maps and climate scenarios I decided to work on a subsample, e.g. 15 squares (or classes). At the end I will have around 50 squares and present just these squares.

One last question

It is really strange, but when selecting ID_Quadrat between 52000 and 6000 (boxplot values for 3 squares are available, not just for one square) the assignment is still wrong

v.extract Untersuchungsquadrate_2020_2024 out=quadrat_hi where="ID_Quadrat>52000 and ID_Quadrat<600000"

------------------------------------------------------------------------------^^^^^^^----------------------

52000 or 520000 ??

The boxplot should be for 527120, 569118 and 581166, and not for 503150, 563190, 575110 (figure see attachment).

Kind regards

Sibylle

From: Micha Silver  <mailto:tsvibar at gmail.com> <tsvibar at gmail.com> 
Sent: Tuesday, June 4, 2024 1:21 PM
To: sibylle.stoeckli at gmx.ch <mailto:sibylle.stoeckli at gmx.ch> ; grass-user at lists.osgeo.org <mailto:grass-user at lists.osgeo.org> ; p.vanbreugel at gmail.com <mailto:p.vanbreugel at gmail.com> 
Subject: Re: [GRASS-user] raster boxplots wrong assignment

Hi Sibylle:

On 04/06/2024 10:07, sibylle.stoeckli at gmx.ch <mailto:sibylle.stoeckli at gmx.ch>  wrote:

Dear Micha

Ah I see your suggestion for

# Create raster for zones, all quadrats with the same value
$ v.to.rast input=quadrat output=quadrat use=val value=1

1.	Is it possible to plot several squares/classes with v.to.rast? So something similar to Fig 1 (all squaraes) or just the squares with values in the raster file “al_pres_apple_spring_merged.tif”?

Yes, sure. But does it make sense to produce 179 (the number of quadrats) separate boxplots?

Here's an example where I choose only the ID_Quadrat between 730000 and 740000 (11 quadrats). Then I converted to raster and did a boxplot.

Note that only one quadrat, 734128, has any pixels with "al_pres_apple_spring" values above 0.

$ v.extract quadrat out=quadrat_hi where="ID_Quadrat>730000 and ID_Quadrat<740000" --o
$ v.to.rast quadrat_hi out=quadrat_hi use=attr attribute_column=ID_Quadrat
$ r.boxplot -o input=al_pres zones=quadrat_hi output="al_pres_boxplot_hi.png" raster_statistics=median --o

2.	 
3.	To challenge with fig 1 is that the assignment of “ID_Quadrat” to the values in the raster file “al_pres_apple_spring_merged.tif” is wrong. Instead of “value 1” I would like the assignment to “ID_Quadrat” similar to fig 1. For example I do not have values for the last one, “737182”. The order is somehow mixed-up.

This quadrat does appear in the attached plot. But with a thin line at the median, since all pixels in that quadrat have al_pres_apple_spring values of zero. In fact I notice that all pixels throughout the al_pres_apple_spring are zero, except for a few of the quadrats. It looks as if this al_pres_apple_spring was either masked, or prepared in advance to cover only certain quadrats. Does that make sense? Do you have the original al_pres_apple_spring map?

4.	 

Kind regards

Sibylle

Fig 1

From: Micha Silver  <mailto:tsvibar at gmail.com> <tsvibar at gmail.com> 
Sent: Monday, June 3, 2024 3:11 PM
To: sibylle.stoeckli at gmx.ch <mailto:sibylle.stoeckli at gmx.ch> ; grass-user at lists.osgeo.org <mailto:grass-user at lists.osgeo.org> ; p.vanbreugel at gmail.com <mailto:p.vanbreugel at gmail.com> 
Subject: Re: [GRASS-user] raster boxplots wrong assignment

Hi Sibylle:

So the problem is getting all the vector quadrats as a single value into the "zones" raster? Then I would use the parameters:

'attr=val value=1' in the v.to.rast step.

Here, below, are the steps I tried. At the start I would point out that the raster "al_pres_apple_spring_merged.tif" has very low variance, with almost all cells at value 0. When you make of boxplot of these values, you get a very narrow band near to the median value, very near to zero. Maybe that's what is causing the confusion.

# Create new GRASS location/mapset based on geotiff
>grass -c al_pres_apple_spring_merged.tif ./CH1903+LV95

# Now in that location Import both layers
$ r.import input=al_pres_apple_spring_merged.tif output=al_pres
$ g.region -ap rast=al_pres
projection: 99 (CH1903+ / LV95)
zone:       0
datum:      ch1903
ellipsoid:  bessel
north:      1270975
south:      1118025
west:       2527000
east:       2804000
nsres:      25
ewres:      25
rows:       6118
cols:       11080
cells:      67787440

# Create raster for zones, all quadrats with the same value
$ v.to.rast input=quadrat output=quadrat use=val value=1
# Prepare boxplot, I add the '-o' flag to also plot outliers
$ r.boxplot -o input=al_pres zones=quadrat output="al_pres_boxplot.png" raster_statistics=median

Have a look at the attached plot. 

Now I check the univariate statistics of the al_pres raster:

# al_pres univariate statistics
$ r.univar al_pres
 100%
total null and non-null cells: 67787440
total null cells: 0

Of the non-null cells:
----------------------
n: 67787440
minimum: 0
maximum: 0.315331
range: 0.315331
mean: 3.23424e-05
mean of absolute values: 3.23424e-05
standard deviation: 0.0022431
variance: 5.03148e-06
variation coefficient: 6935.47 %
sum: 2192.40546066407

Note that the variance is quite small. No wonder that the boxplot shows the median as a very narrow band almost at value zero.

Again, if I'm missing something, let me know.

On 03/06/2024 13:55, sibylle.stoeckli at gmx.ch <mailto:sibylle.stoeckli at gmx.ch>  wrote:

Dear Micha

Dear community

Find enclosed the two files necessary to run the code:

1.	The raster file to plot boxplots: al_pres_apple_spring_merged.tif
2.	The vector/shape file for the different quares (classes, regions…):Untersuchungsquadrate_2020-2024_shp

https://drive.google.com/drive/folders/14UwptmBQTtKYQFGeecRWa0n_lWDhRc9s?usp=sharing

Code based on 

https://ecodiv.earth/post/drawing-boxplots-of-raster-values/

g.region -a vector=Untersuchungsquadrate_2020_2024 

raster= al_pres_apple_spring_merged.tif

r.mask vector=Untersuchungsquadrate_2020_2024

v.to.rast input=Untersuchungsquadrate_2020_2024 at PERMANENT output=Quadrate use=attr attribute_column="ID_Quadrat" label_column="ID_Quadrat"

--> Here is the identified challenge: transformation form “Untersuchungsquadrate” to “Quadrate” (as input for r.boxplot). attribute_column ID_Quadrat or cat provide false assignment. It just takes Number 1 (either cat or ID_Quadrat.

r.boxplot -o input=al_pres_apple_spring_merged.tif zones=Quadrate bx_sort=ascending raster_statistics=median

Kind regards

Sibylle

From: sibylle.stoeckli at gmx.ch <mailto:sibylle.stoeckli at gmx.ch>   <mailto:sibylle.stoeckli at gmx.ch> <sibylle.stoeckli at gmx.ch> 
Sent: Friday, May 31, 2024 12:10 PM
To: 'Micha Silver'  <mailto:tsvibar at gmail.com> <tsvibar at gmail.com>; 'grass-user at lists.osgeo.org <mailto:grass-user at lists.osgeo.org> '  <mailto:grass-user at lists.osgeo.org> <grass-user at lists.osgeo.org>
Subject: RE: [GRASS-user] raster boxplots wrong assignment

Yes: see line 3

From: Micha Silver <tsvibar at gmail.com <mailto:tsvibar at gmail.com> > 
Sent: Friday, May 31, 2024 12:06 PM
To: sibylle.stoeckli at gmx.ch <mailto:sibylle.stoeckli at gmx.ch> ; grass-user at lists.osgeo.org <mailto:grass-user at lists.osgeo.org> 
Subject: Re: [GRASS-user] raster boxplots wrong assignment

On 31/05/2024 12:48, sibylle.stoeckli at gmx.ch <mailto:sibylle.stoeckli at gmx.ch>  wrote:

Dear Micha

Thanks a lot interesting idea: Unfortunately I got a parse error. However the error/challenge is bevore selecting a specific region or site. I was able to plot one boxplot using a different code.

The challenge is that I use this code here to make “Quadrate/Squares” out of the vector layer “Untersuchungsquadrate”, because in r.boxplot you have input raster and zonal raster. Exactly in this code you make the assignment by defining attribute_column. I used both cat and ID_Quadrate, but it seems that my rasterized file “Quadrate” is not yet the one I need.

Can you post the output of: v.info -c Untersuchungsquadrate_2020_2024  ? (Just to verify the ID_Quadrate column)

Ecodiv.earth - Boxplots based on raster data in GRASS GIS <https://ecodiv.earth/post/drawing-boxplots-of-raster-values/> 

v.to.rast input=Untersuchungsquadrate_2020_2024 at PERMANENT output=Quadrate use=attr attribute_column=cat label_column="ID_Quadrat"

OK, after this command you should have a new raster "Quadrate" with pixel values taken from the cat values of the vector. (Why cat and not the ID_Quadrate?)

In the r.mapcalc expression:

1- The full expression needs to be in double quotes, 

2- You need to use an existing **raster** map. Not the vector attribute values.

i.e.

r.mapcalc "MASK = if(Quadrate == 803174), 1, NULL())"

But, as you can see below (since you used the Untersuchungsquadrate_2020_2024 cat values (instead of the ID_Quadrate attrib), the raster values vary from 1 to 179. No value 803174...

I'd suggest to go back to using ID_Quadrate as attribute_column in your v.to.rast .  Then try the r.mapcalc expression to get the mask defined. (Don't forget to remove the mask after: r.mask -r)

HTH

From: Micha Silver  <mailto:tsvibar at gmail.com> <tsvibar at gmail.com> 
Sent: Friday, May 31, 2024 11:09 AM
To: sibylle.stoeckli at gmx.ch <mailto:sibylle.stoeckli at gmx.ch> ; grass-user at lists.osgeo.org <mailto:grass-user at lists.osgeo.org> 
Subject: Re: [GRASS-user] raster boxplots wrong assignment

Hi Sibylle:

I'm not sure I fully understand the problem. If the below example is not relevant, just ignore...

I did a simple test using the nc_spm_08 location. The PERMANENT mapset includes an elevation raster 'elev_srtm_30m' and a categorical landcover raster 'landclass96'. I used the landcover raster as zones, and elevation for the values in the boxplots.

g.region -ap rast=elev_srtm_30m

# All landcover zones

r.boxplot input=elev_srtm_30m zones=landclass96 raster_statistics=median output=~/work/tmp/boxplot.png --o

# See attached boxplot image

# Now limit to only one zone, the forests, class 5 in this raster:

r.mapcalc "MASK = if(landclass96 == 5, 1, null())"

r.boxplot input=elev_srtm_30m zones=landclass96 raster_statistics=median output=~/work/tmp/boxplot_forest.png

# See attached boxplot_forest

r.mask -r

Hope this is somehow helpful.

On 31/05/2024 10:35, sibylle via grass-user wrote:

Dear community

I tested the r.boxplot addon and it is definitively possible to run the addon with more than 20 classes/regions/sites/squares.

Furthermore I realised the reason for the wrong assignment of class to raster, but I am not yet able to fix it.

1.	The challenge is, that (at the moment) I have raster data for one class (one square), but in total I have 150 squares. So at the moment it just selects the first square (either by cat oder ID_Quadrat).
2.	It seems however, if I have raster data for all squares, r.boxplot is working fine.
3.	My question: Is it possible to assign raster data for one individual class/squares, of you have missing data for other classes/squares?

Kind regards

Sibylle

First using this code here:

v.to.rast input=Untersuchungsquadrate_2020_2024 at PERMANENT output=Quadrate use=attr attribute_column="ID_Quadrat" label_column="ID_Quadrat"

r.boxplot -o input=U803174_abgeschnitten zones=Quadrate bx_sort=ascending raster_statistics=median

Second using this code here:

v.to.rast input=Untersuchungsquadrate_2020_2024 at PERMANENT output=Quadrate use=attr attribute_column=cat label_column="ID_Quadrat"

r.boxplot -o input=U803174_abgeschnitten zones=Quadrate bx_sort=ascending raster_statistics=median

From: sibylle.stoeckli at gmx.ch <mailto:sibylle.stoeckli at gmx.ch>   <mailto:sibylle.stoeckli at gmx.ch> <sibylle.stoeckli at gmx.ch> 
Sent: Thursday, May 30, 2024 9:27 AM
To: 'grass-user at lists.osgeo.org <mailto:grass-user at lists.osgeo.org> '  <mailto:grass-user at lists.osgeo.org> <grass-user at lists.osgeo.org>
Subject: RE: raster boxplots wrong assignment

Dear community

I am still wondering, if the r.boxplot addon is working with just a limited number of areass/sites/regions?

Kind regards

Sibylle

From: sibylle.stoeckli at gmx.ch <mailto:sibylle.stoeckli at gmx.ch>  <sibylle.stoeckli at gmx.ch <mailto:sibylle.stoeckli at gmx.ch> > 
Sent: Saturday, May 25, 2024 3:48 PM
To: 'grass-user at lists.osgeo.org <mailto:grass-user at lists.osgeo.org> ' <grass-user at lists.osgeo.org <mailto:grass-user at lists.osgeo.org> >
Subject: raster boxplots wrong assignment

Dear community

I would like to plot boxplots from a raster _pres_apple_spring_merged  for different 1 km2 squares (38 squares, Fig. 1, layer “Untersuchungsquadrate”). I was using the same code as used before for 12 regions:

https://ecodiv.earth/post/drawing-boxplots-of-raster-values/ (Author Paulo van Breugel). All layers have the same projection (EPSG:2056, CH1903+/LV95).  However, when loading the layer squares (Untersuchungsquadrate) GRASS GIS was asking for reprojection.

1.	I got now a warning message, that I have a lot of categories, however, the modelling time was fast (Fig 2)
2.	The error is now, that the boxplot is not showing all squares  (e.g. ID_Quadrat 803174) and the boxplots are assigned to the wrong square (ID_Quadrat) (Fig. 3)

Kind regards

Sibylle Stöckli

Fig. 1

Fig. 2

Figure 3

_______________________________________________
grass-user mailing list
grass-user at lists.osgeo.org <mailto:grass-user at lists.osgeo.org> 
https://lists.osgeo.org/mailman/listinfo/grass-user

-- 
Micha Silver
Ben Gurion Univ.
Sde Boker, Remote Sensing Lab
cell: +972-523-665918

-- 
Micha Silver
Ben Gurion Univ.
Sde Boker, Remote Sensing Lab
cell: +972-523-665918

-- 
Micha Silver
Ben Gurion Univ.
Sde Boker, Remote Sensing Lab
cell: +972-523-665918

-- 
Micha Silver
Ben Gurion Univ.
Sde Boker, Remote Sensing Lab
cell: +972-523-665918

-- 
Micha Silver
Remote Sensing Lab, Sde Boker
Ben Gurion University
+972-523-665918

-- 
Micha Silver
Ben Gurion Univ.
Sde Boker, Remote Sensing Lab
cell: +972-523-665918
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.osgeo.org/pipermail/grass-user/attachments/20240608/8c18e79c/attachment-0001.htm>
-------------- next part --------------
A non-text attachment was scrubbed...
Name: image001.png
Type: image/png
Size: 88282 bytes
Desc: not available
URL: <http://lists.osgeo.org/pipermail/grass-user/attachments/20240608/8c18e79c/attachment-0010.png>
-------------- next part --------------
A non-text attachment was scrubbed...
Name: image002.png
Type: image/png
Size: 29806 bytes
Desc: not available
URL: <http://lists.osgeo.org/pipermail/grass-user/attachments/20240608/8c18e79c/attachment-0011.png>
-------------- next part --------------
A non-text attachment was scrubbed...
Name: image003.png
Type: image/png
Size: 39093 bytes
Desc: not available
URL: <http://lists.osgeo.org/pipermail/grass-user/attachments/20240608/8c18e79c/attachment-0012.png>
-------------- next part --------------
A non-text attachment was scrubbed...
Name: image004.png
Type: image/png
Size: 37615 bytes
Desc: not available
URL: <http://lists.osgeo.org/pipermail/grass-user/attachments/20240608/8c18e79c/attachment-0013.png>
-------------- next part --------------
A non-text attachment was scrubbed...
Name: image005.png
Type: image/png
Size: 117720 bytes
Desc: not available
URL: <http://lists.osgeo.org/pipermail/grass-user/attachments/20240608/8c18e79c/attachment-0014.png>
-------------- next part --------------
A non-text attachment was scrubbed...
Name: image006.png
Type: image/png
Size: 128422 bytes
Desc: not available
URL: <http://lists.osgeo.org/pipermail/grass-user/attachments/20240608/8c18e79c/attachment-0015.png>
-------------- next part --------------
A non-text attachment was scrubbed...
Name: image007.png
Type: image/png
Size: 55410 bytes
Desc: not available
URL: <http://lists.osgeo.org/pipermail/grass-user/attachments/20240608/8c18e79c/attachment-0016.png>
-------------- next part --------------
A non-text attachment was scrubbed...
Name: image008.png
Type: image/png
Size: 42731 bytes
Desc: not available
URL: <http://lists.osgeo.org/pipermail/grass-user/attachments/20240608/8c18e79c/attachment-0017.png>
-------------- next part --------------
A non-text attachment was scrubbed...
Name: image009.png
Type: image/png
Size: 48805 bytes
Desc: not available
URL: <http://lists.osgeo.org/pipermail/grass-user/attachments/20240608/8c18e79c/attachment-0018.png>
-------------- next part --------------
A non-text attachment was scrubbed...
Name: al_pres_boxplot_3.png
Type: image/png
Size: 129245 bytes
Desc: not available
URL: <http://lists.osgeo.org/pipermail/grass-user/attachments/20240608/8c18e79c/attachment-0019.png>