[GRASS-user] How to implement 2D lookup table?

Rainer_Krug · May 23, 2013, 8:26am

Hi

I have two raster layers

,----
| dc: CELL, 1, ..., 7, NULL
| ageClass: CELL, 1,2,3,NULL
`----

And I want to create a third one, depending on the value in a lookup table. dc corresponds to the row, ageClass to the column,

The lookup table looks as follow:

,----
| seedling young adult
| dc1 425000 87500 37500
| dc2 270000 55000 23500
| dc3 127500 26000 11250
| dc4 32500 7100 3200
| dc5 5550 1220 500
| dc6 605 132 55
| dc7 56 12 6
`----

Now I want to have something like (using r.mapcalc as an example)

,----
| r.mapcalc expression = "result = lookupTable[col=ageClass, row=dc]"
`----

and e.g. for one cell

,----
| ageClass = 3
| dc = 5
| ==>
| result = 500
`----

I haven't found anything like this in GRASS.

I could use MASKS and then r.reclass(), or generate dynamically a
complex if() construct, but both options sound a little bit
awkward to me.

I am scripting GRASS from R, so there is always the option of doing it
in R, but there is the issue of reading the data into R and writing it
back to GRASS, which takes time. As I am doing this in a simulation
which runs several times, this is not an unimportant issue.

Is there better way of achieving this?

Thanks,

Rainer

--
Rainer M. Krug

Centre of Excellence for Invasion Biology
Stellenbosch University
South Africa

Tel : +33 - (0)9 53 10 27 44
Cell: +33 - (0)6 85 62 59 98
Fax : +33 - (0)9 58 10 27 44

Fax (D): +49 - (0)3 21 21 25 22 44

email: Rainer@krugs.de

Skype: RMkrug

mlennert · May 23, 2013, 8:43am

On 23/05/13 10:26, Rainer M Krug wrote:

Hi

I have two raster layers

,----
| dc: CELL, 1, ..., 7, NULL
| ageClass: CELL, 1,2,3,NULL
`----

And I want to create a third one, depending on the value in a lookup table. dc corresponds to the row, ageClass to the column,

The lookup table looks as follow:

,----
| seedling young adult
| dc1 425000 87500 37500
| dc2 270000 55000 23500
| dc3 127500 26000 11250
| dc4 32500 7100 3200
| dc5 5550 1220 500
| dc6 605 132 55
| dc7 56 12 6
`----

Now I want to have something like (using r.mapcalc as an example)

,----
| r.mapcalc expression = "result = lookupTable[col=ageClass, row=dc]"
`----

and e.g. for one cell

,----
| ageClass = 3
| dc = 5
| ==>
| result = 500
`----

I haven't found anything like this in GRASS.

I could use MASKS and then r.reclass(), or generate dynamically a
complex if() construct, but both options sound a little bit
awkward to me.

I am scripting GRASS from R, so there is always the option of doing it
in R, but there is the issue of reading the data into R and writing it
back to GRASS, which takes time. As I am doing this in a simulation
which runs several times, this is not an unimportant issue.

Is there better way of achieving this?

I don't see a possibility to include the lookup table into r.mapcalc, but anyway, IIUC, the result you are looking for is not a raster map, but a numerical value, or ?

My best first guess is:

r.stats -c in=dc,ageClass

possibly combined with a little awk to do the lookup:

r.stats -c in=dc,ageClass | awk '{if($1==5 && $2==3) print $3}'

Or, if the maps are heavy, feed the data resulting from the r.stats call into a new db table and then use sql queries to do the lookup. Or feed the results of r.stats into an R dataframe.

Moritz

Rainer_Krug · May 23, 2013, 9:04am

Moritz Lennert <mlennert@club.worldonline.be> writes:

On 23/05/13 10:26, Rainer M Krug wrote:

Hi

I have two raster layers

,----
| dc: CELL, 1, ..., 7, NULL
| ageClass: CELL, 1,2,3,NULL
`----

And I want to create a third one, depending on the value in a lookup
table. dc corresponds to the row, ageClass to the column,

The lookup table looks as follow:

,----
| seedling young adult
| dc1 425000 87500 37500
| dc2 270000 55000 23500
| dc3 127500 26000 11250
| dc4 32500 7100 3200
| dc5 5550 1220 500
| dc6 605 132 55
| dc7 56 12 6
`----

Now I want to have something like (using r.mapcalc as an example)

,----
| r.mapcalc expression = "result = lookupTable[col=ageClass, row=dc]"
`----

and e.g. for one cell

,----
| ageClass = 3
| dc = 5
| ==>
| result = 500
`----

I haven't found anything like this in GRASS.

I could use MASKS and then r.reclass(), or generate dynamically a
complex if() construct, but both options sound a little bit
awkward to me.

I am scripting GRASS from R, so there is always the option of doing it
in R, but there is the issue of reading the data into R and writing it
back to GRASS, which takes time. As I am doing this in a simulation
which runs several times, this is not an unimportant issue.

Is there better way of achieving this?

I don't see a possibility to include the lookup table into r.mapcalc,

Me neither - that was just the easies to show what I want to do.

but anyway, IIUC, the result you are looking for is not a raster map,
but a numerical value, or ?

No - the lookup would be for each cell, therefore resulting in a CELL
raster map.

My best first guess is:

r.stats -c in=dc,ageClass

possibly combined with a little awk to do the lookup:

r.stats -c in=dc,ageClass | awk '{if($1==5 && $2==3) print $3}'

Or, if the maps are heavy, feed the data resulting from the r.stats
call into a new db table and then use sql queries to do the lookup. Or
feed the results of r.stats into an R dataframe.

Hmm - sounds interesting, but also for other purposes. But as the result
will be a raster map, I don't think this approach is usable here (unless
I can feed the numbers back into a raster map, i.e. the inverse of
r.stats -c)

Rainer

Moritz

--
Rainer M. Krug

mlennert · May 23, 2013, 12:14pm

On 23/05/13 11:04, Rainer M. Krug wrote:

Moritz Lennert<mlennert@club.worldonline.be> writes:

On 23/05/13 10:26, Rainer M Krug wrote:

Hi

I have two raster layers

,----
| dc: CELL, 1, ..., 7, NULL
| ageClass: CELL, 1,2,3,NULL
`----

And I want to create a third one, depending on the value in a lookup
table. dc corresponds to the row, ageClass to the column,

The lookup table looks as follow:

,----
| seedling young adult
| dc1 425000 87500 37500
| dc2 270000 55000 23500
| dc3 127500 26000 11250
| dc4 32500 7100 3200
| dc5 5550 1220 500
| dc6 605 132 55
| dc7 56 12 6
`----

Now I want to have something like (using r.mapcalc as an example)

,----
| r.mapcalc expression = "result = lookupTable[col=ageClass, row=dc]"
`----

and e.g. for one cell

,----
| ageClass = 3
| dc = 5
| ==>
| result = 500
`----

I haven't found anything like this in GRASS.

I could use MASKS and then r.reclass(), or generate dynamically a
complex if() construct, but both options sound a little bit
awkward to me.

I am scripting GRASS from R, so there is always the option of doing it
in R, but there is the issue of reading the data into R and writing it
back to GRASS, which takes time. As I am doing this in a simulation
which runs several times, this is not an unimportant issue.

Is there better way of achieving this?

I don't see a possibility to include the lookup table into r.mapcalc,

Me neither - that was just the easies to show what I want to do.

but anyway, IIUC, the result you are looking for is not a raster map,
but a numerical value, or ?

No - the lookup would be for each cell, therefore resulting in a CELL
raster map.

My best first guess is:

r.stats -c in=dc,ageClass

possibly combined with a little awk to do the lookup:

r.stats -c in=dc,ageClass | awk '{if($1==5&& $2==3) print $3}'

Or, if the maps are heavy, feed the data resulting from the r.stats
call into a new db table and then use sql queries to do the lookup. Or
feed the results of r.stats into an R dataframe.

Hmm - sounds interesting, but also for other purposes. But as the result
will be a raster map, I don't think this approach is usable here (unless
I can feed the numbers back into a raster map, i.e. the inverse of
r.stats -c)

r.cross + r.reclass ?

moritz

Rainer_Krug · May 23, 2013, 1:43pm

Moritz Lennert <mlennert@club.worldonline.be> writes:

On 23/05/13 11:04, Rainer M. Krug wrote:

Moritz Lennert<mlennert@club.worldonline.be> writes:

On 23/05/13 10:26, Rainer M Krug wrote:

Hi

I have two raster layers

,----
| dc: CELL, 1, ..., 7, NULL
| ageClass: CELL, 1,2,3,NULL
`----

And I want to create a third one, depending on the value in a lookup
table. dc corresponds to the row, ageClass to the column,

The lookup table looks as follow:

,----
| seedling young adult
| dc1 425000 87500 37500
| dc2 270000 55000 23500
| dc3 127500 26000 11250
| dc4 32500 7100 3200
| dc5 5550 1220 500
| dc6 605 132 55
| dc7 56 12 6
`----

Now I want to have something like (using r.mapcalc as an example)

,----
| r.mapcalc expression = "result = lookupTable[col=ageClass, row=dc]"
`----

and e.g. for one cell

,----
| ageClass = 3
| dc = 5
| ==>
| result = 500
`----

I haven't found anything like this in GRASS.

I could use MASKS and then r.reclass(), or generate dynamically a
complex if() construct, but both options sound a little bit
awkward to me.

I am scripting GRASS from R, so there is always the option of doing it
in R, but there is the issue of reading the data into R and writing it
back to GRASS, which takes time. As I am doing this in a simulation
which runs several times, this is not an unimportant issue.

Is there better way of achieving this?

I don't see a possibility to include the lookup table into r.mapcalc,

Me neither - that was just the easies to show what I want to do.

but anyway, IIUC, the result you are looking for is not a raster map,
but a numerical value, or ?

No - the lookup would be for each cell, therefore resulting in a CELL
raster map.

My best first guess is:

r.stats -c in=dc,ageClass

possibly combined with a little awk to do the lookup:

r.stats -c in=dc,ageClass | awk '{if($1==5&& $2==3) print $3}'

Or, if the maps are heavy, feed the data resulting from the r.stats
call into a new db table and then use sql queries to do the lookup. Or
feed the results of r.stats into an R dataframe.

Hmm - sounds interesting, but also for other purposes. But as the result
will be a raster map, I don't think this approach is usable here (unless
I can feed the numbers back into a raster map, i.e. the inverse of
r.stats -c)

r.cross + r.reclass ?

There is one aspect in r.cross which makes this approach quite
difficult to implement: the value of the result map layer can not easily
be traced back to the values of the two input layers:

,----
| Note: The actual category value assigned to a particular combination in
| the result map layer is dependent on the order in which the combinations
| occur in the input map layer data and can be considered essentially
| random. The example given here is illustrative only.
`----

But this gave ma an idea: As I know the *theoretical* maximum values of
each map (as defined in the lookup table), I can do the following:

,----
| r.mapcalc expression="lookup = dc + (ageClass-1)*7"
`----

and flatten in the same way the lookup table
and convert it to reclass
rules:

Now I can use

,----
| r.reclass input=lookup output=finalLookup rules=TheRulesAsAbove
| r.mapcalc "finalLookup = finalLookup"
| r.remove rast=lookup
`----

and then I should have it.

Thanks a lot,

Rainer

moritz
_______________________________________________
grass-user mailing list
grass-user@lists.osgeo.org
http://lists.osgeo.org/mailman/listinfo/grass-user

--
Rainer M. Krug

Vincent_Bain · May 23, 2013, 1:47pm

>>> ,----
>>> | r.mapcalc expression = "result = lookupTable[col=ageClass,
row=dc]"
>>> `----

Not a mapcalc expert, but what about turning your lookup table in a true
raster, I mean create a 3colsx7rows LUT raster with values from your
table, then use this "map" in an r.mapcalc expression, calling LUT[i,j]
to retrieve the value corresponding to ageClass i and dc j
Sorry to interfere if my proposition is stupid

Vincent

Rainer_Krug · May 23, 2013, 2:21pm

Vincent Bain <bain@toraval.fr> writes:

>>> ,----
>>> | r.mapcalc expression = "result = lookupTable[col=ageClass,
row=dc]"
>>> `----

Not a mapcalc expert, but what about turning your lookup table in a true
raster, I mean create a 3colsx7rows LUT raster with values from your
table, then use this "map" in an r.mapcalc expression, calling LUT[i,j]
to retrieve the value corresponding to ageClass i and dc j
Sorry to interfere if my proposition is stupid

Not at all this sounds interesting. But could you give some more details
on how I could do this? and what do you mean by "calling LUT[i,j]"?

Cheers,

Rainer

Vincent

_______________________________________________
grass-user mailing list
grass-user@lists.osgeo.org
http://lists.osgeo.org/mailman/listinfo/grass-user

--
Rainer M. Krug

Vincent_Bain · May 23, 2013, 2:47pm

Not at all this sounds interesting. But could you give some more details
on how I could do this? and what do you mean by "calling LUT[i,j]"?

Rainer,
I fear my solution is hopeless.
I first reminded in r.mapcalc algebra, one could point at a particular
(row,col) position within a raster. But I was wrong. The bracketed
notation is intended to be a so-called "neighborhood modifier" relative
to a given cell : e.g. map[-2,-1] refers to the cell two rows above and
one column to the left of the current cell (within a single map or
across multiple maps).

In a nutshell: it was a stupid proposal!

Bye,
Vincent.

Rainer_Krug · May 23, 2013, 2:57pm

Vincent Bain <bain@toraval.fr> writes:

Not at all this sounds interesting. But could you give some more details
on how I could do this? and what do you mean by "calling LUT[i,j]"?

Rainer,
I fear my solution is hopeless.
I first reminded in r.mapcalc algebra, one could point at a particular
(row,col) position within a raster. But I was wrong. The bracketed
notation is intended to be a so-called "neighborhood modifier" relative
to a given cell : e.g. map[-2,-1] refers to the cell two rows above and
one column to the left of the current cell (within a single map or
across multiple maps).

In a nutshell: it was a stupid proposal!

Not stupid - probably not workable here, but not stupid.

Cheers,

Rainer

Bye,
Vincent.

_______________________________________________
grass-user mailing list
grass-user@lists.osgeo.org
http://lists.osgeo.org/mailman/listinfo/grass-user

--
Rainer M. Krug

Glynn_Clements1 · May 23, 2013, 11:37pm

Rainer M. Krug wrote:

>> Hmm - sounds interesting, but also for other purposes. But as the result
>> will be a raster map, I don't think this approach is usable here (unless
>> I can feed the numbers back into a raster map, i.e. the inverse of
>> r.stats -c)
>
> r.cross + r.reclass ?

There is one aspect in r.cross which makes this approach quite
difficult to implement: the value of the result map layer can not easily
be traced back to the values of the two input layers:

If you only have a small number of classes, you can just create the
cross-product manually, e.g.:

r.mapcalc 'result = ageClass * 10 + dc'

So a category of 15 would be ageClass=1, dc=5.

r.cross is more useful when the total number of pairs which actually
occur is much fewer than the product of the numbers of categories, as
it only creates a category for each combination which actually occurs,
not for each potential combination.

--
Glynn Clements <glynn@gclements.plus.com>

Rainer_Krug · May 24, 2013, 7:11am

On Friday, May 24, 2013, Glynn Clements wrote:

Rainer M. Krug wrote:

Hmm - sounds interesting, but also for other purposes. But as the result
will be a raster map, I don’t think this approach is usable here (unless
I can feed the numbers back into a raster map, i.e. the inverse of
r.stats -c)

r.cross + r.reclass ?

There is one aspect in r.cross which makes this approach quite
difficult to implement: the value of the result map layer can not easily
be traced back to the values of the two input layers:

If you only have a small number of classes, you can just create the
cross-product manually, e.g.:

r.mapcalc ‘result = ageClass * 10 + dc’

Yes - This is effectively what I am doing now.

So a category of 15 would be ageClass=1, dc=5.

r.cross is more useful when the total number of pairs which actually
occur is much fewer than the product of the numbers of categories, as
it only creates a category for each combination which actually occurs,
not for each potential combination.

True. The problem I have with r.cross is that I have to manually decode afterwords what combination the value refers to. I could use a she’ll script to translate the category values, but this sounds a little tedious, especially as the format is not that easily parsable.

Cheers,

Rainer

–
Glynn Clements glynn@gclements.plus.com

–
NEW GERMAN FAX NUMBER!!!

Rainer M. Krug, PhD (Conservation Ecology, SUN), MSc (Conservation Biology, UCT), Dipl. Phys. (Germany)

Centre of Excellence for Invasion Biology
Natural Sciences Building
Office Suite 2039
Stellenbosch University
Main Campus, Merriman Avenue
Stellenbosch
South Africa

Cell: +27 - (0)83 9479 042
Fax: +27 - (0)86 516 2782
Fax: +49 - (0)321 2125 2244
email: Rainer@krugs.de

Skype: RMkrug
Google: R.M.Krug@gmail.com