!kSAxcBUchQDkBtKcuV:matrix.org

Fatiando a Terra - verde

19 Members
+fatiando:matrix.org1 Servers

Load older messages


SenderMessageTime
20 Nov 2020
@_slack_fatiando_U01F18B993Q:matrix.orgMat Sadowskiimage.png
Download image.png
18:00:31
@_slack_fatiando_U01F18B993Q:matrix.orgMat Sadowski I'm wondering if anyone would have some pointers for me. I'm trying to use Verde on a very small scale project compared to the data you are processing in the tutorials (surveying lakes/ponds). Scipygridder seemed to have quite nicely interpolated the points without going into the free space (https://www.fatiando.org/verde/latest/gallery/scipygridder.html). When I tried the same approach on my data I have the interpolation is quite far away from the surveyed data (see the screenshot). Does anybody know if I can easily tune this? Is this an issue of scale? 18:00:31
@_slack_fatiando_UMSRSPEMA:matrix.orgleouieda
In reply to@_slack_fatiando_U01E7LSJCHH:matrix.org
So, from the JOSS article, this would be the difference, right? "Some of these interpolationand data processing methods already exist in GMT ... . However, there are no model selection tools in GMT and it can be difficult to separate parts of the processing that are done internally by its modules. Verde is designed to be modular, easily extended, ... .It can be used to implement new interpolation methods".
Exactly. We are missing some options from GMT’s greenspline still, like the splines in tension. But those are relatively easy to add and we’ve been focusing on the bigger picture of the API for Verde v1.*. Now we’re getting performance down and making a viable alternative to surface which is still the best gridder out there in my opinion.
18:02:44
@_slack_fatiando_UMSRSPEMA:matrix.orgleouieda
In reply to@_slack_fatiando_U01F18B993Q:matrix.org
I'm wondering if anyone would have some pointers for me. I'm trying to use Verde on a very small scale project compared to the data you are processing in the tutorials (surveying lakes/ponds). Scipygridder seemed to have quite nicely interpolated the points without going into the free space (https://www.fatiando.org/verde/latest/gallery/scipygridder.html). When I tried the same approach on my data I have the interpolation is quite far away from the surveyed data (see the screenshot). Does anybody know if I can easily tune this? Is this an issue of scale?
Scipy applies something like vd.convexhull_mask by default, it will interpolate inside of the convex hull formed by the data. You might want to try using vd.Spline and then masking with vd.distance_mask instead.
18:19:03
@_slack_fatiando_UMSRSPEMA:matrix.orgleouieda
In reply to@_slack_fatiando_UMSRSPEMA:matrix.org
Scipy applies something like vd.convexhull_mask by default, it will interpolate inside of the convex hull formed by the data. You might want to try using vd.Spline and then masking with vd.distance_mask instead.
Or you could also apply distance_mask on top of the Scipy results.
18:19:56
@_slack_fatiando_U01F18B993Q:matrix.orgMat Sadowski
In reply to@_slack_fatiando_UMSRSPEMA:matrix.org
Or you could also apply distance_mask on top of the Scipy results.
thanks! I will give this a go 🙂
18:38:24
21 Nov 2020
@_slack_fatiando_U01E7LSJCHH:matrix.orgFederico Esteban
In reply to@_slack_fatiando_UMSRSPEMA:matrix.org
Exactly. We are missing some options from GMT’s greenspline still, like the splines in tension. But those are relatively easy to add and we’ve been focusing on the bigger picture of the API for Verde v1.*. Now we’re getting performance down and making a viable alternative to surface which is still the best gridder out there in my opinion.
Do you have any material where I can undestarnd the idea behind Green function without many formulas? I am not a (geo)physicists so it is hard for me to understand the idea from the fornulas.
01:47:00
@_slack_fatiando_UMSRSPEMA:matrix.orgleouieda
In reply toundefined
Don’t really have many resources without maths. We should include this in the Verde docs actually. The closest I could find are these slides by santisoler https://santisoler.github.io/talks/egu2020.html it’s a similar idea but for gravity and magnetics
09:23:58
@_slack_fatiando_UMSRSPEMA:matrix.orgleouieda
In reply toundefined
(edited) ... and magnética => ... and magnetics
09:24:11
2 Dec 2020
@_slack_fatiando_U01G12N5SAW:matrix.orgStavros joined the room.21:35:17
@_slack_fatiando_U01G12N5SAW:matrix.orgStavros changed their profile picture.21:35:18
7 Dec 2020
@_slack_fatiando_U01E7LSJCHH:matrix.orgFederico Esteban Hi. I am trying to understand the remove-compute-restore method that I see in verde Tutorial (in Transform 2020). Thanks in advance. 18:38:09
8 Dec 2020
@_slack_fatiando_UMSRSPEMA:matrix.orgleouieda
In reply to@_slack_fatiando_U01E7LSJCHH:matrix.org
Hi. I am trying to understand the remove-compute-restore method that I see in verde Tutorial (in Transform 2020). Thanks in advance.
Federico Esteban anything in particular?
08:34:55
@_slack_fatiando_U01E7LSJCHH:matrix.orgFederico Esteban
In reply to@_slack_fatiando_UMSRSPEMA:matrix.org
Federico Esteban anything in particular?
Just the idea used in verde. With some graph/animations would be ideal.
15:35:01
@_slack_fatiando_U0156QCM6AH:matrix.orgRichard Scott Here's a possible Verde public dataset http://auspass.edu.au/research/AusMoho.html 23:48:46
9 Dec 2020
@_slack_fatiando_UMSRSPEMA:matrix.orgleouieda
In reply to@_slack_fatiando_U0156QCM6AH:matrix.org
Here's a possible Verde public dataset http://auspass.edu.au/research/AusMoho.html
Awesome! Thanks for sharing 👍 That’ll be a good one for testing the new spherical interpolations a student is currently implementing
08:01:29
4 Jan 2021
@_slack_fatiando_UT0BDHKPE:matrix.orgLorenzo Hi all, i wonder if there is a way to BlockReduce categorical data. I.E. i would like to decimate a dataset, but maintin their categorical features…. 15:06:39
@_slack_fatiando_UMGLPTLAW:matrix.orgCraig Miller
In reply to@_slack_fatiando_UT0BDHKPE:matrix.org
Hi all, i wonder if there is a way to BlockReduce categorical data. I.E. i would like to decimate a dataset, but maintin their categorical features….
Perhaps run through pandas groupby then blockreduxe?
18:22:12
5 Jan 2021
@_slack_fatiando_UMSRSPEMA:matrix.orgleouieda
In reply toundefined
I guess it would depend on the reduction function. If it can handle categorical data, then BlockReduce with center_coordinates=True should work. Please let us know if it does actually. We’ve never had a use case for this.
07:13:55
@_slack_fatiando_UMSRSPEMA:matrix.orgleouieda
In reply toundefined
(edited) ... with `center_coordinates=True` => ... with center_coordinates=True should work. Please let us know if it does actually. We’ve never had a use case for this.
07:15:03
@_slack_fatiando_UT0BDHKPE:matrix.orgLorenzo
In reply to@_slack_fatiando_UMSRSPEMA:matrix.org
I guess it would depend on the reduction function. If it can handle categorical data, then BlockReduce with center_coordinates=True should work. Please let us know if it does actually. We’ve never had a use case for this.
Many thanks, I will try these options and let know
07:41:11
@_slack_fatiando_UMSRSPEMA:matrix.orgleouieda
In reply to@_slack_fatiando_UT0BDHKPE:matrix.org
Many thanks, I will try these options and let know
Out of curiosity, what kind of reduction do you want to do on categorical data?
07:45:01
@_slack_fatiando_UT0BDHKPE:matrix.orgLorenzo
In reply toundefined
actually my problem is not really a reduction, I just want to decimate data. I convert a lines in vertices (it represent a seismic line), but instead of having a vertices every 10 meters I want to reduce the vertices to a 50 m spacing. And I want to keep all the attributes (columns of the dataframe) associate with the vertices (that contains some categoricals). Since I use Blockreduce with success in the past, but never with categorical variables I wonder if there is an option. Honestly I’m not sure this is the best approach to do that…
07:49:44
@_slack_fatiando_UT0BDHKPE:matrix.orgLorenzo
In reply toundefined
(edited) ... the certices (that ... => ... the vertices (that ...
07:50:28
@_slack_fatiando_UMSRSPEMA:matrix.orgleouieda
In reply to@_slack_fatiando_UT0BDHKPE:matrix.org
actually my problem is not really a reduction, I just want to decimate data. I convert a lines in vertices (it represent a seismic line), but instead of having a vertices every 10 meters I want to reduce the vertices to a 50 m spacing. And I want to keep all the attributes (columns of the dataframe) associate with the vertices (that contains some categoricals). Since I use Blockreduce with success in the past, but never with categorical variables I wonder if there is an option. Honestly I’m not sure this is the best approach to do that…
Right, that seems reasonable. What you can do is pass in a custom function to BlockReduce that preserves the categories the way you want it to. BlockReduce takes care of assigning coordinates to blocks and passing in the data values that fall on each block to your function.
13:28:27
24 Feb 2021
@_slack_fatiando_U01PJ0DJXRP:matrix.orgSina Balkhi joined the room.07:51:14
23 Mar 2021
@_slack_fatiando_U01S09PTB1U:matrix.orgmassimo di stefano joined the room.20:01:59
@_slack_fatiando_U01S09PTB1U:matrix.orgmassimo di stefano changed their profile picture.20:02:01
24 Apr 2021
@_slack_fatiando_U01UXA6PHT9:matrix.orgmycarta joined the room.19:50:43
30 Apr 2021
@_slack_fatiando_U020SC29PB6:matrix.orgMiguel Alejandro Alzate Gutierrez joined the room.17:50:34

There are no newer messages yet.


Back to Room List