PyGMT vs Plotly

Lounge
1

Hi there,

I have to switch to Python(3) using VSCode in Interactive mode for my work and I’m having quite a hard time with figures. I’ve worked on a small exercise in which I simply plot stations.

For this I got a geojson file with

index | on/off | name | number | geometry   | area
0     |   0    |  bla |   0    | POINT(x,y) | zone_1
1     |   0    |  ble |   1    | POINT(x,y) | zone_1
...
128   |   1    |  bly |   64   | POINT(x,y) | zone_8

So far, the cleanest code I have to display those stations is

import plotly as px
import geopandas as gpd

data = gpd.read_file(file)
fig=px.scatter(data,
		x=data["geometry"].values.x, 
		y=data["geometry"].values.y,
		color="area",
		symbol="on/off")

fig.show(renderer="browser") #for the external window

It’s nice, interactive and all … But I’m not satisfied because I don’t have a map in the background.

I’ve tried a little bit to play with folium, geopandas.plot and such… but it’s convoluted and not necessarily what I’m looking for.

ANYWAY ! Back to GMT, or should I say PyGMT !

  • First question:
    Is there a way to achieve what I’ve done just above without adding too many variables (I try to avoid `x=data[“this”] ; x2 = x[“that”] ; so on …)?

  • Second question:
    Can I add a background such as CartoDB positron or OpenStreetMap ?

What I’ve done with PyGMT is :

fig.basemap(region=region, 
			projection="M12c", 
			frame=["af", f'WSen+t"{title}"'])

fig.coast(land="gray", 
		  water="lightblue",
		  shorelines="thinner,black")

fig.plot(x=data2["geometry"].values.x,
		 y=data2["geometry"].values.y,
		 style="c0.3c",
		 color="red")

But of course, it’s not as nice like this.

Thanks,
Guillaume

2

You can plot images as background. You’d have to find the equivalent of the plot!() command in
https://www.generic-mapping-tools.org/GMTjl_doc/examples/art/art_examples/#example_14772517729074949607
(the one who fills the flower)

EDIT: But on a second thought. Why no plot the image with grdimage and overlay the lines on it?

3

Have you considered using Cartopy? I have some simple scripts with it and it works great. I usually use it for my “working” figures and then pygmt when it’s a figure for publication.

4

I am only starting to try PyGMT, but I think it knows how to plot data from a GeoPandas object without extracting the variables.

When I want to make maps with OpenStreetMap background, I use QGIS.

5

It seems to me that some non trivial steps would be involved to co-locate points no?

Looking into it right now.

In my case, that would implies to segregate the data… that’s ok. But assigning colors manually is a little too much… no ?

x_on  = data [geometry].x & [on]  > color [area = 1]
x_off = data [geometry].x & [off] > color [area = N]

That’s an efficient point and click solution indeed, but I need to be able to script it (+ learn some python).

6

Hadn’t understood that you want a referenced image. Bu no problem, see grdimge -Dr

7

Well, I’m doing a scatter plot of stations that I want to display over a map (with roads, streets,…).

Does using grdimage -Dr mean that my static map background needs to correspond exactly to the -R domain on which I plot my stations ?

EDIT : I actually managed with plotly and this line

fig.update_layout(mapbox_style="carto-positron")
8

It means that the -R that you pass to grdimage must be that of the image, not of your data points. But next you plot them with image’s -R

9

Makes sense, thanks :slight_smile:
I’ll play a little more with these

10

Try using contextily (Introduction guide to contextily — contextily 1.1.0 documentation). They have CartoDB.positron and OpenStreetMap listed at A short look into providers objects — contextily 1.1.0 documentation.

As an aside, I’ve actually been thinking whether it’s worth integrating PyGMT with contextily so that users can get a choice of different basemaps. They have a bounds2img function that returns an RGBA numpy array, which we could georeference into an xarray.DataArray and possibly use as a basemap.

11

Thanks @weiji14. I’ll have a look at it.
To be honest, for simple purposes QGIS or python MAPBOX tools seem more compact to code with the bonus of having interacting vizualisation.