- Last 7 days
-
notebooksharing.space notebooksharing.spaceNotebook26
-
xr.DataArray from 'cities' gpd.GeoDataFra
Delete.
-
it
Mention and link to Weatherbench2 here.
-
a cliamte model
replace climate with atmosphere in this paragraph. ERA5 is an atmospheric reanalysis.
-
are excited to announce the extension of the Xarray data model to support vector geometries
The announcement has already happened with Xvec :) . Instead I would phrase as "describing how the Xarray ecosystem supports vector geometries".
Also I'd mention xvec in this first paragraph.
-
(geomet
Can you install
jupyterlab-code-formatterandruffand/orblack. It'll auto-format these cells to multiple lines for you. -
era5_europe_cities['time'].dt.season
"time.season" -
Convert Xarray objects to geopandas GeoDataFrame
Plotting
-
Spatial indexing
This is quite powerful, I'd add a bit about
.xvec.queryin the introduction.AND here, you are showing how the geometry dimension is special, and enables nice interfaces
-
Computation and grouping along a time dimension
The time dimension is a speicfic example here. More importantly, you can do the "usual" multi-dimensional array things along the non-geometry dimensions. I would make that point.
-
where(era5_eu
.idxmax() -
Sample raster data cube with geometries from vector data cube
"Creating a vector data cube from a raster data cube"
-
assign_coords({'city':era5_europe_cities['city'], 'country':era5_europe_cities['country'] })
set_coords(["city", "country"]) -
The above operation interpolated the ERA5 data onto the coordinates from europe_ds but in the process we lost the data variables describing the name and country of each city. Add those onto the interpolated vector data cube and drop the level coordinate variable, which we don't need.
If you set them as coords, I bet you wouldn't need this.
-
city(geometry)
I would also rename geometry -> city, and city->name maybe?
-
lat(geometry)float6441.33 41.32 41.11 ... 50.18 50.57array([41.3275 , 41.3230556, 41.1125 , ..., 50.75 , 50.1833333, 50.5666667])lon
lat, lon are redundant with "geometry", I'd drop them to illustrate your points better
-
load_dataset
I couldn't run this notebook because I didn't have this function
-
points
To me, Polygons of interest are where things get interesting.
Points are reasonably handled without these geometries. But a polygon is a meaningful increase in metadata associated with an observation.
-
Vector datasets are frequently treated as 'flat' or where the spatial dimension is the only required functional dimension; but what happens when vector datasets contain additional dimensions like time?
nice
-
What about situations
I would consider illustrating the poiints here with a 5 row DataFrame.
-
T
"In memory, geometries are commonly represented as Shapely geometries"
-
shapely geometry objects.
This column can also be the index no?
-
data frames
Use "table" instead of "data frame"
-
raster data is viewed as a cube, while vector data is discussed as a data frame.
When you use this kind of construction, always use the same verb i.e. "viewed", and the same ordering. For example - BAD: "A is better than B, while D is worse than C" - GOOD: "A is better than B, and C is better than D".
When you use the same ordering and minimize differences, it becomes a lot easier for the brain to parse.
-
Vector data cubes
You'll need to explain what a vector data cube is up top.
-
, pushing the limitations of existing tools for working with these data types
Long sentence, I would delete this.
-
:tada
Should check that these will render properly. Usually you need
:tada:
-
- Jun 2024
-
notebooksharing.space notebooksharing.spaceNotebook1
-
idx_variables.update(index.create_variables(variables))
indexis a tuple, but should be PandasIndex. You've copied too much code over :) Write it from scratch.
-
- Jul 2022
-
notebooksharing.space notebooksharing.spaceNotebook7
-
these are the resources I've produced or contributed to this summer,
I agree with moving this up. This is your main "product" IMO. And to write these you developed the skills in (1)
It'd also be nice to mention the NASA project, so it shows that your work is feeding in to a bigger effort in the area.
-
what's next
Could consider linking up with ITS_LIVE analysis software development efforts (just a thought)
-
contribution
I would group c,d,e under "Contributing directly to a vibrant open-source project" or something like that.
-
working with xarray,
and "designing analysis pipelines for nD array datasets".
Usually the word thrown around for "using less for loops" is "vectorization" but that's a bit too much jargon.
-
raster format,
Need to explain this too.
-
remote sensing
Good to start with "what is remote sensing data"
-
specialized knowledge is passed down from advisor to advisee, or in small university classes
Nice place to bring in a personal story if you have one.
-