extract returns

[mmann1123]

Running gw.extract with polygons the returned dataframe has

id | point | geometry | index

Just wondering how these are generated. None seems related to the polygon index. Maybe worth adding to docs?

[jgrss]

Good idea.

The id column is a unique numerical identifier for each polygon and is determined by the id_column parameter in extract(), point is a unique identifier for each point within each polygon, geometry is the shapely.geometry auto-generated by GeoPandas, and index must be from your polygon file.

Using some of the built-in data

from geowombat.data import l8_224078_20200518_polygons, l8_224078_20200518_B4
import geowombat as gw
import geopandas as gpd

The polygons

These will be converted to points at the native cell resolution. If id_column is not specified or the the column values are strings, then the extract function will generate a unique numerical identifier.

gpd.read_file(l8_224078_20200518_polygons).head()

        name                                           geometry
0      water  POLYGON ((737544.502 -2795232.772, 737544.502 ...
1       crop  POLYGON ((742517.658 -2798160.232, 743046.717 ...
2       tree  POLYGON ((742435.360 -2801875.403, 742458.874 ...
3  developed  POLYGON ((738903.667 -2811573.845, 738926.586 ...

The extract function

The id column was generated from the polygon name column. The points are all unique, but they can be grouped by the polygon using the id column. The original polygon column name (name) is appended. I didn't specify a band name, so 1 is the raster band default.

with gw.open(l8_224078_20200518_B4) as src:
    df = gw.extract(src, l8_224078_20200518_polygons)
    print(df.head())

   id  point                         geometry   name     1
0   0      0  POINT (737535.000 -2795205.000)  water  6283
1   0      1  POINT (737565.000 -2795205.000)  water  6294
2   0      2  POINT (737595.000 -2795205.000)  water  6295
3   0      3  POINT (737625.000 -2795205.000)  water  6284
4   0      4  POINT (737655.000 -2795205.000)  water  6266

[jgrss]

Following up on the previous comment...

The id column should be a column of integers. If id_column is specified and the column has strings, extract() should fail. If id_column is not specified (the default is id_column='id') and the column doesn't exist, it will be created from the index values by:

if id_column not in df.columns.tolist():
    df.loc[:, id_column] = df.index.values

If you want your own unique ids for each polygon, you could do:

from geowombat.data import l8_224078_20200518_polygons
import geopandas as gpd
from sklearn.preprocessing import LabelEncoder

le = LabelEncoder()

df_poly = gpd.read_file(l8_224078_20200518_polygons)
df_poly['unique'] = le.fit_transform(df_poly['name'])

gw.extract(src, df_poly, id_column='unique')

[mmann1123]

Ok great. Thanks. Oddly the index didn't match my polygon index, odd... I'll play around some more.

…

On Tue, Feb 2, 2021 at 8:19 PM Jordan ***@***.***> wrote: Following up on the previous comment... The id column should be a column of integers. If id_column is specified and the column has strings, extract() should fail. If id_column is not specified (the default is id_column='id') and the column doesn't exist, it will be created from the index values by: if id_column not in df.columns.tolist(): df.loc[:, id_column] = df.index.values If you want your own unique ids for each polygon, you could do: from geowombat.data import l8_224078_20200518_polygonsimport geopandas as gpdfrom sklearn.preprocessing import LabelEncoder le = LabelEncoder() df_poly = gpd.read_file(l8_224078_20200518_polygons)df_poly['unique'] = le.fit_transform(df_poly['name']) gw.extract(src, df_poly, id_column='unique') — You are receiving this because you authored the thread. Reply to this email directly, view it on GitHub <#66 (reply in thread)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/ABHR6VE4R7CXYUY4ONE67I3S5CQBPANCNFSM4W63CTOA> .

[jgrss]

Do you have an example? If you didn't provide the id_column value then it could be that your polygon index is not unique. ??