How to do a union rolling window on a GeoDataFrame

Question

I want to get a new GeoDataFrame in which each element is the union of N consecutive elements in my original GeoDataFrame. I'd be very happy if this code worked, but it doesn't: the last step fails.

from shapely.geometry import Polygon
from shapely.ops import unary_union
import geopandas as gpd
p1 = Polygon([(1,1),(1,2),(2,2)])
p2 = Polygon([(1,1),(2,2),(2,1)])
p3 = Polygon([(1,1),(2,1),(2,0)])
gdf = gpd.GeoDataFrame(geometry=[p1, p2, p3])
gdf.rolling(window=2).apply(unary_union)

I would expect something that behaves like this

import pandas as pd
df = pd.DataFrame({'numbers': [1,2,3]})
df.rolling(window=2).apply(sum)
Out[1]: 
   numbers
0      NaN
1      3.0
2      5.0

score 1 · Accepted Answer · answered Jun 08 '23 at 05:21

You can solve it like this:

from shapely.geometry import Polygon
from shapely.ops import unary_union
import geopandas as gpd
import numpy as np
p1 = Polygon([(1,1),(1,2),(2,2)])
p2 = Polygon([(1,1),(2,2),(2,1)])
p3 = Polygon([(1,1),(2,1),(2,0)])
gdf = gpd.GeoDataFrame(geometry=[p1, p2, p3])
#gdf.rolling(window=2).apply(unary_union)
#DataError: Cannot aggregate non-numeric type: geometry
rolling_window = gdf.rolling(window=2)
#Create a list of lists with geometry pairs in each sublist
geometry_pairs = [[*l.geometry] for l in rolling_window if len(l) == 2] #https://stackoverflow.com/questions/70670079/get-indexes-of-pandas-rolling-window
unions = [unary_union(pair) for pair in geometry_pairs]
unions.insert(0, np.nan)
[nan,
<POLYGON ((1 2, 2 2, 2 1, 1 1, 1 2))>,
<POLYGON ((2 0, 1 1, 2 2, 2 1, 2 0))>]

How to do a union rolling window on a GeoDataFrame

1 Answers1

[nan,

<POLYGON ((1 2, 2 2, 2 1, 1 1, 1 2))>,

<POLYGON ((2 0, 1 1, 2 2, 2 1, 2 0))>]