A spatial join uses binary predicates
such as intersects
and crosses
to combine two GeoDataFrames
based on the spatial relationship
between their geometries.
A common use case might be a spatial join between a point layer and a polygon layer where you want to retain the point geometries and grab the attributes of the intersecting polygons.
We currently support the following methods of spatial joins. We refer to the left_df and right_df which are the correspond to the two dataframes passed in as args.
In a LEFT OUTER JOIN (how='left'
), we keep all rows from the left and duplicate them if necessary to represent multiple hits between the two dataframes. We retain attributes of the right if they intersect and lose right rows that don't intersect. A left outer join implies that we are interested in retaining the geometries of the left.
This is equivalent to the PostGIS query:
SELECT pts.geom, pts.id as ptid, polys.id as polyid
FROM pts
LEFT OUTER JOIN polys
ON ST_Intersects(pts.geom, polys.geom);
geom | ptid | polyid
--------------------------------------------+------+--------
010100000040A9FBF2D88AD03F349CD47D796CE9BF | 4 | 10
010100000048EABE3CB622D8BFA8FBF2D88AA0E9BF | 3 | 10
010100000048EABE3CB622D8BFA8FBF2D88AA0E9BF | 3 | 20
0101000000F0D88AA0E1A4EEBF7052F7E5B115E9BF | 2 | 20
0101000000818693BA2F8FF7BF4ADD97C75604E9BF | 1 |
(5 rows)
In a RIGHT OUTER JOIN (how='right'
), we keep all rows from the right and duplicate them if necessary to represent multiple hits between the two dataframes. We retain attributes of the left if they intersect and lose left rows that don't intersect. A right outer join implies that we are interested in retaining the geometries of the right.
This is equivalent to the PostGIS query:
SELECT polys.geom, pts.id as ptid, polys.id as polyid
FROM pts
RIGHT OUTER JOIN polys
ON ST_Intersects(pts.geom, polys.geom);
geom | ptid | polyid
----------+------+--------
01...9BF | 4 | 10
01...9BF | 3 | 10
02...7BF | 3 | 20
02...7BF | 2 | 20
00...5BF | | 30
(5 rows)
In an INNER JOIN (how='inner'
), we keep rows from the right and left only where their binary predicate is True
. We duplicate them if necessary to represent multiple hits between the two dataframes. We retain attributes of the right and left only if they intersect and lose all rows that do not. An inner join implies that we are interested in retaining the geometries of the left.
This is equivalent to the PostGIS query:
SELECT pts.geom, pts.id as ptid, polys.id as polyid
FROM pts
INNER JOIN polys
ON ST_Intersects(pts.geom, polys.geom);
geom | ptid | polyid
--------------------------------------------+------+--------
010100000040A9FBF2D88AD03F349CD47D796CE9BF | 4 | 10
010100000048EABE3CB622D8BFA8FBF2D88AA0E9BF | 3 | 10
010100000048EABE3CB622D8BFA8FBF2D88AA0E9BF | 3 | 20
0101000000F0D88AA0E1A4EEBF7052F7E5B115E9BF | 2 | 20
(4 rows)
Let's take a look at how we'd implement these using GeoPandas
. First, load up the NYC test data into GeoDataFrames
:
%matplotlib inline
from shapely.geometry import Point
from geopandas import GeoDataFrame, read_file
import geodatasets
# NYC Boros
zippath = geodatasets.get_path("nybb")
polydf = read_file(zippath)
# Generate some points
b = [int(x) for x in polydf.total_bounds]
N = 8
pointdf = GeoDataFrame(
[
{"geometry": Point(x, y), "value1": x + y, "value2": x - y}
for x, y in zip(
range(b[0], b[2], int((b[2] - b[0]) / N)),
range(b[1], b[3], int((b[3] - b[1]) / N)),
)
]
)
# Make sure they're using the same projection reference
pointdf.crs = polydf.crs
pointdf
polydf
pointdf.plot()
polydf.plot()
join_left_df = pointdf.sjoin(polydf, how="left")
join_left_df
# Note the NaNs where the point did not intersect a boro
join_right_df = pointdf.sjoin(polydf, how="right")
join_right_df
# Note Staten Island is repeated
join_inner_df = pointdf.sjoin(polydf, how="inner")
join_inner_df
# Note the lack of NaNs; dropped anything that didn't intersect
We're not limited to using the intersection
binary predicate. Any of the Shapely
geometry methods that return a Boolean can be used by specifying the predicate
kwarg.
pointdf.sjoin(polydf, how="left", predicate="within")
We can also conduct a nearest neighbour join with sjoin_nearest
.
pointdf.sjoin_nearest(polydf, how="left", distance_col="Distances")
# Note the optional Distances column with computed distances between each point
# and the nearest polydf geometry.