Lab 01: Dealing with Spatial Data in Python

In this tutorial, we will work on dealing with non-spatial and spatial data using Python libraries: pandas and geopandas. Many concepts/techniques will be echoed to the ones you have learned from Lecture 02: Spatial Database I and Lecture 03: Spatial Database II.

To follow this tutorial, you should have installed Jupyter Notebook (or Jupyter Lab) on your own computer (note: computers in the lab should have already installed it).

Requred packages include:

• pandas / geopandas

• shapely

• pysal

• pyproj

• osmnx / pyrosm

• matplotlib (visualization)

Mostly, you can use the commond like pip install pandas to install the package.

Basics in Pandas (for non-spatial data)

Pandas is currently the mostly important tool for data scientists working in Python. It is the backbone of many state-of-the-art techniques like machine learning and visualization. Here we cover the basics of using Pandas. For more comprehensive tutorial, follow this video and/or this post

First of all, the two key components in Pandas are Series and DataFrame. A Series is essentially a column, and a DataFrame is a multi-dimensional table made up of a collection of Series. These two components are quite similar in that many operations that you can do with one, you can do with the other too, such as filling in null values and calculating the mean.

Create a dataframe from scratch

There are many ways to create a DataFrame from scratch, but a great option is to just use a simple dict (this is a common data structure called dictionary, which is composed by a key:value pair). Each key:value pair corresponds to a column in the resulting DataFrame.

Let’s say we have a fruit stand that sells apples and oranges. We want to have a column for each fruit and a row for each customer purchase. To organize this as a dictionary for pandas we could do something like:

import pandas as pd
data = {
    'apples': [3, 2, 0, 1], 
    'oranges': [0, 3, 7, 2]
}
purchases = pd.DataFrame(data)
purchases

/var/folders/xg/5n3zc4sn5hlcg8zzz6ysx21m0000gq/T/ipykernel_83912/2875102215.py:1: DeprecationWarning: 
Pyarrow will become a required dependency of pandas in the next major release of pandas (pandas 3.0),
(to allow more performant data types, such as the Arrow string type, and better interoperability with other libraries)
but was not found to be installed on your system.
If this would cause problems for you,
please provide us feedback at https://github.com/pandas-dev/pandas/issues/54466
        
  import pandas as pd

	apples	oranges
0	3	0
1	2	3
2	0	7
3	1	2

The Index of this DataFrame was given to us as the numbers 0-3. We could also create our own when we initialize the DataFrame.

purchases = pd.DataFrame(data, index=['June', 'Robert', 'Lily', 'David'])
purchases

	apples	oranges
June	3	0
Robert	2	3
Lily	0	7
David	1	2

So now we could locate a customer’s order by using their name:

purchases.loc['June']

apples     3
oranges    0
Name: June, dtype: int64

Loading data

We can also load in data with formats like csv, json, txt, and so on. For example, if you downloaded purchase.csv to your local directory, you should be able to load the data by running:

purchases_loaded = pd.read_csv('purchase.csv')
purchases_loaded

	Unnamed:0	apples	oranges
0	June	3	0
1	Robert	2	3
2	Lily	0	7
3	David	1	2

Note here that CSVs don’t have indexes like DataFrames, so we need to designate the index_col when reading:

purchases_loaded = pd.read_csv('purchase.csv', index_col=0)
purchases_loaded

	apples	oranges
Unnamed:0
June	3	0
Robert	2	3
Lily	0	7
David	1	2

Viewing your data

There are many operations to view/describe your data. For example, you can use .head() to check the first several rows of your dataframe, .tail() to see the last severl rows, .info() to have a list of information about your dataframe (you are suggested to always run it first after your data is loaded), .shape to see the dimension of your dataframe (i.e., how many rows and columns are there?), etc. Let’s try .info() here and you could also try the others yourself.

purchases_loaded.info()

<class 'pandas.core.frame.DataFrame'>
Index: 4 entries, June to David
Data columns (total 2 columns):
 #   Column   Non-Null Count  Dtype
---  ------   --------------  -----
 0   apples   4 non-null      int64
 1   oranges  4 non-null      int64
dtypes: int64(2)
memory usage: 96.0+ bytes

From the output, you can see our loaded purchases_loaded dataframe has 4 entries (rows), and there are two columnes, each has 4 non-null values and their data types are both int64 (i.e., integer with 64 digits). Data type here is an important concept (we have covered it in our lecture too). Different data types might imply various operations/analysis that are available. See the table below for a full list of data types in Pandas, and Python and NumPy (another important package in Python).

Querying (selecting, slicing, extracting) Dataframe

Similar to the complex DBMS, Pandas also support selecting, slicing or extracting data from the Dataframe.

Select by column

We can extract a column using square brackets like this:

purchases_apple = purchases_loaded['apples']
purchases_apple

Unnamed:0
June       3
Robert     2
Lily       0
David      1
Name: apples, dtype: int64

type(purchases_apple)

pandas.core.series.Series

Notice that the returned purchases_apple is a Series. To extract a column as a DataFrame, we need to pass a list of column names. In our case that’s just a single column of “apples”, but is within a double bracket:

purchases_apple = purchases_loaded[['apples']]
type(purchases_apple)

pandas.core.frame.DataFrame

Select by row

For rows, we can use two ways to extract data:

• loc: locates by name

• iloc: locates by numerical index

For example, we can select the row of June (how many apples and oranges June has got?) from our purchase_loaded dataframe.

purchases_June = purchases_loaded.loc["June"]
purchases_June

apples     3
oranges    0
Name: June, dtype: int64

Conditional selection

So far, We’ve gone over how to select columns and rows, but what if we want to make a conditional selection?

For example, what if we want to filter our purchases_loaded DataFrame to show only people who bought apples less than 2?

To do that, we take a column from the DataFrame and apply a Boolean condition to it. Here’s an example of a Boolean condition:

condition = (purchases_loaded['apples'] < 2)
condition

Unnamed:0
June       False
Robert     False
Lily        True
David       True
Name: apples, dtype: bool

A little bit more complex, how about showing people who bought apples less than 2 but oranges larger than 2? Can you try it? Hint, you need to use the logic operator & to connect two conditions.

GeoPandas for Spatial Data

Geopandas is designed to process spatial data in Python. Geopandas combines the capabilities of data processing library pandas with other packages like shapely and fiona for managing and visualizing spatial data. The main data structures in geopandas are GeoSeries and GeoDataFrame which extend the capabilities of Series and DataFrame from pandas, but with spatial speciality.

The key difference between GeoDataFrame and DataFrame is that a GeoDataFrame contains at least one column as geometry so that the data entry is spatially referenced. By default, the name of this column is 'geometry'. The geometry column is a GeoSeries which contains the geometries (points, lines, polygons, multipolygons etc.) as shapely objects.

Loading spatial data

Spatial data that are in the format of geojson, shp, etc. can all be loaded as GeoPandas’ Dataframe (GeoDataFrame) by using the function read_file(). Let’s use a shapefile (shp) downloaded from OpenStreetMap as an example here. You can also find the specific data (building in Bristol) from Blackboard.

import geopandas as gpd

# Filepath
bristol_building_file = "./bristol-buildings.shp/gis_osm_buildings_a_free_1.shp" # make sure the directory is correct in your case

# Read the file
bristol_building = gpd.read_file(bristol_building_file)

# How does it look?
bristol_building.head()

	osm_id	code	fclass	name	type	geometry
0	4309554	1500	building	Bristol City Hall	None	POLYGON ((-2.60242 51.45242, -2.60241 51.45244...
1	4315318	1500	building	Clifton Cathedral	church	POLYGON ((-2.61673 51.45965, -2.61673 51.45965...
2	4315809	1500	building	Clifton Down Shopping Centre	retail	POLYGON ((-2.61133 51.46431, -2.61070 51.46441...
3	4317900	1500	building	Za Za Bazaar	None	POLYGON ((-2.59869 51.45029, -2.59865 51.45042...
4	4317901	1500	building	Mackenzies Café Bar	commercial	POLYGON ((-2.59851 51.45107, -2.59850 51.45110...

As can be seen here, the GeoDataFrame bristol_building contains various attributes in separate columns. The geometry column contains the spatial information (it is WKT format, which is implemented by the shapely library). We can next take a look of the basic information of bristol_building using the following command:

bristol_building.info()

<class 'geopandas.geodataframe.GeoDataFrame'>
RangeIndex: 149805 entries, 0 to 149804
Data columns (total 6 columns):
 #   Column    Non-Null Count   Dtype   
---  ------    --------------   -----   
 0   osm_id    149805 non-null  object  
 1   code      149805 non-null  int64   
 2   fclass    149805 non-null  object  
 3   name      7587 non-null    object  
 4   type      98915 non-null   object  
 5   geometry  149805 non-null  geometry
dtypes: geometry(1), int64(1), object(4)
memory usage: 6.9+ MB

What kind of information can you get from this output?

Since our data is intrinsically spatial (it has a geometry column), we can visualize it to better understand its spatial distribution. plot() is the function for it:

bristol_building.plot()

<Axes: >

Saving spatial data

Once you are done with your process/analysis, you can also save your GeoDataFrame into files (e.g., .shp, .geojson, etc). Here, since we loaded data from .shp, let’s now try to save our data to .geojson (in the example below, we only save a subset of bristol_building):

bristol_building.iloc[:100].to_file('osm_bristol_buildings.geojson', driver='GeoJSON') 
## this will save your data to the current directory same to this notebook. 
## you can check the current directory by ruing cwd = os.getcwd()

Retrieving data directly from OSM

We have so far seen how to read spatial data from your computer disk (i.e., the data is downloaded and saved on your local directory). Next, let’s see how we can retrieve data from OSM directly using a library called pyrosm. With pyrosm, you can easily retrieve data from anywhere in the world based on OSM.PBF files (a specific data format for OSM) that are distributed by Geofabrik. In fact, this is where the Bristol buildings data were downloaded. The package aims to be a more efficient way to parse OSM data covering large geographical areas (such as countries and cities).

Note that if you would like to be flexible about your download, e.g., selecting a bounding box by yourself rather than by administrative regions, you can consider using OSMnx library.

from pyrosm import OSM, get_data# Download data for Bristol

bristol = get_data("bristol")

# Initialize the reader object for Bristol
osm = OSM(bristol)

In the first command, we downloaded the data for “Bristol” using the get_data function. This function in fact automates the data downloading process and stores the data locally in a temporary folder. The next step was to initialize a reader object called osm. The OSM() function takes the filepath of a given osm.pbf file as an input. Notice that at this point we actually haven’t yet read any data into a GeoDataFrame.

OSM contains a lot of information about the world, which is contributed by citizens like you and me. In principle, we can retrieve information under various themes from OSM using the following functions.

• road networks –> osm.get_network()

• buildings –> osm.get_buildings()

• Points of Interest (POI) –> osm.get_pois()

• landuse –> osm.get_landuse()

• natural elements –> osm.get_natural()

• boundaries –> osm.get_boundaries()

Try them yourselves! You might consider using them in your dissertation project too. Here, let’s extract the road network at Bristol from OSM:

bristol_roadnetwork = osm.get_network()

/Users/gy22808/opt/anaconda3/envs/ox/lib/python3.12/site-packages/pyrosm/networks.py:37: FutureWarning: ChainedAssignmentError: behaviour will change in pandas 3.0!
You are setting values through chained assignment. Currently this works in certain cases, but when using Copy-on-Write (which will become the default behaviour in pandas 3.0) this will never work to update the original DataFrame or Series, because the intermediate object on which we are setting values will behave as a copy.
A typical example is when you are setting values in a column of a DataFrame, like:

df["col"][row_indexer] = value

Use `df.loc[row_indexer, "col"] = values` instead, to perform the assignment in a single step and ensure this keeps updating the original `df`.

See the caveats in the documentation: https://pandas.pydata.org/pandas-docs/stable/user_guide/indexing.html#returning-a-view-versus-a-copy

  edges, nodes = prepare_geodataframe(

---------------------------------------------------------------------------
KeyboardInterrupt                         Traceback (most recent call last)
Cell In[17], line 1
----> 1 bristol_roadnetwork = osm.get_network()

File ~/opt/anaconda3/envs/ox/lib/python3.12/site-packages/pyrosm/pyrosm.py:249, in OSM.get_network(self, network_type, extra_attributes, nodes, timestamp)
    246 self._read_pbf(timestamp)
    248 # Filter network data with given filter
--> 249 edges, node_gdf = get_network_data(
    250     self._node_coordinates,
    251     self._way_records,
    252     tags_as_columns,
    253     network_filter,
    254     self.bounding_box,
    255     slice_to_segments=nodes,
    256 )
    258 if edges is not None:
    259     # Add metadata
    260     edges._metadata.append(network_type)

File ~/opt/anaconda3/envs/ox/lib/python3.12/site-packages/pyrosm/networks.py:37, in get_network_data(node_coordinates, way_records, tags_as_columns, network_filter, bounding_box, slice_to_segments)
     34     return None, None
     36 # Prepare GeoDataFrame
---> 37 edges, nodes = prepare_geodataframe(
     38     nodes,
     39     node_coordinates,
     40     ways,
     41     relations,
     42     relation_ways,
     43     tags_as_columns,
     44     bounding_box,
     45     parse_network=True,
     46     calculate_seg_lengths=slice_to_segments,
     47 )
     49 return edges, nodes

File ~/opt/anaconda3/envs/ox/lib/python3.12/site-packages/pyrosm/frames.pyx:136, in pyrosm.frames.prepare_geodataframe()

File ~/opt/anaconda3/envs/ox/lib/python3.12/site-packages/pyrosm/frames.pyx:143, in pyrosm.frames.prepare_geodataframe()

File ~/opt/anaconda3/envs/ox/lib/python3.12/site-packages/pyrosm/frames.pyx:72, in pyrosm.frames.prepare_way_gdf()

File ~/opt/anaconda3/envs/ox/lib/python3.12/site-packages/pyrosm/distance.py:78, in calculate_geom_length(geom)
     77 def calculate_geom_length(geom):
---> 78     return calculate_geom_array_length(geom).sum().round(0)

File ~/opt/anaconda3/envs/ox/lib/python3.12/site-packages/pyrosm/distance.py:82, in calculate_geom_array_length(geom_array)
     81 def calculate_geom_array_length(geom_array):
---> 82     coords = get_coordinates(geom_array).T
     84     # Only every second element should be taken from the coordinates
     85     lon1, lat1 = coords[0][:-1:2], coords[1][:-1:2]

File ~/opt/anaconda3/envs/ox/lib/python3.12/site-packages/shapely/coordinates.py:136, in get_coordinates(geometry, include_z, return_index)
     93 def get_coordinates(geometry, include_z=False, return_index=False):
     94     """Gets coordinates from a geometry array as an array of floats.
     95 
     96     The shape of the returned array is (N, 2), with N being the number of
   (...)
    134     ([[2.0, 2.0], [4.0, 4.0], [0.0, 0.0]], [0, 0, 1])
    135     """
--> 136     return lib.get_coordinates(
    137         np.asarray(geometry, dtype=np.object_), include_z, return_index
    138     )

KeyboardInterrupt: 

We can get the lenth of this DataFrame (how many road network do we have in Bristol from OSM?) and some basic descritions of it by running:

len(bristol_roadnetwork) 

102212

bristol_roadnetwork.describe() # note that it only provides a statistical summary for columns whoes data type is numeric

	id	timestamp	version	length
count	1.022120e+05	102212.0	102212.0	102212.000000
mean	4.555495e+08	0.0	-1.0	108.314650
std	3.824359e+08	0.0	0.0	193.418633
min	1.900000e+02	0.0	-1.0	0.000000
25%	8.795292e+07	0.0	-1.0	20.000000
50%	3.684994e+08	0.0	-1.0	48.000000
75%	7.817397e+08	0.0	-1.0	111.000000
max	1.245598e+09	0.0	-1.0	7572.000000

Likewise, we can also plot it. Please try it yourself.

Coordinate Reference System for GeoDataFrame

Another difference between GeoDataFrames and DataFrames is that the former has intrinsic coordinate reference system (CRS) as it has the geometry column. To check this information, we can call its attribute crs:

bristol_roadnetwork.crs

<Geographic 2D CRS: EPSG:4326>
Name: WGS 84
Axis Info [ellipsoidal]:
- Lat[north]: Geodetic latitude (degree)
- Lon[east]: Geodetic longitude (degree)
Area of Use:
- name: World.
- bounds: (-180.0, -90.0, 180.0, 90.0)
Datum: World Geodetic System 1984 ensemble
- Ellipsoid: WGS 84
- Prime Meridian: Greenwich

It shows that coordinates in geometry column are using the WGS 84 with a EPSG code 4326. In fact, it is the mostly used coordinate reference system (CRS) in spatial data science as it is a global coordinate system and has been used for GPS as well. However, as we covered in the lecture, those global CRSs are not that accurate for local regions. For the UK, or Bristol, a more commonly used CRS is EPSG:27700 (National Grid for Great Britain), and this CRS is also projected. Let’s then transfer bristol_roadnetwork from EPSG:4326 to EPSG:27700:

bristol_roadnetwork_reprojected = bristol_roadnetwork.to_crs(epsg=27700)
bristol_roadnetwork_reprojected.crs

<Projected CRS: EPSG:27700>
Name: OSGB36 / British National Grid
Axis Info [cartesian]:
- E[east]: Easting (metre)
- N[north]: Northing (metre)
Area of Use:
- name: United Kingdom (UK) - offshore to boundary of UKCS within 49°45'N to 61°N and 9°W to 2°E; onshore Great Britain (England, Wales and Scotland). Isle of Man onshore.
- bounds: (-9.01, 49.75, 2.01, 61.01)
Coordinate Operation:
- name: British National Grid
- method: Transverse Mercator
Datum: Ordnance Survey of Great Britain 1936
- Ellipsoid: Airy 1830
- Prime Meridian: Greenwich

Now we have an projected CRS for the road network data in Bristol. To confirm the difference, let’s take a look at the geometry of the first row in our original road network bristol_roadnetwork and the projected bristol_roadnetwork_reprojected.

orig_geom = bristol_roadnetwork.loc[0, "geometry"]
projected_geom = bristol_roadnetwork_reprojected.loc[0, "geometry"]

print("Orig:\n", orig_geom, "\n")
print("Proj:\n", projected_geom)

Orig:
 MULTILINESTRING ((-2.6097211837768555 51.3657112121582, -2.609546422958374 51.36564636230469), (-2.609546422958374 51.36564636230469, -2.609264612197876 51.365482330322266), (-2.609264612197876 51.365482330322266, -2.6088733673095703 51.365318298339844), (-2.6088733673095703 51.365318298339844, -2.608508825302124 51.36508560180664), (-2.608508825302124 51.36508560180664, -2.6083426475524902 51.3649787902832), (-2.6083426475524902 51.3649787902832, -2.608126163482666 51.364810943603516), (-2.608126163482666 51.364810943603516, -2.607820749282837 51.36458206176758)) 

Proj:
 MULTILINESTRING ((357648.974429959 163137.70370458602, 357661.0804417359 163130.39013390895), (357661.0804417359 163130.39013390895, 357680.5471925497 163111.98427999223), (357680.5471925497 163111.98427999223, 357707.6323758529 163093.51498389826), (357707.6323758529 163093.51498389826, 357732.79552590376 163067.42507962266), (357732.79552590376 163067.42507962266, 357744.2656479743 163055.45009858347), (357744.2656479743 163055.45009858347, 357759.1817168629 163036.65820931632), (357759.1817168629 163036.65820931632, 357780.23268873093 163011.0270668068))

As we be seen, the coordinates that form our road segments (MULTILINESTRING) has changed from decimal degrees to meters. Next, let’s visualize it:

orig_geom

projected_geom

As you can see, the shape of the two road segments are quite different (e.g., the lenth, where the curve occures, etc.). This is exactly due to the difference between the two CRSs.

It is also worth noting here, the data type, MultiLineString, of the variables orig_geom and projected_geom are defined by shapely. It enables us to conduct these kind of spatial operations and visializations.

type(orig_geom)

shapely.geometry.multilinestring.MultiLineString

Computation on GeoDataFrame

There are many operations embeded in GeoDataFrame that can be directly called to do some spatial computations. For example, we can get the area of buildings for our bristol_building dataframe:

bristol_building["building_area"] = bristol_building.area
bristol_building["building_area"].describe()

/var/folders/xg/5n3zc4sn5hlcg8zzz6ysx21m0000gq/T/ipykernel_11376/2251981216.py:1: UserWarning: Geometry is in a geographic CRS. Results from 'area' are likely incorrect. Use 'GeoSeries.to_crs()' to re-project geometries to a projected CRS before this operation.

  bristol_building["building_area"] = bristol_building.area

count    1.498050e+05
mean     1.341308e-08
std      6.657677e-08
min      5.667000e-11
25%      6.084315e-09
50%      7.575225e-09
75%      1.033238e-08
max      1.038017e-05
Name: building_area, dtype: float64

Here, you can see a warning that the current Geometry is in a geographic CRS, hence the results of computing area mighht not be accurate. Can you project the dataframe to a projected coordicate reference system (e.g.,EPSG:27700 in our case)? After your projection, do the area computation again. What do the results look like? What is the unit of the area?

Spatial join

As we have discussed in the lecture, joining tables using keys is a core operation for DBMS. Regarding spatial data, spatial join is somewhat similar to table join but with the operation being based on geometries rather than keys.

In this tutorial, we will try to conduct a spatial join and merge information between two GeoDataFrames. First, let’s read all restaurants (a type of Point of Interests (POI)) at Bristol from the OSM. Then, we combine information from restaurants to the underlying building (restaurants typically are within buildings). We will again use pyrosm for reading the data, but this time we will use the get_pois() function:

# Read Points of Interest (POI) using the same OSM reader object that was initialized earlier
# The custom_filter={"amenity": ["restaurant"]} indicates that we want only "restaurant", a type of POI
bristol_restaurants = osm.get_pois(custom_filter={"amenity": ["restaurant"]})
bristol_restaurants.plot()

<Axes: >

bristol_restaurants.info()

<class 'geopandas.geodataframe.GeoDataFrame'>
RangeIndex: 677 entries, 0 to 676
Data columns (total 31 columns):
 #   Column            Non-Null Count  Dtype   
---  ------            --------------  -----   
 0   id                677 non-null    int64   
 1   timestamp         677 non-null    uint32  
 2   tags              640 non-null    object  
 3   changeset         143 non-null    float64 
 4   lon               142 non-null    float32 
 5   version           677 non-null    int32   
 6   visible           676 non-null    object  
 7   lat               142 non-null    float32 
 8   addr:city         501 non-null    object  
 9   addr:country      53 non-null     object  
 10  addr:housenumber  488 non-null    object  
 11  addr:housename    101 non-null    object  
 12  addr:postcode     620 non-null    object  
 13  addr:place        27 non-null     object  
 14  addr:street       583 non-null    object  
 15  email             29 non-null     object  
 16  name              672 non-null    object  
 17  opening_hours     103 non-null    object  
 18  operator          14 non-null     object  
 19  phone             107 non-null    object  
 20  website           258 non-null    object  
 21  amenity           677 non-null    object  
 22  bar               4 non-null      object  
 23  internet_access   10 non-null     object  
 24  source            20 non-null     object  
 25  geometry          677 non-null    geometry
 26  osm_type          677 non-null    object  
 27  building          524 non-null    object  
 28  building:levels   75 non-null     object  
 29  start_date        3 non-null      object  
 30  wikipedia         3 non-null      object  
dtypes: float32(2), float64(1), geometry(1), int32(1), int64(1), object(24), uint32(1)
memory usage: 153.5+ KB

From the info(), we can see that there are 676 restaurants in Bristol according to OSM (you might see a different number depending on which version of OSM data you have downloaded). Note that OSM is a valunteered geographic information platform. So the quality, accuracy, and completness of the data might be low. Next, let’s join data from bristol_buildings to bristol_restaurantsusingsjoin()` function from geopandas:

# Join information from buildings to restaurants
bristol_join = gpd.sjoin(bristol_restaurants, bristol_building)

# Print column names
print(bristol_join.columns)

# Show rows
bristol_join

Index(['id', 'timestamp', 'tags', 'changeset', 'lon', 'version', 'visible',
       'lat', 'addr:city', 'addr:country', 'addr:housenumber',
       'addr:housename', 'addr:postcode', 'addr:place', 'addr:street', 'email',
       'name_left', 'opening_hours', 'operator', 'phone', 'website', 'amenity',
       'bar', 'internet_access', 'source', 'geometry', 'osm_type', 'building',
       'building:levels', 'start_date', 'wikipedia', 'index_right', 'osm_id',
       'code', 'fclass', 'name_right', 'type', 'building_area'],
      dtype='object')

	id	timestamp	tags	changeset	lon	version	visible	lat	addr:city	addr:country	...	building:levels	start_date	wikipedia	index_right	osm_id	code	fclass	name_right	type	building_area
8	853556896	0	None	0.0	-2.611020	0	False	51.458733	None	None	...	NaN	NaN	NaN	24609	451622999	1500	building	The Clifton	None	9.942395e-09
10	1207448023	0	None	0.0	-2.625203	0	False	51.453415	None	None	...	NaN	NaN	NaN	1164	104679655	1500	building	Avon Gorge Hotel	hotel	8.777610e-08
12	1386051923	0	None	0.0	-2.619858	0	False	51.455570	Bristol	None	...	NaN	NaN	NaN	23228	444980827	1500	building	Rodney Hotel	None	2.753519e-08
17	1881624837	0	{"cuisine":"regional","outdoor_seating":"yes",...	0.0	-2.585233	0	False	51.451546	Bristol	None	...	NaN	NaN	NaN	2138	125956114	1500	building	Hilton Garden Inn Bristol City Centre	None	1.725836e-07
21	2900424207	0	{"cuisine":"pizza"}	0.0	-2.589893	0	False	51.456440	None	None	...	NaN	NaN	NaN	1945	124113904	1500	building	The Galleries	None	6.674781e-07
...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...
671	1069613264	0	{"house":"terraced"}	NaN	NaN	-1	False	NaN	Bristol	None	...	2	None	None	145381	1069613265	1500	building	None	house	5.491100e-09
671	1069613264	0	{"house":"terraced"}	NaN	NaN	-1	False	NaN	Bristol	None	...	2	None	None	145380	1069613264	1500	building	None	house	6.423220e-09
674	1188184241	0	{"addr:suburb":"Clifton","fhrs:id":"1584632"}	NaN	NaN	-1	False	NaN	Bristol	None	...	None	None	None	10652	275376189	1500	building	Ruby & White	None	1.343296e-08
674	1188184241	0	{"addr:suburb":"Clifton","fhrs:id":"1584632"}	NaN	NaN	-1	False	NaN	Bristol	None	...	None	None	None	10640	275376177	1500	building	Everyman	None	1.508399e-07
676	13571356292	0	{"addr:suburb":"Clifton","cuisine":"indian","f...	0.0	NaN	-1	NaN	NaN	Bristol	NaN	...	NaN	NaN	NaN	138593	13450709	1500	building	The Mint Room	None	1.037321e-08

862 rows × 38 columns

Now with this joined table, you can check which building each restaurant is locatd in. Note that after joining information from the buildings to restaurants, geometries of the left-side GeoDataFrame, i.e. restaurants, were kept by as the default geometries. So if we plot bristol_join, you will only see restaurants, rather than buildings + restaurant. Please try!

Also by default, sjoin() use the topological relation - intersects. You can also specify this parameter as other types of topological relatoons (e.g., contains and within) in the function. More details can be found at: https://geopandas.org/en/stable/docs/reference/api/geopandas.sjoin.html

Visualization

So far, we simply used the plot() function to visualize GeoDataFrame. These maps are less appealing compared to the ones generated via GIS softwares. In fact, the package: matplotlib is very powerful in providing us beautiful visualization in Python. Let’s try it.

First, let’s add some legends to the bristol_building data using its building type:

ax = bristol_building.plot(column="type", cmap="RdYlBu", figsize=(12,12), legend=True)

Here, we used the parameter column to specify the attribute that is used to specify the color for each building (it can be categorical or continuous). We then used cmap to specify the colormap for the categories and we added the legend by specifying legend=True. Note that since the type of buildings for Bristol is very diverse, we see a long list of legend. There are ways to make it into two columns, for example. Can you explore how to achieve it? Feel free to use Google search!

Another issue is that the map is in a very large map scale. Next, we would like to zoom in a little bit. To do so, we can use set_xlim() and set_ylim() functions:

# Zoom into city center by specifying X and Y coordinate extent
# These values should be given in the units that our data is presented (here decimal degrees)
xmin, xmax = -2.65, -2.55
ymin, ymax = 51.44, 51.48

# Plot the map again
ax = bristol_building.plot(column="type", cmap="RdYlBu", figsize=(12,12), legend=True)
 

# Control and set the x and y limits for the axis
ax.set_xlim([xmin, xmax])
ax.set_ylim([ymin, ymax])

(51.44, 51.48)

As you can see, we now zoomed in to the city center quite a lot. You can adjust the parameters yourself and test more!

Meanwhile, you may wonder whether we can overlay multiple dataframes into the map? The answer is yes. Here is a sample code:

# Zoom into city center by specifying X and Y coordinate extent
# These values should be given in the units that our data is presented (here decimal degrees)
xmin, xmax = -2.65, -2.55
ymin, ymax = 51.44, 51.48

# Plot the map again
ax = bristol_building.plot(column="type", cmap="RdYlBu", figsize=(12,12), legend=True)

# Plot the roads into the same axis
ax = bristol_roadnetwork.plot(ax=ax, edgecolor="gray", linewidth=0.75)

# Control and set the x and y limits for the axis
ax.set_xlim([xmin, xmax])
ax.set_ylim([ymin, ymax])

(51.44, 51.48)

Congrats! You have now finished the very first lab of using Python to process spatial data. I hope you enjoyed it and have seen the power of GeoPandas, and Python in general, for processing and studying spatial data. It is also worth highlighting that the functions introduced in this tutorial are selective. There are way more interesting and useful functions provided by these aforementioned packages. I highly recommended you to explore them by yourself. Learning never stops!

Finally, let’s go back to the task of asking you to figure out how to better organize the long legend box. Below is the solution I found. There might be other ways of doing it. How is yours?

Basically, I used the parameter legend_kwds to set up the number of columns (ncol) to be 4. For more details, check its official documentation. Note that knowing how to read these kinds of documentations would be very helpful for your programming, so it is an important skill/experience you should develop.

from mpl_toolkits.axes_grid1 import make_axes_locatable
import matplotlib.pyplot as plt

xmin, xmax = -2.65, -2.55
ymin, ymax = 51.44, 51.48

fig, ax = plt.subplots(figsize=(12, 12))


bristol_building.plot(column='type',
                categorical=True,
                cmap="RdYlBu",
                ax=ax, 
                legend=True,
                legend_kwds={'loc': 'lower left',
                             'ncol': 4,
                             'bbox_to_anchor': (0, 0, 0.5,0.5)})
#ax.set_axis_off()
ax.set_xlim([xmin, xmax])
ax.set_ylim([ymin, ymax])

plt.show()