Overture Maps Buildings¶
This notebook will give a quick overview of using the new Overture Maps Python library with Lonboard.
+We'll pass in a bounding box covering New York City and the Overture Python API will fetch only the data inside that bounding box. While Overture's buildings dataset contains 2.3 rows, by using a relatively small bounding box, we can download data for our query relatively quickly (around 30 seconds on my internet connection).
+This uses the latest advancements of the GeoParquet specification. Data is fetched directly from the cloud with no server in the middle!
+You can view a hosted version of this notebook on Notebook Sharing Space (26MB download).
+Dependendencies¶
This notebook requires Lonboard (tested with version 0.8) and overturemaps (tested with version 0.4). To install overturemaps, run:
pip install overturemaps
+Imports¶
+import numpy as np
+import overturemaps
+from matplotlib.colors import LogNorm
+from palettable.colorbrewer.sequential import Oranges_9
+
+from lonboard import Map, PolygonLayer
+from lonboard.colormap import apply_continuous_cmap
+Using this bounding box tool we can find a bounding box covering Manhattan. Choose "CSV" in the drop down box, then paste the result in the following cell:
+bbox = -74.02169, 40.696423, -73.891338, 40.831263
+Ensure you pass the bbox into overturemaps.record_batch_reader as in the next line.
record_batch_reader returns a RecordBatchReader, an iterator of Arrow RecordBatches. We call .read_all() to materialize all batches into memory. This cell fetches the relevant data from the cloud (AWS S3), so it may take a minute to download.
table = overturemaps.record_batch_reader("building", bbox).read_all()
+
+# Temporarily required as of Lonboard 0.8 to avoid a Lonboard bug
+table = table.combine_chunks()
+Let's color buildings by their heights. First, convert the "height" column to a numpy array, then replace any null values with 1.
heights = table["height"].to_numpy()
+heights = np.nan_to_num(heights, nan=1)
+Building heights tend to scale non-linearly. That is, the vast majority of buildings are relatively low, but a few buildings are very tall. So that the low buildings aren't completely washed out, we'll use matplotlib's LogNorm to normalize these values according to a log scale.
normalizer = LogNorm(1, heights.max(), clip=True)
+normalized_heights = normalizer(heights)
+Then we can apply a colormap to these heights. For this case, we'll use the Oranges_9 colormap.
Oranges_9.mpl_colormap
+Using apply_continuous_cmap, we can apply our values onto this colormap.
colors = apply_continuous_cmap(normalized_heights, Oranges_9)
+We create a PolygonLayer with our data, passing in the heights and colors from before.
layer = PolygonLayer(
+ # Select only a few attribute columns from the table
+ table=table.select(["id", "height", "geometry", "names"]),
+ extruded=True,
+ get_elevation=heights,
+ get_fill_color=colors,
+)
+We manually set the view state here for the original NYC data so that the map will start pitched. Remove or change this view_state call if you change the input dataset.
view_state = {
+ "longitude": -73.98416810282863,
+ "latitude": 40.72651721370669,
+ "zoom": 12.726630492730596,
+ "pitch": 59.80465353190481,
+ "bearing": 13.243243243243244,
+}
+m = Map(layer, view_state=view_state)
+m
+
+