Multi-station NO2 — the traffic diurnal cycle¶

Fetch hourly-rollup NO2 for several stations in one metro area, then group by hour-of-day split into weekday vs. weekend to reveal the traffic-driven diurnal pattern. Shows multi-station handling and using the lat/lon columns to tell stations apart.

Needs a free OPENAQ_API_KEY. The live cell is guarded so the notebook stays --nbval-lax-safe. Hourly rollups over a month are a few hundred rows per sensor — keep the window modest.

Setup¶

Imports up front: os reads the API key from the environment, matplotlib draws the diurnal plot at the end, and earthlens provides the unified EarthLens entry point.

In [ ]:

import os

import matplotlib.pyplot as plt

from earthlens import EarthLens

import os import matplotlib.pyplot as plt from earthlens import EarthLens

The request¶

Pick a bounding box over the Los Angeles basin and a one-month window. These define the area of interest and date range for the OpenAQ query.

In [ ]:

bbox_lat = [34.0, 34.3]  # Los Angeles basin
bbox_lon = [-118.5, -118.1]
start, end = "2024-03-01", "2024-03-31"

bbox_lat = [34.0, 34.3] # Los Angeles basin bbox_lon = [-118.5, -118.1] start, end = "2024-03-01", "2024-03-31"

Fetch the hourly NO2 rollups¶

Build the EarthLens request first, then call download() on its own line — the server-side hourly rollup keeps each sensor to a few hundred rows. The whole fetch is guarded behind OPENAQ_API_KEY so the notebook stays safe to run without credentials.

In [ ]:

df = None
if os.environ.get("OPENAQ_API_KEY"):
    openaq = EarthLens(
        data_source="openaq",
        variables=["no2"],
        start=start,
        end=end,
        aoi=[bbox_lon[0], bbox_lat[0], bbox_lon[1], bbox_lat[1]],
        temporal_resolution="hourly",  # server-side hourly rollup
        max_locations=8,
        path="out/openaq",
    )
    df = openaq.download(progress_bar=False)
    print(df.shape, "stations:", df['station_id'].nunique() if df is not None else 0)
else:
    print("set OPENAQ_API_KEY to run the live cell")

df = None if os.environ.get("OPENAQ_API_KEY"): openaq = EarthLens( data_source="openaq", variables=["no2"], start=start, end=end, aoi=[bbox_lon[0], bbox_lat[0], bbox_lon[1], bbox_lat[1]], temporal_resolution="hourly", # server-side hourly rollup max_locations=8, path="out/openaq", ) df = openaq.download(progress_bar=False) print(df.shape, "stations:", df['station_id'].nunique() if df is not None else 0) else: print("set OPENAQ_API_KEY to run the live cell")

Group into a weekday/weekend diurnal cycle¶

Bucket each reading by hour-of-day and a weekend flag, then average so each hour gets one weekday and one weekend mean. The result is a 24-row table with two columns ready to plot.

In [ ]:

diurnal = None
if df is not None and not df.empty:
    local = df.copy()
    local["hour"] = local["datetime_utc"].dt.hour
    local["is_weekend"] = local["datetime_utc"].dt.dayofweek >= 5
    diurnal = (
        local.groupby(["is_weekend", "hour"])["value"].mean().unstack("is_weekend")
    )
    diurnal.columns = ["weekday" if not c else "weekend" for c in diurnal.columns]
    display(diurnal.head())

diurnal = None if df is not None and not df.empty: local = df.copy() local["hour"] = local["datetime_utc"].dt.hour local["is_weekend"] = local["datetime_utc"].dt.dayofweek >= 5 diurnal = ( local.groupby(["is_weekend", "hour"])["value"].mean().unstack("is_weekend") ) diurnal.columns = ["weekday" if not c else "weekend" for c in diurnal.columns] display(diurnal.head())

Plot the diurnal pattern¶

Plot both curves over the 24-hour axis: the weekday line should show the morning and evening rush-hour NO2 peaks that the weekend line flattens out.

In [ ]:

if diurnal is not None and not diurnal.empty:
    fig, ax = plt.subplots(figsize=(9, 4))
    diurnal.plot(ax=ax, marker="o")
    ax.set_xlabel("hour of day (UTC)")
    ax.set_ylabel("NO2 (µg/m³)")
    ax.set_title("NO2 diurnal cycle — weekday vs weekend (metro mean)")
    ax.set_xticks(range(0, 24, 3))
    plt.show()

if diurnal is not None and not diurnal.empty: fig, ax = plt.subplots(figsize=(9, 4)) diurnal.plot(ax=ax, marker="o") ax.set_xlabel("hour of day (UTC)") ax.set_ylabel("NO2 (µg/m³)") ax.set_title("NO2 diurnal cycle — weekday vs weekend (metro mean)") ax.set_xticks(range(0, 24, 3)) plt.show()