Skip to content

Landsat 8 (Microsoft Planetary Computer)

Overview

landsat8_mpc streams Landsat 8 Collection 2 Level-2 surface reflectance directly from the Microsoft Planetary Computer (MPC) STAC API. The verb signs COG URLs using the planetary_computer SDK and stacks the requested bands into a lazy, dask-backed DataArray with dimensions (time, band, y, x).

Quickstart

Get the stream (CubeDynamics grammar)

import cubedynamics as cd
from cubedynamics import pipe, verbs as v

bbox = [-105.35, 39.9, -105.15, 40.1]
cube = (
    pipe(None)
    | v.landsat8_mpc(bbox=bbox, start="2019-07-01", end="2019-07-10")
).unwrap()

ndvi = (cube.sel(band="nir") - cube.sel(band="red")) / (cube.sel(band="nir") + cube.sel(band="red"))
pipe(ndvi) | v.plot(time_dim="time")

Preview plot

Landsat 8 MPC preview

!!! note Image placeholder — after running the code below locally, save a screenshot to docs/assets/datasets/landsat8_mpc-preview.png.

Regenerate this plot

  1. Run the Quickstart snippet to stream a small Landsat 8 stack and compute NDVI.

  2. Capture the viewer returned by the plotting verb:

    python viewer = (pipe(ndvi) | v.plot(time_dim="time")).unwrap() viewer.save("docs/assets/datasets/landsat8_mpc-preview.html")

  3. Open docs/assets/datasets/landsat8_mpc-preview.html in a browser and save a 1200×700 px PNG screenshot to docs/assets/datasets/landsat8_mpc-preview.png.

Required arguments

  • bbox – [min_lon, min_lat, max_lon, max_lat] geographic bounding box.
  • start – ISO start date string (e.g., `"2019-07-01").
  • end – ISO end date string (e.g., `"2019-08-01").

Optional arguments

  • band_aliases – tuple of band aliases to pull (default ("red", "nir")).
  • max_cloud_cover – maximum allowed eo:cloud_cover percentage (default 50).
  • chunks_xy – optional dask chunk sizes for x and y (e.g., { "x": 1024, "y": 1024 }).
  • stac_url – STAC API endpoint, defaults to the MPC service.

Return structure

A lazy :class:xarray.DataArray backed by dask with dimensions (time, band, y, x). Data are not loaded until you compute or plot.

Basic usage

from cubedynamics import pipe, verbs as v

bbox = [-105.35, 39.9, -105.15, 40.1]
cube = (
    pipe(None)
    | v.landsat8_mpc(
        bbox=bbox,
        start="2019-07-01",
        end="2019-08-01",
    )
).unwrap()

NDVI example

After streaming, compute NDVI from the red and near-infrared bands:

red = cube.sel(band="red")
nir = cube.sel(band="nir")
ndvi = (nir - red) / (nir + red)

# Visualize the time series interactively
pipe(ndvi) | v.plot(time_dim="time")

For visualization, you can avoid loading huge rotated footprints by using the downsampling helper:

from cubedynamics.verbs import landsat_ndvi_plot

landsat_ndvi_plot(
    bbox=bbox,
    start="2019-07-01",
    end="2019-08-01",
)

The landsat_vis_ndvi/landsat_ndvi_plot helpers crop all-NaN borders and coarsen the cube so v.plot remains responsive, while the raw v.landsat8_mpc output stays available for analysis at full resolution.

Notes

  • This verb uses MPC-signed STAC assets; no additional credentials are required.
  • Because the DataArray is dask-backed, subsequent analysis and visualization remain lazy until execution.

Back to Datasets Overview
Next recommended page: Which dataset should I use?