Note

You can view & download the original notebook on GitHub.

Running on a cluster#

We’ll use a Dask cluster in the cloud—in this case, using Coiled—to use many machines to process the data in parallel. We can also run in the data center where the data is stored for better performance.

To use Coiled, you’ll need a Coiled account (free), and you’ll need to connect Coiled to your cloud provider account (AWS or GCP) so it can launch clusters on your behalf.

Another option for Dask clusters is Microsoft’s Planetary Computer.

[2]:
import coiled
import distributed

cluster = coiled.Cluster(
    name="stackstac",
    n_workers=8,
    package_sync=True,
    backend_options={"region": "us-west-2"},  # where the data is
)
client = distributed.Client(cluster)
client
[2]:

Client

Client-6a095c68-6f7d-11ed-91e2-acde48001122

Connection method: Cluster object Cluster type: coiled.ClusterBeta
Dashboard: http://35.84.199.219:8787

Cluster Info

[3]:
import stackstac
import pystac_client
from rasterio.enums import Resampling

Search for a full year of Sentinel-2 data over Santa Fe, New Mexico.#

We’ll look at 2019-2020.

[4]:
%%time
items = pystac_client.Client.open(
    "https://earth-search.aws.element84.com/v1"
).search(
    intersects=dict(type="Point", coordinates=[-106, 35.7]),
    collections=["sentinel-2-l2a"],
    datetime="2019-01-01/2020-01-01",
    max_items=10_000,
).item_collection()
len(items)
CPU times: user 1.16 s, sys: 81.1 ms, total: 1.24 s
Wall time: 7.91 s
[4]:
292

Set a coarser resolution to speed up the computation#

We’ll have stackstac retrieve the data at 100m resolution, instead of its native 10m. Since the data is stored in Cloud-Optimized GeoTIFFs with internal overviews, fetching lower-resolution data is very efficient and requires processing an order of magnitude less data.

(Internally, stackstac is just telling rasterio/GDAL to build a VRT at this resolution. GDAL then automatically figures out which overview level to fetch data from.)

We also set bounds_latlon to just the area we want to look at (additionally, this drops any items that don’t intersect that bounding box), and set the resampling method to bilinear to produce a nicer-looking image.

[5]:
%%time
stack = stackstac.stack(
    items,
    resolution=100,
    bounds_latlon=(-106.2, 35.6, -105.6, 36),
    resampling=Resampling.bilinear
)
CPU times: user 279 ms, sys: 9.37 ms, total: 288 ms
Wall time: 295 ms
[6]:
stack
[6]:
<xarray.DataArray 'stackstac-efb5ba84853ae19e5a8b478f6f73560f' (time: 292,
                                                                band: 32,
                                                                y: 450, x: 547)>
dask.array<fetch_raster_window, shape=(292, 32, 450, 547), dtype=float64, chunksize=(1, 1, 450, 547), chunktype=numpy.ndarray>
Coordinates: (12/52)
  * time                                     (time) datetime64[ns] 2019-01-02...
    id                                       (time) <U24 'S2A_13SDV_20190102_...
  * band                                     (band) <U12 'aot' ... 'wvp-jp2'
  * x                                        (x) float64 3.913e+05 ... 4.459e+05
  * y                                        (y) float64 3.985e+06 ... 3.94e+06
    s2:unclassified_percentage               (time) object 0.828183 ... 15.96...
    ...                                       ...
    title                                    (band) <U31 'Aerosol optical thi...
    gsd                                      (band) object None 10 ... None None
    common_name                              (band) object None 'blue' ... None
    center_wavelength                        (band) object None 0.49 ... None
    full_width_half_max                      (band) object None 0.098 ... None
    epsg                                     int64 32613
Attributes:
    spec:        RasterSpec(epsg=32613, bounds=(391300, 3939700, 446000, 3984...
    crs:         epsg:32613
    transform:   | 100.00, 0.00, 391300.00|\n| 0.00,-100.00, 3984700.00|\n| 0...
    resolution:  100

For comparison, this is how much data we’d be processing if we’d used full 10m resolution:

[7]:
import dask
dask.utils.format_bytes(stackstac.stack(items).nbytes)
[7]:
'15.66 TiB'

Prepare monthly RGB composites#

Now, use standard xarray methods to select out the red, green, and blue bands, then make monthly median composites.

[8]:
rgb = stack.sel(band=["red", "green", "blue"])
monthly_rgb = rgb.resample(time="MS").median(dim="time")
monthly_rgb
[8]:
<xarray.DataArray 'stackstac-efb5ba84853ae19e5a8b478f6f73560f' (time: 12,
                                                                band: 3,
                                                                y: 450, x: 547)>
dask.array<stack, shape=(12, 3, 450, 547), dtype=float64, chunksize=(1, 2, 450, 547), chunktype=numpy.ndarray>
Coordinates: (12/22)
  * band                                     (band) <U12 'red' 'green' 'blue'
  * x                                        (x) float64 3.913e+05 ... 4.459e+05
  * y                                        (y) float64 3.985e+06 ... 3.94e+06
    proj:epsg                                int64 32613
    constellation                            <U10 'sentinel-2'
    mgrs:latitude_band                       <U1 'S'
    ...                                       ...
    gsd                                      (band) object 10 10 10
    common_name                              (band) object 'red' 'green' 'blue'
    center_wavelength                        (band) object 0.665 0.56 0.49
    full_width_half_max                      (band) object 0.038 0.045 0.098
    epsg                                     int64 32613
  * time                                     (time) datetime64[ns] 2019-01-01...
Attributes:
    spec:        RasterSpec(epsg=32613, bounds=(391300, 3939700, 446000, 3984...
    crs:         epsg:32613
    transform:   | 100.00, 0.00, 391300.00|\n| 0.00,-100.00, 3984700.00|\n| 0...
    resolution:  100

Compute in parallel on the cluster#

[9]:
client.wait_for_workers(8)
[10]:
%time rgb_ = monthly_rgb.compute()
CPU times: user 3.32 s, sys: 1.22 s, total: 4.54 s
Wall time: 1min 42s

Using 8 Coiled workers (2 CPU, 8 GiB memory each), we processed the ~10 GB of data into median composites in a couple minutes.

Go to the dashboard at https://cloud.coiled.io to watch the computation in progress.

[11]:
rgb_.plot.imshow(row="time", rgb="band", robust=True, size=6);
../_images/examples_cluster_17_0.png

According to the Coiled dashboard, this cost about 74 cents. Was this picture worth 74 cents to you?