Colab Github

Image Collection to xarray

In this tutorial, we’ll look at how we can use wxee and the wx accessor to convert an ee.ImageCollection of gridMET weather data to an xarray.Dataset for visualization and further processing.

Setup

[ ]:

!pip install wxee

[1]:

import ee
import wxee

ee.Authenticate()
wxee.Initialize()

Converting to xarray

Load the gridMET image collection.

[2]:

gridmet = ee.ImageCollection("IDAHO_EPSCOR/GRIDMET")

Select a date range of the collection to download to xarray. gridMET has 1 image per day, so this will give us 6 images.

[3]:

collection = gridmet.filterDate("2020-09-05", "2020-09-11")

Set the download parameters. We’ll use the full gridMET image extent of the continental US, but if we wanted to select a specific region we could pass an ee.Geometry.Polygon to the region argument.

[4]:

# The coordinate reference system to use (NAD83 Albers CONUS)
crs = "EPSG:5070"
# Spatial resolution in CRS units (meters)
scale = 20_000

The wx accessor extends images and image collections with additional functionality including to_xarray. We’ll use that to download the collection, storing it in memory as an xarray.Dataset.

[5]:

arr = collection.wx.to_xarray(scale=scale, crs=crs)

Tip: If the download fails due to a DownloadError, retry or increase the max_attempts argument.

The downloaded array has 6 time coordinates (1 for each image) and 16 variables (1 for each band).

[7]:

arr

[7]:

<xarray.Dataset>
Dimensions:  (time: 6, y: 154, x: 292)
Coordinates:
  * time     (time) datetime64[ns] 2020-09-05T06:00:00 ... 2020-09-10T06:00:00
  * y        (y) float64 3.27e+06 3.25e+06 3.23e+06 ... 2.5e+05 2.3e+05 2.1e+05
  * x        (x) float64 -2.91e+06 -2.89e+06 -2.87e+06 ... 2.89e+06 2.91e+06
Data variables: (12/16)
    pr       (time, y, x) float32 nan nan nan nan nan ... nan nan nan nan nan
    rmax     (time, y, x) float32 nan nan nan nan nan ... nan nan nan nan nan
    rmin     (time, y, x) float32 nan nan nan nan nan ... nan nan nan nan nan
    sph      (time, y, x) float32 nan nan nan nan nan ... nan nan nan nan nan
    srad     (time, y, x) float32 nan nan nan nan nan ... nan nan nan nan nan
    th       (time, y, x) float32 nan nan nan nan nan ... nan nan nan nan nan
    ...       ...
    eto      (time, y, x) float32 nan nan nan nan nan ... nan nan nan nan nan
    bi       (time, y, x) float32 nan nan nan nan nan ... nan nan nan nan nan
    fm100    (time, y, x) float32 nan nan nan nan nan ... nan nan nan nan nan
    fm1000   (time, y, x) float32 nan nan nan nan nan ... nan nan nan nan nan
    etr      (time, y, x) float32 nan nan nan nan nan ... nan nan nan nan nan
    vpd      (time, y, x) float32 nan nan nan nan nan ... nan nan nan nan nan
Attributes:
    transform:               (20000.0, 0.0, -2920000.0, 0.0, -20000.0, 328000...
    crs:                     +init=epsg:5070
    res:                     (20000.0, 20000.0)
    is_tiled:                1
    nodatavals:              (-32768.0,)
    scales:                  (1.0,)
    offsets:                 (0.0,)
    AREA_OR_POINT:           Area
    TIFFTAG_RESOLUTIONUNIT:  1 (unitless)
    TIFFTAG_XRESOLUTION:     1
    TIFFTAG_YRESOLUTION:     1

Visualizing

Let’s calculate the max wind velocity over the time dimension and plot it.

[8]:

arr.vs.max("time").plot(figsize=(10, 5), cmap="magma")

[8]:

<matplotlib.collections.QuadMesh at 0x7fa51c759f40>
../_images/examples_image_collection_to_xarray_18_1.png

Going a little further, let’s calculate the Hot-Dry-Windy (HDW) index and see how it changed over the 6 days.

[9]:

hdw = arr.vs * arr.vpd

hdw.plot(col="time", col_wrap=3, cmap="magma", figsize=(15, 6))

[9]:

<xarray.plot.facetgrid.FacetGrid at 0x7fa51c5b0cd0>
../_images/examples_image_collection_to_xarray_20_1.png

Note

This page was auto-generated from a Jupyter notebook. For full functionality, download the notebook from Github and run it in a local Python environment.