Data product
The EO workstream generated a data cube of the model outputs across Great Britain. The final data cube including predictions for the years 2016 to 2021 for 7 and 12 classes is available in two different flavours, both stored as a Zarr file (version 3) available from the links below.
Generic data cube:
https://s3.cl4.du.cesnet.cz/4f4743b6_4043_4e02_a3ba_0452aa7523a2:uscuni-public/eurofab/eurofab_predictions_datacube.zarrVector data cube:
https://s3.cl4.du.cesnet.cz/4f4743b6_4043_4e02_a3ba_0452aa7523a2:uscuni-public/eurofab/eurofab_predictions_vector_datacube.zarrThe size of each is approximately 1.8GB. See examples below for differences.
Metadata
The metadata YAML file describing the data of both:
name: urban_fabric_prediction
description: Urban fabric prediction from EO data
metadata:
product_type: urban_fabric_map
class_labels:
k7:
'0': Countryside agriculture
'1': Open sprawl
'2': Compact suburbia
'3': Urban
'4': Urban buffer
'5': Warehouse/Park land
'6': Wild countryside
k12:
'0': Accessible suburbia
'1': Connected residential neighbourhoods
'2': Countryside agriculture
'3': Dense residential neighbourhoods
'4': Dense urban neighbourhoods
'5': Disconnected suburbia
'6': Gridded residential quarters
'7': Open sprawl
'8': Urban buffer
'9': Urbanity
'10': Warehouse/Park land
'11': Wild countryside
cluster_numbers: [7, 12]
time_range:
start: 2016
end: 2021
measurements:
- name: prediction
dtype: int32
nodata: -1
units: "1"
aliases: [cluster]
- name: probabilities
dtype: float32
nodata: -9999
units: "1"
aliases: [class_probs]
storage:
crs: EPSG:4326The dataset is chunked and can be read only partially, if needed.
Example code
Basic data cube
The data can be read with xarray and similar tools.
import os
import xarray as xr
os.environ["ZARR_V3_EXPERIMENTAL_API"] = "1"
path = "https://s3.cl4.du.cesnet.cz/4f4743b6_4043_4e02_a3ba_0452aa7523a2:uscuni-public/eurofab/eurofab_predictions_datacube.zarr"
ds = xr.open_zarr(path, zarr_format=3)
ds<xarray.Dataset> Size: 4GB
Dimensions: (k: 2, obs: 3634816, year: 6, class: 12)
Coordinates:
* k (k) int64 16B 7 12
lon (obs) float64 29MB dask.array<chunksize=(2000,), meta=np.ndarray>
* obs (obs) int64 29MB 0 1 2 3 ... 3634812 3634813 3634814 3634815
lat (obs) float64 29MB dask.array<chunksize=(2000,), meta=np.ndarray>
geometry (obs) object 29MB dask.array<chunksize=(3634816,), meta=np.ndarray>
* year (year) int64 48B 2016 2017 2018 2019 2020 2021
Dimensions without coordinates: class
Data variables:
prediction (obs, year, k) int32 174MB dask.array<chunksize=(2000, 1, 1), meta=np.ndarray>
probabilities (obs, class, year, k) float64 4GB dask.array<chunksize=(2000, 12, 1, 1), meta=np.ndarray>Vector data cube
Vector data cube is encoded using CF conventions. The example code using xarray and xvec:
import os
import xarray as xr
import xvec
os.environ["ZARR_V3_EXPERIMENTAL_API"] = "1"
path = "https://s3.cl4.du.cesnet.cz/4f4743b6_4043_4e02_a3ba_0452aa7523a2:uscuni-public/eurofab/eurofab_predictions_vector_datacube.zarr"
ds = xr.open_zarr(path, zarr_format=3).xvec.decode_cf()
ds<xarray.Dataset> Size: 4GB
Dimensions: (geometry: 3634816, class: 12, year: 6, k: 2)
Coordinates:
* geometry (geometry) object 29MB POLYGON ((6220 901450, 6220 901700,...
* k (k) int64 16B 7 12
* year (year) int64 48B 2016 2017 2018 2019 2020 2021
Dimensions without coordinates: class
Data variables:
probabilities (geometry, class, year, k) float64 4GB dask.array<chunksize=(2000, 12, 1, 1), meta=np.ndarray>
prediction (geometry, year, k) int32 174MB dask.array<chunksize=(2000, 1, 1), meta=np.ndarray>
Indexes:
geometry GeometryIndex (crs=EPSG:27700)See the documentation of Xvec for more details on vector data cubes.
For details on the contents, see the relevant Technical notes.