Usage
This section explains how to use the Cube Builder application to generate data cubes from Sentinel 2, Landsat 8 and CBERS collections using API REST interactions. You may consider to install the Data Cube Manager which provides a GUI for data cube creation. Keep in mind that it must follows the compatibility table to work properly.
If you have not read yet how to install or deploy the system, please refer to Installation or Deploying documentation. You may also read the Brazil Data Cube Documentation for further details about Brazil Data Cube Project.
Temporal Compositing Functions
Before create any data cube, we strongly recommend you to read the Temporal Compositing
to understand how these functions are defined in Cube-Builder. In short, we have the following functions supported:
Average(AVG): consist in the average of the observed values.Median(MED): consist in the median value of the observations.Least Cloud Cover First(LCF): consists in aggregating pixels from all images in the time interval according to each image quantity of valid pixels considering the image cloud cover efficacy.
Creating a Grid for the Data Cubes
A Data Cube must have an associated grid as mentioned in BDC Grid. For this example, we will create 3 hierarchical grids for different data cubes:
BRAZIL_SM: used by collectionSentinel-2which has resolution10 meters;BRAZIL_MD: used by collectionLandsat-8which has resolution30 meters;BRAZIL_LG: used by collectionCBERSwhich has resolution64 meters(or55 metersforCBERS4A).
In order to do this, we must understand a few concepts:
names: A list consisting the Grid names to generate.tile_factor: An ordered list of tile factor for the hierarchical grid scaling. We recommend you to use divisible values like 10, 20 and 40.shape: A pivot shape to represent the tile factor. It usually consists in a desirable image shape for grid.meridian: The central pivot for first grid. This will be reference/center value used to scale the other grids.
Note
Keep in mind that the values shape and tile_factor are int values representing in meters.
Note
Make sure that names and tile_factor MUST HAVE the same dimension. Each entry consists in
a individual grid.
To create the new grids use the API Resource /create-grids as following:
curl --location \
--request POST '127.0.0.1:5000/create-grids' \
--header 'Content-Type: application/json' \
--data-raw '{
"names": [
"BRAZIL_SM",
"BRAZIL_MD",
"BRAZIL_LG"
],
"description": "Brazil Grids - Albers Equal Area",
"projection": "aea",
"meridian": -54,
"tile_factor": [
[10, 10],
[20, 20],
[40, 40]
],
"shape": [10560, 10560],
"bbox": [-73.98318215899995, -33.75117799399993, -28.847770352999916, 5.269580833000035],
"srid": 100001
}'
The response will have status code 201 and the body:
"Grids ['BRAZIL_SM', 'BRAZIL_MD', 'BRAZIL_LG'] created with successfully"
Note
You may create non-hierarchical Grid, just specify your own arguments like shape, bbox, shape and use single name and tile_factor:
curl --location \
--request POST '127.0.0.1:5000/create-grids' \
--header 'Content-Type: application/json' \
--data-raw '{
"names": [
"BRAZIL_SM"
],
"description": "Brazil Grids - Albers Equal Area",
"projection": "aea",
"meridian": -54,
"tile_factor": [
[10, 10]
],
"shape": [10560, 10560],
"bbox": [-73.98318215899995, -33.75117799399993, -28.847770352999916, 5.269580833000035],
"srid": 100001
}'
Creating data cube Landsat-8
In order to create data cube Landsat-8 monthly using the composite function Least Cloud Cover First (LC8-1M), use the following command to create data cube metadata:
curl --location \
--request POST '127.0.0.1:5000/cubes' \
--header 'Content-Type: application/json' \
--data-raw '
{
"datacube": "LC8-1M",
"datacube_identity": "LC8",
"grs": "BRAZIL_MD",
"title": "Landsat-8 (OLI) Cube Monthly - v001",
"resolution": 30,
"version": 1,
"metadata": {
"license": "proprietary",
"platform": {
"code": "Landsat-8",
"instruments": "OLI"
}
},
"temporal_composition": {
"schema": "Continuous",
"step": 1,
"unit": "month"
},
"composite_function": "LCF",
"bands_quicklook": [
"sr_band4",
"sr_band3",
"sr_band2"
],
"bands": [
{"name": "sr_band1", "common_name": "coastal", "data_type": "int16", "nodata": -9999},
{"name": "sr_band2", "common_name": "blue", "data_type": "int16", "nodata": -9999},
{"name": "sr_band3", "common_name": "green", "data_type": "int16", "nodata": -9999},
{"name": "sr_band4", "common_name": "red", "data_type": "int16", "nodata": -9999},
{"name": "sr_band5", "common_name": "nir", "data_type": "int16", "nodata": -9999},
{"name": "sr_band6", "common_name": "swir1", "data_type": "int16", "nodata": -9999},
{"name": "sr_band7", "common_name": "swir2", "data_type": "int16", "nodata": -9999},
{"name": "Fmask4", "common_name": "quality", "data_type": "uint8", "nodata": 255}
],
"indexes": [
{
"name": "EVI",
"common_name": "evi",
"data_type": "int16",
"nodata": -9999,
"metadata": {
"expression": {
"bands": ["sr_band5", "sr_band4", "sr_band2"],
"value": "(10000. * 2.5 * (sr_band5 - sr_band4) / (sr_band5 + 6. * sr_band4 - 7.5 * sr_band2 + 10000.))"
}
}
},
{
"name": "NDVI",
"common_name": "ndvi",
"data_type": "int16",
"nodata": -9999,
"metadata": {
"expression": {
"bands": ["sr_band5", "sr_band4"],
"value": "10000. * ((sr_band5 - sr_band4)/(sr_band5 + sr_band4))"
}
}
}
],
"quality_band": "Fmask4",
"description": "This datacube contains the all available images from Landsat-8, with 30 meters of spatial resolution, reprojected and cropped to BDC_MD grid, composed each 16 days using the best pixel (LCF) composite function.",
"parameters": {
"mask": {
"clear_data": [0, 1],
"not_clear_data": [2, 3, 4],
"nodata": 255,
"saturated_data": []
}
}
}'
Note
If you would like to create a data cube with temporal composition with 16 days which reset the time line per year, you may change the JSON key temporal_composition:
..
"temporal_composition": {
"schema": "Cyclic",
"step": 16,
"unit": "day",
"cycle": {
"step": 1,
"unit": "year"
}
}
..
The property mask inside parameters represents how the Cube Builder will deal with Clear Data and Not Clear Data pixels.
The Clear Data pixels are considered to identify the Best Pixel (LCF) and it is count on the Clear Observation Band (ClearOb).
In order to trigger a data cube, we are going to use a collection LC8_SR-1 made with Surface Reflectance using LaSRC 2.0 with cloud masking Fmask 4.2. In this example, we are going to use the official Brazil Data Cube STAC. To do so, you will need to have an account in Brazil Data Cube environment. If you don’t have any account, please, refer to Brazil Data Cube Explorer.
Once the data cube definition is created, you can trigger a data cube using the following command:
SQLALCHEMY_DATABASE_URI="postgresql://postgres:postgres@localhost/bdc" \
cube-builder build LC8-1M-1 \
--stac-url https://brazildatacube.dpi.inpe.br/stac/ \
--collections=LC8_SR-1 \
--tiles=011009 \
--start=2019-01-01 \
--end=2019-01-31 \
--token <USER_BDC_TOKEN>
Note
If you would like to trigger data cube generation using API call instead commandline use as following:
# Using curl (Make sure to execute cube-builder run)
curl --location \
--request POST '127.0.0.1:5000/start' \
--header 'Content-Type: application/json' \
--data-raw '{
"stac_url": "https://brazildatacube.dpi.inpe.br/stac/",
"token": "<USER_BDC_TOKEN>",
"datacube": "LC8-1M",
"collections": ["LC8_SR-1"],
"tiles": ["011009"],
"start_date": "2019-01-01",
"end_date": "2019-01-31"
}'
Note
The command line cube-builder build has few optional parameters such
bands, which defines bands to generate data cube.
You may also pass --stac-url=URL_TO_STAC (command line) or "stac_url": "URL_TO_STAC" (API only)
if you would like to generate data cube using a different STAC provider. Remember that the --collection must exists.
Creating data cube Sentinel 2
In order to create data cube Sentinel 2, use the following command to create data cube metadata:
curl --location \
--request POST '127.0.0.1:5000/cubes' \
--header 'Content-Type: application/json' \
--data-raw '
{
"datacube": "S2-16D",
"datacube_identity": "S2",
"grs": "BRAZIL_SM",
"title": "Sentinel-2 SR - Cube LCF 16 days -v001",
"resolution": 10,
"version": 1,
"metadata": {
"license": "proprietary",
"platform": {
"code": "Sentinel-2",
"instruments": "MSI"
}
},
"temporal_composition": {
"schema": "Cyclic",
"step": 16,
"unit": "day",
"cycle": {
"unit": "year",
"step": 1
}
},
"composite_function": "LCF",
"bands_quicklook": [
"B04",
"B03",
"B02"
],
"bands": [
{"name": "B01", "common_name": "coastal", "data_type": "int16", "nodata": 0},
{"name": "B02", "common_name": "blue", "data_type": "int16", "nodata": 0},
{"name": "B03", "common_name": "green", "data_type": "int16", "nodata": 0},
{"name": "B04", "common_name": "red", "data_type": "int16", "nodata": 0},
{"name": "B05", "common_name": "rededge", "data_type": "int16", "nodata": 0},
{"name": "B06", "common_name": "rededge", "data_type": "int16", "nodata": 0},
{"name": "B07", "common_name": "rededge", "data_type": "int16", "nodata": 0},
{"name": "B08", "common_name": "nir", "data_type": "int16", "nodata": 0},
{"name": "B8A", "common_name": "nir08", "data_type": "int16", "nodata": 0},
{"name": "B11", "common_name": "swir16", "data_type": "int16", "nodata": 0},
{"name": "B12", "common_name": "swir22", "data_type": "int16", "nodata": 0},
{"name": "SCL", "common_name": "quality","data_type": "uint8", "nodata": 0}
],
"indexes": [
{
"name": "EVI",
"common_name": "evi",
"data_type": "int16",
"nodata": -9999,
"metadata": {
"expression": {
"bands": [
"B8A",
"B04",
"B02"
],
"value": "(10000. * 2.5 * (B8A - B04) / (B8A + 6. * B04 - 7.5 * B02 + 10000.))"
}
}
},
{
"name": "NDVI",
"common_name": "ndvi",
"data_type": "int16",
"nodata": -9999,
"metadata": {
"expression": {
"bands": [
"B8A",
"B04"
],
"value": "10000. * ((B8A - B04)/(B8A + B04))"
}
}
}
],
"quality_band": "SCL",
"description": "This data cube contains all available images from Sentinel-2, resampled to 10 meters of spatial resolution, reprojected, cropped and mosaicked to BDC_SM grid and time composed each 16 days using LCF temporal composition function.",
"parameters": {
"mask": {
"clear_data": [4, 5, 6],
"not_clear_data": [2, 3, 7, 8, 9, 10, 11],
"nodata": 0,
"saturated_data": [1]
}
}
}'
In order to trigger a data cube, we are going to use a collection S2-16-1 made with Surface Reflectance using Sen2Cor:
# Using cube-builder command line
SQLALCHEMY_DATABASE_URI="postgresql://postgres:postgres@localhost/bdc" \
cube-builder build S2-16D-1 \
--stac-url https://brazildatacube.dpi.inpe.br/stac/ \
--collections=S2_L2A-1 \
--tiles=017019 \
--start=2019-01-01 \
--end=2019-01-31 \
--token <USER_BDC_TOKEN>
Creating data cube CBERS-4 AWFI
In order to create data cube CBERS4 AWFI, use the following command to create data cube metadata:
curl --location \
--request POST '127.0.0.1:5000/cubes' \
--header 'Content-Type: application/json' \
--data-raw '
{
"datacube": "CB4-16D",
"datacube_identity": "CB4",
"grs": "BRAZIL_LG",
"title": "CBERS-4 (AWFI) SR - Data Cube LCF 16 days - v001",
"resolution": 64,
"version": 1,
"metadata": {
"license": "cc-by-sa-3.0",
"platform": {
"code": "CBERS-4",
"instruments": "AWFI"
}
},
"temporal_composition": {
"schema": "Cyclic",
"step": 16,
"unit": "day",
"cycle": {
"unit": "year",
"step": 1
}
},
"composite_function": "LCF",
"bands_quicklook": [
"BAND15",
"BAND14",
"BAND13"
],
"bands": [
{"name": "BAND13", "common_name": "blue", "data_type": "int16", "nodata": -9999},
{"name": "BAND14", "common_name": "green", "data_type": "int16", "nodata": -9999},
{"name": "BAND15", "common_name": "red", "data_type": "int16", "nodata": -9999},
{"name": "BAND16", "common_name": "nir", "data_type": "int16", "nodata": -9999},
{"name": "CMASK", "common_name": "quality","data_type": "uint8", "nodata": 0}
],
"indexes": [
{
"name": "EVI",
"common_name": "evi",
"data_type": "int16",
"nodata": -9999,
"metadata": {
"expression": {
"bands": [
"BAND16",
"BAND15",
"BAND13"
],
"value": "(10000. * 2.5 * (BAND16 - BAND15) / (BAND16 + 6. * BAND15 - 7.5 * BAND13 + 10000.))"
}
}
},
{
"name": "NDVI",
"common_name": "ndvi",
"data_type": "int16",
"nodata": -9999,
"metadata": {
"expression": {
"bands": [
"BAND16",
"BAND15"
],
"value": "10000. * ((BAND16 - BAND15)/(BAND16 + BAND15))"
}
}
}
],
"quality_band": "CMASK",
"description": "This data cube contains the all available images from CBERS-4/AWFI resampled to 64 meters of spatial resolution, reprojected and cropped to BDC_LG grid, composed each 16 days using the best pixel (LCF) composite function.",
"parameters": {
"mask": {
"clear_data": [127],
"not_clear_data": [255],
"nodata": 0,
"saturated_data": []
}
}
}'
Trigger data cube generation with following command:
# Using cube-builder command line
SQLALCHEMY_DATABASE_URI="postgresql://postgres:postgres@localhost/bdc" \
cube-builder build CB4-16D-1 \
--stac-url https://brazildatacube.dpi.inpe.br/stac/ \
--collections=CBERS4_AWFI_L4_SR \
--tiles=005004 \
--start=2019-01-01 \
--end=2019-01-31 \
--token <USER_BDC_TOKEN>
Restarting or Reprocessing a Data Cube
When the Cube-Builder could not generate data cube for any unknown issue, you may restarting the entire process
with the same command you have dispatched:
SQLALCHEMY_DATABASE_URI="postgresql://postgres:postgres@localhost/bdc" \
cube-builder build CB4-16D-1 \
--stac-url https://brazildatacube.dpi.inpe.br/stac/ \
--collections=CBERS4_AWFI_L4_SR \
--tiles=005004 \
--start=2019-01-01 \
--end=2019-01-31 \
--token <USER_BDC_TOKEN>
It will reuse most of files that were already processed, executing only the failed tasks. If you notice anything suspicious or want to re-create theses files again, use the option --force:
SQLALCHEMY_DATABASE_URI="postgresql://postgres:postgres@localhost/bdc" \
cube-builder build CB4-16D-1 \
--stac-url https://brazildatacube.dpi.inpe.br/stac/ \
--collections=CBERS4_AWFI_L4_SR \
--tiles=005004 \
--start=2019-01-01 \
--end=2019-01-31 \
--token <USER_BDC_TOKEN> \
--force
Data Cube Parameters
The Cube-Builder supports a few parameters to be set during the data cube execution.
In order to check the parameters associated with data cube CB4-16D-1, use the command:
SQLALCHEMY_DATABASE_URI="postgresql://postgres:postgres@localhost/bdc" \
cube-builder show-parameters CB4-16D-1
The following output represents all the parameters related with the given data cube:
mask -> {'clear_data': [127], 'not_clear_data': [255], 'nodata': 0}
quality_band -> CMASK
stac_url -> https://brazildatacube.dpi.inpe.br/stac/
token -> ChangeME
You can change any parameter with the command cube-builder configure with DataCubeName-Version:
SQLALCHEMY_DATABASE_URI="postgresql://postgres:postgres@localhost/bdc" \
cube-builder configure CB4-16D-1 --stac-url=AnySTAC
Note
Once parameter is set, it only be affected in the new execution.
Be aware of what you are changing to do not affect the integrity of data cube.
For example, changing the masking clear_data when there is a already area generated.
Make sure to re-generate all the periods and tiles again.
List of supported values:
sentinel_safe(STRING): Flag to determine the usage of cube builder and generation data cube from Sentinel-2 Safe Format. Use to specify which key described in STAC Item Asset contains thezipfile. Defaults toNone.channel_limits(LIST): Customize the range limit for thumbnail files. Defaults toNone.no_post_process(BOOL): Flag to skip data cube post processing step. Defaults toFalse.reference_day(INT): Value representing a day used as priority forLeast Cloud Cover(LCF) algorithm. This value can be used to improve day gap between the best pixels.skip_vi_identity(BOOL): Flag to skip data cube identity vegetation index generation. Defaults toFalse.
Advanced User Guide
Generate data cubes from local dir
New in version 1.0.0.
Note
To proceed this step, you will need to have a set of files in disk. We will not provide this files since its just a briefing of this feature. You may consider to have own files individually.
With latest change of Cube-Builder (1.0), the user can generate data cubes using local directories containing
images. This feature is useful to generate data cubes when the user has a bunch of image files locally and would like
to apply temporal composition function over these files. In this case, a STAC Server is not required.
This feature can be achieved using parameters --local DIRECTORY and --format PATH_TO_FORMAT.json.
It follows the signature of GDALCubes Formats to read
directories.
Essentially, a format contains the following properties:
images(REQUIRED): Object context representing how to seek for any image in disk.pattern(REQUIRED): Regex value for data discovering.
datetime(REQUIRED): Object context describing how to identify data times from any directory path or file path.pattern(REQUIRED): A regex expression describing how to match datetime.format(REQUIRED): ISO Format to get data time from str.
bands(REQUIRED): The data set bands that will be captured while recurring disk. You can also add extra fields to increment metadata of band. The following internal props are required:pattern: Regex pattern to identify band in disk.nodata: No data value for band.
tags(OPTIONAL): List of keywords describing the given format.description(OPTIONAL): A detailed multi-line description to fully explain the format.
You can check a minimal example in examples/formats/bdc-sentinel-2-l2a-cogs.json, which offers support to
locate Sentinel-2 Cloud Optimized GeoTIFF files. You may also take a look in GDALCubes Formats
for others formats.
For this example, lets create a simple sentinel-2 data cube called S2-LOCAL-16D. The signature is similar from
Creating data cube Sentinel 2. We just need to change the cube parameters to something like:
...
"parameters": {
"mask": {
"clear_data": [4, 5, 6],
"not_clear_data": [2, 3, 7, 8, 9, 10, 11],
"nodata": 0,
"saturated_data": [1]
},
"local": "/path/to/local/files",
"recursive": true,
"format": "examples/formats/bdc-sentinel-2-l2a-cogs.json",
"pattern": ".tif"
}
So you can create a data cube with command:
curl --location \
--request POST '127.0.0.1:5000/cubes' \
--header 'Content-Type: application/json' \
--data-raw '
{
"datacube": "S2-LOCAL-16D",
"datacube_identity": "S2-LOCAL",
"grs": "BRAZIL_SM",
"title": "Sentinel-2 SR - Cube LCF 16 days -v001",
"resolution": 10,
"version": 1,
"metadata": {
"license": "MIT",
"platform": {
"code": "Sentinel-2",
"instruments": "MSI"
}
},
"temporal_composition": {
"schema": "Cyclic",
"step": 16,
"unit": "day",
"cycle": {
"unit": "year",
"step": 1
}
},
"composite_function": "LCF",
"bands_quicklook": [
"B04",
"B03",
"B02"
],
"bands": [
{"name": "B01", "common_name": "coastal", "data_type": "int16", "nodata": 0},
{"name": "B02", "common_name": "blue", "data_type": "int16", "nodata": 0},
{"name": "B03", "common_name": "green", "data_type": "int16", "nodata": 0},
{"name": "B04", "common_name": "red", "data_type": "int16", "nodata": 0},
{"name": "B05", "common_name": "rededge", "data_type": "int16", "nodata": 0},
{"name": "B06", "common_name": "rededge", "data_type": "int16", "nodata": 0},
{"name": "B07", "common_name": "rededge", "data_type": "int16", "nodata": 0},
{"name": "B08", "common_name": "nir", "data_type": "int16", "nodata": 0},
{"name": "B8A", "common_name": "nir08", "data_type": "int16", "nodata": 0},
{"name": "B11", "common_name": "swir16", "data_type": "int16", "nodata": 0},
{"name": "B12", "common_name": "swir22", "data_type": "int16", "nodata": 0},
{"name": "SCL", "common_name": "quality","data_type": "uint8", "nodata": 0}
],
"indexes": [
{
"name": "EVI",
"common_name": "evi",
"data_type": "int16",
"nodata": -9999,
"metadata": {
"expression": {
"bands": [
"B8A",
"B04",
"B02"
],
"value": "(10000. * 2.5 * (B8A - B04) / (B8A + 6. * B04 - 7.5 * B02 + 10000.))"
}
}
},
{
"name": "NDVI",
"common_name": "ndvi",
"data_type": "int16",
"nodata": -9999,
"metadata": {
"expression": {
"bands": [
"B8A",
"B04"
],
"value": "10000. * ((B8A - B04)/(B8A + B04))"
}
}
}
],
"quality_band": "SCL",
"description": "This data cube contains all available images from Sentinel-2, resampled to 10 meters of spatial resolution, reprojected, cropped and mosaicked to BDC_SM grid and time composed each 16 days using LCF temporal composition function.",
"parameters": {
"mask": {
"clear_data": [4, 5, 6],
"not_clear_data": [2, 3, 7, 8, 9, 10, 11],
"nodata": 0,
"saturated_data": [1]
},
"local": "/path/to/local/files",
"recursive": true,
"format": "examples/formats/bdc-sentinel-2-l2a-cogs.json",
"pattern": ".tif"
}
}'
After cube definition created, you can just use the command line cube-builder build-local:
SQLALCHEMY_DATABASE_URI="postgresql://postgres:postgres@localhost/bdc" \
cube-builder build-local S2-LOCAL-16D-1 \
--tiles 003011 \
--start-date 2021-08-29 \
--end-date 2021-09-13 \
--directory /path/to/local/files \
--format examples/formats/bdc-sentinel-2-l2a-cogs.json
Note
This example just illustrate how to trigger the data cube using local directory. You may need to
change these values like directory, format, start-date, end-date and tiles.
Right now, it only supports using --tiles as parameter. It will be replaced in the next release
to support any Region of Interest (ROI) or shapefile.