Skip to main content

ECOSTRESS Collection 3 Level 2 STARS Vegetation Index & Albedo

This is the main repository for the ECOsystem Spaceborne Thermal Radiometer Experiment on Space Station (ECOSTRESS) collection 3 level 2 STARS NDVI and albedo data product. This product will utilize the Spatial Timeseries for Automated high-Resolution multi-Sensor (STARS) data fusion system to produce normalized difference vegetation index (NDVI) and albedo estimates corresponding to ECOSTRESS surface temperature measurements, to support the evapotranspiration product.

The ECOSTRESS collection 3 level 2 vegetation index and albedo data product is the pre-cursor to the Surface Biology and Geology (SBG) collection 1 level 2 vegetation index and albedo data product algorithm.

Gregory H. Halverson (they/them)
gregory.h.halverson@jpl.nasa.gov
NASA Jet Propulsion Laboratory 329G

Margaret C. Johnson (she/her)
maggie.johnson@jpl.nasa.gov
NASA Jet Propulsion Laboratory 398L

Evan Davis (he/him)
evan.w.davis@jpl.nasa.gov
NASA Jet Propulsion Laboratory 397K

Kerry Cawse-Nicholson (she/her)
kerry-anne.cawse-nicholson@jpl.nasa.gov
NASA Jet Propulsion Laboratory 329G

Claire Villanueva-Weeks (she/her)
claire.s.villanueva-weeks@jpl.nasa.gov
NASA Jet Propulsion Laboratory 329G

The code for the ECOSTRESS level 2 STARS PGE will be developed using open-science practices based on the ECOSTRESS collection 2 gridded and tiled product generation software.

This software will produce estimates of:

  • Normalized Difference Vegetation Index (NDVI)
  • albedo

NDVI and albedo are estimated at 70 m ECOSTRESS standard resolution for each daytime ECOSTRESS overpass by fusing temporally sparse but fine spatial resolution images from the Harmonized Landsat Sentinel (HLS) 2.0 product with daily, moderate spatial resolution images from VIIRS. The VCD43NRT package uses Suomi NPP first, with JPSS-1 and JPSS-2 available as mission fallbacks, to provide those VIIRS inputs. The data fusion is performed using a variant of the Spatial Timeseries for Automated high-Resolution multi-Sensor data fusion (STARS) algorithm developed by Dr. Margaret Johnson and Gregory Halverson at the Jet Propulsion Laboratory. STARS is a Bayesian timeseries methodology that provides streaming data fusion and uncertainty quantification through efficient Kalman filtering.

Operationally, each L2T STARS tile run loads the means and covariances of the STARS model saved from the most recent tile run, then iteratively advances the means and covariances forward each day updating with fine imagery from HLS and/or moderate resolution imagery from VIIRS up to the day of the target ECOSTRESS overpass. VCD43NRT applies the near-real-time BRDF workflow to daily VIIRS observations to produce the coarse NDVI and albedo inputs.

1. Introduction to Data Products

The data format for the ECOSTRESS products is described in the ECOSTRESS Collection 3 landing page.

2. L2T STARS NDVI and Albedo Product

flowchart TB
    subgraph VCD43NRT[VCD43NRT Python/Rust]
        VIIRS_I[VIIRS<br>I-Band<br>Surface<br>Reflectance]
        VIIRS_M[VIIRS<br>M-Band<br>Surface<br>Reflectance]
        VIIRS_downscaling[VIIRS<br>Downscaling]
        VIIRS_downscaled[Downscaled<br>500m<br>VIIRS<br>Surface<br>Reflectance]
        VIIRS_BRDF[VCD43NRT<br>BRDF<br>Correction]
        VIIRS_corrected[VIIRS<br>BRDF-Corrected<br>500m<br>Surface<br>Reflectance]
        VIIRS_NDVI[VIIRS<br>500m<br>NDVI]
        VIIRS_albedo[VIIRS<br>500m<br>Albedo]
    end

    subgraph HLS_aquisition[HLS.jl]
        direction TB
        Landsat_reflectance[HLS<br>Landsat<br>30m<br>Surface<br>Reflectance]
        Landsat_upsampled[Upsampled<br>Landsat<br>70m<br>Surface<br>Reflectance]
        Landsat_NDVI[Landsat<br>70m<br>NDVI]
        Sentinel_reflectance[HLS<br>Sentinel<br>30m<br>Surface<br>Reflectance]
        Sentinel_upsampled[Upsampled<br>Sentinel<br>70m<br>Surface<br>Reflectance]
        Sentinel_NDVI[Sentinel<br>70m<br>NDVI]
        Landsat_albedo[Landsat<br>70m<br>Albedo]
        Sentinel_albedo[Sentinel<br>70m<br>Albedo]
    end

    subgraph bayesian_state[Bayesian State]
        NDVI_covariance_prior[NDVI<br>Fine-Coarse<br>Covariance<br>Prior<br>from<br>Previous<br>Overpass]
        NDVI_covariance_posterior[NDVI<br>Fine-Coarse<br>Covariance<br>Posterior<br>for<br>Next<br>Overpass]
        albedo_covariance_prior[Albedo<br>Fine-Coarse<br>Covariance<br>Prior<br>from<br>Previous<br>Overpass]
        albedo_covariance_posterior[Albedo<br>Fine-Coarse<br>Covariance<br>Posterior<br>for<br>Next<br>Overpass]
    end

    fine_NDVI_input[NDVI<br>70m<br>Composite]
    NDVI_data_fusion[STARS.jl<br>NDVI<br>Data<br>Fusion]
    fine_NDVI_output[Fused<br>30m<br>NDVI]
    fine_NDVI_uncertainty[NDVI<br>Uncertainty]

    fine_albedo_input[Albedo<br>70m<br>Composite]
    albedo_data_fusion[STARS.jl<br>Albedo<br>Data<br>Fusion]
    fine_albedo_output[Fused<br>30m<br>Albedo]
    fine_albedo_uncertainty[Albedo<br>Uncertainty]

    SBG_L2T_STARS(ECOSTRESS<br>L2T<br>STARS<br>NDVI<br>&<br>Albedo<br>Product)

    VIIRS_I --> VIIRS_downscaling
    VIIRS_M --> VIIRS_downscaling
    VIIRS_downscaling --> VIIRS_downscaled
    VIIRS_downscaled --> VIIRS_BRDF
    VIIRS_BRDF --> VIIRS_corrected
    VIIRS_corrected --> VIIRS_NDVI
    VIIRS_corrected --> VIIRS_albedo

    Landsat_reflectance --> Landsat_upsampled
    Sentinel_reflectance --> Sentinel_upsampled

    Landsat_upsampled --> Landsat_NDVI
    Sentinel_upsampled --> Sentinel_NDVI

    Landsat_upsampled --> Landsat_albedo
    Sentinel_upsampled --> Sentinel_albedo

    Landsat_NDVI --> fine_NDVI_input
    Sentinel_NDVI --> fine_NDVI_input
    fine_NDVI_input --> NDVI_data_fusion
    VIIRS_NDVI --> NDVI_data_fusion
    NDVI_covariance_prior --> NDVI_data_fusion
    NDVI_data_fusion --> fine_NDVI_output
    NDVI_data_fusion --> fine_NDVI_uncertainty
    NDVI_data_fusion --> NDVI_covariance_posterior

    Landsat_albedo --> fine_albedo_input
    Sentinel_albedo --> fine_albedo_input
    fine_albedo_input --> albedo_data_fusion
    VIIRS_albedo --> albedo_data_fusion
    albedo_covariance_prior --> albedo_data_fusion
    albedo_data_fusion --> fine_albedo_output
    albedo_data_fusion --> fine_albedo_uncertainty
    albedo_data_fusion --> albedo_covariance_posterior

    fine_NDVI_output --> SBG_L2T_STARS
    fine_NDVI_uncertainty --> SBG_L2T_STARS
    fine_albedo_output --> SBG_L2T_STARS
    fine_albedo_uncertainty --> SBG_L2T_STARS

    click VIIRS_BRDF "https://github.com/JPL-Evapotranspiration-Algorithms/VCD43NRT"
    click NDVI_data_fusion "https://github.com/STARS-Data-Fusion/STARS.jl"
    click albedo_data_fusion "https://github.com/STARS-Data-Fusion/STARS.jl"

Figure 1. Flowchart of the ECOSTRESS Collection 3 L2T STARS processing workflow.

NDVI and albedo are estimated at 70 m ECOSTRESS standard resolution with uncertainty for each UTC day in which there is an ECOSTRESS overpass by fusing temporally sparse but fine spatial resolution images from the Harmonized Landsat Sentinel (HLS) 2.0 product with daily, moderate spatial resolution VIIRS observations. Suomi NPP is preferred; JPSS-1 and JPSS-2 can fill dates without a preferred-mission observation.

Landsat and Sentinel surface reflectances are collected using the HLS.jl package.

VIIRS surface reflectance is processed by the VCD43NRT package. It selects observations per date using the configured mission priority (SNPP, JPSS-1, JPSS-2 by default), then applies the near-real-time BRDF workflow to produce VIIRS NDVI and albedo. HLS retrieval, STARS time windows, fusion, and ECOSTRESS output generation are unchanged.

The data fusion is performed with a variant of the Spatial Timeseries for Automated high-Resolution multi-Sensor data fusion (STARS) algorithm developed by Dr. Margaret Johnson and Gregory H. Halverson at the Jet Propulsion Laboratory using the STARS.jl package. STARS is a Bayesian timeseries methodology that provides streaming data fusion and uncertainty quantification through efficient Kalman filtering. Operationally, each L2T STARS tile run loads the means and covariances of the STARS model saved from the most recent tile run, then iteratively advances the means and covariances forward each day updating with fine imagery from HLS and/or moderate resolution imagery from VIIRS up to the day of the target ECOSTRESS overpass.

The layers of the L2T STARS product are listed in Table 2. All layers of this product are represented by 32-bit floating point arrays. The NDVI estimates and 1σ uncertainties (-UQ) are unitless from -1 to 1. The albedo estimates and 1σ uncertainties (-UQ) are proportions from 0 to 1.

Name Description Type Units Fill Value No Data Value Valid Min Valid Max Scale Factor Size
NDVI Normalized Difference Vegetation Index float32 Index NaN N/A -1 1 N/A 13.4 mb
NDVI-UQ Normalized Difference Vegetation Index Uncertainty float32 Index NaN N/A -1 1 N/A 13.4 mb
albedo Albedo float32 Ratio NaN N/A 0 1 N/A 13.4 mb
albedo-UQ Albedo Uncertainty float32 Ratio NaN N/A 0 1 N/A 13.4 mb

Table 2. Listing of L2T STARS data layers.

Prerequisites

This is a Python package that calls Julia code. Julia must be installed in order to run this package.

Authentication

This package requires an EarthData account and reads EarthData credentials from ~/.netrc in the following format:

machine urs.earthdata.nasa.gov
login <USERNAME>
password <PASSWORD>

Environment

On macOS, there are issues with installing pykdtree using pip, so it's better to use a mamba environment and install the pykdtree mamba package.

mamba create -y -n ECOv003-L2T-STARS -c conda-forge python=3.11 jupyter pykdtree 
mamba activate ECOv003-L2T-STARS

Installation

Install this package from PyPi using the name ECOv003-L2T-STARS with dashes:

pip install ECOv003-L2T-STARS

You can also install development versions of this package directly from a clone of this repository:

pip install .

Julia packages

pip does not install Julia. After pip install, run:

ECOv003-L2T-STARS-install-julia

Or call the same installer from Python:

import ECOv003_L2T_STARS

ECOv003_L2T_STARS.install_julia()

The installer finds the exact Julia version recorded in the packaged Manifest.toml (currently 1.11.6). If necessary, it installs Juliaup and that Julia version, then instantiates and precompiles the pinned Julia packages and checks the Julia imports. Repeat calls reuse an existing installation. This may download software and write to the user's Julia installation and depot; it does not change an existing Juliaup default or directory override. Installing Juliaup on Windows requires winget. The function raises an error if installation or verification fails.

To check an existing installation without installing anything, run ECOv003-L2T-STARS-install-julia --verify-only. Julia is used for STARS data fusion; VIIRS retrieval and BRDF processing use the Python/Rust VCD43NRT package.

Usage

Import this package with the name ECOv003_L2T_STARS with underscores:

import ECOv003_L2T_STARS

VCD43NRT example

To generate a VIIRS NRT granule for a date and MODLAND tile, edit the settings in example_VCD43NRT.py and run:

python example_VCD43NRT.py

The example requires Earthdata credentials for VIIRS downloads and access to GEOS5-FP data. Mission selection is performed per observation date. Set VCD43NRT_VJ209GA_CONCEPT_ID when the deployment needs a JPSS-2 CMR concept-ID override.

Command Line Usage

Command-Line Entry-Point for the ECOv003-L2T-STARS Product Generating Executable

ECOv003-L2T-STARS <runconfig> [--date YYYY-MM-DD] [--spinup-days DAYS] [--target-resolution METERS] [--ndvi-resolution METERS] [--albedo-resolution METERS] [--viirs-mission-priority SNPP,JPSS1,JPSS2] [--calibrate-fine] [--sources-only] [--no-remove-input-staging] [--no-remove-prior] [--no-remove-posterior] [--threads COUNT] [--num-workers COUNT] [--version]

Command-Line Entry-Point for the ECOv003-DL Product Generating Executable

ECOv003_DL <runconfig_filename> [--tiles TILE_ID [TILE_ID ...]] [--version]

References

Schaaf, C. B. et al. (2017). Algorithm Theoretical Basis Document for MODIS Bidirectional Reflectance Distribution Function and Albedo (MOD43) Products. NASA. Link to source

Metadata

Release files for ECOv003-L2T-STARS 1.15.0

For a detailed explanation of source distributions (sdists) and built distributions (wheels), please see the package formats documentation.

Source distribution (sdist)

Source distribution for ECOv003-L2T-STARS 1.15.0
File Size Uploaded Provenance Actions
ecov003_l2t_stars-1.15.0.tar.gz 53.0 MB Without provenance Details for ecov003_l2t_stars-1.15.0.tar.gz

Built distribution (wheel)

Table of built distributions (wheels) for ECOv003-L2T-STARS 1.15.0
File Interpreter ABI Platform Actions
ecov003_l2t_stars-1.15.0-py3-none-any.whl Python 3 none any Details for ecov003_l2t_stars-1.15.0-py3-none-any.whl

Total release size: 53.1 MB

Release files / ecov003_l2t_stars-1.15.0.tar.gz

Download URL ecov003_l2t_stars-1.15.0.tar.gz
Size 53.0 MB
Tags Source
SHA-256 checksum
How to use checksums
33b4534ec54b97558f3a6abe68dda3d1074a8b1bed292a081c0959333d380315
BLAKE2b-256 checksum
How to use checksums
2e996e5c37bcc39aa6b257c3f705baa56447c4e414a60ce3bc506282f394e143
Upload date
Uploaded using Trusted Publishing?
What is trusted publishing?
No
Uploaded via twine/6.2.0 CPython/3.9.25

Release files / ecov003_l2t_stars-1.15.0-py3-none-any.whl

Download URL ecov003_l2t_stars-1.15.0-py3-none-any.whl
Size 97.5 kB
Tags Python 3
SHA-256 checksum
How to use checksums
32a3a86ab8ded52e89c70a56c9a0214c9d7bbae6bb694d76a70329210c58bb3f
BLAKE2b-256 checksum
How to use checksums
2f2dfa2cb91634061a5f2ac86a606b37b7b89123b9f8e61527988d4b970fe0c4
Upload date
Uploaded using Trusted Publishing?
What is trusted publishing?
No
Uploaded via twine/6.2.0 CPython/3.9.25

Security for ECOv003-L2T-STARS 1.15.0

Provenance

Provenance describes where a file came from. On PyPI, provenance is shared via attestations, which provide a verifiable record of the build or publishing details. View details, limitations and caveats.

No provenance found

None of the files in this release were uploaded with provenance metadata.

A note on package security

PyPI is a repository for distributing packages, not a service for auditing them. While third-party security researchers and our staff do their best to detect and remove malicious packages, PyPI cannot guarantee the quality, security or suitability of any given package.

To learn more about protecting yourself from malicious actors, visit our package security documentation.

Anthropic, PBC Visionary sponsor Bloomberg Visionary sponsor Hudson River Trading Visionary sponsor Meta Visionary sponsor NVIDIA Visionary sponsor Microsoft Sustainability sponsor Depot Continuous Integration AWS Cloud computing and Security Sponsor Datadog Monitoring Fastly CDN Google Download Analytics Sentry Error logging StatusPage Status page