National Aeronautics and Space Administration

Focus Area on Biophysical Product Validation

Luke Brown, University of Salford, UK
Richard Fernandes, Natural Resources Canada
Hao Tang, National University of Singapore 

 

LAI and fAPAR Definitions

Proposal for additional Vegetation Biophysical Parameters

Highest Validation Stage Currently Reached for Satellite-Derived LAI and fAPAR Products

Validation Good Practices

The LPV Subgroup has just published a Good Practices Protocol document for validation of LAI products. The document is available for download along with associated documents:

Fernandes, R., Plummer, S., Nightingale, J., Baret, F., Camacho, F., Fang, H., Garrigues, S., Gobron, N., Lang, M.,   Lacaze, R., LeBlanc, S., Meroni, M., Martinez, B., Nilson, T., Pinty, B., Pisek, J., Sonnentag, O., Verger, A., Welles, J.,   Weiss, M., & Widlowski, J.L. (2014). Global Leaf Area Index Product Validation Good Practices. Version 2.0.  In G.   Schaepman-Strub, M. Román, & J. Nickeson (Eds.), Good Practices for Satellite-Derived Land Product Validation (p.   76): Land Product Validation Subgroup (WGCV/CEOS), doi:10.5067/doc/ceoswgcv/lpv/lai.002

  (Download CEOS_LAI_PROTOCOL_Aug2014_v2.0.1.pdf, 5MB)

• Reviewer Responses
• Recommendations
• Annex I

Validation Reference Datasets and Tools

• DIRECT

• ImagineS

• GBOV

• FRM4VEG

• GROUNDED EO

• Other Datasets

• Tools

DIRECT V2.1

For validation of hectometric and kilometric products, the DIRECT V2.1 database compiles leaf area index (LAI), fraction of absorbed photosynthetically active radiation (FAPAR) and fraction of vegetation cover (FCOVER) averaged values over a 3 km x 3 km area for 176 sites. In situ reference measurements were upscaled using high spatial resolution imagery following CEOS WGCV LPV LAI validation good practices (Fernandes et al., 2014) to account for the spatial heterogeneity of the sites.  

Data included in the original DIRECT database compiled by Garrigues et al. (2008) were the result of several international activities, including Validation of Land European Remote Sensing Instruments (VALERI), BigFoot, Southern African Regional Science Initiative (SAFARI-2000), as well as campaigns conducted by the University of Alberta, Canada Centre for Remote Sensing (CCRS), Earth Observation Laboratory (EOLAB), Environmental Protection Agency (EPA), and European Space Agency (ESA). These were later ingested in the (no longer accessible) Online Validation Experiment (OLIVE) tool (Weiss et al., 2014) for accuracy assessment. The DIRECT sites were revised by Camacho et al. (2013), retaining only those with both overstory and understory measurements.. In an effort several years later, the sites were expanded and incorporated into a database with Implementation of Multiscale Agricultural Indicators Exploiting Sentinels (ImagineS) sites (Camacho et al., 2021). DIRECT V2.1 is the latest update, that includes 44 new sites from China (Fang et al., 2019; Song et al., 2021) and two new sites from ESA’s Fiducial Reference Measurements for Vegetation (FRM4VEG) programme (Brown et al., 2021).

DIRECT V2.1 constitutes a major effort of the international community to provide reference data for the validation of LAI and FAPAR products, with a total of and 280 LAI, 128 FAPAR, and 122 FCOVER values spanning the period from 2000 to 2021 and covering seven main biome types.

Whilst DIRECT V2.1 follows good practices for use and upscaling of in situ measurements, it is subject to limitations that should be acknowledged when used:

1. Many in situ reference measurements, and all those for forested plots, do not correct for woody area. This is likely to result in substantial overestimation of LAI, FAPAR, and FCOVER (Gower et al. 1999; Brown et al., 2024).
2. In situ reference measurements at some sites use approaches that do not (fully) correct for canopy clumping (Ryu et al., 2010). This is likely to result in an underestimation of LAI.
3. DIRECT V2.1 does not include quality layers corresponding to regions in each image represented by in situ measurements, as is recommended by good practice (Fernandes et al., 2014), or use FRM approaches to propagate in situ reference measurement uncertainties.
4. Only 3 km x 3 km averages (as opposed to upscaled reference maps at the native spatial resolution, and/or in situ reference measurements at the elementary sampling unit (ESU) scale) are provided. This limits utility for decametric product validation.

These limitations should be acknowledged when using DIRECT V2.1 until future studies are able to address them quantitatively.

Additional details and access to the DIRECT V2.1 database can be found on the dedicated CEOS Cal/Val Portal page.

Implementation of Multi-scale Agricultural Indicators Exploiting Sentinels (ImagineS)

Based on the success of the VALERI project, more recent validation activities have been conducted under the Implementation of Multiscale Agricultural Indicators Exploiting Sentinels (ImagineS) project, part of which focussed on the development and validation of 300 m LAI, FAPAR and FCOVER products for the Copernicus Global Land Service (CGLS). Following the VALERI methodology and sampling scheme, 46 campaigns were performed by a network of international partners over 20 sites between 2013 and 2016. In situ data collection protocols were followed by local teams, and as in the VALERI project, these data were processed centrally to derive 3 km x 3 km high spatial resolution reference maps from OLI data (Camacho et al., 2021). Both ESU scale in situ reference measurements and upscaled high spatial resolution reference maps are provided by ImagineS.

Whilst ImagineS follows good practices for use and upscaling of in situ reference measurements, it is subject to limitations that should be acknowledged when used:

1. Correction for woody area is not applied. This is likely to result in overestimation of LAI, FAPAR, and FCOVER for agricultural sites with substantial woody material (e.g. orchards).
2. In situ reference measurements at some sites use approaches that do not (fully) correct for canopy clumping (Ryu et al., 2010). This is likely to result in an underestimation of LAI.
3.  FRM approaches are not used propagate in situ reference measurement uncertainties.

These limitations should be acknowledged when using ImagineS until future studies are able to address them quantitatively.

Additional details and access to the ImagineS data can be found on the project website.

Ground Reference Observations for Validation (GBOV)

One of the most recent validation initiatives is the Ground Based Observations for Validation of Copernicus Global Land Products (GBOV) project, which was initiated by the European Commission’s Joint Research Centre. Its aim is to develop and distribute robust datasets for validation of satellite-derived land products. The project is primarily leveraging in situ measurements collected through recent environmental monitoring networks such as the National Ecological Observatory Network (NEON) in the United States, the Terrestrial Ecosystem Research Network (TERN) in Australia, and the Integrated Carbon Observation System (ICOS) in Europe, though it also plans to establish additional sites in underrepresented areas such as the tropics and semi-arid regions (Brown et al., 2020). In addition to LAI, FAPAR and FCOVER, other parameters, including soil moisture, albedo, and land surface temperature are also considered within the project. Both ESU scale in situ reference measurements and upscaled high spatial resolution reference maps are provided by GBOV.

Whilst GBOV follows good practices for use and upscaling of in situ measurements, it is subject to limitations that should be acknowledged when used:

1. Many in situ reference measurements, and all those for forested plots, do not correct for woody area. This is likely to result in overestimation of LAI, FAPAR, and FCOVER (Gower et al. 1999; Brown et al., 2024).
2. Calibrated radiative transfer model based retrievals are used for upscaling. This may reduce independence for products that make use of the same leaf and canopy radiative transfer models, i.e. Leaf Optical Properties Spectra (PROSPECT) and Scattering by Arbitrarily Inclined Leaves (SAIL).

These limitations should be acknowledged when using GBOV until future studies are able to address them quantitatively.

Additional details and access to the GBOV database can be found on the GBOVservice website.

Fiducial Reference Measurements for Vegetation (FRM4VEG)

The ESA-initiated FRM4VEG programme is focused on establishing the protocols required for traceable in situ measurements of vegetation-related parameters, to support the validation of Copernicus products from Sentinel-2, -3, and PROBA-V. Fiducial reference measurements are ‘the suite of independent ground measurements that provide the maximum return on investment for a satellite mission by delivering, to users, the required confidence in data products, in the form of independent validation results and satellite measurement uncertainty estimation, over the entire end-to-end duration of a satellite mission’.  They should:

• Have documented SI traceability (or conform to appropriate international community standards).
• Be independent from the satellite geophysical retrieval process.
• Be accompanied by an uncertainty budget for all instruments and derived measurements.
• Adhere to community-agreed, published and openly-available measurement protocols/procedures and management practices.
• Be accessible to other researchers allowing independent verification of processing systems.

In FRM4VEG Phases 1 and 2, campaigns were conducted at Las Tiesas – Barrax and Wytham Woods in 2018 and 2021, including the collection of LAI, FAPAR, and FCOVER in situ reference measurements (Brown et al., 2021; Camacho et al., 2024). Both ESU scale in situ reference measurements and upscaled high spatial resolution reference maps are available. Uniquely, these reference datasets are accompanied by uncertainty estimates, which were derived according to International Standards Organisation (ISO) Guide to the Expression of Uncertainty in Measurement (GUM) principles. This process involved quantifying the uncertainties associated with individual in situ reference measurements and incorporating these uncertainties within the upscaling procedure (as well as those associated with the high spatial resolution imagery used for upscaling).

Additional details and access to the FRM4VEG data can be found on the project website.

Ground Reference Observations Underlying Novel Decametric Vegetation Data Products from Earth Observation (GROUNDED EO)

The ESA-funded GROUNDED EO project (Brown et al., 2025) focussed on empirical data-driven retrieval of LAI and FAPAR from Sentinel-2, and involved the generation of an extensive fiducial reference database at decametric (i.e. ESU) scale. The database contains over 16,000 LAI, FAPAR, and FCOVER fiducial reference measurements covering 81 NEON, ICOS, and TERN sites between 2013 and 2022. The sites span cultivated crops, deciduous forest, evergreen forest, grasslands, mixed forest, pasture/hay, shrub/scrub and woody wetlands over the United States, Europe, and Australia.

Additional details and access to the GROUNDED EO database can be found in the associated Zenodo record.

Other sources of in situ reference data

A non-exhaustive list of established environmental monitoring networks and initiatives providing in situ reference data is given below:

• NEON: National Ecological Observatory Network (US)
• ICOS: Integrated Carbon Observation System (Europe)
• TERN: Terrestrial Ecosystem Research Network (Australia)
• CCRS: Canada Centre for Remote Sensing - Landsat and Sentinel-2 Validation
• EnviroNet

Tools

Note that many networks and initiatives (e.g. NEON, ICOS, TERN, etc.) provide only raw data in the form of digital hemispherical photography (DHP) or digital cover photography (DCP), requiring further processing to derive LAI, FAPAR and FCOVER. Free tools for processing these data include:

• CAN-EYE: GUI-based software for interactive processing of DHP and DCP images (Weiss and Baret, 2017).
• coveR2: An R package for processing digital cover photography images to retrieve forest canopy attributes (Chianucci et al., 2022).
• CoverPy: Automated estimates of plant area index, vegetation cover, crown cover, crown porosity, and uncertainties from digital cover photography in Python (Brown et al., 2024).
• HemiPy: A Python module for automated estimation of forest biophysical variables and uncertainties from digital hemispherical photographs (Brown et al., 2023).
• hemispheR: an R package for fisheye canopy image analysis (Chianucci et al., 2023).

Of these, HemiPy and CoverPy adopt FRM4VEG recommendations, allowing uncertainties in derived values to be quantified according to the ISO GUM.