Validation of baseline and modified Sentinel-2 Level 2 Prototype Processor leaf area index retrievals over the United States
Why this work is in the frame
A frame that forgets how it found something cannot be audited. These are the routes that admitted this work.
Bibliographic record
Abstract
The Sentinel-2 Level 2 Prototype Processor (SL2P) is made available to users for the retrieval of vegetation biophysical variables including leaf area index (LAI) from Multispectral Instrument (MSI) data within the Sentinel Application Platform (SNAP). A limited number of validation exercises have indicated SL2P LAI retrievals frequently meet user requirements over agricultural environments, but perform comparatively poorly over heterogeneous canopies such as forests. Recently, a modified version of SL2P was developed, using the directional area scattering factor (DASF) to constrain retrievals as an alternative to regularisation (SL2P-D). Whilst SL2P makes use of prior information on expected canopy conditions, SL2P-D is trained using uniform distributions of input parameters to define radiative transfer model (RTM) simulations. Using in situ measurements available through the Copernicus Ground Based Observations for Validation (GBOV) service, we performed an extensive validation of SL2P and SL2P-D LAI retrievals over 19 sites throughout the United States. For effective LAI (LAIe), SL2P demonstrated good overall performance (RMSD = 0.50, NRMSD = 31%, bias = −0.10), with all LAI retrievals meeting the Sentinels for Science (SEN4SCI) uncertainty requirements over homogeneous canopies (cultivated crops, grasslands, pasture/hay and shrub/scrub), whilst underestimation occurred over heterogeneous canopies (deciduous forest, evergreen forest, mixed forest, and woody wetlands). SL2P-D retrievals demonstrated reduced bias, slightly improving overall performance when compared with SL2P (RMSD = 0.48, NRMSD = 30%, bias = −0.05), indicating its retrieval approach appears to offer some advantages over regularisation using prior information, especially at LAIe > 3. Additionally, SL2P-D resulted in 32% more valid retrievals than SL2P, with the largest differences observed at LAIe < 1. Validation against in situ measurements of LAI as opposed to LAIe yielded similar patterns but poorer performance (RMSD = 1.08 to 1.13, NRMSD = 49% to 52%, bias = −0.64 to −0.68) because the RTM used by SL2P and SL2P-D does not account for foliage clumping. In addition to the retrievals themselves, we examined the relationship between predicted uncertainties and observed differences in retrieved and in situ LAI. With respect to LAIe, SL2P’s predicted uncertainties were conservative, underestimating observed differences in only 35% of cases, whilst those for LAI were unbiased.
Fetched live from OpenAlex and de-inverted. Abstracts are not stored in this database: the inverted indexes are 8.6 GB of the frame’s 9.3 GB of text, and the host has 13 GB free.
Full frame distilled prediction
Teacher imitationNot calibrated prevalence, not ground truth. Human validation pending. Learned from the 10,348 direct Codex labels and 10,348 direct Gemma labels. Candidate is the union of thresholded teacher heads; consensus is their intersection. These outputs are machine_predicted_unvalidated and are not human labels or direct frontier model labels.
Codex and Gemma teacher scores by category
| Category | Codex | Gemma |
|---|---|---|
| Metaresearch | 0.001 | 0.000 |
| Meta-epidemiology (narrow) | 0.000 | 0.000 |
| Meta-epidemiology (broad) | 0.000 | 0.000 |
| Bibliometrics | 0.000 | 0.001 |
| Science and technology studies | 0.000 | 0.000 |
| Scholarly communication | 0.000 | 0.000 |
| Open science | 0.000 | 0.000 |
| Research integrity | 0.000 | 0.000 |
| Insufficient payload (model declined to judge) | 0.000 | 0.000 |
Machine scores (provisional)
The two teacher heads of the student model, read on this work. A score orders the frame for review; it never asserts a category, and the validation status ships verbatim with every row.
Baseline scores from an immature model (maturity gate not passed, 7 training rounds). Scores rank; they never assert a category.
score_only:v0-immature-baseline · verbatim from the scoring run: score_only means the number may rank works, and no category label ships from it