Improving Satellite Leaf Area Index Estimation Based On Various Integration Methods
Bibliographic record
Abstract
Leaf Area Index (LAI) is an important land surface biophysical variable that is used to characterize vegetation amount and activity. Current satellite LAI products, however, do not satisfy the requirements of the modeling community due to their large uncertainties and frequent missing values. Each LAI product is currently generated from only one satellite sensor data. There is an urgent need for advanced methods to integrate multiple LAI products to improve the product's accuracy and integrality for various applications. To meet this need, this study proposes four methods, including the Optimal Interpolation (OI), Bayesian Maximum Entropy (BME), Multi-Resolution Tree (MRT) and Empirical Orthogonal Function (EOF), to integrate multiple LAI products. Three LAI products have been considered in this study: Moderate Resolution Imaging Spectroradiometer (MODIS), Multi-angle Imaging SpectroRadiometer (MISR) and Carbon cYcle and Change in Land Observational Products from an Ensemble of Satellites (CYCLOPES) LAI. As the basis of data integration, this dissertation first validates and intercompares MODIS and CYCLOPES LAI products and also evaluates their geometric accuracies. The CYCLOPES LAI product has smoother temporal profiles and fewer spatial variations, but tends to produce spurious large errors in winter. The Locally Adjusted Cubic-spline Capping algorithm is revised to smooth multiple years' average and variance. Although OI, BME and MRT based methods have been used in other fields, this is the first research to employ them in integrating multiple LAI products. This dissertation also presents a new integration method based on EOF to solve the problem of large data volume and inconsistent temporal resolution of different datasets. High resolution LAI reference maps generated with ground measurements are used to validate these algorithms. Validation results show that all of these four methods can fill data gaps and reduce the errors of the existing LAI products. The data gaps are filled with information from adjacent pixels and background. These algorithms remove the spurious large temporal and spatial variation of the original LAI products. The combination of multiple satellite products significantly reduces bias. OI and BME can reduce the RMSE from 1.0 (MODIS) to 0.7 and reduce the bias from +0.3 (MODIS) and -0.2 (CYCLOPES) to -0.1. MRT can produce similar results with OI but with significantly improved efficiency. EOF also generates the results with the RMSE of 0.7 but zero bias. Limited ground measurement data hardly prove which methods outperform the others. OI and BME theoretically produce statistically optimal results. BME relaxes OI's linear and Gaussian assumption and explicitly considers data error, but bears a much higher computational burden. MRT has improved efficiency but needs strict assumptions on the scale transfer function. EOF requires simpler model identification, while it is more "empirical" than "statistical". The original contributions of this study mainly include: 1) a new application of several different integration methods to incorporate multiple satellite LAI products to reduce uncertainties and improve integrality, 2) an enhancement of the Locally Adjusted Cubic-spline Capping by revising the end condition, 3) a novel comprehensive comparison of MODIS C5 LAI product with other satellite products, 4) the development of a new LAI normalization scheme by assuming the linear relationship between measurement error and LAI natural variance to account for the inconsistency between products, and finally, 5) the creation of a new data integration method based on EOF.
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.
How this classification was reachedexpand
Full frame machine prediction
Teacher imitationNot calibrated prevalence, not ground truth. Human validation pending. The Gemma side is a direct model label for every work in the frame, read from the title-only record. The Codex side is a classifier learned from the 10,348 direct Codex labels and calibrated to design-weighted sample rates; fields without enough sample support carry no Codex call. Candidate is the union of the two sides; consensus is their intersection. These outputs are machine_predicted_unvalidated and are not human labels.
Distilled classifier scores by category (both heads)
| Category | Codex | Gemma |
|---|---|---|
| Metaresearch | 0.001 | 0.003 |
| Meta-epidemiology (narrow) | 0.001 | 0.000 |
| Meta-epidemiology (broad) | 0.001 | 0.001 |
| Bibliometrics | 0.002 | 0.001 |
| Science and technology studies | 0.000 | 0.000 |
| Scholarly communication | 0.001 | 0.001 |
| Open science | 0.001 | 0.001 |
| Research integrity | 0.000 | 0.000 |
| Insufficient payload (model declined to judge) | 0.001 | 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 itClassification
machine, unvalidatedMachine predicted; a candidate call from one source (direct Gemma or distilled Codex), not a consensus.
How this classification was reached, model by model and score by score, is at the end of the page under "How this classification was reached".