GIMMS LAI4g: Global Leaf Area Index from 1982 to 2020

GIMMS LAI4g version 1.2 is a spatiotemporally consistent global Leaf Area Index product providing data at half-month intervals and a 1/12° spatial resolution. It addresses uncertainties from NOAA satellite orbital drift and AVHRR sensor degradation by using biome-specific BPNN models trained on 3.6 million Landsat LAI samples and the PKU GIMMS NDVI product. The dataset is consolidated with Reprocess MODIS LAI via a Random Forests fusion method to extend coverage to 2020, presenting a more reasonable long-term global vegetation trend.

Use Cases

• Analyzing long-term global vegetation change trends based on a consistent 39-year time series.
• Validating and calibrating terrestrial ecosystem and climate models using a globally sampled LAI product.
• Studying the impact of sensor degradation and orbital drift corrections on historical satellite data records.
• Investigating biome-specific phenology and productivity using the biome-specific neural network models.

Strengths

• Covers a long temporal range from 1982 to 2020, providing nearly four decades of data.
• Incorporates 3.6 million high-quality global Landsat LAI samples for model training, suggesting a large reference dataset.
• Uses a pixel-wise Random Forests consolidation method to merge AVHRR and MODIS data, aiming for temporal consistency.
• Addresses specific known issues in historical LAI products, namely NOAA satellite orbital drift and AVHRR sensor degradation.

Limitations

• Specific column names, file size, and row counts are not provided in the available metadata.
• The dataset's update history and the exact 'last updated' date are unclear from the sources.
• Users must rely on the quality control (QC) layer for adequate use, implying potential data quality variations that require careful handling.

Provenance

Source

Sen Cao, Peking University

Collection Method

Generated using biome-specific Back-Propagation Neural Network (BPNN) models employing the PKU GIMMS NDVI product and Landsat LAI samples, then consolidated with Reprocess MODIS LAI.

Time Range

1982 to 2020

Geography

Global coverage at 1/12° resolution