GLAB-VOD: Global AI-Based Vegetation Data from Space

GLAB-VOD is a global, long-term vegetation optical depth dataset derived from L-band satellite brightness temperatures using a neural network. It provides an 18-day temporal and 25 km spatial resolution product from 2002 to 2020, created by researchers at Columbia University. The methodology also produced a side-product, GLAB TB, a daily-resolution global brightness temperature dataset starting in 2002.

Use Cases

• Monitoring global vegetation biomass trends based on the 2002-2020 time series.
• Analyzing sub-seasonal vegetation cycles and extremes using the residual signal components.
• Calibrating or validating other satellite-derived soil moisture or vegetation products using the machine-learning fusion approach.
• Studying the relationship between vegetation optical depth and climate variables across different biomes.

Strengths

• Provides a consistent, long-term global record from 2002 to 2020, enabling trend analysis.
• Uses a multi-staged neural network to merge data from SMOS, AMSR-E, and AMSR-2 missions without introducing bias, ensuring internal consistency.
• Offers a high-quality side-product (GLAB TB brightness temperatures) with daily temporal resolution matching SMOS quality.

Limitations

• Spatial resolution is coarse (25 km), limiting analysis of fine-scale vegetation patterns.
• Key metadata such as exact file size, row count, column names, and specific last update date are unavailable across all sources.
• The primary VOD target product (SMOSMAP-IB) used for training only covers 2015-2020, potentially affecting model performance outside that period.

Provenance

Source

Olya Skulovich, Columbia University

Collection Method

Created using a neural network trained on SMOSMAP-IB VOD product (2015-2020) with inputs from SMOS, AMSR-E, and AMSR-2 brightness temperatures and the CASM soil moisture dataset.

Time Range

2002 to 2020

Geography

Global