VODCA2GPP: High Resolution GPP Estimation in the Mediterranean basin from Vegetation Optical Depth Using Machine Learning

Gross Primary Production (GPP) serves as a critical indicator of ecosystem function and its response to climate change. While numerous GPP estimates from different sources (model data, satellite derived, flux tower) exist, discrepancies between these datasets remain, emphasizing the need for new datasets and continuous improvement, particularly in understanding carbon-climate feedbacks and ecosystem resilience.
By refining the VODCA2GPP product  [Wild et al., 2022], which estimates GPP using Vegetation Optical Depth (VOD) from microwave remote sensing, we present an enhanced version with four key methodological improvements.
First, we enhanced the spatial resolution from 0.25° to 0.1°, enabling finer scale detection of spatial heterogeneity in vegetation productivity and improving representation of local ecosystem dynamics. Additionally, we transitioned from X-band to Ku-band VOD observations due to their superior signal to noise ratio in the Mediterranean region, enhancing data quality while maintaining overall model performance.
Second, we integrated land cover information to improve model generalizability across different biomes, addressing the imbalanced distribution of in-situ validation stations and enhancing the model's ability to capture ecosystem specific carbon uptake patterns. 
Third, we incorporated soil moisture data to account for water availability, which is the primary constraint on vegetation productivity in many biomes and is particularly crucial for understanding drought responses and ecosystem stress. 
Fourth, we utilized ESA CCI Biomass observations to better capture biomass accumulation patterns.
The enhanced model was validated using an expanded set of in-situ measurements, including data from WARM Winter, AmeriFlux, JapanFlux, and CH4 datasets, which significantly extends our validation capabilities across different climatic zones and ecosystem types. Validation against FLUXNET in situ measurements and comparisons with leading datasets, including MODIS and FLUXCOM, demonstrate that the finer spatial resolution better captures local scale variability while maintaining strong model accuracy and reliability. This updated VODCA2GPP version offers a valuable resource for analyzing global vegetation dynamics, enabling better monitoring of ecosystem responses to environmental change and improving our understanding of the terrestrial carbon cycle.

This research has been funded through the GLANCE project.

References:

Bernhard Wild, Irene Teubner, Leander Moesinger, Ruxandra-Maria Zotta,
Matthias Forkel, and Wouter A Dorigo. Vodca2gpp–a new, global, long-
term (1988–2020) gross primary production dataset from microwave re-
mote sensing. Earth System Science Data, 14(3):1063–1082, 2022. doi:
10.5194/essd-14-1063-2022