Assessing the impact of LAI data assimilation in a multidecadal global reanalysis

This study presents a global reanalysis of land surface variables from 1981 to 2022 at a 0.25° x 0,25° spatial resolution, conducted using the ISBA land surface model within the SURFEX v9 modeling platform. Two experiments were performed: (1) an open-loop (OL) simulation with the ISBA land surface model and (2) a data assimilation analysis (ANA) experiment incorporating Leaf Area Index (LAI) observations in ISBA using the LDAS-Monde system. The simulations were forced offline with ERA5 reanalysis atmopsheric variables. The assimilated LAI dataset consists of THEIA AVHRR-derived observations from 1981 to 2018, and of Copernicus Land Monitoring Service (CLMS) GEOV2 from 2019 to 2022.

The study analyzes the impact of LAI assimilation on surface variables, with a particular focus on soil temperature anomalies at deep layers. By comparing the OL and ANA runs, we assess how improved vegetation representation influences soil thermal dynamics and energy exchanges over multiple decades.

To validate the impact of LAI assimilation, an evaluation is conducted over France, where in situ temperature observations at 1m depth are available from more than 70 automatic weather stations. This comparison provides insight into how LAI assimilation affects subsurface thermal conditions and helps quantify its added value in the reanalysis. Additionally, on a global scale, we assess the impact of LAI assimilation on Gross Primary Production (GPP), highlighting the improvements observed after integrating LAI observations into the model.

This long-term global reanalysis, spanning over 40 years, provides a unique dataset for studying historical land surface dynamics and their interactions with climate variability. Having both OL and ANA simulations is crucial, as the OL experiment serves as a benchmark for understanding long-term trends driven by atmospheric forcing alone, while the ANA experiment allows us to quantify the impact of observational constraints on land surface processes. This dual approach is essential for improving our confidence in land surface modeling, enabling better representation of vegetation-atmosphere interactions, and providing a solid foundation for future climate studies, hydrological assessments, and ecosystem monitoring.