Joint retrieval of aerosol optical depth and leaf area index from FY-3D/MERSI measurements

Leaf area index (LAI) is a key indicator of vegetation structure and plays a crucial role in global ecological and climate systems. The accuracy of satellite-based LAI retrievals critically depends on the effective separation of land surface and atmospheric signals in radiation measurements. Most traditional remote sensing algorithms retrieve LAI from land surface reflectance (LSR) products after atmospheric correction. In such schemes, uncertainties in aerosol optical depth (AOD) retrievals are implicitly propagated into LSR and subsequently into LAI estimates, leading to degraded accuracy and reduced spatiotemporal consistency, particularly under conditions of rapidly varying or heavy aerosol loadings.

In this study, we propose a joint retrieval framework that simultaneously estimates AOD, LSR, and LAI directly from FY-3D/MERSI top-of-atmosphere (TOA) reflectance observations. The approach employs an ensemble machine learning–based model, thereby avoiding the conventional decoupled treatment of atmospheric correction and vegetation parameter retrieval. By jointly optimizing atmospheric state variables and land surface biophysical parameters under a unified observational constraint, the method effectively suppresses the propagation of atmospheric correction uncertainties into LAI estimates.

Global retrievals for the period 2020–2023 demonstrate robust performance across a wide range of aerosol loading and observation conditions. The retrieved LAI captures reasonable spatial patterns and seasonal dynamics and shows good consistency with GLASS and MODIS LAI products. This study advances a novel land–atmosphere integrated inversion strategy and establishes a global-scale coupled aerosol-vegetation remote sensing dataset, which can serve as an important technique and data source for improving Earth system models and investigating ecosystem responses to global climate change.