Bridging the latency gap in satellite data: impact of random forest-based leaf area index on biogenic emissions and air quality over East Asia

Abstract Biogenic volatile organic compounds (BVOCs), key ozone (O 3 ) precursors, have emission estimates highly sensitive to leaf area index (LAI). Despite its high fidelity, the Global Land Surface Satellite (GLASS) LAI often relies on outdated data for current air quality simulations due to update delays, introducing considerable uncertainty. Therefore, this study aims to develop a Random Forest (RF) model that generates near-real-time LAI for East Asia using multi-source satellite and reanalysis data. The model had high predictive accuracy (R > 0.90) across all seasons. Applying the March 2024 LAI predicted by the same RF model to the Model of Emissions of Gases and Aerosols from Nature –CMAQ simulations reduced total isoprene emissions over East Asia by 9.17 Gg and total monoterpene emissions by 7.03 Gg compared to simulations using 2021 GLASS LAI. Consequently, aircraft observations from the ASIA-AQ campaign showed a reduced model overestimation, with normalized mean bias for isoprene and monoterpenes decreasing from 50.58% to 16.18% and from 120.46% to 70.97%, respectively. Furthermore, RF-based LAI simulations showed performance similar to those using the officially released 2024 GLASS LAI. These findings indicate that the RF-based LAI proposed in this study provides up-to-date LAI information during GLASS data update delays and serves as a reliable complementary input for BVOC emission estimates and air quality modeling.