Improving East Asian BVOC Emission Estimates via Machine Learning-Based Plant Functional Types Mapping

Biogenic volatile organic compounds (BVOCs) are dominant precursors of tropospheric ozone and secondary organic aerosols, making their accurate representation critical for air quality modeling. Current BVOC emission estimates rely heavily on satellite-derived Plant Functional Type (PFT) maps, which often exhibit substantial discrepancies and fail to capture detailed land-use characteristics in heterogeneous agricultural and urban landscapes. To address these limitations, this study developed a new regional PFT dataset by integrating field-surveyed forest information from the Korea Forest Service’s Forest Geographic Information System (FGIS). We applied a machine learning approach to extrapolate the high-resolution, observation-based PFT characteristics of the Korean Peninsula to the broader East Asian region, utilizing meteorological variables and satellite land-cover products as predictors. The generated PFT dataset was implemented into the biogenic emission module of the Community Multiscale Air Quality (CMAQ) model to evaluate its impact on air quality simulations. We validated the model performance by comparing simulated BVOC mixing ratios with airborne observations from the ASIA-AQ campaign. The results demonstrate that the proposed PFT dataset yields distinct spatial emission patterns compared to conventional satellite-based maps. Notably, the simulation showed improved consistency with observations, particularly over complex terrain and mixed land-use areas. These findings suggest that regionally optimized PFT inputs, grounded in field observations and machine learning, significantly reduce uncertainties in BVOC emission inventories and subsequent chemical transport modeling over East Asia.

Acknowledgment: This work was supported by the National Research Foundation of Korea(NRF) grant funded by the Korea government (MSIT) (No. RS-2025-16070879).