[A-CG39] Global Carbon Cycle Observation and Analysis

[ACG39-P12] Variability of Terrestrial Carbon Fluxes in Asia: Multiple Bottom-up Approaches

*Kazuhito Ichii¹, Misaki Hase¹, Ruci Wang¹, Daniel Joseph Henri¹, Shuai Shao¹ (1.Chiba University)

Keywords:Terrestrial Carbon Cycle, Asia, Terrestrial Biosphere Model, Remote Sensing

Asia is undergoing significant changes in its land surface environment due to factors such as its large population, industrial development, and climate change, and there have also been significant fluctuations in the CO2 budget and flux in the region. Many studies have been conducted on the analysis of changes in the terrestrial carbon budget in Asia, focusing on the results of process models and atmospheric inverse models. In addition, various analyses have been conducted using remote sensing based model (RS-model) and machine learning based upscaling (ML-upscaling). By conducting integrated analysis that spans various estimation methods, including process models, RS-model and ML-upscaling, it is possible to identify the characteristics of each method and to make robust estimates. In this study, we used outputs of process-based model by TRENDY, multiple RS-models, and ML-upscaling products, and analyzed changes in the gross primary productivity (GPP), net ecosystem exchange (NEE), and net biome productivity (NBP) in Asia from the 1980s to recent years (e.g. 2020). In the three regions of North Asia (Siberia), East Asia, and South Asia, the trends in the interannual variability of photosynthesis showed good agreement among the three estimation approaches. On the other hand, we found large differences in GPP, NEE, and NBP among different approaches in Southeast Asia. In addition, with regard to NEE and NBP, there was good agreement on the trend of interannual variability in North, East and South Asia, despite the differences between NEP and NBP. On the other hand, the absolute magnitude of NEE and NBP showed large differences among the results of RS-based, ML-upscaling, and process models. Since RS-based models and ML-upscales have the great advantage of being able to capture detailed changes because the spatial resolution of the input data is high (1-5 km in this case), further efforts are expected to improve magnitude of NEE and NBP for RS-based model and ML-upscaling.