AI Model Uses High-Frequency Satellite Data to Monitor Hourly Carbon Absorption Across East Asia

Approximately 30% of global carbon dioxide (CO2) emissions are absorbed by terrestrial vegetation through photosynthesis. Recently, researchers from UNIST have developed an innovative artificial intelligence (AI) method that predicts this CO2 uptake with unprecedented temporal precision. This breakthrough could substantially assist in climate change mitigation and inform carbon-neutral strategies. The research team, led by Professor Jungho Im from the Department of Earth Environmental Urban Construction Engineering at UNIST, introduced an AI model capable of estimating daily gross primary production (GPP) at hourly intervals. GPP is a critical metric that indicates the amount of carbon actively absorbed by plants during photosynthesis, essential for quantifying ecosystem carbon sequestration. Their findings were published in the journal Remote Sensing of Environment. The new model utilizes high-frequency observations from the Himawari-8 geostationary satellite, which captures data every 10 minutes. First author Sejeong Bae emphasized the advantage of this frequent data collection: "Unlike polar-orbiting satellites that only observe a given location one to four times per day, our model can track diurnal changes in photosynthesis with remarkable precision." This enhanced temporal resolution provides detailed insights into how plants absorb CO2 throughout the day. The model integrates various meteorological data, including Aerosol Optical Depth (AOD), a measure derived from satellite observations that reflects the concentration of particulate matter like fine dust. AOD plays a significant role in GPP by influencing the amount and quality of sunlight that reaches the Earth's surface. Sunlight is often absorbed or scattered by aerosols, affecting photosynthetic activity. To better understand the model's prediction mechanisms, the researchers used SHapley Additive exPlanations (SHAP), an advanced explainable AI technique. The results indicated that AOD is the most influential factor during the morning and evening when the sun is lower in the sky. At these times, more scattered light reaches the ground, making the photosynthetic response of vegetation more sensitive to atmospheric aerosols. This aligns with existing knowledge about how lower solar elevation intensifies the impact of scattered light. Professor Im highlighted the model's practical applications: "Our method can estimate the spatial and temporal dynamics of carbon absorption over East Asia at a 2km resolution and across all 24 hours of the day. This makes it a powerful tool for ecosystem carbon flux analysis, vegetation monitoring, and photic environment-based carbon modeling." This AI-based approach marks a significant step forward in environmental monitoring, offering more accurate and timely data to support policy-making and scientific research on climate change and carbon cycling. The model's ability to integrate high-frequency satellite data and account for the effects of aerosols could lead to more effective carbon management and conservation strategies, ultimately contributing to global efforts to mitigate climate impacts.