STGF-AGB: a machine learning and spatiotemporal fusion framework for seamless monthly grassland biomass mapping

The resulting 30 m continuous AGB dataset for 2001-2024 revealed pronounced spatial heterogeneity in peak AGB trends across Chinese grasslands. Significant increases occurred across 60.16% of the valid grassland area, concentrated mainly in the central and eastern Tibetan Plateau and the central and western Inner Mongolian Plateau. In contrast, significant increases in growing season evapotranspiration (ET) derived from PML V2 occurred across only 17.90% of the grassland area, indicating widespread spatial asynchrony between biomass gains and changes in total ET.

ET component analysis further showed significant increases in vegetation transpiration and canopy interception evaporation of 0.64 and 0.20 mm yr^-1, respectively, whereas soil evaporation showed a nonsignificant declining trend of -0.15 mm yr^-1. The relationship between AGB and ET also varied along the aridity gradient, with concurrent increases in biomass and ET most pronounced in semiarid grasslands, while stronger compensation among ET components occurred toward wetter regions.

Overall, the proposed framework provides a practical approach for generating long term seamless monthly grassland AGB. The dataset improves characterization of seasonal biomass dynamics and reveals that widespread biomass gains are not accompanied by proportional increases in total ET, providing a new data basis for grassland monitoring and assessment of regional water use responses</p>