Past and present hydrothermal regimes shape microbial resource allocation for soil C, N, and P cycling: insights from machine-learning predictions

Understanding and predicting the feedback between climate change and soil carbon dynamics remains a major scientific challenge. A key uncertainty lies in our limited knowledge of how changing hydrothermal conditions influence microbial functional dynamics and their contributions to soil carbon emissions. In particular, the microbial functions that respond to hydrothermal variability—and their interactions with functions involved in soil carbon and nutrient cycling—remain poorly characterized. It is still unclear how both historical and current hydrothermal conditions affect the relative abundances of these microbial functions and how these shifts impact the dynamics of soil carbon emission in response to changing hydroclimate. To fill these knowledge gaps, we combined gene-to-ecosystem data from key ecological networks to develop artificial intelligence models to identify and quantify microbial resource allocation strategies in response to past and present hydrothermal properties. Our findings indicated that microbial communities acclimated to reduced soil moisture by lowering investment in recalcitrant-C decomposition and monomer nutrient mineralization. This drought-mitigation response was amplified by drying legacy but dampened by nutrient limitation. Elevated soil temperature, in contrast, generally increased microbial investment in N acquisition, while thermal legacy strengthened the thermal resistance of N-acquisition allocation and promoted reallocation of C- and P-acquiring functions toward adaptation to current hydrothermal dynamics. Finally, we will show how the identified resource optimization strategies can be applied to interpret observed soil carbon dynamics under climate change and to advance earth system modeling of soil carbon emissions.