Uncovering time-lagged causal mechanisms in hyperthermophilic composting using Peter-Clark momentary conditional...

Composting is a multivariate time-series process jointly shaped by pile structure, microbial succession, and physicochemical conditions, in which time-lagged effects and feedback loops are common yet often overlooked. Here, we applied the Peter-Clark Momentary Conditional Independence (PCMCI) causal-discovery framework to hyperthermophilic composting to reconstruct multi-lag causal networks for rice husk hyperthermophilic composting (RHC) and corn stalk hyperthermophilic composting (CSC) from multi-source time-series data. The inferred networks revealed fundamentally different regulatory architectures. In RHC, free air space (FAS) repeatedly emerged as a persistent upstream driver across multiple lags, regulating moisture, pH, and NH₄⁺-N and steering microbial community shifts toward trajectories that supported sustained carbon degradation and humic substance accumulation. In contrast, CSC exhibited a more compact, constraint-dominated network in which system regulation depended more on microbial succession and nitrogen transformation; notably, NH₄⁺-N exerted a significant delayed inhibitory effect on the germination index, revealing a key bottleneck that constrained maturity development. Compared with correlation analysis and structural equation modeling, PCMCI explicitly resolves causal directionality, time delays, and feedback structures within a unified multivariate framework. Overall, these causal insights explain the superior performance of rice husk composting relative to corn stalk composting and provide a transferable framework for time-resolved mechanistic inference and process optimization in composting systems.