Decoding the spring-initiated fermentation of northern sauce-flavor baijiu: Integrating physicochemical dynamics, microbial succession, and interpretable machine learning

Sauce-flavor baijiu production in northern China is constrained by a seasonal mismatch, whereby the traditional autumn-initiated fermentation regimen suffers from insufficient thermal accumulation and impeded microbial colonization upon northward relocation. Spring-initiated fermentation has thus been adopted to better align microbial establishment with seasonal warming, yet its ecological trajectory remains poorly resolved. In this study, a single spring-initiated production cycle at one commercial distillery was characterized across the Xiasha (XS) and Zaosha (ZS) stages and four recycling rounds (R1–R4) by integrating physicochemical profiling, microbial ecological analyses, and interpretable machine learning. Over the production cycle, moisture rose from 41.2% to 53.3%, starch declined to 17.25 ± 0.82%, reducing sugar transiently peaked at 4.37 ± 0.33% in R3, and ethanol reached 3.30 ± 0.10% (v/v) at R2. Bacterial communities underwent pronounced stage-dependent succession, converging toward Lactobacillus -enriched assemblages in later rounds, whereas fungal communities maintained broader compositions dominated by Pichia , Aspergillus , Thermomyces , and Rasamsonia . Ecological analyses indicated that bacterial assembly was governed predominantly by deterministic selection, whereas fungal assembly was more strongly influenced by stochastic processes. Machine-learning analysis revealed that microbial succession was most tightly coupled with ethanol and reducing sugar transitions (R 2 = 0.787 and 0.633), and SHapley Additive exPlanations (SHAP) attribution further identified transition-sensitive microbial indicators, including Pichia , Aspergillus , and Rasamsonia . Collectively, these findings describe the covariation of physicochemical changes and microbial succession within a single spring-initiated production cycle, and provide a data-driven framework for process monitoring and quality optimization in northern sauce-flavor Baijiu.