Historical Retrospect of Aquatic Microbial Communities Based on Water–Carbonate Equilibrium and pH Values in Dianchi Lake Using the Random Forest Algorithm

Aquatic microbial communities are crucial for the dynamics of algal blooms. Due to the limited development history and high cost of high-throughput sequencing, microbial information is commonly unavailable in past aquatic quality surveys. This study comprehensively analyzed 19 years of historical data from alkalitrophic Dianchi Lake to explore the interrelationships among pH, carbonate equilibrium (CO32– and HCO3–), and physicochemical/microbial variables, and developed predictive models of carbonate equilibrium and pH using the Random Forest (RF) algorithm. Microbial taxa, particularly Proteobacteria, Cyanobacteria, and Bacteroidetes, were more predictive of pH and carbonate equilibrium than physicochemical variables. The Mean Absolute Percentage Error (MAPE) was 1.5%, 26.9%, and 30.1% for pH, CO32–, and HCO3–, respectively. Different microorganisms played predominant roles across varying pH ranges, with photoautotrophs (e.g., Cyanobacteria) determining carbonate equilibrium at pH < 9.0 and heterotrophs (e.g., Proteobacteria and Bacteroidetes) at pH > 9.0. Furthermore, a backtracking algorithm was used to reconstruct the historical microbial community structure based on pH values, with the backtracked cyanobacterial abundance matching well with historical chlorophyll-a data from 2000 to 2018. Our results highlight the strong correlations between microbial community structure, aquatic pH, and carbonate equilibrium and provide a powerful backtracking algorithm to obtain historical microbial information for lake management.