ABSTRACT Soil bacteria play essential roles in nutrient cycling, organic matter decomposition and ecosystem stability. However, conventional methods for assessing bacterial abundance and diversity, such as DNA sequencing, are costly, labor‐intensive and limited in spatial coverage. Our study aimed to evaluate the feasibility of using mid‐infrared (MIR) spectroscopy integrated with spectrotransfer functions (STFs) to predict soil bacterial relative abundance and community diversity (ASV richness) in an alpine ecosystem. We analysed 121 soil samples collected along an elevational gradient in the Sygera Mountains on the Qinghai‐Tibet Plateau, China. Bacterial phylum‐level abundance and community diversity were predicted using MIR spectra, soil and environmental properties, and integrated MIR‐environmental STFs. Six machine learning algorithms were assessed using a 10‐fold cross validation, including Partial Least Squares Regression (PLSR), Gaussian Process Regression (GPR), Support Vector Machines (SVM), Random Forest (RF), Extreme Gradient Boosting (XGBoost) and Cubist. The results demonstrated that, for summer topsoils of the Sygera Mountains, integrated STFs achieved R 2 values of 0.42–0.73 for phylum/class abundance and 0.43 for diversity using RF, XGBoost or Cubist algorithms, which consistently outperformed linear methods. Key predictive spectral regions identified corresponded to organic functional groups (e.g., C‐H, O‐H, N‐H) and clay minerals (O‐H), indicating that MIR spectra effectively captured both biochemical and mineralogical soil properties that shaped bacterial communities. Our findings confirmed that MIR‐STFs provide an efficient approach of estimating soil bacterial abundance and diversity, offering strong potential for large‐scale microbial monitoring and soil health assessment in remote alpine environments.
Yang et al. (Wed,) studied this question.