A study by Xiaofei Ge, Tao Zhang, and colleagues, published in BiocharBiochar is a carbon-rich material created from biomass decomposition in low-oxygen conditions. It has important applications in environmental remediation, soil improvement, agriculture, carbon sequestration, energy storage, and sustainable materials, promoting efficiency and reducing waste in various contexts while addressing climate change challenges. More (2025), investigated the use of machine learning (ML) to predict the optimal conditions for the hydrothermal treatment (HT) of biowastes, focusing on the fate of phosphorus (P). The work specifically explored how the addition of calcium (Ca) and iron (Fe) ions affects the distribution of P between the solid product, hydrochar, and the liquid phase. Effectively managing biowastes like swine manure (SM) is urgently needed to recover valuable nutrients and immobilize toxic elements, making this research significant for sustainable resource management.

The study compared the predictive performance of three ML models—XGBoost, Decision Tree, and Random Forest—using a dataset of 423 entries related to P fate in hydrochar production from biowastes. Th XGBoost model proved to be the most accurate, showing outstanding performance on the testing set with a coefficient of determination (R2) of 1.0 for predicting inorganic P in the liquid phase (IPL). For predicting total P in the hydrochar (TPS), the XGBoost model achieved an R2 of 0.77, outperforming the Decision Tree model’s R2 of 0.49 on the testing set. This strong predictive capability highlights the potential of AI-based methods to optimize HT processes for safe recycling of bioresources.

Feature importance analysis, using the XGBoost model, helped identify the most critical parameters influencing the output. The results indicated that feedstockFeedstock refers to the raw organic material used to produce biochar. This can include a wide range of materials, such as wood chips, agricultural residues, and animal manure. More characteristics generally had a greater overall impact on P distribution than operational conditions. The oxygen (O) content in the raw material was the most influential factor, with its feature importance nearly reaching 50%, significantly surpassing all other factors. The reaction time was found to have a more significant effect on the output than the reaction temperature. The concentration of calcium and iron ions ranked just after reaction time, indicating their significant role in regulating P distribution in the solid and liquid phases.

The addition of Ca and Fe ions was observed to positively correlate with an upward trend in TPS concentration. This is attributed to Ca2+ ions binding with phosphate to form precipitates, promoting P transfer into the hydrochar. Similarly, Fe2+ facilitates the incorporation of P into the hydrochar, potentially by infiltrating the inner pores rather than simply depositing on the surface. Further analysis using Partial Dependence Plots ( PDP) showed that P distribution was more substantially affected under strongly acidic (pH<2) and strongly alkaline (pH>12) conditions compared to neutral conditions. The formation of metal phosphate precipitates is favored under alkaline conditions, promoting P transfer to the solid phase. It was also found that the effect of time on TPS and IPL was minimal when the reaction time was less than 200 minutes.

Experimental validation confirmed the XGBoost model’s effectiveness in predicting the P distribution and the impact of Ca and Fe ions. Moreover, advanced analyses of the hydrochar revealed that lower reaction intensities promoted crystal formation, but an increase in reaction severity led to crystal degradation. The forms of organic P detected in the hydrochar became more homogenized as the reaction severity increased.

This research successfully demonstrates that machine learning can precisely control key HT conditions to effectively regulate P distribution in hydrochar and the liquid phase, offering a powerful predictive tool for optimizing agricultural waste management and potentially contributing to carbon neutrality strategies.

Source: Ge, X., Zhang, T., Mukherjee, S., Chen, Y., Wang, X., Chen, X., Liu, M., Ali, E. F., Rinklebe, J., Lee, S. S., & Shaheen, S. M. (2025). Optimizing the conditions of biowastes hydrothermal treatment and predicting phosphorus fate in the hydrochar and liquid phase using machine learning. Biochar, 7(96).