Prediction of nano, fine, and medium colloidal phosphorus in agricultural soils with machine learning

Predicting the governing factors for the release of colloidal phosphorus using machine learning

Predicting the parameters that influence colloidal phosphorus (CP) release from soils under different land uses is critical for managing the impact on water quality. Traditional modeling approaches, such as linear regression, may fail to represent the intricate relationships that exist between soil qualities and environmental influences. Therefore, in this study, we investigated the major determinants of CP release from different land use/types such as farmland, desert, forest soils, and rivers. The study utilizes the structural equation model (SEM), multiple linear regression (MLR), and three machine learning (ML) models (Random Forest (RF), Support Vector Regression (SVR), and eXtreme Gradient Boosting (XGBoost)) to predict the release of CP from different soils by using soil iron (Fe), aluminum (Al), calcium (Ca), pH, total organic carbon (TOC) and precipitation as independent variables. Results show that colloidal-cations (Fe, Al, Ca) and colloidal-TOC strongly influence CP release, while bioclimatic variables (precipitation) and pH have weaker effects. XGBoost outperforms the other models with an R2 of 0.94 and RMSE of 0.09. SHapley Additive Explanations described the outcomes since XGBoost is accurate. The relative relevance ranking indicated that colloidal TOC had the highest ranking in predicting CP. This was supported by the analysis of partial dependence plots, which showed that an increase in colloidal TOC increased soil CP release. According to our research, the SHAP XGBoost model provides significant information that can help determine the variables that considerably influence CP contents as compared to RF, SVM, and MLR.

Microbially Induced Soil Colloidal Phosphorus Mobilization Under Anoxic Conditions

Understanding the behavior of colloidal phosphorus (Pcoll) under anoxic conditions is pivotal for addressing soil phosphorus (P) mobilization and transport and its impact on nutrient cycling. Our study investigated Pcoll dynamics in acidic floodplain soil during a 30-day flooding event. The sudden oxic-to-anoxic shift led to a significant rise in pore-water Pcoll levels, which exceeded soluble P levels by more than 2.7-fold. Colloidal fractions transitioned from dispersed forms (<220 nm) to colloid-associated microaggregates (>220 nm), as confirmed by electron microscopy. The observed increase in colloidal sizes was paralleled by their heightened ability to form aggregates. Compared to sterile control conditions, anoxia prompted the transformation of initially dispersed colloids into larger particles through microbial activity. Curiously, the 16S rRNA and ITS microbial diversity analysis indicated that fungi were more strongly associated with anoxia-induced colloidal release than bacteria. These microbially induced shifts in Pcoll lead to its higher mobility and transport, with direct implications for P release from soil into floodwaters.

Restricted colloidal-bound phosphorus release controlled by alternating flooding and drying cycles in an alkaline calcareous soil

Colloid-facilitated phosphorus (P) migration plays an important role in P loss from farmland to adjacent water bodies. However, the dynamics of colloidal P (Pcoll) release as influenced by irrigation in alkaline calcareous soil remains a knowledge gap. The present study, monitored the dynamic change of Pcoll under different water management strategies: 1) control, 2) flooding, and 3) alternating flooding and drying cycles. Soil water-dispersible colloids (0.6 nm-1 μm) were extracted by combining filtration and ultrafiltration methods. The contents of P, cation and organic carbon in the water-dispersible colloids were determined and the stability and mineral composition of colloidal fractions were characterized. The results showed that Pcoll ranged from 16.5 to 25.5 mg kg-1 and represented 42.8%-64.9% of the water-extracted P in the control. Flooding significantly decreased the Pcoll content by 16.0%-62.1% (mean 32.7%) and it may be attributed to the dissolution of colloidal iron (Fe) bound P. The alternating flooding and drying treatment significantly reduced the Pcoll content by 11.6%-88.0% (mean 67.6%). The Pcoll content of the flooding event was always greater than the Pcoll content of the drying event during flooding and drying cycles. Redundancy analysis and random forest modeling showed that the colloidal calcium (Ca) and ionic strength in soil solutions had negative correlations with the Pcoll content, and pH, ionic strength and truly dissolved P were the critical factors affecting Pcoll. Drying of the flooded soil led to the decrease of pH and the increase of ionic strength, colloidal Ca content and positive charges of colloid surfaces, which promoted colloid aggregation and enhanced soil P sorption capacity. This restricted the loss potential of Pcoll. In summary, controlled flooding and drainage when managed correctly have a role to play in mitigating Pcoll loss from P-enriched calcareous soils.

Contribution of Biogas Slurry-Derived Colloids to Plant P Uptake and Phosphatase Activities: Spatiotemporal Response

The bioavailability for varied-size phosphorus (P)-binding colloids (Pcoll) especially from external P sources in soil terrestrial ecosystems remains unclear. This study evaluated the differential contribution of various-sized biogas slurry (BS)-derived colloids to plant available P uptake in the rhizosphere and the corresponding patterns of phosphatase response. Keeping the same content of total P input (15 mg kg-1), we applied different size-fractioned BS-derived colloids including nanosized colloids (NCs, 1-20 nm), fine-sized colloids (FCs, 20-220 nm), and medium-sized colloids (MCs, 220-450 nm) respectively to conduct a 45-day rice (Oryza sativa L.) rhizotron experiment. During the whole cultivation period, the dynamics of chemical characteristics and P fractions in each experimental rhizosphere soil solution were analyzed. The spatial and temporal dynamics examination of P-transforming enzymes (acid phosphatases) in the rice rhizosphere was visualized by a soil zymography technique after 5, 25, and 45 days of rice transplantation. The results indicated that the acid phosphatase activities and its hot spot areas were significantly 1) correlated with the relative bioavailability of colloidal P (RBAcoll), 2) increased with the colloid-free (truly dissolved P) and BS-derived NC addition, and 3) affected by the plant growth stage. With the nanosized BS colloid addition, the RBAcoll and plant biomass were respectively found to be the highest (64% and 1.22 g plant-1), in which the acid phosphatase-catalyzed hydrolysis of organic Pcoll played an important role. All of the above suggested that nanosized BS-derived colloids are an effective alternative to conventional phosphorus fertilizer for promoting plant P uptake and P bioavailability.

Size and composition of colloidal phosphorus across agricultural soils amended with biochar, manure and biogas slurry

The long-term application of organic amendments like manure, biochar and biogas slurry can increase phosphorus (P) levels in agricultural soils; however, at present, it's not clear how this affects the P association with different mobile water-dispersible colloidal particles (Pcoll). Thus, this study aimed to assess the effects of the long-term application of different organic amendments on the abundance, size and compositional characteristics of Pcoll. For this purpose, a total of 12 soils amended with the above three organic amendments were sampled from the Zhejiang Province, China, and Pcoll were fractionated into nano-sized (NC; 1–20 nm), fine-sized (FC; 20–220 nm), and medium-sized (MC; 220–450 nm) by a combination of differential centrifugation and ultrafiltration steps. These three Pcoll forms together accounted for 74 ± 14% of the total soil solution dissolved P content, indicating that Pcoll release was a key process in the overland P transport from these soils. Soils treated with biochar showed lower Pcoll contents than those treated with manure or slurry alone; this effect should be further explored in a controlled inductive research approach. Compositional analysis showed that inorganic P was the predominant Pcoll form in the NC (54 ± 20%) and FC (63 ± 28%) fractions, but not in the MC (42 ± 26%) fraction. Among the three fractions, the organic carbon (OC)–calcium (Ca) complex was the major carrier of NC-bound Pcoll, MC-bound Pcoll was better correlated with OC–manganese/iron/aluminium colloids than with OC–Ca colloids, and both of these phenomena co-occurred in the FC fraction. The current study provides novel insights into the impact of various carbon amendments on the propensity for P loss associated with different soil mobile colloidal fractions, and will therefore, inform future agronomic and environmental-related policies and studies.