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

Soil colloids have been shown to play a critical role in soil phosphorus (P) mobility and transport. However, identifying the potential mechanisms behind colloidal P (P) release and the key influencing factors remains a blind spot. Herein, a machine learning approach (random forest (RF) coupled with partial dependence plot analyses) was applied to determine the effects of different soil physicochemical parameters on P content in three colloidal subfractions (i.e., nano- (NC): 1-20 nm, fine- (FC): 20-220 nm and medium-sized colloids (MC): 220-450 nm) based on a regional dataset of 12 farmlands in Zhejiang Province, China. RF successfully predicted P content (R = 0.98). Results showed that colloidal- organic carbon (OC) and minerals were the major determinants of total P content (1-450 nm); their critical values for increasing P release were 87.0 mg L for OC, 11.0 mg L for iron (Fe) or aluminium (Al), 2.6 mg L for calcium (Ca), 9.0 mg L for magnesium (Mg), 2.5 mg L for silicon (Si), and 1.4 mg L for manganese (Mn). Among three colloidal subfractions, the major factors determining P were soil Olsen-P (P; 125.0 mg kg), Ca (2.5 mg L), and colloidal P saturation (21.0%) in NC; Mn (1.5 mg L), Mg (6.8 mg L), and P (135.0 mg kg) in FC; while Mn (1.5 mg L), Al (2.5 mg L), and Fe (3.8 mg L) in MC, respectively. OC had a considerable effect in the three fractions, with critical values of 80.0 mg L in NC or FC, and 50.0 mg L in MC. Our study concluded that the information gleaned using the RF model can be used as crucial evidence to identify the key determinants of different size fractionated P contents. However, we still need to discover one or more easy-to-measure parameters that can help us better predict P.