Phosphate leaching in the excessively fertilised Flemish region: a geochemical and GIS approach.

The intensification of agriculture has been associated with excessive addition of phosphorus (P) fertilisers to agricultural soils. The excess P in soil is a legacy for the environment since P slowly migrates downward by leaching towards the groundwater. Once P reaches surface water, eutrophication is induced and a limiting natural resource is lost. The current understanding of the leaching of P states that P leaching is slow and that P is strongly retarded by the sorption of the orthophosphate (PO4) anions on iron (Fe) and aluminium (Al) oxyhydroxides in soil or by precipitation as calcium phosphates in neutral or calcareous soils. Phosphorus only becomes highly mobile once the sorption sites are saturated. The current understanding predicts that the mobility of P is related to the Degree of Phosphate Saturation of the Fe and Al oxyhydroxides (DPS) and that there is very limited P migration in most European soils because the sorption saturation is only rarely reached. However, this prediction is contradicted by observations of significant subsurface P accumulation in fertilised soils and loss of P via drainage pipes, the major reasons being unclear. The general objective of this thesis was, therefore, to understand and quantify the biogeochemical processes that affect the migration of P in agricultural soils. The underlying idea was that the current DPS-model needs to be validated for more contrasting soil types (since it was established for noncalcareous sandy soils from the Netherlands) and that three factors, not yet embedded in the current concept required attention i.e. 1) sorption kinetics, 2) colloid-facilitated P transport and 3) effects of anaerobic redox reactions on P migration.

First, the sorption of PO4 on Fe and Al oxyhydroxides was confirmed to control leachate P concentrations from 120 unsaturated columns, filled with agricultural soils with contrasting properties, including pH-neutral soils (Chapter two). Leachate P concentrations ranged from 0.7 to 240 µM and increased as the ratio of P to Fe and Al in acid oxalate soil extracts (i.e. DPS) increased and as the PO4-distribution-coefficient (determined by a radiotracer experiment) decreased. This shows that the DPS-model can be extended to pH-neutral soils as well. Surprisingly, leachate P increased with increasing leachate Fe and Al concentrations. Surface complexation modelling was done to describe PO4 sorption to ferrihydrite (using the CD-MUSIC model). This yielded a reasonable description of leachate P concentrations, but only when reactive PO4 was described from isotopically exchangeable PO4, when organic matter was included as the main competing adsorbate and when mobile colloidal ferrihydrite was included. The model revealed that PO4-binding by colloids enhanced leachate PO4 concentrations up to a factor 50 (in comparison to the leachate PO4 concentrations in absence of colloids), mostly at small DPS and at small calcium (Ca2+) concentration in solution, likely because a high Ca2+ concentration causes flocculation of the organic-matter stabilised Fe and Al colloids.

A second deviation from the current model was revealed in a Luvisol after 19-years of application of different organic fertilisers where P migrated more than expected from the DPS concept (Chapter four). Anaerobic respiration mobilised P and colloids, even in the unsaturated zone of this upland soil. Extracts (10-3 M calcium chloride, CaCl2) of field-moist soil, sampled at distinct depths, and in situ wick samplers indicated a markedly strong correlation between soluble (< 0.45 µm) P and manganese (Mn) and both peaked in and below the compacted plough pan, where a high water-filled porosity and a high organic carbon (OC) content coincided. Waterlogged soil incubations confirmed that anaerobic respiration co-mobilises Mn and colloidal P. The long-term applications of farmyard manure and an immature compost enhanced soluble Mn, Fe, and Al in the subsurface with respect to the mineral N treatment, whereas less such effect was found under the application of more stable organic fertilisers. Farmyard manure application significantly enhanced soil P stocks below the plough layer despite a small net positive soil P balance. Overall, multiple lines of evidence confirm that anaerobic respiration, sparked by labile organic matter, mobilises P in this seemingly well-drained soil.

Additional soil cores (57 in total) were sampled down to 160 cm depth at contrasting locations to identify the controlling soil properties of redox-related mobilisation of P and its role on P migration (Chapter five). Field-moist soil extracts (10-3 M CaCl2) showed that reductive dissolution of Mn and Fe oxyhydroxides generally occurred and this was enhanced at low pH and a high OC content. This process enhanced soluble (< 0.45 µm) Fe(III) and Al concentrations, that are most likely colloidal particles, which enhanced, in turn, soluble P concentrations. The depth distributions of P in these 57 soil cores were empirically fitted with a sigmoidal curve to identify the soil factors that explain actual P migration. In fertilised plots, P enrichments down to 70 cm depth were detected. The depth distributions showed much more P dispersion than what reasonably can be estimated by a transport model under nonlinear equilibrium sorption. The slopes of the DPS-depth sigmoidal lines were more spread out, i.e. more disperse, with increasing OC content and decreasing pH of the topsoils, suggesting that anaerobic-induced P mobilisation explains the large dispersion.

Thirdly, the potential disequilibrium of PO4 sorption creates another possible source of error for models assuming local equilibrium sorption (e.g. the CD-MUSIC model; Chapter three). A new rate constant distribution (RCD) model, that assumes frequency distributions of both adsorption and desorption rate constants, was developed and compared with other kinetic models. Batch radiolabeled PO4 (33PO4) sorption was measured in agitated suspensions between two minutes and 20 days after spiking in thirteen contrasting types of soil and two iron oxyhydroxides. Overall, the RCD model, with three adjustable parameters, describes the data better than the other models tested. The so-called slow reactions, denoted as the factor change in soluble 33PO4 between one and 20 days after spiking, were described better by the RCD model and ranged from 1.0 (i.e. no change) to 6.9. Orthophosphate sorption on ferrihydrite and on some soil samples with a high ratio of poorly crystalline iron oxyhydroxides to total iron did not cease within 20 days. Thus, the lack of sorption equilibrium may be a source of error for equilibrium models.

Taken together, this study confirms that the interaction between PO4 and Fe and Al oxyhydroxides, i.e. the DPS,  is crucial to understand the mobility of P in fertilised soils. However, this interaction is more dynamic than previously modelled and thought; 1) the PO4 sorbing Fe and Al oxyhydroxides can detach from the soil’s solid phase as mobile colloids and mediate vertical PO4 transport;  2) PO4 sorption on Fe(III) and Al oxyhydroxides is not instantaneous and 3) reductive dissolution of Mn and Fe oxyhydroxides triggers the mobilisation of  P, even in the unsaturated zone of agricultural soil. Models that succeed at incorporating these dynamic interactions will offer an improved understanding of the fate of legacy P in agricultural soil.