Global phosphorus shortage will be aggravated by soil erosion

Phosphorus is essential for all life, forming key components of DNA, RNA, ATP, and phospholipids. In natural ecosystems, soil-plant cycling and slow mineralization regulate P availability, while in managed systems fertilizer inputs replenish losses. Unlike nitrogen, there is no biological fixation that can offset P deficits at scale; mineral P fertilizers are derived from finite geological deposits. Global P demand and dependence on imported phosphate rock have increased, raising concerns about resource depletion and price volatility. A critical pathway of P loss from agricultural systems is soil erosion by water because most soil P is particle-bound. Despite recognition that erosion is a major P loss pathway, many global and regional P cycle assessments have not explicitly and spatially accounted for erosion-driven P losses. This study aims to quantify spatially explicit global P losses from cropland soils due to water erosion, evaluate continental and regional P balances with and without mineral fertilizer inputs, and assess how soil erosion aggravates P depletion risks under scenarios of constrained fertilizer supply.

Previous global assessments have examined P inputs (atmospheric deposition, weathering, chemical fertilizer) and outputs (harvest, runoff/erosion) with varying conclusions on regional deficits and surpluses (e.g., MacDonald et al.). Several studies highlighted opportunities to improve P-use efficiency and recycle manure and human waste, but often lacked spatially explicit treatment of erosion. Earlier extrapolations (e.g., Pimentel, Smith) and FAO averages suggested substantial erosion-driven P loss but with high uncertainty. Reported global cropland P losses span roughly 1–20 kg ha−1 yr−1 depending on method and scale. There is also debate on the role of atmospheric P deposition relative to erosional outputs at large scales. Overall, the literature agrees erosion is crucial to the global P cycle but quantification has been uncertain and spatially coarse; this work addresses that gap by integrating global erosion modeling with soil P fraction data.

• Study scope and spatial data: The analysis covers approximately 1.04 billion ha of global cropland at 0.5° × 0.5° resolution, consistent with the LUH v1 land-use harmonization dataset.
• Erosion modeling: Soil loss by water was modeled using RUSLE-type approaches focusing on sheet and rill erosion (inter-rill and rill processes). Gully erosion, landslides, and tillage erosion were not included. Resulting soil loss represents on-site soil displacement, not net exported sediment.
• Soil phosphorus pools: Global distributions of soil P fractions were compiled from published datasets (Hedley-type fractionation framework). Pools include labile inorganic P, stable organic P, mineral-associated inorganic P (e.g., apatite), occluded P, and total P. Plant-available P was defined as the sum of the most bioavailable fractions over timescales relevant to crop uptake (months). Due to uncertainties linking fractions to plant uptake, results are presented for total and plant-available pools.
• Fluxes and balances: Annual P inputs considered include atmospheric deposition and chemical fertilizer; outputs include plant uptake (harvest) and erosion-driven P loss. Manure and crop residue inputs were included to represent organic P management. Organic P management was defined as manure + residue inputs − plant uptake. Two balance conditions were assessed: with chemical fertilizer and a hypothetical scenario without chemical fertilizer (to emulate constrained mineral P supply). The P balance equation: Balance = atmospheric deposition + organic P management (+ chemical fertilizer, if included) − P loss due to soil erosion.
• Computation of P loss due to erosion: For each grid cell, P loss (kg ha−1 yr−1) was computed by combining modeled soil loss (t ha−1 yr−1) with soil P content in relevant pools, using P_loss ≈ P fraction concentration × soil mass loss. Results were aggregated to regional and continental scales.
• Verification with riverine P exports: To compare on-site erosion-derived P loss with off-site river loads, literature values for sediment delivery ratios (SDR) were applied (11–30% for large basins) to estimate the fraction of eroded P reaching rivers. Modeled exports were compared to observed or modeled riverine P loads from global and regional studies, recognizing contributions from non-agricultural sources and the exclusion of gullies/landslides in the erosion model.
• Error estimation: Relative, non-symmetric uncertainties were propagated from the variability in soil P pools and input datasets. Global-scale comprehensive uncertainty quantification is limited due to unknown uncertainties in multiple input layers.
• Data sources: P inputs (chemical fertilizer, manure, residues), plant uptake, and atmospheric deposition were based on published global databases and modeling frameworks referenced in the study; spatial soil P pools followed global compilations; erosion factors used high-resolution environmental layers consistent with prior global erosion risk assessments.

• Global and continental P balances: Nearly all continents show negative P balances (net P loss from agricultural systems), except Asia, Oceania, and Australia; Asia is approximately balanced despite high fertilizer inputs.
• Drivers of depletion: Africa exhibits high depletion due to low affordability and use of mineral fertilizers; South America due to inefficient organic P management combined with high erosion; Eastern Europe due to a combination of low fertilizer use and management deficits.
• Magnitude of losses: Under a hypothetical scenario with no chemical fertilizer inputs, agricultural soils globally would be depleted by approximately 4–19 (up to ~20) kg P ha−1 yr−1, highlighting strong dependence on non-renewable mineral P.
• Erosion contribution: Average P losses due to water erosion account for over 50% of total P losses in many regions. Global average P loss from arable soils due to water erosion is on the order of 5.9–17.9 kg P ha−1 yr−1. Total global losses due to water erosion from arable soils are estimated at roughly 6.3–19.7 kg P ha−1 yr−1, within the lower range of literature estimates.
• Hotspots: Extremely high erosion-driven P losses (>20 kg ha−1 yr−1) occur in eastern China, many Indonesian regions, parts of eastern and south-eastern Africa (e.g., Ethiopia, Eritrea, Mozambique), Central America, and parts of South America (south-eastern Brazil, southern Chile, Peru). Very high losses (10–20 kg ha−1 yr−1) occur in parts of Southern Africa (South Africa, Madagascar, Tanzania) and Bolivia; high losses (5–10 kg ha−1 yr−1) are widespread in India, Angola, Zambia, and Uruguay.
• Comparison with river loads: When applying plausible sediment delivery ratios (11–30%), modeled on-site erosion P losses produce river export estimates within ranges reported by global and regional studies, though generally at the lower end, consistent with the exclusion of gully/landslide processes and mixed-source river loads.
• Implication: Fertilizer inputs currently mask substantial erosion-driven depletion; without them, soil fertility would rapidly decline. Erosion control and improved P recycling/management are critical to avoid aggravating global P scarcity and eutrophication risks downstream.

The integration of spatially explicit erosion modeling with soil P pools demonstrates that water erosion is a dominant pathway of P loss from croplands, significantly contributing to negative P balances across most continents. This finding directly addresses the research question by quantifying erosion-driven P depletion and showing its magnitude relative to other fluxes. The dependence on mineral P fertilizers is evident: where inputs are high, balances may be neutral or slightly positive; without fertilizer, widespread depletion of 4–19 kg ha−1 yr−1 would occur. Verification against riverine P exports supports the plausibility of on-site estimates, acknowledging that only a fraction of eroded P reaches rivers, with substantial redeposition within catchments. However, redeposition does not mitigate the agronomic loss of P from fields and contributes to off-site eutrophication risks. Regional patterns link biophysical drivers (rainfall erosivity, extreme events), land use intensity, and management practices to P loss hotspots. The results underscore that minimizing erosion is indispensable but insufficient alone; improving organic P management, recycling, and P-use efficiency must complement erosion control to sustain soil fertility and reduce environmental impacts.

Soil erosion by water substantially accelerates depletion of phosphorus from agricultural soils worldwide, with erosion often responsible for more than half of total P losses. While some regions maintain near-balanced budgets through high mineral fertilizer inputs, global agriculture remains vulnerable due to reliance on finite phosphate rock. In scenarios with constrained fertilizer supply, widespread P depletion (4–19 kg ha−1 yr−1) would occur. Therefore, reducing soil erosion to the feasible minimum, combined with integrated soil fertility management—optimizing mineral and organic inputs, recycling organic residues and manures, and incorporating legumes/cover crops—is essential. Future research should refine global estimates by incorporating gully, tillage, and mass-movement processes; improve sediment delivery and source apportionment to rivers; reduce uncertainties in soil P fraction distributions and plant availability; and evaluate region-specific management interventions to close P cycles sustainably.

• Erosion processes: The erosion model includes only sheet and rill (inter-rill and rill) erosion; gully erosion, landslides, and tillage erosion are excluded, likely underestimating P loss in some regions.
• Sediment delivery and validation: Sediment delivery ratios are uncertain at continental to global scales; comparisons to river loads include mixed sources (agricultural, urban, industrial), complicating validation.
• Soil P pools and availability: Associations between measured P fractions and plant availability carry significant uncertainty; some cropland areas lack sufficient data and were excluded.
• Process coverage: Leaching losses of P were neglected, assumed minor relative to particulate loss, but may be non-negligible in specific settings.
• Input data uncertainty: Global-scale uncertainty quantification is limited due to unknown uncertainties in multiple inputs (erosivity, soil properties, management data).