. 2024 Jan 2;11:17. doi: 10.1038/s41597-023-02751-6

Table 1.

FARM driver and its representation in (data = GPASOIL-v0) and (data = GPASOIL-v1).

Driver name	FARM (farming practices)
Description of its effect on the soil P	Soil P input/output related to farming practices
Corresponding variables in our approach	Soil P input/output corresponding to: chemical fertilizer ( $f P_{c h e m}^{o u t \to i - l a b}$ ), manure reaching cropland/grassland soils ( $f P_{m a n u}^{o u t \to m}$ with m in {i-lab, o-lab, o-sta}), crop/grass residues ( $f P_{r e s i}^{o u t \to m}$ with m in {i-lab, o-lab, o-sta}), and plant uptake ( $f P_{u p t a}^{i - l a b \to o u t}$ ). Here, ‘residues’ refer to plant biomass that remains on/within the soil after the harvest (for cropland) or harvest/grazing (for grassland). It includes root biomass.
Dataset ref used in (data = GPASOIL-v0)	Chemical fertilizer and manure were taken from^{Bouwman et al.13}. Residues and plant uptake were derived from P harvest/withdrawal provided by Bouwman et al.¹³ following basic assumptions.
Issues related to the use of this dataset in GPASOIL-v0	The approach used in Bouwman et al.¹³ (based on IMAGE⁶⁷,) excluded ~20% of cropland area (due to some boolean treatment about land-use). This lead to some inconsistency with LUCC in GPASOIL-v0. In Bouwman et al.¹³, the treatment in grassland was simple: P withdrawal was estimated as ~90% of the total applied fertilizer. In GPASOIL-v0, we used few global numbers, without crop distinction, to derive uptake and residues from withdrawal/harvest (see Supp.Inf of Ringeval et al.¹¹).
Representation in (data = GPASOIL-v1)	For both cropland and grassland, P in manure reaching the soil was derived from a combination between N in manure reaching grassland soil given by Xu et al.³¹, a global P:N ratio (0.20), and P manure produced at country-scale based on livestock category population and manure production rate⁹. Grassland: P in chemical fertilizer was derived from N in chemical fertilizer applied to grassland given by Xu et al.³¹ and a global P:N ratio (0.22). P plant uptake was derived from the actual NPP (including above and belowground NPP) in grassland and some spatially constant parameters (P concentration, etc.). P in residues was derived from P plant uptake and (1-GI) with GI: grazing intensity given by Kastner et al.⁴¹ and corresponding to the ratio of the Human appropriation of net primary production through harvest and the actual NPP. The spatial distribution of NPP provided by Kastner et al.⁴¹ was used but global average NPP was made varying between the values of Kastner et al.⁴¹ and Wang et al.⁴². Cropland: P in chemical fertilizer was provided by Lu and Tian³⁴. The latter ref. ³⁴ assumed that all chemical fertilizer is applied to cropland and here, we corrected the chemical fertilizer applied to cropland by subtracting the chemical fertilizer applied on grassland described above. P uptake and P in residues were computed as functions of yield and crop specific parameters. Crop-dependent, spatially explicit and temporal varying yields are get by combining spatially-explicit yield per crop for year 2000 by Monfreda et al.³⁹ and country-averaged yield (but temporal-varying) given by FAOSTAT. The yield of FAOSTAT at country-scale is extended for years before 1961 by scaling the 1961 yield with the country-sale human population provided by HYDE 3.2²³. Yield is then combined with different crop dependent parameters (concentration of P in different organs -root, harvest, aboveground biomass excluding the harvest, root:shoot ratio and harvest index) and fraction of total residus remaining on/within soil (equal to 1/2 or 0, following Smil⁴⁴).
Characteristics of the spatially explicit dataset used in (data = GPASOIL-v1)	Xu et al.³¹ provided for grassland, chemical N fertilizer for years between 1961 and 2016, and N in manure for years between 1860 and 2016, both in kgN.km⁻².yr⁻¹, and at half-degree spatial resolution⁶⁸ (10.1594/PANGAEA.892940). Demay et al.⁹ provided P manure produced at country-scale per livestock category for years between 1950 and 2017 (⁶⁹10.57745/LEPJCS). Kastner et al.⁴¹ provided NPP (in gC m⁻² yr⁻¹) and GI (no unit) for grassland for some years (1910, 1930, 1950, 1960, 1970, 1980, 1990, 2000, 2010) at 0.083° spatial resolution(⁷⁰10.5281/zenodo.7313791). ^{Lu and Tian34} provided chemical P fertilizer applied on cropland for any years between 1900 and 2013, in gP m⁻² yr⁻¹, at half-degree spatial resolution, (⁷¹10.1594/PANGAEA.863323). Zhang et al.³⁰ provided N in manure applied on cropland for any years between 1860 and 2014, in in kgN.km⁻².yr⁻¹ at half-degree spatial resolution (⁷²10.1594/PANGAEA.871980). Monfreda et al.³⁹ provided crop-dependent and spatially explicit (half-degree resolution) yield for years ~2000 (Earthstat dataset, http://www.earthstat.org/harvested-area-yield-175-crops/). FAOSTAT provided country-averaged and crop-dependent yield for any years over 1961–2019 (FAOSTAT. License: CC BY-NC-SA 3.0 IGO. Extracted from: https://www.fao.org/faostat/en/#data/QCL and https://www.fao.org/faostat/en/#definitions Data of Access: 20-01-2023.). Klein Goldewijk²³ provided the country-sale human population (HYDE 3.2⁷³, 10.17026/dans-25g-gez3).