3.1 Crop yield, harvest phosphorus content, and exported phosphorus
In LTFE1, the average maize grain yield was 8.4 ± 0.3 t ha-1 and ranged from a minimum of 4.8 ± 0.1 t ha-1 in 1977 to a maximum of 11.4 ± 0.3 t ha-1 in 1989 (Fig. 2), without differences among treatments (P = 0.85). With OWP fertilisation, the yields in LTFE2 were not significantly different despite the discrepancies in P inputs among the different organic products (P = 0.65). Without adding P fertilisation (except a low rate in 2014, 16 years from the start of the trial), yields (average: 7.5 ± 0.4 t ha-1) have decreased significantly than those in the other treatments (8.4 ± 0.1 t ha-1; P = 0.002). In LTFE3, average crop yields were 8.8 ± 0.4 t ha-1. Despite the difference in P input through OWP and the control, no difference was observed in yield among the treatments (P = 0.99).
For the LTFE1 experiment, the average maize grains P content was 3.2 ± 0.1 g P kg-1 and was significantly affected by P rate (P = 0.02), with a maximum for TSP79 in 1984 (4.0 g P kg-1) and a minimum for CON0 in 1992 (2.6 g P kg-1). The control treatment differentiated between the two TSPs in 1988 and remained significantly lower (P = 0.003; Fig. 2). In LTFE2 and LTFE3, average harvest P contents were 3.1 ± 0.1 g P kg-1, and no differences were observed among treatments (P = 0.43 and 0.99 for LTFE2 and LTFE3, respectively).
In LTFE1, the average exported P in harvest was 26.5 ± 0.8 kg P ha-1, and all treatments presented a positive increase over time. However, this increase was lower for CON0, which began to differentiate from the other two treatments in 1984 (Fig. 2). In LTFE2, the exported P in CON3 was slightly lower (23.0 ± 1.2 kg P ha-1) than the four other treatments (OWP average: 27.5 ± 0.7 kg P ha-1; P < 0.0001). Conversely, no difference among treatments was observed in LTFE3 with an average of 25.3 ± 0.6 kg P ha-1 (P = 0.70).
Figure 2 Annual yield, its P concentration and annual P exported as affected by fertilization treatments and years for the three studied LTFE. Values are means ± std. err. (n = 4). CON: no P applied; MWS: municipal solid waste compost; FYM: farmyard manure; FYMC: FYM compost; SLU: dehydrated urban sewage sludge; GWS: compost of green waste and SLU; BIOW: bio-waste compost; TSP: triple superphosphate. Subscript values in treatment abbreviations are the average annual rate of P application, in kg P ha-1 yr-1.
3.2 The OWP main composition
On average, OWP applied at LTFE2 contained 280 ± 10 g C kg-1, 20 ± 1 g N kg-1 and 6.7 ± 0.6 g P kg-1. In LTFE3, OWPs were in the same concentration range with 332 ± 11 g C kg-1, 32 ± 3 g N kg-1, and 12.3 ± 1.6 g P kg-1. In LTFE2 and LTFE3, the C, N, and P contents were significantly different from the OWPs (Table 2).
The C:P ratios of the OWPs applied at LTFE2 were 1060 (MSW24), 117 (FYM40), 706 (BIOW46), and 135 (GSW112). At LTFE3, the N:P ratios were 65 (BIOW17), 11 (FYM21), 10 (FYMC21), 12 (SLU36), and 12 (GWS45). Because of the highly different C:N:P ratios and fertilisation reasoning based on C or N, P inputs were highly variable among OWPs. Cumulative over 17 years of trial (LTFE2), ten OWP applications of 420, 699, 819, and 2008 kg P ha-1 were added to the soil for MSW24, FYM40, BIOW46, and GSW112. Cumulative organic P inputs (% of TPHF) were respectively 37 (9%), 241 (34%), 64 (8%), and 325 (16%) kg P ha-1. In LTFE3, the cumulative TPHF inputs were 250, 317, 317, 545, and 667 kg P ha-1 for BIOW17, FYM21, FYMC21, SLU36, and GWS45, respectively, including 18 (7%), 114 (36%), 114 (36%), 93 (17%), and 107 (16%) kg P ha-1 of organic P.
3.3 P incorporation to SOP and the release of phosphate ions in solution
On average for all sites, years, and treatments, 39 ± 2 kg P ha-1 was applied to the soil through crop residues and OWPs, with significant differences between sites (P < 0.0001), i.e., 15 ± 1 (LTFE1), 57 ± 7 (LTFE2), and 35 ± 4 (LTFE3) kg P ha-1 yr-1. At all sites, the sum of all P inputs (i.e., aboveground, belowground, grain residues, and total P in OWPs) was significantly different among treatments (LTFE1, P = 0.02; LTFE2, P < 0.00001; LTFE3, P < 0.0001). Details on the average values, significant differences across treatments, and P inputs across years, treatments, and LTFEs are shown in SI: Tables 4-SI to 6-SI and Fig. 3-SI.
In LTFE1, aboveground residues represented the most important P flux from maize returned to the soil (average of 7.8 ± 0.3 kg P ha-1 yr-1; 52%). Average P returned to aboveground and belowground residues increased significantly with increasing P addition (14.2 ± 0.5; P = 0.02), while P returned to grain residues was homogeneous among P fertilisation rates (0.8 ± 0.1 kg P ha-1 yr-1; P = 0.11). Therefore, cumulative P return within TSP79 crop residues was higher than in TSP27 or CON0 (315 ± 2, 267 ± 5, and 224 ± 9 kg P ha-1, respectively).
In LTFE2 and LTFE3, the most important P input was P applied as OWP (44 ± 7 and 23 ± 3 kg P ha-1 yr-1; 77% and 66%, respectively). Within crop residues, belowground residues represented the main P input flux (8.8 ± 0.5 and 7.4 ± 0.5 kg P ha-1 yr-1; 63% and 60%, respectively for LTFE2 and LTFE3). In the LTFE2 grain residues, FYM40 was slightly higher than CON3 (P = 0.02); otherwise, no discrepancies were observed among treatments, regardless of the site or residue type. Therefore, cumulative crop residues P input in LTFE2 were 201 ± 6 (CON3), 232 ± 6 (MSW24), 256 ± 3 (FYM40), 241 ± 6 (BIOW46), and 257 ± 5 (GWS112). In LTFE3, the cumulative P input in crop residues were 164 ± 3 (CON8), 168 ± 1 (BIOW17), 182 ± 4 (FYM21), 180 ± 1 (FYMC21), 176 ± 2 (SLU36), and 175 ± 1 (GWS45).
On average, decomposition of crop residues and OWPs over one year released 12.5 ± 0.5, 46 ± 6, and 27 ± 3 kg P ha-1 yr-1 as phosphate ions in solution for LTFE1, LTFE2, and LTFE3, respectively, i.e., 83% (LTFE1), 80% (LTFE2) and 71% (LTFE3) of P applied as OWPs and crop residues. Rates of P incorporated to SOP stock were 2.5 ± 0.1, 11.0 ± 1.4, and 7.9 ± 0.7 kg P ha-1 yr-1 for LTFE1, LTFE2, and LTFE3, respectively.
3.4 Multiannual evolution of SOP and SIP stocks compared to the cumulative P budget
Averaged over treatments and years, TPHF stocks were 1500 ± 74, 2577 ± 89, and 4307 ± 27 kg P ha-1 in LTFE1, LTFE2, and LTFE3, respectively, whose 451 ± 7, 635 ± 14, and 1209 ± 16 kg P ha-1 were SOP stocks and the remainder were SIP stocks (1048 ± 73, 1941 ± 85, and 3104 ± 30 kg P ha-1).
In the LTFE1 treatment, 487 and 1448 kg P ha-1 were cumulatively applied to TSP27 and TSP79, respectively. Therefore, SIP stock dynamics were significantly affected by treatments at all sites (Fig. 3; Table 3-SI); SIP stock decreased when no P was applied, and increased significantly in LTFE1 and LTFE2 for treatments with the highest P inputs.
Figure 3 Evolution of soil inorganic P (SIP, dotted line) and soil organic P (SOP, solid line) stocks in the ploughed layer as affected by plot cumulative P budget (Bcum). Values are means ± std. err. (n = 4). Lines are linear regressions and grey areas are confidence interval (95%) of the linear regressions. CON: no P input; MWS: municipal solid waste compost; FYM: farmyard manure; FYMC: FYM compost; SLU: dehydrated urban sewage sludge; GWS: compost of green waste and SLU; BIOW: biowaste compost; TSP: triple superphosphate. Subscript values in treatments abbreviations are the average annual rate of P application, in kg P ha-1 year-1.
The initial SOP stocks varied significantly between sites (P < 0.0001; Table 5; Fig. 3). In LTFE1, the SOP stocks remained stable over the 17 years of cropping (Fig. 3) because there were no significant differences among treatments and years (ANOVA: treatment P = 0.64; year P = 0.97). The SOP stocks in LTFE2 increased significantly (Fig. 3), regardless of the OWP treatment (P = 0.30). LTFE2 and LTFE3 showed a significant positive increase in SOP stocks (Fig. 3). No treatment differences were observed (P = 0.13).
The average SOP mineralisation k coefficient, was 0.005 ± 0.001 year-1, 0.018 ± 0.004 year-1 and 0.004 ± 0.001 year-1 for LTFE1, LTFE2, and LTFE3, respectively. The corresponding average SOP residence times (1/k) were 217, 56, and 227 years, respectively.
$${SOP}_{t+1}=\left(\frac{\sum \left({P}_{j}\times {h}_{j}\right)}{k}\right)\times \left(1-{e}^{-k}\right)+{SOP}_{t}\times {e}^{-k}$$
The k value at LTFE2 was significantly higher than that at the other two sites (P < 0.0001), with a residence time of less than 100 years. LTFE2 was also the only site where the fertilisation treatment affected k optimisation, with MSW24 at the lower limit set for optimisation (Table 5). Thus, the SOP residence time in the GWS112 treatment was much shorter than the other treatments (approximately 30 years). However, the k differences in LTFE2 were not reflected in the gross annual rate of SOP mineralisation, and none of the sites had a treatment effect on this variable. On average, rates of SOP mineralisation were 2.1 ± 0.1, 11.2 ± 0.5, and 5.4 ± 0.3 kg P ha-1 yr-1, for LTFE1, LTFE2, and LTFE3, respectively, with significant differences between sites (Table 5).
Table 5
Main features of the mineralisation of soil organic P (SOP) stocks (kg P ha-1 yr-1) in the plough layer of the three studied LTFEs. Parameters (± std. dev.) of initial stock; coefficient of SOP mineralisation: k (yr-1); 1/k: residence time (yr); annual rates of SOP mineralisation and incorporation (kg P ha-1 yr-1). Values are means ± std. err. Different lower-case letters (a, b, c) indicate significant (α = 0.05) differences between treatments within a given LTFE. Different upper-case letters (A, B, C) indicate significant (α = 0.05) differences between LTFEs. CON: no P input; MWS: municipal solid waste compost; FYM: cattle dairy farmyard manure; FYMC: FYM compost; GWS: compost of green waste and SLU; SLU: dehydrated urban sewage sludge; BIOW: biowaste compost, TSP: triple superphosphate. Subscript values in treatment abbreviations are the average annual rate of P application, in kg P ha-1 yr-1.
Treatment | Initial SOP stock | Coefficient k | 1/k | Mineralisation rate | Incorporation rate (OWP and residues) | Decomposition rate (OWP and residues) |
| kg P ha-1 | yr-1 | yr | kg P ha-1 yr-1 | kg P ha-1 yr-1 | kg P ha-1 yr-1 |
LTFE1 |
CON0 | 446 ± 13 B | 0.0046 ± 0.0011 B | 217 | 2.1 ± 0.1 C | 2.1 ± 0.1 b | 10.4 ± 0.4 b |
TSP27 | 2.5 ± 0.1 ab | 12.4 ± 0.7 ab |
TSP79 | 2.9 ± 0.2 a | 14.6 ± 1.0 a |
Treatment effect P-value | 0.59 | 0.81 | | 0.81 | 0.02 | 0.02 |
Mean over Treatments | | | | | 2.5 ± 0.1 C | 12.5 ± 0.5 C |
LTFE2 |
CON3 | 595 ± 28 B | 0.011 ± 0.006 ab | 91 | 11.2 ± 0.5 A | 3.1 ± 0.4 b | 8 ± 1 c |
MSW24 | 0.001 ± 0 b | 1000 | 5.3 ± 0.8 b | 31 ± 5 bc |
FYM40 | 0.030 ± 0.003 a | 33 | 17.8 ± 3.6 a | 35 ± 6 b |
BIOW46 | 0.015 ± 0.008 ab | 67 | 6.9 ± 1.2 b | 52 ± 10 b |
GWS112 | 0.031 ± 0.011 a | 32 | 21.8 ± 4.7 a | 104 ± 22 a |
Treatment effect P-value | 0.93 | 0.03 | | 0.10 | < 0.0001 | < 0.0001 |
Mean over Treatments | | 0.018 ± 0.004 A | 56 | | 11.0 ± 1 A | 46 ± 6 A |
LTFE3 |
CON8 | 1145 ± 22 A | 0.0044 ± 0.0009 B | 227 | 5.4 ± 0.3 B | 3.6 ± 0.5 c | 7 ± 1 c |
BIOW17 | 4.9 ± 0.5 c | 23 ± 5 b |
FYM21 | 11.6 ± 2.3 a | 22 ± 4 bc |
FYMC21 | 9.4 ± 1.6 ab | 24 ± 5 b |
SLU36 | 6.8 ± 0.9 bc | 41 ± 10 a |
GWS45 | 11.0 ± 2.1 a | 45 ± 11 a |
Treatment effect P-value | 0.22 | 0.48 | | 0.52 | < 0.0001 | < 0.0001 |
Mean over Treatments | | | | | 7.9 ± 0.7 B | 27 ± 3 B |
LTFE effect P-value | < 0.0001 | 0.0009 | | < 0.0001 | < 0.0001 | < 0.0001 |
3.5 Model precision and accuracy
In the three study sites, model precision and accuracy were high, with a regression slope line of the SMA of the simulated versus observed SOP close to 1 (Fig. 4) and an RMSE of 129 kg P ha-1. The PLA was 7% and the PLP was (93%). Therefore, the apparent model with low precision and high accuracy was attributed to the initial SOP stocks, which were highly variable among and within sites (Table 5), limiting model precision. This statement was supported by an RRMSE (15%) that was twice as low as the coefficient of variation of the observed SOP (40% across all sites) but higher than the AMG model applied to C (approximately 5%) (Clivot et al., 2019). However, the model was still accurate because its predicted SOP values remained the same magnitude as the observed SOP values. The management history of the plots may explain the origin and difference among the initial SOP stocks, but this could not be verified.
Figure 4 Simulated vs observed SOP stocks in the ploughed soil layer by the Hénin and Dupuis (1945) model transposed to SOP.
The solid line is the standardized major axis regression line (SMA). The dashed line is the 1:1 line. CON: no P input, MWS: municipal solid waste compost, FYM farmyard manure, FYMC: FYM compost; GWS: compost of green waste and SLU: SLU: dehydrated urban sewage sludge; BIOW: bio-waste compost; TSP: triple superphosphate. Numbers in treatments names are P annually applied on average (kg P ha-1 year-1). Numbers in treatments names are P annually applied on average (kg P ha-1 year-1). LTFE1: n = 48, LTFE2: n = 140, LTFE3: n = 144.