The availability of phosphorus in a soil phosphorus depleting scenario

Phosphorus (P) is an essential nutrient for plant growth. The development of intensive agriculture in Europe during the last century has been associated with an era of excess P fertilizer use to warrant an optimal food production. This was mainly due to the excessive input of P to agricultural soils by animal manure in areas with intensive animal husbandry. This caused an accumulation of P in soil (legacy P). The non-renewability of P resources and the concerns regarding eutrophication of surface waters have prompted the demand for a more sustainable use of P fertilizers. This is put into practice by decreasing P fertilizer application on agricultural fields and even targeting a negative soil P balance, i.e. when the annual removal of P with the harvested crop exceeds the annual P returns to the soil, for agricultural soils with an excessively large soil P status. It is well established that the sorption and desorption of phosphate anions in soils have a pronounced hysteresis suggesting that the plant availability of this legacy P is below that of freshly added P. The goal of this study is to determine crop and soil factors affecting the availability of P in soils in a soil P depleting scenario (i.e. in a soil P mining scenario) and to identify soil tests which adequately predict this. In a soil P depleting scenario, the main P source for plants is coming from the legacy P rather than the annual fertilizer application, and the desorption rate of P might become a limiting factor controlling the amount of available P in soil solution. It is speculated that in such scenario, the crop growth rate, which affects the P uptake rate, is a critical factor in determining whether plants suffer from P deficiency or not. In practice, this was approached by, first, evaluating soil P tests which most optimally predicts crop responses to P in European soils. Secondly, experiments and modelling were used to determine the extent to which legacy P is available for uptake by plants under a P mining scenario. The evaluation of soil P test's capacity to predict crop response, e.g. yield compared to maximal yield (i.e. yield in a situation with optimal P availability), to soil P was done on soils from European long-term field trials and on Flemish soils, the latter were first subjected to a P mining experiment. The five soil P tests evaluated in this study were: extraction with ammonium oxalate (P-Ox), extraction with ammonium lactate (P-AL), extraction with 0.5 M NaHCO3 (P‑Olsen), extraction with 0.01 M CaCl2 (P-CaCl2) and the diffusive gradient in thin film technique (P-DGT). The first three indicate the soil P quantity (Q, i.e. the P which is reversible adsorbed on the solid phase; Q-tests) whereas the latter are a measure of the soil P intensity (I, i.e. soil solution concentration; I-tests). First the soil P tests were evaluated on soils from European long-term field trials (some > 100 years old) (Chapter 2). The Olsen and AL extraction performed best in predicting relative crop yield based on the R² of fitted Mitscherlich curves. The soil P tests had a similar but small 95% confidence interval surrounding the general critical P value, i.e. the soil P test value for which the relative yield equaled 95%. In contrast, when evaluating the variance in critical P values among different soils and years, the DGT technique performed worst and the Ox extraction was the most robust of the five soil P tests in distinguishing the fertile from the P deficient soils. A combination of two tests was also evaluated and the relative yield was related to the combination of an I- and a Q-test. This analysis showed that there was little added value of including the I-tests in predicting crop response on the soils of the European long-term field trials, i.e. the P bioavailability in European soils is most affected by the P quantity of the soil. Secondly, the soil P tests were evaluated on Flemish soils from an accelerated soil P mining experiment (Chapter 3). Perennial ryegrass (Lolium perenne, Melpetra tetra) was grown for two years in a greenhouse on 5-cm deep soil layers of eight contrasting soils with periodic grass clipping. Each soil was split into four fertilizer treatments, i.e. no P and adequate P and two nitrogen levels, the latter to alter the rate of P uptake. The difference in P-Ox and P-AL values between the initial and final soil samples of the depletion experiment corresponded best to the cumulative P uptake by the grass during that period, i.e. they close the mass balance best. None of the soil P tests could adequately diagnose yield losses based on the final soil P content of the sample. In contrast, the relationship between the relative P uptake (more sensitive to P deficiency than yield data, but agronomical less relevant than relative yield) and the soil P tests was best for P-DGT and worst for P-Ox. All soil P tests (except for the Ox extraction) performed well in predicting the cumulative P uptake by plants at which the relative yield decreased below 90%, further denoted as the critical cumulative P uptake (CCP), based on the initial soil P values. This indicates the predictive properties of the soil P tests. The largest R² value for the relationship between CCP and the initial soil P values was observed for P-DGT and P-AL, and for P-Olsen the relation followed an almost a 1:1 line. The general outcome of the evaluation of the soil P tests is that none of the tests consistently outperformed other tests in all evaluation criteria and the results showed that the soil P quantity is currently the main factor in determining the P availability of European soils. The effect of the crop growth rate and the effect of mining on the P availability in a P depleting scenario were determined by experiments and modelling. The effect of the crop growth rate was experimentally studied by including different N fertilizer applications rates in the P mining trial to induce different crop growth rates, i.e. different P removal rates from the soil. The grass grown at adequate N supply (+N) had a faster growth and a faster decrease in shoot P content compared to the grass grown at limited N supply (-N). Such results are rather trivial, however more importantly was that the cumulative P uptake where yield declines due to P deficiency was, on average, a factor 1.7 larger for the -N treatment than for the +N treatment. This illustrates that for a certain soil P stock, faster growing plants can access less P before P deficiency occurs than slower growing plants. The effect of mining on the chemical P availability was illustrated in Chapter 4 by laboratory desorption tests during 77 days on the initial and final soil samples from the P depletion trial. The P desorption as a function of time was described with a fast reacting P pool (Q1) and a slow reacting P pool (Q2). The ratio of the size of the fast P pool to the slow P pool was, on average, 0.16 in the initial soil samples. Upon P uptake by the plant, this ratio decreased to 0.08, on average, indicating that more P was taken up from the fast P pool than is resupplied by the slow P pool. The effect of the crop growth rate and of mining on the P uptake (which is used as a measure of the P availability) was modelled by a mechanistic nutrient uptake model (Chapter 5), based on the model of Barber-Cushman that was extended with nutrient uptake kinetics which depends on the plant growth rate and with P desorption kinetics that control the solid-liquid partitioning of P. Simulations by the mechanistic nutrient uptake model showed that faster growing plants rely more on the desorption kinetics of the slow reacting solid phase pool than slowly growing plants. The modelling confirmed the experimental data that faster growing plants can take up less of the soil P stock than slower growing plants before they experience yield loss. The mechanistic model also showed that the reduction in P desorption between the initial and final soil samples of the mining trial was mainly caused by a reduction of the fast reacting P pool, which showed a decreasing trend similar to the soil solution P upon mining. In contrast, the slow reacting P pool decreased upon decrease of the fast reacting P pool but showed a different decreasing trend. This implies that the availability of P decreases with a decreasing size of the Q1 pool, and that, once the Q1 pool is depleted, the availability of P is controlled by the desorption kinetics of the Q2 pool. The P availability on the long-term in a P depleting scenario was determined based on (i) the P mining experiment and (ii) based on an empirical model. The results from the P mining experiment, extrapolated to field conditions and assuming that only the plough layer (0-23 cm layer) provides P to the crop, suggest that crops could grow maximally for eight years before 10% yield losses are observed (Chapter 3). The empirical model predicted P uptake over a longer period (Chapter 5): upon cessation of fertilization of an even highly fertile soil, the annual P uptake drops by 13% from current values after 10 years. Thereafter, almost another 20 years of mining is possible with an almost constant sustained crop offtake before the crop P uptake sharply drops by over 25% from current values. Since reduced P uptake is possible without associated crop yield loss (luxury P uptake), this suggests that current soil stocks allow soil P mining in arable fields for about 30 years with no considerable yield decline. The long-term modelling, however, showed a considerable hysteresis in the relationship between crop-offtake and soil P stock between the era of build-up (excessive P fertilization) of the soil P stock and a future scenario of soil P mining where the soil P stock is less available because of the slow desorption of P from the solid phase. This is, in theory, related to the lack of fresh P addition which constantly feeds the fast P pool in soil. In conclusion, this study showed that none of the current soil P tests outperformed the others in their capacity to predict P related yield losses on European soils. The results showed that the available P in European soils is currently controlled by the soil P quantity. Moreover, it was shown that slower growing crops can take up more legacy P in a depleting scenario than faster growing plants before they experience yield losses. Modelling suggested that the P availability in a depleting P scenario depends on the size of the fast desorbing P pool and that, upon depletion of this pool, the P availability is controlled by the desorption kinetics of the slow desorbing P pool. It is suggested that the soil P tests might be improved by taking the P desorption kinetics into account, i.e. dynamic P tests are suggested, and that the mechanistic modelling should be further expanded to identify the processes involved.status: publishe