Phosphorus (P) supply from soils is crucial to crop production. Often, crop yields are constrained if no external P is supplied. The limitation of P on biomass production will become more severe under increasing nitrogen (N) global loading in the future [1
]. Phosphorus fertiliser is mainly produced from rock phosphate, which is not only a finite resource, but also a geographically unevenly distributed one, meaning that continued supply is subject to geopolitical threats [2
]. To support the demand for food with the projected increase in population this century [3
], an increase in P fertiliser application will be expected [4
]. Both limited supply and increased demand implies the importance of using the existing resources wisely in order to make agricultural production sustainable. The recovery of added P by crops is often low [5
], due to sorption reactions of P which form poorly soluble or insoluble compounds [7
]. Hence, there is a challenge in agriculture on how crop P use efficiency can be improved. P fertilisers can also contribute to P surpluses in soils and to P exports from land to surface waters. Therefore, an efficient way of maintaining optimum crop production through adding P fertilisers while reducing P losses to the environment is to manage P applications properly to ensure maximum crop acquisition of soil and fertiliser P [8
Total P in soils consists of many different fractions, including crystalline, occluded and adsorbed particulate organic, soluble organic and soluble inorganic P. Soil organic P compounds have been largely overlooked in P agronomy, but can represent up to 90% of soil total P [9
]. Methods have been developed to characterise soil organic P mainly focused on the identification and quantification of inositol phosphates [10
] and other phosphomonoesters [11
]. However, a small proportion of soil organic P compounds with potentially very different properties remain unidentified [11
]. For the identified compounds, their cycling, mobility and bioavailability generally are poorly understood [10
]. On the other hand, inorganic P exists in soils in various forms that are categorised in terms of accessibility (accessible to plants) and extractability (extractable in different reagents) in agricultural soils [13
]. Furthermore, mobility, stability and transformations between P compounds are affected by soil environmental conditions, e.g., soil acidity [14
] and land use [15
]. Soil P dynamics are also influenced by complex interactions between physical, chemical and biological processes that occur within the rhizosphere.
Given the complexity of P-cycling, a model that is able to simulate the major P-cycling processes and link with other nutrients and factors, would be a useful tool to quantify the impact of nutrients, soil moisture, field management practices and climatic conditions on crop growth and yields. There are several field-scale and process-based P-cycling models which are either independent ICECREAM [16
] or modules integrated into other nutrient cycling models, e.g., ANIMO [18
], APSIM [19
], EPIC [20
], CENTURY [21
] and its daily version DAYCENT [22
] and other unnamed models [24
]. Daroub et al. [26
] also developed a soil P module that operates with two comprehensive crop simulation models within the DSSAT (Decision Support System for Agrotechnology Transfer) software [27
]. Lewis and McGechan [28
] reviewed some of these models in terms of the concepts and constituent processes and concluded that a comprehensive description of all processes relevant to P in soil should consider transport of both soluble and particulate P, inorganic and organic P, as well as transformations from one form of P to another following applications of both mineral fertiliser and manure P. However, they also concluded that a mechanistic approach is too complex for incorporation into a systems model for the whole range of P processes.
] is a field scale model with a flexible soil layer definition (layer number and thickness of each defined layer), weather-driven, process-based and daily-time-step dynamic simulation on plant growth and development, soil carbon (C) and N cycling, with water movement and heat transfer. It has been used to investigate various issues, including nitrate leaching [31
], root systems [33
], greenhouse gas emissions [30
], soil C and N stock change [35
] and the responses of cropping/grassland systems to environmental changes [36
]. However, it would be impossible to accurately assess the impacts of general management practices (i.e., combined application of N and P fertilisers) on agricultural systems using the SPACSYS model without incorporating P-cycling. The objectives of this study were to describe a process-based P module added to the SPACSYS model and to evaluate its predictive capability on the dynamics of P content in crops and the impact of soil P status on crop growth via validation with a winter wheat field experiment conducted at Rothamsted Research, Harpenden, UK, over two consecutive growing seasons between 2012 and 2014.
Accurately estimating the dynamics of soil moisture is essential for quantifying crop growth and nutrient cycling correctly. Simulated soil water content in separate soil layers over the growing seasons were compared with discrete samplings (Figure 1
). Both simulated results and observed data show that soil moisture in the field of the Control treatment was higher than that of the P-added treatment, especially in the sub-soil layer (23–46 cm). The model performance statistical analysis demonstrate that the simulations followed the observed trend and reasonably matched the observed data (Table 1
). It was noted that there were abnormal observed values (over 60% in the topsoil layer and between 45–63% in the subsoil layer on 25 February 2014). The reason is that soil was too wet to be sieved. Removal of the point, statistical analysis showed significantly improved model performance (Table 1
Similarly to soil water content, both simulated and measured nitrate and ammonium content in the topsoil showed the same trend through the growing seasons. In general, the contents from the P added treatment were lower than those from the control, especially for nitrate (Figure 2
). The measured nitrate content in the Control treatment was much higher than that in the P-added treatment. However, the simulations did not show this. In general, the simulation of nitrate content for the P added treatment was over-estimated but it was under-estimated for the Control treatment.
Overall, dynamics of aboveground dry matter for both treatments were simulated reasonably well compared to the observed data (Table 2
). The SPACSYS model over-estimated aboveground dry matter accumulation for the Control treatment in the first growing season but slightly under-estimated it for the P added treatment in both seasons (Figure 3
). However, the relative errors in final grain yield were ± 15%, with the smallest being only 5.5% for the P added treatment in the 2013–2014 growing season. Furthermore, both simulations and observations showed a difference in dry matter accumulation, especially after GS35 (stem elongation), between the treatments, which suggests that P supply from the soils affected the growth rate. The model slightly over-estimated dry matter accumulation before GS35.
Simulated P accumulation in aboveground biomass generally followed the observed trend (Table 2
) although it showed large discrepancies during the later growing stages. The observed P content in the crop in the P added treatment showed an increase in P accumulation during the later growth stages in the first growing season, but the simulation showed a decrease in rate of accumulation (Figure 4
). For the Control treatment, the model over-estimated P uptake before anthesis and under-estimated it after. The phosphorus content in leaves was over-estimated, which might be caused by the defined partitioning coefficients of nutrients absorbed into various organs.
This study describes the testing of a P module for the SPACSYS model and demonstrates that the SPACSYS model was able to investigate the interactions between C, N, P and water in a single process-based model after the tested P module implemented. The advantage is to investigate not only the dynamics of a single element but also the interactions between them, and in turn, their effects on crop growth using a systems approach. The simulation results show reasonable agreement with observed data in terms of aboveground dry matter accumulation and P content, dynamics of soil water, soil inorganic forms of N and final grain yield under different soil P supply levels under varying climatic conditions. However, large errors between the simulated and observed data occurred when weather was abnormal, e.g., warmer, wetter and more sunshine hours over the 2013–2014 growing season. In the model, however, the negative impact of excess soil water (or waterlogging) on crop growth and partitioning was not considered. It was noted that soil ammonium content was over-estimated in the field of the Control treatment (Figure 2
B), which might imply that the soil mineralisation process and the preference of ammonium update by winter wheat were exaggerated under lower soil P concentration. The interaction of soil N and P dynamics under various environmental conditions should be further investigated.
A review on the effects of waterlogging on wheat growth reported that root growth and physiology are adversely affected by soil waterlogging [38
]. A field experiment conducted in the UK showed that winter waterlogging affected dry matter accumulation, shoot:root ratio and final yield of cv. Xi-19 [39
]. In our study, wheat plants were subjected to saturated conditions during the winter in both growing seasons (Figure 1
), which should reduce plant net photosynthesis. Furthermore, plant growth might be subject to the impacts of pests and diseases. However, the model does not consider these. As a result, the model may have over-estimated crop biomass accumulation during the period. During the later stages of the second growing season, the net photosynthesis rate may have been over-estimated because of the establishment of a larger canopy due to favourable weather conditions, which results in enhanced nutrient uptake, accumulated dry matter and P content (Figure 3
; Figure 4
). There are many processes involved in nutrient cycling further complicated by the interactions between the processes and intricate relationships between these and environmental factors. It is inevitable that the simulations generated from the model created some discrepancies in simulated data compared with observed data. In order to understand and quantify the interaction between nutrients and the impact of nutrient supply on crop growth and partitioning of photosynthate and absorbed nutrients, it might be necessary to design new experiments to monitor nutrient contents in crop and soils under different combinations of N and P supply.
The evidence showed that P plays a fundamental role in controlling resource allocation of plants in response to nutrient enrichment [40
]. The model captured the influence of P stress on wheat growth and P content, and final grain yield and P content in grains were reasonably simulated. The simulations suggest that the model estimated the dynamics of P uptake reasonably well (Table 2
), which indicates that soil P-cycling was correctly represented in the model. In the observed data, P content in grains rapidly increased at the early stages of grain filling, which is similar to reported observations in rice [41
]. Another independent experiment concluded that P deficiency affected harvest index and the root-shoot ratio of barley after anthesis [42
]. However, the rate of P remobilised from vegetative tissues to grains was set to a constant in our model. This gave rise to a discrepancy between observed and simulated P contents in grain. As this difference was generated within plants, it could be addressed in further investigations into partitioning of P within the plant and its subsequent remobilisation.