Response of wheat to phosphorus-enriched ironstone gravel

Soils and properties

An IG soil (0–30 cm) was collected from West Dale (−32.327233 S 116.509133 E), around 90 km south east of Perth, WA (Weaver et al. 2022). Under the Western Australian soil groups classification (Schoknecht and Pathan 2013) it is described as a very gravelly deep sandy soil, and is colloquially known as Yalanbee soil. It is typical of 45% of the 2.7 million ha of the gravelly soils used for crop and pasture production in SWWA. The soil is described as a sesqui-nodular tenosol (Isbell and National Committee on Soil and Terrain 2021) using the Australian soil classification and is classified as a Endopetric Pisoplinthic Plinthosol (Arenic) (IUSS Working Group WRB 2015) under the World Reference Base. The soil had not been previously fertilised nor used for agricultural production and was collected from a cleared area amongst remnant vegetation [Eucalyptus accedens (powderbark wandoo) with an understory of Banksia sessilis (parrotbush)].

Soil preparation

The IG soil (soil A; Table 1) was sieved using <2, 2–4, 4–6, 6–8 and 8–10 mm fractions in preparation for use in experiments. The >2 mm fractions of the IG soil were gently washed with water and air dried prior to use. In addition, a white, low P retention, nutrient deficient washed sand (soil B; Table 1) (https://www.richgro.com.au/product/soft-washed-play-sand/) was used as the <2 mm fraction or supporting soil matrix (sand culture) for combination with >2 mm ironstone gravels in pot experiments. The <2 mm fractions were analysed for Colwell P (Colwell 1963), Colwell K (Rayment and Lyons 2011), P buffering index (PBI; Burkitt et al. 2002), pH (CaCl₂) (Rayment and Lyons 2011), KCl-40S (Blair et al. 1991), particle size (Indorante et al. 1990) and specific surface area (SSA) using N adsorption (Brunauer et al. 1938) and are shown in Table 1. Ammonium oxalate extractable Fe and Al (Tamm 1922) of the <2 mm fraction were estimated using an established pedotransfer function for Fe (Weaver and Summers 2021) and the following pedotransfer function for Al (log ammonium oxalate Al = 0.799 × logPBI + 148; n = 3484; R² = 0.54). Other properties of the <2 and >2 mm fractions of the IG soil (soil A) are described in Weaver et al. (2022).

Phosphorus adsorption and desorption

Adsorption and desorption of P was examined for each size fraction of the IG soil A (<2, 2–4, 4–6, 6–8 and 8–10 mm). The P adsorption isotherms were developed at a 10:1 solution:soil ratio by incubating each size fraction in triplicate, and quiescently, with occasional gentle mixing to avoid abrasion of gravel surfaces (Weaver et al. 2022). Initial P concentrations of 0, 5, 10, 20, 40, 100 and 400 mg P L⁻¹ were used for the <2 mm fraction, and 0, 5, 10, 20, 40 and 100 mg P L⁻¹ for the >2 mm fractions. At a 10:1 solution:soil ratio this resulted in P application of 50, 100, 300, 400, 1000 and 4000 mg P kg⁻¹ for initial P concentrations of 5, 10, 20, 40, 100 and 400 mg P L⁻¹, respectively. The P solutions were prepared using potassium dihydrogen orthophosphate (KH₂PO₄), and final concentrations were determined after 1, 2, 5, 8 and 16 days of adsorption. Following the adsorption phase, each sample was air dried at ambient temperature and subjected to five sequential extractions of 1 day duration with de-ionised (DI) water at a 10:1 ratio. All solutions were filtered at <0.45 μm prior to analysis for filterable reactive P (Murphy and Riley 1962). Adsorption isotherms were constructed by plotting P adsorbed vs equilibrium P concentration and fitting the Freundlich equation. Desorption isotherms were constructed by plotting P that remained adsorbed (by difference) vs equilibrium P concentration and fitting the Freundlich equation.

Volumetric water content

To assist with interpretation of the results from pot experiments, volumetric water content (VWC) measurements were determined on oven dried (105°C) mixes of soil B (Table 1) and IG 2–4, 4–6, 6–8 and 8–10 mm in size. These VWC measurements were determined at drained upper limit (DUL) or field capacity (Cassel and Nielsen 1986). The VWC was determined for gravel contents of 0, 25, 50, 75 and 100% by volume following cessation of drainage under gravity (Govindasamy et al. 2023), an approach shown to be comparable to the pressure plate method (Cresswell et al. 2008). Estimated gravimetric water contents were then adjusted to VWC using bulk density of each of the mixes (Moorberg and Crouse 2017).

Pot experiments

A series of five pot experiments were undertaken to test the stated hypotheses. This involved the preparation of P-enriched IG (IG_+P) of different size ranges and mixing IG_+P or IG without P enrichment (IG_−P) in different proportions with a <2 mm fraction (Table 1). Following enrichment with P, the IG was air dried prior to mixing with the <2 mm fraction. Dependent on the experiment, the <2 mm fraction may also have been enriched with P, but separately from the IG. All incubations of <2  and >2 mm fractions with P were undertaken using KH₂PO₄, and pre- and post-incubation P concentrations were measured (Weaver et al. 2022). The test crop used in all pot experiments was wheat (var. Cobalt). This variety was selected for its consistent growth and high yield. More contemporary varieties are favoured for their relevance to current agricultural conditions and yield potential (Anderson et al. 2015).

Data analysis and curation

The available data were presented as box and violin plots (McGill et al. 1978; Hintze and Nelson 1998) and scatterplots, stratified and symbolised by IG amount, size, weighted Colwell P and total P added to experimental pots. Weighted Colwell P was calculated from separate Colwell P of gravel and matrix fractions and the mass of these fractions within a pot. In addition, linear, non-linear regression, analysis of variance (ANOVA), analysis of covariance (ANCOVA) and Spearman rank correlation were undertaken using the exploratory data analysis and presentation tools in RStudio (rstudio.com), DataDesk 8.3 (datadesk.com) and Igor Pro v9 (wavemetrics.com). The threshold level of significance was set at P < 0.05.

The responsiveness of wheat to applied P, either added as IG_+P for different gravel size fractions or directly to the <2 mm fraction, was examined using the Mitscherlich equation (Eqn 1):

where y is the yield of DM (g pot⁻¹), X is the amount of P applied (mg P pot⁻¹), a is maximum DM (g pot⁻¹), b quantifies the yield response (a−Y0a, where Y₀ is DM when X = 0; b ranges within 0–1) and c describes the curvature of the response. As c increases, the response curve moves to the left, becomes steeper and less P is required to produce the same relative yield. Hence smaller values of c are regarded as less responsive than larger values of c. The DM responses were re-expressed as RY by dividing y by a(RY=ya×100), refitting Eqn 1 and setting a = 100 and b = 1 which resolves to Eqn 2.

Parallel curves analysis using accumulated ANOVA in Genstat V22.1 from VSN International Ltd was used to determine whether equation parameters for different treatments were justified (Powers 2021). Treatment differences explored included whether more than or less than 50% of the weighted Colwell P or total P added to a pot was derived from gravel or matrix, or whether treatments included 25% or 50% gravel.

Phosphorus adsorption and desorption

The amount of P adsorbed increased in the size order 8–10, 6–8, 4–6, 2–4 and <2 mm. The amount of P adsorbed increased with increasing contact time for all size fractions at all incubating P concentrations (Fig. 1a–e). The amount of P adsorbed increased at a faster rate over time as the incubating P concentration increased.

Fig. 1.

Net P adsorption and desorption dynamics for the (a) <2, (b) 2–4, (c) 4–6, (d) 6–8 and (e) 8–10 mm fractions of soil A. Closed symbols show adsorption phase for 1, 2, 5, 8 and 16 days at 5, 10, 20, 40, 100 and 400 mg P L⁻¹ initial P concentrations. Open symbols show desorption of five sequential 24 h extractions at the end of the adsorption phase. Thick dashed grey line and grey shading show fitted adsorption curve (Freundlich) after 16 days with 95% confidence band. Thin solid black lines connect treatments with the same initial P concentration through increasing time for the adsorption phase, followed by sequential extracts through the desorption phase (thin dashed black line).

More P was adsorbed than desorbed. The <2 mm fraction adsorbed 29–1050 mg P kg⁻¹ after 16 days and desorbed 5–172 mg P kg⁻¹ after five sequential extractions depending on the amount of P applied. Correspondingly, the 2–4 mm fraction adsorbed 40–210 mg P kg⁻¹ and desorbed 2–36 mg P kg⁻¹, the 4–6 mm fraction adsorbed 40–120 mg P kg⁻¹ and desorbed 2–25 mg P kg⁻¹, the 6–8 mm fraction adsorbed 36–77 mg P kg⁻¹ and desorbed 2.5–21 mg P kg⁻¹, and the 8–10 mm fraction adsorbed 31–49 mg P kg⁻¹ and desorbed 2–18 mg P kg⁻¹.

There was a significant interaction between size, incubate P concentration and extraction on the amount of P desorbed. Most P was desorbed in the first extraction for all size fractions, and the amount of P desorbed decreased with increasing size. The amount of P desorbed increased with increasing incubate P concentration in the following size order: 8–10, 6–8, 4–6, 2–4 and <2 mm.

The P adsorption decreased with increasing particle size (Fig. 2a), with most P adsorbed by the <2 mm fraction and the least by the 8–10 mm fraction of soil A. The percentage of applied P retained after 16 days decreased with increasing P applied and ranged within 5–80% (Fig. 2b). The 2–4, 4–6, 6–8 and 8–10 mm fractions retained more P (60–80%) than the <2 mm fraction (60%) at the lowest rate (50 mg P kg⁻¹) of P application. The rate of decline in the percentage of P adsorbed with P applied was greater for the >2 mm fractions than the <2 mm fraction. At 1000 mg P kg⁻¹ applied, the >2 mm fractions retained 5–20% of the applied P, whilst the <2 mm fraction retained about 35%. The percentage of adsorbed P that was subsequently desorbed remained constant (~20%) with P applied for the <2 mm fraction (Fig. 2c). In contrast, the percentage of adsorbed P that was subsequently desorbed increased with increasing P applied and with increasing size for the >2 mm fractions, and ranged within 5–35%. At 1000 mg P kg⁻¹ applied, the 4–6, 6–8 and 8–10 mm fractions desorbed a greater percentage of the retained P than the <2 mm fraction, whilst for below 400 mg P kg⁻¹ applied the >2 mm fractions desorbed a smaller percentage (5–20%) of the retained P than the <2 mm fraction.

Fig. 2.

(a) Freundlich isotherms fitted to 16-day P adsorption data for the <2, 2–4, 4–6, 6–8 and 8–10 mm fractions. Fitted Freundlich curves (Padsorbed = A × Equilibrium P concentration^B) had the following A and B coefficients and R² values: A = 18.06, B = 0.72, R² = 0.96; A = 29.08, B = 0.45, R² = 0.99; A = 25.13, B = 0.35, R² = 0.99; A = 19.85, B = 0.29, R² = 0.98; and A = 19.37, B = 0.20, R² = 0.93 for the <2, 2–4, 4–6, 6–8 and 8–10 mm fractions, respectively. (b) Percentage of applied P adsorbed after 16 days for the <2, 2–4, 4–6, 6–8 and 8–10 mm fractions for different amounts of P applied. (c) Percentage of the P retained after 16 days that was desorbed after five extractions with DI water for the <2, 2–4, 4–6, 6–8 and 8–10 mm fractions for different amounts of P applied.

Volumetric water content

When soil B was used as the fine fraction, VWC decreased linearly with increasing gravel percentage. The ANCOVA showed that the effect of gravel size was not significant (P > 0.05), and there was no significant difference in intercept among the groups. Therefore, a single linear equation described VWC as a function of gravel percentage (Fig. 3). The VWC decreased from around 40% (equivalent to 65 mm rainfall in a 4-L pot) when no gravel was present, to <5% (equivalent to <10 mm rainfall) when the sample was 100% gravel.

Fig. 3.

Volumetric water content as a function of gravel percentage. Moisture (%) = −0.343 × Gravel % + 39.18; r² = 0.99. Gravel size increases with increasing symbol size.

Pot experiments

Experiment 1

The treatment effect of IG_+P had a 30-fold increase in g DM pot⁻¹ (11.1) over IG_−P (0.37). There was no significant treatment effect of gravel size on DM (P = 0.47). However, there was a significant effect (P < 0.001) of gravel amount (Fig. 4a), with 75% IG_+P producing lower amounts of DM than both 25% and 50% IG_+P. For DM, there was significant interaction (P < 0.001) between gravel amount and gravel treatment (IG_+P, IG_−P), but not between gravel size and gravel treatment (P = 0.31). Soil solution P concentrations of the IG_+P treatments were much greater (mean of 0.65 mg P L⁻¹) than the IG_−P treatments (<0.01 mg P L⁻¹) (Fig. 4b). Soil solution P concentrations decreased significantly over time (P < 0.001) and with IG_+P amount (P < 0.001), but no significant effect of gravel size was identified (P = 0.85). Measured P concentrations were roughly in proportion to the mass of IG_+P present.

Fig. 4.

(a) Boxplots of dry matter (g pot⁻¹) for 2–4, 4–6 and 6–8 mm IG_+P and IG_−P (control) treatments at 25%, 50% and 75% (v/v). (b) Boxplots of the weekly change in soil solution P concentration (mg L⁻¹) for for 2–4, 4–6 and 6–8 mm IG_+P and IG_−P (control) treatments at 25%, 50% and 75% (v/v). The LSD is based on a four way interaction between treatment, size, amount and sampling time.

Experiments 2 and 3

The combined RY from the first harvest of Experiments 2 and 3 is shown in Fig. 5a. Wheat grown on 100% sand (soil B; Table 1) was the most responsive to P application, whilst wheat grown on 100% <2 mm fraction (soil A; Table 1) was the least responsive to P application. Between soils B and A, wheat response to P application systematically decreased with decreasing gravel size. The c coefficients of the fitted curves were 0.0257, 0.0153, 0.0113, 0.0091, 0.0085 and 0.0053 for the sand and 8–10, 6–8, 4–6, 2–4 and <2 mm fractions, respectively. For IG of the same size, greater levels of RY and DM were achieved when either IG_+P was present in greater amounts relative to the <2 mm fraction, or the IG_+P was enriched with a higher P concentration (Fig. 5a, b). Wheat DM was greater in the first harvest than in subsequent harvests. Some treatments from the first harvest, particularly IG_+P either with smaller size (2–4 and 4–6 mm) and a higher level of P enrichment (100 mg P L⁻¹), or higher gravel amount (50%) of any size with higher levels of P enrichment (40 and 100 mg P L⁻¹) could support as much DM as soil B with P applied to maximise wheat growth (Fig. 5b). The mean DM pot⁻¹ of harvests 2 and 3 fell to 37% and 33%, respectively, of harvest 1. The P content of harvested plant material fell with successive harvests (Fig. 5c). For harvest 1 of experiment 2, plant tissue P content increased significantly with increasing P enrichment (P < 0.0001), gravel amount (P < 0.0001) and decreasing gravel size (P < 0.0001). From harvest 1, the highest tissue P contents were found in sand treatments with P applied to maximise growth, followed by treatments with 50% IG_+P enriched to the highest level and with smaller sizes of IG (2–4 and 4–6 mm), and then treatments with 25% IG_+P, with the lowest P content for wheat grown in soil B with no P added. By harvest 3, the P contents of wheat grown on all IG_+P treatments were similarly low and below the whole-shoots critical range estimated from Reuter and Robinson (1997) for all treatments except where soil B had P applied to achieve maximum growth. For IG treatments, the amount of P removed in harvested biomass decreased with consecutive harvests for similar levels of P in the pot at the start of a growth cycle (Fig. 5d), as indicated by the decreasing slope of the relationship for consecutive harvests. This reduction occurred despite the relatively small removal of P in harvested biomass (tens of milligrams) compared to the quantity of P available at the start of each growth cycle (hundreds of milligrams). The amount of P removed in harvest for IG treatments was also smaller than for wheat grown in soil B. The slope of the relationship between P removed in harvested biomass and P in the pot at the start of the growth cycle (Fig. 5d) was highest when P was added in a soluble form to soil B, than when P was added in the form of IG_+P.

Fig. 5.

(a) Relationship between the RY (%) of wheat to the added mg P pot⁻¹ either as IG_+P or directly for the <2 mm fraction of the IG soil (; c = 0.0053), and for 2–4 mm (●; c = 0.0085), 4–6 mm (●; c = 0.0091), 6–8 mm (; c = 0.0113), 8–10 mm (; c = 0.0153) gravels, and for sand (□; c = 0.0257). The c coefficient from curvefit (Eqn 2) in parenthesis following symbol. Data compiled from the first harvest of Experiments 1, 2 and 3. (b) Violin plots of DM from three sequential harvests (Experiment 2) compared to DM of wheat grown in sand without (□) and with P (■) to maximise growth. (c) Violin plots of DM P content from three sequential harvests (Experiment 2) compared to DM of wheat grown in soil B without (□) and with P () to maximise growth. Grey shaded area shows typical whole shoot critical tissue concentrations for wheat from Reuter and Robinson (1997). (d) P removed in harvested material as a function of the P available at the start of each growth cycles (closed coloured symbols) for three sequential harvests (harvest 1: , harvest 2: ■, harvest 3: ▲) compared to soil B (open symbols) without (origin) and with P to maximise growth. Linear fit through combined soil B data for three harvests also shown. Symbol size shows gravel size, and colour and colour intensity represent gravel % (v/v) and concentration of P enriching solution (mg P L⁻¹), respectively.

Experiment 4

The DM and soil solution P concentrations from Experiment 4 are shown in Fig. 6. Most DM was produced in soil B when P was applied, and the least was produced in soil B when P was not applied. The IG_+P treatments with the largest quantity of P per pot and greatest surface concentration of P also showed high DM production (Fig. 6a). However, within each treatment and size fraction, the range of DM values remained relatively narrow. The ANOVA for the IG_+P treatments separated into different IG sizes showed significant increases in DM as the amount of P per pot increased and concentration of P on the surface of IG increased (Fig. 6a). Further, for a fixed level of P per pot and IG_+P surface P concentration, DM amount varied little and showed small but significant differences.

Fig. 6.

(a) Boxplots of dry matter (g pot⁻¹) for soil B with and without P, IG_+P of various sizes incubated with 2 mg P L⁻¹ (green), 20 mg P L⁻¹ (orange) and 100 mg P L⁻¹ (red) at different solution:soil ratios to achieve similar surface P concentrations. Quantity of P added with differing amounts of IG_+P or as KH₂PO₄ for each treatment following incubation shown in mg P pot⁻¹. (b) Boxplots of log soil solution P for soil B with and without P, IG_+P of various sizes incubated with 2 mg P L⁻¹ (green), 20 mg P L⁻¹ (orange) and 100 mg P L⁻¹ (red) at different solution:soil ratios to achieve similar surface P concentrations. Week of sampling shown by symbol size. Treatments with the same letter are not significantly different and increase alphabetically.

Like DM, soil solution P concentrations were highest for soil B with P applied, and lowest for soil B with no P applied (Fig. 6b). The highest soil solution P concentrations for IG_+P treatments occurred with the largest quantity of P per pot and highest surface concentration of P, and lowest soil solution P concentrations occurred for the lowest quantity of P per pot and surface concentration of P. For a fixed level of P per pot and IG_+P surface P concentration, soil solution P concentrations varied little and showed some small but significant differences. The ANOVA showed no significant effect of IG size (P = 0.97), but there was a significant effect due to the quantity of P per pot and surface P concentration (P < 0.0001). Soil solution P concentrations decreased with increasing time (P < 0.0001) (Fig. 6b).

Experiment 5

The DM for harvest 1, harvest 2 and grain followed Mitscherlich responses when expressed as a function of weighted Colwell P (Fig. 7a–c) and total P added per pot (Fig. 7d–f). Much more DM was observed at harvest 2 and grain yield compared to harvest 1. While typical Mitscherlich responses were followed, differentiation by the proportion of weighted Colwell P or total P added per pot showed systematic increases in the amounts of DM when a greater proportion of the P was supplied by the soil matrix (<2 mm) than by the gravel (>2 mm). Separation of the data into groups where less than or more than 50% of the P derived from the gravel showed significant increases in DM when <50% of the P was derived from gravel (>50% of the P was derived from the soil matrix). Relationships of RY as a function of median soil solution P for harvest 1 (Fig. 7g), harvest 2 (Fig. 7h) and grain yield (Fig. 7i) followed a Mitscherlich response; however, greater RY was achieved for 25% than for 50% gravel treatments. Similarly, relationships of RY as a function of weighted Colwell P for harvest 1 (Fig. 7j), harvest 2 (Fig. 7k) and grain yield (Fig. 7l) followed a Mitscherlich response. However, greater RY was achieved at 25% than at 50% gravel only for harvest 1, whilst the fitted curves could not be separated for harvest 2 or grain and were not significantly different (P > 0.05). Relationships of RY as a function of P content measured in harvested biomass for harvest 1 followed a Mitscherlich response (Fig. 7m). For the same P content, greater RY was achieved for 25% than 50% gravel.

Fig. 7.

Dry matter (g pot⁻¹) as a function of weighted Colwell P (mg kg⁻¹) for (a) harvest 1, (b) harvest 2 and (c) grain; dry matter (g pot⁻¹) as a function of total P added (mg pot⁻¹) for (d) harvest 1, (e) harvest 2 and (f) grain. Relative yield as a function of median soil solution P (mg L⁻¹) for (g) harvest 1, (h) harvest 2 and (i) grain. Relative yield as a function of weighted Colwell P (mg kg⁻¹) for (j) harvest 1, (k) harvest 2 and (l) grain. Relative yield as a function of P content (%) for (m) harvest 1. For a–f point colour represents the proportion of weighted Colwell P (a–c) or total P added (d–f) derived from gravel or soil matrix. Dashed black line and grey shading show the fitted curve and 95% confidence band for the full dataset. For a–f red solid line shows the fitted curve for points where the proportion of weighted Colwell P (a–c) or total P (d–f) added from gravel is <50% (matrix > 50%), and the blue solid line shows the fitted curve for points where the proportion of weighted Colwell P (a–c) or total P (d–f) added from gravel is >50% (matrix < 50%). For g–i (median soil solution P) and j–l (weighted Colwell P) and m (P content) red points and line from pots with 25% gravel and blue points and line for pots with 50% gravel. Top left corner of each panel indicates whether curve parameters (a, b or c from Eqn 1) are significantly different (●) or not significantly different (○) between groups (red and blue lines).

Weighted Colwell P was linearly related with total P added (Fig. 8a), and the slope of these relationships decreased as the percentage of gravel increased. Soil solution P collected in rhizons during the first 46 days decreased weekly, and much higher concentrations were found in treatments with higher weighted Colwell P (Fig. 8b). The highest mean concentrations were in treatments where both IG and the soil matrix were enriched to achieve 95% RY, and the lowest observed mean concentrations where neither IG nor the soil matrix had any P enrichment. Mean soil solution P concentrations were higher when the soil matrix was enriched to achieve 95% RY and IG 0% RY than other RY combinations of IG and soil matrix, such as 50% IG 95% soil matrix, 95% IG 50% soil matrix, 95% IG 0% soil matrix and 50% IG 50% soil matrix (Fig. 8b). Median soil solution P from rhizons was linearly related with weighted Colwell P, and the slope of this relationship was greater for treatments with 50% gravel than for 0% or 25% gravel (Fig. 8c). Median soil solution P from rhizons was linearly related with total P added; however, this relationship was not differentiated by gravel percentage (Fig. 8d).

Fig. 8.

(a) Linear relationships between weighted Colwell P (mg kg⁻¹) and total P added (mg pot⁻¹) for treatments with 0% (black triangles), 25% (orange circles) or 50% (red squares) gravel by weight per pot. Size fraction increases with increasing symbol size. Fitted lines based on significant terms from ANCOVA. (b) Changes in soil solution P (mg L⁻¹) over sampling week. Gaussian noise added in x direction to reveal overlapping data points. Colour scale shows weighted Colwell P (mg kg⁻¹) of treatment at commencement. Triangles show treatments with 0% gravel, circles with 25% gravel and squares with 50% gravel. Size fraction increases with increasing symbol size. Overlays of cubic splines fitted to mean values for treatments where IG and soil matrix had been incubated with P prior to being combined in a pot to achieve 95% RY from IG and 95% RY from matrix (solid black line), 0% RY from IG and 95% RY from matrix (dotted black line), 50% RY from IG and 95% RY from matrix (black dash dot dot line), 95% RY from IG and 50% RY from matrix (dark grey dash dot line), 95% RY from IG and 0% RY from matrix (dark grey dashed line), 50% RY from IG and 50% RY from matrix (light grey short dash dotted line), and 0% RY from IG and 0% RY from matrix (light grey long dashed line). (c) Log median soil solution P as a function of weighted Colwell P (mg kg⁻¹) of treatment at commencement. Black triangles show treatments with 0% gravel, orange circles with 25% gravel and red squares with 50% gravel. Size fraction increases with increasing symbol size. Fitted lines based on significant terms from ANCOVA. (d) Log median soil solution P as a function of total P added (mg pot⁻¹) of treatment at commencement. Black triangles show treatments with 0% gravel, orange circles with 25% gravel and red squares with 50% gravel. Size fraction increases with increasing symbol size. Fitted lines based on significant terms from ANCOVA.

Estimated VWC varied over time during the first 46 days of plant growth according to plant growth stage, watering regime, percentage gravel and weighted Colwell P of each treatment (Fig. 9). Small-scale diurnal changes in VWC were observed in all treatments due to daily automated applications of water and evapotranspiration. Large-scale changes in VWC were observed due to larger manual applications of water when plant observations and extractions of soil moisture from rhizons indicated that additional moisture was required. For a given gravel percentage, VWC decreased as weighted Colwell P increased. Pots containing both high gravel content and high weighted Colwell P levels exhibited the lowest VWC and the most rapid declines in VWC between manual watering events.

Fig. 9.

Estimated volumetric water content (%) as a function of time for pots with differing amounts of gravel (0%, 25% and 50%) and differing levels of weighted Colwell P.

Phosphorus retention and release

Our findings demonstrate that the trends in P adsorption and desorption of the IG fractions (>2 mm) and fine earth fraction (<2 mm) are similar but vary in degree (Figs 1, 2). The adsorption capacity of IG was lower due to their small surface area (Weaver et al. 2022), while the desorption behaviour exhibited similar trends to the fine earth fraction but tended to be lower for IG than fine earth fraction. In addition, the percentage of added P that was retained (Fig. 2b), and the percentage of retained P that was desorbed (Fig. 2c) were similar to values reported by Clarendon et al. (2019). The results from this work support hypothesis 1, confirming that the larger size and lower surface area of IG reduces their P adsorption capacity and potential for P desorption.

This behaviour aligns with previous research indicating that, dependent on mineralogy, larger soil particles (such as gravel) have a reduced capacity for P adsorption due to their lower surface area-to-volume ratio (Clarendon et al. 2019; Weaver et al. 2022). This characteristic makes IG less efficient in retaining applied P (Fig. 2b), leading to lower overall P retention in soils where IG is the dominant soil fraction. In addition, IG enriched with high P concentrations showed a greater percentage of retained P desorbed. This indicated that the IG surface of larger fractions with lower surface area was likely saturated with P that could be more rapidly desorbed. Despite lower overall P retention, IG can be sufficiently enriched to provide a significant proportion of the Colwell P and DM from a soil (Figs 7, 8a). In support of hypothesis 3, wheat can take up P from the <2 or >2 mm fractions. However, reduced RY when IG contents are 50% compared to 25% (Fig. 7) suggests some negative impacts due to the presence of more IG. While sufficient P may be supplied, other factors such as reduced VWC may override the growth potential that P fertility from IG may provide.

Soil solution concentrations

The presence of IG_+P influenced soil solution P concentrations in the following ways. As the amount of similarly enriched IG_+P increased, the P concentration in the soil solution increased and was maintained at higher levels for longer (Fig. 4b), likely due to the greater surface area enriched with P exposed to soil water. These increases in soil solution P led to commensurate increases in DM, supporting hypothesis 2 that wheat growth can be supported when P-enriched IG is the only source of P, and that wheat biomass will increase with increases in P enrichment of IG or the amount of P-enriched gravel present. While soil solution P concentrations are supported to some degree by IG_+P, the soil matrix seems to support higher soil solution P concentrations than IG. This is evident in Fig. 8b where much higher soil solution P concentrations were maintained when the whole soil comprised soil matrix (<2 mm fraction of soil A) enriched to 95% RY and IG not enriched compared to soil matrix (<2 mm fraction of soil A) not enriched and IG enriched to 95% RY. This is consistent with the P adsorption and desorption behaviour of the different size fractions shown in Fig. 1, and similar to those reported by Clarendon et al. (2019). The soil matrix was able to retain and release more P than IG, and hence whole soils containing P-enriched fine fractions are more likely to support higher soil solution P concentrations than P-enriched IG. However, even when higher soil solution P concentrations can be achieved, DM can decrease with high gravel amounts (Fig. 7g–i). Presumably this occurs because of lower soil moisture, and because higher soil solution P concentrations will encourage more biomass and evapotranspiration, depleting soil moisture even more rapidly. The IG_+P enriched with higher initial P concentrations led to higher soil solution P concentrations (Fig. 6b), but these concentrations tended to be lower than when a soluble P source was applied directly to soil B. This points to differences in short- and long-term P retention and release characteristics of IG compared to soil B and the form in which P was added to pots. In the case of IG_+P added to a pot, P was pre-adsorbed onto the IG and must then be released into soil solution via desorption processes (Fig. 1) whereas soluble P was added directly to soil B as a control. Soil B had a lower PBI than IG, and was less influenced by adsorption–desorption processes, hence P was more likely to move to plant roots via mass flow. In contrast, when IG_+P is the P source, P is likely to move to plant roots via diffusion (Bhat and Nye 1973) as its release is controlled by desorption processes.

In addition, the reduction in the slope of the positive relationship between the log of median soil solution P and weighted Colwell P as gravel content decreased (Fig. 8c) suggests that whole-soil PBI is influenced by gravel content. Weaver et al. (2023) demonstrated that the slope of linear relationships between calcium chloride extractable P (Rayment and Lyons 2011), the benchmark measure of dissolved reactive P, and Colwell P decreased as soil PBI increased. The PBI of the >2 mm fractions examined ranged within 5–10 (Weaver et al. 2022) compared to the <2 mm fraction of soil A which had a PBI of 45.4 (Table 1). Increasing soil gravel content would reduce whole-soil PBI, leading to higher soil solution concentrations for the same weighted Colwell P in comparison to a soil with a low gravel content and consistent with the findings of Weaver et al. (2023).

Impact on wheat growth

Wheat growth experiments revealed that higher gravel content generally resulted in lower absolute plant biomass, despite providing sufficient P fertility. This reduction in biomass is primarily attributed to low plant available water (Figs 3, 6, 7 and 9). In support of hypothesis 3, gravel size influences P responsiveness of wheat (Fig. 5a) in a similar way that soil PBI (Burkitt et al. 2002) influences responsiveness of pastures to P (Gourley et al. 2019; Weaver et al. 2024). Soils with low PBI have response curves with large curvature coefficients (c value), and soils with high PBI have small c values. This effect is evident in Fig. 5a where soil B has a low PBI, a low surface area and is coarse textured (Table 1). In comparison, soil A has a much higher PBI and comprises finer textured materials. Ranging between these extremes are the mixtures of soil B and IG fractions in which the c value systematically increases as gravel size increases. Both surface area and mineralogy are key factors influencing soil PBI, and PBI of IG increases as gravel size decreases (Weaver et al. 2022). Hence the amount of gravel, and the size distributions of gravel in a soil, along with PBI of each size fraction will influence the overall P retention of the whole soil and its responsiveness to P applications.

The decline in DM and P content of plant tissue in sequential harvests for Experiment 2 (Fig. 5b, c) is indicative of removal of P in plant tissue along with P that may have been leached from the pot. In addition, the decline in slope of the relationship between P removed in harvest and the amount of P at the start of the growth cycle for IG treatments compared to the much greater slope for the same relationship for soil B where P was supplied in a soluble form, points to both differences in short- and longer-term P retention and release characteristics of soil materials in different treatments, and the form in which P was added. For example, soil B has a PBI of 2.5 (Table 1) compared to PBI values of 4.7–10.6 for intact IG dependent on size (Weaver et al. 2022). These slightly higher PBI values will lead to ongoing adsorption and penetration of P into the minerals on the gravel surface (Barrow 2021). The incubation process to prepare IG_+P materials and time for plant growth for each harvest would contribute to ongoing adsorption and penetration of P. This is evident in Fig. 5d through the reduction in slope in subsequent harvests. In comparison, when soluble P is added to soil B with a lower PBI, adsorption and penetration of P is unlikely because surface area is lower, and soil B does not contain minerals that contribute to these processes. Hence soil B does not provide competition for P resources for plant growth, whilst the addition of IG_+P to soil B in sufficient quantity provides P in an adsorbed form subject to ongoing slow reactions (Barrow 1974; Barrow and Shaw 1975) that competes with plant growth for P resources.

The lower DM and RY in the presence of 50% compared to 25% gravel (Experiment 5) (Fig. 7) is indicative of reduced soil moisture, leading to lower biomass for the same level of weighted Colwell P or total P added. This is reinforced by the relationship between RY and wheat P content for harvest 1 of Experiment 5 (Fig. 7m). Reduced RY when 50% gravel is present compared to 25% gravel for the same P content in the plant biomass is suggestive of concentration effects due to differences in biomass when 25% or 50% gravel is present. Previous research indicates that drought-induced biomass reduction can lead to elevated concentrations of P in plant tissues, even when the total amount of P absorbed by the plant decreases. Reduced biomass can lead to a higher concentration of P within the remaining tissue (Liu et al. 2021; Jiang et al. 2023). These drought-induced effects may help to explain some of the variability in relationships in Fig. 7, and why the relationship between RY and P content (Fig. 7m) could be differentiated by gravel content using parallel curves analysis.

Reduced volumetric water content

High gravel content significantly affects the soil’s physical properties, particularly its ability to retain water (Figs 3, 9). Gravelly soils typically have larger pore spaces, which leads to increased drainage and reduced water-holding capacity. This reduced VWC can limit the availability of water to plant roots, especially during critical growth stages. It is well established that plant available water capacity (PAWC) reflects soil texture variations and subsequent constraints on water and nutrient uptake (He et al. 2021). Even though sufficient P may be present, the lack of adequate soil moisture can impede the plant’s ability to absorb and utilise nutrients effectively, leading to reduced biomass. The pot experiments show that despite sufficient P supply from IG, plant biomass and RY were diminished when higher gravel contents were present. For example, in Experiment 1 the least biomass resulted from 75% IG_+P which generated the highest soil solution concentrations compared to 25% and 50% IG_+P (Fig. 4). In Experiment 5 (Fig. 7), RY was lower in the presence of 50% compared to 25% gravel, and when more than 50% of the Colwell P or total P added was derived from the gravel fraction for the same weighted Colwell P, or total P added (Fig. 8). Hence, despite sufficient P available in a pot measured either as weighted Colwell P, total P added or median soil solution P, RY was lower for higher gravel amounts or when IG was supplying most P. To supply most P, gravel must either have higher P contents than the matrix or contribute most of the mass. Adsorption experiments (Fig. 1) showed that IG fractions retain less P than the matrix, hence it is more likely for gravel to supply most P when gravel amounts are high, which will lead to lower PAWC (Figs 3, 9) and reduced plant biomass.

The amount of IG seems to act as a control on VWC (Figs 3, 9) and subsequent plant growth even when P supply from IG and/or the fine earth fraction maintains high soil solution P concentrations (Fig. 4) or soil Colwell P (Fig. 7).

Reduced soil moisture retention in gravelly soils poses a challenge for crop growth. Water is a crucial medium for desorption, diffusion and hence nutrient transport to plant roots, and its scarcity can limit the plant’s capacity to take up essential nutrients, including P. The decreased water availability in gravel-rich soils may lead to water stress, further inhibiting plant growth and productivity (He and Dijkstra 2014; Suriyagoda et al. 2014). In addition, the fluctuating water availability can create an environment where nutrients are either leached away during heavy rainfall or become unavailable during dry periods, complicating nutrient management strategies.

Practical implication on agriculture

The study underscores the importance of considering soil IG content and its influence on P fertility and transient VWC over and above water limited yield potential (French and Schultz 1984). Depending on the nature of the soil matrix, soils with higher gravel content may have lower soil P retention than is estimated from soil tests on the soil matrix alone. Since soils with lower P retention are more P responsive due to lower c values, and lower P application rates are likely required to optimise crop yield.

The lack of influence of gravel content on Mitscherlich responses of RY as a function of weighted Colwell P (Fig. 7k and l) suggests that whole-soil Colwell P is a more suitable measure than soil solution P to estimate RY of wheat across a range of soil moisture conditions, whether that be influenced by climatic conditions and/or gravel content.

The impact of reduced VWC on wheat growth highlights the need for strategies that account for the increased drainage in gravel-rich soils. Techniques such as mulching or incorporating organic matter could be explored to improve soil moisture retention in these conditions. Mechanical separation of the gravel into the upper soil profile may act as a mulch and protect evaporation of the separated underlying matrix. Organic matter can enhance soil structure, and may lead to an increase in water-holding capacity, although this is uncertain (Minasny and McBratney 2018). Mulching can help reduce water evaporation from the soil surface, maintaining higher soil moisture levels and mitigating the adverse effects of high gravel content.

Future research should explore the potential benefits of mulching gravelly soils to improve soil moisture retention so that nutrient supply from IG can be optimised. Long-term field trials could provide valuable insights into the practical applications of these strategies and their impact on crop yield and sustainability. Additionally, investigating the interactions between gravel content, soil moisture and nutrient dynamics can help develop more effective soil management practices tailored to specific soil compositions and crop requirements.

The experiments conducted here included a <2 mm fraction with PBI values classified as extremely low to medium P sorption capacity (Gourley et al. 2019). Choosing a soil with a <2 mm fraction with a higher PBI than was used here may exaggerate some findings and moderate others. For example, a <2 mm fraction with higher PBI would be less P responsive and have lower c values, requiring greater additions of P to achieve RY targets. Given the stronger P retention of the <2 mm fraction of a higher PBI soil, whole-soil PBI would likely be dominated by the P sorption capacity of the <2 mm fraction despite dilution of whole-soil PBI by addition of gravel with low PBI characteristics, reducing soil solution concentrations. Despite this, increased P fertilisation should allow critical Colwell P levels to be achieved without a deficit in RY targets. Ultimately, the PBI of the soil’s constituent parts, the amounts of each and the P fertility of each fraction will determine the potential response of wheat to the characteristics of the whole soil. But these factors may be inconsequential if the whole-soil characteristics result in yield limitation by a soil moisture deficit. Furthermore, these experiments used soil fractions with very low P fertility that were artificially enriched with P, achieving Colwell P levels up to 30 ppm for the <2 mm fraction, probably slightly higher than critical level for a soil with PBI of 45 (Bell et al. 2013). Choosing soil fractions that have been enriched with P under standard agricultural practices, separating these into different size fractions and recombining in different ways may provide additional understanding of the effects of different soil P fertility levels, PBI, soil organic carbon and influences of non-wetting. This may lead to improved understanding of P dynamics in relation to gravel content and assist to refine fertilization practices for sustained P availability and optimal plant growth.

Further effort is required to find practical solutions to apply these findings to soil sampling strategies and commercial laboratory settings where the >2 mm fraction is routinely discarded, yet the wider industry is reliant on soil test results from laboratories to make evidence-based fertiliser decisions. Foremost, however, simpler and routine methodologies are required to assess and understand how plant available water capacity may limit growth even when soil fertility is adequate (Foley 2017).

It may be necessary to reconsider soil sampling approaches that consider the collection of samples that are representative of whole soils, notwithstanding the spatial variability of soil gravel content at a range of scales (Weaver et al. 1992; Holmes et al. 2021).

After the collection of representative samples, laboratory processes need to be considered. Previous research has shown that for some measures such as PBI, grinding of soil greatly increases P retention (Clarendon et al. 2019; Weaver et al. 2022) as does end-over-end tumbling of gravels, but may be a suitable approach for measures like Colwell P. Introduction of additional preparation (splitting, grinding of some size fractions and not others) and analytical steps is unlikely to be favoured by commercial laboratories, hence other approaches, including adaptation of current, or new testing procedures need consideration. This may include upscaling (greater sample mass and extractant volumes) of diffuse gradient thin film technology (Mason et al. 2013, 2010) to cater for increased variability and heterogeneity introduced by whole-soil samples containing gravel (Buchter et al. 1994).

This study highlights the agronomic and analytical importance of accounting for gravel content in soil assessments. For farmers, recognising the influence of gravel amount, size and P enrichment on P availability and soil moisture retention can support more targeted fertilisation and water management strategies, ultimately improving wheat yield outcomes. For policymakers, including those with oversight of guiding laboratory analysis, soil test interpretation and decision support systems, the findings underscore the need to revisit soil sampling and testing protocols to ensure they reflect whole-soil properties. This includes gravel fractions typically excluded from routine analysis. Consideration also needs to be given to temporally variable characteristics such as rainfall, which along with soil PBI and pH have been shown to be of critical importance influencing soil moisture, P availability and RY (Scanlan et al. 2024). Incorporating these insights can enhance the reliability of soil test interpretations and inform evidence-based nutrient management, ensuring that both agronomic advice and policy guidelines are better tailored to the realities of gravelly agricultural soils.