Root morphology and phosphorus requirements of 12 tropical pasture species grown in a controlled environment

Plant material

Twelve tropical pasture species were grown to investigate shoot yield and root morphology in response to soil P supply. There were five grasses, including Bambatsi Panic (Panicum coloratum L.), Floren Bluegrass (Dichanthium aristatum (Poir.)), Gatton Panic (Panicum maximum), Katambora Rhodes (Chloris gayana Kunth cv. Katambora) and Premier Digit (Digitaria eriantha Steud cv. Premier). There were seven legumes, including Bundey Centro (Centrosema pascuorum cv. Bundey), Butterfly Pea (Clitoria ternatea), Cardillo Centro (Centrosema molle cv. Cardillo), Common Stylo (Stylosanthes guianensis var. V8), JCU 7 Desmanthus (Desmanthus leptophyllus cv. JCU 7), JCU 9 Desmanthus (Desmanthus pernambucanus cv. JCU 9) and Shrubby Stylo (Stylosanthes scabra cv. Seca). These species were selected because they are commonly grown in the pastures of northern Australia.

Soil and nutrient treatments

A sandy soil (Grey Tenosol; Isbell 1996) was collected from the upper 2–15 cm soil layer of a field at Newholme SMART Farm, Armidale, New South Wales (NSW), Australia (30°26′21.4″S, 151°39′55.5″E). This soil was used because it allows easy removal of plant roots for analysis. The soil had a Colwell-extractable P concentration of 3 mg P kg⁻¹ (as measured by the method of Colwell (1963)), a phosphorus buffering index (PBI) of 29 (as measured by the modified method of Burkitt et al. (2008)), and a pH (1:5 w/v; 0.01 M CaCl₂) of ~4.7. The soil was passed through a 5 mm sieve and a basal nutrient solution was applied, and included 34 mg kg⁻¹ soil MgSO₄.7H₂O, 36 mg kg⁻¹ CaSO₄.2H₂O, 141 mg kg⁻¹ KNO₃, 23 mg kg⁻¹ (NH₄)₂SO₄, 14 mg kg⁻¹ NH₄NO₃, 99 μg kg⁻¹ H₃BO₃, 635 μg kg⁻¹ MnCl₂.4H₂O, 301 μg kg⁻¹ ZnSO₄.7H₂O, 28 μg kg⁻¹ CuSO₄.5H₂O, 60 μg kg⁻¹ (NH₄)₂MoO₄, 17 μg kg⁻¹ CoCl₂·6H₂O and 1283 μg kg⁻¹ FeNa-EDTA. Six P-amended soil treatments (0, 7.5, 15, 30, 60 and 120 mg P kg⁻¹ soil, hereafter referred to as P0, P7.5, P15, P30, P60 and P120 respectively) were prepared by adding KH₂PO₄ to the nutrient solution before it was applied to the soil. Potassium chloride was also added to the nutrient solution that was applied to the P0–P30 treatments, to balance the K so that it was equivalent to that of the P60 treatment. On the basis of previous experimentation, the P120 treatment was expected to allow maximum growth, whereas the P0–P60 treatments were predicted to be within the P-responsive range of the legumes (McLachlan et al. 2021). After the addition of all nutrients, the Colwell extractable P concentrations of the P0–P120 soils were 3, 7, 14, 25, 38 and 96 mg Colwell P kg⁻¹ soil. Cylindrical PVC pots (87 mm internal diameter; 200 mm height) were filled with 1.3 kg (oven-dry basis) of the P-amended soils. The total depth of soil was ~190 mm and the bulk density was ~1.15 g cm⁻³.

Plant growth conditions and experimental design

Micro-swards of each species were established by sowing seed (~5 mm depth) to achieve a population of six plants per pot. The seed of the two Desmanthus spp. was heat treated by submersing in 80°C water for 10 s, while the seed of Gatton Panic was mechanically scarified using sandpaper. These methods have been previously identified to satisfactorily break seed dormancy prior to sowing (Hopkinson and English 2004; McLachlan et al. 2021). Four replicate pots of each species in each P treatment were prepared. After planting, the pots were watered and moved to a glasshouse (natural daylight ~1800 μmol m⁻² s⁻¹ peak intensity; 30°C/25°C, day/night) in Armidale, NSW, Australia. Humidifiers were used to maintain the humidity within the glasshouse at ~50%. Pots were arranged in a randomised complete block design (blocks comprised the different replicates), which was generated using DiGGer (Coombes 2006). Soil moisture was maintained at 80–100% field capacity by watering daily to a predetermined weight. An additional 50 mg N kg⁻¹ soil as CH₄N₂O was applied to the surface of each pot 4 weeks after planting because of early signs of N deficiency in some of the grasses.

Harvest and measurements

Plants were harvested after 5 weeks of growth because a longer growing period would likely have restricted the root development and growth of the most productive species because of the small volume of soil in the pots. Shoots were cut at the soil surface, oven-dried at 75°C for 72 h and weighed. Shoot samples were finely cut before a ~50 mg subsample was pre-digested in a glass tube with 1 mL deionised water and 4 mL 70% (v/v) nitric acid for at least 4 h. Samples were then digested using a Milestone UltraWAVE 640 (Milestone Srl, Sorisole, Italy). The P concentration of the digested samples was determined colorimetrically at 630 nm by using the malachite green method (Irving and McLaughlin 1990). Shoot P content was calculated by multiplying shoot P concentration and shoot dry mass. Shoot samples were also assessed for N concentration by using a LECO Trumac Analyser. Shoot N content was calculated by multiplying shoot N concentration and shoot dry mass, while the N:P ratio of the shoots was calculated as the shoot N concentration divided by the shoot P concentration.

The soil from each pot was removed intact and cut longitudinally into the following three sections: (1) one-quarter for measurement of root hair length and mycorrhizal colonisation, (2) one-half for drying, and (3) one-quarter for root length scanning and drying, meaning that three-quarters of the soil core would be dried for determination of root dry mass. The roots were then washed from the soil over a 2 mm sieve. There was no evidence of any disease on the roots at the time of collection. Root samples for the measurement of root hair length and mycorrhizal colonisation were stored in 70% (v/v) ethanol at 4°C. Root samples for drying were oven-dried at 75°C for 72 h and weighed. Root samples for scanning were scanned using an Epson Perfection V700 Photo flatbed scanner (Seiko Epson Corporation, Suwa, Japan) at 600 dpi. When a root sample was too large for scanning, a representative subsample was scanned and the remaining roots were dried. Root length and average root diameter were determined using WinRHIZO™ software (Regent Instruments Inc., Quebec, Canada) (Bouma et al. 2000). Scanned root samples were then oven-dried at 75°C for 72 h and weighed. The estimated total mass of roots in the entire soil core was calculated as 4/3 multiplied by the combined dry mass of the unscanned half and scanned quarter. Total root mass fraction was calculated as the estimated total mass of roots divided by total plant mass (i.e. combined dry mass of shoots and roots). Total root length was calculated by multiplying the specific root length of roots from the scanned quarter by the estimated total mass of roots. Root length densities were calculated as root length per unit soil volume; the total soil volume was 1129 cm³.

Root hairs were measured on two lengths of root selected at random from each of the stored samples. Roots were imaged using a Nikon SMZ25 stereomicroscope fitted with a high-resolution Nikon DS-Ri2 digital camera (Nikon Corporation, Tokyo, Japan). Root hairs approximately perpendicular to the line of vision were randomly selected and measured using ImageJ 1.52 (FIJI) (Schindelin et al. 2012). The lengths of five root hairs per image (i.e. 10 root hairs per pot) were measured. Root hair coverage (i.e. the proportion of root length covered with root hairs) was estimated for roots in the P7.5 treatment, using the root sample that had been scanned at 600 dpi. The percentage of root length covered by root hairs was estimated using the gridline intersect method (Giovannetti and Mosse 1980) within ImageJ (FIJI) (Schindelin et al. 2012).

The surface area of the root hair cylinder (SARHC) was calculated as follows:

where ARD = average root diameter, RHL = root hair length, and RL = root length.

Root length colonisation by mycorrhizal fungi was measured on the plants grown in the P7.5 treatment. Roots were cleared in 10% (w/v) KOH for 50 min at 90°C, rinsed in water and 0.1 M (v/v) HCl, and stained using a 5% (v/v) Schaeffer black ink/white vinegar solution for 2 h (Vierheilig et al. 1998). Colonisation % was determined using the gridline intersect method, by examining the presence/absence of mycorrhizal fungi at 100 intersects on the stained root samples (Giovannetti and Mosse 1980). Root hair coverage and mycorrhizal colonisation were assessed only in the P7.5 treatment. It was expected that the soil P supply in this treatment would be below the critical external P requirement of all the species and that their reliance on P acquisition strategies would be highest, whereas the soil P supply in the P0 treatment was expected to cause severe P deficiency that could limit the expression of their P acquisition strategies.

Statistical analyses

Measured parameters were analysed in R (ver. 4.0.2; R Core Team 2020) by fitting linear models and using an ANOVA with ‘species’ and ‘P treatment’ as predictor variables. When appropriate, the effect of ‘replicate’ from the randomised complete block design was included in the most parsimonious model. When included, the effect of ‘replicate’ accounted for the error associated with spatial variation in the glasshouse. Shoot yield responses to P application were determined by fitting a self-starting Weibull growth function (y = a − b × exp (−exp (c) × x^d), where x is the P application rate and y is the shoot dry mass), as described by Crawley (2013). Critical external P requirements were calculated as the amount of P applied to achieve 90% of maximum yield on the basis of fitted Weibull growth functions, and the 95% confidence intervals of the critical P requirements were determined by bootstrapping residuals as described by Crawley (2013). Critical internal P requirements were calculated as the shoot P concentration that corresponded with the critical external P application rate (as defined by 90% relative shoot yield). The means and standard errors for root hair length, root hair coverage and mycorrhizal colonisation were calculated from the fitted linear models (R package emmeans, ver. 1.5.0, https://CRAN.R-project.org/package=emmeans; Lenth 2020) and means were compared using Tukey’s honest significant differences (HSD). The relative importance of root morphological traits for P uptake in the P7.5 treatment was determined using random forest modelling (R package randomForest, https://CRAN.R-project.org/doc/Rnews/; Liaw and Wiener 2002). The importance of each factor was estimated by the change in out-of-bag prediction error calculated when each factor was dropped from the model. The sensitivity of the proposed model to any one species was tested by sequentially removing individual species and assessing the scale and rank of importance factors. Importance factors were not influenced by species. Correlations between root length and shoot P content were analysed using linear regression (Crawley 2013). Normal quantile–quantile plots and Shapiro–Wilk tests were used to test the normality of the residuals for all fitted models. When required, these models were log-transformed to meet the assumptions of normality. A 5% level of significance was applied for all statistical tests.

Biomass production and critical external P requirements

The shoot dry mass of each pasture species increased in response to applied P, with the largest shoot yields being generally produced in the P30–P120 treatments (P < 0.001; Fig. 1). The exceptions were Shrubby Stylo and Floren Bluegrass, both of which had lower shoot yields when grown above the P30 and P60 treatments respectively. Average canopy height also increased in response to soil P supply (P < 0.001), and reflected the shoot yields of the pasture species (R² = 0.62, P < 0.001) (Supplementary material Fig. S1). In general, the grasses produced taller canopies than the legumes, with the relative difference being more pronounced at lower levels of soil P supply. Although increased soil P supply generally had a positive effect on shoot yield, there were substantial differences in how the species responded to P (P < 0.001; Fig. 1). In particular, maximum shoot yield ranged from 1.7 g DM pot⁻¹ for Shrubby Stylo to 9.8 g DM pot⁻¹ for Gatton Panic (5.7-fold difference; Table 1), and critical external P requirements ranged from 12.8 mg P kg⁻¹ soil for Katambora Rhodes to 38.0 mg P kg⁻¹ soil for Floren Bluegrass (3.0-fold difference; Table 1). On average, the shoot yields of the grasses were 1.6-fold higher than those of the legumes (P < 0.001), whereas the critical external P requirements of the grasses, except Floren Bluegrass, were equal to or lower than those of the legumes.

Fig. 1.

(a–f) The shoot dry mass of 12 tropical pasture species when grown in response to six rates of applied phosphorus (P) (0, 7.5, 15, 30, 60 and 120 mg P kg⁻¹ soil). Values show the mean ± s.e. (n = 4). The curves that were fitted to the shoot yield data show Weibull growth functions (Crawley 2013). ANOVA results were as follows: species P < 0.001, P treatment P < 0.001, species × P treatment interaction P < 0.001. The yield achieved by the Shrubby Stylo in the P60 and P120 treatments was significantly lower than in the P30 treatment. To enable the Weibull growth function to be fitted, the shoot yield achieved in the P60 and P120 treatments was assumed to be equivalent to that of the P30 treatment. Similarly, the yield achieved by Floren Bluegrass in the P120 treatment was significantly lower than in the P60 treatment. To enable the Weibull growth function to be fitted, the shoot yield achieved in the P120 treatment was assumed to be equivalent to that of the P60 treatment. The dashed lines represent the assumed P response curves.

Shoot P concentration and critical internal P requirements

The shoot P concentration of the pasture species increased in response to applied P, particularly between the P30 and P120 treatments, with the highest concentrations achieved in the P120 treatment (P < 0.001; Fig. 2). Across the entire P treatment range, there were significant differences in shoot P among the pasture species (3.1-fold range in the P0 treatment to 4.7-fold range in the P120 treatment) (P < 0.001). There were also differences in the critical internal P requirement of the species. The grasses had critical internal P requirements that ranged between 0.67 and 1.55 mg P g⁻¹ DM, whereas the requirements of the legumes ranged between 1.41 and 2.80 mg P g⁻¹ DM (Table 1). Shoot P content increased with soil P supply (P < 0.001) and varied significantly among the pasture species (P < 0.001) (data not shown because they reflect the shoot dry mass and shoot P concentration results).

Fig. 2.

(a–f) The shoot phosphorus (P) concentration of 12 tropical pasture species when grown in response to six rates of applied P (0, 7.5, 15, 30, 60 and 120 mg P kg⁻¹ soil). Values show the mean ± s.e. (n = 4). ANOVA results were as follows: species P < 0.001, P treatment P < 0.001, species × P treatment interaction P < 0.001.

Shoot N concentration and shoot N:P ratio

The shoot N concentration of the pasture species was highest in the P0 treatment and declined as soil P supply was increased (P < 0.001; Fig. S2). On average across the soil P supply treatments, shoot N concentrations were higher for the legumes (average 28 mg N g⁻¹ DM) compared to the grasses (average 16 mg N g⁻¹ DM) (P < 0.001). The N and P concentration data were used to calculate shoot N:P ratio. The N:P ratio of the shoots was inversely proportional to soil P supply (P < 0.001; Fig. 3). On average across the P treatments, the two Desmanthus spp. had the highest shoot N:P ratio, whereas Premier Digit and Katambora Rhodes had the lowest.

Fig. 3.

(a–f) The shoot nitrogen:phosphorus (P) ratio of 12 tropical pasture species when grown in response to six rates of applied P (0, 7.5, 15, 30, 60 and 120 mg P kg⁻¹ soil). Values show the mean ± s.e. (n = 4). ANOVA results were as follows: species P < 0.001, P treatment P < 0.001, species × P treatment interaction P < 0.001.

Root traits

Root mass fraction was lowest when the pasture species were grown with an adequate supply of P (i.e. at or above their critical external P requirement) that allowed maximum growth (i.e. the P30–P120 treatments) (Fig. 4). At lower rates of soil P supply, the proportion of plant biomass allocated to roots increased significantly (P < 0.001), so that most of the species allocated the largest proportion of plant biomass to roots when grown in the P0 treatment. This pattern of response was the same for the grasses and the legumes. On average across the P treatments, there was a substantial difference in the proportion of biomass allocated to roots by the different pasture species (P < 0.001). In general, the grasses achieved higher root mass fractions than did the legumes.

Fig. 4.

The root mass fraction of (a) five tropical grasses and (b) seven tropical legumes when grown in response to six rates of applied phosphorus (P) (0, 7.5, 15, 30, 60 and 120 mg P kg soil⁻¹). Values show the mean ± s.e. (n = 4). ANOVA results were as follows: species P < 0.001, P treatment P < 0.001, species × P treatment interaction P < 0.001.

Specific root length was influenced by both the level of applied P (P < 0.001) and pasture species (P < 0.001; Fig. 5). Most of the pasture species achieved the highest specific root lengths in the low-P treatments (i.e. P0–P7.5 treatments), particularly the grasses (e.g. Gatton Panic, Katambora Rhodes and Premier Digit). Specific root lengths of some of the legumes were relatively unresponsive to P treatment. For example, the two Stylosanthes spp. had relatively high specific root lengths across the range of P treatments, whereas that of Butterfly Pea was relatively low. Average root diameters varied among the species, ranging from 0.15 mm for Premier Digit to 0.46 mm for Butterfly Pea (P < 0.001; data not shown). On average across the species, average root diameter declined at lower levels of soil P supply (P < 0.001) and was negatively correlated with specific root length (R² = 0.25, P < 0.001).

Fig. 5.

The specific root length of (a) five tropical grasses and (b) seven tropical legumes when grown in response to six rates of applied phosphorus (P) (0, 7.5, 15, 30, 60 and 120 mg P kg soil⁻¹). Values show the mean ± s.e. (n = 4). ANOVA results were as follows: species P < 0.001, P treatment P < 0.001, species × P treatment interaction P < 0.001.

Root length was generally lowest under severe P constraint (in the P0–P7.5 treatments) and increased in response to soil P supply, with the longest roots generally being achieved in the P15–P120 treatments where P was less limiting (P < 0.001; Fig. 6). The development of root length varied significantly among the grasses and legumes (P < 0.001), because Gatton Panic, Katambora Rhodes and Premier Digit produced relatively long roots in the P7.5–P120 treatments, whereas all of the legumes produced relatively short roots in each of the P treatments.

Fig. 6.

The root length of (a) five tropical grasses and (b) seven tropical legumes when grown in response to six rates of applied phosphorus (P) (0, 7.5, 15, 30, 60 and 120 mg P kg soil⁻¹). Values show the mean ± s.e. (n = 4). ANOVA results were as follows: species P < 0.001, P treatment P < 0.001, species × P treatment interaction P < 0.001.

Root hair length did not change substantially across the soil P treatments (P = 0.043). Consequently, the root hair lengths of each of the species were averaged across the P treatments (Table 2). Average root hair length varied by 3.4-fold among the pasture species (P < 0.001). The longest root hairs were produced by the two Stylosanthes spp. and the shortest were produced by the two Desmanthus spp. Root hair coverage ranged between 3% and 55% (P < 0.001) and mycorrhizal colonisation ranged between 3% and 51% (P < 0.001) among the pasture species (Table 2). There was a slight negative correlation between mycorrhizal fungi colonisation and root hair length (R² = 0.12, P = 0.011), whereas there was no correlation between mycorrhizal fungi colonisation and either root hair coverage or average root diameter (P > 0.05) (data not shown).

Phosphorus acquisition

The relative importance of the different root traits for determining shoot P content was analysed using random forest modelling when the 12 pasture species were grown in the P7.5 treatment (Fig. 7). The modelling indicated that surface area of the root hair cylinder, root length and root dry mass were the three most important traits that influenced the P uptake of the grasses (Fig. 7a), whereas root dry mass, root length and mycorrhizal fungi colonisation were the three most important traits that influenced the P uptake of the legumes (Fig. 7b). Because root length had the highest shared relative importance for both the grasses and legumes, it was correlated with shoot P content at each level of soil P supply (Fig. 8). In the P0–P30 treatments, in which the tropical pasture species would have been foraging for P, the R² coefficients of determination were strong for the grasses (R² = 0.70–0.79; Fig. 8a) and generally moderate for the legumes (R² = 0.14–0.67; Fig. 8b). There was an increase in the slope of the correlations with increasing soil P supply, which was more pronounced for the legumes, particularly in the P60–P120 treatments. This increase could be attributed to more P uptake per unit root length, yet the R² coefficients of determination also indicated that there were significant differences in shoot P content per unit root length among both groups of species (i.e. the five grasses and the seven legumes; Fig. 8). Indeed, shoot P content per unit root length varied significantly (P < 0.001) among the pasture species (data not shown).

Fig. 7.

The relative importance of root morphological traits for phosphorus (P) acquisition for (a) five tropical grasses and (b) seven tropical legumes when grown with 7.5 mg P kg⁻¹ applied to a low-P soil. The relative importance of each trait was determined using random forest modelling, by estimating the change in out-of-bag prediction error calculated when each factor was dropped from the model. This indicated the traits that were more closely linked among the species of grass and legume. The sensitivity of the proposed model to any one species was tested by sequentially removing individual species and assessing the scale and rank of importance factors. Importance factors were not influenced by species.

Fig. 8.

The relationship between root length and shoot phosphorus (P) content for (a) five tropical grasses and (b) seven tropical legumes that were grown in response to six rates of applied P (0, 7.5, 15, 30, 60 and 120 mg P kg soil⁻¹). The regression lines and R² coefficients of determination were fitted separately for each P treatment and include all data points of each species.

Shoot growth and critical external P requirements

The tropical pasture grasses were generally more productive and, with the exception of Floren Bluegrass, had comparable or lower critical external P requirements than did the tropical pasture legumes examined in this study. In addition, the grasses had taller canopies than did the legumes, particularly at lower levels of soil P supply. These differences in height and productivity suggest that, if the legumes were grown in mixed pasture swards with the grasses, the legume component may be less competitive, which could lead to a grass-dominant pasture. Indeed, maintaining the balance between grasses and legumes often requires some form of ongoing management even in low-input systems (e.g. maintaining adequate soil fertility and limiting selective grazing) (Dear and Virgona 1996). Because the soils of northern Australia are generally nutrient-deficient and receive little nutrient input, plus grazing management is limited because of extensive paddock areas, these potential differences in competitiveness between grasses and legumes are likely to be more pronounced.

Differences in critical external P requirements among the tropical pasture species indicate that there is potential to use P-efficient selections in inherently low-P soil. For example, Katambora Rhodes had the lowest critical external P requirement and was the most P-efficient species; therefore, it may be more suited to relatively infertile soils compared to Floren Bluegrass, which had a relatively high critical external P requirement. This difference in P requirement is supported by known differences in preferred soil type, as Katambora Rhodes is suited to light sandy soils, whereas Floren Bluegrass is suited to heavier clay soils (Boschma et al. 2010). Differences in critical P requirements also allow species with similar requirements to be identified and paired for use in mixed-pasture swards. For example, Premier Digit and Bambatsi Panic are likely to be more suitable for pairing with tropical pasture legumes such as Desmanthus spp., because their critical P requirements are more similar and therefore the relative difference in competition for soil P is less likely to negatively affect the productivity and persistence of the legume component. Premier Digit and Bambatsi Panic are commonly-grown species on medium- and heavier-textured soils respectively (Boschma and McCormick 2008); so, the ability to pair these grasses with tropical pasture legumes would be advantageous. By comparison, Katambora Rhodes with its low critical external P requirement is likely to be highly competitive under low to moderate P conditions and outcompete other species. Indeed, Rhodes grass is known to establish quickly and be highly competitive, and will even outcompete other tropical grasses such as Bambatsi Panic and Premier Digit (Lodge et al. 2009).

Previous research has demonstrated that several Stylosanthes spp. can persist in very low-P soils (Jones 1974; Smith et al. 1990). In the present study, Shrubby Stylo had the lowest critical external P requirement of the legumes, but it was not substantially lower than that of the other legume species. This may be a result of the sandy soil that was used in the present experiment. Indeed, factors such as texture, pH and PBI are likely to influence the expression of root traits that facilitate P acquisition and improve plant productivity, and previous work has demonstrated that pH and PBI influence the critical P requirements of Premier Digit and JCU 9 Desmanthus (McLachlan et al. 2024a). Alternatively, the similarity between the Stylosanthes spp. and the other legumes may be because our assessment was conducted on plants during the early establishment phase and it is possible that Stylosanthes spp. either take longer to mature, or implement other P acquisition strategies later in their maturity. Nevertheless, their early lifecycle root traits indicated that they should have been more efficient at acquiring P. For example, their specific root length and root hair length were both relatively high compared with the other legumes. These are important traits for developing a large surface area for P acquisition (Richardson et al. 2009), so their relative P-acquisition efficiency may have improved with time. Indeed, the sensitivity of Shrubby Stylo to the P60 and P120 treatments would suggest that it has a low requirement for P.

Shoot P and N concentrations

Critical internal P requirements were relatively low across the tropical pasture species, which suggests that they may require less available P than temperate pasture species to be productive. Though this is well known for Stylosanthes spp., our findings now include a more diverse range of tropical grasses and legumes.  Low critical internal P requirements indicate that the tropical pasture species are suited to relatively P-deficient soils, the trade-off is that low tissue P concentrations can be below the levels required for grazing livestock (Dixon et al. 2020). The critical internal P requirements could be used to inform the need for, and management of, fertiliser application in tropical pasture swards. This is because previous work has suggested that the critical internal P requirements of tropical pasture species (specifically Premier Digit and JCU 9 Desmanthus) are relatively stable across a range of soil pH and PBI conditions (McLachlan et al. 2024a). We also observed significant differences in shoot N:P ratio of the species, with the legumes having a higher N:P ratio, even though they had not been inoculated with Rhizobia. This result highlights the importance of the legume component for improving the quality of C₄ grass-dominant pasture swards.

Root traits

Random forest modelling was used to identify the most important traits for P acquisition among the 12 tropical pasture species. The species were divided into grasses and legumes because it was expected that these pasture components would differ in their dependence on the various root traits assessed in our study. The most important trait for P acquisition among the tropical pasture grasses was the surface area of the root hair cylinder. This trait combines average root diameter, root length and root hair length to determine the total surface area developed by the plants for P acquisition, as described by Haling et al. (2016b). In contrast, the most important trait for the tropical pasture legumes was root dry mass, which suggests that traits such as average root diameter and root hair length are not as important for P acquisition. These contrasting results suggest that tropical pasture legumes may rely on other factors such as mycorrhizal fungi or organic exudates when acquiring P from relatively low-P soil. Indeed, mycorrhizal fungi colonisation was the third-most important trait for the legumes, whereas root exudates were not considered in the present work due to the complexity of analysis. Previous work has shown that organic acid exudates are important for other pasture species (Kidd et al. 2016, 2018); so, it is plausible that organic exudates were also important for the tropical legumes examined in our study. Nevertheless, for both pasture components, the second-most important trait was root length. Previous work by McLachlan et al. (2021) identified root length as the primary determinant of P acquisition among nine Desmanthus spp. genotypes. This result suggests that, although there may be differences in root morphology between grasses and legumes, and even among different grass and legume species, the key driver of P acquisition is the development of long roots that effectively forage the soil for P.

The importance of root length for P acquisition is well documented (Lynch 1995; Lynch and Brown 2001), and similarly we observed moderate to strong correlations between root length and shoot P content across most of the P treatments when the species were separated into grasses and legumes. Yet, there was some variation in the relationship between root length and shoot P content, which may be due to a species reliance on other traits that influence P acquisition, or because of differences in root architecture through the development of root length that influences the effectiveness of root placement for P acquisition. Specifically, there are two main ways that root length is developed. Root proliferation is a localised response whereby root branching leads to lateral root production that maximises root surface area within a relatively small area (e.g. in response to a patch of P). In contrast, root elongation occurs when roots explore the soil profile for more mobile nutrients and water, thus generally increasing the extent of the root system (Hodge 2004, 2006; Lynch and Wojciechowski 2015). These different, yet complimentary, strategies are important for the productivity and persistence of pasture species because of the varying location of nutrients and water, particularly in nutrient-deficient soils that are often dry, and to the generally patchiness of relatively immobile nutrients such as P. The importance of root length for P acquisition has the following two implications:

Root hair coverage varied significantly among the tropical pasture species, and the proportion of root length covered with root hairs was generally relatively low. Similar observations have been made among both temperate (McLachlan et al. 2019) and tropical (Crush 1974; McLachlan et al. 2021) pasture legumes. Because root hair coverage was assessed only at the end of the experiment, it is not known whether root hairs were produced but later lost (turned over), or whether the production of root hairs was generally patchy. Indeed, among some of the legumes, such as the two Desmanthus spp. and two Stylosanthes spp., the proportion of root length covered in root hairs was <13%. This is likely to have significantly reduced their root surface area for P acquisition. Because root hairs are known to improve P acquisition at a relatively limited metabolic cost (Richardson et al. 2011), they might be a reasonable target for improving P-acquisition efficiency.

Practical implications

There were considerable differences in critical external P requirements among the tropical pasture species, which is something that could be exploited for improved pasture productivity and persistence in P-limited environments. It may be possible to select for species that have relatively low critical P requirements (e.g. Rhodes grass) and thereby improve the productivity of low-input systems, but it is also suggested that the relative efficiency of the different pasture components should be considered because if there is a large difference between grasses and legumes then the grass is likely to dominate in mixed swards over time. Of course, consideration of a single trait such as P efficiency may result in a one-dimensional view and the observed differences in shoot yield and canopy height could likewise lead to the legume component being overwhelmed by the large amount of biomass produced by the highly productive C₄ grasses.

Differences in critical external P requirements primarily occurred between the grasses and legumes, whereas there were fewer differences among the legumes. This contrasts with some of our previous work that showed larger differences in critical external P requirement among nine Desmanthus spp. genotypes (McLachlan et al. 2021), including JCU 7 and JCU 9 Desmanthus, which were grown in the present experiment. Further work is therefore required to validate differences in P efficiency among tropical pasture species, particularly for plants grown under field conditions that are influenced by other factors such as moisture stress and grazing pressure.

Differences in P efficiency were associated, to varying degrees, with the development of root length and related traits. Future improvements in P acquisition efficiency could be achieved by selecting and breeding for plants with longer roots that forage the soil for available nutrients. Indeed, the efficient development of root length is likely to be important in northern Australia where soils are nutrient deficient and seasonally dry, meaning that surface-foraging roots for nutrient acquisition and deeper roots for water acquisition could be important in driving productivity and persistence.