Changes in pasture and soil properties with liming and superphosphate application on five soils in the Central Tablelands of New South Wales over 12 years

Site details

Experimental sites were located on the Central Tablelands of NSW, Australia on soils that represent a range of the more acidic types in the area. The five sites were: (1) Mt. David; (2) Vittoria; (3) Browns Creek; (4) Neville; and (5) Black Springs (Fig. 1). Site locations, soil classifications (Stace et al. 1968; Northcote 1979; Isbell 1996), and selected soil properties are in Table 1. The experiment was conducted in two phases to accommodate changes in pasture composition and to improved pasture sampling methods towards the end of the period. Site meteorological data was inferred using spatial interpolation for each site with records dating from 1930 using Datadrill (see Fig. 2; Jeffrey et al. 2001).

Fig. 1.

Locations of the five experimental sites (Mt David, Vittoria, Brown’s Creek, Neville, and Black Springs) on the Central Tablelands of New South Wales, Australia.

Fig. 2.

Total annual rainfall, and average annual rainfall received at each site over the experimental period. Values were derived from datadrill (Jeffrey et al. 2001) using records dating from 1930.

Phase I (1978–1986)

Sites were ploughed (15 cm depth) in December 1978 and lime (NV = 97; 95% <150 μm) at five rates (0, 0.63, 1.25, 2.5, and 5.0 t/ha) was incorporated by rotary-hoe to a depth of 12.5–15.0 cm. Superfine lime (95% <150 μm) was locally sourced from a plant at Bathurst, NSW. Two experiments were sown at each site in April 1979 with seed of species with moderate tolerance (Experiment 1) and high sensitivity to acid soils (Experiment 2), respectively. Species sown and their seed weights (kg/ha) in Experiment 1 were: red clover (Trifolium pratense) cv. Redquin (3.4); white clover (Trifolium repens) cv. Haifa (1.1); subterranean clover (Trifolium subterraneum) cv. Woogenellup (4.5); perennial ryegrass (Lolium perenne) cv. Victorian (5.0); and cocksfoot (Dactylis glomerata) cv. Currie (1.1). The acid-sensitive species sown in Experiment 2 included lucerne (Medicago sativa) cv. WL318 (6.7 kg/ha) and phalaris (Phalaris aquatica) cv. Sirosa (4.5 kg/ha). Each experiment was a randomised block design and replicated twice. Plot size was 5 × 3 m.

All plots received a basal dressing of copper (Cu), manganese (Mn) and zinc (Zn) as sulfates plus borax (all four at 8 kg/ha each), flowers of sulfur (S; 33 kg/ha) and magnesium (Mg) sulfate (125 kg/ha) prior to seeding with a tyne-seeder in April 1979. Molybdenum (Mo) single superphosphate (8.8% P, 11% S, 19% Ca, 0.025% Mo) was also applied at 250 kg/ha with the seed. In spring each year potassium (K) chloride (250 kg/ha) was broadcast on all plots, and single superphosphate (SSP, 8.8% P, 11% S, 19% Ca) was applied annually at 0, 125, and 250 kg/ha. All legume seed was appropriately inoculated and lime pelleted. The experiments were disrupted by the drought of 1980–1982, causing plots to be reseeded and trace elements reapplied.

Measurements

Soil

Soil samples (0–7.5 cm depth) were collected in November in the years 1979–1986 from every plot immediately before reapplication of the SSP treatments. Each sample consisted of a composite of 35 cores, diameter 2.25 cm. The samples were dried at 40°C and passed through a 2-mm sieve prior to analysis. Exchangeable cations (0.0125 M BaCl₂ leach) (Vimpany et al. 1985), pH (0.01 M CaCl_2; pH_Ca) (Rayment and Lyons 2011; Method 4B1), P-sorption (Bache and Williams 1971; Rayment and Lyons 2011; Method 9I1) and organic carbon (Leco total C) (Rayment and Lyons 2011; Method 6B2b). Specific detail of the procedures are described by Vimpany et al. (1985). Effective cation exchange capacity (ECEC) was calculated as the sum of cations measured (Ca²⁺, Mg²⁺, K⁺, Na⁺, Al³⁺), and Ca (% Ca) and Al (%Al) saturation were calculated as proportions of the ECEC. Phosphorus (Bray-1) was also measured on all plots in all years (Bray and Kurtz 1945; Rayment and Lyons 2011; Method 9E).

Pasture

Pasture harvests were conducted whenever treatment effects were visually apparent and are presented as at the sampling time, irrespective of the intervening period since the previous sampling (see Fig. 3 for actual sampling times at each site). Plant material from the central part of the plot (1 × 2 m) was cut to 5 cm height with a cutter bar mower and weighed in the field on portable scales. A sub-sample (~500 g) was collected from each to calculate plot dry matter yield (DM). The remaining pasture was cut and all the material removed from the plots after harvesting, as plots were not grazed.

Fig. 3.

Biomass harvests (solid dot) and level of significance (1 = P < 0.05, 2 = P < 0.01, 3 = P < 0.001) of lime, P, and lime × P interaction in E1 and E2 at each site. The lower figure represents actual monthly rainfall (bars) and long-term monthly average rainfall (line) for each site.

Phase II (1987–1990)

Phase II commenced in 1987 and utilised the sites originally sown in 1979 (Phase I). Immediately after soil cores were taken in spring each year, SSP was applied at the same three rates. Experiment 1 (E1) continued unaltered but Experiment 2 (E2) was extensively modified by re-seeding the pasture. To ensure similar botanical composition across treatments during Phase II, the perennial species T. subterraneum cv. Woogenellup (susceptible to clover scorch) and Vulpia spp. were removed by herbicide (glyphosate, 360 g/L) in spring 1987 and autumn 1988. All plots were reseeded with T. subterraneum cv. Karridale (10 kg/ha) resulting in a legume dominant pasture that was more sensitive to pH in comparison to the grass dominant pasture in E1. Later, germinating seedlings of Vulpia spp. were controlled by an application of carbetamex herbicide (carbetamide, 700 g/kg). After each harvest, there was a heavy grazing of both experiments (50–200 sheep/ha) for periods of 4–7 day, depending on seasonal conditions.

Statistical analysis

Because of the change in soil sampling depths between Phases I and II, different methods of analysis were used for the data. Analysis of soil pH_Ca detected no significant difference between the two sampling depths (0–7.5 cm, 0–10 cm) for soil collected in 1988. Regression analysis was then used to determine time trends in pH_Ca from 1979 to 1989, where 1979 was assigned time = 0. There were no co-incident measurements at the two sampling depths for the other measured soil parameters, and these parameters were analysed by individual ANOVA for each sampling period, with the change in levels of significance between lime and SSP rates recorded. Long-term DM responses were analysed by spline regression analysis, while individual DM harvests and composition data were analysed by ANOVA (GenStat 2000). Trends and differences presented are at the 5% level of significance unless otherwise indicated. Botanical composition was analysed by ANOVA.

Meteorological data

Long-term (1930–1989) average annual rainfall for each site (Fig. 2) was greatest at Neville (938 mm) and lowest at Vittoria (796 mm) with Browns Creek (903 mm), Mt David (845 mm) and Black Springs (831 mm) between these extremes. The higher long-term average rainfall at Neville and Browns Creek compared with the other sites was due to historically higher rainfall in the May–October period (data not shown). Total annual rainfall for each site during the experimental period was variable with well above average rainfall in 1978 and 1990; above average in 1981, 1983 and 1984 (except Mt David in 1981 and 1983); below average over the period 1985–1989 (except Mt David in 1988; Vittoria in 1988–1989; and Black Springs in 1986, 1988); and well below average in 1979, 1980 and 1982 (Fig. 3), a period of widespread drought across much of eastern Australia (e.g. Archer and Rochester 1982).

Monthly rainfall across sites followed the same trends over the course of the experiment (Fig. 3). Generally, rainfall was well below average in 1980, preceding a wet winter in 1981 followed by little rainfall until the start of 1983. Monthly rainfall in 1983 was close to average, concluding with above average rainfall in November and December. The years 1984 and 1986 were characterised by low summer/autumn rains, while in 1984 and 1985, there were above average winter rains. Spring rains were above average in 1985 and 1986. From 1987, monthly rainfall was generally close to the average, with only the occasional month where rainfall was below average. Monthly temperature maxima and minima for each site followed similar trends with Browns Creek being generally warmer and Black Springs cooler (data not shown).

Site differences in soil characteristics

Soils at Mt David, Vittoria, and Browns Creek were similar under the following classifications: Northcote (1979) Gn 2.34, Stace et al. (1968) yellow earth and Isbell (1996) yellow kandosol (Table 1), with Neville (Gn 2.14, red earth, red kandosol) and Black Springs (Uf 6.12, kraznozem, red dermosol) being distinctly different. Soil texture (0–15 cm) was similar for all sites (silty clay) except Black Springs (light clay). Initial pH_Ca values were similar across all sites and ranged from 4.15 (Browns Creek) to 4.38 (Vittoria). Initial ECEC values were highest at Vittoria (5.16) and Black Springs (4.51), and lowest at Browns Creek (2.34), while %Al was lowest at Vittoria (10.2%) and highest at Black Springs (31%). The ECEC was dominated by Ca at all sites. with %Ca from 47.6% (Mt David) to 64.8% (Vittoria). P sorption and organic C values were greatest at Black Springs (506 mg/kg, 4.32% C) and lowest at Browns Creek (166 mg/kg, 2.30% C).

Effect of lime on key soil parameters

Lime rate and soil pH

In 1979, 1 year after lime application, soil pH_Ca was increased (Fig. 4a) by an across site average of 0.3 units/t of lime applied (pH_Ca = 4.29 + 0.30 × lime (t/ha); R² = 0.88), but this response had declined to 0.1 pH_Ca units by 1989 (pH_Ca = 4.04 + 0.10 × lime (t/ha); R² = 0.55; P < 0.001). Responses in E1 and E2 at each site were similar over time.

Fig. 4.

Long term effects of (a) lime and (b) superphosphate on soil pH over the experimental period (1979–1989). Data are averaged over all experiments and sites. Lime (L) rates: 0 = nil; 1 = 0.63 t/ha; 2 = 1.25 t/ha; 3 = 2.5 t/ha; 4 = 5.0 t/ha. Superphosphate (P) rates: 0 = nil; 1 = 125 kg/ha; 2 = 250 kg/ha.

Lime rate and %Al

For all sites, application of 1.25 t lime/ha halved Al saturation, and the 2.5 t/ha rate decreased Al saturation to less than one-fifth of its initial values. Lime at the highest rate (5 t/ha) decreased Al saturation effectively to zero at 2 years after application (Fig. 5a).

Fig. 5.

Effect of lime rate on the decrease on (a) Al saturation (see Fig. 4a for legend) and (b) the relationship between %Al saturation and pH_Ca in E1 for the five experimental sites 2 years after application of lime (1980).

For the duration of the experiment, significant (P < 0.001) differences in Al saturation occurred between the lime rates, except in 1979 and 1980 where there was no significant difference between 2.5 and 5.0 t/ha lime, and in 1989 between 0.63 and 1.25 t/ha lime. In both E1 and E2, the early results showed no differences in Al saturation between 2.5 and 5.0 t/ha lime. There was no significant difference between nil and 0.63 t/ha lime in E1 in 1987. In E2, there were no significant differences in 1984 and 1987 (0.63 and 1.25 t/ha), 1988 (nil and 0.63 t/ha), and 1989 (nil, 0.63, and 1.25 t/ha). Black Springs was the most responsive site over time, with differences (P < 0.001) in %Al occurring in all years except for between the two highest rates in 1979. There were no differences between nil and 0.63 t/ha rates in E1 at Mt David in any year. For the other sites, this comparison was only significant until 1984.

Effect of superphosphate

Plant responses to lime and superphosphate application

Individual harvests

A total of 87 harvests was taken on the cocksfoot pastures (E1) and 68 on the subterranean clover pastures (E2) from the five sites. Significant DM responses to lime were more evident in E2 than in E1 (Fig. 3). Mt David and Neville were the most lime responsive sites, while Vittoria was the least. In E1, the proportion of responses was similar in the early and late phases of the experiment. However, there was a proportionally greater number of positive responses during the early phase of the experiment in E2 (Fig. 3). Harvests were taken when differences in growth were visible. Consequently, harvests tend to overlap with periods of favourable rainfall, especially during spring. The absence of harvests in 1982–1983 for both experiments was due to a protracted drought, and in E2 in 1988, when the pasture was modified and reseeded to subterranean clover. Superphosphate application generally increased DM for all harvests (Fig. 3), except at the start of the experiment. Analysis using untransformed data indicated that significant lime × SSP interactions occurred in four instances in E1 and in 14 instances in E2. However, data were skewed, and a log transformation of the data reduced the number of significant interactions from 18 to 10. Vittoria was again the least responsive site.

Grouping of significant responses on the basis of non-significant harvest × lime × SSP interactions resulted in the formation of three groups with similar responses to lime and SSP (Fig. 6). In Group 1, there was a positive DM response to lime at the higher levels of SSP, and a positive DM response to SSP at each lime rate. Most of the SSP response occurred between P0 and P1. Most of these responses occurred at Mt David (n = 4) but also at Neville and Black Springs, and most responses were in E2. Group 2 was represented by E2 at Browns Creek and Neville, lime rate increased DM at the lowest rate of SSP, but at the higher rates of SSP, this effect was only evident at the two highest lime rates. Group 3 was represented by E1 at Browns Creek and Neville, and showed a limited lime effect where lime tended to decrease the DM response particularly at the higher rates of SSP.

Fig. 6.

Patterns of DM response to lime and superphosphate (P) application where lime × superphosphate interactions were significant (n = 10). Similar responses (non-significant sample × L × P interactions) resulted in formation of three groups representing six, two, and two samplings for Groups 1, 2 and 3 respectively. Group 1 comprised harvests from Mt David, Black Springs, and Browns Creek. Group 2 comprised harvests from E2 from Browns Creek and Neville. Group 3 comprised harvests from E1 from Browns Creek and Neville. Responses described are mean responses for each group.

Trends in botanical composition (1987–1990)

The significance of the effect of lime and SSP rate on the proportion of the various pasture components, together with the percentages at L0 and P0 levels and the mean values for each experiment and site are in Tables 5 and 6. Also included is mean DM. Comparisons showing significant trends are only discussed further if mean DM exceeded 0.1 t/ha. Species composition across sites was variable as indicated by the significant (P < 0.001) site × species interaction.

Relationship between relative plant yield and %Al and %Ca in soil

Soil acidity parameters exhibited a strong exponential relationship between Al (as exchangeable and % of ECEC) and pH_Ca (Fig. 5b). The increasing %Al values at decreasing lime rates for each site is presented in Table 7.

Overall, relative annual pasture yield (derived from maximum annual yield for each site for each experiment for each year) was not affected by increasing %Al (Table 8) but was increased by %Ca (P < 0.001; Table 8). The relative annual yield response to increasing %Ca was greater (P < 0.01) in E1 compared with E2 and increased more rapidly in E1 as %Ca increased. Responses were significantly (P < 0.001) affected by site. Relative yield at Vittoria was increased with increasing %Al (P < 0.01), while at Black Springs relative yield declined with increasing %Al (P < 0.05; Table 9). Relative yield at Browns Creek and Vittoria (P < 0.001) were less responsive than the other sites to %Ca (P < 0.001; P < 0.05), respectively. These relationships were affected by year (P < 0.001), with 1981 the most responsive year for both %Al and %Ca, with relative yield decreasing and increasing significantly, respectively. The latter years of the experiment tended to be less responsive to increasing %Al. While relative yield increased (P < 0.001) with SSP application (P0 < P1 < P2), it declined at a greater rate as %Al increased on the P1 and P2 treatments (P < 0.05) compared to P0. However, for increasing %Ca, there were no significant effects between the SSP treatments on relative yield, with relative yield increasing (P < 0.001) at a similar rate for all SSP application rates.

Across all sites with data restricted to 5–45 Bray P mg/kg, and 0–45%Al, relative yields generally increased with increasing Bray-1 P concentrations on both E1 and E2 (Fig. 7). Compared with E1, the relative reduction in yield at very low soil P concentrations was greater in E2, depicted by more blue along the Y-axis (Fig. 7). In addition, the yield advantage at lower %Al was greater in E2 than in E1, depicted by a greater proportion of the lower right quadrant, which is coloured red in Fig. 7.

Fig. 7.

3D response of relative yield (all years; Z-axis, coloured) to Bray-1 P (X-axis) and %Al (Y-axis) in E1 and E2. Dots represent individual points. Data restricted to 5–45 Bray P mg/kg, and 0–45% Al.

Lime

This study provides rare insight on the legacy benefit of lime on soil pH on different soils. Averaged across all sites, an increase of 0.1 pH units per t/ha lime applied was still evident in 1989, 12 years post-application. The longevity of response reported here exceeds expectations established by previous, shorter-run pasture studies (e.g. Holford and Crocker 1994). Soil acidified in all treatments, including in the nil lime plots, but the rate of pH change was greater as lime rate increased. This is in line with expectations (Scott et al. 1999) and is likely explained by a combination of increased product removal (Slattery et al. 1991; Li et al. 2019), increased rates of nitrification (Paul and Clark 1996; Kaveney et al. 2020) and increased vertical movement of lime (Conyers and Scott 1989; Conyers et al. 2003) with higher pH values at the soil surface.

In the unlimed soil, pH generally declined by 0.01–0.03 pH units per year over the 12-year study period. This is less than the decline of 0.04 pH units per year reported over 18 years in a study in southern NSW (Helyar et al. 1997), probably explained by a higher initial pH_Ca (~5.2) compared to the five sites in the present study (pH_Ca 4.1–4.4). In contrast, Li et al. (2019) reported a marginal pH increase (~0.006 pH units/year) in surface soil in the unlimed plots in a separate experiment also conducted over 18 years in southern NSW, perhaps reflecting the lower initial pH_Ca (~4.0) at that site. However, Li et al. (2019) reported a decline of 0.005 pH units per year in the 10–20 cm layer in the unlimed treatments. Soil pH was not monitored at layers below 10 cm depth in the present study, highlighting the possibility that the true rate of acidification in the profile is underestimated (Condon et al. 2021).

The experiments reported here pre-dated modern recommendations targeting surface soil pH_Ca > 5.5 to ameliorate subsurface acidity (Conyers and Scott 1989; Li et al. 2019; Condon et al. 2021). For each site, the highest lime rate (5.0 t/ha) proved adequate to raise pH_Ca to 5.5–6.0 although for the Browns Creek site, which had a lower initial pH_Ca (4.15), a higher rate (5.5–6.0 t/ha) was probably required to account for the spatial and temporal variability in soil pH that is known to exist in these soils (Conyers and Davey 1990; Conyers et al. 1997; Hayes et al. 2022). Such high rates of lime may have seemed extreme when these experiments were established but are now known to be required to adequately manage soil acidity in these environments.

The increase in exchangeable cations in the surface soil at all sites following lime application was expected and largely attributed to the large addition of Ca, plus a small addition of Mg, in the lime applied. The increase in exchangeable Mg lasted throughout the 12-year period at all sites. However, at the Vittoria, Browns Creek and Black Springs sites (interestingly, the sites least responsive to lime), there was a suggestion that the increase in exchangeable Ca was diminishing by the end of the experimental period. At face value this seems surprising because a lower concentration of Mg (1.5–2.0%) existed in the lime source used compared with Ca (~40%) and at all sites, Ca comprised 48–65% of ECEC following lime application. The decline in Ca over the experimental period may be explained by plant uptake and removal, and by leaching below the soil test layer. In a recent study of the effect of long-term liming on the mineral profile of pastures, Li et al. (2024) demonstrated an increase in Ca concentrations with lime, but no increase in Mg concentrations, noting however that the source of lime was different to that used in this study. Increased Ca concentrations in plant tissues may explain the gradual depletion of exchangeable Ca over time in the limed compared to nil lime treatments.

Calculating lime rates to achieve a target pH can be problematic for farmers as it requires integration of a range of soil attributes, such as organic matter content, texture and the nature of the clay minerology (Hochman et al. 1989; Condon et al. 2021). Equations have been established to calculate the pH change for given lime rates where the pH buffering capacity (pH_BC) is known (Hayes et al. 2023) but pH_BC is not commonly provided on standard soil tests. Across the five sites in the present study, where lime was incorporated and well mixed with the soil using a rotary hoe, soil pH 12 months following application was increased by 0.3 units per t/ha lime applied. This may prove a useful rule of thumb on similar soil types where the pH_BC of the soil is not known and sophisticated decision support tools are not available. This rule of thumb is expected to be most relevant on soils with a starting pH_Ca ranging between 4.1 and 4.4, and with ECEC in the range 2.3–5.2. It is noted that in commercial situations initial increases in pH are expected to be lower than was achieved in this study, where high quality lime was well mixed with the soil.

Superphosphate

The application of SSP led to a small but significant decline in soil pH with differences in pH between the nil and plus SSP treatments increasing as the experiment progressed. This result is likely attributable to indirect effects of greater pasture biomass and removal, rather than to direct effects of the fertiliser itself as there was no difference in pH between the different rates of SSP. Direct effects of P-fertiliser on soil pH are known to be localised and proportional to phosphate retention (Saunders 1958). The small decrease in pH due to SSP resulted in a small increase in both exchangeable and %Al, indicative of the strong negative correlation between pH and Al.

The increase in exchangeable Ca due to SSP is likely explained by the Ca concentration of the fertiliser (~20%), as differences were only observed later in the experiment after several years of fertiliser top-dressing, and mainly with the higher fertiliser rate. By contrast, SSP application was associated with a decline in both exchangeable K and Mg in the surface soil, the decline being greater at the higher SSP rate. This result is likely explained by greater pasture uptake and removal of both Mg and K in herbage. Although it differs with plant species, K concentration of herbage can be around double that of Ca, and ~10-fold greater than the Mg concentration in pasture herbage (Hayes et al. 2008; Li et al. 2024). Strong DM responses to SSP over the duration of the experiment would explain a gradual depletion of both nutrients in the soil. The explanation for the decline in both Mg and K due to SSP where no such response was observed with liming reflects the substantially greater pasture DM response observed across all sites over the experimental period with SSP compared with lime. Some redistribution of these macro-nutrients in deeper soil layers also cannot be ruled out, especially some forms of K, which have previously been shown to be redistributed in soil by plants (Hayes et al. 2020). By 1989, the nil SSP treatment was depleted of P with Bray P declining to values that were just ⅕ – ½ of starting values and at all sites, generally below levels assumed to be optimal (10–12 mg/kg) for pasture production (Moody and Bolland 1999).

Bray P was used in this study as it was believed at the time the experiments commenced that Bray-1 was the more reliable index of P availability on acidic, mineral soils due to the sensitivity of Colwell P to variations in oxalate-extractable Fe (Vimpany et al. 1997). Various buffering indices have been developed to improve the interpretation of Colwell P, in addition to oxalate-extractable Fe, but the early workers at these sites considered that the use of Bray-1 was a simpler solution. When the Australian database was compiled for pasture responses to P, K and S (Gourley et al. 2019) there were 10 soil P tests in use in Australia. Of the 3638 experimental site-years available, 2244 (62%) used Colwell P and 209 (6%) used Bray-1. Therefore, Colwell P was adopted through weight of data whilst Bray 1 calibrations were restricted to particular regions and land uses such as those reported here. The question of which test is ‘better’ remains open.

Pasture response

Most pasture responses to lime in this study were observed in E2 where T. subterraneum dominated the swards. In E1 where D. glomerata dominated swards at all sites, significant lime effects were confined to species that were only minor components. It has been established that in short-term ecosystems such as pastures, total production is overwhelmingly driven by the dominant species (Grime 1998; Sasaki and Lauenroth 2011; Hayes et al. 2024). Therefore, altering the production of minor species with lime may have little effect on overall production in the short-term.

This study highlights the complexity of response to soil treatments of mixed pasture swards. First, when two soil treatments are being assessed, Liebig’s law of the minimum dictates that the relative response to either superphosphate or lime will be driven firstly by the relative limitation posed to pasture production by low pH or deficiency in P or S. In this network of sites, there was a greater number of species that responded positively to superphosphate than to lime, suggesting that at all sites the deficiency of P and/or S was a greater constraint to pasture production than was soil acidity.

Second, not only are pasture species responding directly to the soil treatment, but there is also an indirect effect of competition from neighbouring species that may themselves be responding even more strongly to the soil treatment. This competitive relationship is driven in part by the relative tolerance of the individual species to either a deficiency in P or S, or to the toxicities that exist in these low pH soils. A case in point is the negative response to superphosphate observed almost universally in Vulpia spp. It is unlikely that superphosphate inhibits the growth of Vulpia spp. when grown alone. Rather, the negative response observed is a competition effect where Vulpia growth was depressed by the many other species in the present study that grew more strongly in response to superphosphate (Dowling et al. 1997). The most consistent response to lime was the positive response of M. sativa, which is renowned for its sensitivity to acidic soil conditions (Hayes et al. 2011, 2012). P. aquatica, although also renowned for its sensitivity to aluminium toxicity (Culvenor et al. 1986) rarely responded positively to lime in the present experiment but often responded positively to superphosphate. This may be explained by a combination of factors such as: (1) soil fertility being a bigger constraint to phalaris growth than acidity; (2) phalaris was adequately able to tolerate the levels of Al toxicity (~15–30%) at this range of sites (Hayes et al. 2015); and (3) phalaris was only a minor component (1–7%) of most swards.

It is also noted that myriad other factors impact the competitive relationship beyond superphosphate and lime, to which individual species will respond differently, such as grazing management and drought stress. One also cannot overlook the overriding effects of nitrogen on pasture growth (Cocks 1980; Paul et al. 2003), which in the present experiments would have been indirectly affected by any factor that would impact legume growth, such as soil treatment, drought or grazing. Taken together, whilst this study demonstrated good relationships between soil treatment and pasture response, it would not be expected that superphosphate and lime would account for the full pasture response observed due to these other factors contributing to the competitive relationship in a mixed pasture sward.

For the farmer weighing up the merits of investing in either superphosphate or lime, this study highlights a common conundrum. On the one hand, superphosphate provided a greater immediate pasture response and so is easier to justify using standard financial analyses (Ellington 1984; Khairo et al. 2010). On the other hand, it is increasingly acknowledged that productive leguminous pastures used for agriculture inevitably acidify soils (Williams 1980; Condon et al. 2021) and that acidity will only become harder to address as time elapses and the acidity appears deeper in the soil profile (Li et al. 2019). As our study demonstrates, lime was unquestionably effective at ameliorating soil acidity, with the effects of one application still evident in a range of soil properties 12 years later. Logically, for productive and sustainable pasture systems in these Central Tableland environments, both lime and superphosphate will be required. If farmer decisions of whether lime is or is not applied continue to be driven by partial budgeting analyses that seek positive responses in the short-term, then the case for improved pastures becomes stronger (Li et al. 2006b; Hayes et al. 2016). In the present study, a greater pasture response was observed in E2 compared with E1, reflecting the more responsive ‘improved’ pastures in the E2 experiments.