Fertilisation with P, N and S requires additional Zn for healthy plantation tree growth on low fertility savanna soils

Statistical analysis

Factorial analyses of variance were carried out using the statistical package GenStat (VSN International Inc.; https://vsni.co.uk/software/genstat), testing three levels each of N, P and S, the 27 treatments replicated twice and plots allocated randomly in each replication on the three sites in randomised complete block designs, resulting in 22 degrees of freedom for the residual term (Cochran and Cox 1957). Linear regression was used to test whether stem fresh biomass, a commercially important indicator of plantation yield, could be estimated from standard stem dimensions (height and D_BH) as combined into V_c,BH.

Second phase experiments

The N, P and S experiments described above were terminated after the 3.81-year measurements. At that time, late in the dry season, termites were killing trees at GP, and trees had become unhealthy at all three sites with symptoms ranging from short needles to fasciation and stem tip die-back, similar to symptoms on other plantation species nearby (Rance et al. 2022). The incidence of affected trees was highest at TW and lowest at HD, although occasionally a tree or two appeared to be healthier than those surrounding. The experiments became test beds for other work.

Three Ways

A patch of unhealthy trees, approximately a third of the experiment and extending across N, P and S treatments, was defined from healthier trees by a clear boundary, indicating consistent soil differences without obvious differences of soil colour or texture. Severely unhealthy trees had very short needles, commonly approximately 5 cm long, compared with 15–20 cm on healthy trees, some with no extended needles, premature needle fall, some with fasciation and dead tops (Fig. 1). Stunted tip growth hinted at Zn, Mn or B deficiency.

Fig. 1.

An unhealthy tree with a dead stem tip and multiple leaders, and uneven needle lengths at Three Ways, some with no extended needles. Photo: S. Rance.

KCl (28 g tree⁻¹) was applied in a ring around two rows of trees in each net plot (2.91 years). Twelve matched trios, ranging from mildly unhealthy to very unhealthy, were selected and measured. One of each trio was sprayed to wet all needles with a commercial 9.8 g Zn chelate L⁻¹ solution, a second matching set had 0.5 L of ZnSO₄ solution (10 g ZnSO₄ L⁻¹, 9 kg Zn ha⁻¹) applied to the soil surface in a ring around the trees, the third untreated.

At age 5.7 years (6.2 years after clearing the site for plantation establishment), soil cores of 4.8 cm internal diameter by 0–5 cm deep were collected along two transects, the first in the undisturbed native forest adjacent to the experiment, the second between trees along the diagonals of plots, starting in the healthier part of the experiment, through three plots in the badly affected area, and back into the healthier zone. Samples were taken at intervals between 1 and 2 m along the transects, spaced to minimise damage to tree roots. Ten samples were taken in the native forest, nine in the better part of the experiment and 19 among the severely affected trees. Exchangeable cations were leached with 1N NH₄Cl (pH 7) solution, shaken and filtered, and cations (Ca, Mg, K, Na, and Mn) were determined by atomic absorption spectroscopy (AAS).

Second phase experiments

After fire had removed the litter layer at the start of the second phase experiments at HD, interrupting nutrient cycling, responses to factorial N, P and S treatments remained similar to the first phase (Supplementary Tables S1 and S2). The effect of S₁ applications was perhaps more pronounced, but with little further growth to S₂. The additive effects of the main treatments were substantial.

Responses to K applications at TW increased foliar K concentrations (K₀ = 2.5 mg/g, K₁ 3.2 mg/g, P < 0.001), but there was no significant height increase. Unhealthy trees with Zn chelate sprayed onto the needles improved rapidly in visual health, but there was no apparent response to 9 kg Zn ha⁻¹ applied to the soil surface around the trees within 1 year. Height increments after soil-surface Zn and Cu was applied at HD at 5.91 years were not significantly different by age 6.54 years, but diameter increments were increased by Zn applications (P < 0.01; Table 9), and the response persisted. By age 7.57 years, some trees were over 12 m tall. When expressed as V_c,BH increments from 5.91 to 7.57 years, Zn applications tended to suppress growth on N₀P₀S₀ treatment, had little consistent effect on combinations with one or two of the main treatments, and increased growth in N + P + S combinations (Table 9). There was no significant growth response to Cu applications.

A striking feature of exchangeable cation concentration measurement at TW was the wide range of values over short distances. Soil samples a metre apart had very different cation profiles, both in absolute concentrations and in ratios between cations (Table 10). Concentrations of cations along the transects were examined in relation to land preparation, fertiliser treatment, and tree growth and health. Variation of Ca and Na concentration along the transects was only partly related to applied treatments in plots (S was applied as CaSO₄ and P as NaH₂PO₄.2H₂O). Mean exchangeable Mg concentrations were significantly lower on the cleared (now plantation) site than those on the undisturbed forest (P < 0.001), and mean concentrations of Ca, Na and the total of the five concentrations in the poor patch were significantly lower than the undisturbed forest soil (P < 0.01). Mean K and Mn concentrations were not significantly different. High Ca concentrations were associated with low K/Ca ratios, and low Ca concentrations with high K/Ca ratios (Fig. 4).

Fig. 4.

K/Ca ratios by Ca concentration (cmol kg⁻¹ of soil) from all samples in the transect plots.

N, P and S fertilisation

The combined primary growth response was to P, N and S, and probably in that order, the best growth on N + P + S treatments. The main responses were consistent with other work in the Northern Territory (Cameron et al. 1981, 1982; Rance et al. 1982, 1983), and with preliminary glasshouse experiments (Rance et al., 2020). Indeed, multiple element responses are common across a variety of savanna and grassland ecosystems (Devi Kanniah et al. 2010; Fay et al. 2015), indicating that these responses were widespread. Growth in the field experiments was poorest at GP, yet best in the glasshouse on GP soil where it was not water stressed.

Despite positive growth responses to all main experimental treatments, the emergence of symptoms of ill health, most prominently on better-performing trees, provided strong evidence of further nutrient limitation. Symptoms of ill health started to appear near the tops of stems as the second dry season progressed; indicating uptake of a key element(s) was restricted in the drying soil profile, as was translocation from old growth to new. Accordingly, foliar sampling was standardised to the youngest fully expanded needles near the top of the tree, and newly extended growth flushes (branch tips) before needle extension/formation, collected near the end of the dry season when trees were stressed and conditions stable. Mean height increments increased in response to fertiliser applications when needle N concentrations were less than approximately 7 mg g⁻¹ (Tables 3–5), and when P and S concentrations were less than approximately 0.6 mg g⁻¹. In most cases, the best growth was on the N₂P₂S₂ treatment, although other N + P + S treatments were often not significantly different (Table 7) and could have been more economical.

It was standard practice to measure tree heights to 6 m and D_BH on trees >6 m tall as it was anticipated that the two would be correlated, but height appeared to be the least sensitive parameter to examine treatment effects. Weed competition restricted height growth in the first year, and branch growth and the proportion of foxtails (trees with few or no branches) appeared to partly reflect treatment effects and branching affected stem taper. Diameters could be combined with heights for an estimation of stem volume. A diameter low in the stem but above the butt swell would appear the most appropriate. This raised a question of where and when to standardise diameter measurement of plantation P. caribaea trees to assess growth responses as early as possible in the life of a research program, and the associated question about normalising diameters measured at different heights (Rance et al. 2011). A less than threefold increase in height at 3.58 years at HD represented a 13-fold increase in V_c,BH (Table 9), recognising that V_c,BH ignored the stem below 1.3 m.

Addressing causes of tree ill health

Termites (M. darwiniensis) became more numerous and active after clearing (Fox and Clarke 1972), and Mirex baits eliminated an estimated 200 kg of termites in each colony at GP, with minimal exposure to predator species such as ants or aboveground species. A colony at HD at 6.66 years hinted at borderline soil conditions for termites to establish colonies, possibly related to soil pH (Table 1). In the glasshouse, applications of N + P increased growth of P. caribaea var. hondurensis on soil with pH >5.2, but not at lower pH (Rance et al. 2020). The two sites initially free of termites, TW and HD, had surface soil pH ~5.2 (Table 1), while the two with active termites, GP and Howard Springs (Cameron et al. 1981), had soil pH >5.4. This is the range where some elements become more or less available in soil solution, Al more soluble, perhaps to concentrations toxic to the soil microbiome and the gut fauna of termites. Whatever the factors limiting termite occurrence, the risk that fertilisation to improve plantation tree performance may also enhance termite activity needs investigation. Mirex is too toxic to be used as a routine insecticide and termites have an important role in nutrient cycling.

Most of the minor and trace elements analysed were within or near ranges associated with healthy growth (Mengel and Kirkby 1982). Fewer trees at HD displayed acute symptoms of ill health like dead tops and fasciation than TW and GP, indicating that conditions were closer to a critical value. Stunted tip growth was consistent with the observation that concentrations of Zn in the new growth of Zn-deficient Pinus radiata were lower than older foliage and that Zn is relatively immobile (McGrath and Robson 1984). Stunted growth, short needles, yellowing and stem tip death are also consistent with Al toxicity in some species (Foy et al. 1978), and soil pH at the three sites was in the range where changes in pH could influence Al solubility. The sites free of termites, TW and HD, also had the lowest soil pH (Table 1) and highest foliar Al concentrations (Table 8). NaH₂PO₄.2H₂O applications on these soils increased soil pH in the glasshouse, and NH₄NO₃ and CaSO₄ applications reduced pH (Rance et al. 2020), so part of the growth responses could have been related to changes in pH and Al availability.

As a guide, most agricultural species with foliar Zn concentrations of 16–20 μg g⁻¹ were likely to respond to Zn fertiliser applications, those with concentrations of 15 μg g⁻¹ or less were probably Zn deficient (Loneragan et al. 1979; Mengel and Kirkby 1982). Foliar concentrations of Zn less than 10 μg g⁻¹ (Table 8) were consistent with a likely Zn deficiency at all three sites. Needle Zn concentrations in P. caribaea above ~6–7 μg g⁻¹ were associated with short needles and abnormal stem tip growth but trees still maintained reasonable growth; those <4 μg g⁻¹ were less healthy. Foliar applications of Zn chelate at TW improved the appearance and increased growth within weeks of application, confirming a Zn response (Rance et al. 1982). However, the foliar spray resulted in a negligible change in needle Zn concentrations, indicating the response would be transient, but there seemed to be little uptake from 9 kg ZnSO₄ ha⁻¹ applied to the soil surface at TW by the following year. This application rate was commonly used in agriculture, so the application rate on the second phase experiment at HD was increased to 73.4 kg ZnSO₄ ha⁻¹. This returned a small mean diameter increment increase (P < 0.01) within 0.63 years (Table 9), a larger response a year later. A more careful examination indicated V_c,BH increments were only increased with Zn in N + P + S treatments. In the second year, increments were reduced by 10% in N₀P₀S₀ treatments, increased in N + P + S treatments, intermediate or inconsistent in treatments with one or two of N, P or S. Zn applications at HD indicated that the response was delayed and secondary to other responses. The role of P fertilisation in inducing or enhancing symptoms of Zn deficiency by depressing Zn absorption or diluting available Zn supplies in greater growth (Loneragan et al. 1979), also need to be considered.

A study of grassland sites across six continents identified nutrients other than N and P that were important for plant production, Zn and Fe well represented (Radujković et al. 2021). Unlike other countries, Australia does not appear to have a widespread response to Fe, perhaps related to the lack of large herds of herbivores migrating with the seasons, exporting Fe from the grasslands in haemoglobin and myoglobin. In the short term, foliar sprays seem the most efficient method of Zn application, until nutrient cycling through litterfall is re-established.

The transect soil sampling at TW indicated that clearing and root raking the sites before the trees were planted removed nutrient capital in the tree and shrub fraction off-site, concentrating it in windrows, interrupting internal nutrient cycling from old growth to new, nutrient cycling through litter fall, and exposed surface soil to erosion and further nutrient loss (Leuning et al. 1991; Saur et al. 2000; O’Connell et al. 2004). Site preparation and tree harvesting were recognised as the major mode of nutrient removal from the plantation system (Mendham et al. 2003; O’Connell et al. 2004). A significant reduction of soil Mg concentration (Table 10), and probably in chlorophyll in foliage, indicated a likely response to Mg fertilisation in later years. Clearing trees and shrubs without burning and with minimum soil disturbance, leaving the litter/soil exposed to full sun, removed litter fall from the growth cycle (Cunningham 1963). Unshaded soil experienced higher day temperatures and lower moisture contents, reduced organic C, organic P and total N concentrations, lowered cation exchange capacity, exchangeable K and soil pH within 3 years compared with shaded soil (Cunningham 1963). Shaded soil also contained more N fixing and nitrifying bacteria than exposed soil. Litter removal reduced soil moisture content down to 90 cm and increased water stress in longleaf pine (Ginter et al. 1979), and interrupted nutrient cycling (Stocker 1981; West 1984). Further, repeated harvest residue removal reduced productivity of E. globulus plantations (O’Connell et al. 2003; Mendham et al. 2014).

The variability of cation concentrations across the site must be considered for any soil sampling strategy to assess nutrient stocks for site description (e.g. Table 1), assessing tree performance to experiment treatments and determining optimal fertilisation rates and composition. The mean of the five highest readings of Ca concentration in the poor patch was 0.606 cmol.kg⁻¹ soil (Table 10), and had no overlap with the mean of the five lowest readings (0.169). The sum of the five measured exchangeable cations (Ca, Mg, Mn, K and Na) represented approximately 40% of the cation exchange capacity in the native forest, 31% in the better-performing parts of the experiment and 20% in the poor-performing patch.

Large variation of soil conditions over small spatial scales has been attributed to historical processes affecting soil development (Phillips and Marion 2004). In our study, the substantial variation in cations at small scales might be echoes of past events such as ash beds from burning large structures like tree stems (Fig. 5), or foraging ants or grass-eating termites concentrating nutrients at nest sites. The K/Ca ratios (Fig. 4) are consistent with a long history of recurrent fires. K is volatilised at 779°C and Zn at 907°C (Latimer and Hildebrand 1956), temperatures experienced in burning wood and likely to be lost to the atmosphere, while Ca is volatilised at 1482°C, so likely to remain in ash beds. In these experiments, most of the wood and crowns in the windrows containing a significant proportion of the site nutrient capital burned during the first 2 days, but some smouldered for more than a week. Much of the site N, P, S, K and Zn, and possibly B nutrient capital would have been volatilised in high-temperature combustion in the first 2 days or lost in the smoke fraction when the windrow smouldered (Raison et al. 1985; Cook 1994; Alves et al. 2011; Pereira et al. 2011; Radujković et al. 2021), leaving refractory cations of Al, Fe, Ca and Mn in the ash bed on the exposed soil surface where they could be washed downhill or down the profile during high-intensity storms in the build-up to the wet season.

Fig. 5.

A tree felled during a severe wet season storm and burned during the dry season. In the following wet season, grass seed blown into the ash bed germinated but died soon after. Photo: S. Rance.

Conclusion

The close spacing in these experiments reduced the stands to a near monoculture and revealed the main treatment responses but exposed faster growing trees to environmental stresses. Establishing new crops in savanna regions of northern Australia, including exotic plantation species, are likely to need broad-spectrum fertilising and protection from fire, especially in the establishment phase. Nutrient additions will need to encompass minor and trace elements in addition to N, P and S, specifically Zn, and consider the mechanism of application. Fertiliser application rates will need to be low to minimise induced nutrient deficiencies and site run off, and consider small-scale variation in cation exchange capacity and composition. Changes of soil pH or increased risks of insect attack also need to be considered.