Reduced irrigation in high rainfall years and winter application of nitrogen reduce granulation in Imperial mandarin (Citrus reticulata cv. Imperial)

Trial site and design

The trial was situated within a commercial orchard block in Wallaville, Central Queensland, Australia (25.1219°S, 151.9955°E), using trees on ‘Benton’ rootstocks that were 7 years old in Year 1, on a sandy soil, at 7 × 3 m spacing. Rows were on a slight slope, aligned east–west. The trial had a split plot design with irrigation treatments as the main plots and N treatments as the subplots. The subplots were of five trees, with data collected from the central three trees. There were five replicates. The trial design was balanced using two interlinked Latin squares, one for each irrigation treatment. Apart from the irrigation and N treatments detailed below, all trees were subject to the same fertiliser, pest management and harvesting practices (Hardy et al. 2017).

Treatments

Treatments were applied for 5 consecutive years (2017/18–2021/22). The two irrigation treatments were a ‘control’ treatment, that is, standard commercial irrigation as practised for Imperial blocks (Hardy et al. 2017), and a ‘deficit’ treatment. Deficits were imposed by turning off taps installed in irrigation pipes in ‘deficit’ treatment rows when the ‘control’ treatment was irrigated. The level of deficit each year was thus not an independent amount but was always relative to the ‘control’. Decisions to irrigate the ‘control’ treatment were made by the grower. Decisions to irrigate the ‘deficit’ treatment were made by the researchers, based on a visual assessment of stress in the trees and soil moisture monitoring using tensiometers. In Year 1, there was frequent rainfall in Stage I of fruit development so deficits could not be applied. We therefore continued treatments in this and subsequent years into mid-Stage II (mid- to late-January). Deficits were applied from the end of flowering (early September). Table 1 shows estimated evapotranspiration (ET₀), hours of irrigation, rainfall, and total water (rain + irrigation) as a percentage of ET₀ during treatment periods for each treatment. ET₀ is reference crop evapotranspiration for Bundaberg as calculated by the Australian Bureau of Meteorology using the Penman–Monteith equation (Allen et al. 1998). Irrigation rate was calculated as sprinkler output (L/h)/[row spacing (m) × sprinkler spacing (m)] that is, 3.3 mm/h.

Nitrogen nutrition treatments were five rates of winter applications of N: 850, 700, 550, 400 and 250 g N/tree or 405, 333, 262, 190 and 119 kg N/ha (Treatments 1–5 respectively). The recommended rate for 7-year-old trees is 700 g/N tree, increasing to 800 g/N tree from Year 8 onwards (Vock et al. 1997). In Year 2, the lowest N rate (Treatment 5) was 340 g/tree due to an application error. N applications were single broadcast applications of granular urea (46% N) in early winter with minor adjustments for fertiliser blends applied at approximately the same time. No additional N applications were made at other times of the year.

Crop load assessment

From Year 2, the crop load on each data tree was visually rated at harvest using an 11-point scale. Ratings are relative within the 1 year and years are not directly comparable.

Flush growth and fruit set measurements

From Year 2, we tagged eight twigs per tree, one at each of the 45° points of the circumference, at a reachable height on one data tree per sub-plot. In Year 2, only three of the five subplots for each treatment combination were tagged. Twigs were single growth units from the previous year’s spring or summer flush. Twigs were tagged in late winter before any spring growth was evident.

We measured the length and number of new spring flush shoots on the tagged twigs approximately every 2 weeks until fully elongated. At harvest, fruit that remained on tags were cut and rated as for the main sample (see ‘Fruit sampling and assessment’ below). Data on flush growth from twigs with no fruit at harvest were included in analysis of treatment effects on flush but excluded from analyses of flush effects on granulation. When more than one fruit was harvested from a twig, the same flush data was used for each fruit, not divided by the number of fruit.

Fruit sampling and granulation assessment

The effects of treatments were assessed at fruit maturity, approximately a week before commercial picking. Twenty fruit were picked from each data tree in a stratified random sample, five from each quadrant. Samples included fruit from both outside and inside the canopy, to a maximum of an arm’s length (~0.8 m) within the canopy, and to a height within reach of the sampler (~2 m). Where crop load was unevenly distributed around the tree, we adjusted sampling to represent the crop. In Year 5, crop loads on most trees were extremely low so we reduced sample numbers to avoid picking fruit deep inside the canopy.

Each fruit was cut in the field in half along the equator and visually rated from 0 to 5, in steps of 0.5 (Fig. 1). Fruit rated 0 had no visible granulation, that is, vesicles had thin translucent walls, and colour was deep orange. Fruit rated 5 were opaque and white, and when squeezed had no extractable juice.

Fig. 1.

Example of visual ratings of granulation severity with ratings in bottom row (L to R) 0, 1, 2, 3, 4; second row (L to R), ratings 0.5, 1.5, 2.5, 3.5, 4.5; top row, rating 5.

To calibrate the juice percentage of ratings, we selected 162 cut fruit in Year 2 and 159 fruit in Year 4 to cover as wide a range as possible in ratings. The juice percentage of each cut half of the subsample was measured. ‘Juice percentage’ was calculated as the volume (mL) of readily extractable juice (using a domestic juicer) as a percentage of weight (g) of the half fruit, including peel. The correlation coefficient (r) of the linear relationship between rating and juice percentage for the combined sample data was −0.711 (P < 0.001) (Fig. 2). On the basis of this data, we report fruit rated ≥3 or approximately <33% juice as ‘severely granulated’. Fruit rated 0, 0.5, 1 or 1.5 (mean juice percentage of 48.7, 47.7, 48.0 and 43.6, respectively) were categorised as ‘good’. Fruit rated as 2 or 2.5 were categorised as ‘crunchy’ (mean juice percentage of 37.8 and 37.9, respectively).

Fig. 2.

Juice percentages of individual fruit by rating in a sample of 162 fruit in 2020 and 159 fruit in 2022, showing the range present in each rating (r = −0.7111, P < 0.001).

Statistical analysis

Genstat (18th – 20th edition) was used for analysis. Treatment means are separated by analysis of variance using one- or two-way analysis as appropriate, using α = 0.05. We report statistical error as estimated standard error of the means (‘ese’).

Irrigation treatment effects on granulation

The deficit treatment reduced granulation in 3 out of 5 trial years, although in Year 5 treatment means were not significantly different at α = 0.05 (Table 2). The effectiveness of the treatment varied each year with differences in overall or seasonal crop load, rainfall and the volume of irrigation applied, as follows:

In Year 1, total water over the treatment period in the deficit treatment was approximately 20% less than the control, with the differential applied mostly in early Stage II (Table 1). The deficit irrigation treatment reduced the mean granulation rating from 2.72 to 2.43 (P = 0.013). The mean proportion of severely granulated fruit was reduced from 47.7% to 37.5% (P = 0.017) (Table 2).

In Year 2, there was less rainfall than Year 1 and a high crop load. In the deficit treatment, total water was reduced by approximately half, the mean granulation rating was reduced from 1.55 to 1.06 (P = 0.009), and the mean proportion of severely granulated fruit from 8.5% to 1.1% (P = 0.012) (Table 2).

In Year 5, a year with high rainfall and very low crop load, reducing total water by 10–15% reduced the mean granulation rating from 3.06 to 2.75 (P = 0.147), and the proportion of severely granulated fruit from 62.3% to 47.6% (P = 0.080) (Table 2).

In Years 3 and 4, the deficit treatment (total water reductions of 50% and 40% respectively) had no additional benefit. We attribute this to low rainfall and the reduced volumes in the control irrigation regime (Table 1). After good results from the deficit irrigation treatment in the first 2 years of the trial, the grower reduced control irrigation volume so that total water in this treatment dropped from >100% of ET₀ in the previous 2 years to 56% of ET₀ in Year 3 and 77% of ET₀ in Year 4. High seasonal crop load in Year 4 also meant minimal granulation in all treatments. Fruit in the deficit treatment samples were on average smaller in diameter in all years. In most years, the mean treatment differences were <3 mm but in Year 3, a year with low crop load and the lowest level of total water in the deficit treatment, the means differed by 7 mm (Table 2).

These results indicate that overwatering is an important cause of granulation. The current advice to Australian growers is that mandarin water requirements are high, 7–12 ML/ha per annum, and that it is critical to avoid water stress in flowering, fruit set and Stage I of fruit development (Hardy et al. 2017). Our results suggest these recommendations may need to be revised for Imperials. In our trial, the total water for the control treatment during the 16–18 week treatment period in the wet years 1, 2 and 5 was 6.3, 7.2 and 8.6 ML/ha (101%, 98% and 129% of ET₀), compared to 5.5, 4.1 and 7.4 ML/ha in the deficit treatment (87%, 56% and 112% of ET₀) (Table 1). Low levels of granulation in both irrigation treatments were achieved in the low rainfall years (3 and 4) within the range 1.8–5.8 ML/ha, or 27–77% of ET₀.

Although the number of data points is small, our results show a clear trend in increasing granulation when the ratio of total water received to evapotranspiration increases, particularly in low crop load years (Fig. 3). Variability in crop load makes it difficult to suggest specific irrigation parameters, but our data shows that applying total water to 50–60% of ET₀ or 3–4 ML/ha in the 16–18 week treatment period would be a good experimental starting point. The suggested crop factor for citrus at 70% canopy cover as calculated by the UN’s Food and Agriculture Organization is 65–67% of ET₀ (Allen et al. 1998). When water in the deficit treatment was reduced as low as 27% of ET₀ or 1.8 ML/ha in the treatment period in Year 3, the effect on granulation was slightly detrimental (P = 0.147) and fruit size declined (P = 0.002), so this may indicate the lower level of the range for experimentation.

Fig. 3.

Relationships between the ratio of total water/estimated ET₀ and mean granulation ratings (left) and % of fruit rated ≥3 (right) for irrigation treatments in 5 years of study, showing high and low crop load years separately. For mean granulation ratings in low crop load years r² = 0.876 (P = 0.004) and in high crop load years r² = 0.344 (P = 0.250). For mean % of fruit rated ≥3 in low crop load years, r² = 0.917 (P = 0.002) and in high crop load years, r² = 0.541 (P = 0.167).

Although our earlier research suggested that Stage I of fruit development was more critical (Hofman 2010), in this new trial we carried deficits through to mid- to late-January (mid-Stage II). We suggest that granulation could be triggered at any stage of fruit development but is likely to be most severe when triggered early, as it appears that, once triggered, it continues to exacerbate without the possibility of reversal. It may be that a deficit all season is beneficial: this remains to be tested.

Nitrogen treatment effects on granulation

There was lower mean granulation in the higher winter N treatments in the first 3 years of the trial (P = 0.324, 0.022, 0.084, respectively) (Table 2). Although differences between means were only significantly different at α = 0.05 in 1 year, the trend is clear. We note that Treatment 2 (700 g N/tree) tended to show higher levels of granulation than would be expected from the general trend. We attribute this anomaly to poorer tree health and shallow soil profiles in 3 of the 10 plots for this treatment. In Year 4, a very high crop load year, granulation was minimal and all the N treatments had similar granulation levels (P = 0.376). In Year 5, the trend was reversed: the higher N treatments tended to have higher granulation than the lower N treatments (P for mean granulation = 0.065; P for % of fruit severely granulated = 0.195). The reason for this reversal is unclear, as flush growth was lower for the higher N treatments (Table 3) and there were no clear differences between N treatments in fruit set (Table 4) or crop load at harvest (Table 2) (see discussion in ‘Flush growth and granulation relationships’ below).

These results indicate that a factor in managing granulation is to ensure sufficient application of winter N. Our study suggests that the current recommendation of 800 g N/tree for mature trees is appropriate (Vock et al. 1997).

Flush growth and granulation relationships

Our data shows different patterns in Years 2, 3 and 4 for flush and granulation relationships at the individual twig level (Figs 4, 5 and 6). In Year 5, there were insufficient numbers of fruit remaining on tagged twigs at fruit maturity to allow analysis of associations between flush and granulation.

Fig. 4.

Mean maximum length of spring flush at six dates by rating of individual fruit in Year 2 (high crop load year). All treatments combined. Error bars show standard deviations. Fruit of different ratings may come from the same twig. Data includes zero values (i.e. where twigs set fruit but had no spring flush). The numbers of fruit in each category were 0 = 25, 0.5 = 41, 1 = 33, 1.5 = 24, ≥2 = 14 (total 137 fruit). Mean fruit diameter (mm) in each category for rating 0 = 49.5, 0.5 = 51.4, 1 = 52.2, 1.5 = 54.0, ≥2 = 57.4.

Fig. 5.

Mean maximum length of spring flush at four dates by rating of individual fruit in Year 3 (low crop load year). All treatments combined. Error bars show standard deviations. Fruit of different ratings may come from the same twig. Data includes zero values (i.e. where twigs set fruit but had no spring flush). The numbers of fruit in each category were 0 = 10, 0.5 = 13, 1 = 45, 1.5 = 22, 2 = 14, 2.5 = 8, ≥3 = 8 (total 120 fruit). Mean fruit diameter (mm) in each category for rating 0 = 48.6, 0.5 = 53.8, 1 = 55.7, 1.5 = 57.7, 2 = 58.3, 2.5 = 62.4, ≥3 = 57.4.

Fig. 6.

Mean maximum length of spring flush at five dates by rating of individual fruit in Year 4 (very high crop load year). All treatments combined. Error bars show standard deviations. Fruit of different ratings may come from the same twig. Data includes zero values (i.e. where twigs set fruit but had no spring flush). The numbers of fruit in each category were 0 = 21, 0.5 = 70, 1 = 93, ≥1.5 = 23 (total 207 fruit). The highest rating of fruit in this sample was 2 (two fruit). Mean fruit diameter (mm) in each category for rating 0 = 49, 0.5 = 52, 1 = 56, ≥1.5 = 59.

In Year 3, a low crop load year, there was a high incidence of granulation. Mean maximum flush length was higher than in high crop load years. Stronger early flush growth was associated with higher granulation ratings; stronger later flush was associated with lower granulation ratings (Fig. 5). We suggest that in low crop load years, carbohydrate and nutrient reserves in the tree have likely been depleted by the previous season’s large crop, so there is competition between fruit development and early flush growth. Later flush seems to support better fruit development. This may be because in low crop load years there are fewer fruit and more flush to supply carbohydrates to the lower numbers of fruit, once the flush matures (Goldschmidt and Golomb 1982).

In Years 2 and 4, both high crop load years, there was less granulation and mean maximum flush length was shorter than in the low crop load year. In Year 2, there appeared to be no detrimental effect of early flush growth but stronger later flush meant fruit were more granulated (Fig. 4). In Year 4, a very high crop load year, there appeared to be no association between granulation and flush growth at any measurement date (Fig. 6). Carbohydrate and nutrient reserves are known to accumulate in low crop years (Jones et al. 1975; Goldschmidt and Golomb 1982) and our results suggest that, in the subsequent high crop load years, these reserves are sufficient to support both early fruit development and spring flush growth. In addition, heavy flowering and fruit set suppresses flush growth, reducing potential competition. The observed detrimental effect of stronger later spring flush in Year 2 may be due to earlier depletion of reserves in that year.

The extent to which fruit set and new flush growth in citrus rely on carbohydrate reserves is under dispute, with some studies suggesting growth relies on current photosynthesis rather than stored resources (Kriedemann 1969; Monerri et al. 2011). However, our observations of the different patterns in low and high crop load years for granulation suggest that reserves play a significant part early in fruit development for Imperials. This reinforces the need for adequate application of winter N.

Treatment effects on spring flush growth and fruit set

The deficit irrigation treatment had a lower mean maximum spring flush length than the control in all years, and a lower mean number of spring flush shoots per tag in all years except Years 4 and 5 (Table 3). The data on fruit set for the irrigation treatments show an inverse pattern to flush growth, with the deficit treatment setting and retaining more fruit, although P values were high in Years 2 and 5 (P = 0.689 and 0.591, respectively, for October measurements) (Table 4). These results may mean that overirrigation, in fostering greater flush growth, exacerbates flush and fruit competition. However, data for overall granulation incidence were not fully aligned with these trends, with the deficit irrigation treatment only reducing granulation in some years, despite having shorter flush and setting more fruit in all years. Other factors, such as water availability to fruit, may have more influence on overall granulation incidence than flush competition.

The fruit set data for Years 2, 3 and 4 show a weak pattern of better early (October) fruit set from the higher N treatments, with variable plot means (P = 0.281, 0.363 and 0.970, respectively) (Table 4). However, the patterns of flush growth for N treatments differ with seasonal crop load. The higher N treatments had less maximum flush growth per twig than the lower N treatments when the seasonal crop load was low (Years 3 and 5) (P = 0.027 and 0.011, respectively) (Table 3). In high crop load years (Years 2 and 4), the trend is reversed, although variability was high (P = 0.680, 0.303, respectively). This was despite the pattern of reduced granulation from the higher N treatments in all years except Year 5. Thus the driving factor in N treatment differences is more likely to be crop load and its effects on reserves, rather than flush growth per se.

The flush and fruit set data show that variability in flush growth from twig to twig may explain some of the variability in granulation severity from fruit to fruit within the one tree but suggest that other factors such as crop load and water availability may be more important. Treatments that regulate flush growth, such as plant growth retardants, are worth further testing, but are likely to be less effective than managing crop load. Strategies to better manage crop load include fruit thinning, timely picking, preflowering drought stress and preflowering gibberellic acid sprays.

Possible role of juice cell water potential

The effectiveness of both irrigation and N treatments suggests that cell turgor and soluble solute concentrations may be key physiological factors underlying granulation. Turgor (or ‘hydrostatic potential’) and soluble solute concentration (or ‘osmotic potential’) are components of water potential, a measure of the free energy of water per unit volume, which drives water movement in plant cells. Juice cells in citrus fruit are not directly connected to the vascular system, so maintaining moisture content depends on low water potentials within the cell (Kaufmann 1970; Mantell et al. 1980). Low cell turgor and high levels of soluble solutes enable continuing water diffusion through permeable cell walls or channels in cell walls into juice vacuoles. Results from the current study are consistent with our earlier suggestion that granulation may be linked to relatively high water potential in juice cells, that is, to high cell turgor and low soluble solute concentrations (Hofman and Smith 2008).

Differences in cell water potential could be an explanation for the variability in granulation severity between seasons, between trees, between fruit within the one tree, and between vesicles within the one fruit. Cell turgor and solute concentration could differ at all these levels as a result of factors such as availability, competition, distribution patterns and transport efficiencies. Water potential could also be an explanation for many of the environmental and physiological conditions associated with granulation. For example, the association of granulation with vegetative vigour and mineral deficiencies could be explained by their detrimental effects on soluble solid content in juice cells. The associations with larger fruit size, rootstock vigour, coastal regions and higher rainfall could be explained by higher cell turgor. The associations of low crop loads and shaded positions in the canopy with granulation could be due to either or both components.