This study was initiated in spring 2016 and ended in spring 2018 in Quebec, Canada (46˚14'16" to 46˚19'41.4" N, 72˚02'13.4" to 71˚44'19.7"W), at three sites as follows: 1) Notre-Dame-de-Lourdes (conventional site 9), 2) St-Louis-de-Blandford (organic site A9), and 3) Laurierville (conventional site 10). Sites 9, A9 and 10 were 21, 12, or 9 years old, respectively. The cultivar was “Stevens”. Sites Soil series were Saint-Jude (sandy, mixed, acid, mesic Aquic Haplorthod) at site A9, Saint-Samuel (sandy, mixed, acid, mesic Typic Humaquept) at site 10, and Sainte-Sophie (sandy over loamy, mixed, acid, mesic Typic Haplorthod) at site 9.

Permanent plots of 4 m × 3 m (12 m²) in size were installed in the cranberry basins. The experimental design was a randomized complete block with two replications and four treatments per site, totaling 24 plots per year. Sulfur was applied annually as granular (260 SGN = 2.8 mm ∅) Tiger 90CR S (90% split-pea-shaped pastilles S regular grade manufactured by Tiger Resources Technology, Calgary, AB, Canada). Tiger 90CR is an S-bentonite product claimed to disperse and degrade rapidly into sulphate throughout the growing season. The S amendment (0-0-0-90) was applied at rates of 0, 250, 500, and 1000 kg S ha⁻¹·year⁻¹, within the range suggested in literature [6] [7] . The effect of elemental sulfur was monitored during two consecutive years.

The sites were irrigated to maintain soil matric potential between −3 and −7 kPa [19] . The N dosage was 45 kg N ha⁻¹ as ammonium sulfate (21% N) on conventional sites or certified fish emulsions (6-1-1) on the organic site. The P dosage was 15 kg P ha⁻¹ as triple superphosphate (0-46-0) or bone meal (0-13-0), respectively. The K dosage was 80 kg K ha⁻¹ as potassium sulfate (40% K) or sulfate of K and Mg (18% K and 9% Mg). The Cu and B rates were 2 kg Cu ha⁻¹ as Cu sulfate and 1 kg B ha⁻¹. Sulfur and NPK fertilizers were applied at the same time.

From the last week of May to early June 2016, 2017 and 2018, soils were randomly sampled in the root zone (0 - 15 cm) before fertilization. Samples were air-dried at 50˚C for 24 h and screened to less than 2 mm before conducting analyses. Soil texture was measured using the hydrometer method [20] . Soil texture varied slightly among sites (Figure 1).

Soil nutrients were extracted using the Mehlich III method [21] and quantified by inductively coupled plasma emission spectroscopy. Organic C was quantified by combustion using the Leco CNS-2000 analyzer (LECO Corp., Joseph, MI). Soil pH was measured in a 1:2 soil-solution (0.01 M CaCl₂) volumetric ratio. Soil pH values before sulfur treatments in June 2016 are presented in Table 1. The pH_CaCl2 was converted into pH_CaCl2 as follows [22] :

p H w a t e r = 1.0205   p H C a C l 2 + 0.2941 ,   r 2 = 0.995

As a result, pH_water of 4.20 would correspond to pH_CaCl2 of 3.83, indicating need for acidification to tackle nitrification at the three sites.

Two hundred tissue samples of fruit-bearing and non-fruiting uprights were collected in each plot between late August and early September [23] . Tissues were dried at 55˚C in a forced-air oven and ground to less than 1 mm. Samples were analyzed for N using the Leco CNS-2000 analyzer. After digestion in a mixture of nitric and perchloric acids, P, K, Mg, S, Cu, Fe, Mn, Al, and Zn were quantified by plasma emission spectroscopy. The B was quantified by the azomethine-H colorimetric method after tissue calcination.

Fruits were hand-harvested in four 30.5 by 30.5 cm areas per plot before flooding the basins at the beginning of October, then counted and weighed after discarding fruits mechanically bruised, infected by rot, or damaged by insects. Samples were stored at 4˚C after field harvest then frozen at −18˚C for a minimum of 1 month for TAcy and Brix analyses. Berry TAcy (total anthocyanin content) [24] and Brix (refractometer) were analyzed at the Ocean Spray Laboratory of Quality Operating Standard in Warren, WI. Fifty berries for firmness detection were refrigerated per plot overnight and then stored at room temperature for 1 - 2 h. Berry firmness was measured with the TA.TX2 Texture Analyzer (Texture Technologies Inc., Scarsdale, NY) [25] . The trigger force was 0.1 N. Pre-test speed was 1 mm·sec⁻¹, test speed, 2 mm·sec⁻¹, and post-test speed was 10 mm·sec⁻¹. Dried fruit samples (65˚C) were ground to <0.2 mm for mineral analysis. Nitrogen and carbon concentrations were quantified by the Leco CNS-2000 analyzer. The P, K, Mg, Ca, S, Cu, Fe, Mn, Zn, B, and Al were quantified by inductively coupled plasma spectroscopy after nitric acid digestion of 0.2 g dry samples.

Plant nutrients are interrelated and multivariate. Nutrient relationships are commonly expressed as pairwise ratios [26] . Pairwise ratios can be integrated into a single multi-ratio formulation using the centered log ratio transformation (clr) as follows [13] .

c l r x i = x i g ( x i )

where x i is the i^th D-part component of tissue composition, and g ( x i ) is geometric meanacross components, computed as follows:

g ( x i ) = C × N × P × K × Mg × Ca × Al × S × Mn × Zn × Fe × Cu × F v 13

Each clr expression is a linear combination of pairwise ratios, exemplified by the N balance as follows:

c l r N = ln N g ( x i ) ln ( N 13 C × N × P × K × Mg × Ca × Al × S × Mn × Zn × Fe × Cu × F v ) 1 13

where F_v is the filling value computed by difference between measurement unit and the sum of nutrient concentrations.

We used the open-source statistical software R [27] version 4.2.2 to analyze the data and draw figures. Soil, upright and berry compositions were clr-transformed to run statistical analysis. Data were standardized to express slope coefficients on a common scale. The R packages were compositions [28] (version 2.0.4) for clr transformations, vegan [29] (version 2.5.7) for ordination, tidyverse [30] (version 1.3.1) for data wrangling, ggtern [31] (version 3.3.5) for soil texture triangle and nlme [32] (version 3.1.155) to compare the coefficients [33] . Significance was tested for each primary outcome at the 95% compatibility interval to avoid converting the p-value into a Bayes factor [34] . Sulfur and year effects were calculated as linear coefficients using a mixed model. Random effect was computed for each site within years and for each block within sites or years. Model accuracy was measured by the R² coefficient.

Sulfur dosage impacted soil pH over two years (Table 2). The regression model relating pH_CaCl2 to sulfur dosage through years was linear as follows:

Final   p H C a C l 2 − Initial   p H C a C l 2 = − 0.0001586429 × S d o s a g e − 0.0779807692 × Δ y e a r

where ∆year is elapsed time in year and dosage is the annual S application rate (0 to 1000 kg of S ha⁻¹). The R² coefficient of the mixed model was 0.53, indicating high variability where sulfur was surface applied in field trials compared to the thorough mixing of elementary S and soils monitored in traditional incubation studies [35] . Indeed, soil acidification rate must vary widely with initial pH, sulfur source and rate, the method of application, soil buffering capacity and the abundance or activity of S oxidizers over time [35] [36] [37] [38] .

Two years after S applications, the pH_water of 4.2 was reached by applying 250 kg S ha⁻¹ at site #10, the most acidic soil condition, 500 kg S ha⁻¹ at site #A9 of intermediate soil acidity, and 1000 kg S ha⁻¹ at site #9, the less acidic soil (Figure 2). In comparison, S dosage up to 1120 kg S ha⁻¹ split in two applications has been recommended to decrease soil pH by one unit [6] [7] [8] , likely in soils showing much higher initial pH values than in the present study (upper limit recommended for pH_water is 5.5) [2] . The vertical distribution of soil pH_water values must be highly variable where sulfur is surface applied at the various rates and under various initial soil pH values. The stratification of soil pH over time should be further investigated in relation with plant rooting depth.

Elemental sulfur tended to increase the S concentration in soils but to decrease that of Ca (Table 3). There was a comparable pattern for the clr values reflecting soil S and Ca balances (Figure 3). The fact that soil S increased markedly with S additions indicated effective conversion of elemental sulfur into sulfuric acid. The oxidation rate depends on soil moisture, aeration, surface area of S particles, microbial population and the contact between S particle and microbes [39] that are site-specific.

The S treatments apparently impacted upright nutrient composition more than berry nutrient composition as shown by average values (Table 4). The N, S, P, K, Cu and Al concentrations in uprights were highest for the highest S treatment. However, the clr values was shown to increase significantly for the S balance only, to decrease significantly for the Zn nutrient balance only (Figure 4).

Because cranberry grows in strongly acidic soils where aluminum [40] , iron [41] , and Mn [42] toxicity may occur, the bioavailability of Fe and Al may increase and that of S decrease with the application of elemental sulfur to lower soil pH [6] . The Al toxicity in plant tissues may be problematic in acid soils at pH values lower than 5.5 [40] . Soil pH_water below 4.8 may cause Mn toxicity to sensitive crops [43] [44] but this was apparently not a problem for cranberry. The S treatments did not impact significantly fruit Brix, firmness, TAcy, weight and yield (Figure 5). Possibly, crosstalks between S, Mn, Fe and Al may tackle metal toxicity [45] .

While excessive N inputs may redirect C allocation and produce vegetative overgrowth [46] [47] , the nitrogen appeared to be properly balanced with other tissue components to reach high berry yield and quality in the present study. The sulfur increased berry S and Mn, but decreased berry B significantly (Figure 6). Indeed, the bioavailability of Fe and Al may increase, but negatively related to berry B by adding elemental sulfur as treatment to lower soil pH [6] [48] . Total B demand depends on berry yield [49] and is also impacted by N additions [50] . Boron nutrition, if not addressed properly, could decrease fruit resistance to diseases [51] [52] . The indirect impact of S amendments on nutrient availability could be addressed in future studies in relation with boron in cell structure and integrity [53] .