Yield, fruit quality, and survival of 'Pêra' sweet orange on eight rootstocks in tropical cohesive soils

– The objective of this work was to determine the influence of eight rootstocks on the yield, fruit quality, and survival of 'Pêra CNPMF D6' sweet orange ( Citrus sinensis ) trees grown under rainfed conditions in a cohesive soil of the Brazilian Northeast. In 2014–2018, the yield, fruits, and survival of trees planted in 1997 were evaluated in a randomized complete block design. Yield was assessed using annual fruit production, yield efficiency, and the alternate bearing index. Tree survival was considered the percentage of plants that remained alive until 2018. Fruit quality was evaluated by physicochemical attributes. 'Cravo' confirmed its good yield performance. Despite the lower number of trees, the 'Mazoe' rough lemon rootstock induced the highest fruit yield. The 'Cravo' x 'Cleópatra' hybrid and 'Volkameriano' lemon favored the highest trees survival, whereas 'Indio' citrandarin and 'Cravo' lime resulted in intermediate levels. The 'San Diego' and 'Riverside' citrandarins induced shorter tree with a high yield efficiency and a higher alternate bearing index, respectively. Fruits with less juice and mass were induced, respectively, by 'Riverside' citrandarin and 'Volkameriano' lemon. It can be concluded that the 'Mazoe' rootstock induces a higher fruit yield and 'Cravo' x 'Cleópatra' and 'Volkameriano', a greater survival. In addition, all rootstocks, except 'Riverside' and 'Volkameriano', induce fruit that meet the minimum quality requirements.


Introduction
Brazil holds about 23% of the world's production of sweet oranges [Citrus sinensis (L.) Osbeck] and more than half of its juice production (FAO, 2020).Throughout the country, 'Pêra' sweet orange, with mid-season maturity, is the most frequently grown variety, as it is suitable for both juice processing and fresh consumption (Carvalho et al., 2019b), and the most propagated rootstock for it has been 'Cravo' lime (Citrus limonia Osbeck) (Carvalho et al., 2019b;Conceição et al., 2019).Of the Brazilian regions, the Southeast is the major citrus-producing one, covering 468,563 ha (71%), followed by the Northeast, with 108,527 ha (16%) and a fruit production of 1,154,225 tons (6%) (IBGE, 2020).
In the Northeast, sweet oranges are mainly cultivated by smallholders under rainfed conditions in cohesive soils that are highly susceptible to seasonal drought, which contributes to low yields (Martins et al., 2016).About 90% of the citrus orchards in Northeastern Brazil use the combination of 'Pêra CNPMF D6' sweet orange grafted onto 'Cravo' lime (Prudente et al., 2004;Amorim et al., 2018).This rootstock has excellent agronomical characteristics, including the induction of early production and drought tolerance, but is susceptible to diseases such as citrus decline or blight and citrus sudden death, which reduce fruit yields and shorten orchard longevity (Pompeu Junior & Blumer, 2019).For this reason, there is an increasing recognition of the risks posed by using a single rootstock for 'Pêra' sweet orange orchards (Sampaio et al., 2016;Amorim et al., 2018).
Despite the substantial body of knowledge on citrus rootstocks, there is relatively little information on their long-term survival (Sanderson & Treeby, 2014), particularly in tropical cohesive soils (Prudente et al., 2004;Carvalho et al., 2019a).However, it is known that tropical trees grow, on average, two times faster than trees from temperate biomes and live for a significantly shorter time (Locosselli et al., 2020).
The objective of this work was to determine the influence of eight rootstocks on the yield, fruit quality, and survival of 'Pêra CNPMF D6' sweet orange trees grown under rainfed conditions in a cohesive soil of the Brazilian Northeast.

Materials and Methods
The study evaluated the influence of eight citrus rootstocks on the horticultural performance of 'Pêra CNPMF D6' sweet oranges from 17 to 21 years after planting in a tropical cohesive soil of the Brazilian Northeast.
The experiment was conducted in a commercial orchard of the cultivar, planted in 1997, in the municipality of Rio Real, in the north of the state of Bahia, Brazil (11°29'7"S, 3756'4"W, at 170 m above sea level).The local climate is As according to Köppen-Geiger's classification, with a rainy season from April to September, followed by a dry summer.According to the historical series of 1998-2015, the average minimum, maximum, and median air temperatures for this municipality are, respectively, 20.70, 29.83, and 25.27°C, and the average annual rainfall is 1,208 mm (Embrapa, 2019).From planting in 1997, rainfall was recorded daily using a rain gauge installed in the orchard, showing an average of 1,033 mm per year.The driest years were 2004, 2008, 2012, 2016, and 2018, with 885, 907, 639, 942, and 779 mm, respectively, whereas the most humid were 1997 and 1999, with 1,530 and 1,415 mm.The soil is a sandy clay loam, classified as an Argissolo Amarelo distrófico (Santos et al., 2018), i.e., an Ultisol (Soil Survey Staff, 2014).
The trial was carried out in a randomized complete block design, with eight treatments (rootstocks), four replicates, and three trees per plot, spaced at 7.0x4.0m (357 trees per hectare), without irrigation.Management practices included chemical fertilization, pest and weed control, as well as pruning.A soil sample collected prior to planting indicated: 0.57 cmol c dm -3 Mg, 0.25 cmol c dm -3 K, and 0.18 cmol c dm -3 Na, cation exchange capacity of 4.33 cmol c dm -3 , and base saturation of 74.03%.Soil acidity was corrected every two years by the application of dolomitic limestone.The orchard was fertilized annually based on soil analysis, using 2,400 g per plant of the N-P 2 O 5 -K 2 O formula (20-0.5-12).In addition, foliar fertilizers were applied twice a year, supplying 8 kg Zn 200 L -1 , 6 kg Mg 200 L -1 , and 1 kg B 200 L -1 .Pests, such as the citrus black fly Aleurocanthus woglumi Ashby (Hemiptera: Auchenorrhyncha: Aleyrodidae) and the leaf miner Phyllocnistis citrella Stainton (Lepidoptera: Gracillariidae), were controlled by spraying registered pesticides four times a year (Agrofit, 2019).
The effects of rootstocks on plant development and productive performance began to be assessed in 2014 when the orchards were 17 years old.Tree height (m) and canopy volume (m 3 ) were measured in 2017 according to Carvalho et al. (2019a).Fruit from all plants were counted and weighed in each harvest.With the data of yield and planting density, annual fruit yield (Mg ha -1 ) was estimated for each cultivar.Yield performance was evaluated through fruit yield, mean fruit yield, and fruit yield efficiency (kg m -3 ); the latter was recorded only in 2017 and estimated by the quotient between the production of fruits per plant (kilogram per tree), as in Teodoro et al. (2020).In addition, the alternate bearing index (ABI) was calculated based on the fruit yield recorded in 2014 and 2018, following the methodology of Carvalho et al. (2020).Survival rate (%) was estimated in 2018 as the percentage of remaining plants in relation to initial planting density.
From 2014 to 2018, the mass of the fruits harvested annually was weighed and its mean value was estimated for each year.Fruit quality was evaluated using six fruits per plant harvested in 2015, the eighteenth year after planting, considering: fruit mass (g); juice content (%); titratable acidity (TA, gram of citric acid per 100 mL of juice), measured by titration with NaOH 0.1 N; total soluble solids (TSS, °Brix), estimated by a refractometer; and vitamin C (mg of ascorbic acid per liter), measured by redox titration using iodate solution and the maturity index or the TSS/TA ratio.All these attributes were determined according to methodologies described in Carvalho et al. (2020) and Teodoro et al. (2020).
The univariate analysis of variance, followed by Scott-Knott's test, at 5% probability, for grouping means was performed on all measured and estimated data.In addition, Fisher's discriminant analysis was also used to check whether or not the groups (rootstocks) differed from each other considering the set of all significant variables.The properties of each group were identified through these explanatory variables, which included plant height and canopy volume for tree size; average fruit yield, fruit yield efficiency, and ABI for production; and juice content and fruit mass for fruit quality.The following statistical packages were used: Real Statistics Resource Pack for the analysis of variance, DSAASTAT for Scott-Knott's test, and XLSTAT 2014 for the discriminant analysis, all add-ins for Excel.

Results and Discussion
The smallest sweet orange trees were induced by 'San Diego' citrandarin, and the greatest ones by 'Mazoe' rough lemon (Table 1).Similar results were reported by Carvalho et al. (2016) for the 'BRS Piemonte' mandarin grafted onto 'San Diego'.The smaller size of the plants grafted onto 'San Diego' favors management practices and harvesting.Moreover, trees with a small canopy volume and high productive efficiency allow obtaining a high yield by increasing planting density.
All rootstocks induced annual fruit yields surpassing 18,000 kg ha -1 for 'Pêra CNPMF D6' 17 to 21 years after planting (Table 2).These numbers suggest that there was a higher yield in the eighteenth harvest regardless of the rootstock, followed by reductions in the other harvest years.However, the annual and average yields obtained in the present study are in line with those found for sweet oranges by Carvalho et al. (2019a).
'Mazoe' rough lemon induced the highest 'Pêra CNPMF D6' fruit yield throughout the harvests, surpassing in 28% those obtained on 'Cravo' lime (Table 2).These results are attributed to the largest canopy volume and highest height of the trees induced by 'Mazoe' rough lemon (Table 1).These trees had a fruit yield efficiency similar to that on 'Cravo' lime, but lower than that on 'San Diego' (Table 2).The high fruit yield induced by 'Mazoe' rough lemon explains why some farmers in this region maintain this combination in their orchards despite great tree losses (Table 1).Likewise, Amorim et al. (2018) found a greater fruit yield in sweet oranges grafted onto Florida rough lemon than on 'Cravo' lime.Prudente et al. (2004) and Pompeu Junior & Blumer (2019) also observed a high fruit yield when 'Pêra' sweet orange was grafted onto rough lemon.
'Cravo' lime induced high yields, although lower than those obtained with 'Mazoe' rough lemon, combined with a good regularity of production throughout the harvests (Table 2).The outstanding performance of 'Pêra' grafted onto 'Cravo' lime has also been highlighted in other studies carried out in Northeastern Brazil (Carvalho et al., 2016(Carvalho et al., , 2019a;;Martins et al., 2016) and in the state of São Paulo (Pompeu Junior & Blumer, 2014).
From the seventeenth to the twenty-first year after planting, 'Indio' citrandarin induced relatively low fruit yields, in contrast to 'Cravo' lime and 'Mazoe' rough lemon (Table 2).These results differ from those reported by other authors for the same citrusproducing region in younger orchards (Carvalho et al., 2016;Sampaio et al., 2016. Amorim et al., 2018) and for the state of Acre, also in Brazil (Rodrigues et al., 2019).It should be pointed out that 'San Diego' citrandarin induced shorter trees, the lowest canopy volume (Table 1), and the highest fruit yield efficiency (Table 2).According to Schinor et al. (2013), hybrids involving the 'Cleópatra' and 'Sunki' mandarins, such as citrandarins, are good alternatives to 'Cravo' lime.
The present study showed variations in fruit yield throughout harvests.The highest yield was recorded in 2015 and the lowest in 2018 (Table 2).In short, low yields were observed after high production harvests.Girardi et al. (2017) andSilveira et al. (2020) reported similar variations in fruit production and suggested alternate bearing as one of the possible explanations.In this study, the rootstocks induced changes on the degree of alternate bearing, as indicated by the ABI, which ranged from 0.13 to 0.29.Trees with a higher ABI also presented a lower fruit yield.The Table 1.Tree height (TH), canopy volume (CV), and survival rate (SR) of 'Pêra CNPMF D6' sweet orange (Citrus sinensis) grafted onto eight rootstocks in the municipality of Rio Real, in the state of Bahia, Brazil (1) .
No noticeable disease symptoms were observed from the seventeenth to the twenty-first years after planting.Although the causes of tree mortality were not identified in this study, some researchers have reported symptoms of incompatibility between 'Pêra' sweet orange and rough lemons, including 'Mazoe' (Salibe et al., 2002). However, Mourão Filho et al. (1991), Prudente et al. (2004), andPompeu Junior &Blumer (2019) found no symptoms of graft incompatibility between the 'Pêra' sweet orange scion and the rough lemon rootstock.
The symptoms leaf brightness loss, chlorosis, and leaf fall, including stem dieback and tree death, have usually been verified in sweet orange trees grown in the Brazilian Northeast.It is well known that the 'Cravo' lime and 'Volkameriano' rough lemon rootstocks are highly susceptible to citrus decline, while the 'Cleópatra' and 'Sunki' mandarins would be asymptomatic (Mourão Filho et al., 1991).The high survival of trees grafted onto 'Volkameriano' might be related to a relatively good resistance to Phytophthora root rot, a common disease in the Brazilian Northeast (Salibe & Cereda, 1984).
Table 2. Annual and mean fruit yield (FY), from the seventeenth to the twenty-first harvests (2014-2018), fruit yield efficiency (FYE), and alternate bearing index (ABI) for 'Pêra CNPMF D6' sweet orange (Citrus sinensis) grafted onto eight rootstocks in the municipality of Rio Real, in the state of Bahia, Brazil (1) .

Conclusions 1 .
The 'Mazoe' rough lemon (Citrus jambhiri) rootstock induces the highest fruit yield for 'Pêra CNPMF D6' sweet orange (Citrus sinensis), even 21 years after planted under rainfed conditions in a cohesive soil of the Brazilian Northeast.

Figure 1 .
Figure 1.Graphical representation of Fisher's discriminant analysis of eight rootstocks onto which 'Pêra CNPMF D6' sweet orange (Citrus sinensis) was grafted in the municipality of Rio Real, in the state of Bahia, Brazil.A, observation cloud with confidence circles for the groups in the space of the two main factors (F1 and F2); and B, correlation circle obtained by Fisher's discriminant analysis.Red arrows, variables mostly correlated with Factor 1; blue arrow, variable mostly correlated with Factor 2; and black arrows, variables associated with other factors.