Root system distribution and vegetative characteristics of Prata type bananas under different irrigation strategies

This study aimed to analyze the root system distribution and vegetative characteristics (pseudo-stem perimeter, plant height, number of leaves and the leaf area) in ‘Dwarf-Prata’ and ‘BRS-Platina’ in their third production cycles under different five irrigation strategies. The vegetative characteristics were measured at the flowering stage and the root sampling was done after the harvesting in the third production cycle. The irrigation depths (ID) were obtained by the model ID= K × AF × ETo, where K is an empirical transpiration coefficient, AF is the leaf area of Dwarf-Prata’ plants and ETo is the reference evapotranspiration. Irrigation strategy 5 was based on the crop evapotranspiration, ETc = ETo × Kc, where Kc is the crop coefficient. Drip irrigation was used, with two laterals per plant row and emitters with flow rate of 8 L h -1 , which were spaced out at 0.5 m, totaling 10 emitters per plant. The irrigation strategies based on crop evapotranspiration and on the model ID = K × AF × ETo, with K ranging from 0.2 to 0.65, exhibited similar values for vegetative characteristics, as well as for the root length density (RLD) in Prata type banana; however, higher RLD is found in deeper layers when using a lower K coefficient. The ‘Dwarf-Prata’ displays taller plants, longer pseudo-stem perimeter, higher number of leaves and larger leaf area than the ‘BRS-Platina’, although both exhibit similar root distribution.


INTRODUCTION
The Brazilian semiarid exhibits high productive potential for banana production.However, shortage of rainfall and its inconsistency limit the banana production.This makes irrigation necessary if banana production is to be successful.The water shortage is a universal phenomenon and represents a great challenge for producing bananas (Ravi et al., 2013).Therefore, the irrigation efficiency should be increased.This justifies carrying out studies that deal with local specificities by involving strategies that allow the farmer to manage the irrigation to increase water productivity.
Author(s) agree that this article remain permanently open access under the terms of the Creative Commons Attribution License 4.0 International License based on the crop evapotranspiration (ETc), which is the product of the reference evapotranspiration (ETo) and crop coefficient (Kc).Nevertheless, the Kc of a given crop is not available in literature for every condition.
Accordingly, the use of strategies that base on the local climate, crop development and crop physiology might be more accurate.In this context, the volume of applied water per plant (ID) can be determined by the product of the ETo, leaf area (LA) and a coefficient (K) to adjust the model.This improves the water use efficiency of a given crop, since the LA refers to the crop and ETo is local.The use of different irrigation strategies causes changes in the soil water condition, which when associated with the climate, may influence the plant water status, resulting in changes in crop growth and development.The knowledge of root system distribution and shoot growth in bananas allows a better understanding of how to arrange the irrigation system for an increase in water use efficiency.
Plant height is an important vegetative characteristic from a crop production standpoint, as it interferes with plant spacing and density and ultimately impacts on yield.A negative feature of taller plants is the higher vulnerability to breaking the pseudo-stem or toppling of the plant due to strong winds (Nomura et al., 2013).Conversely, larger pseudo-stem perimeter gives the plant higher capacity of sustaining the mass of the bunch and exhibits higher resistance to breaking of the pseudo-stem and toppling, as well as showing higher vigor.
Banana plants reproduce asexually by shooting suckers (daughter) from an underground stem.Harvesting the mother plants defines the transition from a production cycle to another, as daughters replace the mothers in the following cycle as the main plant.An adequate number of leaves during the flowering stage is important for the development of daughter plants and the bunch (Rodrigues et al., 2009;Rodriguez et al., 2012) since there is no sprouting of leaves after flowering (Donato et al., 2015).Leaf area is important because it is related to the surface responsible for transpiration and assimilation of carbon dioxide (CO 2 ) (Turner et al., 2007).
Knowing the highest root length density (RLD) in the soil profile permits the development of irrigation projects and indicates, under different irrigation system arrangements, the correct positioning of moisture sensors in the soil (Sant'ana et al., 2012) and where to fertilize.This results in an increase in yield in the banana grove (Borges et al., 2011).According to Azevedo et al. (2008), the RLD is a fundamental parameter to determine the potential of crops to absorb water and nutrients.
The distribution of RLD is influenced by the water content in the soil as a result of several factors, such as the adopted irrigation management practice (Santos et al., 2014), the irrigation system (Sant'ana et al., 2012) and the chemical, physical and microbiological conditions in the soil (Segura et al., 2015).These factors are associated with the growth, development and production of banana plants.In literature, there is still need for information about the root system distribution of Prata type bananas under different irrigation strategies.Therefore, this study aimed to evaluate the root system distribution and vegetative characteristics (pseudo-stem perimeter, plant height, number of leaves and the leaf area of 'Dwarf-Prata' and 'BRS-Platina' bananas grown under different irrigation strategies.

MATERIALS AND METHODS
The study was conducted in a banana plantation that was planted in March, 2012, at 2.5 × 3.0 m spacing, in a medium-textured Red-Yellow Latosol (Hapludox).Its physical characteristics are shown in Table 1.The area is located in the Irrigated District of Ceraíma, municipality of Guanambi, state of Bahia, Brazil, 14° 13' 30" S, 42° 46' 53" W and altitude of 545 m.The average annual rainfall and temperature of the study area are 680 mm and 25.78°C, respectively.
The following variables were assessed in the third productive cycle of Dwarf Prata (AAB) and BRS Platina (AAAB) cultivars subjected to the different treatments of irrigation strategies: root distribution and vegetative characteristics of the shoot, plant height, pseudo-stem perimeter at 0.30 m in height, leaf area and number of leaves.
Five irrigation strategies were used.Four strategies are based on leaf area and an empirical transpiration coefficient -K (Coelho Filho et al., 2004;Oliveira et al., 2009) following the model: where ID is the volume of applied water (L plant -1 ), K is a coefficient of 0.20, 0.35, 0.50, and 0.65 for the strategies S1, S2, S3 and S4, respectively; LA is the leaf area in m 2 of Dwarf-Prata, the most planted cultivar, measured every fifteen days during the cycle.ETo is the reference evapotranspiration in mm, daily determined during the cycle, as shown in Figure 1.
The choice of the coefficients 0.20, 0.35, 0.50 and 0.65 for the strategies based on the leaf area considered the work of Oliveira et al. (2013), who, by evaluating the growth of 'Grand Nain' using the same model with coefficients varying from 0.0 to 0.8, verified that 0.57 displayed the best performance.
To determine the leaf area, the mother and daughter of 'Dwarf-Prata' banana were used, as it is the most planted cultivar in Brazil.The leaf area (LA), in m 2 , was estimated every fifteen days from the  length and width measurements of the third leaf and from the total number of leaves in the plant, as Oliveira et al. (2013).
The strategy 5 (full irrigation) is based on the crop evapotranspiration, ETc, obtained by using Penman Monteith, FAO 56 and Kc, crop coefficient, of banana plant for the northern region of the Brazilian state of Minas Gerais (Borges et al., 2011) (Figure 1).Irrigation was performed daily by drip irrigation, with two lateral lines per row of plants and pressure compensating emitters with flow rate of 8 L h -1 spaced every 0.50 m, totaling 10 emitters per plant.The crop evapotranspiration (ETc), in mm, for managing the irrigation strategy 5, was calculated by the product of the ETo and crop coefficient.Since drip irrigation is a localized irrigation system, ETc was adjusted by the location coefficient (Kl) (Bernardo et al., 2006).
where, P is either the percentage of wetted area or shaded area, whichever is higher.As the assessments were done in the third cycle, the leaf area of banana plants covers 100% of the surface, resulting in a unitary location coefficient.This condition is reached after 275 days after planting, DAP (Figure 1).The cumulative gross irrigation depth in the different strategies and the occurrence of rain over the four cycles are depicted in Figure 2. In the first cycle, all plants from every strategy were fully irrigated until the 144 th day after planting.It can be seen in Figure 2 that the duration of the second cycle is shorter than the remaining cycles.The harvest was used to separate one cycle from the other; however, when harvesting in the first cycle, the daughter plant was already under development, in the bunch emergence phase, which justifies the shorter duration of this cycle.It should also be noted that there was a small occurrence of rain at the end of the second cycle because it matched with the dry season (drought) of the region.
The evaluation of vegetative characteristics was performed at flowering stage of the third cycle of 'Dwarf-Prata' and 'BRS-Platina' banana plants for all the irrigation strategies, with three replicates and six plants per experimental unit (Figure 3).
After harvesting in the third cycle, roots of the plants were sampled from the different treatments to assess the root system distribution.For each irrigation strategy, roots of three plants were collected longitudinally and perpendicularly to the laterals lines, in which, for each plant, twenty samples were collected, with five distances from the pseudo-stem: 0.15, 0.40, 0.65, 0.90 and 1.20 m and four depths: 0.00 to 0.20; 0.20 to 0.40; 0.40 to 0.60; and 0.60 to 0.80 m.Each sample was pushed out by a cylindrical core sampler with 20 cm in height, corresponding to 623.45 cm 3 in volume (Vr).
After removing the roots from the soil, the samples were placed in plastic bags and taken to a laboratory where the roots were separated from soil by washing with water.Once separated, the roots from each position of the soil profile were scanned and converted into Tagged Image File Format (TIFF).These images were analyzed using Adobe Photoshop to clean dark edges caused by the scanning process and submitted to the application Rootedge (Kaspar and Ewing, 1997) for determining geometric characteristics: length and diameter of the roots.The root length, Lr (cm) was used to determine the root density length, RDL (cm cm -3 ) in a sample volume Vr (cm 3 ) by: RDL = Lr Vr -1 (3) Root density length was analyzed considering all roots per treatment, very fine roots (diameter below 0.55 mm), fine roots (diameter between 0.55 and 2.00 mm), small roots (diameter between 2.00 and 5.00 mm) and medium to very large roots (diameter above 5.05 mm) as described by Santos et al. (2014).As for this study, only very fine roots and fine roots were observed.
Concerning the data for vegetative characteristics, two cultivars and five irrigation strategies were arranged in a 2 × 5 factorial experiment in completely randomized design.Conversely, for the analysis of root distribution, two factorial experiments were used: (a) 2 × 2 × 5, two cultivars, two sampling direction and five irrigation strategies in a completely randomized design; (b) 5 × 4 × 5, five irrigation strategies, four sampling depths and five distances from the plant in a completely randomized design.The data for vegetative characteristics and very fine, fine and total root density length (RDL) were subjected to analysis of variance and the interactions were deployed according to their significance.The means of these variables were compared to one another using the Tukey's test for the irrigation strategies, cultivars and sampling direction factors.As for distances and depths factors, regression was used instead.

RESULTS AND DISCUSSION
According to the analysis of variance, the vegetative characteristics varied only for the cultivars (p<0.05),regardless of the irrigation strategy (Table 2).The 'Dwarf-Prata', despite being parent of 'BRS-Platina', exhibited greater plant height, pseudo-stem perimeter, number of leaves during the flowering stage and leaf area.The average height of the plants was 3.67 and 3.52 m; the pseudo-stem perimeter, 1.11 and 1.01 m; number of leaves for the mother plant, 20.30 and 16.28; number of leaves for the daughter plant, 4.11 and 3.41; total leaf Plant height in m (PHT), plant perimeter in m (PSP), total leaf area in m 2 (TLA), number of leaves in the mother plant (NLM), number of leaves in the daughter plant (NLD), source of variation (SV), degrees of freedom (DF), mean square (MS), coefficient of variation (CV), F value (Fc) and P value (Pr).
area, 22.98 and 17.58 m 2 for the 'Dwarf-Prata' and 'BRS-Platina', respectively.Marques et al. (2011), who evaluated the vegetative characteristics of these two cultivars, recorded in the first and second production cycles higher values of pseudo-stem diameter and leaf area for the 'Dwarf-Prata' as well.The same authors found out higher number of leaves at flowering stage for 'Dwarf-Prata' in comparison with 'BRS-Platina', which is consistent with the results of this study.Therefore, the parent (Dwarf-Prata) exhibits higher number of leaves than the progeny (BRS-Platina).On the other hand, Donato et al. (2009) noticed similarities in plant height, pseudo-stem perimeter and number of leaves in the first and second production cycles, between the 'Dwarf-Prata' and 'BRS-Platina'.Nonetheless, the study of Donato et al. (2009) was carried out in a region with incidence of Yellow Sigatoka, leaf disease to which 'Dwarf-Prata' is susceptible and the 'BRS-Platina' is resistant, which contributes to the decrease in quantity of functional leaves and vigor in the parent.
From the analysis of variance, the root length density (RLD) of fine roots was influenced by the double interaction between the distance and the depth, and the very fine and total RLD varied in an independent way for distance and depth (Table 3).
There were no differences in RLD between the cultivars, which may be due to the fact that 'Dwarf-Prata' and 'BRS-Platina' are parent and progeny, in spite of the differences in the shoot.Likewise, there were no differences in RLD between the longitudinal and perpendicular directions to the laterals, which can be explained by the wet bulb created by two irrigation laterals, one of each side, favoring the moisture to reach higher distances perpendicularly to the plant row and, consequently, the root development.
Although there are no differences within the means of RLD among the irrigation strategies, it can be seen in Figure 4 different root system distribution in the soil profile.The root system distribution of Prata banana plants when subjected to lower irrigation depth, obtained by the model ID = 0.2 × ETo × LA, results in higher expansion of roots in greater depths (Figure 4 1A).These results may imply that there is a tendency to increase the RLD when the plant is subjected to lower water availability, creating a partial water-deficit, and as is discussed by Taiz and Zeiger (2013), in water-deficit condition, there may be a greater investment in roots and reduction in leaf area as a result of changes in the ratio shoot/root and preferential drain, which depends on the intensity and length of the drought season.In this study, there was no reduction in leaf area in relation to the irrigation strategy for any cultivar, though there was a change in the ratio shoot/root through the greater root expansion; therefore, indicating a higher sensitivity to root development regarding the depletion of water in the soil.
The root length density in the strategy S4 (K = 0.65) is higher up to 0.50 m deep, reaching values above 0.045 cm cm -3 of roots up to the distance of 1.20 m.The application of higher volume of water in this strategy, perhaps, resulted in greater moisture distribution in the superficial layer, which favored the development of roots in these layers, The root system growth towards deeper layers associated to irrigation deficit (K = 0.20) (Figure 4 1A) could be attributed to physiologic mechanisms to cope with abiotic stress, involving investment in roots under lower availability of water.However, the greater root deepening in water-deficit condition is more related to the survival of the plant than its yield (Pereira, 2011).The observed results with K = 0.65 agree with San'Ana et al. ( 2012) who observed predominance of roots near Source of variation (SV), degrees of Freedom (DF), mean square (MS), coefficient of variation (CV), F value (Fc) and P value (Pr).
the soil surface, as 80% were concentrated at 0.61 m deep for the 'Dwarf-Prata' cultivar irrigated by dripping, with one lateral line per plant row and irrigation management based on the crop coefficient Kc, closer to that coefficient.Similar results were also found by Santos et al. (2014), who, by studying the root system distribution of 'Tommy Atkins' mango tree under regulated deficit irrigation, found out a tendency to increase the root length density with partial deficit of 50% of the ETc when the deficit is applied in the stage when the water requirement of the crop is the highest.The results also agree with those found by Boni et al. (2008) who verified higher root distribution of cashew tree, both vertically and horizontally in the soil for the conditions with no irrigation, in comparison with irrigated cashew trees.These authors point out that the search for water is more intense by smaller roots in conditions of reduced availability of water in the soil.It is widely accepted in the literature that the plant roots under moderate water deficit grow more than those that receive water adequately (Kramer and Boyer, 1995), as it was observed by He et al. (2014) who reported that the RLD in rice at 0.2 to 0.6 m soil layer was approximately 2.5 to 5 times higher under water-limiting condition than flooding irrigation.Santos et al. (2014) mention that this behavior is explained by the greater allocation of photo-assimilates to roots, allowing the absorption of water in deeper soil layers, and by the availability of water in the soil, sufficient to maintain the turgor and root growth.
The roots of 'Dwarf-Prata' and 'BRS-Platina' banana plants decrease as the depth in the soil and the distance from the pseudo-stem increase (Figure 5).Considering the very fine roots and all roots (very fine and fine), there is a linear decrease as it becomes deeper (Figure 5A) regardless of the distance from the pseudo-stem, and there is a reduction as the distance from the pseudo-stem increases, regardless of the depth.It can be explained by modeling a quadratic function, with lower RLD at 1.0 m.Since the irrigation was performed by dripping, making a continuous wet strip, it possibly favored the development of roots in the longitudinal direction to the wet strip, occurring overlapping roots after 1.0 m from the pseudo-stem between two plants in the row, thus explaining the quadratic behavior of the model for estimating the RLD as a function of the distance from the pseudo-stem.
It is noted, still, in Figure 5, by the distribution of RLD, the predominance of very fine roots in the whole profile, while the fine roots exhibit greater concentrations near the pseudo-stem.Sant'ana et al. (2012) observed predominance of very fine roots in the whole root zone, in which the zones with higher root length density, up to 0.40 m deep, were the regions with higher water extraction for plants watered by drip, micro-sprinkler and conventional sprinkler irrigation systems.As for the very fine roots, which are those more related to the absorption of water and nutrients, even with no differences among the treatments in RLD, the influence of the irrigation strategies on the root distribution is observed, in which, in the irrigation strategy 1 (K = 0.20) and based on ETc, the highest RLD is located between 0.30 and 0.70 m in depth, and, in the irrigation strategies 2, 3 and 4 (K of 0.35, 0.50 and 0.65, respectively), it is located between 0.10 and 0.60 m.In these cases, these depths are indicated for the installation of moisture sensors.
Concerning the fine roots (diameters between 0.50 and 2.00 mm), there was interaction between the distance from the pseudo-stem and the RLD, which is adjusted by the response surface (Figure 5C).It is observed that as the distance from the pseudo-stem and depth increase, there is a reduction in RLD, indicating that these roots are those more related to the plant anchoring and they are located more closely to the pseudo-stem.As it can also be seen in the Figure 4 1B, 2B, 3B, 4B, 5B and in the Figure 5A and 5B, the increase in depth and distance from the pseudo-stem, the estimative model tends to approach the values of very fine and total RLD.The percentage of cumulative root length in the longitudinal and perpendicular direction to the lateral line as function of the depth and the distance from the pseudo-stem is depicted in Figure 6.Regardless of the cultivar and irrigation strategy, 80% of roots are concentrated up to 0.51 m in depth, both in the longitudinal and perpendicular direction.On the other hand, by considering the distance from the pseudo-stem, 80% of all roots are concentrated up to 0.87 m and 0.84 m in the longitudinal and perpendicular direction, respectively.Sant'ana et al. (2012), who evaluated the distribution of the root system of 'Dwarf-Prata' banana plants watered by different irrigation systems, verified that, under dripping at the end of the second cycle, there was predominance of roots near the soil surface, with 80% at 0.61 m in depth and at 0.63 m from the pseudostem of the plant.This is a common characteristic in the root system's design of banana plants irrigated by dripping, as its wet bulb limits the root distribution up to about 0.50 m from the pseudo-stem (Sant'ana et al., 2012).As Pereira (2011) discusses, localized irrigation that creates wet bulb in the soil induces the concentration of roots around points where water is applied.By watering the papaya plants with drip irrigation, Coelho et al. (2005) verified that 80% of roots were concentrated in the first 0.45 m in depth, while Lopes et al. (2014) reported that 90% of the roots of peach-palm watered by drip irrigation are found at depths from 0 to 0.30 m, indicating that 0.3 m is the effective depth of the root system for irrigation purposes.

Conclusions
The irrigation strategies based on the crop evapotranspiration and based on the model ID = K × LA × ETo, with K varying from 0.20 to 0.65, allow the exhibition of similar values for vegetative characteristics, as well as the root length density (RLD) in Prata type banana plants; however, soil profiles with higher RLD are created in deeper layer when lower values of K are used.The 'Dwarf-Prata' exhibits taller plants, longer pseudostem perimeter, higher number of leaves and larger leaf area than 'BRS-Platina', despite both exhibiting similar root distribution.
In the third production cycle, the application of fertilizers to the soil can be performed within a radius of 0.84 cm from the pseudo-stem and the installation of moisture sensors should be done within the same radius, up to 0.50 m in depth.

Figure 5 .
Figure 5. Root length density (RLD) when considering all diameters and diameters lower than 0.5 mm in relation to the depth (A) and in relation to the distance from the pseudo-stem (B), and in relation to the depth and distance from the pseudo-stem when considering roots with diameter between 0.5 and 2.0 mm (C).

Figure 6 .
Figure 6.Percentage of cumulative root length in the longitudinal direction to the lateral line for depth (A) and for distance from the pseudo-stem (B) and perpendicularly to the lateral line for depth (C) and for distance from the pseudo-stem (D).

Table 1 .
Physical characteristics of the soil in the study area.

Table 2 .
Analysis of variance of the arrangement of the 2 × 5 factorial experiment, with two cultivars and five irrigation strategies.

Table 3 .
Analysis of variance of the arrangement in 2 × 2 × 5 factorial, for two cultivars (VAR), two directions (DIREC) and five irrigation strategies (IS) in a 5 × 5 × 4 factorial experiment, for five irrigation strategies, five distances (DIST) from the plant and four depths.