Management of trickle irrigation for banana: Hydrodynamic processes and sensor placement at the root zone1

Information on soil hydrodynamic processes assists in explaining the soil-water-plant relationship and has practical applications to irrigation management, such as the definition of soil water sensor placement. The objective of this study was to detail the hydrodynamic process in the soil root zone and to define the location for placement of soil water sensor under different configurations of trickle irrigation in banana crops. Three microsprinkler emitters with flow rates of 70 (T1), 53 (T2), 35 L h-1 (T3), and two drip system, one with one drip line per row of plants (T4), and another with two drip lines per row of plants (T5) were evaluated. The experiment was conducted in a randomized block design with five repetitions. Higher water extraction was found for irrigation systems with higher flow rates for all configurations of trickle irrigation systems. Soil moisture sensors in drip systems should be placed at distances of 0.75 to 0.81 m from the pseudo stem and at depths of 0.33 to 0.44 m. Under micro-sprinkler systems, soil water sensors should be placed at 0.75, 0.77 and 0.83 m from the pseudo stem towards to the emitter and at depths of 0.33, 0.48 and 0.55 m for emitter flow rates of 35, 53 and 70 L h-1, respectively.


Introduction
Water is important for agricultural production and the rational management of water is decisive for plant development and sustainability of water resources. Geronimo et al. (2015) pointed out that an inefficient use of the available water resources may result in water shortage in the future. Information on water distribution in the soil is important and directly affects crop yields (Souza & Matsura, 2004). Thus, it is fundamental for irrigation design and management, including the determination of number of drippers per plant and their position in relation to the plant or plant row (Asawa, 2014;Lopes et al., 2009).
Water flow and distribution in irrigated soils differ in the different irrigation systems. Levien (2014) reported that drip irrigation differs from other systems in water flow because of the formation of a wet bulb below each emitter or, when emitters are arranged in rows, a continuous wet strip. Assessments of water distribution in the soil and definitions of root-zones have been carried out using time-domain reflectometry (TDR) (Santos & Martinez, 2013;Silva et al., 2015b;Santos et al., 2017).
Information about positioning of sensors is scarce and relies on root distribution studies (Bassoi et al., 2003;Sant' Ana et al., 2012). The effective root depth and distance (Sant' Ana et al., 2012) combined with the effective root-extraction depth and distance have been recommended as limits for sensor placement in irrigation systems (Santos et al., 2005;Coelho et al., 2007). However, root uptake varies with time and space in the effective root zone (Silva et al., 2018). Thus, the effective root zone for sensor placement also depends upon the crop type, soil type, irrigation system, and age of plants (Coelho et al., 2007). The objective of this study was to detail the hydrodynamic process in the root zone and define the location for placement of soil moisture sensor under different configurations of trickle irrigation in banana crops.

Material and Methods
The study was carried out in 2011 at the Experimental Station of Gorutuba, in northern Minas Gerais State, Brazil (15° 47' S, 43° 18' W, and 516 m of altitude), which belongs to the Empresa de Pesquisa Agropecuária de Minas Gerais (EPAMIG). The region presents a BSwh, hot semi-arid climate with dry winter, according to the Köppen classification (Alvares et al., 2013). The rainfall depth in 2011 was 695 mm, but only March and October presented more than 100 mm per month; the rainfall depth was 10.3 mm during the data collection period (June to September). The soil of the area was classified as a Typic Hapludox; the soil physical and hydraulic attributes is presented in Table 1. The soil water retention curve parameters were: α = 0.6826; n = 1.3345; m = 0.2569, saturation water content = 0.3200 m 3 m -3 , and residual water content = 0.2100 m 3 m -3 .
The study was conducted using banana crops of the Prata cultivar grown in single plant rows, with spacing of 3.0 × 2.5 m. The experiment was carried out in a randomized block design, with five repetitions and five treatments: T1 -Micro-sprinkler; 70 L h -1 and one emitter for four plants located in one lateral line between two plant rows; T2 -Micro-sprinkler; 53 L h -1 and one emitter for four plants located in one lateral line between two plant rows; T3 -Micro-sprinkler; 35 L h -1 and one emitter for four plants located in one lateral line between two plant rows; T4 -Dripper; 4 L h -1 and one lateral line per plant row with on-line emitters spaced 0.7 m apart; and T5 -Dripper; 4 L h -1 and two lateral lines per plant row with on-line emitters spaced 0.7 m apart. Irrigation water depths were the same in all treatments, performed on a daily basis, and calculated from daily evaporation data measured by using a Class A pan (EVA), taking into account the pan correction factor (Bernardo et al., 2006), location coefficient (Kl) (Bernardo et al., 2006), and crop coefficient (Kc) (Coelho et al., 2004). The total calculated crop evapotranspiration during the year was 1038 and 593 mm from June to September.
Trenches (1.0 m deep and 1.0 m wide) were opened from the pseudo-stem along the plant row for the drip-irrigation treatments, and from the plant towards the emitter for the micro-sprinkler irrigation treatments. Soil water content (θ) in the root zone of the banana plants were measured at the flowering and beginning of fruit formation stages from June to September. TDR (time-domain reflectometry) probes, containing three 0.1 m long rods (Silva et al., 2015a), which were previously calibrated (Soncela et al., 2013;Batista et al., 2016), were inserted into each trench, at distances of 0.25, 0.50, 0.75, and 1.00 m from the pseudo-stem (r) and at depths of 0.20, 0.40, 0.60, and 1.00 m (z). Each trench wall was considered a two-dimensional plane for the purpose of measuring the hydrodynamic processes.
Soil moisture content (θ) at each distance from the pseudostem and soil depth (r i ; z i ) was measured with TDR by using multiplexers and continuously and automatically recorded in a datalogger every 15 min.
The limited number of multiplexers and TDR allowed evaluating the hydrodynamic processes in only one trench of each treatment. Infiltrated water (IW), water extraction (WE), and percolating water (PW) were determined for each treatment by using soil moisture data measured over irrigation cycles. IW (Eq. 1) was determined by the difference between the amount of water stored in the root zone limited by 1.0 m distance from plant (r = 1.0 m) and by 1.0 m soil depth (z = 1.0 m) at the time t+1 and t; t refers to the time immediately after an irrigation, and t+1 refers to the time the irrigation water Table 1. Soil physical characteristics of the experimental area reached the zone limited by r = 0.25 m and z = 0.60 m. WE (Eq. 2) was quantified by the difference between the amount of water stored in the root zone at time t+1 and at t+2 (time immediately after the subsequent irrigation).
This study considered the suitable location in the root zone for placing soil moisture sensors as the intersection between the zones of effective root distribution and effective water extraction. The effective rooting zone was assumed as that with at least 80% of total root length found in the trench. The effective water uptake zone was assumed as that with at least 80% of total water extraction in the trench (Santos et al., 2005).
The average soil moisture content data collected on the bi dimensional plane of each treatment from June to September was used to evaluate the infiltrated water (IW), extracted water (WE), percolated water, and available water. Analysis of variance was performed for root data, focusing on root distribution.

Results and Discussion
The use of one-drip lateral lines per plant row resulted in higher water depths than two lateral lines per plant row. This larger water depth is a result of a smaller wetted area under the soil surface when using one lateral line, considering that the same volume of water was applied using two lateral lines per plant row. Available water contents were above the upper limit, which corresponds to the field capacity of the soil ( Figure  1). The highest water availability in the soil irrigated by onedrip lateral line is due to the area and volume watered by the drippers, which were spaced 0.70 m apart. The use of two lateral lines resulted in a greater total wetted area per plant. A single lateral line per plant row wetted an area that corresponded to half the area and soil volume watered when using two lateral lines with drippers under the same conditions and applying the same amount of water. Micro-sprinklers provided larger wetted areas than drippers (Figure 1).
Both irrigation systems (micro-sprinkler and drip) showed losses by percolation (Figure 2). The losses found for the one-drip lateral line per plant row were higher than those found for two-drip lateral lines per plant row. The microsprinkler irrigation showed lower percolation losses for flow rates of 35 and 53 L h -1 than for the rate of 70 L h -1 . Water extraction from the system (WE), which includes percolation, evaporation, and water uptake by roots, were affected by the irrigation system and its configurations, i.e., flow rates of emitters and number of lateral lines (drip system). The highest water extraction rates occurred under higher flow rates of micro-sprinklers and drip irrigation with one lateral line per plant row (Figure 2). Micro-sprinklers applied lower where: θ t (r i , z i ) -soil moisture content at (r i , z i ) and at time t, m 3 m -3 ; θ t+1 (r i , z i ) -soil moisture content at (r i , z i ) and at time t+1, m 3 m -3 ; and, θ t+2 (r i , z i ) -moisture content at (r, z) and at time t+2, m 3 m -3 .
The integrals were solved numerically by using the trapezium rule.
Water extraction (WE) included soil evaporation, deep percolation (PW), and root extraction. Deep percolation was assumed as the water in the soil layer below the effective rooting depth between two irrigation events. Effective rooting depth was assumed as that with 80% of total root length (Kanber et al., 1996), which was 0.40 m based upon the results of Sant' Ana et al. (2012). Thus, deep percolation was assumed as the difference between water content stored in the soil layer limited by depths 0.40 and 1.00 m at time t (immediately before an irrigation event) and at time t+2 (time before the next irrigation event). The water stored (WS) in the layer at any time (t i ) limited by depths 0.40 and 1.00 m was given by Eq. 3.
Root distribution was evaluated by using the monolith method, using monoliths of 0.1 × 0.1 × 0.1 m (Bohm, 1979). The samples were collected at the distances of 0.25, 0.50, 0.75, and 1.0 m, and depths of 0.1, 0.2, 0.4, 0.6, and 0.8 m. The roots were separated from the soil by washing and digitalized as TIFF (Tagged Image File Format). The files were then processed in the software Rootedge to determine total root length at each position (r i ,z i ). The individual root length data were processed to obtain the effective root distance and depth according to Sant' Ana et al. (2012). Root distribution in depth and distance from the plant was evaluated in a randomized block statistical design using a split-split plot arrangement, with five treatments, five depths, and four distances from the plant.
Water extraction was related to soil water availability (AW) that was obtained in each point (ri, zi) of the trench, using the soil moisture content θ (ri, zi) (Eq. 4). water depths than drippers due to their larger wetted area and unevenly water distribution. The one-drip line waters a smaller area than two lateral lines, which results in a higher water depth. According to Silva et al. (2009), increasing the number of the drip emitters in lateral lines results in larger root zone distribution, larger water extraction, smaller losses by percolation flow, and higher water application efficiency in banana crops. This difference is even greater when compared with micro-sprinklers.
Regions of higher water extraction in the root zone of the banana plants overlapped zones in the soil where water was more readily available (Figure 1). This result was also reported by Silva et al. (2009) andCarvalho (2011) and confirmed what was expected, since zones with the highest water availability are those in which plants can uptake water more easily, as long as water-uptake roots are present. Additionally, Silva et al. (2013) concluded that available soil water distribution in the root zone affects the water extraction rate by the plant, deep percolation, and irrigation efficiency when the same volume of water is applied by micro-sprinklers with different flow rates on the lateral line. After irrigation, treatment 1 exhibited average available water content of 57.20%, at the distance of r = 0.25 m. The highest water extraction rates occurred up to the depth of 0.4 m, where water is more readily available (Figure 1), and the effective rooting depth occurred up to 0.36 m ( Figure 3A and Figure 4A).
The remaining treatments exhibited higher soil available water contents, water extraction at depths higher than 0.40 m, and effective rooting depths, which were 0.48, 0.55, 0.33, and 0.44 m, for treatments T2, T3, T4, and T5, respectively ( Figure  3 and Figure 4A).
Available soil water distribution affects water extraction zones, as seen in treatment T1, whose average available water content at the distance of 0.25 m from the plant after irrigation was 57.20% and the percentage of total extracted water from the effective rooting zone was 23.37%. Regarding the distance from the plant of 0.50 m, where the average available water content increased 269%, the percentage of total extracted water from the effective rooting zone was 37.41%.
Effective rooting depths ( Figure 3) and distances from the plant matched the effective water extraction zone (Figure 1), as also reported by Carvalho (2011). Eighty percent of roots in treatment T1 were up to 0.36 m deep and 0.75 m away from the plant, whereas 80% of the total water extraction occurred up to 0.33 m deep and 0.77 m away from the pseudo-stem, denoting that most absorbing roots were at the effective depth of water extraction. The effective depth of water extraction in treatment T5 was 0.60 m from the soil surface, which did not match the effective rooting depth measured at 0.44 m. In this case, the effective distance of water extraction, was found at 0.86 m from the pseudo-stem of the banana trees, which somewhat relates to the effective distance of the root system, 0.81 m.
The optimal soil moisture sensor positioning was determined by the intersection between the effective zones of higher water extraction and the soil region with 80% of total roots.
The depths and distances from the plants for an adequate positioning of sensors in the soil irrigated with drip and microsprinkler irrigation systems are shown in Table 2. The ideal regions of sensor placement found was up to 0.75 m, from the pseudo-stem and up to 0.33 m deep, towards the microsprinkler for treatment T1; up to 0.80 m from pseudo-stem and up to 0.48 m deep for treatment 2; and up to 0.83 m from pseudo-stem towards the micro-sprinkler, and up to 0.55 m deep for treatment 3. These results are in agreement with those proposed by Silva et al. (2018), who applied the concept of Time Stable Representative Proposition (TSRP) and found two patterns of water uptake for banana crops as a function of the developmental stages: one characterizes the initial and vegetative growth stage, when the effective water extraction occurs up to the distance of 0.70 m from the plant, and the other characterizes the flowering and fruit growth stages, when the effective extraction occurs up to 0.90 m from    Table 2. Distances from plant and soil depth for placement water content or potential sensors in the root zone of banana crops