In the analysis of variance (ANOVA) (P < 0.05), it was observed that in cycle I, the gas exchange variables represented by stomatal conductance (gs) and leaf transpiration (E) were statistically significant based on the applied irrigation treatments, showing significance at a 1% probability level. The same was observed for the other evaluated variables, including leaf water potential (ΨLeaf), soil moisture (SMO), and thermal indices (∆Tcanopy−ar, CWSI, and IG), at all three assessment times.
In cycle II, the effect of irrigation treatments observed in the ANOVA yielded identical results to those of cycle I, with all monitored variables showing statistical significance at a 1% probability level.
3.1 Leaf water potential (Ѱleaf) and soil moisture (SMO) versus accumulated irrigation depths (IR)
Correlating accumulated irrigation depths (IR) with ΨF and SMO (Fig. 2), it can be observed that in both cycles, this relationship showed a strong correlation, with the IR x SMO interaction in cycle I having an R2 value of 0.7 with a significant linear model at 1%, while in the other interactions, the coefficients were greater than 0.8, and the models showed significance at P < 0.001.
Biweekly measurements of ΨLeaf and SMO in melon plants conducted by Aragão et al. (2023) - representing one measurement per phenological phase in the Brazilian semiarid region - demonstrated that the greatest differences between irrigation regimes are observed at 45 DAT, and these differences are more pronounced in soils without dead plant cover, which is consistent with the data presented here. However, when comparing the ΨLeaf averages obtained here with those observed by Aragão et al. (2023), it is noted that the excess irrigation beyond what is required by the crop tends to yield similar results, with a smaller difference (approximately − 0.02 MPa) compared to water deficit (-0.24 MPa), where the difference between deficit treatments in the studies was 10% of ETc. Similarities are observed regarding SMO, thus agreeing with Buckley (2019) regarding the relationship between ΨLeaf gradients in the soil-plant system.
In addition to SMO, it is expected that the water content in the effective root layer of melon plants in the RDI treatments will remain in a condition of moderate stress, between the severe deficit applied and the adequate condition, approaching more towards these extremes as the percentage of ETc applied in the phase increases, as was the case in this study, where averages close to the corresponding treatments were obtained. It is worth noting that changes in replenishment would also lead to changes in soil moisture content, as observed by Lagos et al. (2017), Cotrim et al. (2017), and Blanco et al. (2019), with this variation being more noticeable in the top soil layers. In this study, it was in the top 0.1 m, while in kiwi, Lagos et al. (2017) observed it in the top 0.3 m.
3.2 Meteorological variables and gas exchange
For the days monitored in this study, the observed Tmin and Tmax were 22.7 and 37.5°C, respectively, at 45 DAT in cycle I, and 22.9 and 36.8°C, respectively, at 42 DAT in cycle II. These temperature values are within the melon's tolerance range, which ideally falls between 25°C and 35°C (Bezerra et al., 2017; Oliveira et al., 2017a).
Based on the variation in vapor pressure deficit (Δea), it was observed that the minimum values occurred between 5:00 and 6:00 h, and the maximum values between 14:00 and 15:00 h. The decline in Δea after 15 hours was more pronounced in cycle I. Observing these daily and periodic variations helped in choosing the data collection time for future studies.
Analyzing the variation of meteorological variables with the plant-atmosphere water relations, it is understood that this variation directly influences the analogous process between stomatal conductance (gs) and leaf transpiration (E). However, gs values also depend on stomatal opening and/or closing, which is determined by the movement of the guard cell walls adjacent to the stomatal pore, due to changes in cell turgor and the occurrence of gradients in leaf water potential both between and within plant tissues (Buckley, 2019). The decline in soil water potential induces a decline in leaf water potential throughout the plant, consequently reducing stomatal conductance as a result of negative feedbacks (Buckley, 2019).
The gas exchange rates obtained (Fig. 3) show partially similar behaviors to those found by Vieira et al. (2019) with yellow and pele de sapo melons in the São Francisco Valley region. Relating stomatal conductance (gs) and leaf transpiration (E) rates to the FDI irrigation condition, the authors observed lower gs rates at the lowest water replacement (50% of ETc), which disagrees with the results obtained here, where the lowest water replacement provided gs values higher than in the T1 and T2 replacements, which were 125% and 100% of ETc, respectively. However, the same did not happen with the leaf transpiration rate (E), where a higher rate was found at the lowest water replacement tested (50% of ETc), in agreement with the results obtained. There is also a noticeable increase in gs and E rates for cycle II compared to cycle I, indicating a seasonal difference between the cycles. Seasonal variations in gs were also observed by Rodrigues et al. (2011).
It is worth noting that during the data collection period (Phase III of melon), the water replacements for the irrigation treatments from T1 to T9 corresponded to 125%, 100%, 75%, 50%, 100%, 75%, 75%, 100%, and 50% of ETc, respectively. Treatments from T1 to T4 correspond to fixed deficit irrigation (FDI), meaning that the replacement is the same in the other phases of melon cultivation. For treatments from T5 to T9, the replacements varied in the other phases of cultivation.
Regarding the RDI treatments, differences in the mean values were observed both between cycles and between corresponding treatments with FDI. This was evident in the responses of treatments T2 (FDI), T5, and T8 (RDI), which in the evaluated phase received 100% of ETc replenishment, as well as treatments T3, T6, and T7, which received 75% of ETc, and treatments T4 and T9, which received 50% of ETc. It's important to note that these correspondences differed in other physiological phases. Therefore, the differences in the results of RDI treatments are likely due to plant adaptations to water availability. García-Tejero et al. (2018b), working with irrigation deficits in almond trees, observed that under sustained deficit (100% of ETc, except during seed filling and pre-harvest, when this treatment was irrigated with ~ 50% of ETc), stomatal conductance (gs) changed according to the amount of water applied.
Although stomatal responses to environmental variations are complex, it is presumed that the environmental conditions (Ta, RH and Δea) at 8:00 h still provided viability for vapor - CO2 flux in the plants of treatment T4, explaining the high gs rates for this treatment in both cycles, especially in cycle I.
In treatments T1 (FDI − 125% ETc) and T2 (FDI − 100% ETc), the responses to low Δea, due to low losses to the atmosphere, resulted in mesophyll saturation because of the high potential in these treatments, causing partial stomatal closure, as mentioned by Buckley (2019). This led to low gs and E rates, which later, with an increase in Δea, would exhibit rates considered normal and consistent with water replenishments.
The same can be applied to RDI treatments; however, the influence of plant morphology should be considered. The plant's larger structure was developed in a specific phase (vegetative growth - phase II) with different replenishments than applied in the phase from which the data were collected. In addition, the stomatal responses obtained for both FDI and RDI should also be related to the selected leaf, which was the fourth from the apex. Choosing a more morphologically mature leaf with established stomatal density could potentially yield more reliable results. According to Driesen et al. (2020), stomatal density is considered an adaptive response of plants to environmental conditions.
To support and reinforce the points made, several factors should be considered. Firstly, plants, being sessile, have the ability to adapt their physiology to external factors, influencing stomatal distribution and function (Driesen et al., 2020). In the short term, this response involves adjusting stomatal closure, while in the long term, it involves regulating stomatal density in developing leaves (Pillitteri and Torii, 2012; Zoulias et al., 2018; Haworth et al., 2011). Also, according to Grossiord et al. (2020), Δea is one of the main determinants of plant water relations, leading plants to close their stomata to minimize water loss and avoid critical tension within the xylem (Running, 1976). Due to the relationship between air warming and the amount of vapor air can hold when saturated (es) (Bohren et al., 2000), the impacts of Δea on plant physiology are associated with temperature, solar radiation, and CO2 concentration (Grossiord et al., 2020). Another factor that Buckley (2019) mentioned is that in angiosperms, the passive inflation of epidermal cells at high leaf water potential pushes guard cells, causing stomata to close instead of open. Therefore, the author explains that the simplest and most intuitively obvious conceptual model of stomatal response to environmental factors is fundamentally incorrect due to ignoring the opposite effect of adjacent epidermal cells. To achieve stomatal opening at high leaf water potential, the epidermal effect must be overcome. As Buckley (2005; 2019) suggests, a likely mechanism to achieve this is hydroactive feedback, which hypothetically unifies stomatal responses to any factor that influences leaf water potential, including changes in humidity, soil moisture, and water transport by plants, under a single mechanism. Finally, Buckley (2019) and Grossiord et al. (2020) discuss the fact that stomata open transiently before closing in response to increased loss or reduced water supply. In angiosperms, this transient opening occurs after any hydraulic disturbance (Buckley, 2019), including a change in evaporative demand (Mott et al., 1997), source water potential (Comstock and Mencuccini, 1998), hydraulic conductance (Saliendra et al., 1995), or leaf excision (Powles et al., 2006).
3.3 Relationship between thermal indices versus leaf water potential (ΨLeaf) and soil moisture (SMO)
Studies related to the imposition of water stress on plants have established leaf water potential (ΨLeaf) as an excellent response variable of plants to water stress, especially in situations with deficit irrigation. However, variations in ΨLeaf only occur due to changes in water availability to the plant, represented here by SMO, explaining the relationship between these two variables.
Variations in leaf water potential in the soil-plant system directly influence leaf turgor, which, in turn, is reflected in the thermal response observed in thermography. This fact supports the correlations obtained between thermal indices based on ΨLeaf and SMO (Fig. 4). The relationships presented for the 12:00 and 16:00 h time points were established using ΨLeaf and SMO values obtained at 8:00 h. However, this does not affect the analyses, as can be observed in the relationships with ∆Tcanopy−ar, as there is a coincidence in the hydraulic responses.
The trend of ∆Tcanopy−ar between the cycles shows similarity only at 12:00 h, with the means concentrated on a scale between 0.0 and − 5.5. At 8:00 h, it was observed that the means of cycle II are distributed on a larger scale than cycle I, while at 16:00 h, the range of scales is similar, but with the means of cycle II being lower, reaching values below − 8.0. These results are attributed to: 1) the climatic season - Winter solstice in cycle I and spring equinox in cycle II; 2) the interaction and variation of meteorological variables, which provide more stressful conditions at 12:00 h; 3) the local microclimate - different between the cycles, drier in cycle II, due to the deciduous characteristic of the caatinga in the non-rainy season (Albuquerque and Bandeira 1995), similar to the observations made by Fitter and Fitter (2002), Morison and Morecroft (2008), and Shoa et al. (2021).
In the literature, studies by Gonzalez-Dugo et al. (2012; 2014); Ballester et al. (2013a), García-Tejero et al. (2016; 2017a; 2017b; 2018a; b), Egea et al. (2017), and Costa et al. (2020) have explored the use of relationships between meteorological variables and thermal indices (∆Tcanopy-ar, CWSI, and IG) to estimate physiological responses like ΨLeaf or soil water content across different crops, measurement times, irrigation strategies, and environmental conditions.
As presented in Table 1, the significance of the adjusted model (linear or quadratic polynomial, selected based on the p-value) was similar for all variables, being equal to or less than 1.0%. This justifies the fact that the variables consistently express the plant's water status and the potential use of thermal indices for irrigation management. These relationships are supported by the works of Ben-Gal et al. (2009) with olive trees, García-Tejero et al. (2011) with citrus, Ballester et al. (2013b) with citrus and persimmon, Costa et al. (2019) with vineyards, and Blaya-Ros et al. (2020) with cherry trees. In melon cultivation, correlations conducted by Aragão et al. (2023) showed R2 values of 0.83 and 0.88 for ∆Tcanopy−ar in relation to ΨLeaf and SMO, respectively.
Table 1
Regressions of the thermal indices of the temperature difference between the canopy and air (∆Tcanopy−ar), crop water stress index (CWSI), stomatal conductance index (IG) as a function of leaf water potential (ΨLeaf) and soil moisture content (SMO), obtained for two cycles of melon crop subjected to fixed irrigation (FDI) and regulated deficit irrigation (RDI) treatments, corresponding to 45 DAT (August-September 2021) and 42 DAT (November-December 2021), Sobral, Ceará, Brazil, 2021. The equations were selected based on significance (p-value).
|
|
CYCLE I
|
|
Interaction
|
08 h
|
12 h
|
16 h
|
ΨLeaf x ∆T
|
Y = -3.967–24.930x
R2 = 0.82; P = 0.0008
|
Y = -9.906–65.647x
R2 = 0.68; P = 0.006
|
Y = -9.334–64.494x
R2 = 0.93; P < 0.0001
|
SMO x ∆T
|
Y = 3.38–44.312x
R2 = 0.85; P = 0.0004
|
Y = 65.83–1126.61x + 4480.12x2
R2 = 0.97; P < 0.0001
|
Y = 9.120–109.852x
R2 = 0.88; P = 0.0002
|
ΨLeaf x CWSI
|
Y = -0.010–1.819x
R2 = 0.83; P = 0.0006
|
Y = -0.152–3.210x
R2 = 0.68; P = 0.0059
|
Y = -0.184–4.562x
R2 = 0.92; P < 0.0001
|
SMO x CWSI
|
Y = 0.525–3.222x
R2 = 0.86; P = 0.0004
|
Y = 3.482–53.839x + 213.459x2
R2 = 0.97; P < 0.0001
|
Y = 1.126–7.810x
R2 = 0.89; P = 0.0001
|
ΨLeaf x IG
|
Y = 13.610 + 84.951x
R2 = 0.85; P = 0.0004
|
Y = 23.40 + 171.394x
R2 = 0.91; P < 0.0001
|
Y = 45.014 + 796.091x + 3676.977x2
R2 = 0.89; P = 0.0015
|
SMO x IG
|
Y = -9.214 + 131.865x
R2 = 0.67; P = 0.0067
|
Y = -24.805 + 284.687x
R2 = 0.83; P = 0.0007
|
Y = 66.27–1320.472x + 6743.149x2
R2 = 0.49; P = 0.1338
|
|
|
CYCLE II
|
|
ΨLeaf x ∆T
|
Y = 0.774 + 89.692x + 1145.88x2
R2 = 0.98; P < 0.0001
|
Y = -7.643–56.677x
R2 = 0.81; P = 0.001
|
Y = -12.169–70.081x
R2 = 0.83; P = 0.0006
|
SMO x ∆T
|
Y = 126.64–1892.29x + 7017.24x2
R2 = 0.93; P = 0.0003
|
Y = 91.646–1450.84x + 5468.71x2
R2 = 0.84; P = 0.0044
|
Y = 105.402–1708.62x + 6414.33x2
R2 = 0.85; P = 0.0031
|
ΨLeaf x CWSI
|
Y = 0.149 + 4.795x + 58.62x2
R2 = 0.98; P < 0.0001
|
Y = -0.131–3.339x
R2 = 0.79; P = 0.0013
|
Y = -0.164–5.454x
R2 = 0.83; P = 0.0006
|
SMO x CWSI
|
Y = 6.405–94.459x + 351.019x2
R2 = 0.93; P = 0.0003
|
Y = 5.547–82.651x + 310.626x2
R2 = 0.82; P = 0.0060
|
Y = 8.824–130.311x + 488.332x2
R2 = 0.85; P = 0.0033
|
ΨLeaf x IG
|
Y = 57.49 + 1047.13x + 5000.988x2
R2 = 0.96; P < 0.0001
|
Y = 48.62 + 780.511x + 3357.278x2
R2 = 0.90; P = 0.0009
|
Y = 10.756 + 90.681x
R2 = 0.94; P < 0.0001
|
SMO x IG
|
Y = -57.628 + 54.010x
R2 = 0.98; P < 0.0001
|
Y = -53.279 + 520.533x
R2 = 0.94; P < 0.0001
|
Y = -17.731 + 177.598x
R2 = 0.95; P < 0.0001
|
According to Gonzalez-Dugo and Zarco-Tejada (2022), the Crop Water Stress Index (CWSI) is the most commonly used thermal indicator to assess the water status of plants. It is considered to be a robust indicator compared to ∆Tcanopy−air due to its method of calculation, which involves the normalization of lower (Twet) and upper (Tdry) temperature limits, ranging from 0 (no water stress) to 1 (fully stressed). For more details, please refer to Maes and Steppe (2012).
As observed in ∆Tcanopy−air, the response of CWSI was consistent with the limits of ΨLeaf and SMO, but without significant variation between cycles or times of day. Values obtained at 8:00 h and 12:00 h remained within the range of 0.0 to 0.3, while those at 16 hours reached a maximum of 0.39 in T4 (FDI) in cycle I and 0.48 in T9 (RDI) in cycle II, both with 50% ETc replenishment during the referenced phase (phase III of melon crop). Similar values for CWSI in melon crops were also obtained by Silva et al. (2007). In their study, using fixed irrigations - with a control treatment corresponding to 75% of Class "A" pan evaporation and the others corresponding to 90%, 80%, and 70% of the control treatment - they found an average CWSI ranging from 0.23 to 0.39 in the cycle. Values were around 0.2 between 28 and 50 days after sowing for each treatment. Ferreira (1995) suggests that a CWSI of 0.3 can be considered a threshold value to initiate irrigation in melon crops without harming the plants.
The linear model was the best fit for CWSI as a function of ΨLeaf at 12:00 h and 16:00 h in both cycles. In terms of SMO, the quadratic model provided the best fit for the 12:00 h data. For the other relationships, the models were different between cycles, and in all cases, they were significant at a level of 1% or lower (Table 1).
The literature has shown that the use of CWSI has provided satisfactory results in monitoring plant water status, with values varying according to daily seasonality and the crop cycle. This is justified by the plant's response to the imposed conditions, as observed by Erdem et al. (2005), Agam et al. (2013), Sezen et al. (2014), García-Tejero et al. (2017a), Ghazouani et al. (2017), Blaya-Ros et al. (2020), among others.
Comparing the thermal indices ∆Tcanopy−air and CWSI, and considering the results obtained in this study, the following observations are made: 1) the differences between cycles observed in this study are more pronounced with ∆Tcanopy−air, especially at times of lower stress (8:00 h and 16:00 h), possibly because this index is more sensitive to environmental variation and influence; 2) CWSI minimizes the effect of local microclimate differences between cycles due to its robustness, which considers the normalization of upper and lower temperature limits, especially when irrigation is adequate or has a moderate deficit (100% and 75% of ETc); 3) for management purposes, 12:00 h can be considered a suitable time for thermal data collection. Similar results were obtained for citrus, with the best time between 11:20 h and 12:00 h according to Gonzalez-Dugo et al. (2014); for almonds (García-Tejero et al., 2018a), the best time was from 11:30 h to 14:00 h, and for Gutiérrez-Gordillo et al. (2020), the best time was from 12:00 h to 13:30 h.
Intrinsic to the IG results obtained here, in previous studies, values for well-irrigated conditions did not exceed 5.0, as reported by Grant et al. (2016), Costa et al. (2012), Pou et al. (2014), García-Tejero et al. (2016), Belfiore et al. (2019), and Gautam and Pagay (2020). In this study, values closest to those in the literature were obtained at 16:00 h, with the averages concentrated between 1.0 and 5.5 for treatments with 50% and 100% of ETc, respectively, in both cycles. At other times, for treatments with equal replenishment, IG ranged, in cycle I, between 3.0 and 6.0 at 8:00 h and between 2.0 and 10.0 at 12:00 h. In cycle II, IG ranged between 2.0 and 11.0 at 8:00 h and between 3.0 and 12.0 when monitored at 12:00 h. For treatment T1, with surplus replenishment, the highest values were observed at 8:00 h and 12:00 h in cycle II, with an average close to 20.0. In cycle I, the highest average was obtained at 12:00 h, with IG close to 14.0.
Like ∆Tcanopy−air and CWSI, IG as a function of ΨLeaf and SMO showed significant correlations of less than 1% with R2 values exceeding 0.6 (Table 1). The exception was at 16:00 h in cycle I when it was not significant, and the R2 was 0.49. This result may be related to the fact that the IG for T1 was more than double that obtained for treatments with 100% of ETc (T2, T5, and T8).
Maes and Steppe (2012) state that the input data for IG are the same as those used for calculating CWSI, and although the IG values obtained in this study are higher than those in previous work, they still align with the relationship presented by the authors op cit, with CWSI = (1 + IG)−1. It is worth noting that the CWSI used by the authors is obtained through a direct approach, but it is analogous to the analytical (theoretical) (Jackson et al., 1981) and empirical (Idso et al., 1981) approaches. This fact does not have implications when using the data obtained in this research in the mentioned relationship, such as the averages of IG and CWSI for treatment T4 at 8:00 h in cycle I, which were 3.29 and 0.23, respectively, or treatment T2, also at 8:00 h in cycle II, with averages of 11.43 and 0.08. For all the averages obtained in all treatments, times, and cycles, the relationship was consistent with the literature.
The average IG values obtained in this study are specific to melon cultivation in a hot region. It's important to note that the values may differ for other crops or under different environmental conditions. For example, in cucumber cultivation, which belongs to the same family as melon, Kaukoranta et al. (2005) found IG values of up to 6.0 in a protected environment with supplemental lighting from 18 to 22 hours during the winter solstice in Finland. It's worth emphasizing that, prior to this study, there were no IG results available for melon crop subjected to different irrigation strategies and treatments under field conditions in the study region.
Thermal images, yield and water use efficiency (WUE) of melon under fixed deficit irrigation (FDI) and regulated deficit irrigation (RDI)
In cycle I, the infrared thermal pattern shown in the images (Fig. 5) had less variation at 8:00 h, unlike what was observed in cycle II. In the other monitored times, in both cycles, there is a noticeable difference between the deficit levels, as well as the similarity between the FDI and RDI treatments with the same replenishment. Based on the ∆T, it was also observed in cycle II that higher values were recorded at 8:00 h, indicating that the plants were warmer at that time compared to cycle I. This may be related to a higher occurrence of winds and a possible decrease in temperature, which in cycle II, due to the surrounding vegetation being drier, led to the so-called oasis effect, a phenomenon resulting from the difference in humidity between the surroundings and the experimental area (Allen et al., 1998).
In the other two time intervals evaluated, the difference in the infrared thermal pattern in the images is more visible, and by the ∆T values at 12:00 h., where the averages are similar between cycles, it appears to be a good time for data collection for the development of irrigation management models via thermography. Regarding the thermal information observed at 16:00 h, the difference in ∆T between cycles is related to the cooling of the air at 16:00 h, which in cycle I was 3ºC lower than the temperature at 15:00 h. While in cycle II, this reduction was less than 1ºC. Therefore, the warmer air in cycle II at 16:00 h resulted in lower ∆T. It should be noted that conclusions based on thermal image patterns can lead to errors, and the same can happen in the interpretations of ∆T when the data refer to more than one crop cycle.
Regarding productivity, it is observed in cycle I that treatments T5, T6, T7, and T8 (RDI) provided yields higher than the average productivity of the Brazilian semiarid region (29 ton. ha− 1), according to IBGE data (2021) for the year 2019, repeating this achievement in cycle II. For the FDI treatments, only T2 in cycle I and T1 in cycle II showed average productivity higher than that of the Brazilian semiarid region.
Comparing the irrigation strategies, the average productivity with RDI was higher than with FDI, with emphasis on treatment T6 in cycle I, which had the highest yield (36.1 ton. ha− 1) and had a replenishment of 75% of ETc during the evaluated phase (phase III of melon cultivation). When compared to T7 (RDI), which also received the same percentage of replenishment during phase III, and to T3 (FDI), with 75% of ETc throughout the entire cultivation, both in cycle I, the productivity was 13.7% and 52.8% higher than the average, respectively. When this comparison is made using the productivities of cycle II, the superiority of T6 (32.28 ton. ha− 1) over T7 (29.1 ton. ha− 1) and T3 (20.6 ton. ha− 1) was 11.1% and 41.4%, respectively. Statistically, treatment T6 remained the same as T7 in both cycles.
Table 2
– Total accumulated irrigation dephts (IRT), productivity, and water use efficiency (WUE) of melon crops subjected to fixed deficit irrigation (FDI) and regulated deficit irrigation (RDI) in two growing cycles. Cycle I (August-September 2021), Cycle II (November-December 2021), Sobral, Ceará, Brazil, 2021. Means in the vertical followed by the same letter do not differ significantly from each other according to the Tukey test at a 5% significance level.
Treatments
|
Cycle I
|
|
Cycle II
|
IRT
(mm)
|
Prod.
(ton. ha− 1)
|
WUE
(kg m− 3)
|
|
IRT
(mm)
|
Prod.
(ton. ha− 1)
|
WUE
(kg m− 3)
|
T1 – FDI
|
368.80
|
27.8 ab
|
7.5 b
|
|
331.95
|
29.5 ab
|
8.7 c
|
T2 – FDI
|
295.04
|
34.2 a
|
11.6 ab
|
|
265.56
|
26.4 abc
|
9.8 bc
|
T3 – FDI
|
221.28
|
23.6 ab
|
10.7 ab
|
|
199.17
|
20.6 bc
|
10.2 bc
|
T4 – FDI
|
147.52
|
16.0 b
|
10.8 ab
|
|
132.78
|
14.8 c
|
11.0 abc
|
T5 - RDI
|
264.31
|
29.4 ab
|
11.1 ab
|
|
244.30
|
29.8 ab
|
12.0 abc
|
T6 – RDI
|
230.75
|
36.1 a
|
15.0 a
|
|
209.69
|
32.3 ab
|
15.3 a
|
T7 – RDI
|
199.19
|
31.7 a
|
15.9 a
|
|
188.85
|
29.1 ab
|
15.1 a
|
T8 – RDI
|
267.46
|
33.2 a
|
12.6 ab
|
|
233.58
|
34.2 a
|
14.4 ab
|
T9 - RDI
|
179.43
|
27.5 ab
|
15.3 a
|
|
170.22
|
21.6 bc
|
12.4 abc
|
In relation to water use efficiency (WUE), T6 stood out from the other treatments, with 15.6 kg m− 3 in cycle I and 15.3 kg m− 3 in cycle II, being inferior only to T7 in cycle I (15.9 kg m− 3), although they are statistically equal.
Previous studies have demonstrated that lower irrigation levels result in higher WUE values, which is consistent with the results obtained here when considering the FDI treatments, with the WUE of T4 (50%) being higher than that of T2 (100% ETc). As observed, the results of the RDI treatments differed from those with FDI, resulting in higher productivity and WUE, supporting the purpose of RDI. According to the results, the strategy of T6 (RDI) appeared to be more effective.
The feasibility of adopting RDI irrigation strategies has been demonstrated in various crops, including sugar beet by Fabeiro et al. (2003), onions by Ollala et al. (2004), coffee by Silva et al. (2009), soybeans by Gava et al. (2015), pears by Molina-Ochoa et al. (2015), tomatoes by Nangare et al. (2016), mangoes by Cotrim et al. (2017), kiwis by Lagos et al. (2017), peppers by Yang et al. (2018), sugarcane by Simões et al. (2018), and industrial tomatoes by Válcarcel et al. (2020). For melon, similar results were obtained by Fabeiro et al. (2002) and Azevedo et al. (2016).
3.5 Relation between water use efficiency (WUE) and thermal indices
The average WUE values did not show a clear trend or sequential order (direct or inverse) based on accumulated irrigation depths, as the highest WUE values were observed with RDI when the accumulated irrigation depths ranged from 50–100% of ETc. This fact, when correlated with thermal indices, resulted in disordered distributions of averages (Fig. 6), as the thermal response obtained showed concurrent averages with water replenishment. Therefore, when considering the treatments in general, there was no considerable fit in the regression (linear or quadratic).
In Fig. 6, it can be observed that the regression performed, highlighting the FDI and RDI strategies, had more consistent fits, with most of them conforming to the quadratic polynomial model, except for the linear fit for the FDI relationships in cycle II with ∆Tcanopy−ar and CWSI at 8:00 h, and with IG for both evaluated times (Table 3).
Table 3
Regression equations resulting from the interactions between Water Use Efficiency (WUE) in two melon crop cycles subjected to fixed deficit irrigation (FDI) and regulated deficit irrigation (RDI), as a function of the thermal indices of the temperature difference between the canopy and air (∆Tdossel−ar), crop water stress index (CWSI), stomatal conductance index (IG), obtained at 8:00 h, 12:00 h, and 16:00 h, at 45 DAT in cycle I (Aug-Sep 2021) and 42 DAS in cycle II (Nov-Dec 2021). The equations were chosen based on significance (P-value < 0.05) and the coefficient of determination R2, Sobral, Ceará, Brazil, 2021.
|
|
CYCLE I
|
|
Interaction
|
8 h
|
12 h
|
16 h
|
WUEFDI x ∆T
|
Y = 7.398–7.044x – 2.720x2
R2 = 0.79; P = 0.459
|
Y = 9.761 -2.526x – 0.504x2
R2 = 0.44; P = 0.7496
|
Y = 7.976–2.755x – 0.492x2
R2 = 0.73; P = 0.5169
|
WUERDI x ∆T
|
Y = -27.787–55.629x – 17.716 x2
R2 = 0.99; P = 0.0124
|
Y = -38.549–30.811x – 4.204x2
R2 = 0.95; P = 0.0554
|
Y = 4.065–8.667x – 1.586x2
R2 = 0.75 P = 0.2489
|
WUEFDI x CWSI
|
Y = -5.643 + 190.775x – 517.436x2
R2 = 0.80; P = 0.4432
|
Y = 4.029 + 83.488x – 198.914x2
R2 = 0.42; P = 0.7650
|
Y = 4.075 + 55.845x – 100.031x2
R2 = 0.72; P = 0.5296
|
WUERDI x CWSI
|
Y = -79.98 + 1171.123x – 3574.112x2
R2 = 0.99; P = 0.0024
|
Y = -21.949 + 519.226x – 1702.104x2
R2 = 0.90; P = 0.0965
|
Y = − 9.617 + 180.427x – 318.452x2
R2 = 0.74; P = 0.2574
|
WUEFDI x IG
|
Y = 6.161 + 1.998x – 0.191x2
R2 = 0.90; P = 0.3182
|
Y = 9.16 + 0.799x – 0.064x2
R2 = 0.76; P = 0.4898
|
Y = 9.666 + 0.704x – 0.076x2
R2 = 0.95; P = 0.2237
|
WUERDI x IG
|
Y = -28.681 + 17.057x – 1.635x2
R2 = 0.99; P = 0.0035
|
Y = 2.533 + 4.574x – 0.373x2
R2 = 0.96; P = 0.0355
|
Y = 15.953 + 0.570x – 0.279x2
R2 = 0.66; P = 0.3422
|
|
|
CYCLE II
|
|
WUEFDI x ∆T
|
Y = 9.50 + 0.364x
R2 = 0.87; P = 0.0696
|
Y = 10.080–1.088x – 0.292x2
R2 = 0.96; P = 0.2068
|
Y = 6.585–1.795x – 0.180x2
R2 = 0.95; P = 0.2214
|
WUERDI x ∆T
|
Y = 14.043 + 1.554x – 0.506x2
R2 = 0.73; P = 0.2688
|
Y = 8.569–5.050x – 0.933x2
R2 = 0.55; P = 0.4472
|
Y = -8.702–7.918x – 0.650x2
R2 = 0.64; P = 0.36
|
WUEFDI x CWSI
|
Y = 8.785 + 7.426x
R2 = 0.86; P = 0.0723
|
Y = 7.516 + 33.884x – 80.567x2
R2 = 0.98; P = 0.1514
|
Y = 6.434 + 23.326x – 29.467x2
R2 = 0.96; P = 0.1941
|
WUERDI x CWSI
|
Y = 8.966 + 73.224x – 213.014x2
R2 = 0.75; P = 0.2555
|
Y = 7.633 + 98.337–300.565x2
R2 = 0.60; P = 0.3976
|
Y = 5.045 + 67.215x – 108.766x2
R2 = 0.65; P = 0.3529
|
WUEFDI x IG
|
Y = 11.002–0.113x
R2 = 0.94; 0.0307
|
Y = 11.272–0.120x
R2 = 0.99; P = 0.0066
|
Y = 11.446–0.366x
R2 = 0.97; P = 0.0151
|
WUERDI x IG
|
Y = 7.948 + 2.092x – 0.134x2
R2 = 0.87; P = 0.1285
|
Y = 10.246 + 1.068x – 0.063x2
R2 = 0.34; P = 0.6560
|
Y = 9.547 + 3.269x – 0.495x2
R2 = 0.49; P = 0.5061
|
Analyzing the obtained models, it can be observed that most of them do not show statistical significance, although they have acceptable R2 values, above 0.7, and some below 0.5. Of all cases, the relationships in which the regression showed statistical significance in cycle I were WUE of RDI as a function of ∆Tcanopy−ar, CWSI, and IG at 8:00 h, and as a function of IG at 12:00 h, all adjusted to quadratic polynomial regression, with R2 above 0.9. In cycle II, only FDI WUE as a function of IG at both evaluated times, linearly adjusted and with R2 also above 0.9.
Among similar results found in the literature, reference is made to studies where regression is performed with productivity as a function of thermal indices, highlighting Silva et al. (2007) with melon, Sezen et al. (2014) with red pepper, Çolak et al. (2015) with eggplant, Delgado et al. (2017) with lettuce, Silva et al. (2018) with tomatoes, Camoglu et al. (2018) with pepper, and Khan et al. (2022) with wheat, applying irrigations with fixed replacements and correlating productivity with CWSI. The R2 values obtained in these studies ranged from 0.68 to 0.99.
Based on the regression results and considering the R2 values, it can be considered that in each cycle, WUE of one of the irrigation strategies correlates better with the thermal indices: RDI in cycle I and FDI in cycle II. The exact reason is unknown, but it is suggested that the reason may be related to differences in the microclimate surrounding the experimental area between the two cycles. However, when the goal is to increase productivity with less water, the suggestion is to adopt RDI strategies, which showed the highest WUE.