Water Stress Index and Stomatal Conductance under Different Irrigation Regimes with Thermal Sensors in Rice Fields on the Northern Coast of Peru

Abstract

In the face of the climate change crisis, the increase in air temperature negatively impacts rice crop productivity due to stress from water scarcity. The objective of this study was to determine the rice crop water stress index (CWSI) and stomatal conductance (Gs) under different irrigation regimes, specifically continuous flood irrigation treatments (CF) and irrigations with alternating wetting and drying (AWD) at water levels of 5 cm, 10 cm, and 20 cm below the soil surface (${\mathrm{AWD}}_5{, \mathrm{AWD}}_{10}$, and ${\mathrm{AWD}}_{20}$) in an experimental area of INIA-Vista Florida and in six commercial areas of the Lambayeque region using thermal images captured with thermal sensors. The results indicated that AWD irrigation generated more water stress, with CWSI values between 0.4 and 1.0. Despite this, the yields were similar in CF and ${\mathrm{AWD}}_{20}.$In the commercial areas, CWSI values between 0.38 and 0.51 were obtained, with Santa Julia having the highest values. Furthermore, a strong Pearson correlation (R) of 0.91 was established between the CWSI and Gs, representing a reference scale based on Gs values for evaluating water stress levels.

1. Introduction

Rice plays a crucial role in providing sustenance for more than half of the global population and requires a significant amount of water for its growth [1]. Approximately 70% of the world’s water resources are utilized for agricultural purposes, with 34% to 43% of irrigation water being allocated to rice cultivation [2,3]. In the 2022/23 season, worldwide rice production reached a staggering 519.8 million tons [4]. However, the escalating water crisis in rice-producing nations poses a severe threat to food security [1].

To address this pressing issue, innovative agricultural technologies are necessary to optimize irrigation practices and reduce water consumption. This would involve moving away from the conventional method of continuous flooding (CF), which demands a substantial volume of water. Several alternative techniques have been advocated for adoption by farmers, including the alternating wetting and drying (AWD) irrigation method, the aerobic rice system, the system of rice intensification (SRI), saturated soil cultivation (SSC), direct seeding of rice (DSR), drip irrigation, and automated surface gravity irrigation systems [1,5]. The implementation of AWD stands out due to its intermittent and controlled application of floods throughout the rice growing cycle. This approach allows for a reduction in overall water usage while ensuring that the rice crop receives an appropriate amount of water during critical stages of production to prevent negative impacts on yield and grain quality [6,7].

Unmanned aerial vehicles (UAVs) have a wide range of applications in agriculture [8]. They can be utilized for monitoring the growth rate and condition of crops, estimating biomass, and detecting water stress [9,10]. Since 1970, the assessment of crop water status has involved the use of plant canopy temperature. The crop water stress index (CWSI) is the most commonly used indicator of thermal origin, which quantifies crop water status using temperature thresholds, also known as baselines. Various types of sensors are available to estimate canopy temperature, with IRT sensors placed within the canopy and the thermography technique being the most frequently employed [11]. In recent times, there has been an increasing use of infrared thermographic cameras to measure crop temperature (${\mathrm{T}}_{\mathrm{c}}$) and plant water stress. These cameras allow the capture of thermal images of the plant surface in a fast and efficient manner. In addition, this technology is characterized by being non-invasive, non-contact and non-destructive [10,12,13,14]. This technology enables the monitoring of CWSI variations, providing valuable information on the spatial and temporal distribution of crop water requirements. Consequently, farmers are able to take appropriate measures to ensure and potentially enhance crop yields [15,16].

Stomatal conductance, which plays a crucial role in processes like photosynthesis and transpiration, is primarily influenced by environmental factors such as temperature, soil moisture, solar radiation, and water vapor pressure deficit. These factors determine the variability in stomatal conductance and act as regulators for vegetation equilibrium [17]. Olalekan et al. [18] found a positive correlation between relative leaf water content and stomatal conductance under optimal water conditions. Conversely, Alvarez et al. [19] demonstrated that most plants reduce their stomatal conductance in response to a decrease in soil water content.

In 2022, the production of paddy rice in Peru reached 3 million tons, with the regions of Lambayeque and La Libertad being the highest producers [20]. In the coastal valleys, 82% of the rice area is cultivated using irrigation, while 18% relies on rainfed irrigation [21]. However, the high water consumption associated with flooded irrigation (12,000 to 20,000 ${\mathrm{m}}^3{ \mathrm{ha}}^{-1}$) is leading to a rise in the water table and subsequent soil salinization [22]. This, in turn, is reducing agricultural land and decreasing productivity in the coastal valleys. Additionally, water scarcity is becoming a growing challenge, particularly for rice cultivation, where water deficit is a major form of abiotic stress [23]. Nonetheless, it is important to note that with proper management, there is potential to enhance both crop quality and yield. In Peru, there have not been studies on CWSI, stomatal conductance (Gs), and UAVs, so this study is a pioneer in evaluating various irrigation regimes CF and AWD. In this way, it seeks to address water deficit conditions in Lambayeque.

The objective of this study was to evaluate water stress and Gs in rice crops under different AWD and CF irrigation regimes in experimental and commercial areas of the Lambayeque region. For this purpose, the collection of field data (radiometer and portable weather station) allowed for the estimation of a more robust and accurate CWSI. In addition, a relationship between the CWSI and Gs was established to develop a technological tool to monitor water stress in rice crops in order to improve irrigation management efficiency.

2. Materials and Methods

Figure 1 depicts the research methodology employed in this study, which took place during the rice growing season from December to June.

2.1. Study Area

The research was conducted in the Lambayeque region, situated on the northern coast of Peru. An experimental area was designated for the application of AWD irrigation, while the commercial areas employed the CF technique. The experiment was carried out at the rice plots of the Agricultural Experimental Station (EEA) “Vista Florida”, which is part of the National Institute for Agrarian Innovation (INIA) and is located in the district of Picsi, province of Chiclayo (Figure 2c). Six commercial rice areas were chosen for the study: Caballito 1, Caballito 2, Garcia, Santa Julia, Totora, and Zapote. These areas are all situated in the district and province of Ferreñafe (Figure 2d).

Table 1. Characteristics of the study area.

Area	Longitude	Latitude	Altitude (m.a.s.l)	Area (ha)	N° Plots	Variety
INIA-Vista Florida	06°43′56.55″S	79°47′27.55″W	35	0.11	4	INIA 515—Capoteña
Caballito 1	06°35′25.18″S	79°47′8.53″W	46	2.74	3	INIA 508—Tinajones
Caballito 2	06°35′38.82″S	79°47′5.32″W	47	11.45	12	INIA 508—Tinajones (9) INIA 515—Capoteña (3)
García	06°35′2.51″S	79°47′3.50″W	47	5.23	3	INIA 508—Tinajones
Santa Julia	06°36′32.02″S	79°47′30.64″W	41	6.46	5	INIA 510—Mallares
Totora	06°35′35.16″S	79°47′32.74″W	44	5.38	6	INIA 513—Puntilla
Zapote	06°35′44.20″S	79°47′8.04″W	46	6.01	6	Pakamuros

provides information on the five rice varieties used in the different study areas.

The experimental area at INIA-Vista Florida has a soil with a sandy loam texture, consisting of 26% sand, 39% clay, and 35% silt. The soil has a bulk density (da) of 1.41 g${\mathrm{cm}}^{-3}$ and a real density (dr) of 2.67 g${\mathrm{cm}}^{-3}$ (

Table 2. Soil physical characteristics.

Area	Soil Texture			Soil Texture	Bulk Density $(\mathbf{g} {\mathbf{cm}}^{-3}$)	Real Density $(\mathbf{g} {\mathbf{cm}}^{-3}$)	Field Capacity (%)	Permanent Wilting Point (%)
	% Sand	% Clay	% Silt
INIA-Vista Florida	26	39	35	Sandy loam	1.41	2.67	29.76 ± 1.38	16.27 ± 1.25
Caballito 1	56	16.7	27.3	Sandy loam	1.45	2.33	24.24 ± 0.96	22.54 ± 1.33
Caballito 2	49	25.1	25.9	Silty clay loam	1.37	2.63	22.54 ± 1.33	13.18 ± 0.47
García	58	17.8	24.2	Sandy loam	1.47	2.47	23.31 ± 0.73	12.33 ± 0.27
Santa Julia	37.4	33.3	29.3	Clay loam	1.46	2.56	20.38 ± 0.53	11.79 ± 0.37
Totora	31.2	35.6	33.2	Clay loam	1.53	2.60	17.79 ± 1.01	10.24 ± 0.50
Zapote	33.6	33.7	32.7	Clay loam	1.40	2.53	29.50 ± 1.96	16.58 ± 0.53

). The porosity of the soil is 47.2%. The field capacity (FC) is 29.76%, while the permanent wilting point (PWP) is 16.27%. The soil has an electrical conductivity (EC) of 0.42 dS ${\mathrm{m}}^{-1}$ and a hydrogen potential (pH) of 7.64. The cation exchange capacity (CEC) is 220 meq ${\mathrm{kg}}^{-1}$. The soil contains 1.22% organic matter (OM), 0.11% total N, 4.02% ${\mathrm{CaCO}}_3$, 12 ppm P, and 376 ppm K. These soil characteristics were analyzed at the Soil, Plant, Water, and Fertilizer Analysis Laboratory of the Faculty of Agronomy—UNALM.

2.2. Irrigation Management

At the INIA-Vista Florida experimental station, irrigation water is sourced from a canal that receives water from the Tinajones reservoir, which is under the management of the Capote irrigation commission. Throughout the different growth stages of the plants (vegetative, reproductive, and maturation), a total of sixteen, three, and five irrigation cycles were conducted, respectively (Figure S1). The quantity of water used was monitored by utilizing piezometers positioned at depths ranging from 5 to 20 cm [24]. Within the experimental area, the initial treatment was continuous flooding (CF), involving the use of a permanent or traditional flood irrigation method, while the treatments aimed at water conservation utilized irrigations with alternating wetting and drying (AWD) at water levels of 5 cm, 10 cm, and 20 cm below the soil surface (${\mathrm{AWD}}_5{, \mathrm{AWD}}_{10}$, and ${\mathrm{AWD}}_{20}$). This study was designed as an observational experiment, with four plots measuring $24 \mathrm{m} \times 11$ m each, further divided into three $8 \mathrm{m} \times 11$m subplots.

In the commercial areas, the irrigation system used was CF, which was facilitated by canals managed by the Ferreñafe irrigation commission. These canals were connected to the main hydraulic system of Tinajones.

The chemical characteristics of the irrigation water are the following: pH 7.34, EC 0.31 dS ${\mathrm{m}}^{-1}$, cations (${\mathrm{Ca}}^{2+}$1.91; ${\mathrm{Mg}}^{2+}$0.43; ${\mathrm{Na}}^{+}$ 0.59; ${\mathrm{K}}^{+}$ 0.10) meq ${\mathrm{L}}^{-1}$, and anions (${\mathrm{Cl}}^{-1}$ 1.00; ${\mathrm{HCO}}_3^{-2}$ 1.89; ${\mathrm{SO}}_4^{–2}$ 0.29) meq ${\mathrm{L}}^{-1}$. The irrigation water is classified as C2-S1, indicating low sodium content and salinity. The sodium absorption ratio (SAR) is 0.55.

In the experimental area, the INIA 515—Capoteña variety was planted on 2 January 2023, using the Almacigo technique. Thirty days post sowing (DPS), two seedlings were transplanted per stroke, with a spacing of 0.25 cm × 0.25 cm (

Table 3. Description of flight dates and days post sowing (DPS) in the experimental area.

Flight	DPS	Phenology
		CF	${\mathbf{AWD}}_5$	${\mathbf{AWD}}_{10}$	$\mathit{A}\mathit{W}{\mathit{D}}_{20}$
1	38	Seedling	Seedling	Seedling	Seedling
2	42	Start of tillering	Start of tillering	Start of tillering	Start of tillering
3	61	Tillering	Tillering	Tillering	Tillering
4	65	Tillering	Tillering	Tillering	Tillering
5	75	Tillering	Tillering	Tillering	Tillering
6	79	Tillering	Tillering	Tillering	Tillering
7	88	Tillering	Tillering	Tillering	Tillering
8	92	Maximum tillering	Tillering	Tillering	Tillering
9	103	Panicle initiation	Maximum tillering	Maximum tillering	Maximum tillering
10	107	Panicle initiation	Panicle initiation	Panicle initiation	Panicle initiation
11	123	Heading stage	Heading stage	Heading stage	Heading stage
12	127	Flowering stage	Flowering stage	Flowering stage	Flowering stage
13	147	Dough stage	Dough stage	Dough stage	Dough stage
14	149	Dough stage	Dough stage	Dough stage	Dough stage

). The fertilization N-P-K dose was 250-106-60 ${(\mathrm{Kg} \mathrm{ha}}^{-1})$ in the form of urea, diammonium phosphate, and potassium sulfate, respectively. During transplantation, 100% P and K and 30% N were applied. The remaining nitrogen fertilizer was distributed equally at the early tillering, panicle initiation, early flowering, and peak flowering stages [25]. The crop’s phenology was assessed using degree heating days (GDD), a useful index for tracking crop growth [26].

In the commercial areas, the phenological cycle of the crop began with the transplantation of the seedlings, which occurred between twenty-eight and thirty days after germination. This took place from February 12 to 28 (

Table 4. Description of the flight dates and the corresponding days post sowing (DPS) of the commercial areas.

Flight	Caballito 1		Caballito 2		García		Santa Julia		Totora		Zapote
	DPS	Phenology	DPS	Phenology	DPS	Phenology	DPS	Phenology	DPS	Phenology	DPS	Phenology
1	67	Tillering	67	Tillering	67	Tillering	*	*	83	Heading stage	81	Heading stage
2	71	Panicle initiation	71	Panicle initiation	71	Panicle initiation	*	*	87	Heading stage	85	Heading stage
3	81	Heading stage	81	Heading stage	81	Heading stage	89	Heading stage	97	Flowering stage	95	Flowering stage
4	85	Heading stage	85	Heading stage	85	Heading stage	93	Flowering stage	101	Flowering stage	99	Flowering stage
5	94	Flowering stage	94	Flowering stage	94	Flowering stage	102	Flowering stage	110	Milk stage	108	Milk stage
6	98	Flowering stage	98	Flowering stage	98	Flowering stage	106	Milk stage	114	Milk stage	112	Milk stage
7	109	Milk stage	109	Milk stage	109	Milk stage	117	Milk stage	125	Dough stage	123	Dough stage
8	113	Milk stage	113	Milk stage	113	Milk stage	121	Milk stage	129	Dough stage	127	Dough stage
9	129	Dough stage	129	Dough stage	129	Dough stage	137	Dough stage	145	Maturity	*	*
10	133	Dough stage	133	Dough stage	133	Dough stage	141	Maturity	149	Harvest	*	*

* There were no flights conducted using UAVs.

). The reproductive development phase started when the first signs of flowers appeared, at 89 DPS (Figure S1). Harvesting took place in May 2023, between days 145 and 150 DPS. Fertilizer rates differed for the Zapote crop (265-72-50) and the other areas (263-92-75). Fertilizer application was divided as the following: before transplantation, 5% N, 85% P, and 100% K were applied; eighteen days after transplantation, 42% N and 15% P were administered; and the remaining nitrogen fertilizer was distributed equally during the tillering stage and at the cotton point. Figure S1 provides more details on the flights conducted in both the experimental and commercial areas, according to crop phenology.

During the growth of the crops, we conducted insecticide applications to control the presence of Chironomus sp. And Hydrellia wirthii in the vegetative stage [3]. As the crops reached the milky grain stage, we applied fungicides to prevent diseases caused by the fungus Villosiclava virens.

2.3. Field Measurements

2.3.1. Thermal Imaging and Processing

High-resolution thermal images were gathered using a Zenmuse H20T multifunctional digital thermal imager (precision range of $\pm$2 °C) from 11 February to 2 June 2023. This camera can capture both thermal and optical images in RGB, as well as radiometric TIF formats. Its temperature measurement range spans from 8 to 14 $\mathsf{\mu }\mathrm{m}$, with a thermal resolution of 640 × 512 pixels. The camera is equipped with a built-in uncooled Vox microbolometer sensor, 40.6° DFOV lens, focal length of 13.5 mm (equivalent to 58 mm), an aperture scale of f $1{.0}^{-1}$, and a focus of 5 m at $\infty$. It also features a laser rangefinder with an accuracy of ±(0.2 m + D × 0.15%), with D denoting the distance to a vertical surface. The H20T thermal camera was mounted on a DJI gimbal on the Matrice 300 RTK UAV. The camera and the UAV both operate at a temperature range of −20 °C to 50 °C. The UAV Matrice 300 RTK offers a GPS displacement precision of $\pm$0.1 m vertically and $\pm$0.3 m horizontally, as well as RTK positioning accuracy and a maximum tilt angle of 30 °C.

The indirect calibration of the thermal images obtained by the H20T camera was performed by comparing them with the data measured with a radiometer in nine coverages. To rescale the temperature values in the thermal images, an Apogee MI-210 radiometer (MI-210; Apogee Instruments, Inc., Logan, UT, USA) was used. This radiometer was used on nine known coverslips, including aluminum, dry leaves, green leaves, expanded polystyrene, yellow cloth, black cloth, red cloth, green cloth, and bare soil. These covers were placed on a 1 × 1 m PVC square. Thermal images were captured at a height of 45 m in the experimental area and at 100 m in the commercial areas. Flights in the experimental area took place between 09:00 and 15:00 h, while in the commercial areas they were carried out from 10:30 to 17:00 h. All flights were scheduled on clear days to minimize the presence of shadows.

The average of the coefficients of determination (${\mathrm{R}}^2$) were obtained in comparison with the radiometer data and the thermal images, with the objective of identifying the possible combinations of the 8, 7, 6, 5, 4, 3, 2, and 1 coverage. This procedure was carried out for INIA-Vista Florida (blue boxplot) and the commercial areas (green boxplot) (Figure S2). It was observed that the use of multiple coverages, in this case 9, contributed to establishing a temperature gradient that improved the robustness and accuracy of crop temperature in rice fields.

However, the use of four coverages generated a minimum percentage difference of 3% to 4% in the average ${\mathrm{R}}^2$ with respect to nine coverages. This would allow a reduction in data collection time. This is in agreement with Machaca et al. [27], who used four coverages (bare soil, leaf litter, grass, and thawed water) with an ${\mathrm{R}}^2$ of 0.97, and Torres-Rua [28], who used three coverages (bare soil, short green canopy, and tall canopy) with an ${\mathrm{R}}^2$ between 0.58 and 0.95.

For each flight day, an adjustment equation was derived (Figure 3). For the images collected at INIA with meteorological conditions (Figure 4), a general equation was established ${\mathrm{Y}}_{\mathrm{radiometer} }=0{.6638\mathrm{X}}_{\mathrm{H}20\mathrm{T}}+12.615$, with a Pearson’s coefficient® of 0.959 and a root mean square error (RMSE) of 6.309 °C. Similarly, for the nearby commercial areas (Caballito 1, Caballito 2, Totora, and Zapote) (Figure 2d), a second general equation was established ${\mathrm{Y}}_{\mathrm{radiometer}}=0{.9921\mathrm{X}}_{\mathrm{H}20\mathrm{T}}+4.416)$, with a Pearson’s coefficient® of 0.960 and a root mean square error (RMSE) of 5.544 °C. This general equation was also applied to the commercial area of Garcia, which experienced similar climatic conditions (Figure 5).

In the commercial area of Santa Julia, the final area in the UAV data collection, small fluctuations in wind speed (WS) and solar radiation (SR) were observed, influenced by the order in which the data were collected. It is noted that the air temperature (${\mathrm{T}}_{\mathrm{a}}$) and relative humidity (RH) in Santa Julia are similar to the conditions recorded in the other commercial areas. In this regard, Dong et al. [29] analyzed the impact of various meteorological variables (${\mathrm{T}}_{\mathrm{a}}$, RH, WS, and SR) from 1961 to 2020, revealing that the values of ${\mathrm{T}}_{\mathrm{a}}$ and RH used to calculate the vapor pressure deficit (VPD) are the variables with the greatest impact in determining the crop water stress index (CWSI).

The temperature orthomosaics were generated using Pix4Dmapper Pro 4.4.12 software (Pix4D S.A., Prilly, Switzerland). Mainly, three standard procedures were performed consisting of the following: alignment of the photographs, generation of a point cloud, and creation of orthomosaics with digital surface models (DSMs). Then, the resolution was homogenized by the resample method at 3.5 and 1.5 cm ${\mathrm{pixel}}^{–1}$ for the experimental and commercial areas, respectively. In addition, a supervised classification (maximum likelihood) of the images was performed using QGIS 3.32 software to obtain the vegetation mask for each orthomosaic generated.

2.3.2. Stomatal Conductance

A direct measurement sensor, Decagon Model SC-1 porometer (METER Group, Pullman, WA, USA), was used to measure stomatal conductance (Gs) and crop temperature (${\mathrm{T}}_{\mathrm{c}}$). This device allows the measurement of Gs in a range from 0 mmol ${\mathrm{m}}^{-2} {\mathrm{s}}^{-1}$ to 1200 mmol ${\mathrm{m}}^{-2 }{\mathrm{s}}^{-1}$, with an accuracy of 10%. In addition, it is capable of operating in a temperature range of 5 °C to 40 °C, with a maximum relative humidity of 100%. The measurements were performed minutes prior to the execution of the UAV flight plan between 9:00 a.m. and 12:00 p.m. The sensor head was placed on the leaf surface of three plants in each subplot with three replicates. From the data collected in the field, a specific value scale relating rice stomatal conductance to the crop water stress index was developed. This approach is important as there are reference scales for other crops, such as beans [30], to monitor whether the crop is within the range of optimal water conditions.

2.3.3. Meteorological Conditions

The collection of meteorological data is fundamental to calculating the crop water stress index (CWSI), highlighting the importance of air temperature and relative humidity for this calculation [11]. Therefore, meteorological data for the UAV flight dates were obtained from the ATMOS 41 portable automatic weather station in the experimental area (Table S1) and in the commercial areas (Table S2), which have SDI-12 digital communication capabilities to transmit data from environmental sensors. This weather station, installed at an altitude of 2 m, was connected to a model ZL6 data logger. Data collected included air temperature (${\mathrm{T}}_{\mathrm{a}}$, °C), relative humidity, (RH, %), solar radiation, (SR, W ${\mathrm{m}}^{-2}$), and wind speed (WS, m ${\mathrm{s}}^{-1}$). The station was located at the center of each study area, at six different points during the course of the day.

2.3.4. Determination of the Crop Water Stress Index (CWSI)

For the temperature data collection, we employed the H20T thermal camera integrated into the UAV, following the methodology used in Section 2.3.1. A mask was subsequently created to isolate leaf-only temperatures, excluding water bodies.

The atmospheric vapor pressure deficit (VPD, kPa) was calculated from air temperature and relative humidity using a portable weather station [31], Equation (1):

$$\mathrm{VPD}=\left(0.6108{ \times 10}^{\left(\frac{17.27{ \times \mathrm{T}}_{\mathrm{a}}}{237.3{+\mathrm{T}}_{\mathrm{a}}}\right)}\right)\times \left(1-\frac{\mathrm{HR}}{100}\right)$$

(1)

where ${\mathrm{T}}_{\mathrm{a}}$ is the air temperature (°C) and RH is the relative humidity (%).

Reference baselines were calculated according to [32]. The lower limit (LL) values were calculated from the baseline of the unstressed crop, which was derived from the linear relationship of the temperature difference and the vapor pressure deficit (VPD), according to Equation (2), and the upper limit (UL) values in conditions of minimal or non-existent transpiration were determined with different maximum temperatures or remained constant at different levels of VPD [33]. The coefficients a and b were determined by the linear regression of the scatterplot between $\left({\mathrm{T}}_{\mathrm{c}}-{\mathrm{T}}_{\mathrm{a}}\right)$ vs. VPD Equation (2).

$${\left({\mathrm{T}}_{\mathrm{c}}{ - \mathrm{T}}_{\mathrm{a}}\right)}_{\mathrm{LL}}=\mathrm{a} \times \mathrm{VPD}+b$$

(2)

$${\left({\mathrm{T}}_{\mathrm{c}}{ - \mathrm{T}}_{\mathrm{a}}\right)}_{\mathrm{UL}}=\mathrm{K}$$

(3)

The CWSI according to [34] is defined as the following:

$$\mathrm{CWSI}=\frac{{\left({\mathrm{T}}_{\mathrm{c}}-{\mathrm{T}}_{\mathrm{a}}\right)}_{\mathrm{m}}-{\left({\mathrm{T}}_{\mathrm{c}}-{\mathrm{T}}_{\mathrm{a}}\right)}_{\mathrm{LL}}}{{\left({\mathrm{T}}_{\mathrm{c}}-{\mathrm{T}}_{\mathrm{a}}\right)}_{\mathrm{UL}}-{\left({\mathrm{T}}_{\mathrm{c}}-{\mathrm{T}}_{\mathrm{a}}\right)}_{\mathrm{LL}}}$$

(4)

where $\left({\mathrm{T}}_{\mathrm{c}}-{\mathrm{T}}_{\mathrm{a}}\right)$ is the difference between crop temperature and air temperature.

Various studies on the CWSI in rice cultivation, considering diverse environmental conditions, irrigation methods, soil types, thermal sensors, and threshold temperatures, are presented in

Table 5. Rice crop water stress index (CWSI) values from various studies.

Climate Classification Köppen–Geiger	Irrigation System	Soil	Thermal Sensor	Threshold Temperature	CWSI	Study Area	Reference
Wet continental	Controlled non-flooded irrigation (CI)	Sandy clay	Infrared camera (FLIR E8, EE. UU.)	Twet: rice leaves were sprayed with water 1 min before measuring Tdry: T + 5 °C	0.1–0.98	Heilongjiang, China	[14]
Warm summer	Flooded, AWD	–	Infrared camera (Ti-125 FLUKE, EE. UU.)	Twet, Tdry estimated according to estimated according to [35]	0.2–0.8	Shenyang, China	[36]
Coastal desert	Drip	Sandy loam	Infrared camera (FLIR E60, FLIR Systems Inc., Suecia) “T” type thermocouple (OMEGA, model TT-T-36-SLE-500)	${\mathrm{T}}_{\mathrm{wet}}$: 15.3 °C ${\mathrm{T}}_{\mathrm{dry}}$: 33.7 °C	0.1–0.7	La Molina, Peru	[37]
Coastal desert	Drip, AWD	Sandy loam	Infrared camera (FLIR E60, FLIR Systems Inc., Suecia) “T” type thermocouple (OMEGA, model TT-T-36-SLE-500)	${\mathrm{T}}_{\mathrm{wet}}$: measured from 07:40 a.m. to 08:15 a.m. ${\mathrm{T}}_{\mathrm{dry}}{=\mathrm{T}}_{\mathrm{a}}+18.8$ °C	0.2–0.7	La Molina, Peru	[23]
Semiarid	Flooded, AWD	–	Infrared thermal camera R300 (Nippon Avionics Co., Ltd., Japan)	$\mathrm{VPD} \mathrm{vs}. {\mathrm{T}}_{\mathrm{c}}-{ \mathrm{T}}_{\mathrm{a}}$	0.17–0.768	Montpellier, France, and Banfora and Farako-Ba, Burkina Faso	[38]
Humid subtropical	Flooded, AWD	Clayey	Infrared camera (Therma CAM SC20008, EE.UU.)	${\mathrm{T}}_{\mathrm{wet}}$$: \mathrm{average} \mathrm{of} \mathrm{lowest} {\mathrm{T}}_{\mathrm{c}}$ ${\mathrm{T}}_{\mathrm{dry}}{=\mathrm{T}}_{\mathrm{a}}+5$ °C	0.3–0.58	Kunshan, Jiangsu, China	[39]
Humid subtropical	Flooded, AWD	Clay loam	Infrared camera (FLIR C2, FLIR Systems, Wilsonville, OR EE. UU.)	$\mathrm{VPD} \mathrm{vs}. {\mathrm{T}}_{\mathrm{c}}-{ \mathrm{T}}_{\mathrm{a}}$	−0.015–0.349	Tokio, Japan	[40]
Humid subtropical	Flooded, AWD	–	Infrared camera (Tau2-640, FLIR Systems Inc., EE.UU.)	${\mathrm{T}}_{\mathrm{wet}}$$: {\mathrm{T}}_{\mathrm{c}}$$({\mathsf{\theta }}_{\mathrm{cc}}$) ${\mathrm{T}}_{\mathrm{dry}}$$: {\mathrm{T}}_{\mathrm{c}}$$(30-40\% {\mathsf{\theta }}_{\mathrm{cc}}$)	0.0–1.0	Zhejiang, China	[41]
Humid subtropical	Flooded, AWD	–	Infrared camera (Jenoptik VarioCam, Alemania)	${\mathrm{T}}_{\mathrm{wet}}$$: 5\% \mathrm{lower} {\mathrm{T}}_{\mathrm{c}}$ ${\mathrm{T}}_{\mathrm{dry}}$$: 5\% \mathrm{of} \mathrm{highest} {\mathrm{T}}_{\mathrm{c}}$	0.0–1.0	Nueva Delhi, India	[42]
Tropical monsoon	Flooded, AWD	Sandy clay loam	Handheld infrared camera (FLIR i7, FLIR Systems Inc.)	$\mathrm{VPD} \mathrm{vs}. {\mathrm{T}}_{\mathrm{c}} –{ \mathrm{T}}_{\mathrm{a}}$	−0.25–1	Laguna, Philippines	[15]
Tropical monsoon	Flooded, AWD	–	MODIS (Land Surface Temperature) sensor	${\mathrm{T}}_{\mathrm{wet}}$$: {\left({\mathrm{T}}_{\mathrm{c}} –{ \mathrm{T}}_{\mathrm{a}}\right)}_{\mathrm{LL}}$$, {\mathrm{T}}_{\mathrm{dry}}$$: {\left({\mathrm{T}}_{\mathrm{c}}{ - \mathrm{T}}_{\mathrm{a}}\right)}_{\mathrm{UL}}$	0.2–0.8	Gemawang, Indonesia	[43]
Tropical savanna	Drip, AWD	Clayey	Infrared radiometer SI-111-SS (Apogee Instruments In., Logan, EE.UU.) in UAVs	${\mathrm{T}}_{\mathrm{wet}}$$, {\mathrm{T}}_{\mathrm{dry}}$ estimated according to [34]	0.22–0.71	Kununurra, Australia	[44]

. The highest CWSI values were observed in regions with humid subtropical climates, specifically in Kunshan, Jiangsu, Heilongjiang (China), and New Delhi (India).

3. Results

3.1. The Weather Conditions in the Field

Figure 4 illustrates the meteorological conditions during the study period in the experimental area, while Figure 5 provides information on the commercial areas. The ambient temperature ranged from 25 °C to 31 °C. The wind speed varied between 1 m ${\mathrm{s}}^{-1}$ and 5 m ${\mathrm{s}}^{-1}$, with fluctuations primarily caused by wind gusts in the study area. Relative humidity exhibited high variability, ranging from 55% to 95%. Solar radiation showed a range of 100 W ${\mathrm{m}}^{-2}$ to 1200 W ${\mathrm{m}}^{-2}$, with the lowest values observed in Santa Julia due to the specific time interval of the UAV flight, which occurred between 14 and 17 h.

3.2. Baseline Condition with the Presence of Water Stress (UL) and without the Presence of Water Stress (LL)

The highest temperature was recorded between 12:00 and 14:00, yielding the maximum CWSI value. The timing of field measurements significantly influences results [37,40]. Scatter plots established the relationship between the VPD and $\left({\mathrm{T}}_{\mathrm{c}}-{\mathrm{T}}_{\mathrm{a}}\right)$ to determine limits ${\left({\mathrm{T}}_{\mathrm{c}}-{\mathrm{T}}_{\mathrm{a}}\right)}_{\mathrm{LL}}$ and ${\left({\mathrm{T}}_{\mathrm{c}}{-\mathrm{T}}_{\mathrm{a}}\right)}_{\mathrm{UL}}$, which were obtained through measurements between 9:00 and 10:30 and between 11:00 and 13:00, respectively (Figure 6).

The data used for this analysis were gathered from both experimental and commercial areas in the Lambayeque region. To ensure the accuracy of the measurements, additional experiments were conducted in La Molina, where thermocouple sensors were placed in pots. The UP was determined to be 8.5 °C, while the LL is represented by the equation $\left[\mathrm{LL}=-2.043 \times \mathrm{VPD}-1.434\right]$. By analyzing the images captured during flights 9 and 10 in the experimental and Zapote commercial areas, respectively, the LL representing optimal conditions without stress was identified. On the other hand, the UL, which represents maximum stress conditions, was obtained from images collected in treatment ${\mathrm{AWD}}_{20}$ during Flight 06-INIA.

3.3. CWSI

The variations in average crop temperatures during different phenological stages at various flight dates are depicted in Figure 7a,b. These data reveal significant fluctuations in environmental conditions throughout the crop cycle. In Figure 7a, it can be observed that crop temperatures increased during flights 5 and 6. This was attributed to the occurrence of Cyclone Yaku [45] and the insufficient water supply from irrigation canals. However, a decrease in temperature values was observed starting from flight 7, which occurred towards the end of the vegetative phase and the beginning of the reproductive phase. This decline can be attributed to the occurrence of rainfall, reaching up to 26.8 mm da${\mathrm{y}}^{-1}$ in the days preceding the monitoring, as well as the application of irrigation in the experimental plots. Conversely, Figure 7b illustrates an increase in crop temperatures from flight 8 onwards. This rise in temperatures is connected to the ripening phase of the crop, which necessitates the drying process of the grains before harvesting.

The boxplots in Figure 8a–d display the CWSI values for the experimental area, ranging from 0 to 1. Figure 8a illustrates the lowest CWSI values, which resulted from continuous flood irrigation CF. Conversely, Figure 8d depicts the highest CWSI values, attributed to treatment ${\mathrm{AWD}}_{20}$, which had a reduced water sheet.

The maximum average CWSI values for the commercial areas are displayed in Figure 9a–f. The Caballito 1 and Caballito 2 areas had maximum values of 0.41 and 0.38, respectively, at 113 DPS. Zapote had a maximum value of 0.47 at 127 DPS. In the Garcia area, the maximum value was 0.40 at 84 DPS. The Santa Julia area recorded a maximum value of 0.51 at 93 DPS. The range of maximum CWSI averages varied from 0.38 to 0.51, with a DPS range of 84 to 113. Notably, the highest value was observed in Santa Julia with the INIA 510—Mallares variety, while the lowest value was recorded in Caballito 2 with the INIA 515—Capoteña variety.

3.4. The Relationship between Stomatal Conductance and the Crop Water Stress Index (CWSI)

Figure 10a depict the correlation between crop temperature (${\mathrm{T}}_{\mathrm{c}}$) and stomatal conductance (Gs), both measured with a porometer, obtaining a coefficient of determination (${\mathrm{R}}^2$) of 0.92. In Figure 10b, the correlation between the CWSI and Gs is presented, resulting in an ${\mathrm{R}}^2$ value of 0.83. Both graphs demonstrate a pronounced Pearson correlation, with coefficients of 0.96 and 0.91, respectively.

Based on the evaluation of the data collected, a reference scale is proposed to determine water stress levels from stomatal conductance (Gs) values (Figure 11). This scale was elaborated upon by taking into account previous research [30]. The equation presented in Figure 10b was used to calculate the reference scale. This proposal is based on the relationship between the crop water stress index (CWSI) and Gs values. In this context, higher Gs values, obtained in the experimental area and in the commercial areas with continuous flood irrigation (CF), indicate luxury transpiration. In contrast, the lowest values, obtained in the ${\mathrm{AWD}}_{20}$ treatment, indicate moderate to extreme stress.

The variation in Gs in the INIA-Vista Florida experimental area, indicating crop transpiration levels, is depicted in Figure 12. A total of 14 flights were conducted, with the first six flights occurring during the vegetative phase on days 38, 42, 61, 65, 75, and 79 DPS. Flights 7–10 correspond to the reproductive phase on days 88, 92, 103, and 107 DPS, while flights 11–14 were conducted during the maturation phase on days 123, 127, 147, and 149 DPS. Figure 12a–d show that transpiration levels remained within the optimal range due to adequate rainfall and irrigation prior to each monitoring. However, Figure 12e,f clearly indicate that the crop experienced water stress during the transition from the vegetative to the reproductive phase, primarily due to insufficient water availability. Flights 13 and 14 (Figure 12m,n) captured the crop in the ripening phase, and it faced a period of drought before harvest, leading to increased temperatures.

The graph presented in Figure 13 depicts the changes in Gs in the commercial area of Santa Julia. In the initial four flights (89, 93, 102, and 106 DPS), most of the plots within this area maintained an optimal range. However, two plots were identified as experiencing water stress due to soil salinity issues, resulting in a lack of crop transplantation. In the final four flights (117, 121, 137, and 141 DPS), which correspond to the crop maturation stage, a noticeable shift in Gs was observed. The levels reached moderate and mild stress, particularly during the waxy grain development phase of the crop.

The additional information shows that there is observable diversity in stomatal conductance within the commercial areas of Caballito 1, Caballito 2, Garcia, Totora, and Zapote, as depicted in Figures S3–S6.

3.5. The Yields of Crops in Both the Experimental and Commercial Areas

According to Figure 14, it can be observed that significantly higher rice yields ${(\mathrm{t} \mathrm{ha}}^{-1})$ were obtained in the experimental area compared to the commercial areas. The treatments that recorded the highest and lowest yields for the experimental area are CF and ${\mathrm{AWD}}_5$, respectively.

4. Discussions

Variations in thermal data obtained from infrared thermal cameras can occur due to factors such as wind speed (WS), solar radiation (SR), and atmospheric interferences, resulting in temperature differences in several degrees Celsius [46]. To ensure the accurate determination of the crop water stress index (CWSI), a portable weather station was utilized to obtain reliable meteorological data. This is important because, as highlighted by Gonzalez-Dugo et al. [11], precise measurement of environmental conditions is crucial for calculating the CWSI, especially since meteorological data in many studies are often obtained from remote stations. Furthermore, the temperature values obtained from thermal images were adjusted using information from an Apogee MI-210 radiometer (Figure S5), demonstrating a favorable correlation between the temperatures captured by the H20T thermal camera and the radiometer, as recommended by Torres-Rua [28]. Moreover, Ribeiro-Gomes et al. [47] discuss the impact of UAV flight altitude on thermal images, as the measured temperature is negatively correlated with altitude.

A humid temperate climate is characterized by a high level of relative humidity (RH), resulting in low vapor pressure deficit (VPD) conditions. It is crucial to establish a reference baseline for water stress (LL) based on the specific environmental conditions [40]. When determining the LL empirically, there is an observed increase in the uncertainty of the crop water stress index (CWSI), considering the sensitivity of rice varieties to VPD when selecting the LL. In an arid desert climate, the humidity influences the LL, leading to higher uncertainty in the CWSI [23]. In our study, the LL was calculated using morning data and days with a lower RH were taken into account. It was found that the highest daily water stress occurred at midday in the water deficit treatments ${\mathrm{AWD}}_5{, \mathrm{AWD}}_{10}$, and ${\mathrm{AWD}}_{20}$, which aligns with a previous study conducted by Ref. [39].

The irrigation method known as alternate wetting and drying (AWD) has a direct impact on the temperature of crops, and it was expected that different levels of water stress would elicit a response. As a result, the highest crop temperature (${\mathrm{T}}_{\mathrm{c}}$) values were recorded with ${\mathrm{AWD}}_{20}$ irrigation, while the lowest value was observed with traditional irrigation. Furthermore, the values of the crop water stress index (CWSI) exhibited variations ranging from 0 to 1. During the tillering phase, CWSI values remained high (0.2 and 1), indicating water stress. Conversely, during the reproductive phase, CWSI values were lower. However, during the maturation phase, particularly at the time of a pasty and waxy grain, CWSI values increased to 0.8, indicating a water deficit. It is worth noting that the behavior of the CWSI was different in the commercial areas of Caballito 1, Caballito 2, and García. When comparing the CWSI values obtained with those documented by the authors in Table 5, it is evident that areas with AWD irrigation regimes and similar soil texture exhibited a comparable range of CWSI values.

The findings indicate that the diurnal pattern of the CWSI is influenced by both the evaporative demand of the air and the significant increase in solar radiation (SR) and air temperature $\left({\mathrm{T}}_{\mathrm{a}}\right)$ during midday. These conditions resulted in a higher rate of crop transpiration compared to the rate of soil moisture uptake by roots [39]. Consequently, a water deficit occurred to some extent, leading to high CWSI values in both the experimental and commercial areas. It is worth noting that the lowest CWSI values were observed during the cotton point and initial flowering phase in both the experimental and commercial areas, which was attributed to the impact of rainfall from cyclone Yaku [45] and abnormal precipitation before [48] the UAV flight.

In the experimental area, various irrigation regimes were implemented. These different regimes had a significant impact on stomatal conductance (Gs) and crop yield. The treatment with the lowest water stress displayed a higher Gs and yield, as evidenced by Figure 8, Figure 12, and Figure 14, when compared to the other treatments. In contrast, the treatment with the highest water stress, ${\mathrm{AWD}}_{20}$ (Figure 12f), exhibited a lower Gs. These findings contradict a previous study on high-yielding rice crops [18,49], where a decrease in Gs and yield was observed. It is worth noting that this decrease occurred at different levels of volumetric soil water content (VWC) for each variety, indicating varying responsiveness to stress conditions. These variations in crop responses highlight the complexity of adaptation mechanisms to water stress, suggesting the existence of diverse coping strategies.

As a result of this research, the integration of UAV thermal imagery and CWSI maps in agriculture offers several practical applications that can improve irrigation and crop management practices in rice fields. By utilizing these technologies, agriculturalists can apply specific irrigation methods that are customized to the unique requirements of each section of their fields. This optimizes water consumption and enhances the productivity of their crops.

Early detection of water stress becomes a tool for timely decision making. By monitoring CWSI variations in crops, growers can detect signs of water stress at an early stage. Also, understanding the relationship between the CWSI and stomatal conductance gives growers a quick understanding of plant physiological processes related to transpiration. This allows for quick interventions, such as adjustments in irrigation schedules or the application of water conservation techniques to reduce potential yield losses.

5. Conclusions

According to these research findings, the implementation of the AWD irrigation technique results in a noteworthy elevation in the rice crop water stress index (CWSI) in contrast to the continuous flooding (CF) method (Figure 8). Despite this rise in CWSI in the experimental area utilizing AWD irrigation, a greater yield was observed compared to the commercial areas employing CF irrigation (Figure 14). These results imply that AWD irrigation may present a more sustainable and efficient alternative for water conservation purposes.

These results also indicate a significant correlation between the crop temperature ${\mathrm{T}}_{\mathrm{c}}$ and stomatal conductance Gs, as well as between the CWSI and Gs (Figure 10). This suggests that as water stress increases, stomatal conductance decreases. These findings, combined with the use of UAVs to capture high-quality images, can serve as a technological tool for better monitoring and managing water stress in rice crops. By utilizing the relationship between the CWSI and Gs, the developed tool has the potential to greatly enhance irrigation management efficiency, enabling more precise irrigation decisions and optimizing water usage in rice crops.

In future work, the frequency of UAV flights should be increased and their application extended to a larger number of rice fields to obtain a better characterization of the heterogeneity of these fields. In addition, it would be beneficial to integrate satellite data so that farmers can access information on their crops in real time. On the other hand, an economic evaluation is needed for the adoption of these technologies in order to achieve cost-effectiveness.