ASABE Logo

Article Request Page ASABE Journal Article

Diurnal Trends of Maize Canopy Cover Under Water Stress

Kendall DeJonge1,*, Huihui Zhang1, Liam Cummins1†,2, Tyler Gilkerson1†,3, Katherine Ascough1†,4, Tyler Pokoski1

Published in Journal of Natural Resources and Agricultural Ecosystems 2(2): 77-89 (doi: 10.13031/jnrae.15792). 2024 American Society of Agricultural and Biological Engineers.

1    Water Management and Systems Research Unit, USDA ARS, Fort Collins, Colorado, USA.

2    Colorado Division of Water Resources, Denver, Colorado, USA.

3    National Park Service, Fort Collins, Colorado, USA.

4    Wyoming Department of Transportation, Cheyenne, Wyoming, USA.

*    Correspondence: kendall.dejonge@usda.gov

†    Former affiliation.

The authors have paid for open access for this article. This work is licensed under a Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 International License https://creative?commons.org/licenses/by-nc-nd/4.0/

Submitted for review on 24 August 2023 as manuscript number NRES 15792; approved for publication as a Research Article and as part of the “Digital Water: Computing Tools, Technologies, and Trends” Collection by Associate Editor Dr. Sushant Mehan and Community Editor Dr. Kati Migliaccio of the Natural Resources & Environmental Systems Community of ASABE on 15 January 2024.

Mention of trade names or commercial products in this publication is solely for the purpose of providing specific information and does not imply recommendation or endorsement by the USDA. The USDA is an equal opportunity provider and employer.

Highlights

Abstract. By differentiating green and non-green pixels of RGB images of crop canopy, the fractional canopy cover (fc,%) can be estimated and subsequently used to estimate crop transpiration demand (e.g., Kcb, basal crop coefficient, following FAO-56 methodology). While unstressed maize crops maintain consistent fc through the day, leaves of heat or water stressed crops will curl, thereby reducing the instantaneous fc and the associated transpiration. This paper reports the effects of diurnal leaf curl and subsequent fc reduction on full and limited water treatments of irrigated maize through nadir images obtained every 15 minutes. Results suggest that fc images obtained in the early morning (fcNS) are typically representative of a non-stressed condition and can be used to estimate Kcb. Reduction of fc throughout the day (i.e., fc/fcNS) was compared to water stress coefficient Ks obtained from neutron probe and TDR measurement and showed that fc/fcNS from hours 11 a.m. to 2 p.m. had a high correlation (R2 = 0.773), indicating that mid-day fractional canopy cover reduction could be used as a proxy for Ks. This method creates new opportunities for estimating crop water stress during vegetative growth but also highlights potentially unknown challenges regarding lower resolution remote sensing during the same growth stages.

Keywords.Basal crop coefficient, Canopy cover, Crop coefficient, Evapotranspiration, Stomatal conductance, Stress crop coefficient, Transpiration, Water stress.

Given the rising demand for agricultural products and the uncertainty of water supplies, advancing methods and tools for estimating evapotranspiration (ET) is a critical priority to ensure future crop productivity, as highlighted by Fischer and Connor (2018) and Harmel et al. (2020). One of the most established methods to estimate or calculate ET in agricultural systems is the crop coefficient method (Allen et al., 1998), where a reference crop ET (ETref) is multiplied by a crop coefficient (Kc) to estimate crop ET (ETc). Reference ET (ETref) can either be for a short (e.g., grass) surface or a tall (e.g., alfalfa) surface, represented by EToand ETr. In the expanded but accurate dual crop coefficient method, Kcis partitioned into a basal crop coefficient (Kcb) and a wet soil evaporation component (Ke). The basal crop coefficient (Kcb) is defined as the ratio of crop ET over the reference ET (ETc/ETref) when the soil surface is dry and Ke = 0, but the current availability of soil water does not limit plant growth or transpiration. The Kcb is based on the current state of growth, plant development, and full water availability and is independent of past conditions. To account for conditions of water stress, Kcb can be multiplied by a water stress factor (Ks), with values of 1 indicating the absence of stress and 0 indicating the complete shutdown of transpiration. When adjusting for nonstandard conditions, which consider the effects of water stress, ETc is denoted as ETcact. In alignment with the tall crop reference standard used in Colorado and other regions, the equation can be represented as:

        (1)

Past studies have shown the basal crop coefficient Kcb to be related to the interception of solar radiation, as a function of fractional canopy cover (fc) (Allen and Pereira, 2009; Pereira et al., 2020), and models such as AquaCrop use fc as a basis for the calculation of a crop transpiration coefficientand subsequent crop transpiration (Steduto et al., 2009; Raes et al., 2023). This concept is fairly simple under well-watered conditions, where even a trapezoidal time series curve can be useful to generalize Kcb, and some versions of equation 1 are provided where Ks = 1 and the crop experiences no water stress (Allen et al., 1998; ASCE, 2016). Current water stress can reduce transpiration in the short term (i.e., Ks < 1), and prior accumulated water stress can reduce above ground biomass and fc, subsequently rendering the trapezoidal “well-watered” FAO-56 curve less accurate than Kcb calculated from direct measurement of fc (Trout and DeJonge, 2017). Thus, water stress can reduce crop transpiration both in the short and long-term, and time series of Kcb for well-watered and water-limited crops can be fundamentally different.

More recent studies have used fc and/or other remote sensing measurements of the canopy to inform the estimation of ET under water stress (DeJonge et al., 2016; Kullberg et al., 2017; Bellvert et al., 2018; Tang et al., 2019; Zhang et al., 2023). Kcbcurves as a function of fc have been defined for maize (Trout and DeJonge, 2018) and various other crops (Pereira et al., 2020). Kcbvalues from this technique were used to estimate the reduction of transpiration as the water stress term Ks(Trout and DeJonge, 2021). Formulations to calculate Ksinclude a linear reduction in transpiration as deficit exceeds the Readily Available Water (RAW) as suggested by FAO-56 (Allen et al., 1998), but more recently an exponential decay function was used, e.g., as defined by AquaCrop (Steduto et al., 2009) and reinforced by other studies (Trout and DeJonge, 2021; Terán-Chaves et al., 2022).

The response of plants to water stress is mainly reflected in leaves and roots, with the former being highly variable in terms of the environment, showing signs of etiolation, atrophy, curling, senescence, and even abscission (Wu et al., 2022). Plant-based indicators for irrigation scheduling often require an understanding of the dynamics of plant water status and can be challenging to determine amongst diurnal changes (Fereres and Goldhamer, 2003). For example, past studies have documented diurnal changes in chlorophyll fluorescence in sunflower and Chinese hibiscus (Amoros-Lopez et al., 2008), and potato and sugar beets (Wang et al., 2021), and diurnal transpiration of maize showed marked reductions under drought stress as compared with fully-watered plants (Gleason et al., 2017). Canopy temperature also follows diurnal patterns that converge at night and diverge during the day based on varied water stress (DeJonge et al., 2015), and methods have been developed to model the diurnal pattern using one time per day measurements and a reference temperature curve (Peters and Evett, 2004). However, unlike canopy temperature, prior accumulated water stress can cause reduced canopy growth (Stevens et al., 2020) and accelerated plant senescence (Wu et al., 2022), resulting in reductions in fc and subsequently Kcb (Trout and DeJonge, 2021).

It is well understood that water stress can significantly vary crop vegetation canopy structure (Kimes and Kirchner, 1982). While many examples in the literature quantify fc from imaging, few exist that specifically quantify sub-daily diurnal changes in fc due to water stress (Stevens et al., 2020). One recent study used a qualitative system to score visual observations of maize leaf curl throughout the day using digital hemispherical photography from ground-mounted cameras facing upward, finding significant diurnal leaf curl for water-stressed maize and no leaf curl for well-watered maize (Baret et al., 2018). The authors only found one recent example that specifically quantifies diurnal fc under water stress, albeit with bananas (Stevens et al., 2020), an interesting but vastly different crop than maize. A recent maize study emphasized the need for Kcbestimates based on fc to be obtained during non-stressed or at least low stress periods, such as night or early morning, where available soil water can always meet the low evaporative demands regardless of soil moisture status (DeJonge and Zhang, 2021).

Many tools and technologies now exist that simplify the estimation of fc. As imaging and processing have become less expensive, both in cost and computing resources, the advent of smartphones and apps has led to easy and user-friendly quantification of fc, such as Canopeo (Patrignani and Ochsner, 2015), which estimates fc using a free smartphone app without requiring a computational background from the user. Image acquisition and processes for estimating fc using UAVs (unmanned aerial vehicles) have become more common (Niu et al., 2021; Yan et al., 2019), and UAV use is expanding as the rate of new agricultural UAV operators is outpacing the rate of new manned agricultural aircraft operators (Rodriguez, 2023).

Thus, the tools exist to estimate the effects of water stress on fc and, subsequently, ET. However, just as prior studies have indicated that the diurnal fluctuation of canopy temperature is consistent at night but varied among water treatments during the day (DeJonge et al., 2015), fc is also expected to have a varied response between water treatments due to water stress. Because FAO-56 (Allen et al., 1998) indicates that Kcbis related to the canopy of a well-watered (i.e., non-stressed) crop, then what are the best practices when estimating Kcbusing fc when a crop actually is stressed? Can a reduction in fc through the day be used to estimate the reduction in crop transpiration, such as the Ks term? The overarching objective of this study is to demonstrate how fc fluctuates during the day and over the season based on crop water status and provide recommendations on how this information can be used in practice for the estimation of ET.

Materials and Methods

The field experiment of maize was conducted in the 2016 field season at the USDA-ARS Limited Irrigation Research Farm (LIRF) in Greeley, CO (40°26 ´57 ? N, 104°38 ´12 ? W, 1427 m asl), as part of a longer ongoing experiment that ran from 2012-2016. The farm is surrounded by irrigated farmland and has soils of sandy and fine sandy loam. Maize (Zea mays L.) was planted on May 5 with a population of 85,250 plants ha-1. The field was strip tilled on April 15 with a sidedress of 29.1 kg ha-1 N and starter fertilizer at planting of 4.5 kg ha-1 N. Subsequent N applications were injected into the irrigation system to ensure no crop yield losses due to nutrient deficiency (total 186.6 and 141.5 kg ha-1 N for Full and Limited Irrigation, respectively). The maize field was divided into four replicate blocks, with each block divided into 12 randomly assigned plots (9 x 43 m) containing 12 crop rows (0.76 m spacing) oriented north-south. Measurements were collected from the central six rows of twelve to avoid edge effects.

Irrigation was supplied via surface drip irrigation tubing with 30 cm in-line emitter spacing (1.1 L h-1 per emitter), and irrigation amounts were controlled and recorded with Campbell Scientific CR1000 dataloggers (Campbell Scientific, Logan, UT), and measured independently with turbine flowmeters (Badger Recordall Turbo 160, Badger Meter, Milwaukee, WI). The 12 irrigation treatments had varying levels of deficit irrigation; the treatments discussed in this paper are Full Irrigation which met 100% of crop ET demand throughout the season, and Deficit Irrigation, which met 40% of ETc during late vegetative (V8-VT) and maturation (R4-R6) growth stages with 100% ETc demand during reproductive growth, between those two periods (i.e., 40/40 Treatment as described in Zhang et al., 2019a). A Colorado Agricultural Meteorological Network (CoAgMet; http://www.coagmet.com) automated weather station (GLY04) is located on a 0.4 ha irrigated grass lawn adjacent to the research plots, which was used to determine ASCE (2005) Standardized Reference ET (ETr) on an hourly basis. This ETris used with the FAO-56 dual crop coefficient method (Allen et al., 1998) to determine plot water balance based on ETc. Precipitation was measured with three onsite tipping bucket rain gauges. More specific experimental details can be found in Zhang et al. (2019a), which reported the results from 2012-2015, but the same experiment was continued in 2016.

Soil water content was measured at the soil surface (0 cm to 15 cm layer) by a portable time domain reflectometer (MiniTrase, SoilMoisture Equipment Corp., Santa Barbara, CA) and at subsequent depths (30 cm increments to 2 m) using neutron attenuation (neutron moisture meter, CPN-503 Hydroprobe, InstroTek, San Francisco, CA). The neutron moisture meter was calibrated using gravimetric samples in 2008 (R2 = 0.92) and was verified annually with gravimetric cross-calibration samples to ensure that the best fit line remained within 95% confidence intervals of the original calibration. All soil water measurements were taken in the middle row of each plot, typically before and often after irrigation, two to three times per week. Field capacity was estimated for each plot and soil layer from previously observed soil water content measurements following large rainfall or irrigation events, following gravity drainage. Soil water deficit (SWDRZ) of the root zone (RZ) was calculated as follows:

        (2)

where SWDRZ is the SWD in the root zone (assumed as 105 cm), L is an identifier for each soil layer (with L=1 beginning at the soil surface and L=LRZ ending at the root zone), DL is the depth of the layer, FCL is the FC of the layer, and ?L is the volumetric water content of the layer. More details on this process can be found in Trout and DeJonge (2017).

In 2016, images of crop canopy were obtained by a waterproof Crenova brand RGB “Trail Camera,” commonly used for capturing images in remote or harsh environments. A custom-made frame cantilevered over the crop, and the camera was mounted facing nadir downward so that no frame and only crop and soil were included in the images. The frame was raised through the season to maintain an approximate camera height of 1 m above the crop canopy during the observation period. The RGB camera with 5 Megapixel resolution (2592 x 1944) was set to acquire images every 15 minutes, day and night, from one fully irrigated plot and one limited irrigation plot. Nighttime images were illuminated by LED lights mounted on the camera. Camera lenses were cleaned, and images were downloaded weekly. Images were processed by an internally developed Python program, which evaluated the color hue of each pixel in the image. Pixels that were classified as green are counted as active canopy cover, and the sum of the green pixels divided by the total pixels equates to the fractional canopy cover (fc). While the process was internally developed to automate the processing of multiple images, it is very similar to that used by the Canopeo phone app (Patrignani and Ochsner, 2015).

This study evaluates the ability of decline in fc under limited water to quantify or represent the actual water availability and/or reduction in transpiration. Because soil water observation is critical to verifying crop water status, analysis was initially limited to days when soil water was collected, typically the day before an irrigation and sometimes following an irrigation event. Figure 1 illustrates the daily diurnal fluctuation of fc, which can help illustrate decision making in data screening and analysis. The results later expand on how these changes occur over a 6-day period, with irrigation occurring in the middle of the period. Finally, boxplots of all fc values during the daytime are given throughout the entire observation period (28 June-19 August, or DOY 180-232), which spanned maize growth periods V8 to R4.

The overall dataset used in this study included both full and limited irrigation treatments through several growth periods of varying canopy, environment, and level of water stress. As an initial analysis, the Regression Trees (Partition) model in JMP statistical software (JMP Version 16.2, SAS Institute Inc., Cary, NC) was used to identify key categories for evaluation that predict measured Ks. This method is a type of decision tree model used for predicting a numeric outcome variable based on multilinear regression of one or more predictor variables. The process of creating a regression tree involves recursively partitioning the data into subsets based on the values of the predictor variables.

The main categories identified for the separation of images were:

  1. Before and after irrigation. In the data collection schedule, measurement of soil water the day before (or sometimes the day of) irrigation was critical to the irrigation decisions, as it would be in practice for any farmer. Occasionally, soil water data were obtained the day following irrigation, sometimes used to verify or recalibrate estimates of field capacity. In these situations, modeling of soil water deficit would typically indicate a deficit below RAW, thus no water stress and Ks= 1. However, there is some uncertainty in these measurements since soil water under saturation may still be under flux.
Figure 1. Diurnal fractional canopy cover (fc) on 8 July 2016, in Full Irrigation (a-e) and Limited Irrigation (f-j) at 2-hour segments during the day. fc calculated from images every 15 minutes are shown in (k), and reference evapotranspiration (ETr) and solar radiation (Rs) are shown in (l). While images are available but impractical to show for all plotted points in (k), larger symbols in (k) indicate times images shown (a-j) were taken. Grey vertical lines (4:00-6:00 and 18:00-20:00) in (k-l) indicate dawn and dusk hours, respectively, where 0.001 < Rs < 0.5 MJ m-1 hr-1, when oversaturation of images causes errors. Animation of figures is available at DeJonge et al. (2023).
  1. Vegetative vs. reproductive growth. Most importantly, in the experimental structure, all treatments were irrigated to field capacity closely prior to anthesis (R1) to ensure that any yield loss was not attributed to water stress during this critical growth period (Comas et al., 2019). Thus, both treatments were essentially treated as full irrigation from growth stages VT/R1 through R3 and would have little observed leaf curl as a result. Significant hail was observed on DOY 232 as well, ending any ability to observe leaf wilt in later growth stages.
  2. Irrigation treatment. Because the full irrigation treatment was meant to maintain field capacity at each irrigation, water stress was expected to be minimal (e.g., consistency in blue dots in fig. 1), even immediately prior to irrigation. Although some soil water observations indicated a small amount of water stress for the full irrigation treatment the day before irrigation, the majority of the observations indicated Ks = 1, giving very little variability in the model evaluation.
  3. Night and low light. While some nighttime fc estimates were included in the initial analysis and show results very similar to early/late daytime fc estimates, the Python program required different threshold values for pixel color analysis (i.e., a separately calibrated program), and it is acknowledged that the collection of high-resolution RGB imaging at nighttime is impractical on most scales. Also, low sun angles create undersaturation or oversaturation of images taken at dawn and/or dusk (see fig. 1). Thus, all timestamps where the measured solar radiation (Rs) was less than 0.5 MJ m-2 hr-1 were eliminated from the analysis.
  4. Clouds. Clouds can directly and quickly reduce solar radiation, and the crop can be very responsive, as shown in the initial analysis. Most satellite-based remote sensing methods cannot even report data under cloudy conditions, and cloudy conditions are commonly used as a filtering criterion for canopy temperature studies for the same reasons (DeJonge et al., 2015). Thus, all timestamps were removed from analysis where the ratio of solar radiation to clear sky radiation (i.e., Rs/Rso) was less than 80%, indicating clouds or haze.

To obtain an overall impression of the ability of environmental inputs to predict Ks, a multilinear model was run in JMP software with several environmental variables as input variables: reduction in fractional canopy cover (fc/fcNS), fraction of clear sky solar radiation (Rs/Rso), closest hour the data was obtained (Hour), relative humidity (RH,%), solar radiation (Rs, MJ m-2 d-1), air temperature (Ta, °C), tall reference ET (ETr, mm h-1), and vapor pressure deficit (VPD, kPa).

The field capacity (FC) to hold water following drainage is typically estimated as soil water potential at 30 kPa, with the permanent wilting point (PWP) estimated as soil water potential at 1,500 kPa and assumed to be 50% of FC (Trout and DeJonge, 2021). The difference between these values (FC-PWP) is defined as the total available water (TAW). It is useful to define soil water deficit (SWD) as the amount of water required to bring current conditions back to FC. These ideas have been expanded to define a water stress coefficient Ks as a function of the relative soil water deficit (rSWD = SWD/TAW). While earlier documents such as FAO-56 (Allen et al., 1998) used a linear curve to relate Ks to SWD, this study, aligning with Trout and DeJonge (2021), adopts the curvilinear model recommended by AquaCrop (Raes et al., 2023), defined as:

        (3)

where sf, the shaping factor, is assumed to be 1.5 in this study, and the relative depletion (Drel) is defined as:

        (4)

where p is defined as a threshold rSWD above which Ksis not affected, assumed as a base value of p = 5 mm d-1 but adjusted with fluctuations in estimated ETcbased on FAO-56 (Allen et al., 1998).

As defined earlier, the basal crop coefficient (Kcb) is used to estimate the transpiration component of ET under conditions when the soil surface layer is dry (i.e., Ke = 0) but where the average soil water content of the root zone is adequate to sustain full plant transpiration (i.e., no water stress where Ks = 1). Because plant transpiration is negligible at night and minimal at the beginning and end of the day (fig. 1), preliminary observations indicated that the upper values of each day’s fc observations could represent a plant status where transpiration is not limited, even in limited irrigation prior to the decline of fc due to water stress (fig. 1). Thus, the 90th percentile of all daytime observations for the treatment of interest was used to represent an ideal non-stressed fractional canopy (fcNS) for that treatment. The 90th percentile was used rather than the maximum to allow for bias due to high-side errors. It is important to define a fcNS to be specific to each treatment rather than just using the full irrigation treatment since previous studies have shown that prior water stress can reduce canopy cover and subsequent transpiration capacity, e.g., different values typically exist for the Kcbbetween treatments (Trout and DeJonge, 2021).

Some diurnal plant measurements, such as canopy temperature, normalize at night (Peters and Evett, 2004), regardless of water status (DeJonge et al., 2015) or genotype (Balota et al., 2007), and even among plant species (Vo and Hu, 2021). However, canopy cover measured at night often varies between irrigation treatments based on prior water stress (fig. 1), and a reduction in canopy cover (canopy cover represented as Kcb) could be represented to be correlated to a reduction in transpiration (represented as Ks). Thus, the main factor assumed in this study to represent Ks(reduction in transpiration) as a function of canopy cover was a measure of the decline in fractional cover, represented by fc/fcNS. This representation allows for theoretical boundary conditions of Ks = 0 when fc = 0 (i.e., no transpiration when there is no canopy), and Ks = 1 when fc = fcNS (full transpiration with full canopy), thus any estimates of fc/fcNS > 1 (e.g., greater than the 90th percentile) were set to equal 1.

It is desirable to have simple parameters for each outcome, but the understanding in this context is that fc/fcNS can be highly variable throughout the day, which of course is due to water stress but further impacted by evaporative demand (fig. 1), which is a function of many factors influencing ETr (i.e., Rs, T, RH, and wind). ETrresulting from these factors also has a diurnal fluctuation, driven by the sun (i.e., Rs). But another goal of this study is to indicate the best times of the day to observe fc/fcNS for estimation of water stress (Ks), and it is intuitive that fc/fcNS should reach its minimum value when Rs and ETr peak, typically at mid-day. Thus, for each day of observation, the trend of Ks vs. fc/fcNS was evaluated, but by separating fc/fcNS on an hourly basis. The best model of Ksvs. fc/fcNS was then aggregated into a simple model that could be repeated for estimation of Ks, with results demonstrated herein.

Results

Similar to the findings of Baret et al. (2018), maize canopy cover is often fairly consistent throughout the day under full irrigation where there is an absence of water stress, but under deficit irrigation the crop leaves will curl during the day to slow transpiration (fig. 1). Because deficit irrigation maize had accumulated water stress prior to this day, the overall amount of biomass is less than full irrigation, as evidenced by the peak values of fc observed at late night and early day, in this case just over 80% and 70% for full and deficit irrigation, respectively. However, through the day, deficit irrigated fc dropped by around 35%, reducing canopy cover in half (from 70% to 35%), whereas fully irrigated fc only dropped by 5% (from 80% to 75%). Deficit irrigation showed a much more pronounced mid-day effect from mid-day peak transpiration demand than full irrigation treatment.

Figure 2. Diurnal canopy cover images from 12:00 pm each day 13-18 July 2016, in Full Irrigation (a-f) and Limited Irrigation (g-l). Fractional cover (fc) calculated from images every 15 minutes are shown in (m), with data omitted where solar radiation (Rs < 0.5 MJ m-1 hr-1, and irrigation and rain events noted by bars with colors associated with treatment. Reference evapotranspiration (ETr) is shown in (n). Animation of figures is available at DeJonge et al. (2023).

Expanding figure 1 to span a week instead of a day, there are not only diurnal fluctuations and effects of cumulative stress on fc but also recovery of fc following a water event (fig. 2). Like in figure 1, the fc for full irrigation stays fairly consistent, typically between 75% and 80%, whereas for deficit irrigation, there are significant reductions in fc during mid-day. The role of day-to-day water availability becomes clearer here, as indicated by mid-day fc typically declining in subsequent days without rainfall (e.g., from 13 July to 14 July and from 16 July to 17 July), as the plant extracts water from the soil, thereby incrementally increasing water stress.

Likewise, plant response from increased water through an irrigation event is observed on 15 July, where the deficit fc recovers immediately and the drop is less substantial in the days that follow. Both treatments show less response on 18 July, a lower ETr day in comparison, and following a rainfall event the night of 17 July. The general trajectory of the daytime values for fc also increased during this time of active vegetative growth, moving from approximately 80% to 85% for full irrigation and 72% to 80% for limited irrigation, of course boosted by irrigation and rain events.

Expanding figure 2 to encompass the entire image observation period (DOY 182-232), some of these same trends are observed over the course of the growing season (fig. 3). Boxplots are shown to indicate both the general value of fc throughout the day and the distribution, which indicates the amount of leaf curl during the day. Because both crops were taken out of stress at VT stage (DOY 207, indicated by the darker vertical line in fig. 3), the discussion will focus primarily on vegetative growth prior to VT.

Generally, full irrigation shows more consistent daily values for fc than limited irrigation, as shown in figures 1 and 2. Limited irrigation, however, generally has a much larger distribution in fc unless it follows a large irrigation or rainfall event and/or is paired with low ETr days. The distribution for limited irrigation generally broadens, and the overall value drops with subsequent days lacking irrigation (e.g., DOY 185-192, 193-196, 200-207). These trends are sometimes observed in full irrigation, for example, DOY 189-192, although less severe than under limited irrigation. Both treatments are responsive to irrigation and rain, with narrower boxplots at higher fc generally the day of or following a watering event. After the large irrigation event on DOY 207, fc in deficit irrigation is generally equal to fc in limited irrigation, with some days having slightly higher fc with smaller distributions in limited irrigation.

It is noteworthy that fc is responsive to not only water status but environmental variables as well, even on this timescale. For example, note the leaf curl for full irrigation between DOY 189-192, when high ETr values were generally above 8 mm d-1 during this time. A deeper look shows very high temperatures during this time, with each day from DOY 189 to 195 having a maximum temperature above 30°C, and maximum temperatures of 36.2°C and 37.4°C on DOY 191 and 192, respectively (fig. 3). Also note that on DOY 217-219, which are very low ETr days following an irrigation event, both treatments are responsive to low ET demand as well, having high values of fc with little variability (above 80%, near the peak values of previous and following days).

Figure 3. Boxplots of fractional canopy cover (fc) during daylight hours where Rs > 0.5 MJ m-2 d-1, showing daily irrigation and rainfall events (mm), ETr (mm d-1), and daily maximum Ta (°C). Vertical dashed line at DOY 207 (July 25) indicates the beginning of anthesis growth stage (VT), when both treatments were taken out of stress through high levels of irrigation.

After observing how fc can change under full and limited irrigation under daily (fig. 1), weekly (fig. 2), and seasonal (fig. 3) timescales, the influence on environmental variables and time to predict Ks was determined by running a multiple linear regression model using daily data where measured Ks values were available (table 1). As suggested by the initial JMP Regression Tree analysis, which identifies the most important factors for a continuous outcome, these models were run separately by treatment and growth stage.

Limited irrigation under vegetative growth, the model of interest, had the best fit of the four models evaluated (R2 = 0.559), whereas full irrigation under vegetative growth had a poorer fit (R2 = 0.213). This result is intuitive considering that in a full irrigation treatment the goal is to maintain Ks at or near 1, thus less variability in the dependent variable. In this study, Ks = 1 for 55% of observations in full irrigation but only 16% of observations in limited irrigation (data not shown). In all four combinations of model runs (limited or full irrigation in combination with vegetative or reproductive growth), fc/fcNS was a significant parameter, and it was the most significant parameter for both treatments under vegetative growth and both growth stages under limited irrigation. The ability of fc/fcNS to predict Ks also had the largest LogWorth value of any single predictor, in any combination of treatments and growth stages, indicating its strong ability to be used for prediction. Under vegetative growth, Rs/Rso was the next most significant parameter, indicating that screening for cloud cover is critical when evaluating these datasets, and this separation was also noted by the JMP Regression Tree Analysis. Hour was also significant under limited irrigation in vegetative growth and should be considered in the analysis.

Because the goal was to evaluate the impact of fc reduction during late vegetative growth under water stress (i.e., limited irrigation), the study focused on modeling the ability of fc/fcNS alone to predict Ks throughout the day, as separated by Hour (fig. 4 and table 2). As data was sorted for clear skies (Rs/Rso > 0.8), the top three most influential parameters from table 1 were accounted for (fc/fcNS, Rs/Rso, and Hour, respectively). During early morning, most of the values for fc/fcNS are near 1, indicating there is little leaf curl in the early morning hours where evaporative demand is barely above zero. As time moves toward mid-day, the data indicating non stress (Ks = 1) typically have little reduction in canopy cover (fc = fcNS and fc/fcNS = 1), whereas values with significant stress (as Ks approaches 0) will have increasing reductions in canopy. All hours at mid-day (between the mid-day hours of 11 a.m. and 2 p.m.) have a very strong fit (0.749 = R2 = 0.820) and an x-intercept less than 0.5 (0.43 = x-intercept = 0.48). This result suggests that a reduction of half the visible fc (i.e., if fc = 0.5 x fcNS) could be associated with a near complete shutdown of plant transpiration (Ks = 0). In later afternoon hours, the x-intercept then increased, indicating a rebound of fc as the peak ET demand hours had passed.

Table 1. Initial screening of environmental parameters to predict Ks (eq. 3), partitioned by irrigation treatment and vegetative/reproductive growth stages. Data was filtered to include daytime only (Rs > 0.5 MJ m-2 d-1). Significant effects in bold (p < 0.05). LogWorth is defined as -log10(p-value), given as a transformation to adjust p-values on an appropriate scale for graphing and other direct comparisons.
Vegetative GrowthReproductive Growth
Sourcep-valueLogWorthSourcep-valueLogWorth
Limited
Irrigation		fc/fcNS<0.0000162.153fc/fcNS<0.000019.383
Rs/Rso<0.000017.134RH (%)0.002222.654
Closest_Hour<0.000015.470Srad (MJ/m2/d)0.017291.762
RH (%)0.001552.810VPD (kPa)0.069011.161
Srad (MJ/m2/d)0.039251.406Ta (C)0.192860.715
Ta (C)0.107830.967Closest_Hour0.199630.700
ETr (mm/h)0.130280.885ETr (mm/h)0.390160.409
VPD (kPa)0.306610.513Rs/Rso0.872130.059
R2 model0.5590.178
RMSE0.2130.104
N401367
Full
Irrigation 		fc/fcNS<0.0000113.799Closest_Hour<0.0000112.417
Rs/Rso0.035561.449Rs/Rso<0.000017.324
RH (%)0.282500.549fc/fcNS<0.000015.508
Closest_Hour0.409130.388Srad (MJ/m2/d)0.000113.955
ETr (mm/h)0.459930.337VPD (kPa)0.000213.669
Ta (C)0.812250.090RH (%)0.001482.831
Srad (MJ/m2/d)0.812980.090Ta (C)0.003612.443
VPD (kPa)0.940380.027ETr (mm/h)0.800880.096
R2 model0.2130.452
RMSE0.2750.114
N408367

As the hours between 11 a.m. and 2 p.m. had similar x-intercepts and slopes, they were grouped into a single model to balance statistical robustness with a greater window of applicable times. This model had an R2 value of 0.773 and an F Ratio of 377.6, leading to its rearrangement as follows:

Figure 4. Crop water stress coefficient Ks vs. fc/fcNS. Data shown is restricted to Limited Irrigation treatment prior to anthesis (DOY 207), with soil water observations taken at least two days following previous irrigation or rain events. Data is limited to daytime (Rs > 0.5 MJ m-2 d-1) and clear sky observations (Rs/Rso > 0.8).

        (5)

which is shown graphically in figure 5. Note that the line crosses Ks = 1 at fc/fcNS = 0.976, likely representing values of fc that are greater than the 90th percentile. The line also crosses Ks = 0 at fc/fcNS = 0.440, indicating that the crop canopy reduction of over 44% from fcNS would be related to an extreme (i.e., near-full) plant reduction of transpiration.

Table 2. Regression results of Ks (eq. 3) vs. fc/fcNS, separated by nearest hour, on the hour. Data shown is restricted to Limited Irrigation treatment prior to anthesis (DOY 207). Data is limited to daytime (Rs > 0.5 MJ m-2 d-1) and clear sky observations (Rs/Rso > 0.8).
HourR2RMSEF-Ratioy interceptSlopex interceptProb > |t|N
60.2990.2182.66.26-5.901.060.16078
70.0490.2821.12.57-2.171.180.300424
80.1210.3093.6-2.232.770.810.070024
90.3510.27513.0-1.091.750.620.001426
100.7800.13692.4-1.141.940.59<0.000128
11[a]0.8150.134114.4-0.841.810.46<0.000128
12[a]0.8200.140105.0-0.721.780.40<0.000125
13[a]0.7490.18583.4-0.821.910.43<0.000130
14[a]0.7860.171102.6-0.992.060.48<0.000130
150.8490.150123.8-1.362.470.55<0.000124
160.6070.21326.3-1.342.270.59<0.000119
170.7140.18440.0-1.432.270.63<0.000118
180.7070.23514.4-1.402.230.630.00908
11-140.7730.163377.6-0.821.870.44<0.0001113

    [a]    Consolidated into a single model containing all measurements for hours 11, 12, 13, and 14, indicated on the last line of the table.

The model from equation 5 was then used for all observations of fc with the closest hour between 11 a.m. and 2 p.m. to predict Ks and compare them with the observed Ks from equation 3 (fig. 6). While the data shown here uses the fc data as shown in figure 3, note that figure 5 uses only the fc values taken at mid-day, when the crop canopy is most responsive to the effects of water stress. While comparisons to Ks taken from soil water are not available every day, the daily changes to Ks values indicate that the method is generally responsive to gradual increases in water stress due to incremental ETc (i.e., declines in Ks) and sharp decreases in water stress (i.e., increases in Ks) following two irrigation and rainfall events (17 mm on DOY 183-184, 16 mm on DOY 193, 21 mm on DOY 197, and 18 mm on DOY 199-200). The new Ks model (eq. 5) was generally in the range of Ks from soil water (eq. 3), but had lower estimates at high Ks values (e.g., DOY 194 and 201). This trend may likely be due to the model (eq. 5 and fig. 5) maxing out at Ks just less than 1, when fc = fcNS. Mathematically, this may be due to fcNS being chosen as the 90th percentile (i.e., 10% of the data would actually be higher), and it is conceivable that a more conservative estimate such as the 95th percentile of the data would shift the Ks = 1 intercept (fig. 5) to the left.

Figure 5. Daily water stress coefficient (Ks) from observed soil water vs. decline in fractional canopy cover (fc/fcNS) for limited irrigation during late vegetative growth, prior to irrigation events, including all hours between 11 a.m. and 2 p.m. Shaded band indicates 95% confidence of the mean, dashed lines indicate 95% confidence of the data.

Discussion

Past studies have shown that fc can be used to determine the transpiration for a crop absent of water stress (i.e., KcbETr), and this study demonstrates that the mid-day reduction in fc can represent the reduction of that transpiration due to water stress (i.e., Ks as a function of fc/fcNS). As a practical consideration, both methods require an observation of fcNS early in the day and/or following a watering event, which may have some limitations in obtaining. Methods that only obtain fc at mid-day, without having a baseline value for fcNS, may be prone to erroneous assumptions about the data. For example, fc values obtained at mid-day may well be representative of fcNS if the crop is non-stressed, and this method could be valuable for estimating Kcb for a fully watered crop. However, if the crop is experiencing even slight water stress resulting in leaf curling, the assumption of this fc equaling fcNS would be erroneous. Additionally, mid-day imaging of an entire field may capture both unstressed and stressed crops, making it impossible to distinguish between them without spatial knowledge of the fcNS. It is suggested that any observations of fc used to represent a non-stressed condition (i.e., fcNS to be converted to Kcb) be taken under time and water status representative of the absence of water stress. Obtaining fc during peak water stress (mid-day) has some flexibility with a larger time window (e.g., 11 a.m. to 2 p.m. in this study), whereas estimation of fcNS is best early in the day. While not thoroughly investigated in this study, the results show that fc is very responsive and recovers considerably to cloud cover, and while fc under cloud cover is not suggested for use in estimating Ks with the methods in this study, fc under cloud cover might be another opportunity to represent fcNS under the right conditions. Moreover, fc under cloud cover conditions could be acquired using remote sensing platforms that offer high-resolution imagery (Fernandez-Gallego et al., 2018).

This work highlights a critical need for further development of remote sensing methods to accurately estimate Kcb for purposes of ET estimation.Interpolation may be suggested as a technique for Kcb calculations between fcNS observations, which is a main driver of the water balance in the dual coefficient method. Additional research is needed to estimate and predict reductions of Kcb due to accumulated water stress. Interpolation of Ks values may be an option during periods absent of watering, but when soil water status changes drastically due to irrigation or rainfall, so will Ks, in which case interpolation would not be recommended. Estimates of daily values for both Kcb and Ks can be valuable in water balance modeling, but having estimates of Ks alone can be good indicators of water stress. When available to use in combination with good estimates of Kcb, canopy temperature is an ideal plant-based measurement of water stress to determine Ks under full or near-full fc (Kullberg et al., 2017; DeJonge and Zhang, 2021) and trigger irrigation (Wanjura et al., 1992; Nakabuye et al., 2022), and has opportunities for integration into decision support systems for irrigation systems such as center pivots (Andrade et al., 2020; Zhang et al., 2021).

Figure 6. Comparison of Ks for the limited irrigation treatment in vegetative growth, determined from equation 3 (blue dots) and equation 5 (red boxplots), with the latter including only daytime measurements (Rs > 0.5 MJ m-2 d-1) with clear sky (Rs/Rso > 0.8) between the hours of 11 a.m. and 2 p.m.

Estimation of Ks has long been done using canopy temperature, but in the past these techniques were limited to full canopy since viewing of sunlit soil in the background can bias canopy temperature measurements. Newer techniques with high-resolution thermal imaging are now able to exclude soil background to isolate canopy temperature (Zhang et al., 2019b), but even these techniques already require RGB imaging. To complement existing thermal techniques to monitor water stress, the techniques developed in the current study rely only on RGB imaging for data inputs. The applications of this technique are largely limited to fc measured under water stress (limited irrigation treatment) in late vegetative growth when full canopy is not realized. Spatial modeling of Ks could feasibly use fc until achieving full canopy cover, then switch to canopy temperature, with verification through soil water measurements when available. Once the crop reaches full canopy cover, reductions in fc may be less indicative or reliable than measurements of canopy temperature. In a world where multiple data streams are readily available, it is conceivable in this context that a combination of data from both visible and thermal bands may be more informative than each alone for estimating water stress.

Some recent studies suggest that for maize, mild water stress during vegetative growth stages can result in high yields with reduced water use (Comas et al., 2019; Chen et al., 2023), and this strategy is a focus in many regulated deficit irrigation strategies. Most importantly for maize, the early reproductive stage is extremely sensitive to yield reduction due to water stress (Westgate and Boyer, 1985), so water stress should be avoided at all costs during this growth stage, and reductions in fc due to water stress would be minimal if following this management objective. That said, if water stress is unavoidable during this critical growth stage, the methods used in this paper may still be considered as an estimation for Ks.

While fc is sensitive to water stress and maximal during mid-day, fc can also be sensitive to heat stress even in the absence of water stress. This phenomenon is likely the reason for fc fluctuations under full irrigation (e.g., DOY 191-193 in fig. 3). Similar to the response to water stress, the plants will close their stomata due to their inability to keep up with the environmental demands of transpiration. It is safe to assume that if a crop under full irrigation is experiencing heat stress, a crop under limited irrigation in the same environment will experience some combination of both water stress and heat stress. Although a recent study investigated the effects of maize pollen viability under heat and drought stress (Li et al., 2022), further investigation is needed to develop methods to separate heat stress from water stress, especially in terms of evapotranspiration reduction and effects on other growth mechanisms. In addition to separating heat stress from water stress, an opportunity exists with this method to quantify cumulative water stress (i.e., Ks) and the ability to predict reductions in Kcb. Finally, while this study is on conventional maize, which is very responsive to leaf curl due to water stress, additional studies are needed to determine the transferability of this method to drought-tolerant maize varieties, as well as other crops.

Lastly, the process developed in this study reveals some potential issues regarding remote sensing for quantification of evapotranspiration, especially under water stress. Balancing resolution and coverage is a persistent consideration with any remote sensing program, and the technique developed in this study uses ground-based, high-resolution images to accurately quantify short-term (i.e., diurnal) changes in fc. It is conceivable that high-resolution images collected by UAVs could quantify fc in a similar fashion, but satellite-based remote sensing methods are incapable of such fine resolution. Considering that most satellite-based remote sensing methods capture images at or near mid-day, during the peak reduction of fc under water stress, these methods may be capturing the effects of water stress without realizing it. Even images taken a few hours apart can show drastically different estimates of instantaneous crop water status (e.g., fig. 1). As regulated deficit irrigation becomes a preferred strategy for managing limited water resources, satellite-based remote sensing methods will need to evolve to consider these drastic changes that the plant presents while under water stress.

Conclusions

This study builds upon the use of fractional canopy cover (fc) as a critical component in accurately estimating crop evapotranspiration using many well-known models and methods. Under conditions of water stress, maize fc can have substantial diurnal fluctuation as leaves wilt and reduce transpiration, which was rarely quantified prior to this study. The primary novelty of the study was the development of a fairly simple temporal and high-resolution imaging process to quantify the ratio of reduced mid-day fc to non-stressed fractional cover (fcNS) from the same plot, and convert that ratio into a water stress (Ks) term related to reduced crop water use. It is crucial to note that methods based solely on mid-day fc observations could introduce errors without a baseline fcNS value, particularly in fields containing both stressed and non-stressed crops. While existing thermal techniques have traditionally required full canopy views, the novel approach developed in this study relies solely on RGB imaging, providing insights into crop water stress during late vegetative growth when full canopy development is not yet achieved. The study highlights potential spatiotemporal challenges in satellite-based remote sensing for evapotranspiration quantification, urging continued advancements to capture the dynamic effects of water stress on crop conditions, particularly under regulated deficit irrigation strategies.

Supplemental Material

The supplemental materials mentioned in this article are available for download from the ASABE Figshare repository at: https://doi.org/10.13031/22568443

Acknowledgments

Kendall DeJonge led the conceptualization of the study, the interpretation of the dataset, and the writing of the manuscript. Huihui Zhang contributed to the interpretation of the dataset and the writing of the manuscript. Liam Cummins fabricated data collection apparatuses and managed image data collection and initial processing. Tyler Gilkerson, Katie Ascough, and Tyler Pokoski all contributed to various subsequent levels of image processing, analysis, and interpretation. Thanks to Gerald Buchleiter and Ross Steward for LIRF management, Garrett Banks for soil water data collection and management, Gavin Roy for preliminary development of the image processing program, and several students for assisting with soil water data collection, which in 2016 included KaMele Sanchez and Ben Choat. Thanks as well to Sara Duke for help with statistical interpretation, Tom Trout for reviewing an early draft of the manuscript, and the anonymous reviewers whose suggestions improved the manuscript. This research was supported by the U.S. Department of Agriculture, Agricultural Research Service.

Nomenclature

fc/fcNS = fraction of fc for each timestep to fcNS for the day

Drel = relative depletion

ET = evapotranspiration (general)

ETc = crop evapotranspiration under ideal (standard) conditions

ETcact = actual crop evapotranspiration, considering effects of non-standard conditions

ETr = tall crop (alfalfa) reference evapotranspiration

ETref = reference evapotranspiration (general)

fc = fractional canopy cover

fcNS = non-stressed fc for each day

Kc = crop coefficient (general), ratio of ETc/ETref

Kcb = basal (transpiration) crop coefficient

Ke = evaporation coefficient

Ks = water stress coefficient, which scales transpiration to Kcb

p = threshold rSWD above which Ks is not affected

Rs= solar radiation

Rso= clear-sky solar radiation

sf = shaping factor for curvilinear Ks model

PWP = permanent wilting point

RAW = readily available water

rSWD = relative soil water deficit

SWD = soil water deficit

TAW = total available water

References

Allen, R. G., & Pereira, L. S. (2009). Estimating crop coefficients from fraction of ground cover and height. Irrig. Sci., 28(1), 17-34. https://doi.org/10.1007/s00271-009-0182-z

Allen, R. G., Pereira, L. S., Raes, D., & Smith, M. (1998). Crop evapotranspiration - Guidelines for computing crop water requirements - FAO Irrigation and drainage paper 56. Rome, Italy: United Nations FAO.

Amoros-Lopez, J., Gomez-Chova, L., Vila-Frances, J., Alonso, L., Calpe, J., Moreno, J., & del Valle-Tascon, S. (2008). Evaluation of remote sensing of vegetation fluorescence by the analysis of diurnal cycles. Int. J. Remote Sens., 29(17-18), 5423-5436. https://doi.org/10.1080/01431160802036391

Andrade, M. A., O’Shaughnessy, S. A., & Evett, S. R. (2020). ARSPivot, a sensor-based decision support software for variable-rate irrigation center pivot systems: Part A. Development. Trans. ASABE, 63(5), 1521-1533. https://doi.org/10.13031/trans.13907

ASCE. Jensen, M. E., and R. G. Allen (Eds.). 2016. Evaporation, evapotranspiration, and irrigation water requirements. ASCE Manual of Practice 70. 2nd ed. Reston, VA: ASCE.

ASCE. 2005. The ASCE standardized reference evapotranspiration equation. Reston, VA: ASCE.

Balota, M., Payne, W. A., Evett, S. R., & Lazar, M. D. (2007). Canopy temperature depression sampling to assess grain yield and genotypic differentiation in winter wheat. Crop Sci., 47(4), 1518-1529. https://doi.org/10.2135/cropsci2006.06.0383

Baret, F., Madec, S., Irfan, K., Lopez, J., Comar, A., Hemmerlé, M.,... Tixier, M. H. (2018). Leaf-rolling in maize crops: From leaf scoring to canopy-level measurements for phenotyping. J. Exp. Bot., 69(10), 2705-2716. https://doi.org/10.1093/jxb/ery071

Bellvert, J., Adeline, K., Baram, S., Pierce, L., Sanden, B. L., & Smart, D. R. (2018). Monitoring crop evapotranspiration and crop coefficients over an almond and pistachio orchard throughout remote sensing. Remote Sens., 10(12), 2001. https://doi.org/10.3390/rs10122001

Chen, X., Xing, H., Liu, B., Wang, Y., Cui, N., Wang, Z., & Zhang, Y. (2023). Changes induced by multi-stage water stress on maize growth, water and nitrogen utilization and hormone signaling under different nitrogen supplies. Agric. Water Manag., 290, 108570. https://doi.org/10.1016/j.agwat.2023.108570

Comas, L. H., Trout, T. J., DeJonge, K. C., Zhang, H., & Gleason, S. M. (2019). Water productivity under strategic growth stage-based deficit irrigation in maize. Agric. Water Manag., 212, 433-440. https://doi.org/10.1016/j.agwat.2018.07.015

DeJonge, K. C., & Zhang, H. (2021). Deficit irrigation management using canopy cover and temperature to inform crop coefficient methods. Proc. 6th Decennial National Irrigation Symposium. St. Joseph, MI: ASABE. https://doi.org/10.13031/irrig.2020-043

DeJonge, K. C., Cummins, L., Gilkerson, T., Ascough, K., & Pokoski, T. (2023). Supplemental videos for “Diurnal trends of maize canopy cover under water stress.”. ASABE. Retrieved from https://figshare.com/s/65abc3176e2beaaba87e

DeJonge, K. C., Mefford, B. S., & Chávez, J. L. (2016). Assessing corn water stress using spectral reflectance. Int. J. Remote Sens., 37(10), 2294-2312. https://doi.org/10.1080/01431161.2016.1171929

DeJonge, K. C., Taghvaeian, S., Trout, T. J., & Comas, L. H. (2015). Comparison of canopy temperature-based water stress indices for maize. Agric. Water Manag., 156, 51-62. https://doi.org/10.1016/j.agwat.2015.03.023

Fereres, E., & Goldhamer, D. (2003). Suitability of stem diameter variations and water potential as indicators for irrigation scheduling of almond trees. J. Hortic. Sci. Biotechnol., 78(2), 139-144. https://doi.org/10.1080/14620316.2003.11511596

Fernandez-Gallego, J., Kefauver, S., Kerfal, S., & Araus, J. L. (2018). Comparative canopy cover estimation using RGB images from UAV and ground - 107830J. Proc. SPIE 10783, Remote Sensing for Agriculture, Ecosystems, and Hydrology XX. SPIE. https://doi.org/10.1117/12.2501531

Fischer, R. A., & Connor, D. J. (2018). Issues for cropping and agricultural science in the next 20 years. Field Crops Res., 222, 121-142. https://doi.org/10.1016/j.fcr.2018.03.008

Gleason, S. M., Wiggans, D. R., Bliss, C. A., Comas, L. H., Cooper, M., DeJonge, K. C.,... Zhang, H. (2017). Coordinated decline in photosynthesis and hydraulic conductance during drought stress in Zea mays. Flora, 227, 1-9. https://doi.org/10.1016/j.flora.2016.11.017

Harmel, R. D., Chaubey, I., Ale, S., Nejadhashemi, A. P., Irmak, S., DeJonge, K. C.,... Mani, I. (2020). Perspectives on global water security. Trans. ASABE, 63(1), 69-80. https://doi.org/10.13031/trans.13524

Kimes, D. S., & Kirchner, J. A. (1982). Diurnal variations of vegetation canopy structure. Int. J. Remote Sens., 4(2), 257-271. https://doi.org/10.1080/01431168308948545

Kullberg, E. G., DeJonge, K. C., & Chávez, J. L. (2017). Evaluation of thermal remote sensing indices to estimate crop evapotranspiration coefficients. Agric. Water Manag., 179, 64-73. https://doi.org/10.1016/j.agwat.2016.07.007

Li, H., Tiwari, M., Tang, Y., Wang, L., Yang, S., Long, H.,... Shao, R. (2022). Metabolomic and transcriptomic analyses reveal that sucrose synthase regulates maize pollen viability under heat and drought stress. Ecotoxicol. Environ. Saf., 246, 114191. https://doi.org/10.1016/j.ecoenv.2022.114191

Nakabuye, H. N., Rudnick, D., DeJonge, K. C., Lo, T. H., Heeren, D., Qiao, X.,... Duan, J. (2022). Real-time irrigation scheduling of maize using Degrees Above Non-Stressed (DANS) index in semi-arid environment. Agric. Water Manag., 274, 107957. https://doi.org/10.1016/j.agwat.2022.107957

Niu, Y., Han, W., Zhang, H., Zhang, L., & Chen, H. (2021). Estimating fractional vegetation cover of maize under water stress from UAV multispectral imagery using machine learning algorithms. Comput. Electron. Agric., 189, 106414. https://doi.org/10.1016/j.compag.2021.106414

Patrignani, A., & Ochsner, T. E. (2015). Canopeo: A powerful new tool for measuring fractional green canopy cover. Agron. J., 107(6), 2312-2320. https://doi.org/10.2134/agronj15.0150

Pereira, L. S., Paredes, P., Melton, F., Johnson, L., Wang, T., López-Urrea, R.,... Allen, R. G. (2020). Prediction of crop coefficients from fraction of ground cover and height. Background and validation using ground and remote sensing data. Agric. Water Manag., 241(1), 106197. https://doi.org/10.1016/j.agwat.2020.106197

Peters, R. T., & Evett, S. R. (2004). Modeling diurnal canopy temperature dynamics using one-time-of-day measurements and a reference temperature curve. Agron. J., 96(6), 1553-1561. https://doi.org/10.2134/agronj2004.1553

Raes, D., Steduto, P., Hsiao, T. C., & Fereres, E. (2023). AquaCrop version 7.1 reference manual. Rome, Italy: United Nations FAO.

Rodriguez, R. (2023). Measuring the adoption of drones: A case study of the United States agricultural aircraft sector. Eng, 4(1), 977-983. https://doi.org/10.3390/eng4010058

Steduto, P., Hsiao, T. C., Raes, D., & Fereres, E. (2009). AquaCrop — The FAO crop model to simulate yield response to Water: I. Concepts and underlying principles. Agron. J., 101(3), 426-437. https://doi.org/10.2134/agronj2008.0139s

Stevens, B., Diels, J., Vanuytrecht, E., Brown, A., Bayo, S., Rujweka, A.,... Swennen, R. (2020). Canopy cover evolution, diurnal patterns and leaf area index relationships in a Mchare and Cavendish banana cultivar under different soil moisture regimes. Sci. Hortic., 272, 109328. https://doi.org/10.1016/j.scienta.2020.109328

Tang, J., Han, W., & Zhang, L. (2019). UAV multispectral imagery combined with the FAO-56 dual approach for maize evapotranspiration mapping in the North China Plain. Remote Sens., 11(21), 2519. https://doi.org/10.3390/rs11212519

Terán-Chaves, C. A., García-Prats, A., & Polo-Murcia, S. M. (2022). Water stress thresholds and evaluation of coefficient Ks for perennial ryegrass in tropical conditions. Water, 14(11), 1696. https://doi.org/10.3390/w14111696

Trout, T. J., & DeJonge, K. C. (2017). Water productivity of maize in the U.S. High Plains. Irrig. Sci., 35(3), 251-266. https://doi.org/10.1007/s00271-017-0540-1

Trout, T. J., & DeJonge, K. C. (2018). Crop water use and crop coefficients of maize in the Great Plains. J. Irrig. Drain. Eng., 144(6), 04018009. https://doi.org/10.1061/(ASCE)IR.1943-4774.0001309

Trout, T. J., & DeJonge, K. C. (2021). Evapotranspiration and water stress coefficient for deficit-irrigated maize. J. Irrig. Drain. Eng., 147(10), 04021044. https://doi.org/10.1061/(ASCE)IR.1943-4774.0001600

Vo, T. T., & Hu, L. (2021). Diurnal evolution of urban tree temperature at a city scale. Sci. Rep., 11(1), 10491. https://doi.org/10.1038/s41598-021-89972-0

Wang, N., Suomalainen, J., Bartholomeus, H., Kooistra, L., Masiliunas, D., & Clevers, J. G. (2021). Diurnal variation of sun-induced chlorophyll fluorescence of agricultural crops observed from a point-based spectrometer on a UAV. Int. J. Appl. Earth Obs. Geoinf., 96, 102276. https://doi.org/10.1016/j.jag.2020.102276

Wanjura, D. F., Upchurch, D. R., & Mahan, J. R. (1992). Automated irrigation based on threshold canopy te    mperature. Trans. ASAE, 35(1), 153-159. https://doi.org/10.13031/2013.28582

Westgate, M. E., & Boyer, J. S. (1985). Carbohydrate reserves and reproductive development at low leaf water potentials in maize. Crop Sci., 25(5). https://doi.org/10.2135/cropsci1985.0011183X0025000500010x

Wu, J., Wang, J., Hui, W., Zhao, F., Wang, P., Su, C., & Gong, W. (2022). Physiology of plant responses to water stress and related genes: A review. Forests, 13(2), 324. https://doi.org/10.3390/f13020324

Yan, G., Li, L., Coy, A., Mu, X., Chen, S., Xie, D.,... Zhou, H. (2019). Improving the estimation of fractional vegetation cover from UAV RGB imagery by colour unmixing. ISPRS J. Photogramm. Remote Sens., 158, 23-34. https://doi.org/10.1016/j.isprsjprs.2019.09.017

Zhang, H., Han, M., Comas, L. H., DeJonge, K. C., Gleason, S. M., Trout, T. J., & Ma, L. (2019a). Response of maize yield components to growth stage-based deficit irrigation. Agron. J., 111(6), 3244-3252. https://doi.org/10.2134/agronj2019.03.0214

Zhang, J., Guan, K., Peng, B., Jiang, C., Zhou, W., Yang, Y.,... Cai, Y. (2021). Challenges and opportunities in precision irrigation decision-support systems for center pivots. Environ. Res. Lett., 16(5), 053003. https://doi.org/10.1088/1748-9326/abe436

Zhang, L., Niu, Y., Zhang, H., Han, W., Li, G., Tang, J., & Peng, X. (2019b). Maize canopy temperature extracted from UAV thermal and RGB imagery and its application in water stress monitoring. Front. Plant Sci., 10. https://doi.org/10.3389/fpls.2019.01270

Zhang, Y., Han, W., Zhang, H., Niu, X., & Shao, G. (2023). Evaluating maize evapotranspiration using high-resolution UAV-based imagery and FAO-56 dual crop coefficient approach. Agric. Water Manag., 275, 108004. https://doi.org/10.1016/j.agwat.2022.108004