Introduction
Intensive agriculture has improved productivity by investing in several crop-man agement practices, including irrigation and fertilization 12,14,36. Since pastures are major fodder sources for cattle in Brazil, intensification of pasture production turns key for the cultivation and management of this crop.
Occupying a vast territory, pastures are subjected to a wide range of environmental conditions, particularly temperature, light, and rainfall. Outranged environmental magni tudes for pasture production compromise vegetative development, expressed as low biomass accumulation and marked production seasonality 20,23,33. Meteorological conditions prevailing during winter impair tropical and subtropical forage crops, whereas spring/summer favor high forage productivity 25,28.
In order to reduce fertilization and counteract low pasture yields during winter, culti vation of temperate forage species has been identified as an alternative for these cultivation systems. Good results have been reported for forage nutritional value and animal production 13. However, mixed cropping systems, where two or more species are co-cultivated using the overseeding method, cause changes in crop water requirements. Therefore, estimating these changes in irrigated pasture systems turns essential.
Adequate information on plant and microclimate interaction constitutes one major factor for accurate irrigation management 6,29. Crop water requirements can be calcu lated after plant water losses through transpiration and soil evaporation, termed crop evapotranspiration (ET c ). Estimated reference evapotranspiration (ET o ) calculated using meteorological data, is then converted to ET c using a crop coefficient (K c ) as a correction factor, while the precise ET c can be determined using a weighing lysimeter 4.
In established pastures, the K c varies depending on crop growth and landscape, such as after sowing, seed germination, or forage grazing 7,17,18,27,31. However, as the grazing cycle is defined by species, climate, and animal characteristics, this approach turns imprecise and avoided for tropical forage crops.
Air temperature is used to calculate degree-days, expressing available thermal units for crop growth. Accumulated degree-days (ADD) are also used to estimate the correct K c for irrigation management. Thus, K c -as estimated by ADD-varies with ambient temperature, influencing crop growth and cycle 16. Therefore, alternative methods for K c estimation may be particularly useful for the dynamics of irrigated pastures. Therefore, in this study and using the weighing-lysimeter method, we aimed to measure K c for growth cycles of ‘Marandu’ forage, a palisadegrass cultivar (Urochloa brizantha). The study took place during 2016 and 2017 and developed an adjustable regression model relating K c and ADD during the four seasons, evaluated when applied to previous and subsequent years.
Materials and methods
Experimental site and soil classification
The field experiment was conducted at the Luiz de Queiroz College of Agriculture (ESALQ/USP), Piracicaba, São Paulo (latitude 22°42ʹ S, longitude 47°38ʹ W, altitude 546 m) in Brazil. The soil is classified as Ferralic Nitisol 38, with 48.6% clay, 32.5% sand, and 18.9% silt (0-0.4 m).
Soil chemical characteristics were determined prior experiment: pH = 5.1, P (resin) = 52 mg dm-3, K = 0.69 cmol c dm-3, Ca = 2.6 cmol c dm-3, Mg = 1.4 cmol c dm-3, H + Al = 3.65 cmol c dm-3, Al = 0.2 cmol c dm-3, and CEC = 8.34 cmol c dm-3. Soil preparation included plowing and harrowing, weed control, Ph correction, and fertilization adjustment.
Forage cultivation
The experimental area included two, 144 m² plots. The area surrounding the experi mental plots was covered with irrigated pastures and sugarcane. Only one close field had bare soil. Reading errors lead by bare soil, affecting evapotranspiration, were identified and excluded from daily calculations after verifying hot air mass entrances.
Initially, both plots were direct-seeded by broadcasting 15 kg ha-1 of ‘Marandu’ palisa degrass. Similar crop management practices were followed in both plots. After crop estab lishment (~70 days), the grass was cut to a standard height of 0.15 m above soil surface using a brush cutter. In autumn, one plot was kept for a single cropping system under ‘Marandu’ palisadegrass. Meanwhile, and to create a mixed cropping system, the other plot was direct-seeded (broadcasting) with 100 kg seeds ha-1 of black oats (Avena strigosa ‘Embrapa 29 (Garoa)’) and 60 kg seeds ha-1 Italian ryegrass (Lolium multiflorum ‘Fepagro São Gabriel’). This created a pasture comprising three different forage crops during autumn and winter.
Cutting cycles were grouped according to seasons: summer (February to March 2016, and December 2016 to January 2017), autumn (April to June 2016), winter (July 2016 to October 2016), and spring (October to December 2016). After each cut, the plots were fertilized with 80 kg ha-1 urea during spring and summer and 50 kg ha-1 during autumn and winter. The Average total forage yield (TFY) was calculated as the sum of leaf and stem dry mass (DM), after kiln drying.
Forage cutting in the single-cropped plot (‘Marandu’ palisadegrass), was performed every 28 days during spring and summer, extended to 40 days during autumn and winter. In the mixed-cropped plot (‘Marandu’ palisadegrass + black oats + ryegrass), harvest was performed at intervals ranging from 24 to 40 days when the canopy had exceeding 95% light interception, measured with a LAI-2000 Plant Canopy Analyzer (LI-COR Environ mental, Lincoln, Nebraska USA).
All experimental plots were irrigated with a conventional sprinkler system, operating via a sectoral mechanism at a flow rate of 590 L h-1 and running at a pressure of 245 kPa. The quantity of water applied was determined by weighing-lysimeter readings (in L/m2 or mm), considering the lysimeter depth (z = 0.58 m) as the maximum crop root depth. In tropical forages, most roots are detected down to 0.60 m, and 70% of all roots are concentrated within the first 0.30 m depth (2, 21, 34).
An irrigation interval kept soil moisture content (SMC) over 70% (recommended for K c estimate), between field capacity (θ fc ) and lowest soil water content (θ wp ). It was measured within 15 bars to standard atmospheric pressure (SMC ≥ 0.7 [Ө fc − Ө wp ]), guaranteeing null soil water deficit, and accurate K c data.
Evapotranspiration and accumulated degree-days
Crop evapotranspiration (ET c ) in the experimental plots was estimated using two lysim eters, made of a PVC box (500 L), of 1.22 m² and 0.58 m depth 26, with an automatic drainage system. Both sets were calibrated and controlled by a datalogger program 26. Lysimeter measurements were recorded every 15 min and converted into 1-day interval data. Values exceeding 0.20 mm in a 15 min interval, caused by excessive rainfall or evapo transpiration, were excluded. Brief periods of high evapotranspiration and consequent ET c fluctuations arose during hot transient air masses originated from exposed soil and arriving mainly from the north, northwest, west, and occasionally, from the southwest.
Daily degree-days (DD i ) were calculated based on temperature 22. Equation 1 was used when crop basal temperature (T b ) was lower than minimum daily temperature (T min ), and eq. 2, when T b exceeded T min . Meteorological data were obtained from the ESALQ Mete orological Station, located 100 m from the experimental area. Data for T min , and T max are shown in Figure 1.
Graphs are divided sequentially according to season (Summer, Autumn, Winter, Spring).
En ambas figuras, los gráficos se dividen secuencialmente según estación (verano, otoño, invierno, primavera).
Reference evapotranspiration (ET o ) (Penman-Monteith model, 2, 3) for different cycles is shown in Table 1 (page 75).
Single cropping cycles are divided sequentially according to season (Su = summer, A = autumn, W = winter, Sp = spring).
Los ciclos para un solo cultivo se dividen secuencialmente según la temporada (Su = verano, A = otoño, W = invierno, Sp = primavera).
In equations 1 and 2:
DD i = daily degree-days,
T max = maximum daily temperature in °C,
T min = minimum daily temperature in °C (Figure 1a),
T b = basal crop temperature for the single-cropped ‘Marandu’ palisadegrass plot, which was considered to be 10.6 °C 26. This same basal temperature was applied for plots with palisadegrass mixed with black oats and Italian ryegrass.
It should be noted that ADD was also estimated for the original cropping cycles (28 days for spring/summer and 40 days for autumn/winter), with 11 productive cycles for single-cropped ‘Marandu’ palisadegrass and five cycles of the same forage in the overseeded mixed-cropping regime (‘Marandu’ palisadegrass + black oats + ryegrass).
The summed DD values for each cropping cycle comprised all accumulated degree-days (ADD). The K c values were obtained as ET c /ET o . Daily values of K c were averaged over 4-day intervals obtaining mean daily values and reducing possible distortions due to water flow within the lysimeters. The K c values of the studied forage crops were grouped as follows: single-cropped pasture during spring-summer (Sp/Su) and autumn-winter (A/W), and the overseeded mixed-cropped pasture during A/W. Subsequently, daily K c means were correlated with ADD for each cutting cycle verifying if ADD could be used in K c determinations.
Statistical analysis
K c and ADD data were correlated using SigmaPlot software for regression analysis, with equations adjusted for the entire experimental period according to seasons. For 2015- 2016 and 2017-2018 periods (prior and after the experiment, respectively), climate data was collected from a weather station located near the experimental area. For both periods, second-order polynomial models (K c vs. ADD) were generated (R 2 ≥ 0.97). K c was estimated according to eq3, based on the calculated ADD in the simulated sequences for 2015-2016 and 2017-2018.
Degree-days were counted from February 11, 2016 to January 31, 2017 (355 days), according to the original experimental period. The same periods were adopted for 2015- 2016 and 2017-2018.
Using the generated regression models, K c values were estimated for these years. Esti mation accuracy was evaluated by the mean error (ME; eq 4), the root mean square error (RMSE; eq 5), Willmott’s concordance index d (37), and confidence index c (10, eq 6 and 7).
In the equations:
n = the number of data points,
E i = the estimated data,
O i = the observed data.
Results
Regression curves for K c means vs. ADD in single-cropped ‘Marandu’ palisadegrass were divided into two periods: Sp/Su and A/W, as shown in Figure 2
For both periods, second-order polynomial models (Kc vs. ADD) were generated (R 2 ≥ 0.97). During A/W, maximum K c (K c max = 0.82) was obtained at an ADD of 160 °C·days (°C × days, degree-days), whereas in Sp/Su, K c max reached 0.90 at an ADD of 280 °C·days (Figure 2).
Mean minimum K c values (K c min ) determined at the beginning of the study periods were 0.50 for Sp/Ss and 0.67 for A/W. ADD values for spring and autumn were 375.82 °C·days and 355.25 °C·days, respectively (Figure 2). ET c was largely ranged during Sp/Su (0.54-6.57) than during the A/W (0.64-5.62), as shown in Table 2, page 77.
Cycles for single cropping are sequentially divided according to season (Su = summer, A = autumn, W = winter, Sp = spring).
Los ciclos para un solo cultivo se dividen secuencialmente según la temporada (Su = verano, A = otoño, W = invierno, Sp = primavera).
K c mean values obtained in the present study were lower than those previously reported, but observed K cmax under ‘Marandu’ palisadegrass in summer, reached 1.20. In addition, except for cycle 1 (establishment), each cycle reached a similar productive potential to that observed by other authors (Table 2, page 77). The TFY values for the single-cropping system in each season were 13927.9 kg DM ha-1 (summer), 5803.6 kg DM ha-1 (spring), 5624.6 kg DM ha-1 (autumn), and 4187.8 kg DM ha-1 (winter). These values sum a total of 29543.9 kg DM ha-1 yr-1 in the single-cropping system. The TFY in the mixed-cropping system was 11343.74 kg DM ha-1 as recorded in a 155-day period (Table 2, page 77).
In Figure 3 (page 78), mean K c for palisadegrass mixed of black oats and Italian ryegrass was similar to that obtained for the single-cropped system.
However, at the end of the cycle, K c values did not decrease, as previously seen for the single-cropped pasture during AW (Figure 2, page 76).
The K c vs. ADDa equation (Figure 2, page 76 and Figure 3, page 78) was used for K c esti mates based on calculated ADD in the simulated sequences for the years 2015-2016 and 2017-2018, as shown in Table 3 (page 78). The estimated K c resulted in R 2 exceeding 0.92 for both Sp/Su and A/W periods, resulting in greater accuracy (higher coefficient d) and lower error in the Sp/Su estimation. Regarding the second year, 2017-2018, R 2 exceeded 0.95, improving the Sp/Su estimation (Table 3, page 78).
Error analysis determined the Willmott’s index (Table 3, page 78). Mean errors (ME) were negative, indicating differences between model estimations and observed data in the field experiment. Observed values were higher than estimated values (with the largest error of 0.0641) in the single-cropped system during A/W. In the same period, the largest root mean square error (RMSE) was 0.11.
During the experimental year, ADD values reached approximately 404 °C·days (DD during spring and summer, and 360 °C·days during autumn and winter. In the previous annual period (2015-2016), mean values for Sp/Su and A/W were 443 and 438 °C·days, respectively, whereas in the subsequent annual period (2017-2018), ADD reached 383 and 379 °C·days, respectively.
Discussion
The decreasing K c values for single-cropped ‘Marandu’ palisadegrass during autumn and winter (Figure 2, page 76) may be related to low biomass accumulation and extensive leaf senescence during this phase, leading to reduced tropical forage production and lower water requirements. In this regard, some authors have shown that even in irrigated cropping systems, biomass yield in colder seasons (A/W) is approximately 50% lower than that produced during warmer periods 1.
The coefficients of determination (R 2) for the observed vs. estimated data in 2017-2018 were higher than those in 2015-2016, keeping high accuracy for both periods. According to the categories for the c index 10, the present model can be considered excellent, except for simulations obtained for A/W, 2015-2016.
The Sp/Su and A/W periods showed a different ADD value for K c max obtention, with a greater maximum for Sp/Su. Other related studies on tropical forages have presented K c max values of 1.04 to 1.25 for Megathyrsus maximus ‘Tanzania’; 1.04 to 1.54 for Megathyrsus maximus ‘Mombaça’; and 1.33 for Urochloa brizantha ‘Piatã’ 5,9,15,19. In our study, curves of K c vs. ADD were constructed by grouping mean K c values for Sp/Su, resulting in an average K c max value of 0.90, while the original data indicated a K c max value of 1.27. This difference can be explained by the plant -atmosphere interaction-a concept known as coupling-observed by other authors 30,32) throughout the year. During the beginning of the Sp/Su cycles, high temperatures may have resulted in the crop-atmosphere decoupling, making evapotranspi ration largely determined by solar radiation, resulting in lower K c values. This differs from the beginning of the A/W period, when temperature, solar radiation, air humidity and cloud cover tend to be lower. These physiological parameters need to be more accurately determined in future studies. Literature on tropical pasture reports the optimum and maximum tempera tures for vegetative development to be 40 °C, and 45 °C, respectively 24. Thus, given that maximum daily temperature recorded over the studied years was between 37.3 °C and 38.7 °C an influence on growth is not considered a significant factor in this study.
During autumn and winter, K c models for the intercropped forage showed different behaviors compared to the models for single-cropped ‘Marandu’ palisadegrass (Figure 3, page 78). In a well-managed intercropped system, water use efficiency (in terms of forage yield and water consumption) may be higher than in a single-cropped system 11,39. Furthermore, the use of climatically adapted species may guarantee stable productivity through the intercropping cycles 35.
Estimation models applied to the 2016-2017 experimental cycle of Kc versus ADD for single-cropped ‘Marandu’ palisadegrass in the Sp/Su and A/W periods showed high precision (R²≥ 0.97). In addition, during winter, the mixed-cropping system reached an R2 ±0.93.
When these equations were tested against two other experimental-year data sets considering ADD value as an input parameter, they showed high precision (R 2 between 0.93 and 0.98), accuracy (d values from 0.86 to 0.99), and confidence (c values between 0.83 and 0.98). A higher error was detected during A/W given that, during this period, the tropical pastures showed major productivity variances. K c and ADD in pasture-irrigation management can be widely adopted given their versatility to cope with grazing rotation 8.
Conclusions
Seasonal Kc was measured by the weighing-lysimeter method for growth cycles of the pali sadegrass ‘Marandu’ cultivar and the overseeded pasture of palisadegrass, black oat, and Italian ryegrass. When tested via regression modeling and estimated by accumulated degree-days (ADD) this model showed accurate results. The equations developed for the spring/summer season and the intercropped system showed higher correlations than autumn/winter. The models can be used to estimate K c with ADD, only requiring minimum and maximum air temperature measurements for Kc estimate, easy in irrigation management.