The experiment was run under field conditions in a Dystric Cambisol [17], from April to August 2018. On May 2017, the cardoon plants (Cynara cardunculus L. var. altilis DC.) were installed in an agricultural field located at Associação Portuguesa de Pais e Amigos do Cidadão Deficiente Mental (APPACDM) of Viseu, Portugal (latitude: 40º37´49´´N, longitude: 7º55´48´´W) (Fig. 1). The physico-chemical properties of the agricultural soil (0-500 mm) were characterized by standard analytical methods [18-22] and are given in Table 1.

Fig. (1). Experimental field of cardoon crop located at Associação Portuguesa de Pais e Amigos do Cidadão Deficiente Mental (APPACDM) of Viseu, Portugal (latitude: 40º37´49´´N, longitude: 7º55´48´´W).

This region has a Mediterranean influence with an annual rainfall of about 1200 mm (76% rainfall between October and March) and a mean annual temperature of 14.0 ºC (mean temperatures of 21.7 ºC in the warmest month and 7.1 ºC in the coldest month). Usually, the agricultural crops are irrigated in the spring/summer seasons (April to September) and rainfed in the autumn/winter seasons. An automatic compact weather station (WS-GP1, Delta-T Devices Ltd, UK) located near the experimental site was used to collect climatic data during the experimental period (Table 2).

The experiment was a randomized complete block design with three replicates and four treatments. The treatments considered were:

(1) Non-amended soil and without irrigation as control (CONTR treatment);

(2) Application of mineral N as basal fertilizer and without irrigation (NITRO treatment);

(3) Non-fertilized and drip irrigation during the growing season (IRRIG treatment);

(4) Application of mineral N dissolved in irrigation water and split by drip irrigation as fertigation (FERTI treatment).

Each plot had an area of 60 m² (10 m width × 6 m length), with 2 m between plots and a density of 10000 plants ha^-1. On April 13, the NITRO treatment received, directly to the soil surface, one application of mineral fertilizer at a rate of 25 kg N ha^-1 (ammonium nitrate (NH₄NO₃)). Then, it was sprayed homogenously and mixed into the upper soil (0-50 mm) using a hoe. From April 13 to August 27, the IRRIG treatment received a total of 460 mm of water by drip irrigation using short irrigation events of 10 mm every 3 days (30% water-filled pore space (WFPS)). For the FERTI treatment, mineral N at a rate of 25 kg N ha^-1 (NH₄NO₃) was dissolved in irrigation water (total=460 mm) and split into short irrigation events of 10 mm every 3 days, similar to those in IRRIG treatment. On April 13, all plots received the same tillage practice of NITRO treatment.

Fluxes of N₂O, CO₂ and CH₄ were measured by the closed chamber technique [18, 23, 24]. Gas measurements were carried out daily during the first 7 days, every 3 days in the next two weeks and then once a week until the end of the experiment. For evaluating the gas fluxes from each treatment, a circular chamber of polyvinyl chloride (Ø=200 mm, h=110 mm), equipped with a septum to sample the interior atmosphere, was inserted into the soil (depth=30 mm). The chambers were kept at fixed locations throughout the sampling dates. After the chamber was closed, a first gas sample (25 mL) was taken (t=0 min) using a plastic syringe and flushed through gas vials (20 mL) and then a second gas sample was taken (t=60 min) from the headspace of the chamber and stored in vials.

The gas samples were analyzed by gas chromatography (GC-2014, Shimadzu, Japan) with the following detectors: electron capture ⁶³Ni detector for N₂O (accuracy=50 ppb to 100 ppm), thermal conductivity detector for CO₂ (accuracy=1 ppm to 1%) and flame ionization detector for CH₄ (accuracy=0.1 ppm to 1%). The N₂O, CO₂ and CH₄ fluxes were determinate by the Eq. (1).

(1)

Where FLU was the N₂O, CH₄ or CO₂ flux on each sampling date (FLU, g N or C m^-2 day^-1), CON was the gas concentration (m³ m^-3 min^-1), MOL was the gas molecular weight (44 g mol^-1 for N₂O or CO₂ and 16 g mol^-1 for CH₄), IDE was the volume of an ideal gas (0.0224 m³ mol^-1), TEM was the temperature during the sampling period (ºC), HEI was the height of the chamber (0.080 m) and TIM was the time corrected for a 1-day duration (1440 min).

The cumulative gas emissions were estimated by averaging the flux between two sampling occasions and multiplying by the time interval between the measurements. The conversion factors of 265 for N₂O and 28 for CH₄ were used to express the Global Warming Potential (GWP) as CO₂-equivalents [18, 23, 24].

The morphological characteristics of cardoon plants were evaluated from July to August 2018, based on international union for the protection of new varieties of plant descriptors [9, 24], being considered the following parameters: Plant Height (PH), Number of Stems (NS), Number of Primary Ramifications (NPR) and Number of Secondary Ramifications (NSR), Number of Inflorescences (NI), Stem Diameter at the Base (SDB) and at the Top (SDT).

In order to evaluate biomass production at the harvest (August 27), all plant materials were collected separately, immediately weighed and oven-dried at 65 ºC (until a constant weight was reached) to determine the dry matter content of each plant component: Leaves Dry Weight (LDW), Stems Dry Weight (SDW), Inflorescences Dry Weight (IDW), Total Plant Dry Weight (TPDW), Total Flowers (TFDW), Total Achens Dry Weight (TADW) and Achenes Yield in the Inflorescences Dry Weight (AYIDW).

Greenhouse gas emissions, morphological characteristics and biomass production were evaluated by analysis of variance (randomized complete block design) to test the effect of treatments (CONTR, NITRO, IRRIG and FERTI). The statistical significance (p<0.05) of the mean differences between treatments was determined by the Tukey Honestly Significant Difference test. The statistical software package used was STATISTIX 10 (Analytical software, Tallahassee, FL, USA).

The daily fluxes of N₂O, CO₂ and CH₄ from treatments are shown in Fig. (2). A significant increase (p<0.05) of N₂O fluxes of amended treatments relative to CONTR treatment was observed during the daily measurements and a first peak was observed on April 13 (114-172 µg N₂O-N m^-2 day^-1) (Fig. 2A). Then, the N₂O fluxes decreased in all treatments followed by a second peak on May 14 (212-310 µg N₂O-N m^-2 day^-1) and a third peak on 18 July (222-245 µg N₂O-N m^-2 day^-1) (Fig. 2A). In the NITRO and FERTI treatments, N₂O fluxes were generally higher than in IRRIG treatment, spite the differences were not statistically significant (p>0.05) (Fig. 2A). Cumulative N₂O emissions from amended treatments were significantly higher (p<0.05) than in the CONTR treatment (three times higher in NITRO and FERTI treatments and one time higher in IRRIG treatment) (Table 3). Thus, the N₂O cumulative emissions did not differ significantly (p>0.05) between NITRO and FERTI treatments but were significantly higher (p<0.05) in ca. 100% relative to IRRIG treatment (Table 3). Previous studies [25-27] reported that the application between 60 and 125 kg ha^-1 of inorganic N led to an increase of N₂O emissions mainly by nitrification and/or denitrification processes, depending on the soil water content, even though nitrifier denitrification can be also a relevant source of this gas, especially under certain soil conditions. Additionally, the increase of the soil WFPS by drip irrigation events have a significant impact on nitrification and denitrification processes [28-30]. Teira-Esmatges et al. [31] reported that nitrification was the dominant source of N₂O in a 30-70% WFPS range. Thus, nitrification may be the major process of N₂O losses in NITRO treatment, but denitrification in the micropores should also be considered in IRRIG and FERTI treatments. In addition, the N₂O cumulative emissions, expressed as a percentage of N applied, were not significantly different (p>0.05) between NITRO and FERTI treatments, representing an average emission rate of 0.5% of N applied (Table 3). The results of this study are in line with Cayuela et al. [32] who suggest an average N₂O emission rate of 0.5% of N input for Mediterranean crops.

Fig. (2). Nitrous oxide (A), carbon dioxide (B) and methane (C) fluxes during the experiment from each treatment. Vertical bars represent standard deviations (n=3).

The CO₂ daily fluxes peaked in all treatments immediately after the beginning of the experiment (19-25 g CO₂ m^-2 day^-1 on April 14) followed by a second peak on July 4 (8-17 g CO₂ m^-2 day^-1) (Fig. 2B). The CO₂ daily fluxes from NITRO treatment were higher in most measurement dates relative to all other treatments, even if differences were not always significant (p>0.05) (Fig. 2B). The CO₂ cumulative emissions present no significant differences (p>0.05) between CONTR and IRRIG treatments (Table 3). However, the same variable from NITRO treatment was significantly higher (p<0.05) in ca. 65% relative to FERTI treatment (Table 3). The application of inorganic N could increase, decrease, or maintain the CO₂ fluxes from soils whereas the incorporation of crop residues enhances the carbon mineralization [33-36]. In the present study, the application of N fertilizer could have increased microbial activity [37] and then increased soil respiration, while the irrigation/rainfall could have reduced the availability of N fertilizer.

The CH₄ daily fluxes varied from negative to positive values (-2107 to 988 mg CH₄ m^-2 day^-1) in all treatments throughout the measurement period (Fig. 2C). Thus, the CH₄ daily fluxes were mainly negative, except in IRRIG and FERTI treatments with positive values (279-988 mg CH₄ m^-2 day^-1) in some measurement dates (May 7-14 and June 6) (Fig. 2C). The CH₄ soil emissions result from the balance between anaerobic methanogenic CH₄ production and aerobic methanotrophic CH₄ consumption, being controlled by soil aeration [29]. Usually, the well-drained agricultural soils act as CH₄ sink. The rainfall between April and June as well as the drip irrigation events should have enhanced the methanogenic activity, resulting in net CH₄ emission [26]. However, the CH₄ cumulative emissions as well as the GWP did not differ significantly (p>0.05) among treatments, with an average value of about -600 kg CH₄ ha^-1 and -6,8 t CO₂ eq. ha^-1, respectively (Table 3).

The morphological characterization and biomass production of cardoon plants per treatment are presented in Tables 4 and 5. On average, plant height was 1561 mm, with the lowest primary ramification obtained at 960 mm and the upper ramification at 1206 mm from the soil. The stalk diameter has an average of 23.7 mm at the base and 7.7 mm at the top. The average number of ramifications on the main stem was discriminated on primary and secondary ramifications with 3.8 and 6.7, respectively. The total number of inflorescences was 17.2, being split in each plant by 2 main inflorescences, 6 primary inflorescences and 9 secondary inflorescences. The length and diameter of the principal inflorescences were lower than secondary inflorescences (61.5 against 51.1 mm for length and 67.2 against 51.5 mm for diameter). The average total production of flowers and seeds per plant was 42.5 g and 77.6 g, respectively. The total average seed weight by inflorescence was 4.5 g; however, the value varied between 8.4 g in the main capitula, 7.5 g and 3.5 g in primary and secondary inflorescences, respectively. The average individual seed weight in the principal, primary and secondary inflorescences was 0.048 g, 0.045 g and 0.041 g, respectively. The seed weight per plant was 77.6 g which represents 13.9% of the inflorescence weight. Each capitulum produced 2.6 g of flowers and 4.4 g of seeds. The flowers harvesting occurred with a frequency of 3-5 days between June 29 till August 10 with a total production of 5338 g. The total seed production was also 7470 g.

The Plant Height (PH) in IRRIG treatment was significantly higher (p<0.05) in about 20% relative to FERTI treatment (Table 4). The Height of Lower Branch (HLB) and Upper Branch (HUB) did not differ significantly (p>0.05) among treatments, although being higher in CONTR treatment and lower in FERTI treatment (Table 4). The number of secondary ramifications (NSR) from the CONTR treatment was significantly higher (p<0.05) in 100% relative to FERTI treatment and 40% than NITRO or IRRIG treatments (Table 4). The number of stems and primary ramifications did not differ significantly (p>0.05) between treatments (Table 4). The CONTR treatment presented the highest number of secondary ramifications and inflorescences followed by IRRIG and the FERTI treatment revealed the lowest number (Table 4). The plants of CONTR treatment showed the lowest height of the first ramification and the highest stem diameter at the base which promoted plants with the highest number of inflorescences (Table 4). The distance between the Lower Branch (HLB) and Upper Branch (HUB) seems to be related with the Number of Secondary Ramifications (NSR) and consequently with the Number of Inflorescences (NI) (Table 4).

The characteristics related with biomass production, namely Leaves Dry Weight (LDW), Stems Dry Weight (SDW), Inflorescences Dry Weight (IDW), Total Plant Dry Weight (TPDW), Total Flowers Dry Weight (TFDW), Total Achenes Dry Weight (TADW), Achenes Yield in the Inflorescences Dry Weight (AYIDW), did not differ significantly (p>0.05) between treatments although numerically higher in IRRIG treatment (Table 5). The CONTR treatment also presented the highest flower production and achenes weight followed by the IRRIG treatment (Table 5). These results are related to the highest number and the largest dimension of inflorescences (Table 5). The plants of IRRIG treatment showed the inflorescences with the highest value of achenes yield (Table 5).

Cardoon biomass production varies considerably with land topography, growing, cultivation management as well as by the genotype [38, 39], being reported a biomass yield that ranges from 1400 to 13700 kg ha^-1 for Mediterranean climate conditions [10, 40]. In Spain, Fernández et al. [1] reported an average partitioning of cardoon was 35.4% basal leaves, 13% cauline leaves, 18.7% stems and 32.9% heads; seeds represented 8-10% of the aboveground biomass (all expressed by dry weight). In the present study, the expected total biomass production per hectare was 10400 kg, namely 4600 kg of stems, 5000 kg of inflorescences and 800 kg of leaves. The plants also produced 425 kg of flowers and 750 kg of seeds. These values are according to data obtained in plants of a second-year growth cycle [9, 24]. The weight of leaves obtained in this study (816 kg) was considerably lower compared with the potential if all leaves were collected (4330 kg). In fact, the weight of flowers collected could vary with the intensity of harvesting. The estimated production of the seeds was 750 kg and could vary significantly between fields and depends on plant genotype. Seed harvesting should be done before the dissemination process when the inflorescences are mature. The seed yield obtained from the inflorescences corresponds to 14%, which is below the values reported by other reference studies (25.5-28.7%) [3, 6]. To maximize seed production, flower harvesting should be done 2-3 days after blossom.