A Sustainability Assessment of the Greenseeker N Management Tool: A Lysimetric Experiment on Barley

A preliminary study was conducted to analyze the sustainability of barley production through: (i) investigating sensor-based nitrogen (N) application on barley performance, compared with conventional N management (CT); (ii) assessing the potential of the Normalized Difference Vegetation Index (NDVI) at different growth stages for within-season predictions of crop parameters; and (iii) evaluating sensor-based fertilization benefits in the form of greenhouse gasses mitigation. Barley was grown under CT, sensor-based management (RF) and with no N fertilization (Control). NDVI measurements and RF fertilization were performed using a GreenSeekerTM 505 hand-held optical sensor. Gas emissions were measured using a static chamber method with a portable gas analyzer. Results showed that barley yield was not statistically different under RF and CF, while they both differed significantly from Control. Highly significant positive correlations were observed between NDVI and production parameters at harvesting from the middle of stem elongation to the medium milk stage across treatments. Our findings suggest that RF is able to decrease CO2 emission in comparison with CF. The relationship between N fertilization and CH4 emission showed high variability. These preliminary results provide an indication of the benefits achieved using a simple proximal sensing methodology to support N fertilization.


Introduction
In recent years, one of the most investigated solutions for improving fertilization efficiency involves taking into account the spatial and temporal crop and soil variability within a field, in an approach called "precision farming management" [1]. This approach is the result of the development and implementation of various technologies, such as the Global Positioning System (GPS), Geographic Information System (GIS), automatic control, variable rate technologies, proximal and remote sensing, computer-driven control devices and telecommunications [2]. Variable rate technologies are applicable to any input factor or crop but have been more frequently adopted for grain crops and fertilization operations [3]. The variable techniques are based on two methods of managing field variability, namely prescription maps based on historical field information and real-time sensing. The first approach is more popular among farmers, although the real-time method is currently receiving attention due to recent developments in proximal sensing technologies. In the past 20 years, several studies have been carried out on the use of precision farming technologies to evaluate crops' agronomic traits, as well as stress detection and quantification. Naser et al. [4] used proximal sensor derived Normalized Difference Vegetation Index (NDVI) to discriminate productivity between different wheat genotypes. Randelović et al. [5] used a machine learning model and vegetation indices extracted by a Unmanned minimum, and maximum temperatures and rainfall) were collected by the Regional Hydrological Service weather station, close to the experimental field [47]. Barley (Hordeum vulgare L.) was sown on 13 November 2017 using seeding density of 400 seeds m −2 and row spacing of 0.1 m in cylindrical tanks of 1 m 3 (height: 0.77 m; area: 1.33 m 2 ). The dimensions of tanks were considered suitable for barley growth based on Poorter et al. [48] and Fan et al. [49]. In fact, Fan et al. [49] reported that, on average, 50% of the total root amount in barley was accumulated within the upper 0.12 m soil profile, 67-76% of roots can be found within the upper 0.3 m, and 95% of total root amount was accumulated within 0.99 m. Further, Poorter et al. [48] suggested that pots with a plant biomass to soil volume ratio of less than 1 g L -1 , such as in our case, had a plant biomass to soil volume ratio of the same order of magnitude as those of plants growing in the field. The tanks were filled in 2015 with a silty clay soil collected from the 0-0.9 m layer of a conventionally cultivated corn field. The soil was placed in the tanks maintaining the original profile's order of layers and taking care to induce only temporary changes in soil structure [50]. Tanks were seeded with barley (Hordeum vulgare L.) on 13 November 2017 using seeding density of 400 seeds m −2 and row spacing of 0.1 m. Three N treatments were carried out in triplicate: a control treatment (Control: 0 kg N ha −1 ), a conventional rate (CF: 150 kg N ha −1 ) and a sensor-based N fertilization rate (RF) based on optical sensor measurements. The method to determine N fertilization rate by proximal sensing was the same as reported by Foster et al. [51], using the specific "Fertilizer Estimation Chart" adapted for Italian barley cultivation. A N-rich strip (150 kg N ha −1 ) was used as no limiting N, to calculate the variable N rate during the second top dress fertilization. Measures were performed on the final phase of stem elongation (BBCH, 39) to establish the N variable dose. The average target yield under the no limiting condition for Italian barley production was set to 7 t ha −1 , whereby the amount of N supplied in the top dressing on RF, considered optimum to reach maximum yield, was 37.5 kg ha −1 . No N was applied at sowing, whilst N was applied as ammonium nitrate (N: 27%) in two spring top dress fertilizations on 26 March and 24 April 2018, respectively. In the first top dressing, N fertilization 75 kg ha −1 (as the fixed rate) and 37.5 kg ha −1 were broadcasted in CF and RF, respectively. The remaining N was scheduled in the second top dress fertilization as a fixed rate for CF, 75 kg ha −1 , and variable rate for RF, 37.5 kg ha −1 . Barley was manually harvested at maturity (15 July 2018).

GHGs Emission Measurements and Flux Estimation
Fluxes of GHGs (CO 2 , CH 4 , N 2 O, NH 3 ) were monitored using a static chamber method as reported in Verdi et al. [52]. Nine static chambers, one per tank, were built according to Parkin and Venterea [53]. Static chambers were comprised of a cylindric anchor (PVC, height of 15 cm and diameter of 20 cm) and a lid (PVC, height of 25 cm and diameter of 20 cm). The static chambers were covered by a reflective Mylar tape to cover solar reflectance. Anchors were inserted into the soil for approximately 5 cm immediately after sowing and remained in that position for the entire growing season to reduce root system disturbance. For each sampling date, the lid was placed on the anchors only during the measurement period (1 h). Measurements were performed by means of a portable gas analyzer Madur MaMoS (Madur, Zgierz, Poland) [52]. The gas analyzer used Nondispersive Infrared technology (NDIR) sensors for the detection of CO 2 , CH 4 and NH 3 (±1 mg kg −1 ) concentrations in the air sample and electrochemical technology for N 2 O (±1 mg kg −1 ). The time response of all sensor was approximately 30 s. Measurements were carried out for a total of 18 days, once daily during the first 6 days after fertilization and then every two days for the remaining time period. Interpolation methodology was adopted to obtain missing data from the days where measurements were not performed. Each measurement was composed of two subsequent samplings per chamber during which the sensor was held within the chamber for 1 min. The first sampling was performed just after chamber closing (T0) and then repeated after a 1 h interval of gas accumulation in the closed chamber (T1). The difference in gas concentration between T1 and T0 was taken as the emission data, which was used for gas flux calculations. Measurements were performed mid-morning, which was representative of a time frame more closely corresponding to average daily temperature [53]. Emission Sustainability 2020, 12, 7303 4 of 16 fluxes were calculated using gas concentration (mg kg −1 ), chamber dimensions (volume and area), molar weight of each gas and closing time. Results of this calculation provided the rate of CO 2 and CH 4 emissions in terms of carbon per unit area (kg C ha −1 ) and the rate of N 2 O and NH 3 emissions in terms of nitrogen per unit area (kg N ha −1 ). Due to the reflective coverage, temperature increases inside the chambers were avoided. Nonetheless, temperature was monitored with two thermocouples. Knowing temperature and atmospheric pressure, air volume contained in each chamber was reduced to standard conditions (22.4 L at 15 • C and 1 bar).

Soil and Crop Analysis
Soil sampling was performed in triplicate for each tank in the 0-40 cm layer. Soil samples were air-dried and then analyzed for the determination of bulk density [54], pH, soil texture [55], total calcium carbonate [56], organic matter content (OM) [57] and total nitrogen (Nsoil) [58] (Table 1). At harvest, whole-plant material was collected from each tank. In the laboratory, the straw and grain were separated, after which fresh biomass and mass of dry matter, dry matter content, and crop N uptake were determined. The mass of dry matter of the straw and grain was measured after oven-drying for 48 h at 80 • C [59]. Thereafter, the dried samples were weighed to determine the accurate by-plot biomass and grain yield. From each by-plot sample, sub-samples were taken and analyzed to detect N content in the straw and grains (Flash EA 1112-ThermoFisher, Waltham, MA, USA). The 3 replicate tanks for each treatment were analyzed separately, and then the average between replicates was determined.

Economic Evaluation Method
For economic evaluation of the different systems, Partial Budgeting was chosen as the method. This is commonly used for precision farming techniques evaluation, and it calculates changes in profit due to changes in inputs. The Partial Budgeting of CF and RF treatments was compared, subdividing economic and environmental partial costs. The net return from different systems, defined as total gains minus variable costs, was used to measure profitability. In this study, the total gains were the grain sale profits per farmer (Sp), considered in relation to 0.18 € kg −1 [60]. On the other hand, the variable costs include only fertilization (Fc), considering a fertilizer cost of 0.28 € kg −1 [61].
The economic partial cost was calculated at a field scale as the difference between the product price and fertilizer cost as reported in Equation (1). However, the environmental partial cost (3) evaluation was also performed, adding to the cost due to CO 2 emission equivalents. The environmental partial cost was subdivided in "Fertilizer production cost" (CO 2 Fpc) and "Fertilization cost" (CO 2 Fc). Regarding the "Fertilization cost" for each treatment, the total net Global Warming Potential (GWP; kg CO 2 -eq ha −1 ) was calculated, as reported in Equation (2). The net GWP was used to estimate the kg CO 2 emissions equivalent for each GHG [62]. The GHG emissions data used were cumulated and measured after fertilization. The price per kg CO 2 emissions equivalent (0.23 € kg −1 of CO 2 ) was obtained from the European Emission Allowances (EUA) website. Thereafter, the "Fertilizer production cost" was calculated from the database "mineral fertilizer carbon footprint reference values: 2011" [63] as the amount of CO 2 emissions produced from using ammonium nitrate fertilizer (3.06 kg CO 2 per kg of ammonium nitrate).

Statistical Analysis
Statistical analysis was performed by using R studio software (version 1.0.143) (RStudio PBC, Waltham, MA, USA). All data were investigated for normality using the Shapiro-Wilk's test. Differences between treatments in grain yield and biomass values resulted in being normally distributed, thus they were statistically evaluated by means of analysis of variance (ANOVA), followed by Tukey's honest significant difference test (Tukey HSD) at significance level p < 0.05. The Coefficient of Determination (R 2 ) was used to analyze the relationship between NDVI, calculated at different growth stages of barley, with the values of the mass of dry matter of the whole biomass, the mass of dry matter of grain, and the nitrogen content in the whole biomass at harvesting. As flux data did not show normal distribution, differences between treatments at each sampling event and in the cumulative emissions were analyzed by means of the Kruskal-Wallis (K-W) test (p < 0.05). Then, Dunn's post-hoc tests, with Bonferroni's p value adjustment method, were used for pairwise multiple comparisons. All experiments were carried out with three replicates as previously indicated in the description of the experiments.

Meteorological Conditions during the Study Period
The dynamics of daily air temperature and rainfall are shown in Figure 1.

Statistical Analysis
Statistical analysis was performed by using R studio software (version 1.0.143) (RStudio PBC, Waltham, MA, USA). All data were investigated for normality using the Shapiro-Wilk's test. Differences between treatments in grain yield and biomass values resulted in being normally distributed, thus they were statistically evaluated by means of analysis of variance (ANOVA), followed by Tukey's honest significant difference test (Tukey HSD) at significance level p < 0.05. The Coefficient of Determination (R 2 ) was used to analyze the relationship between NDVI, calculated at different growth stages of barley, with the values of the mass of dry matter of the whole biomass, the mass of dry matter of grain, and the nitrogen content in the whole biomass at harvesting. As flux data did not show normal distribution, differences between treatments at each sampling event and in the cumulative emissions were analyzed by means of the Kruskal-Wallis (K-W) test (p < 0.05). Then, Dunn's post-hoc tests, with Bonferroni's p value adjustment method, were used for pairwise multiple comparisons. All experiments were carried out with three replicates as previously indicated in the description of the experiments.

Meteorological Conditions during the Study Period
The dynamics of daily air temperature and rainfall are shown in Figure 1. The cumulated rainfall during the growing period was 770 mm. Rainfall was mainly concentrated from November to April (the highest monthly cumulative precipitation was in March, The cumulated rainfall during the growing period was 770 mm. Rainfall was mainly concentrated from November to April (the highest monthly cumulative precipitation was in March, corresponding to the initiation of the tillering stage in barley). Each sampling period corresponded to differing rainfall conditions, namely, 43 mm during the first (tillering stage) and 59.2 during the second (stem elongation stage) measurement, respectively. An average temperature value of 14.3 • C was measured during emissions sampling after the 1st fertilization (constant), while it reached 20 • C after the 2nd fertilization (temperature to decrease at the end of the sampling date).

Yield Responses to N Rates and Economic Benefits
Barley yields at different N application rates are shown in Figure 2. Grain yield was significantly (p < 0.05) improved by N fertilization application with respect to Control. The results showed that the average CF (4.3 ± 0.2 t ha −1 ) and the RF grain yields (3.5 ± 0.5 t ha −1 ) were significantly higher than the Control by 82.6% and 49.3%, respectively. No significant differences (p > 0.05) between the grain yield of CF and RF were observed. Various studies have been performed comparing uniform N fertilization with variable rate fertilization techniques. Bragagnolo et al. [64] reported no significant differences in corn grain yield when applying either a traditional N fertilization application (150 kg ha −1 ) or a variable rate fertilization based on a crop optical sensor. Mayfield and Trengove [65] similarly made a comparison between variable rate fertilization and a uniform application fixed rate, showing non-significant differences. Therefore, results indicated that by applying less N fertilizer based on a sensor-based technique, it was possible to obtain the same production as traditional practice, leading to fertilizer saving.

Yield Responses to N Rates and Economic Benefits
Barley yields at different N application rates are shown in Figure 2. Grain yield was significantly (p < 0.05) improved by N fertilization application with respect to Control. The results showed that the average CF (4.3 ± 0.2 t ha −1 ) and the RF grain yields (3.5 ± 0.5 t ha −1 ) were significantly higher than the Control by 82.6% and 49.3%, respectively. No significant differences (p > 0.05) between the grain yield of CF and RF were observed. Various studies have been performed comparing uniform N fertilization with variable rate fertilization techniques. Bragagnolo et al. [64] reported no significant differences in corn grain yield when applying either a traditional N fertilization application (150 kg ha −1 ) or a variable rate fertilization based on a crop optical sensor. Mayfield and Trengove [65] similarly made a comparison between variable rate fertilization and a uniform application fixed rate, showing nonsignificant differences. Therefore, results indicated that by applying less N fertilizer based on a sensor-based technique, it was possible to obtain the same production as traditional practice, leading to fertilizer saving. Partial Budgeting of the different fertilization approaches (Table 2) resulted in higher fertilizer cost, amounting to 81 € ha −1 for CF compared with RF. Partial Budgeting of the different fertilization approaches (Table 2) resulted in higher fertilizer cost, amounting to 81 € ha −1 for CF compared with RF.
The environmental costs, in terms of CO 2 production, significantly (p < 0.05) increased from Control to RF and then CF. The economic profit of Control was significantly (p < 0.05) lower than RF and CF by about 206.76 € ha −1 and 329.51 € ha −1 , respectively. On the other hand, the profit difference between RF and CF (122.75 € ha −1 ) was not statistically significant (p > 0.05). There was no significant difference (p > 0.05) between Control and RF in term of environmental profit, while the environmental profit for CF was significantly (p < 0.05) lower than RF and Control by about 266.89 € ha −1 and 306.05 € ha −1 , respectively. Therefore, CF and RF showed to have the same cost benefits, attributable to the increase in yield in comparison to the costs spent. Considering the environmental cost that might be paid by farmers for the use of N, the evaluation is in favor of variable rate fertilization due to the high application rate of the conventional practice. It is important to analyze this assumption because new environmental policies are being built up to protect ecosystems and climate, applying variable costs to different losses of N [66,67].

NDVI Trend during Crop Development
No significant differences (p > 0.05) were observed between CF and RF treatments during the monitoring period, even though the fertilization rates were different ( Figure 3).
The environmental costs, in terms of CO2 production, significantly (p < 0.05) increased from Control to RF and then CF. The economic profit of Control was significantly (p < 0.05) lower than RF and CF by about 206.76 € ha −1 and 329.51 € ha −1 , respectively. On the other hand, the profit difference between RF and CF (122.75 € ha −1 ) was not statistically significant (p > 0.05). There was no significant difference (p > 0.05) between Control and RF in term of environmental profit, while the environmental profit for CF was significantly (p < 0.05) lower than RF and Control by about 266.89 € ha −1 and 306.05 € ha −1 , respectively. Therefore, CF and RF showed to have the same cost benefits, attributable to the increase in yield in comparison to the costs spent. Considering the environmental cost that might be paid by farmers for the use of N, the evaluation is in favor of variable rate fertilization due to the high application rate of the conventional practice. It is important to analyze this assumption because new environmental policies are being built up to protect ecosystems and climate, applying variable costs to different losses of N [66,67].

NDVI Trend during Crop Development
No significant differences (p > 0.05) were observed between CF and RF treatments during the monitoring period, even though the fertilization rates were different (Figure 3).
Even during the senescence stage, RF tanks showed higher NDVI than CF. Since NDVI is a way to measure plant health, in this experiment crops responded well to RF management: it thus appears that crops did not require additional N. On the contrary, significant differences were observed between the Control and the two fertilizer treatments starting from the flag leaf stage (BBCH 39). It was shown previously that NDVI is directly dependent on N fertilization [68]. The lower the N level supplied to crops, the lower the chlorophyll content and the lower the absorption of radiation in the visible region, which in turn results in a reduction of NDVI. Therefore, the lack of N in control tanks led to premature yellowing of crops and a stunted vegetative state.

Possible Use of NDVI Measurements to Predict Plant, Grain Weight, and N Content
The correlation coefficients for the analyzed parameters were higher and highly significant (p < 0.05) from the middle of the stem elongation stage (BBCH 32) to the medium milk stage (BBCH 75) ( Table 3). In the present study, no correlation was observed between NDVI (measured either at tillering or at the beginning of senescence) and final total biomass and mass of dry matter of grain. The highest positive and significant correlation (p < 0.05) was performed at the beginning of booting, followed by a slow decrease until the maturity stage, performing GreenSeeker measurements. Regarding the relationship between N content and NDVI, there was a significant correlation (p < 0.05) Figure 3. NDVI values of barley measured at different growth stages (BBCH scale) for control (Control), variable rate fertilization (RF) and conventional fertilization (CF) treatments. Error bars represent standard deviation of means. Different letters indicate significant difference (p < 0.05) between treatments according to the Kruskal-Wallis test, followed by the Dunn's post hoc test.
Even during the senescence stage, RF tanks showed higher NDVI than CF. Since NDVI is a way to measure plant health, in this experiment crops responded well to RF management: it thus appears that crops did not require additional N. On the contrary, significant differences were observed between the Control and the two fertilizer treatments starting from the flag leaf stage (BBCH 39). It was shown previously that NDVI is directly dependent on N fertilization [68]. The lower the N level supplied to crops, the lower the chlorophyll content and the lower the absorption of radiation in the visible region, which in turn results in a reduction of NDVI. Therefore, the lack of N in control tanks led to premature yellowing of crops and a stunted vegetative state.

Possible Use of NDVI Measurements to Predict Plant, Grain Weight, and N Content
The correlation coefficients for the analyzed parameters were higher and highly significant (p < 0.05) from the middle of the stem elongation stage (BBCH 32) to the medium milk stage (BBCH 75) ( Table 3). In the present study, no correlation was observed between NDVI (measured either at tillering or at the beginning of senescence) and final total biomass and mass of dry matter of grain. The highest Sustainability 2020, 12, 7303 8 of 16 positive and significant correlation (p < 0.05) was performed at the beginning of booting, followed by a slow decrease until the maturity stage, performing GreenSeeker measurements. Regarding the relationship between N content and NDVI, there was a significant correlation (p < 0.05) from BBCH 32 (stem elongation) to BBCH 75 (medium milk). In this study, the highest significant (p < 0.05) correlation for estimating the N concentration was found at the beginning of the heading stage (BBCH 51). This was a suitable stage to detect the final characteristics of crops (and is also suitable for biomass and grain weight). However, NDVI values measured during the early (23 and 30 BBCH) and final crop stages (85 and 90 BBCH), respectively, were unsuitable when predicting the final N content of barley. Thus, results suggested that at the beginning of booting (around 39 BBCH) is the most appropriate time to predict barley total biomass and grain biomass, while flowering (51 BBCH) is the best time to predict quality parameters (Figure 4). Table 3. Determination coefficient between mass of dry matter of total biomass, mass of dry matter of grain, nitrogen content in the whole biomass and NDVI value calculated at different growth stages on barley. The symbols * and ** indicate significant levels for p < 0.05 and p < 0.01, respectively. These results were consistent with Hassan et al. [69] who reported that NDVI at booting stage can be used to predict grain yield of rainfed wheat, showing R 2 values ranging from 0.38 to 0.90 from stem elongation to late grain filling across the treatments. Moreover, NDVI at heading stage can also be sensitive to yield changes at different N rates. Similar results were also found for wheat, showing a high positive correlation (R 2 = 0.78) between NDVI and grain yield at the booting stage [70]. Spitkó et al. [71] similarly showed a weak positive relationship between NDVI and corn yield in the early stages and significant (p < 0.05) association in plant development. In this study, the lack of a positive and significant correlation (p > 0.05) between NDVI and biomass at the tillering stage could be due to the low canopy cover and the influence of soil background at early stages [72]. Instead, the lack of a significant relationship (p > 0.05) between NDVI and biomass and NDVI and yield at the senescence stage could be due to the yellowing of crops across the treatments [73]. Moreover, the relationship between biomass and NDVI showed similar results from BBCH stages 32 to 75, probably due to the saturation effect of indices when crop canopy closes over time [74].

Mass of Dry Matter of
Regarding the relationship between N content and NDVI, the present results corroborated previous findings, as a study carried out on wheat that showed correlation between NDVI measured in different growth stages and grain properties, thereby revealing high relationships (R 2 > 0.90 at heading stage) for both grain/biomass yield and N content [75].

Effect of Fertilization Method on GHGs Emissions
During the growing period, no N2O and NH3 emissions were measured. It is likely that N2O and NH3 emissions throughout the whole experimental period were below the minimum detectable level of the gas measuring instrument. In addition, ammonium nitrate is known to have less volatilization and denitrification compared to other N sources. In this context, Chu et al. [43] found the maximum N2O fluxes in winter barley to be less than 50 µg m −2 h −1 . The incorporation of the fertilizer into the soil may have substantially contributed to the prevention of N2O and NH3 emissions, as suggested by Boeckx et al. [76].
CO2 emissions increased approximately one week after the first fertilization treatment ( Figure 5).

Effect of Fertilization Method on GHGs Emissions
During the growing period, no N 2 O and NH 3 emissions were measured. It is likely that N 2 O and NH 3 emissions throughout the whole experimental period were below the minimum detectable level of the gas measuring instrument. In addition, ammonium nitrate is known to have less volatilization and denitrification compared to other N sources. In this context, Chu et al. [43] found the maximum N 2 O fluxes in winter barley to be less than 50 µg m −2 h −1 . The incorporation of the fertilizer into the soil may have substantially contributed to the prevention of N 2 O and NH 3 emissions, as suggested by Boeckx et al. [76]. CO 2 emissions increased approximately one week after the first fertilization treatment ( Figure 5). CO 2 fluxes were low from day 1 to day 6, and no significant differences (p > 0.05) were found between treatments. Then, starting from day 7, CO 2 emissions rapidly increased. Furthermore, significant differences (p < 0.05) between Control and CF were observed for all sampling dates. The differences between Control and RF were only significant (p < 0.05) within a five-day time frame over the monitoring period. Significant differences (p < 0.05) were evident between Control and N fertilization treatments but not (p > 0.05) between CF and RF. CO2 fluxes were low from day 1 to day 6, and no significant differences (p > 0.05) were found between treatments. Then, starting from day 7, CO2 emissions rapidly increased. Furthermore, significant differences (p < 0.05) between Control and CF were observed for all sampling dates. The differences between Control and RF were only significant (p < 0.05) within a five-day time frame over the monitoring period. Significant differences (p < 0.05) were evident between Control and N fertilization treatments but not (p > 0.05) between CF and RF.
After the second fertilization, the CO2 emissions from RF were significantly (p < 0.05) lower than those from CF until day 13, after which no significant differences (p > 0.05) were found. Additionally, no significant differences (p > 0.05) were found between RF and Control until day 10. The emissions might be influenced by differing ecosystem respiration, as well as temperature and soil moisture. During the second sampling, the CO2 emissions from the CF treatment were significantly (p < 0.05) higher than the other treatments for the first 10 days, with a peak at day five, where the highest temperature was also evident (Figure 1). The CO2 emissions from the Control were higher from day 15. During the second fertilization, the average emission measured for CF was significantly higher (p < 0.05) than those measured for RF and Control (Table 4). In the present study, the overall increase in CO2 emissions observed from the first to the second fertilization, respectively, was invariably attributable to higher temperatures. The average air temperature was 20 °C from the end of April to the beginning of May, compared with the average air temperature of 14.3 °C at the time of the first fertilization treatment (Figure 1). Significant differences (p < 0.05) in CO2 emissions between the different tanks, regardless of the fertilization treatment, were found with respect to the initial soil C content. The results showed that for the Control, the organic C content of the soil was negatively correlated with CO₂ emissions. On the contrary, results indicated that emissions were positively correlated with the initial C content when N was supplied through fertilization.
In this study analyzing the dynamics of CH4 emissions from all treatments after both the first and the second fertilization, the results showed lower average emissions with respect to CO2 (Table  4). However, flux was similar with increases occurring during crop growth. Once again, these results might be attributable to weather conditions. The first fertilization treatment was characterized by lower rainfall conditions compared with the second treatment (43 mm F1 and 59.2 mm F2, After the second fertilization, the CO 2 emissions from RF were significantly (p < 0.05) lower than those from CF until day 13, after which no significant differences (p > 0.05) were found. Additionally, no significant differences (p > 0.05) were found between RF and Control until day 10. The emissions might be influenced by differing ecosystem respiration, as well as temperature and soil moisture. During the second sampling, the CO 2 emissions from the CF treatment were significantly (p < 0.05) higher than the other treatments for the first 10 days, with a peak at day five, where the highest temperature was also evident (Figure 1). The CO 2 emissions from the Control were higher from day 15. During the second fertilization, the average emission measured for CF was significantly higher (p < 0.05) than those measured for RF and Control (Table 4). In the present study, the overall increase in CO 2 emissions observed from the first to the second fertilization, respectively, was invariably attributable to higher temperatures. The average air temperature was 20 • C from the end of April to the beginning of May, compared with the average air temperature of 14.3 • C at the time of the first fertilization treatment ( Figure 1). Significant differences (p < 0.05) in CO 2 emissions between the different tanks, regardless of the fertilization treatment, were found with respect to the initial soil C content. The results showed that for the Control, the organic C content of the soil was negatively correlated with CO 2 emissions. On the contrary, results indicated that emissions were positively correlated with the initial C content when N was supplied through fertilization.
In this study analyzing the dynamics of CH 4 emissions from all treatments after both the first and the second fertilization, the results showed lower average emissions with respect to CO 2 (Table 4). However, flux was similar with increases occurring during crop growth. Once again, these results might be attributable to weather conditions. The first fertilization treatment was characterized by lower rainfall conditions compared with the second treatment (43 mm F1 and 59.2 mm F2, respectively, see Figure 1), as well as colder temperatures. In the present study, owing to the variability of emissions during crop growth, it was not possible to carry out an appropriate statistical analysis of CH 4 production. After the second fertilization, CH 4 emissions revealed an opposite trend compared with that observed after the first fertilization treatment. The daily CH 4 emissions varied after the second fertilization, showing a different duration that was related to N treatment. CH 4 emissions were measured until day 12, 3 and 5 for the Control, RF and CF tanks, respectively. Moreover, for Control, daily emissions attained a level of 1.76 kg C ha −1 day −1 , compared with 1.25 and 0.7 kg C ha −1 day −1 for RF and CF, respectively. Grasslands studies showed that the ecosystem respiration in 60 kg N ha −1 plots was not significant compared to the control plots (no N), while ecosystem respiration in 120 and 240 kg N ha −1 treatment plots, respectively, was significantly enhanced relative to the control plots [77]. In general, N content influences soil carbon emissions, thereby affecting microbial activity, root respiration and chemical decay processes [78]. Previous studies have been conducted on the effect of N fertilization on CO 2 emissions and microbiological activity. It was shown that increasing N fertilization was able to depress soil CO 2 fluxes, probably due to a reduction of microbial activity [79]. On the other hand, Sainju et al. [80] demonstrated that N fertilization was able to increase CO 2 flux compared with no N fertilization, probably as a result of accelerated root respiration due to enhanced crop growth. Our results were consistent with those of Tanveer et al. [81], who reported that CO 2 emission fluxes increased with the enhancing of crop growth and air temperature. The positive relationship between CO 2 emission and the initial C content may have been attributable to soil microbial activity, which is strongly influenced by C soil content [82]. Previous results showed that N availability can be substantially influenced by the effect of C availability on CO 2 emissions, depending on the C/N ratio in the soil. Under low N, the C/N ratio is increased, and as a result microbial activity then becomes depressed by the lack of N, which is essential for protein synthesis. Hence, emissions are not shown to increase with increasing C in soil [83]. Instead, if N is available in the soil, microbial activity is stimulated, and an increase in emissions with increasing C is evident. N fertilization stimulated labile soil organic matter (SOM) decomposition, which is derived from soil residues. Hence, N fertilization supplies bioavailable N for microbes to generate hydrolase and achieve resources [84].
In regard to CH 4 , the exchange between croplands and the atmosphere is also affected by N fertilization [85]. Reports from the literature show variable results. In some studies, CH 4 emissions were activated [86,87], whereas in others, they appeared inhibited [88], and in some cases, no significant effects were reported [89]. As reported previously, in wetland soils, warm spring rainfall was shown to promote CH 4 production from the soil [90]. Chu et al. [43], in an experiment carried out on barley, showed that N fertilization application increased CH 4 emissions, probably due to an effect on CH 4 uptake. However, Chu et al. [43] again found positive CH 4 emissions from no fertilized plots after March, linking this to the combination of soil and weather conditions.
Our results indicated that in the experiment, the rate of CO 2 increased significantly with the addiction of N fertilizer, leading to larger emissions. On the other hand, CH 4 emissions seemed to be more related to factors other than N fertilization. Deeper research has to be carried out in other similar studies to validate the results.

Conclusions
More suitable agricultural practices are required in Italian cereals farms, where farmers use techniques affected by tradition. Precision N fertilization techniques are reported to be modern and valid strategies for enhancing productivity and economy and decreasing environmental impact. This study tried to briefly evaluate these aspects, applying a specific sensor-based N rate for Italian barley production and comparing it with conventional N management. Results evidenced that GreenSeeker can be a useful tool for monitoring crop growth and managing N fertilizer in barley. In fact, GreenSeeker proved to be able in predicting barley crop quantity and quality traits at the booting stage. The method used to apply variable N has been proved to be effective, leading to a similar yield to conventional practice (150 kg N ha −1 ) while using less fertilizer (75 kg N ha −1 ). Additionally, the variable rate fertilization method has been proved to be a valid alternative to traditional fertilization, considering environmental impact in addition to economic evaluation and leading a saving of 266 € ha −1 . Nowadays, the issue of the environmental cost of nutrients is spreading, and the possibility of adding taxes to the social cost of N might create serious problems for farmers. Analysis indicated that in the present study, the abatement of N rate for variable fertilization has led to lower CO 2 production from the crop system. Unfortunately, the high variability in the emission data hampered a deep understanding of the relationship between N content and CH 4 production in soil. In addition, it was not possible to detect N 2 O and NH 3 emissions during the experiment. Due to the variability of the results, it is not possible to assert that the specific variable rate technique leads to a GHG emission reduction compared with conventional practice. Further, a more in-depth knowledge of spatial and temporal interactions between the soil-plant-atmosphere system is essential. Therefore, this study showed the potential of a proximal sensor-based fertilization to increase yield and farmers' income, but further analysis is necessary to assess the effect of N fertilizer rates on C emissions dynamics. Funding: The project was financed by the project "PANE + DAYS" co-financed under Tuscany FEASR 2014-2020 Rural Development Programme; Measure 16.2; GO-PEI; the project was also supported by "Fondazione Cassa di Risparmio di Firenze" and "Fondazione per il Clima e la Sostenibilità". The authors also wish to thank Roberto Vivoli from DAGRI for his support during the whole experimentation.

Conflicts of Interest:
The authors declare no conflict of interest.