**3. Results**

#### *3.1. Weather Conditions*

During the 4-year experiment (September 2013–August 2017), the average annual temperature was 15.3 ◦C, and the cumulated yearly precipitation was 865 mm. In all years, most rainfall events were observed from September to April (wet period, on average 602 mm cumulated rainfalls), and during this period, there was a clear gradient going from 2013/14 (extremely wet, 822 mm) to 2014/15 (highly wet, 645 mm), to 2015/16 (dry, 522 mm) and 2016/17 (very dry, 420 mm).

Two extreme events were recorded in the first and last years: a severe hail was recorded on 12 June 2014 and a late frost event with temperature of −1.44 ◦C was recorded on 22 April 2017 (Figure 1).

**Figure 1.** Monthly rainfall (mm) and mean, max and minimum air temperature values recorded at the experimental field station 'FieldLab' (Papiano, Perugia, Italy) during the 4-year experiment.

## *3.2. N Balance*

In order to give an overview of the whole cropping system scale, N budgets were calculated for each cropping system as yearly averages on a 4-year basis (2013/14–2016/17, Table 1; the data for each main crop are reported in Table S2). The INT received a fixed amount of 155 kg N ha−<sup>1</sup> yr<sup>−</sup><sup>1</sup> of extra-farm N fertilizer distributed to durum wheat (160 kg N ha−<sup>1</sup> yr<sup>−</sup>1) and processing tomato (150 kg N ha−<sup>1</sup> yr<sup>−</sup>1). The ORG received a fixed amount of N as poultry manure at durum wheat sowing, and a variable amount of extra-farm N via fertigation to the processing tomato (see Materials and methods section for details). Therefore, the yearly average of extra-farm N fertilizer added to ORG was 69.8 kg N ha−<sup>1</sup> yr<sup>−</sup>1. In ORG+, durum wheat was not fertilized with extra-farm source, however the N rates applied with fertigation to the processing tomato were generally higher than those in ORG; therefore, the yearly amount of N fertilizer added to ORG+ was, on average, 66.16 kg N ha−<sup>1</sup> yr<sup>−</sup><sup>1</sup> (i.e., not statistically different from ORG, *p* < 0.01).


**Table 1.** Annual mean values (2-crop and 4-year basis, kg N ha−<sup>1</sup> yr<sup>−</sup>1) of Nitrogen inputs (Ninput), outputs (Noutput), surplus (Nsurplus) and N lost by leaching (Nleaching) during the experiment for the three cropping systems: integrated (INT), traditional organic (ORG) and innovative organic (ORG+). On the same row, values followed by different letters are statistically different (*p* < 0.05).

The highest Ndfa and N supplied with seeds were observed in ORG+ (legume component of CC + faba bean in TIC) followed by ORG. While the INT had no CC, neither legume was in crop rotation so the Ndfa was zero and the N added with the seeds was the lowest.

The INT was the system with the highest overall N input, while ORG had the lowest input. The average N output ranged from 44 kg N ha−<sup>1</sup> yr<sup>−</sup><sup>1</sup> in ORG+ to 104 kg N ha−<sup>1</sup> yr<sup>−</sup><sup>1</sup> in INT, while ORG showed an intermediate value of 76 kg N ha−<sup>1</sup> yr<sup>−</sup>1. INT and ORG+ had the highest N surplus values (58.1 and 72.8 kg N ha−<sup>1</sup> yr<sup>−</sup>1, respectively) while ORG the lowest (28.6 kg N ha−<sup>1</sup> yr<sup>−</sup>1). The N lost by leaching in INT was twofold compared to that observed in ORG and ORG+ (27.3 kg N ha−<sup>1</sup> yr −1 on average, Table 1).

#### *3.3. Yield and N Leaching*

In all systems, durum wheat yield was low in 2013/14 and 2016/17 (due to a severe hail event and a late frost, respectively). In these two years, the systems did not show any significant difference in terms of yield (Figure 2A). In 2014/15 and 2015/16, the yields were generally higher and the ranking among systems was the same (i.e., INT > ORG > ORG+, *p* < 0.05).

**Figure 2.** Yield (Mg ha−1) and N leaching (kg N ha−1) of durum wheat (top: (**A**) and (**B**), respectively) and processing tomato (bottom: (**C**) and (**D**), respectively) in the three cropping systems: integrated (INT), traditional organic (ORG) and innovative organic (ORG+). Bars represent the standard errors.

The N lost by leaching under durum wheat was significantly (*p* < 0.001) influenced by the systems in the first two (rainy) years, with the INT showing higher values as compared to both ORG and ORG+. In the latter two (drought) years, the N leaching values were generally low and the di fferences among systems were not significant (Figure 2B).

Processing tomato yield was compromised by the above-mentioned hail event in 2013/14, and the damages to the plants were particularly severe in ORG (Figure 2C). During the following years, yield values observed in INT and ORG were not di fferent, while ORG+ always showed the lowest values. From 2014/15, the processing tomato yield showed a decreasing trend, particularly in ORG+, due to attacks of late blight disease (*Phytophthora infestans* (Mont.) de Bary) of increasing severity over time (Figure 2C).

In INT, where the soil was left bare during the autumn and winter seasons, N leaching in processing tomato decreased linearly across years as rainfall amount decreased (Figure 2D). Otherwise, irrespective of CC managemen<sup>t</sup> strategies, both ORG and ORG+ showed similar N leaching values, which were significantly lower as compared to those observed in INT (*p* < 0.001).

#### *3.4. N Loss and Extra-Farm N Input Per Yield Unit*

In 2013/14 and 2014/15, the yield to N loss ratio (Y/Nloss, kg kg−<sup>1</sup> N) of durum wheat was rather low (53 ± 3.8 kg kg−<sup>1</sup> N) and it was not a ffected by the systems (Figure 3A). In 2015/16 and 2016/17, Y/Nloss was statistically similar in INT and ORG, which were significantly higher than ORG+. As for wheat, the Y/Nloss values of processing tomato (Figure 3C) observed in 2013/14 and 2014/15 were low in all systems: in 2013/14, the INT showed halved values as compared to ORG and ORG+, while in 2014/15 the systems did not show any significant di fference (due to the high variability observed in INT). During 2015/16 and 2016/17, the Y/Nloss values observed in ORG were higher (*p* < 0.001) as compared to both the INT and ORG+, which were not significantly di fferent from each other (Figure 3C).

**Figure 3.** Yield per unit of N leached (Y/Nloss, kg kg−<sup>1</sup> N), yield per unit of extra-farm N (Y/Nloss, kg kg−<sup>1</sup> N) in durum wheat (top: ( **A**) and (**B**), respectively) and processing tomato (bottom: ( **C**) and ( **D**), respectively) in the three cropping systems: integrated (INT), traditional organic (ORG) and innovative organic (ORG+). Bars represent the standard errors.

The yield to extra-farm–N input ratio (Y/Nextra, kg kg−<sup>1</sup> N) of durum wheat was calculated just for INT and ORG systems, as the wheat in ORG+ did not receive any extra-farm N input (Figure 3B). Across the entire experimental period, the values observed in ORG were always significantly (*p* < 0.001) higher as compared to the INT. Concerning the Y/Nextra of processing tomato, in 2013/14 and 2014/15

the observed values in INT and ORG+ were not significantly di fferent, but they were both lower than in ORG. In 2015/16 and 2016/17, ORG and INT were similar and higher than ORG+ (Figure 3D).
