**4. Discussion**

The N balance (Table 1) allows the comparison between the three systems as it was computed by averaging the variations related to the conditions in individual years and crops [30]. The adoption of TIC in ORG+ considerably improved the N self-su fficiency of this system [31], raising (+35%) the Ndfa as compared to ORG. As expected, the Noutput values confirmed the gap between organic and conventional farming systems [32]; the Noutput values observed in ORG+ were the lowest in accordance with the finding of Knapp and van der Heijden [33] and Cooper et al. [34]. Nsurplus was high in both INT and ORG+, but such e ffects resulted from di fferent reasons: in INT, there was a very high Ninput and a high Noutput, while in ORG+ there was high Ninput coupled with low Noutput. In contrast, Nsurplus in ORG was the lowest. The relation between N leaching and N surplus was consistent at the crop rotation level only in the INT, confirming that the managemen<sup>t</sup> strategies to retain N in the system (e.g., by using catch crops and organic N fertilizers) are of paramount importance for reducing N leaching risk [16,35].

Wheat production in INT was the most variable across years, and such variability was associated with high N loss from the system when the climatic conditions were favorable to N leaching (high rainfall after side-dress fertilization and/or slow crop growth as recorded in the former two years, Figure 1). TIC was proved to reduce N leaching loss as compared to traditional managemen<sup>t</sup> in organic wheat production [19,22]. Non-inversion soil managemen<sup>t</sup> is known for reducing nutrient loss [36], thus its combination with TIC was expected to significantly improve the N retention in ORG+. However, such reduction (as compared to ORG) was not observed in our experiment, probably because the yield (and Noutput) was remarkably reduced in ORG+.

From 2014/15, a general reduction of processing tomato yield has been observed in all systems, probably due the short biannual rotation [37]. This e ffect was particularly evident in ORG+, where the short rotation problems were attributed to the typical decrease in crop productivity, during the transition from inversion to non-inversion soil till managemen<sup>t</sup> [38]. In our case, the termination e fficiency towards the CC, that is already known to be one of the critical issues in MBNT systems [8], was further reduced by the no-till soil management. However, processing tomato confirmed its good adaptability to organic practice [39] as the yields obtained in ORG were not statistically di fferent from those observed in INT in 3 out of 4 years. As observed by other studies [32], when N-inputs are similar (such as for processing tomato), the yield gap between organic and conventional systems is lower (12% ± 5.0%) than when N-inputs di ffer (such as in durum wheat, 21% ± 4.8%). The introduction of winter-sown CC in crop rotation was confirmed to be a very e ffective practice to prevent N leaching [17,40]; the N loss in INT was proportionate to the rainfall amount during the rainy season (October to March, *R*<sup>2</sup> = 0.854, *n* = 8), while in ORG and ORG+ it was constantly low confirming the essential importance of CC for building agricultural systems with high N self-su fficiency and internal N recycling [41].

Recently, the yield gap in conventional and organic systems has been intensely debated [32,42]. In accordance with Wilbois and Schmidt [43], it is important to reframe this debate by taking into account the appropriate benchmarks. Thus, a comparison of the output in the three cropping systems cannot exclude the extra-farm–N input transformation e fficiency (Y/NExtra) and the environmental cost (in terms of N lost from the system) per unit of yield (Y/NLoss).

In both wheat and processing tomato, ORG showed the best balance between economic output (yield) and water protection service as it showed the highest values of Y/NLoss in five cases out of eight (and in the three remaining cases the di fference among systems was not significant). When favorable conditions for N leaching loss were present, the Y/Nloss observed in durum wheat was not a ffected by systems, while concerning processing tomato, the e ffect of the cover crop on N leaching reduction was predominant [44]. Therefore, although the INT showed the highest yield, the Y/Nloss ratio was not di fferent from the very low yielding ORG+; on the other hand, ORG was the most interesting system in terms of environmental impact of the yield unit [45].

This interesting finding is confirmed by the Y/Nextra ratio (Figure 3B,D): considering durum wheat, the ORG e fficiency in converting an extra-farm–N source to yield was largely above that showed by INT, while ORG+ was not considered, as it did not receive any external N input for wheat production. The e fficiency of ORG+ was tested only in processing tomato, where it showed the lowest Y/Nextra values, due to the very low yield achieved by such system. Thus, at the cropping system level (i.e., considering both cash crops), this fact downplayed the impact of the complete N–self-su fficiency of ORG+ for durum wheat production. Thanks to its high yield, ORG showed Y/Nextra values that were not di fferent (2013/14, 2015/16 and 2016/17) or even higher (2014/15) than those observed in INT.
