Effects of Conservative Tillage and Nitrogen Management on Weed Seed Bank after a Seven-Year Durum Wheat—Faba Bean Rotation

Conservative agriculture includes a range of management strategies with low energy inputs such as no-tillage, minimum tillage, and low application of fertilizers. Weed flora in arable fields is strictly affected by agronomic practices such as tillage and fertilization management. This study was conducted seven years after the beginning of a long-term—durum wheat–faba bean—rotation. It analyzes the combined effects on the soil seed bank of three different tillage systems (conservative, reduced, and conventional tillage) and two levels of nitrogen fertilization. The effects were investigated both using stepwise discriminant analysis and analysis of variance in order to find statistical differences among main factors and their interactions. The seed bank of Conyza canadensis, Papaver rhoeas, Solanum nigrum, Fallopia convolvulus, and Fumaria officinalis was higher in conservative or reduced tillage plots. The magnitude of the response to nitrogen supply varied among weed species. Conyza canadensis seemed to be favored by low nitrogen supply, whereas Sinapis arvensis by higher doses of nitrogen. Anagallis arvensis showed the lowest seed bank in conventionally tilled plots, without distinction of nitrogen supply. The results suggest that different tillage systems and, to a lesser extent, different nitrogen supply, produce changes in the seed bank size and composition, along the soil profile.


Introduction
Conservation agriculture has gained popularity in many agriculture systems all over the world. It is defined by FAO as an "approach to managing agro-ecosystems for improved and sustained productivity, increased profits and food security while preserving and enhancing the resource base and the environment" [1]. Conservation agriculture promotes, mainly, minimal soil disturbance and low application of chemical inputs such as fertilizers and pesticides, therefore reduction of energy inputs. Minimal soil disturbance plays a pivotal role because it reduces the consumption of fossil fuels, avoids soil erosion, maintains soil physical and biological health [2]. In the same way as for tillage, also, nitrogen applications need to be rationally managed because N affects environmental performance of agriculture practices by increasing the emissions of greenhouse gases [3] and the risk of groundwater contamination. Moreover, N coming from agricultural systems is biologically active in soil and water or chemically active in atmosphere; therefore, it can be detrimental for biological diversity, air, and water [4].
A different affinity of wild species to tillage results also in orchards, as shown by Fracchiolla et al. [22]; authors found that Poaceae species (Bronus sterilis, Avenas terilis and Hordeum murinum) are significantly associated with no-tillage systems such as the application of foliar herbicides or mowing.
Weed community and soil seed bank can be affected also by nitrogen content as a consequence of different amounts and types of fertilizers [23]. The response of each species can be very different as shown by many authors such as Blackshaw [24] and Sweeney [25]. Further, Pysek and Leps [26] suggest that weed community is influenced by both a direct effect of the fertilizer and indirect effects such as the increased competition with the crop. To this regard, time of fertilization supply, in relation to time of weed emergence, plays a crucial role.
Nitrogen can also play a role in regulating seed germination; for example, it can stimulate or inhibit germination of dormant seeds as shown by Goudey et al. [27] for Sinapis arvensis.
Moreover, nitrogen fertilizers and tillage systems can interact in modifying weed infestation as shown in barley (Hordeum vulgare) by O'Donovan et al. [28]; particularly, authors report that Setaria viridis was associated with low nitrogen rate and conventional tillage.
Therefore, the extensive literature review concludes that weed species richness, seed bank abundance, and diversity are dependent on disturbance levels of the soil and fertilization rate. The effects are also influenced by the interaction with other factors such as the environmental conditions and crop species [9]. The adoption of conservation tillage provides different environments for weed germination, emergence, growth, and competition by altering physical and chemical properties of soil [18]. Fertilizers, especially nitrogen, interact with the different tillage systems influencing weed germination, emergence, growth, and reproduction.
In investigating the effects of different agronomic managements, and their interactions, on weed density and species composition, many variables (weed species) have to be simultaneously taken into account and analyzed. To this aim, a crucial role in interpreting the results of field monitoring is played by the adoption of multivariate analysis techniques. Multivariate methods, and in particular discriminating techniques, allow to easily summarize the effects observed and extract the main behaviors in relation to the treatments compared [29]. In particular, stepwise discriminant analysis (SDA) is a feature selection technique which allows the identification of the variables enabling maximum discrimination among the treatments (i.e., classes) compared.
Knowledge of long-term effects on weed species and communities in agroecosystems will provide useful information to farmers about possible future scenarios that can be encountered as a consequence of changes in field management.
For all these reasons, this study evaluates, in a fava bean-durum wheat rotation started in 2008, the combined effects on the seed bank of different tillage systems and nitrogen fertilization levels.

Total Number of Species and Seeds in the Seed Bank
A total number of 21 species was found in the soil seed bank, belonging to 13 botanical families. Among all the species, 11 of them, in the experimental area, have the emergence phase typically in Spring or Summer i.e., between May and July, whereas 10 emerge in Autumn or Winter (October-January) ( Table 1).
Regarding total number of seeds m −2 , only the tillage systems caused significant differences ( Table 2) as a consequence of the intensity of soil disturbance; particularly, the number of seeds was the highest in the Conservative Tillage (SS), followed by Reduced Tillage (RT), and Conventional Tillage (CT). The total number of seeds of the species emerging in spring or summer showed the same trend. The total number of seeds emerging in autumn/winter was higher in SS than in CT plots; for these species, the number of seeds m −1 was higher also in the plots fertilized with 90 kg ha −1 of N.  Significant interactions were found within soil depth and tillage for total number of seeds and number of seeds of autumn/winter species. Particularly, both these data were the highest in the layer 0 to 0.20 m of the SS plots (data not shown).

Stepwise Discriminant Analysis
Stepwise Discriminant Analysis allowed a first and overall screening of the effects of the treatments compared on weed seed bank.
Specifically, in the analysis carried out on the upper soil layer (0-0.20 m), SDA identified Anagallis arvensis, Conyza canadensis, and Heliotropium europaeum as the weed species most able in discriminating both different soil tillage and N supply (T × N) ( Table 3 (a)).
By investigating in detail the effect of the three soil managements (T), together with Conyza canadensis, Papaver rhoeas, Solanum nigrum, and Fallopia convolvulus were selected as the species most affected by the treatments compared in the shallower soil layer (Table 3 (b)). Finally, a greater number of species, including also Sinapis arvensis, Fumaria officinalis, Chenopodium album, and Phalaris paradoxa, was identified as the most discriminating the different N supplies (Table 3 (c)).
When the analysis was carried out on the 0.20 to 0.40 m soil layer, Conyza canadensis for the T × N interaction and N, together with Papaver rhoeas and Phalaris paradoxa, respectively for T x N and T, confirmed the behavior observed in the shallower soil layer (0-0.20 m) showing to be among the most sensitive species with regard to soil and N fertilization management (Table 4 (a-c)).
In detail, Conyza canadensis, together with Papaver rhoeas, Portulaca oleracea, and Phalaris paradoxa, showed the highest discriminating capability in the T × N interaction; Papaver rhoeas, Phalaris paradoxa, and Fumaria officinalis showed to be most affected by the different soil tillage, Conyzacanadensis and Sonchusoleraceus by the different N supplies.
Finally, on the analysis performed on the whole layer investigated (0-0.40 m), SDA identified Conyza canadensis among the weed species most able in discriminating both the different soil managements (T) and the whole treatments studied (T × N interaction). Conyzacanadensis was indeed selected as the second and third species, respectively, with partial R-Squared values of 0.4781 (F = 14.65, p < 0.0001) and 0.7085 (F = 13.61, p < 0.0001).
Anagallis arvensis was also selected as first species for both the different soil managements (T) and the whole treatments studied (T × N interaction), with partial R-Squared values of 0.5667 (F = 21.58, p < 0.0001) and 0.9112 (F = 61.53, p < 0.0001), respectively. Table 5 shows the main results of Analysis of Variance performed for each species recorded in the soil samples, both for the main factors and for their interactions.

Analysis of Variance
As regard to effect of soil depth, number of seeds recorded in the 0 to 0.20 m layer was significantly higher for Fallopia convolvulus, Galium aparine, Papaver rhoeas, Polygonum aviculare, and lower only for Veronicahederifolia. For the other species, ANOVA did not show any significant difference.
Except for Polygonum aviculare and Veronica hederifolia, tillage showed to induce significant modifications of the seed bank of all the species considered. Number of seeds found in SS plots was the highest for Conyza canadensis, Fallopiaconvolvulus, Galiumaparine, Heliotropiumeuropaeum, Papaver hybridum, Portulaca oleracea, Setaria viridis, Solanum nigrum, and Silybum marianum. Number of seeds of Anagallis arvensis, Fumaria officinalis, and Papaver rhoeas was the highest in RT plots, whereas only Phalaris paradoxa showed the highest seed bank in the CT plots.
With regard to the effect of nitrogen, in the plots fertilized with 30 kg ha −1 of N, the number of seeds of Anagallis arvensis, Conyza canadensis, and Solanum nigrum was higher than in those fertilized with the greater dose. Opposite results were recorded for Papaver hybridum, Portulaca oleracea, and Sinapis arvensis whose seed bank was higher in the plots with 90 kg ha −1 of N.     (1) Only species whose data are significant for at least a factor are shown. For the main factors, data within a column followed by different letters are significantly different at the 0.05 p-level (Duncan's Test); (2) See Table 1 for weed codes.
Regarding interaction soil depth × tillage, both in the layer 0 to 0.20 m and 0.20 to 0.40 m, the number of seeds m −2 of Conyza canadensis in SS plots was higher than in CT plots. Seeds of Fallopia convolvulus and Galium aparine were found only in the SS plots at both layers of soils studied with a value significantly higher in the upper layer. Papaver rhoeas was found only in both layers of RT plots and in the layer 0 to 0.20 m of SS.
This interaction soil depth × nitrogen was significant only for Conyza canadensis, Papaver rhoeas, and Polygonumaviculare.
The interaction tillage × nitrogen confirmed some behaviors highlighted by SDA, in particular regarding Anagallis arvensis and Portulaca oleracea. Specifically, the lowest number of seeds of Anagallis arvensis was found in all CT plots and SS plots fertilized with 90 kg ha −1 of N. Seed bank of Papaver rhoeas was statistically higher in RT (at both fertilization doses) and in SS plots fertilized with 90 kg ha −1 of N than in CT plots at both fertilization levels. The number of seeds of Portulaca oleracea was higher in SS plots with 90 kg ha −1 of N than in CT plots with 30 kg ha −1 of N; in addition, seeds of this species were not found in the other plots. The highest number of seeds of Setaria viridis was found in the SS plots fertilized with 90 kg ha −1 of N. Seed bank of Solanum nigrum in SS plots with 30 kg ha −1 of N was higher than in the all the other plots.
As for soil depth × tillage × nitrogen, at both depths, number of seeds m −2 of Conyza canadensis was the highest in all the SS plots except for those fertilized with 90 kg ha −1 of N at the layer 0.20 to 0.40 m. Number of seeds of Galium aparine was the highest in the layer 0 to 0.20 m of SS plots fertilized with 90kg ha −1 .

Discussion
Among all the species found in the study area, the results show clear evidence that some species seed banksare particularly affected by tillage systems and the level of nitrogen supply.
In the layer 0 to 0.20 m, tillage systems were discriminated overall by Conyza canadensis, Papaver rhoeas, Solanumnigrum, and Fallopia convolvulus whereas in the layer 0.20 to 0.40 m by Phalaris paradoxa, Papaver rhoeas, and Fumaria officinalis. Except for Phalaris paradoxa whose seed bank was higher in the deeper layer of conventionally tilled plots (moldboard plowing), the seed bank of all the above mentioned species appears promoted by no-tillage or, at least, reduced tillage.
Conyza canadensis seemed to be favored by low cultural inputs in terms of soil disturbance and nitrogen supply. Similar results are reported by other studies particularly regarding the spreading of this weed caused by "no-reverse" tillage systems such as no-tillage [30,31] or chisel plowing [32].
The magnitude of the response to nitrogen supply varied among weed species, as shown also by Blackshaw et al. [24]. In the upper layer, six species were able in discriminating the two different doses of nitrogen whereas only two in the deeper layer. Among these species, Conyza canadensis appears promoted by low nitrogen supply whereas Sinapis arvensis by the higher doses of nitrogen. This last species has been already reported as benefiting from increased nitrogen availability [33].
Anagallis arvensis clearly discriminated, as also highlighted by the SDA results, the whole treatments studied, showing the lowest seed bank in the conventionally tilled plots without distinction of depth or nitrogen supply. The seed bank of Portulaca oleracea, a nitrophilous species [34], appears to be promoted, at all the layers studied, by no-tillage and high levels of nitrogen supply.
On average, the effect of tillage system on most of the species found in the experimental field and on overall weed community was more marked than that of nitrogen supply.
Findings of our research are consistent with those reported by many authors [10,16,18]: reduced or no-tillage caused an increasing infestation of many weed species because of low soil disturbance. Of the species emerging in autumn or winter, the effect of nitrogen supply is more evident; it makes sense to hypothesize that these species are growing at the time of fertilizer distribution and therefore benefit directly from the increased availability of nitrogen.

Materials and Methods
Data were collected in a field trial located in Southern Italy (Policoro-Basilicata Region) at the experimental farm "E. Pantanelli" (University of Bari-40 • 10 19.70" N; 16 • 39 5.05" E).
The site is 15 m above sea level and characterized by subarid climate according to De Martonne classification [35], with average annual rainfall of 560 mm distributed mainly during autumn and winter, and temperature typical of Mediterranean climate.
The soil, according to the Canadian System of Soil Classification [36], is clay-loam with high content of organic matter (Table 6). Since 2008, the field hosted a long-term-durum wheat-faba bean-rotation in a split-plot experimental design. Within the experiment field, an area was isolated and considered to investigate the effects of tillage system (Conservative tillage-SS; Reduced tillage-RT; Conventional tillage-CT) and nitrogen supply (low and high N rate) on weed seedbank size and composition (Table 7).
Specifically, the field had been divided into three strips (each of 30 × 80 m) managed with three different tillage systems (Table 7). Each strip had been subdivided into two plots (30 × 40 m) in which 43.8 or 131.4 kg of Urea (46%) (corresponding to 30 or 90 kg ha −1 of N) were applied when durum wheat was grown. Fertilizer was applied at the end of February (at the end of tillering phase) and spread uniformly all over the plot.
Further details about conduction of the trial can be found in a past paper [3]. Table 7. Tillage systems and related seedbed preparation and sowing systems used in the experimental field.

Tillage System Seedbed Preparation Sowing
Conservative Both in faba bean and durum wheat, weeds were chemically controlled applying the herbicides listed in Table 8.

Soil Sampling
Soil sampling was done in November 2015 (i.e., seven years after the beginning of the rotation) and before the preparation of the seedbed for the sowing of faba bean. Each of the six plots (three tillage systems × 2 N levels) was divided into three parts; in each of them, twenty soil cores of 0.40 m depth were randomly taken within a central area of 10 m × 20 m. A 2.3-cm diameter cylindrical steel probe was used, subdividing each core into two sub-cores of 0.20 m, corresponding to 0-0.2 m and 0.2-0.4 m depths. The twenty samples taken in each part were mixed to form a single sample that represented soil, randomly collected, and deriving from a surface of 8.306 × 10 −3 m 2 .

Seed Bank Determination
For the determination of the seed bank, the direct observation of the plantlets emerging from each soil sample was applied [32,37]. Some species, for which it was not possible to identify at the early stages, were transplanted in pots and allowed to grow to be identified at older stages.
Soil samples were placed into a tray where the soil did not exceed 4 cm depth and was periodically wetted. Trays were kept in the greenhouse for 20 months, under such conditions as not to exceed an internal temperature of 35 • C, during the warmest periods.
In order to favor dormancy breakage, irrigation was suspended periodically for 15 days and samples were stirred before irrigation were resumed; four months before the end of the experiment, samples were irrigated with a solution containing 1000 mg L −1 of KNO 3 to stimulate the germination of residual viable seeds. The plantlets emerging from each tray were identified, counted and removed. The data collected for each species were expressed in terms of number of seeds m −2 of soil [32,38].

Statistical Analysis
Before statistical analysis, all data were square root transformed, in order to increase the homogeneity of variances (assessed with Bartlett test) [32].
The effects of the different soil managements (SS, RT, CT) and N supplies (30, 90 kg N ha −1 ) on weed seedbank composition were first investigated using Stepwise Discriminant Analysis (SDA). In particular, to understand how different soil and N managements affected weed species presence, data analysis was carried out, for each soil layer (0-0.20 m and 0.20-0.40 m), considering both the different managements separately (Tillage and Nitrogen) and their interaction (Tillage × Nitrogen).
Given a classification variable and several quantitative variables, SDA selects a subset of the quantitative variables for use in discriminating among the classes (SAS Institute Inc., Cary, NC, USA, 2012). Wilks' lambda statistics was used as multivariate measure of separability [39][40][41] SDA was performed through the STEPDISC procedure of SAS/STAT using the STEPWISE algorithm (SAS Institute Inc. 2012). Significance level to entry and to stay was set at 0.05.
Analysis of variance was also performed, following a 3-way completely randomized experimental design, in order to find statistical differences among main factors (soil depths, tillage systems, and N levels) and their interactions. Duncan's test was used to compare means resulting from significant factors.

Conclusions
The effect of different tillage systems and nitrogen supply could not be promptly evident on the emerged flora because the seed bank, thanks to seed longevity, acts as a reservoir for weed infestation [32].
Our results, as a whole, suggest that different tillage systems and, to a lesser extent, different nitrogen supply, produce changes in the seed bank size and composition, along the soil profile. Therefore, the characterization of seed bank in terms of size and species composition can be a useful tool, more sensitive than actual flora, for predicting the medium-and long-term effects on weed flora of minimal soil disturbance and other conservative practices such as low application of inputs promoted by conservation agriculture systems.