1. Introduction
No-till (NT) in winter cereals has been adopted over the last 30 years to reduce costs and increase the crop productivity in those areas where precipitation is a limiting factor [
1], which is the case in Mediterranean climates, particularly in the semi-arid Ebro basin in northeastern Spain. However, this modification of the soil management has also promoted a change in the weed flora, with the entrance of species that typically appear in field margins and edges [
2,
3]. Among these species, rigput brome (
Bromus diandrus Roth.) has become the most problematic weed [
4,
5], as it can reduce crop yield from 22% with moderate infestations (12 plants m
−2) up to 71% with severe ones (500 plants m
−2) [
6].
In order to control this noxious weed in NT fields of Mediterranean semi-arid areas, few chemical options are found. Some herbicides offer successful control over
B. diandrus, but most of them can only be applied in wheat. A combination of flufenacet plus diflufenican in PRE and iodosulfuron-methyl plus mesosulfuron-methyl in POST obtained 98% of control efficacy, while the application of iodosulfuron-methyl plus mesosulfuron-methyl alone and pyroxsulam plus florasulam 0.275 kg ha
−1 obtained 92% and 88% efficacy, respectively [
7]. The application of isoproturon plus beflubutamid at 2.5 L ha
−1, an herbicide that can be applied to barley, only obtained 5% efficacy in this same study. Despite the good control of some of these herbicides, the efficacies could be affected by the weed density, if weeds were unable to be adequately sprayed. In this sense, García et al. [
6] observed that the efficacies could be reduced down to 55% with very high densities (>1000 plants m
−2), in contrast to efficacies up to 100% with lower densities. Moreover, staggered emergence [
8] allows some
B. diandrus individuals to escape herbicide application.
B. diandrus control relies on acetolactate synthase (ALS) inhibitors, and consequently the threat of easily developing resistance to this herbicide family is present, as has occurred for
Lolium rigidum Gaudin [
9]. For this reason, integrated weed management (IWM) programs seem to be the best options. The objective in IWM is focused in the cropping system rather than in the crop itself, integrating crop management with direct weed control methods [
10]. Thus, the combination of different measures can guarantee adequate weed management [
11] and prevent the evolution of herbicide resistances.
Sowing delay (SD), crop rotation, crop density, sowing pattern, and stubble removal after harvest have been considered for IWM. In particular, SD has widely been studied for weed control. García et al. [
6] saw that a one-month delay, from mid-October to mid-November, could allow the reduction of
B. diandrus by up to 96% of the initial quantity due to pre-sowing glyphosate spraying, and up to 99% if SD is continued to early-December. García et al. [
12] also saw that the combination of SD with an effective herbicide application in wheat can reduce very high infestations from >500 plants m
−2 to 1 plant m
−2 after three seasons, while these were reduced to 60 plants m
−2 without SD. Crop rotation and diversification are also main approaches to successfully carry IWM strategies (European Union Directive 2009/128/CE). In this sense, crop rotation offers important alternatives that allow interrupting the life cycle of weeds, such as crop sowing date variation, harvest date variation, crop competition, and alternation of the herbicide site of action (SoA) [
13,
14]. Three-year crop rotation programs, alternating wheat, barley, and field pea, have been demonstrated to successfully reduce
B. diandrus infestations [
15]. Crop density has also been studied for the control of many weed species. The competitive effect of the crop itself is able to reduce the fecundity of the surviving weed individuals [
12,
16,
17,
18]. This allows the reduction of the seedbank recruitment, contributing to the decrease of infestation levels. Furthermore, some studies have demonstrated that the competitive effect of the crop is enhanced when the distance between crop seedlings is homogeneous, compared to when the crop is sown in rows [
16,
19,
20]. This way, the competitive effect of the crop is improved and the fitness of the weed diminished. In this aspect, the problem is the lack of specific machinery and the most similar pattern would be a random sowing. Finally, destroying or removing the stubble has also been proposed and effectively used for the control of herbicide-resistant
L. rigidum in Australia [
21,
22]. The seed rain of the most important weed species, including
B. diandrus, occurs at harvest [
23]. Seeds remain on the soil surface with the crop stubble. Thus, destroying or removing the stubble could also help remove the weed seeds and it could partially contribute to the reduction of the seedbank.
All of these methods have been useful for the control of several weeds, but studies of a combination of these methods over time are scarce. For this reason, the objective of this work was to study 12 three-year IWM programs, with different combinations of the abovementioned cultural methods and chemical methods, for the control of B. diandrus in NT.
2. Material and Methods
2.1. Site Description
The experiment was established in a commercial winter cereal field in the province of Lleida, in northeastern Spain. The field, located in Agramunt (41°46′31″ N; 01°04′02″ E, 360 m.a.s.l.), presented a high infestation of B. diandrus and was used for winter cereal production in NT for the last three seasons. Barley was sown the previous season, so there had not been any successful control and the infestation was assured for the season starting the experiment. The field had a 2% slope to the north and the soil structure was 30% sand, 52% clay, and 18% silt, with 2.3% organic matter and a pH of 8.5.
2.2. Integrated Weed Management Assessments
The B. diandrus management was carried out over three seasons, from 2014–2015 to 2016–2017, and the effect of 12 IWM programs on the abundance of the weed was evaluated. The experiment was set as split-plot randomized blocks with three replicates. Each management plot measured 15 m × 6 m to facilitate sowing, herbicide applications, and harvest, leaving a 10-m buffer alley surrounding the trial.
For each IWM program crop rotation, SD, crop density, random sowing, removal of the stubble, chisel plow, and herbicide rotation were combined to reduce the great brome population (
Table 1): (1) traditional cereal monocrop (TRAD-1), two seasons with wheat (
Triticum aestivum L.) and alternation to barley (
Hordeum vulgare L.), with chemical control the third season; (2) wheat (WHEAT), wheat monocrop with SoA rotation; (3) stubble removal (CER-SR), cereal monocrop with wheat-barley-wheat sequence, SD, chemical control, and removal of the stubble after harvest; (4) cultural cereal monocrop (CER-CM), cereal monocrop with wheat-barley-wheat sequence, with chemical control and a combination of cultural managements, SD, stubble removal-high density-random sowing; (5) random sowing (CER-RdS), cereal monocrop wheat-barley-barley, SD, chemical control and random sowing the third season; (6) high density (CER-HD) cereal monocrop wheat-wheat-barley, SD, herbicide SoA rotation and high crop density (250 kg ha
−1) for all three seasons; (7) field pea (
Pisum sativum L.) rotation with cereals (PEA-CER), crop rotation field pea-barley-wheat, with SD and SoA rotation; (8) rapeseed (
Brassica napus L.) rotation (RAPE), crop rotation canola-wheat-barley, with SD the second and third seasons and SoA rotation; (9) camelina (
Camelina sativa L.) rotation (CAME), crop rotation camelina-wheat-barley, with SD the second and third seasons and SoA rotation; (10) traditional cereal monocrop with sowing delay (TRAD-2), wheat-barley-wheat, with SD the second season and SoA rotation; (11) pea rotation with wheat (PEA-WHE), rotation pea-wheat-wheat, with SD and SoA rotation; (12) pea rotation with chisel plow (PEA-CH), rotation pea-wheat-barley with SD, with chisel plow, and without chemical control the first and third seasons.
Sowing was done with a 3-m wide NT disc drill, regulating the sowing depth in the case of the chisel plow program (PEA-CH).
In all IWM programs, a pre-seeding glyphosate application was done, except in PEA-CH (12). In some plots, the
B. diandrus suppression promoted the growth of corn poppy (
Papaver rhoeas L.). For this reason a POST herbicide was applied to control this weed.
Table 2 compiles the chemical products, as well as their characteristics and application doses.
2.3. Estimation of Eliminated Population Due to Sowing Date
As there is an existing hydrothermal time (HTT)-based emergence model developed by García et al. [
8] (1), the proportion of the population that emerged and was then eliminated previous to the sowing date was estimated. This approach employing the HTT emergence model has been successfully applied in previous studies [
6,
24]. For this estimation, HTT had to be calculated as proposed by Spokas and Forcella [
25], using the STM
2 program.
where
y is the percentage of emergence and
x is the cumulated HTT on a certain date.
The weather data (maximum and minimum temperatures and precipitation) were taken from a meteorological station located in Tornabous (Lleida), 7.5 km away from the field (ruralcat.cat).
The chemical management of the programs, applied herbicides and coadjuvants (if necessary), and timings are specified in
Table 3. In all three seasons, Bromoxynil (23.8%) + 2,4-D (23.8%) had to be applied to TRAD-1 for the control of
P. rhoeas. This same herbicide was also applied in CER-CM in season 2016–2017 for the same reason. In season 2015–2016, glyphosate in pre-seeding was not needed as a severe drought prevented weeds from emerging.
2.4. Data Collection
The density of B. diandrus was counted in 10 randomly thrown quadrats at pre-sowing, before herbicide application and 30 and 60 days after application (DAA). Due to the reduction in the weed density in the second and third seasons, and because it was proved that no statistical difference was obtained with a higher number of density samples, the number of quadrats was reduced to five per plot in 2015–2016 and 2016–2017. At flowering, an estimation of the density of panicles was visually performed in two 6-m2 transects in the middle of the plots, with a total of 12 m2 assessed per plot, and transformed to a mean of four panicles/plant accordingly to random observations of 10 plants, in order to estimate B. diandrus density. Harvest of the field was conducted on 18 June 2015, 15 July 2016, and 26 June 2017 with a micro-harvester (Wintersteiger classic plot combine micro-harvester). In 2015, the harvest date was too late for camelina, which should have been done by 20 May, and suffered between 15% and 45% yield loss due to ant predation. For this reason, corrections of the yield results were made with the corresponding yield loss. The estimation of the economic income was performed according to the prices of the crops each year in the agricultural cooperative of Agramunt, which were the following: June 2015, wheat 178 €/Tn, field pea 240 €/Tn, oilseed rape 314 €/Tn, camelina 314 €/Tn; June 2016, wheat 180 €/Tn, barley 156 €/Tn; June 2017, wheat 165 €/Tn, barley 158 €/Tn.
2.5. Statistical Analysis
Statistical analysis was performed with SPSS 15.0 (SPSS Inc., Chicago, IL, USA). Results of the final density (FD) of B. diandrus were each analyzed with a parametric one-way analysis of variance, with the management program being the unique factor, the FD of each season being variable, and—due to the great differences in the initial densities (ID) of the weeds between blocks—the ID being considered as a random effect. This analysis was performed for every season’s final density. Due to the lack of normality of the samples, a transformation of the data into log(x + 0.1) was conducted. The statistical analysis for the reduction of density from the first to the third season was performed with Kruskal-Wallis one-way analysis of variance on ranks due to the lack of normality of these data. For the analysis of the economic income of the yields, this was performed only for the overall income for the three seasons together; the three blocks were considered together and parametric one-way analysis of variance was applied. The yields were not compared season by season is because the aim of the study was to consider each of the three-year managements as a whole.
3. Results
The three growing seasons differed in terms of temperature and precipitation: 2014–2015 was the warmest season (
Table 4) but had quite a large range, with a difference of 18.5 °C between the coldest and warmest months. This contrast was much lower in 2015–2016, only 13.8 °C, and the growing season was the coldest, although it presented the warmest winter. The 2016–2017 growing season was in between the other two (mean 12 °C), but the contrast between the coldest and warmest months was the highest among the three seasons (19.8 °C). With respect to precipitation, 2014–2015 and 2015–2016 showed similar amounts of rain, but differed in the distribution throughout the seasons (
Table 4): 2014–2015 presented a wet autumn (124.5 mm September–November), while it had a dry spring (34.8 mm March–May); in contrast, the autumn of 2015–2016 was extremely dry (34.4 mm September–November), which prolonged into the next two months (December–January), while spring was quite humid (139.7 mm March–May). Finally, 2016–2017 was the wettest season (315 mm), with autumn being reasonably wet (83 mm) and spring being very wet (158 L/m
2). These patterns, mainly those of precipitation, affected the emergence of
B. diandrus and thus the efficacy of the management programs, as will be explained later on.
3.1. Management Programs
The initial density (ID) between blocks varied significantly, from 139 plants m−2 in the first to 812 plants m−2 in the second and 2105 plants m−2 in the third block. The ID variation in 2014–2015 between management programs is explained by the different sowing dates, which allowed for greater emergence in the later sown plots, and by the patchy distribution of the weed. The final densities achieved each season for each of the management programs revealed their effectiveness, which was excellent in most cases. Overall, in the three seasons B. diandrus was almost completely controlled (>99.9%) and no significant differences were found between management programs.
Differences for the control of
B. diandrus between the management programs were found in the first and the third seasons, but not in the second (
Table 5). A significant block effect on the effectiveness of the management programs was observed the first and the second seasons (<0.002;
Table 5), but was not observed the third season.
3.2. Estimation of Eliminated Population Due to Sowing Date
When the emergence characteristics—estimated with the emergence model from García et al. [
8]—were analyzed (
Figure 1) and related to the amount of rainfall and the distribution of each season (
Table 4), a significant variation on the emergence of
B. diandrus was observed. There was also a significant variation in the proportion of the population that was killed each season due to the sowing dates (red arrows). Autumn 2014 was very rainy and around 90% emergence was achieved on 28 November. In relation to the sowing dates that season, by 29 September (RAPE) only 0.18% of
B. diandrus had emerged; but by 29 October 55% of
B. diandrus had emerged and was killed with glyphosate; finally, by 23 January 99.3% had emerged and only 0.7% emerged afterwards. On the other hand, autumn 2015 was extremely dry, and according to the model the real emergence did not start until 12 February. In this season, 90% emergence was not achieved until 27 March.
3.3. Yield Results
Yield was affected by sowing delay and the climatic characteristics of each season (
Table 6). Except in the first season, TRAD-1 obtained lower yields despite being sown earlier than the other programs. Despite this, and due to the variance present between the results, the economic income that these yields represent over all three seasons do not differ between management programs (
Table 6).