On-farm evaluation of integrated weed management in no-till rainfed crops in semi-arid Morocco

Field studies were conducted from 2012-13 to 2014-15 to investigate the dynamics of germinable soil seedbank, density and community composition of weeds in crop rotations of barley (Hordeum vulgare L.) + pea (Pisum sativum L.) and bread wheat (Triticum aestivum L.) in Oued Zem, semi-arid Morocco. In September 2012, the initial seedbank in 6 fields was 2354 seeds m 2. When herbicide-free barley + pea forage mixture (cut for hay) was followed by bread wheat, seedbank reductions were 35% after the two years. When bread wheat was followed by herbicide-free barley + pea forage mixture, seedbank reductions were only 5% in two years. Prior to wheat harvest, weed densities were 82, 8, and 14 plants m 2 in April 2013, 2014 and 2015, respectively. Prior to haying herbicide-free barley + pea, weed densities were 76, 109 and 34 plants m 2 in April 2013, 2014 and 2015, respectively. Weeds identified in bread wheat fields were 49, 36 and 40 species in 2012-13, 2013-14 and 2014-15, respectively. Weeds associated with herbicide-free barley + pea mixture were 68, 51 and 36 species in 2012-13, 2013-14 and 2014-15, respectively. The seedbank prior to planting and weed density prior to harvesting were strongly influenced by the most recent crop. Integrated weed management combining glyphosate before no-till planting, post-emergence herbicide use in bread wheat, haying barley + pea mixture, within the crop rotation (barley + pea/bread wheat) reduced weed seedbank by up to 35%, species richness by up to 47%, and weed density prior to wheat harvest or forage haying by up to 83%. Such changes suggest that integrated weed management practices in no-till system must be continued for more than 3 growing seasons to drastically reduce weed seedbanks and weed densities.


INTRODUCTION
A no-till system has economic, ecological, environmental and social benefits. These include soil conservation, water use efficiency, nutrient cycling, time and fossil fuel saving, and less wear and tear on machinery (Kassam, 2014). Such a system offers Moroccan farmers, who depend primarily on rainfed cereal-livestock production systems, the opportunity to reduce soil degradation and production costs (Boughlala et al., 2011). The estimated area under no-tillage in 2015-16 was only 7000 ha despite the research on tillage reduction in semi-arid Morocco that has been conducted since 1980 (Mrabet, 1993(Mrabet, , 2000(Mrabet, , 2008El Gharras and Idrissi, 2006;El Gharras et al., 2010;Mrabet et al., 2012;Abail et al., 2013;Belmekki et al., 2014;Schwilch et al., 2015).
In Morocco, no-till practices are proposed as a substitute to conventional cropping systems based on tillage using discs and tine tools. However, in crop/livestock integrated production systems, weed weeds are an important challenge since most weeds are considered as free forage for livestock and the weedy fallow is widely practiced (El Brahli and Mrabet, 2000). Abandoning the plow induces a qualitative and quantitative change in the flora which is why no-tillage crop production relies primarily on herbicides, particularly glyphosate, for weed management (Aibar, 2006;Blackshaw et al., 2015;Loss et al., 2015).
Weed seedbank abundance and composition can be used as indicators of the success or failure of cropping systems and management practices (Buhler et al., 2001;Cardina et al., 2002;Rahali et al., 2010;Kleemann et al., 2016). Higher weed seedbank diversity has been reported in more diverse crop rotations (Sosnoskie et al., 2006). Murphy et al. (2006) observed that tillage and crop rotations impact weed seedbank diversity and density, with the highest diversity found in no-tillage systems with a three-year crop rotation and diversity decreasing as tillage intensity increases.
Integrated weed management systems have the potential to provide long-term management of weeds (Blackshaw et al., 2005;Holm et al., 2006;Harker and O'Donovan, 2013;Blackshaw et al., 2015). Agronomic factors such as crop diversification and rotations, combined with herbicides, need to be evaluated for their potential to manage weeds in no-till system in semi-arid Morocco. Therefore, field studies were initiated to determine the combined effects of pre-plant glyphosate, herbicide use in bread wheat, and herbicide-free barley + pea mixture for hay on the weed seedbank, and weed density within a bread wheat/barley + pea mixture rotation under a no-till production system in a semi-arid rain-fed environment.

Study location and agronomic practices
A field study was conducted in Oued Zem (semi-arid region of central Moroccco). Soils were classified as sandy-loam. Annual precipitation varied from 290 mm in 2013-14 to 400 and 480 mm in 2012-13 and 2014-15, with more than 60% occurring from November to April (Table 1). Minimum temperatures were 5-10°C in winter (December to February) and maximum temperatures 25- Six farms were selected in September 2012 and 4 ha on each farm were used to conduct no-till trials. 2 ha were no-till planted to bread wheat (Triticum aestivum L.) and two ha were no-till planted to a forage mixture of barley (Hordeum vulgare L.) + field pea (Pisum sativum L.). The wheat/forage mixture rotation was managed using conservation management practices. Glyphosate at 720 g ha -1 was applied prior to no-till planting in November 2012 and 2014. Before 2012-13, all fields had been planted with cereals (wheat and barley) using a conventional tillage system in rotation with the traditional grazed weedy fallow for several years, and no herbicide applications was practiced. During the rainy seasons, the weedy fallow was cut, faded and hayed by the end of March or mid April as forage.
Bread wheat (cv. "Arrehane") and local barley and pea varieties were planted in November 2012, 2013 and 2014. At each farm, two hectares of land were planted with bread wheat and two hectares for barley + field pea, each growing season. The wheat crop was fertilized at 100 kg ha -1 of diammonium phosphate (18-46-0) and no-till planted at 120 kg ha -1 of seed. The forage mixture was fertilized the same way as the wheat and no-till planted at 25 + 75 kg ha -1 of barley + pea seed, respectively. Nitrogen fertilizer was applied at the rate of 33 kg ha -1 to wheat at tillering. The experiment was done using a tine seed drill of 18 cm row spacing with little chains behind to improve seed covering and no press wheels.
The broadleaf herbicide mixture of florasulam + 2,4-D (3.75 + 180 g ha -1 ) was used at tillering for post-emergent weed control in the wheat. A foliar fungicide (epoxyconazole at 125 g ha -1 ) was sprayed on wheat at the booting stage. The herbicide and fungicide were sprayed with a tractor-mounted sprayer, and volumes of spray were 120 L ha -1 for herbicides and 200 L ha -1 for the fungicide. In November 2012 and 2014, glyphosate (720 g ha -1 ) was applied in all fields prior to no-till planting wheat and barley + pea mixture. In November 2013, no vegetation was present at planting and there was no need to use glyphosate prior to planting. In all cropping seasons, no post-emergent herbicide was used in the barley + pea forage treatments.

Weed seedbank assessment
Soil seedbank was determined in September 2012, 2013, and 2014. Soil sampling was done before the first autumn rain. Each year, ten soil cores (7.2 cm diameter by 10 cm deep) were taken every 10 m along diagonal transects of each field of 2 ha, bulked, air dried, placed in polyethylene bags, and stored in the greenhouse at room temperature (25/10°C) until November. Sampling in the fall has been shown to provide reliable estimates of the viable weed seedbank because it allows natural dormancybreaking mechanisms to operate over the summer (Baskin and Baskin, 2014). Germinable weed seedbank determinations were conducted using the greenhouse emergence method (Smith and Gross, 2006;Gulden et al., 2011;Leon et al., 2015;Kleemann et al., 2016). Soil samples from each field were used to determine weed seedbank by placing soil samples on plastic trays that had been partially filled with potting mix. They were placed in trays in a 2-cm-thick layer, placed in a greenhouse with a day/night temperature of 25/10°C, and watered as necessary to keep moist. Weed emergence counts were made twice a month from November to April. Seedlings that emerged were identified and removed. Those that could not be identified were counted, removed, and indicated as "others". Total seedling counts in each soil sample, from November to April, were considered the germinable fraction of the weed seedbank, and this was converted to seeds per square meter for each field.

Weed density assessment
Weed densities in bread wheat strips were estimated in 10 randomly chosen (0.5 m x 0.5 m = 0.25 m 2 ) quadrats at each site 3 times during the year: 1. In early November prior to application of glyphosate (720 g ha -1 ) at planting time, 2. In mid-February prior to application of florasulam + 2,4-D and 3. In mid-April prior to wheat harvest in May.
In barley + pea mixture fields, weed densities were estimated in 10 randomly chosen (0.5 x 0.5 m = 0.25 m 2 ) quadrats at each site at the same times as reported above.
Weeds were identified using various floras. Scientific names used in this paper are those recommended for the North Africa plants by Dobignard and Chatelain (2010-13).

Data analysis
Since this study was conducted in six farmers' fields in the same region, each field was considered as one replicate. Mean seedbank density values per field were calculated as an average of the six sampled fields. After crop emergence, mean weed densities per field and the standard error of the mean were calculated as an average of the six sampled fields (John and Quenouille, 1977).

Weed seedbank dynamics
In September 2012, the average germinable soil weed seedbank from six farmers' fields was 2354 seeds m -² in the surface layer of 0 to 10 cm (Table 2). Such high seed densities may be due to the lack of weed management in the traditional cereals/weedy fallow rotation. In fact, most farmers were not familiar with herbicide use. Traditionally, weeds in cereal fields are hand pulled for animal feed. In September 2013, the crop rotation that started with bread wheat which was planted in November 2012 and harvested in May 2013 had 1720 seeds m -², while crop rotation starting with the forage mixture barley + pea that was planted in November 2012 and hayed in April 2013 had 2013 seeds/m² (Table 2). In comparison with the initial seedbank which was estimated in September 2012 to be 2354 seeds m -², the reductions in the seedbank in the first growing season (2012-13) were 27% after bread wheat, and 14% after barley + pea, respectively.
In September 2014, the weed seedbank in bread wheat planted in November 2013 and harvested in May 2014 after the barley + pea mixture (2012-13) was 1520 seeds m -², while the forage mixture barley + pea, drilled in November 2013 and hayed in April 2014 after bread wheat (2012-13) was 2238 seeds m -². As compared to the initial seedbank (2354 seeds m -²) estimated in September 2012, seedbank reductions in two years (from September 2012 to September 2014) were 35% in the rotation of barley + pea (2012-13) with bread wheat  and only 5% in the rotation bread wheat (2012-13)/hayed barley + pea . Higher seedbank reductions in crop sequences were essentially due to glyphosate pre-plant and florasulam + 2,4-D (3.75 + 180 g ha -1 ) post-emergence in bread wheat. In a 3-year crop rotation study, Kleemann et al. (2016) found that the use of oaten hay in year 1, followed by effective weed control in field pea and wheat crops, depleted the high initial seedbank (4820 seeds m -²) of rigid ryegrass (Lolium rigidum) to moderate levels (<200 seeds m -2 ) within 3 years. Smith and Gross (2006) found that seedbanks were strongly influenced by the most recent crop.
More than 30 weed species were present in the soil samples ( Table 2). Eleven of these species accounted for 45 to 80% of the seedbank. From the total seedbank, 8 to 43% were from seeds of the endemic annual broadleaf, Diplotaxis assurgens. In two years, seed declined by 76% in the bread wheat/barley + pea rotation and 84% in the barley + pea/bread wheat rotation. In a 4-year cropping rotation of oaten hay/field pea/wheat/barley, rigid ryegrass (Lolium rigidum) seedbank declined by 86% after oaten hay in year 1 (Kleemann et al., 2016). In a seed persistence study, Lutman et al. (2002) found that seed loss rates of 16 weed species over 6 years varied from 8 to 58%.

Weed density
Before applying glyphosate and planting wheat, weed densities were 442, 68 and 261 plants m -² in November 2012, 2013 and 2014, respectively (Table 3). The weed density observed in the fields in November 2013 (68 seedlings m -²) was due to low rainfall (32.6 mm in the whole month, Table 1) and low weed emergence, therefore no glyphosate was applied at planting. Glyphosate was sprayed in November 2012 and 2014, before or at planting at 720 g ha -1 , with excellent control of volunteer crops and annual weeds. Perennial weeds   (Table 3). In two years, reductions in weed density were therefore 83% as compared to the initial situation (2012-13). Such low weed densities found in April, prior to wheat harvest in May, produced a small quantity of seed that shattered onto the soil and contributed little to germinable seedbank. In fact, herbicide use was found to be the main factor reducing arable seedbanks, because it limits both weed growth and weed seed production (Chauhan et al., 2006;Jose Maria and Sans, 2011;Kleemann et al., 2016).
Before applying glyphosate and planting the herbicide-free forage crops, weed densities were 432, 43, and 226 plants m -² in November 2012, 2013 and 2014, respectively (Table 4). As mentioned for bread wheat, the low density of weed seedlings observed in the fields in Table 3. Weed density in 6 no-till bread wheat fields in Oued Zem, Morocco, from 2012-13 to 2014-15. November 2013 was due to low rainfall (32.6 mm in the whole month, Table 1) and low weed emergence; thus, no glyphosate was sprayed. Glyphosate, used in November 2012 and 2014 before or at planting at 720 g ha -1 , provided excellent control of volunteer crops and annual weeds.
In the fields of hay forage, weed densities in herbicidefree barley + pea were 76, 109, and 34 plants m -² in April 2013, 2014 and 2015, respectively (Table 4). The high density observed in April 2014 (109 plants m -²) was due to glyphosate not used pre-planting. After three growing seasons, weed management and crop rotation (barley + pea/bread wheat/barley + pea) reduced weed density by 55%, as measured in April 2015, as compared to April 2013. Most of the weeds at haying had some mature seed. Therefore, a proportion of mature seed shattered on the soil onto the soil, which enabled weeds to re-infest bread wheat the next year. Hill et al. (2016) found that viable seed production was reduced by 64 to 100% when weeds were mowed with immature seeds present as compared to when plants were mowed with mature seeds present.

Weed species richness
In this 3-year study, a total of 68 weed species in wheat fields: 49 in 2012-13, 36 in 2013-14, and 40 species in 2014-15 were identified (Table 5). That is, weed management reduced weed species richness by 27% in one year and 18% in 2 years. The five major weed species were Diplotaxis assurgens, Medicago polymorpha, Scorpiurus muricatus, Bunium fontanesii and Calendula stellata.
The total number of weed species associated with barley + pea mixture during 3 growing seasons was 79 species: 68, 51 and 36 species in 2012-13, 2013-14 and 2014-15, respectively (Table 5). Species richness was reduced by 25% the first year and 47%, the second years. The 5 major weed species were similar to those found in bread wheat fields.

Conclusion
On-farm measurements from 2012-13 to 2014-15 showed   that integrated weed management combining glyphosate before no-till planting, post-emergent herbicide use in bread wheat, haying barley + pea mixture and crop rotation (barley + pea/bread wheat) reduced the weed seedbank by up to 35%, species richness by up to 47%, and weed density prior to harvest by up to 83%. Such changes suggest that integrated weed management practices in no-till system must be continued for more than three growing seasons to greatly reduce weed seedbanks and weed densities. In order to achieve good weed management as well as an integration crop/ livestock production system, grazed weedy fallows should be replaced by forage crops for hay in the form of barley + pea forage mixtures. It is concluded that weed seedbanks and weed densities can be managed effectively when a multi-year weed management program is appropriately implemented into a no-till system.

CONFLICT OF INTERESTS
The authors have not declared any conflict of interest.