Experimental plots were established in a field trial in Agramunt (41°48’N and 1°07’E) (Lleida, Spain). The field is located in the Ebro Valley, 330 m above sea level, and has a semi-arid mixed continental-Mediterranean climate (Papadakis, 1966). The soil was a Fluventic Xerocrept (100-120 cm deep) (SSS, 1994), with 30.1% sand, 51.9% silt and 17.9% clay, 2.3% organic matter and a pH of 8.5. Rainfall at the site during the study period and the long-term average are presented in Table 1.

Trials were conducted over three consecutive growing seasons (2008/09, 2009/10 and 2010/11) in an experimental cereal field under no-tillage. The site had been under no-tillage production for two seasons prior to the initiation of the study and had a natural infestation of B. diandrus. The experiment was arranged as a randomized complete block design with three replications. Each plot was 6 × 30 m and the main factor considered was the emergence time of B. diandrus according to three crop sowing dates (D1, D2 and D3). Respective dates in 2008, 2009 and 2010 were as follows: (D1) 20, 19, and 14 October; (D2) 7, 12, and 18 November; and (D3) 10, 3, and 13 December. Barley cv. ‘Hispanic’ was sown in 2008 and wheat cv. ‘Bokaro’ in 2009 and 2010. Cereals were sown at 180 kg seed/ha (400-450 plants/m²). Crop sowing was performed with a no-till disc drill in rows 19 cm apart.

Doses and dates of herbicides applied in each growing season are detailed in Table 2. Mesosulfuron-methyl plus iodosulfuron-methyl sodium was applied in the 2-5 leaf stage of wheat in 2009/10 and in tillering for D1 and in 2-5 leaf stage for D2 and D3 of wheat in 2010/11. In all plots fertilizer was applied each year from February to March at 48 kg N/ha.

The seedling populations recruited according to the three different crop sowing dates, and the emergence times (ETs) are therefore identified as F1, F2 and F3.

Each season densities of B. diandrus were estimated in each plot 60 days after herbicide application within 0.1 m² quadrats placed at ten locations randomized by blind tossing. Each year in June, when the weeds reached their complete development and fecundity, 20 plants from F1, F2 and F3 were collected and the following biological attributes were estimated: number of stems per plant, number of spikelets per plant and stem, fecundity (caryopses per plant) and number of caryopses per spikelet. Aerial vegetative biomass (VB) and reproductive biomass (RB) per plant were also estimated. For this purpose, plants were cut at ground level, placed in a tray, separated into VB and RB and oven-dried at 65ºC for 24 hours. Reproductive effort (RE) was calculated per plant as RE = RB/VB. At maturity, panicles were collected from several plants from each ET and threshed, and 1000 caryopses were weighed when completely dry. Seed rain was calculated by multiplying final density (last density sampling before the crop harvest) by fecundity.

Before crop harvest 20 panicles were collected from each ET and replication. Caryopses were sampled from both apical (A) and basal (B) spikelet positions on panicles, as well as both apical (a) and basal (b) caryopsis position within spikelets. Thus, the four sampling positions were Aa, Ab, Ba, and Bb. The weight of 20 caryopses situated in the same position for each ET was estimated every season. Weights are expressed per caryopsis.

ANOVA was performed using SAS (SAS Institute Inc., Cary, NC, USA) to detect differences between plants with different emergence times for all the variables analysed. Because different cereals were sown and different weed managements were applied each year, the results were analysed separately for each growing season.

In cases in which ANOVA analyses were significant, LSD post-hoc tests at p=0.05 were done. Before analysis, all characteristics were transformed to satisfy the homogeneity of variance assumptions [log (x+1)], while transformed. Linear regression analyses were performed between biological characteristics (VB vs. RB and fecundity) with the Sigma Plot program 11.0, using log (x+1) transformations.

The densities estimated 60 days after herbicide treatment differed across years and decreased significantly as the ET advanced in each season (Table 3). In 2008/09 density decreased by 71.8% between F1 and F2, by 88.0% between F1 and F3 and by 57.6% between F2 and F3. In 2009/10 density decreased by 96.2% between F1 and F2, by 99.3% between F1 and F3 and by 81.8% between F2 and F3. In 2010/11 density decreased by 95.4% between F1 and F2 and by 98.3% between F1 and F3, with no significant differences between F2 and F3. This consistent decreasing gradient of B. diandrus density as a function of crop sowing date is due to the greater emergence that takes place after autumn rains and the resulting reduction of the soil seed bank. These emerged seedlings are susceptible to pre-sowing control methods (Riba, 1993; Kleemann & Gill, 2006). Therefore, delayed crop sowing reduces the density of weeds competing with the crop.

In 2008/09, no post-emergence effect against B. diandrus was observed with isoproturon plus diflufenican in barley (33.2%, 11.4% and 16.1% for F1, F2 and F3, respectively), and the decrease in weed density observed in F2 and F3 (>71%) can be attributed exclusively to the crop sowing delay. In 2009/10 and 2010/11, after the herbicide treatment, the decreasing density observed in F2 and F3 (>95% in both seasons) reflects the effect of the sowing delay and the efficiency of the chemical control by mesosulfuron-methyl plus iodosulfuron-methyl sodium. These results demonstrate that in direct drilling the control of this species with only a pre-seeding application of glyphosate is limited, but the weed density can be reduced by combining delayed crop sowing with a post-emergence application of an appropriate herbicide (García et al., 2014). After three years a practical depletion of the B. diandrus population was achieved in F2 and F3.

In general, a decreasing gradient of caryopsis weight was observed according to seasons (Table 6). Considering all positions and the three ETs together, the mean caryopsis weight was 15.90, 11.02 and 10.64 mg, respectively, for 2008/09, 2009/10 and 2010/11. An ET effect was only observed in 2008/09 for all the positions considered, with greater values for F1 than for F2 and F3. In 2008/09 and 2009/10 a position effect was observed within each ET (except in F1 in 2009/10), while in 2010/11 no significant differences were found between the four positions considered. In 2008/09 similar patterns of allocation were observed in the three ETs. The heaviest caryopses were those from the bottom of spikelets situated in apical position in the panicle (Ab), followed by those in the lowest position in spikelets situated in basal position in the panicle (Bb), whereas the weight of caryopses in apical position in both apical (Aa) and basal position (Ba) of the spikelet in the panicle was lower than that of caryopses in position Ab, in most cases with a significant difference.

Without a specific herbicide application (season 2008/09), a significant gradient of resource allocation as a function of caryopsis position in the panicle was observed for each ET. Caryopses from apical spikelets were heavier, regardless of their position within them. This could be interpreted as a result of the different moment when the spikelets started to develop and the inflorescence was still inside the plant: spikelet production follows a downward direction, starting the development from the highest (apical) positions of the inflorescence, when the vegetative apex is transformed into the reproductive apex, as has been defined for several grasses (Pujol, 1998). For other grasses, González-Rabanal et al. (1994) and Recasens et al. (2007) confirmed that these differences in seed position in the panicle resulted in further differences in seed dormancy. During development, seeds in different positions may experience different temperatures and/or water content, and differences in resource partitioning, which could influence their dormancy status. This further assessment in B. diandrus could provide new knowledge on how the position of caryopses on the panicle promotes changes in dormancy regardless of environmental factors.

Furthermore, during the first season a decreasing gradient of caryopsis weight was observed from the basal to apical positions in the spikelet. These differences depend on the sequential process of floral development within the spikelet. In rice, these differences in grain weight within the spikelet are attributed to an intra-spikelet competition for assimilates pre- and post-anthesis (Calderini & Reynolds, 2000). The absence of these differences in B. diandrus when the specific herbicide was applied (2009/10 and 2010/11) could be explained by a disruption of the source-sink balance during the period of grain filling. The high number of spikelets without filled grains observed in plants that survived the herbicide (data not shown) is evidence of this process.

Delaying the crop sowing date reduces not only the density of B. diandrus and the weed pressure on the crop, but also the recruitment of new seeds to the seed bank. In the absence of a specific herbicide, plants that emerged late showed higher fecundity, a decrease in reproductive effort and a decrease in 1000-grain weight. However, their contribution to the final seed recruitment was lower than those of earlier-emerged plants, for which higher densities were established. Delaying crop sowing (until mid-November) should be considered as an efficient management tool for controlling this species in a cereal monocrop in no-tillage systems. The application of mesosulfuron-methyl plus iodosulfuron-methyl sodium when wheat sowing is delayed increases efficiency due to delayed plant phenology and reduced weed density, and also reduces the fitness of surviving plants.

One constraint of delaying the crop sowing date is the possible yield reduction. However, in our experiment the delayed sowing did not lead to a yield penalty: on the contrary, higher yields were obtained due to the avoidance of the competitive effect of the weed (data not shown). In all cases, a balance between crop yield and weed control should be established in order to maximize profits.

In our study, after three growing seasons the B. diandrus population was practically depleted in the field. Implementation of cultural and chemical strategies in no-tillage systems may improve integrated weed management programmes in dry-land arable fields where options for growing crops alternative to cereals are limited.