Delayed sowing improved barley yield in a no-till rainfed Mediterranean

ABSTRACT

barley cropping seasons. Similarly, a greater thousand kernel weight and higher WUE y 27 was observed when sowing was delayed. Averaged across years, WEC presented a 28 greater yield and above-ground biomass for D2 and D3 compared to D1, while for 29 WMC there were no grain yield differences seen between the sowing dates, above-30 ground biomass or yield components. Our results demonstrate that, in Western 31 Mediterranean areas, sowing delay under no-till (NT) conditions can increase grain 32 yield, WUE and NUE of winter barley, and also of wheat but only during wet years. NT, no-till; NUE, nitrogen use efficiency; TKW, thousand kernel weight, WUE b , water-36 use efficiency for above-ground biomass; WUE y , water-use efficiency for yield.

38
Core ideas 39 • Sowing delay and cultivar effects on cereal production and water and N use 40 efficiencies were studied.

41
• Sowing delay increased grain yield due to greater number of grains per m2. 42 while reducing the impact of climatic stresses during the grain-filling period as much as 94 possible. To this end, selecting an adequate sowing date and maturity class appear to be 95 key management practices. Moreover, NT bears traffic load better and leads to lower 96 work intensity (Bueno et al., 2006;Soane et al., 2012;Wolf et al., 1989), widening the 97 window of feasible sowing dates to wetter soil conditions. 98 However, interannual rainfall variability, characteristic of the Mediterranean 99 climate, complicates the selection of an optimum sowing date (Mahdi et al., 1998). For October 15 th and 20 th , this treatment being considered a reference as it is typical of the 140 farming regime in the area. D2 was sown between November 5 th and 10 th , and the D3 141 treatment was sown between November 25 th and December 5 th . In the first period (the 142 2006-2007, 2007-2008 and 2008-2009 seasons), barley was grown, comparing two 143 maturity classes: Hispanic (barley early maturity class, BEC) and Sunrise (barley 144 medium maturity class, BMC). In the second period (the 2009-2010, 2010-2011 and (wheat medium maturity class, WMC) and Artur Nick (wheat early maturity class, 147 WEC). These medium and early maturity classes correspond to facultative and spring 148 cultivars, respectively. The experiment was completely randomized in three blocks; 149 individual plot was 6 m wide x 48 m long. Air temperature and rainfall were recorded 150 hourly using an automated weather station located in the experimental area.

151
Crop management practices 152 The experiment was managed under NT, with the use of a 3 m-wide no-till drill 153 with disk openers. Three to five days before sowing, the weeds were controlled by 154 applying 1.5 L ha -1 of glyphosate (N-(phosphonomethyl)glycine). The sowing rate was 155 450 seeds m -2 in rows spaced 17 cm apart for the two crops studied. In the first two Roth.) was carried out with mesosulfuron-methyl plus iodosulfuron-methyl-sodium (15 162 + 3 g a.i. ha -1 ) on 5 March. In 2010-11, a post-emergence control of broadleaf and grass 163 weeds was accomplished with tribenuron-methyl plus metsulfuron-methyl (10 + 5 g a.i. 164 ha -1 ) on 30 March. Mesosulfuron-methyl plus iodosulfuron-methyl-sodium (15 + 3 g a.i. 165 ha -1 ) was applied on 9 February and on 13 April in D1 and D2 and D3, respectively. In 166 2011-2012 herbicide applications aimed at reducing ripgut brome levels and control 167 broadleaf weeds. Tribenuron-methyl plus metsulfuron-methyl (10 + 5 g a.i. ha -1 ) was 168 applied 20 February while mesosulforon-methyl plus iodosulfuron-methyl-sodium (15 + bars. This rate was decided upon according to the potential grain yield of the site (i.e, ≈ 174 2.8 Mg ha -1 ), and the annual N mineralization was estimated to be 30 kg N ha -1 for NT 175 (Angás et al., 2006). Time of application was chosen to minimize N volatilization 176 losses. Traditionally, farmers of the region applied greater N rates than the one used in 177 our experiment and carried out pre-sowing applications. However, more than two  Calculation of water-and nitrogen-use efficiency 211 Water use (WU) during the period between sowing and harvest was calculated as 212 the difference between soil water content (0-90 cm soil depth) at the beginning of 213 October and at the harvest of each treatment plus the amount of rainfall received during 214 that period. As in previous works in the same area, water loss as runoff and deep 215 drainage was considered negligible due to the negligible slope (< 2%) and the severely 216 water-limited conditions (Cantero-Martínez et al., 2007;McAneney and Arrúe, 1993).

217
The above-ground biomass and grain yield at 10% moisture were divided by WU to quantify the agronomic water-use efficiency for above-ground biomass (WUE b ) and 219 water-use efficiency for grain yield (WUE y ), respectively. WUE calculations were 220 based on soil water content in mid-October (right before sowing D1 treatment). This

231
Nitrogen use efficiency was calculated as the ratio of grain yield to N supply. N 232 supply was the sum of soil mineral N at sowing (0-90 cm depth), N applied as fertilizer 233 (i.e., 50 kg N ha -1 ), and mineralized N. This latter was estimated to be 30 kg N ha -1 234 according to the results obtained by Angás et al. (2006) under similar NT conditions. performed for each crop using a general linear model. Differences between treatments 243 were taken to be significant at the 0.05 probability level using a LSD test. Linear 244 relationships between yield components and grain yield were tested using the same 245 software. The slopes of the regressions were tested for differences between sowing 246 dates.   (Table 1). As an average of the two maturity classes 275 studied, D2 and D3 showed greater barley grain yields than D1 in the three cropping 276 seasons studied (Fig. 2). The greatest grain yield of BEC was observed for D2, while for 277 BMC both D2 and D3 presented greater yields than D1 (Table 1).

278
The number of spikes m -2 was significantly affected by sowing date, maturity 279 class and year main effects but not interactions (Table 1). D1 and D2 showed a greater 280 number of spikes m -2 than D3 as an average of maturity classes and cropping seasons.   2006-2007 and 2008-2009, as an average of maturity class (Fig. 2). When 307 distinguishing between maturity classes, the delay of sowing date (D2 and D3 compared 308 to D1) also significantly increased barley NUE (Table 1).  (Table 2).

319
The three wheat yield components studied were significantly affected by the 320 interaction between sowing date and year (Table 2). In 2009-2010 the delay of sowing led to greater number of spikes m 2 and grains per spike, but had no effect on TKW (Fig.   322 3). In contrast, in the 2010-2011 and 2011-2012 seasons lower TKW was observed 323 when sowing was delayed, while in 2010-2011 the delay of sowing led to a lower 324 number of grains per spike (Fig. 3). The wheat HI was significantly affected by the 325 interaction between sowing date and maturity class, and the interaction between sowing 326 date and year (Table 2). However, a delay in sowing produced no consistent trend in 327 this variable. cropping season (Fig. 3). Moreover, greater NUE was observed in WMC than in WEC 338 as an average of cropping seasons (Table 2).

339
The later barley sowings (D2 and D3) showed a significant linear relationship 340 between grain yield and the number of spikes m 2 , no significantly different between 341 them at P < 0.05. Contrarily, no relationship was found in D1 (Fig. 4a) (P = 0.76). As a 342 difference, the three barley sowing dates (D1, D2 and D3) showed the same (P < 0.05) 343 linear relationship between the number of grains per spike and grain yield (Fig. 4b). No 344 relationship was found between TKW and barley grain yield (P = 0.17). In the case of 345 wheat, grain yield was linearly related to the number of spikes m 2 and to the number of grains per spike, with no differences between sowing dates according to the analysis of 347 covariance performed (Fig. 4d, 4e). In contrast, wheat TKW showed a non-significantly 348 different linear relationship with grain yield between D2 and D3, while no relationship 349 was found for D1 at P < 0.05 (Fig. 4f). and 94 kg N ha -1 for the 2006-2007, 2007-2008 and 2008-2009 seasons, as an average of 446 the treatments. The decreased soil N availability resulted from the lower quantities of 447 mineral N rate applied during the experiment (i.e. 50 kg N ha -1 ) compared with the rate applied by the farmer (double or more in some cases). In our experiment that rate was 449 established in order to achieve a soil status that was less susceptible to N losses to the 450 environment.

451
Wheat maturity class choice played a major role in WUE y and NUE. The shorter 452 cycle of WEC than WMC could have reduced the susceptibility to terminal drought, 453 increasing WUE y and NUE.      Table 1 Analysis of variance of barley grain yield, above-ground biomass, spikes m -2 , grains spike -1 , thousand kernel weight (TKW), harvest 616 index (HI), water-use efficiency for above-ground biomass (WUE b ) and grain yield (WUE y ), and nitrogen use efficiency (NUE) as affected by 617 sowing date (D1, October; D2, November, and D3, December), maturity class (BEC and BMC, barley early and medium maturity class, 618 respectively) and year, and their interactions.