Selectivity and weed control efficacy of some herbicides applied to sprinkler irrigated rice ( Oryza sativa L . )

Sprinkler irrigation can reduce the irrigation water needed to grow rice. However, most available information on weed control with herbicides is related to flood irrigated rice because this is the main growing method. Field experiments were conducted at Zaragoza (Spain) during two years to study weed control and tolerance of sprinkler irrigated rice to several herbicides. The main weeds were Atriplex prostrata Bouchér ex DC., Cyperus rotundus L., Echinochloa crus-galli (L.) Beauv. and Sonchus oleraceus L. Rice cv Guadiamar was tolerant to preemergence (PRE) application of clomazone at 0.36 kg ha and oxadiazon at 0.5 kg ha. PRE application of pendimethalin at 1.32 kg ha combined with clomazone at 0.36 kg ha decreased rice yield. Postemergence (POST) application of bentazon at 1.6 kg ha + MCPA at 0.25 kg ha did not injure rice but POST application of azimsulfuron at 0.025 kg ha produced visual crop injury. Only treatments that controlled grassy weeds since rice was planted and by more than 80% at harvest time lead to acceptable rice yield (> 5,000 kg ha). Clomazone applied PRE at 0.36 kg ha provided good control of grassy weeds (> 80%) and the highest rice yield, so it is recommended as a selective and efficacious PRE treatment for weed control of annual weeds in sprinkler irrigated rice. The perennial purple nutsedge was difficult to control at high plant densities (> 150 plants m) and the recommended herbicide is azimsulfuron applied at POST at 0.02 kg ha. Additional key words: irrigation system; upland rice; weed competition.


Introduction
Weed control is one of main cultural changes when rice is grown under sprinkler irrigation instead of under flood irrigation (Akkari et al., 1986;Battilani and Pietrosi, 1990;Spanu et al., 1992).Weed control in rice under flood irrigation is based on the existence of a water layer that precludes the growth of non-aquatic weed species, but allows the growth of aquatic or semiaquatic weeds, which results in a selection of the last type of weeds.Moreover, when grown under flood irrigation rice is usually not rotated with other crops, which makes weed control more difficult and costly herbicide treatments are required (Hill et al., 1994;Flórez et al., 1999;Rao et al., 2007).
Given the relevance of the rice crop worldwide, many studies of weed-crop competition and weed control have been conducted (Rao et al., 2007).However, most available information on weed control with herbicides is related to flood irrigated rice because this is the main growing method.In Spain there is not any relevant area grown under sprinkler irrigation.Worldwide irrigation of rice by sprinkler systems is only relevant in Brasil (Pinheiro et al., 2006).Thus, there is limited information about the competition of weeds with rice and the selectivity and efficacy of herbicides when rice is grown under sprinkler irrigation (Akkari et al., 1986;Spanu et al., 1992).Some studies have been conducted on upland rice (Okafor and De Datta, 1976;Enyinnia, 1992;Esqueda, 2000), but this is mostly rainfed and not a fully irrigated crop.
In the absence of a standing water layer over the soil, no aquatic or semiaquatic weeds emerge (Spanu et al., 1992) and the weed species found in rice fields are the same as those appearing in other irrigated crops such as maize.Thus, strong competition occurs between rice and grass species such as barnyardgrass [Echinochloa crus-galli (L.) Beauv.] and crabgrass (Digitaria spp.) (Akkari et al., 1986;Spanu et al., 1992;Gitsopoulos and Froud-Williams, 2004), as well as with broadleaf weeds, such as common purslane (Portulaca oleracea L.), field bindweed (Convolvulus arvensis L.), burclovers (Medicago spp.), black nightshade (Solanum nigrum L.), common cocklebur (Xanthium strumarium L.) and pigweeds (Amaranthus spp.) (Akkari et al., 1986;Battilani and Pietrosi, 1990;Spanu et al., 1992).Several studies found that barnyardgrass competition with rice increased as the depth and duration of flooding decreased (Agostinetto et al., 2007;Rao et al., 2007).Akkari et al. (1986) reported that control of perennial weeds like purple nutsedge (Cyperus rotundus L.) and johnsongrass [Sorghum halepense (L.) Pers.] is one of the main challenges in sprinkler irrigated rice.Okafor and De Datta (1976) have pointed out the relevance of purple nutsedge as weed in upland rice.
Herbicide weed control can be modif ied by the sprinkler irrigation system because the frequent irrigation events (2 to 3 per week) can wash off foliar acting herbicides from the leaves and/or activate and leach residual soil herbicides (Spanu and Pruneddu, 1998).Changes in herbicide behavior from flood to sprinkler irrigation are expected to be greater for soil residual herbicides because no standing water is present.Thus, in flood irrigated rice leaching below the root zone is usually precluded because an impervious layer is formed by tillage.Besides, soil water content is lower in sprinkler irrigated rice which could affect soil residual activity of soil applied herbicides.
The objective of this work was to study the selectivity and efficacy of several herbicides under sprinkler irrigated rice to establish the most adequate weed control treatments for this system.

Material and methods
Field experiments were conducted in 2003 and 2004 at Zaragoza (Spain) (Lat 42°07' N, Long 1°8' W).The experiments were conducted in two different fields.The soil of the 2003 experiment was a silty clay loam (20% sand, 46% silt, 34% clay) with 1.5% organic matter and a pH of 8.5.The soil of the 2004 experiment was a sandy loam (26% sand, 45% silt, 28% clay) with 1.8% organic matter and a pH of 8.4.
Rice cv.Guadiamar was planted with a commercial cereal seed drill on 28 April 2003 and 27 April 2004 in rows 0.13 m apart at a planting rate of 230 kg ha -1 .Plots were fertilized before planting with 54 kg N ha -1 , 100 kg P 2 O 5 ha -1 and 100 kg K 2 O ha -1 .Top dressing N was applied at tillering (50 kg N ha -1 ) and at panicle initiation (50 kg N ha -1 ).Plots were irrigated with a sprinkler irrigation system with sprinklers separated at 15 × 18 m and located at 1.5 m above the soil.Sprinkler water application rate was 5 mm h -1 .After planting, 5 mm irrigation water was applied every two days to promote the emergence of the crop.Once the crop had emerged, crop water requirements were calculated following the FAO approach (Allen et al., 1998).Reference evapotranspiration was calculated with the FAO Penman-Monteith method (Allen et al., 1998) from meteorological data obtained from an automated weather station positioned over grass and located at the experimental farm.The considered crop coefficient during the growing season was 1.2.The crop irrigation requirement was calculated weekly as the difference between crop evapotranspiration and the effective precipitation, which was estimated as 75% of the precipitation (Dastane, 1978).Precipitation during the crop season was 171 mm in 2003 and 112 mm in 2004.In 2003 the irrigation applied was 750 mm in 72 irrigation events and in 2004 the irrigation applied was 885 mm in 83 irrigation events.
Eight herbicide treatments and a control without herbicide application were compared in each year (Table 1).Results from the first year were taken into account to choose the herbicide treatments to be tested on the second year.In addition, the POST treatments tested in the second year considered the prevailing high purple nutsedge densities.Preemergence treatments were applied within seven days from rice seeding.In both years, POST treatments were applied 7 weeks after planting (WAP) (POST 1) and 10 WAP (POST 2).Herbicide treatments were applied over the crop with a N 2pressurized backpack sprayer calibrated to deliver 300 L ha -1 at 200 kPa.The experimental design was a randomized complete block with 4 replications.The experimental unit area was 28 m 2 (4 × 7 m).
Tolerance of rice plants to herbicides was assessed by counting the number of emerged plants within a 0.28 m 2 grid placed randomly four times in every plot at 6 WAP, and by visually estimating crop injury at 6 WAP and at 11 WAP (2003) or at 9 and 11 WAP (2004).Crop injury was estimated through a visual integration of chlorosis, necrosis, stunting, stand reduction and tissue bleaching using a 0 to 100% scale where 0 = no injury and 100 = plant death.Weed control was evaluated at the same time than selectivity by counting the number of weeds by species within a 0.28 m 2 grid placed randomly four times in every plot.At rice harvest, weed control was assessed visually (0 to 100% scale; 0 = no weed control, 100 = complete weed control) for grass weeds, broadleaf weeds and purple nutsedge separately.A value between 70 and 80 is considered as partial weed control, values between 80 and 90 are considered as moderate control, and values higher than 90 are considered as total weed control.
The rice grain from the central part of each experimental plot was harvested in October (8.4 m 2 ) using a 1.20 m wide combine harvester.In those plots where weed density precluded mechanical harvest, the rice grain in two squares of 1 m 2 was harvested manually.Grain moisture content was determined a For each year, means within a column followed by the same letter are not statistically different according to Fisher's protected LSD test at p = 0.05.b Abbreviations: PRE, preemergence; POST, postemergence; WAP, weeks after planting.c Crop injury was recorded on a 0 (no injury) to 100 (plant death) visual scale after PRE (6 WAP), POST 1 (9 WAP) and POST 2 (11 WAP) treatments.d 14% moisture content.and the grain yield at 14% moisture content was calculated.
Data were subjected to ANOVA and means separated using Fisher's Protected LSD at p = 0.05.Inspection of residuals and scatterplots for the visual rating of crop injury and weed control suggested that assumptions of normality and homoscedasticity held reasonably well.Weed densities were transformed with ͌ʳʲ (x + ʲʲ 0.5) before ANOVA.Means of weed densities shown on Tables 2 and 3 are non-transformed.A stepwise regression analysis was performed to determine the relevance of injury and weed control in grain yield.

Rice tolerance
Rice emergence was not signif icantly affected by any herbicide treatment (Table 1).Pendimethalin at the highest rate reduced the rice stand by 25 to 30% in 2003 although, as stated, the difference was not sig-nificant with the rest of herbicide treated and control plots.
All treatments with clomazone applied PRE visually injured rice 6 WAP (Table 1).Injury symptoms were the typical bleaching of plants (Bollich et al., 2000).The phytotoxicity symptoms of clomazone decreased with time and no injury was found at harvest.Oxadiazon and pendimethalin applied PRE as single treatment did not visually injure rice 6 WAP (Table 1).When oxadiazon was mixed with clomazone rice injury did not increase.When pendimethalin was mixed with clomazone in PRE slight visual injury was found still at harvest in 2003 (Table 1).
Propanil applied alone POST did not injure rice (Table 1).However, POST application of propanil mixed with clomazone in 2003 injured rice by 37%, which suggests that clomazone was mostly responsible for the injury (Table 1).
At rice harvest no relevant visual injury was found (Table 1).Only the mixture of clomazone at 0.36 kg ha -1 and pendimethalin at 1.32 kg ha -1 showed low (10%), but significant injury.

Weed control
The main weeds found in both years were the grass weed barnyardgrass, the broadleaved annual sowthistle (Sonchus oleraceus L.) and triangle orach (Atriplex prostrata Bouchér ex DC.), and the perennial purple nutsedge (Tables 2 and 3).Purple nutsedge at our area is typical for the old irrigated areas of the Ebro valley but is quite rare in the recently-developed irrigated areas.Given that our experimental field is within the old irrigated areas, the purple nutsedge plants found in the plots emerged from tubers.
Only clomazone applied PRE controlled barnyardgrass in both seasons while oxadiazon applied PRE did not control this grassy weed (Tables 2 and 3).Pendimethalin applied PRE provided a 76% control of grass weeds at the highest rate (1.32 kg ha -1 ) only in one of the two years (Tables 2 and 3).When mixed with clomazone, grass weed control increased.No difference in annual weed control was found between 0.66 kg ha -1 and 1.32 kg ha -1 of pendimethalin.

Rice yield
Rice yield was affected by weed control and by herbicide injury (Table 4).Weeds strongly competed with rice under sprinkler irrigation so there was almost no rice yield when no herbicide was applied (Table 1).Consequently, those treatments that did not provide an adequate weed control produced lower yield due to weed competition.The highest yields were obtained when grass weeds were adequately controlled (Tables 1 to 3) and grass weed control was the main variable related with grain yield both years (Table 4).Crop injury decreased the rice yield only in 2003 (Table 4).
In 2003, the highest rice yield was obtained with the application in PRE of clomazone at 0.36 kg ha -1 alone or combined with oxadiazon at 0.25 kg ha -1 , which adequately controlled annual weeds (> 81% at harvest) (Tables 1 and 2).In these treatments the application of azimsulfuron at 0.025 kg ha -1 slightly improved purple nutsedge control compared with the use of bentazon at 1.6 kg ha -1 mixed with MCPA at 0.24 kg ha -1 .The application in PRE of pendimethalin at 1.32 kg ha -1 and propanil in POST at 3.5 kg ha -1 resulted in a yield of 5,030 kg ha -1 , which was not statistically different to the highest yield but almost 2,000 kg ha -1 lower (Table 1).Rice yield in the other treatments was lower than 3,000 kg ha -1 due to low grass control or crop injury (clomazone at 0.36 kg ha -1 + pendimethalin at 1.32 kg ha -1 applied in PRE).
In 2004, the highest yield was less than 6,000 kg ha -1 probably due to the competition of purple nutsedge which was not completely controlled in any case (Tables 1 and 3).All the herbicide treatments that adequately controlled grass weeds (> 80% at harvest) produced more than 5,000 kg ha -1 of rice.Thus, the highest yields were obtained when clomazone was applied in PRE alone at 0.36 kg ha -1 or mixed with oxadiazon at 0.25 kg ha -1 or pendimethalin at 0.66 kg ha -1 .When clomazone was applied mixed with oxadiazon, rice yield was not affected if the rate of clomazone was reduced to 0.27 kg ha -1 (Table 1).The other PRE herbicides lead to lower rice yields due to insufficient grass weed control or due to crop injury (clomazone at 0.36 kg ha -1 + pendimethalin at 1.32 kg ha -1 applied in PRE).If grass weeds were adequately controlled with PRE applied herbicides, it was suff icient to apply azimsulfuron early POST at 0.02 kg ha -1 to provide an adequate control of purple nutsedge.

Rice tolerance
The phytotoxicity symptoms of clomazone applied PRE decreased with time and no injury was found on plants at harvest, similarly to the results reported in other studies under flood irrigation (Webster et al., 1999;Scherder et al., 2004;Zhang et al., 2004Zhang et al., , 2005;;Mudge et al., 2005a) and in upland rice (Esqueda, 2000).However, some reports indicate that at high rates (> 0.8 kg ha -1 ) rice yield can be reduced (Bollich et al., 2000;Mudge et al., 2005a).The moderate visual injury found under sprinkler irrigation agrees with the results of Jordan et al. (1998a) who observed lower rice injury when clomazone was applied in dry seeded rice compared to rice seeded after flooding.Mudge et al. (2005b) reported that when clomazone is mixed with bensulfuron or halosulfuron plant bleaching decreased without decreasing clomazone weed control.Under flood irrigation, Zhang et al. (2004) found cultivar differences in the rice tolerance to clomazone.Thus, long-grain cultivars were less injured than mediumgrain cultivars, such as the one used in our experiment ('Guadiamar').
Although rice plants under flood irrigation have been reported to be injured by oxadiazon when applied PRE (Babiker, 1982) and POST (Smith and Khodayari, 1985), our results showed that this herbicide was selective when applied PRE in sprinkler irrigated rice either alone or mixed with clomazone.Rice injury by pendimethalin has been reported under flood irrigation (Smith and Khodayari, 1985;Street and Lanham, 1996), which agrees with the injury symptoms that were found in 2003 when it was applied mixed with clomazone.

Weed control
Only clomazone applied PRE controlled grasses under sprinkler irrigation.Esqueda (2000) also found good control of junglerice [Echinochloa colona (L.) Link] with clomazone in upland rice but at higher rates than those used in our study.In flood irrigated rice of Arkansas (Webster and Baldwin, 1998) and Louisiana (Jordan et al., 1998a) clomazone provides good control of grasses and is now considered the reference herbicide for annual grasses in Arkansas (Talbert and Burgos, 2007).
Although good broadleaved weed control with oxadiazon has been reported for sprinkler irrigated rice (Akkari et al., 1986), upland rice (Enyinnia, 1992), and flood irrigated rice (Smith and Khodayari, 1985;Prasad and Rafey, 1995), oxadiazon applied PRE did not control grasses in our study, but improved broadleaf weed control.
Pendimethalin applied PRE provided partial control of grass weeds at the highest rate only in one of the two years.Irregular weed control of pendimethalin has been reported due to soil texture (less rapid loss in clay soil) and to lower soil persistence if not soil incorporated (Savage and Jordan, 1980;Zimdahl et al., 1984).Jordan et al. (1998b) reported incomplete control of barnyardgrass in flood irrigated rice with pendimethalin at 1.1 kg ha -1 .
The control of the perennial weed purple nutsedge is one of the main cultural problems of rice under sprinkler irrigation (Akkari et al., 1986).This weed is not common under flood irrigation conditions.Therefore, information about purple nutsedge control with the herbicides used in flood irrigated rice is scarce.When the soil remains dry for some period of the growing season, as in upland rice and dry-seeded rice, purple nutsedge has been reported as a weed of rice (Bhargavi and Reddy, 1992).It was found that when purple nutsedge density was low it was partially controlled with bentazon at 1.6 kg ha -1 plus MCPA at 0.24 kg ha -1 , but completely controlled with azimsulfuron.However, only the application of azimsulfuron provided some control of high density purple nutsedge infestations in our experiments.

Rice yield
Weed control was found to be the main cultural challenge to grow rice under sprinkler irrigation because weed control failure can result in almost complete yield loss, as in our experiments where barnyardgrass competition was particularly strong.Akkari et al. (1986) and Battilani and Pietrosi (1990) reported similar results.Smith (1988) indicated that barnyardgrass is a strong competitor when rice is dry seeded and field flooding is delayed.Thus, grasses should be controlled from the earliest stages of rice development in a sprinkler irrigation system.
Although pendimethalin is the recommended herbicide in Italy for rice under sprinkler irrigation (Spanu and Murtas, 2002) and adequate weed control of grass weeds has been found under flood irrigation (Baltazar and Smith, 1994;Street and Lanham, 1996), our results show less selectivity for rice compared to clomazone and oxadiazon.Thus, it would be more interesting under our growing conditions to use it at low rates (0.66 kg ha -1 ) and mixed with clomazone to decrease rice injury and to increase the control of annual weeds.
The use of clomazone at 0.36 kg ha -1 is recommended as a selective and efficacious PRE treatment for weed control of annual weeds in sprinkler irrigated rice.The addition of oxadiazon at 0.25 kg ha -1 to clomazone at 0.27 to 0.36 kg ha -1 is also recommended due to the increased control of annual broadleaf weeds obtained.If only broadleaf weeds are present, the application of oxadiazon in PRE at 0.5 kg ha -1 would be an adequate choice.The application of azimsulfuron at 0.02 kg ha -1 early POST is recommended to control purple nutsedge in sprinkler irrigated rice.

Table 1 .
Herbicide tolerance and rice yields in 2003 and 2004 a,b

Table 2 .
Weed control by herbicides applied to rice in 2003 a,b

Table 3 .
Weed control by herbicides applied to rice in 2004a,b

Table 4 .
Regression analysis of rice yield as explained by weed control and crop injury a Control GRAM: control of grass weeds at harvest.