Evaluation of Herbicide Efficacy, Injury, and Yield in White Lupin ( Lupinus albus L.)

White lupin is of increasing interest in the southeastern United States (US) as a winter le‐ gume cover crop or as mid-winter forage for ruminants. White lupins are poor weed competitors during early establishment, making effective weed control necessary; howev‐ er, only three herbicides are currently registered for use in lupin. An experiment was con‐ ducted at two Alabama sites in 2007 and 2008 to evaluate herbicide efficacy provided by ten preemergence (PRE) and nine postemergence (POST) herbicides as well as lupin in‐ jury and yield. Overall, PRE applied herbicides, particularly imazethapyr, linuron, and flumioxazin, caused less crop injury than POST herbicides while providing ≥ 86% control of annual bluegrass, corn spurry, heartwing sorrel, henbit, and lesser swinecress six weeks after application. Grass-active herbicides, fluazifop and sethoxydim, provided greater than 95% of annual bluegrass control without causing unrecoverable lupin dam‐ age. Imazethapyr applied POST controlled shepherd’s purse (96% to 98%), cutleaf eve‐ ning-primrose (81% to 96%), and wild radish (71% to 99%) without lupin injury. POST- directed spray applications of glyphosate and flumioxazin provided good weed control of corn spurry (80% to 98%) and winter vetch (71% to 95%) but caused significant crop injury due to drift. In general, grain yields were only reduced with the use of chlorimur‐ on, diclosulam, glyphosate, and thifensulfuron. This research suggests there are several herbicides not currently registered that could be beneficial for use in US lupin produc‐ tion.


Introduction
Conventional agriculture depends on synthetic nitrogen (N) fertilizers and herbicides for high crop performance [1]. Alternative N sources are available in the form of leguminous crops such as Lupinus spp. White lupin is of major interest in the southeastern US because new cultivars exhibit differential vernalization requirements similar to wheat (Triticum aestivum L.) and can be utilized as mid-winter forage. White lupin has been utilized in the southeastern US as a livestock feed, for human consumption and as a winter cover crop in conservation agriculture [2,3]. Since its introduction in the 1930s, until the 1950s the US lupin production reached over 1 million ha; however, production declined with the loss of government support, cold-weather damage to seed nurseries, and the increased availability of inorganic fertilizers [3][4][5].
Lupinus spp. are poor weed competitors during early establishment since canopy development is slow, facilitating light penetration and subsequent weed seed germination and yield loss due to competition. Lupin reaches maximum vegetative growth during flowering when it can successfully compete with newly emerging weeds [6]. Effective weed control is necessary to ensure lupin success under competition with weed species for water, nutrients, and light [6,7].
Previous research has been conducted to compare the effectiveness of herbicides on weed control and potential for crop injury in lupin. A successful preemergence (PRE) herbicide treatment resulting in no crop damage is pendimethalin alone, or in combination with metribuzin [8,9]. Pendimethalin use in white lupin provided 100% control of Russian thistle (Salsola tragus L.) and prostrate knotweed (Polygonum aviculare L.) [10]. The use of PRE applied metolachlor and alachlor, primarily in mixes with other herbicides, successfully controlled annual grasses and some broadleaf weed species greater than 90% in spring-type white lupin [11,12]. Additionally, metolachlor, alone or mixed with linuron, did not cause white lupin injury [13].
Knott [8] found that lupin are especially sensitive to postemergent (POST) herbicides. Fluazifop, as a POST application, provided ≥98% control of wheat (Triticum aestivum L.), triticale (x Triticosecale Wittm ex A. Camus), and annual ryegrass (Lolium multiflorum Lam.) without causing injury to the lupin crop [8,14]. POST application of imazethapyr provided good weed control but resulted in 15% to 24% crop injury and yield reduction [13]. Similarly, Penner et al. [12] found that the use of imazethapyr, as either PRE or POST, caused crop damage of 35% to 60%. Hashem et al. [15] showed that interrow weed control in narrow-leaf lupin provided by paraquat plus diquat increased yields compared to glyphosate alone, glyphosate plus metrabuzin, and glyphosate followed by paraquat plus diquat.
Currently, only three herbicides are registered for use in lupin: S-metolachlor, carfentrazoneethyl, and glyphosate [16]. Therefore, the objective of this experiment is to investigate the use of chemical weed management practices in white lupin and evaluate their effect on weed control, crop injury, and lupin grain yield.

Materials and methods
Experimental treatment and design. A two-year experiment was established at two test sites on the E.V. Smith Research and Extension Center of the Alabama Agricultural Experiment Station in October 2007 and 2008, respectively. The experiment was a 2 (year) x 2 (location) x 3 (cultivar) x 4 (block) x 24 (weed control) factorial treatment arrangement. The experiment design was a randomized complete block design (r = 4) nested within each year x location x cultivar combination. The weed control factor had 20 levels: one nontreated control, ten PREapplied herbicides, and nine POST-applied herbicides ( Table 1). The two locations of the experiment were the Field Crops Unit (FCU), near Shorter, AL (32.42 N, 85.88 W) and the Plant Breeding Unit (PBU), Tallassee, AL (32.49 N, 85.89 W). At FCU, the experiment was established on a Compass soil; a coarse-loamy, siliceous, subactive, thermic Plinthic Paleudults with a loamy sand surface structure. At PBU, the experiment was conducted on a Compass Soil: a fine-loamy, mixed, semiactive, thermic Typic Hapludults with a sandy loam surface structure. The three cultivars used in the experiment were AU Homer (a high-alkaloid, indeterminate cover crop type), AU Alpha (a low-alkaloid, indeterminate forage type), and ABL 1082 (a lowalkaloid, determinate grain type experimental cultivar). Statistical analysis. We used generalized linear mixed models procedures as implemented in SAS 4 PROC GLIMMIX to analyze weed control data. This tool is flexible in the analysis of data with nonnormal distribution and unbalanced designs. Violations of normality and homogeneity of variance issues are often encountered when including a nontreated control treatment or percent control data with a large range. Weed control data were modelled using a binary distribution function or arcsine transformed data. Crop injury data were modelled using arcsine transformed data and then analyzed with a normal distribution function. All treatment factors and their interactions were considered fixed effects except the block factor and its interaction with the various treatment factors. Statistical significance was declared at Dunnett's P < 0.1.

Results and discussion
Weed control. Over the course of the two-year study, 14 weed species were observed. Not all species were present in all environments; therefore, weed control is presented for only those species that appear at both sites in each year of the study. At the first rating after planting, in both years, the following PRE herbicides provided greater than 90% control of all rated weed species when compared to the nontreated included: S-metolachlor 5 /linuron 6 mixture, metri-buzin 7 , diclosulam 8 , flumioxazin 9 , and imazethapyr 10 ( Table 2). Linuron and S-metolachlor alone provided greater than 90% control in most instances except for henbit (Lamium amplexicaule L.), which was controlled by linuron at 86%, as well as lesser swinecress [Coronopus didymus (L.) Sm.] and heartwing sorrel (Rumex hastatulus Baldw.), which were controlled by S-metolachlor at 86% and 88%, respectively ( Table 2). The mixture of S-metolachlor/linuron has been used previously in lupin study, even though linuron is not labeled for use in white lupin production in the southeastern US [17,18]. In this study, at both early weed and late weed ratings, this mixture provided greater than 70% control of all rated weed species. Pendimethalin 11 provided good early season control of all weed species at the 0.5X, 1X, and 2X rate with the exception of lesser swinecress and heartwing sorrel, which were controlled less than 50% by the 0.5X and 1X rates. In 2007, the second weed rating (22 weeks after planting) conducted after POST herbicide applications revealed PRE applied herbicide weed control to be greater than nontreated controls at both FCU and PBU for each rated weed species except for cutleaf evening-primrose (Oenothera laciniata Hill). At PBU, pendimethalin (0.5X rate) provided only 14% weed control and at FCU, cutleaf evening-primrose control was only 23% with the 1X rate of pendimethalin (Table 3). Less than 50% control was achieved for this weed species with the 2X rate of pendimethalin as well as S-metolachlor. The following POST applied herbicides provided greater than 50% control of all rated weed species included: fluazifop 12 , chlorimuron 13 , and imazethapyr. With the exception of black medic (Medicago lupulina L.) and crimson clover (Trifolium incarnatum L.), which were controlled by less than 70% (data not shown), imazethapyr controlled all broadleaf weed species by more than 80%. Ivany and McCully [13] evaluated various herbicides for use in sweet white lupin, they also showed that imazethapyr applied PRE and POST provided good broadleaf weed control (80% to 91%). Sethoxydim 14 provided good control for all weed species except for cutleaf evening-primrose, which was less than 50% at both sites. The grass weed species, annual bluegrass (Poa annua L.), was successfully controlled by the POST-applied grass active herbicides sethoxydim and fluazifop which is in agreement with previous research evaluating grass control in lupin [14,19]. Thifensulfuron 15 did not provide greater weed control than the nontreated for cutleaf evening-primrose at FCU (15%) and provided less than 50% control of this species at PBU (31%) as well as corn spurry (Spergula arvensis L.) at FCU (43%) ( Table 3). Corn spurry control was also less than 50% for fomesafen 16 at both FCU (22%) and PBU (37%) and 2,4-DB at FCU (39%). Glyphosate 17 and flumioxazin, which were both POST-directed spray applications, provided good weed control of all rated weeds at both locations (Table 3).  and cutleaf evening-primrose. At PBU, pendimethalin at the 0.5X rate and full rate did not provide better control of shepherd's purse than the nontreated (Table 4). Similar results were seen at both PBU and FCU for all rates of pendimethalin in cutleaf evening-primrose control with 10% to 12% control with the 0.5X rate, 6% to 23% control with the 1X rate, and 8% to 18% control with the 2X rate. Control of corn spurry at both locations was also lacking for several POST herbicides including: fluazifop (6% to 7%), fomesafen (14% to 19%), 2,4-DB (5% to 6%), and sethoxydim (7% to 45%) (Table 4). Fluazifop, 2,4-DB, and sethoxydim also did not increase control of shepherd's purse compared to the nontreated at PBU with 32%, 21%, and 36% control, respectively. The clove/cinnamon oil mixture achieved poor control of shepherd's purse (29%) and cutleaf evening-primrose (14%) at PBU.
Crop injury. Two-way interactions (herbicide-cultivar and location-herbicide) were significant; therefore, injury ratings are presented by location and cultivar. Injury ratings presented here as the mean crop injury was taken after the POST herbicide applications. PRE-applied herbicides in 2007 resulted in no significant increases in crop injury in comparison to nontreated, with a few exceptions. Metribuzin caused increased white lupin injury (4.45) at FCU in cultivar AU Alpha; pendimethalin at the 2X rate resulted in increased injury (3.95) at FCU for the same cultivar ( (9.43 to 10.00), regardless of cultivar or location; therefore, they were discontinued in 2008 (Table 5). Thifensulfuron and chlorimuron were initially included in this study since they are registered for use in soybean; however, research has shown variable phytotoxicity among soybean cultivars for both herbicides [21,22]. Research conducted by Knott [8] suggests that sulfonylurea herbicides such as metsulfuron cause variable crop injury in white lupin, ranging from limited to severe when applied at the normal field rate. Flumioxazin, as a POST-directed spray, caused significant crop injury at each location for each cultivar (4.50 to 7.84). Significant injury resulted from the use of fomesafen at FCU regardless of cultivar; however, increased injury was not observed with this herbicide at PBU. Glyphosate also resulted in increased lupin injury at FCU for ABL 1082 (6.30) and AU Alpha (5.89) ( Table 5). Glyphosate is registered for POST-directed application in lupin in the US [16]; however, herbicide drift can easily cause significant crop injury. This was the most likely cause of lupin injury in our study. Injury from POST flumioxazin applications may also be attributed to drift since PRE applications of this herbicide did not result in increased crop injury in most cases; although, in drier soil conditions, increased phytotoxicity of flumioxazin has been observed in other crops. This could pose a risk for increased lupin damage [23]. Fluazifop (0.50 to 3.81), 2,4-DB (0.06 to 0.75), sethoxydim (0.26 to 2.28), and imazethapyr (0.94 to 4.45) did not result in increased lupin injury over the nontreated (Table 5).
Crop injury in 2008 resulted in less overall lupin injury than in 2007. PRE applied herbicides did not cause significant injury in comparison to a nontreated at either location for any of the cultivars except for diclosulam (5.26 to 9.00), which caused unacceptable injury, regardless of location or cultivar (Table 6). Diclosulam, which is applied either preplant incorporated (PPI) or PRE, is registered in soybean [Glycine max (L.) Merr.] and peanut (Arachis hypogaea L.) with little injury to either crop [24,25]. Lupin injury from PRE applications of diclosulam was significant for each cultivar included in the experiment. POST-applied herbicides did not increase crop injury over nontreated except for glyphosate (4.49 to 7.76) and fomesafen in AU Alpha at both locations (3.22 to 3.48) and in AU Homer at PBU (4.00) ( Table 6). Crop injury from fomesafen was noted in both years of the study with inconsistent injury for each cultivar.
In other crops, such as soybean and dry beans, previous research has reported negligible fomesafen injury regardless of cultivar [26,27]. In this study, however, it is evident that fomesafen can produce significant injury to lupin.
Grain yield. Mean grain yields (kg ha-1) were much higher for all three cultivars in 2008 as compared to 2007 ( Table 7). The grain type cultivar ABL 1082 yielded highest of the three cultivars in both years. The interaction of treatment and cultivar was statistically significant.  (Table 7). Two POST-applied herbicides, thifensulfuron and chlorimuron, had no measurable yields in 2007. In 2008, glyphosate was the only POST-applied herbicide that caused significant yield losses of 1700 kg ha -1 .    (Table 7). In 2007, none of the PRE-and POST-applied herbicides reduced yield. However, the POST herbicides, thifensulfuron and chlorimuron, yielded 218 kg ha -1 and 0 kg ha -1 , respectively. In 2008, diclosulam, with a mean grain yield of 210 kg ha -1 , was the only PRE herbicide that reduced mean grain yield of this cultivar. Similarly, glyphosate, with a mean grain yield 735 kg ha -1 , was the only POST herbicide that caused significant yield reduction in 2008.
AU Homer. The nontreated control had a mean grain yield of 555 kg ha -1 in 2007 and 1219 kg ha -1 in 2008 ( Table 7). None of the PRE and POST herbicide treatments significantly reduced or increased yield as compared to the control in 2007. In 2008, none of the PRE or POST herbicide applications, with the exception of PRE diclosulam (548 kg ha -1 ), yielded lower than the nontreated control.
Experiments conducted by Payne et al. [4] in the Pacific Northwest showed a maximum white lupin yield of 2128 kg ha -1 , but this yield is not stable. In our study, yield within each cultivar varied greatly between years depending on the treatment. The grain-type cultivar ABL 1082 had the highest mean grain yield, followed by the forage-type cultivar AU Alpha and the covercrop-type cultivar AU Homer. In this experiment, diclosulam, thifensulfuron, chlorimuron, and glyphosate caused major grain yield losses. AU Homer appears to be the least sensitive to herbicide-induced yield reductions, since neither thifensulfuron nor chlorimuron reduced grain yield. Ivany and McCully [13] stated that POST applications of imazethapyr caused severe crop injury and yield loss in sweet white lupin. The results of this study did not confirm their findings. Neither the PRE nor the POST imazethpyr applications caused significant crop injury or subsequent yield reduction. This could be due, in part, to the use of different cultivars than those used by Ivany and McCully [13].
In general, PRE herbicide applications included in this study, excluding diclosulam, could be used in lupin without posing a significant risk of crop injury. Previous observations by Dittman [28] agree with findings that PRE herbicides may cause less lupin injury than POST herbicide options. Certain POST herbicides, such as thifensulfuron, chlorimuron, and fomesafen, are not viable herbicide options for use in lupin. Other POST options, like fluazifop, 2,4-DB, sethoxydim, and imazethapyr, may offer additional options for weed control in lupin without increasing crop injury.
The results of this experiment show that good weed control can be achieved by using a broad spectrum of herbicides that are currently not registered for use in US lupin production such as imazethapyr, flumioxazin, and linuron. With glyphosate and S-metolachlor, which are registered for use in lupin in the US, good weed control in lupin is possible; however, the use of a limited number of active ingredients can potentially increase resistance development in weed species in these systems. Based on these results, it is necessary to expand the number of registered herbicides for use in US lupin production.