Introduction
Understanding weed–crop interactions and the potential for crop loss from weeds allows growers to optimize weed management strategies. Growers need to adopt and apply economic thresholds to minimize yield loss from weeds (Coble and Mortensen Reference Coble and Mortensen1992; Cousins et al. Reference Cousins, Brain, O’Donnovan and O’Sullivan1987). Most published studies of weed and crop competition are for soybean [Glycine max (L.) Merr.], far exceeding studies of the same type for other crops (Barnes et al. Reference Barnes, Jhala, Knezevic, Sikkema and Lindquist2018; Bensch et al. Reference Bensch, Horak and Peterson2003; Song et al. Reference Song, Kim, Im, Lee, Lee and Kim2017; Zimdahl Reference Zimdahl2004). Many of these studies focus on season-long weed interference in soybean.
Palmer amaranth (Amaranthus palmeri S. Watson) and large crabgrass [Digitaria sanguinalis (L.) Scop.] are consistently ranked as two of the most troublesome and common weeds, respectively, in many crops (Van Wychen Reference Van Wychen2016). Amaranthus palmeri has become increasingly troublesome across the U.S. Southeast, Midsouth, and Midwest, with the possibility of further expansion out of these regions (Bagavathiannan and Norsworthy Reference Bagavathiannan and Norsworthy2016; Copeland et al. Reference Copeland, Giacomini, Tranel and Montgomery2018; Davis et al. Reference Davis, Schutte, Hager and Young2015; Kohrt et al. Reference Kohrt, Sprague, Nadakuduti and Douches2017; Korres et al. Reference Korres, Norsworthy and Mauromoustakos2019; Kumar et al. Reference Kumar, Liu, Boyer and Stahlman2019; Rangani et al. Reference Rangani, Salas-Perez, Aponte, Knapp, Craig, Mietzner, Langaro, Noguera, Porri and Roma-Burgos2019; Briscoe Runquist et al. Reference Briscoe Runquist, Lake, Tiffin and Moeller2019; Varanasi et al. Reference Varanasi, Brabham and Norsworthy2018; Webster and Grey Reference Webster and Grey2015). It has documented resistance to eight herbicide mechanisms of action, with some individual biotypes resistant up to five mechanisms of action (Heap Reference Heap2019). Of the Amaranthus species (i.e., waterhemp [Amaranthus tuberculatus (Moq.) J. D. Sauer], redroot pigweed [Amaranthus retroflexus L.], and tumble pigweed [Amaranthus albus L.]), Amaranthus palmeri is considered the most aggressive, because it has the highest growth rate, biomass accumulation, and total leaf area (Guo and Al-Khatib Reference Guo and Al-Khatib2003; Horak and Loughin Reference Horak and Loughin2000) of these species. It can reduce yield in corn (Zea mays L.), peanut (Arachis hypogaea L.), cotton (Gossypium hirsutum L.), sorghum [Sorghum bicolor (L.) Moench ssp. bicolor], and sweetpotato [Ipomoea batatas (L.) Lam.] by up to 91 %, 68 %, 54 %, 63 %, and 79 %, respectively (Basinger et al. Reference Basinger, Jennings, Monks, Jordan, Everman, Hestir, Waldschmidt, Smith and Brownie2019; Burke et al. Reference Burke, Schroeder, Thomas and Wilcut2007; Massinga et al. Reference Massinga, Currie, Horak and Boyer2001; Meyers et al. Reference Meyers, Jennings, Schultheis and Monks2010; Moore et al. Reference Moore, Murray and Westerman2004; Morgan et al. Reference Morgan, Baumann and Chandler2001). Amaranthus palmeri has proven to be a problematic summer annual weed with the capacity to compete with crops for resources while still maintaining high reproductive capacity (Bensch et al. Reference Bensch, Horak and Peterson2003). Several studies have evaluated the impact of A. palmeri interference in soybean (Bensch et al. Reference Bensch, Horak and Peterson2003; Dieleman et al. Reference Dieleman, Hamill, Weise and Swanton1995; Klingman and Oliver Reference Klingman and Oliver1994; Monks and Oliver Reference Monks and Oliver1988). However, only limited information is available on the intraspecific interference of A. palmeri in a soybean cropping system.
Digitaria sanguinalis is recognized as a common weed in many crops (Van Wychen Reference Van Wychen2016) and was originally brought to the United States as a forage grass (Dickinson and Royer Reference Dickinson and Royer2014). Although Digitaria species have declined in importance (Webster and Coble Reference Webster and Coble1997) due to effective herbicide control options, reports exist of resistance to acetyl CoA carboxylase inhibitors in the United States and resistance to acetolactate synthase (ALS) and photosystem II inhibitors abroad (Heap Reference Heap2019; Hidayat and Preston Reference Hidayat and Preston1997; Laforest et al. Reference Laforest, Soufiane, Simard, Obeid, Page and Nurse2017; Volenberg and Stoltenberg Reference Volenberg and Stoltenberg2002). Additionally, resistant biotypes of D. sanguinalis do not show reduced fitness when compared with susceptible biotypes (Wiederholt and Stoltenberg Reference Wiederholt and Stoltenberg1996). Although D. sanguinalis is not ranked highly as a problematic weed, significant yield losses of 6 %, 50 %, 74 %, 76 %, 89 %, and 100 % in cotton, watermelon [Citrullus lanatus (Thunb.) Matsum. & Nakai], grain sorghum, sweetpotato, snap bean (Phaseolus vulgaris L.), and bell pepper (Capsicum annum L.), respectively, have been documented (Aguyoh and Masiunas Reference Aguyoh and Masiunas2003; Basinger et al. Reference Basinger, Jennings, Monks, Jordan, Everman, Hestir, Waldschmidt, Smith and Brownie2019; Byrd and Coble Reference Byrd and Coble1991; Fu and Ashley Reference Fu and Ashley2006; Monks and Schultheis Reference Monks and Schultheis1998; Smith et al. Reference Smith, Murray, Green, Wanyahaya and Weeks1990). Although D. sanguinalis has been studied in several cropping systems, there has been limited focus on interspecific interference in a soybean cropping system (Oreja and Gonzalez-Andujar Reference Oreja and Gonzalez-Andujar2007) and intraspecific interference of D. sanguinalis.
Despite a large amount of research conducted on season-long weed interference in soybean, much of the focus is on yield loss associated with measured weed densities. The present research was conducted to measure yield loss from season-long competition of D. sanguinalis and A. palmeri when seeded at various densities in soybean. A second objective determined intraspecific competition of each weed species under North Carolina climatic conditions. Studying intraspecific competition allows for an understanding of how weeds perform when a crop is not present. Intraspecific interference is not often studied in agricultural systems. However, understanding the effects of intraspecific weed competition can allow for a better understanding of weed population dynamics, weed biomass accumulation, and intraspecies competition. Lack of a crop may be due to poor crop emergence or crop predation and can mimic areas such as turnrows and crop field edges, where weeds can persist without a crop. Limited information is available on intraspecific competition of A. palmeri or D. sanguinalis and interference of D. sanguinalis in soybean in the United States (Schwartz et al. Reference Schwartz, Gibson and Young2016). Yet D. sanguinalis and A. palmeri are pervasive in soybean, and understanding the weed–crop and weed–weed interactions in this system would provide valuable information that growers can use when making weed management decisions.
Materials and Methods
Field studies were conducted in 2016 and 2017 in conventionally grown soybean (Stowe et al. Reference Stowe, Crozier, Bullen, Dunphy, Everman, Hardy, Osmond, Piggott, Rana, Reisig, Roberson, Schrage, Thiessen and Washburn2018) at the Horticultural Crops Research Station (35.1°N, 81.16°W), Clinton, NC. The studies were conducted on a Norfolk loamy sand (fine-loamy, kaolinitic, thermic Typic Kandiudults) with humic matter 0.31 % and pH 5.9 in 2016, and an Orangeburg loamy sand (fine-loamy, kaolinitic, thermic Typic Kandiudults) with humic matter 0.47 % and pH 5.9 in 2017. A preplant fertilizer of 0 (N)–0 (P2O5)–168 (K2O) kg ha−1 was applied on June 7, 2016, and June 6, 2017, and disked to approximately 10-cm deep. Seven days after fertilizer application, soybean ‘AG6536’ (Monsanto Company, St Louis, MO, USA) seeds were planted with a four-row vacuum planter 0.1 m apart within row and 0.3 m between rows, resulting in a seeding rate of 321,000 seeds ha−1. Plots were 1.2-m wide by 5-m long and consisted of four soybean rows. Treatments were combinations of soybean presence or absence, weed species (A. palmeri or D. sanguinalis), and weed density arranged in a randomized complete block design with three replications. At 1 d after soybean planting, A. palmeri or D. sanguinalis seeds were broadcast by hand into treatment plots designated to receive either A. palmeri or D. sanguinalis and then raked to approximately 0.6-cm deep. The same day, 1.3 cm of overhead irrigation water was applied to facilitate soybean and weed seed germination. Supplemental irrigation was not applied in either study year after the initial irrigation event. Amaranthus palmeri seeds used in this experiment were hand harvested from adjacent fields at the Clinton site in 2015, and D. sanguinalis seeds were purchased from Azlin Seed Service (Azlin Seed Service, Leland, MS, USA). Amaranthus palmeri (collected seed) and D. sanguinalis were not screened for resistance. However, A. palmeri populations in the region of the state where seeds were collected have exhibited resistance to glyphosate and ALS herbicides (Heap Reference Heap2019). No resistant biotypes have been reported for D. sanguinalis in the state from which the seed for this species was purchased (Heap Reference Heap2019). Weeds emerged with the crop and were thinned by hand to densities of 1, 2, 4, and 8 for A. palmeri (by 8-cm stage), and 1, 2, 4, and 16 plants m−2 for D. sanguinalis (by 2–expanded leaves stage), using a 1-m2 quadrat to ensure uniform spatial densities, and establishing 5 subsample populations per plot. Within each block, a weed-free treatment was maintained by hand removal for comparison. Weed densities were based on previous research to ensure levels of interference that would allow for estimations of maximum yield loss (Cowan et al. Reference Cowan, Weaver and Swanton1998; Fu and Ashley Reference Fu and Ashley2006; Klingman and Oliver Reference Klingman and Oliver1994; Meyers et al. Reference Meyers, Jennings, Schultheis and Monks2010; Norsworthy et al. Reference Norsworthy, Oliveira, Jha, Malik, Buckelew, Jennings and Monks2008). Soybean was removed immediately after emergence in the plots where intraspecific competition was studied; weed densities were the same as in the interspecific competition plots. All plots were hand weeded weekly to maintain treatment densities and remove other emerged weed species.
At physiological maturity, 5 plants each of soybean (R6 growth stage) (Stowe et al. Reference Stowe, Crozier, Bullen, Dunphy, Everman, Hardy, Osmond, Piggott, Rana, Reisig, Roberson, Schrage, Thiessen and Washburn2018), A. palmeri (seed beginning to ripen, but all foliage still present), and D. sanguinalis (seed set and beginning of seed shatter), were randomly selected from the center two rows of each plot (when crop and/or weed species were present in treatment plot) and cut at the soil surface. Harvested plants were cut into small pieces and placed in separate two-ply paper bags (40 by 30 by 89 cm) by species. The bags containing fresh plant biomass were weighed to determine fresh biomass and subsequently dried in a propane-fueled forced-air heated drier for 96 h at 80 C and weighed to determine plant dry biomass. Fresh and dry weights of each plant sample per plot (individual bag) were divided by the number of plants harvested (5) to determine fresh and dry biomass weight per plant. Biomass per plant was then multiplied by the number of soybean plants per square meter or weed density per square meter to determine crop and weed biomass per square meter.
To determine soybean yield, soybean was cut at the soil surface at full maturity (R8 growth stage) using a hand-held hedge trimmer (HL 100 K, Stihl USA, Virginia Beach, VA, USA), then placed in large two-ply paper bags, as previously described. Soybean plants from each plot were threshed using a small-plot soybean thresher (B-1, Swanson Agricultural Research Equipment, Seymour, IL, USA) and placed in a seed cleaner (ASC-3, Agriculex, Guelph, ON, Canada) to remove any remaining plant material. Clean soybean seeds were then weighed to determine soybean yield for each plot. Yield loss was calculated as a percent of the weed-free control for each replication. Soybean yield reductions were modeled as a percent reduction in yield as compared with weed-free yield using a rectangular hyperbola function (Cousins Reference Cousins1985):
$${{\rm{Y}}_{\rm{R}}} = \left( {ID} \right)/\left[ {{\rm{1}} + \left( {ID/A} \right)} \right]$$
Descriptions of each variable are as follows: Y R is the reduction in yield as a function of weed density, I is the yield loss associated per weed as weed density approaches zero, A is the asymptote of yield loss as weed density approaches infinity, and D is the weed density. Yields from weed-free plots were used as 100 % yield or zero percent yield loss, and were used to calculate yield loss estimates. PROC NLIN in SAS v. 9.4 (SAS Institute, Cary, NC, USA) was used to obtain yield loss estimates as a percent yield loss, using the rectangular hyperbola model.
For dry biomass weight of crop and weed per square meter, weed biomass per plant, and soybean yield, homogeneity of variance was tested before statistical analysis by plotting residuals. Data were subjected to ANOVA using PROC MIXED in SAS v. 9.4. Year, treatment, and the interaction of treatment and year were treated as fixed effects, and replication within each year was treated as a random effect. Contrast statements to test for linear trends were used if the interaction of treatment and year was not significant with α ≥ 0.05, and means were averaged over years. When the year and treatment interaction was significant, response variables were analyzed by year. Weed biomass per square meter, weed biomass per plant, and yield were log transformed for analysis. Log-transformed data for weed biomass per square meter, weed biomass per plant, and crop biomass were subjected to ANOVA using PROC MIXED in SAS v. 9.4.
Interactions for weed density and year were not significant for weed biomass per plant response to weed density; therefore, data were averaged over years. Predictions using linear quadratic or other higher-order regression models did not fit the response of weed biomass per plant for D. sanguinalis or A. palmeri. Weed biomass per plant was compared using differences of least-squares means at each density for weeds grown with and without soybean. Comparisons of weed density were according to Tukey’s honestly significant difference (HSD) for each weed species at α = 0.05. All means reported are nontransformed.
Weed species response in the presence and absence of the soybean crop for weed biomass per square meter was fit to a linear regression model with Equation 2:
$$Y = {y_0} + bx$$
where Y is the estimated biomass per square meter, y 0 is the y intercept for the regression line, and b is the slope for the predicted values of weed biomass per square meter, regressed against weed density.
For comparison to current recommendations for crop loss due to weed interference, weed densities measured in this study were entered into the North Carolina Web Herbicide Application Decision Support System (WebHADSS; Lassiter and York Reference Lassiter and York2009). Loss estimates using WebHADSS are calculated based on the 10-yr average soybean yield (2,320 kg ha−1) (USDA-NASS 2018) and average farm size (68 ha−1) (USDA-NASS 2012) for North Carolina.
Results and Discussion
Interspecific Interference
Interactions of weed density by year were not significant for biomass per plant or biomass per square meter of either weed. Thus, the data for these parameters were averaged across years for D. sanguinalis or A. palmeri. Biomass per square meter of D. sanguinalis or A. palmeri with soybean increased with increasing weed density (Figures 1 and 2). Biomass per square meter increases were due to increasing weed number not to increased weed biomass per plant (Table 1). Klingman and Oliver (Reference Klingman and Oliver1994) also reported increases of A. palmeri biomass per square meter with increasing density in soybean. Biomass of both weed species per square meter was greater in the absence of soybean.
Digitaria sanguinalis dry biomass (kg) per square meter as a function of increasing D. sanguinalis density per square meter. Digitaria sanguinalis was grown with ‘AG6536’ soybean or without soybean at the Horticultural Crops Research Station, Clinton, NC, in 2016 and 2017. Dry biomass per square meter (averaged over years) of D. sanguinalis growing with and without soybean were fit to a linear model (equation: y = y 0 + bx). When grown with soybean, y 0 = 0.023 (0.025), b = 0.0524 (0.003), R 2 = 0.99. When grown without soybean, y 0 = 0.27 (0.057), b = 0.0573 (0.007), R 2 = 0.97, with SEs in parentheses.
Amaranthus palmeri dry biomass (kg) per square meter as a function of increasing A. palmeri density per square meter. Amaranthus palmeri was grown with ‘AG6536’ soybean or without soybean at the Horticultural Crops Research Station, Clinton, NC, in 2016 and 2017. Dry biomass per square meter (averaged over years) of A. palmeri growing with and without soybean was fit to a linear model (equation: y = y 0 + bx). When grown with soybean, y 0 = 0.105 (0.065), b = 0.051 (0.014), R 2 = 0.87. When grown without soybean, y 0 = 0.382 (0.010), b = 0.095 (0.002), R 2 = 0.99, with SEs in parentheses.
Mean weed biomass per plant (kg) followed by SE in parentheses, for Digitaria sanguinalis and Amaranthus palmeri at four densities, grown in the presence and absence of ‘AG6536’ soybean at the Horticultural Crops Research Station, Clinton, NC, averaged over 2016 and 2017.
a P-values are the result of differences in least-squares means comparing crop and no-crop plots for each weed at the given density.
b Different letters within the same column and species indicate significance (P ≤ 0.05) according to Tukey’s HSD.
Biomass per plant of either weed growing with soybean did not vary across density (Table 1). Data for weed biomass per plant did not fit linear or nonlinear models, and thus were compared for each weed using Tukey’s HSD. The lack of weed biomass per plant changing with increasing weed density is in contrast to the findings of Burke et al. (Reference Burke, Schroeder, Thomas and Wilcut2007), who reported decreasing weed biomass per plant for A. palmeri with increasing weed density when grown with peanut. In the present study, soybean biomass per square meter was unaffected by A. palmeri and D. sanguinalis density (data not shown). Loss of soybean biomass has been seen with other weeds [common ragweed (Ambrosia artemisiifolia L.), barnyardgrass [Echinochloa crus-galli (L.) Beauv.], annual sowthistle (Sonchus oleraceus L.), American sloughgrass [Beckmannia syzigachene (Steud.) Fernald], and common lambsquarters (Chenopodium album L.)] at densities greater than those evaluated in this study (up to 140 plants m−2) and at a wider row spacing (70 cm) (Song et al. Reference Song, Kim, Im, Lee, Lee and Kim2017). Soybean biomass may vary and weed species may respond differently with wider rows or at higher planting densities, which are factors that influence soybean and weed plant heights or leaf area index and contribute to crop and weed biomass (Howe and Oliver Reference Howe and Oliver1987; McWhorter and Sciumbato Reference McWhorter and Sciumbato1988; Song et al. Reference Song, Kim, Im, Lee, Lee and Kim2017).
Intraspecific Interference
Biomass per square meter response for intraspecific interference increased with increasing density, similar to the biomass per square meter response when D. sanguinalis and A. palmeri were grown with soybean (Figures 1 and 2). Biomass per square meter of D. sanguinalis was 617 % and 37 % greater for 1 plant m−2 and 16 plants m−2, respectively, when grown without soybean than when grown with soybean. Biomass per square meter of A. palmeri was 272 % and 115 % greater for 1 and 8 plants m−2, respectively, when grown without soybean than when grown with soybean. Weed densities higher than those measured in this study would likely result in additional biomass losses for both weed species. Density of each weed species was maintained throughout each season, and no weed mortality was noted after thinning, indicating the carrying capacity of this system is likely to be greater than the combination of the density of the planted soybeans and the weed densities measured. If the carrying capacity were to be met, biomass per square meter would stabilize, and no further recruitment from seeds would be needed. Additional recruitment would only be possible if weed densities dropped below the carrying capacity.
For D. sanguinalis, biomass per plant was greater when grown without soybean than with soybean for all weed densities, except for the two highest measured densities, 4 and 16 plants m−2 (Table 1). This finding indicates that intraspecific competition for resources may be occurring with D. sanguinalis at a density as low as 4 plants m−2, affecting biomass per plant. Biomass per plant for either weed when grown without soybean was greatest at the lowest density (1 plant m−2). Furthermore, similar slopes seen in Figure 1 indicate that intraspecific competition did not reduce weed biomass per square meter for D. sanguinalis, even in the presence of soybean.
For A. palmeri, biomass per plant was higher across all densities when grown without soybean, and biomass per plant was the lowest at 4 and 8 plants m−2 compared with 1 plant m−2. Weed biomass per plant was similar across all densities of A. palmeri when grown with soybean, indicating that soybean was competitive with A. palmeri, reducing biomass per plant despite increases in weed density. Amaranthus palmeri results for biomass per square meter were similar to results for D. sanguinalis, showing increasing biomass per square meter with increasing weed density (Figure 2). Additionally, different slopes between soybean presence and absence for A. palmeri indicate that there is both intraspecific and interspecific competition occurring at 4 and 8 plants m−2 when soybean is present.
Lower biomass of D. sanguinalis or A. palmeri with soybean relative to without soybean suggests that interspecific interference is occurring between soybean and both weed species across all densities. Within density, D. sanguinalis or A. palmeri biomass per plant was higher when grown without soybean, except for D. sanguinalis at 4 and 16 plants m−2 (Table 1). These results suggest that D. sanguinalis may be able to tolerate higher intraspecies densities than A. palmeri. Results from the present study suggest that increased weed biomass per square meter is due to increasing weed density, not increased weed biomass per plant (Table 1). In previous studies, similar responses were seen in E. crus-galli in the absence of tomato (Solanum lycopersicum L.), spiny amaranth (Amaranthus spinosus L.) in lettuce (Lactuca sativa L.), and with A. palmeri and D. sanguinalis in absence of sweetpotato, with increasing biomass per meter of row and decreasing biomass per plant with increasing density (Basinger et al. Reference Basinger, Jennings, Monks, Jordan, Everman, Hestir, Waldschmidt, Smith and Brownie2019; Norris et al. Reference Norris, Elmore, Rejmanek and Akey2001; Shrefler et al. Reference Shrefler, Shilling, Dusky and Brecke1994). Previous work done by the authors showed greater biomass accumulation per plant at the same densities of D. sanguinalis and A. palmeri in sweetpotato compared with the present study (Basinger et al. Reference Basinger, Jennings, Monks, Jordan, Everman, Hestir, Waldschmidt, Smith and Brownie2019). Although a direct comparison cannot be made, these results suggest that the cropping system may have an effect on weed intraspecific interference. The authors believe that greater weed biomass accumulation per plant in sweetpotato may be due to weed spatial distribution due to tillage events that coincide with potassium fertilizer application. Previous work done by Norris et al. (Reference Norris, Elmore, Rejmanek and Akey2001) indicated that weeds that were clumped reduced plant biomass and seed production compared with random or evenly spaced plants at the same density.
Yield
Soybean yield responses to densities of A. palmeri and D. sanguinalis lacked year by treatment interactions and therefore were combined over years. Rainfall and growing degree days were higher in 2016 than in 2017 but did not affect yield between years (Table 2). Weed-free yields averaged 3,093 kg ha−1, above the 10-yr NC soybean average of 2,320 kg ha−1 (USDA-NASS 2018). Soybean yield reductions were fit to a rectangular hyperbola model (Cousins Reference Cousins1985) for D. sanguinalis (Figure 3) and A. palmeri (Figure 4), with yield reduction increasing as weed density increased. Yield loss ranged from 0 % at 1 D. sanguinalis plant m−2 to 37 % at 16 D. sanguinalis plants m−2, and from 19 % at 1 A. palmeri plant m−2 to 37 % at 8 A. palmeri plants m−2. The I parameter, yield loss as weed density approaches zero, for D. sanguinalis and A. palmeri was calculated as 9 % and 20 % respectively. Yield loss as weed density approaches zero estimates for D. sanguinalis averaged 33 % for snap bean and 40 % for bell pepper (Aguyoh and Masiunas Reference Aguyoh and Masiunas2003; Fu and Ashley Reference Fu and Ashley2006). Lower yield loss as weed density approaches zero values indicated that D. sanguinalis is less competitive with soybean compared with A. palmeri. Amaranthus palmeri interference estimates for yield loss were 118 % for corn (Massinga et al. Reference Massinga, Currie, Horak and Boyer2001), 87 % for soybean (Bensch et al. Reference Bensch, Horak and Peterson2003), and constrained to 100 % in peanut (Burke et al. Reference Burke, Schroeder, Thomas and Wilcut2007). Yield loss as weed density approaches zero for A. palmeri was reported as 11.8 % to 104.6 % in soybean (Bensch et al. Reference Bensch, Horak and Peterson2003), 39 % in peanut (Burke et al. Reference Burke, Schroeder, Thomas and Wilcut2007), and 90 % in corn (Massinga et al. Reference Massinga, Currie, Horak and Boyer2001).
Monthly rainfall (mm) and growing degree days (GDD; base 10 C) at the Horticultural Crops Research Station, Clinton, NC, from May to September 2016 and 2017.a
a Data were collected from an on-site weather station.
‘AG 6536’ Soybean yield loss ( %), based on weed-free soybean maximum yield, from Digitaria sanguinalis. Mean soybean yield loss for 2016 and 2017 is plotted as a function of increasing large crabgrass density per square meter at the Horticultural Crops Research Station, Clinton, NC. Data were fit to a rectangular hyperbola model (equation: Y R = (ID)/[1+(ID/A)]), with I = 9.17 (4.75), A = 50.47 (17.34), R 2 = 0.84; SE is given in parentheses.
‘AG 6536’ soybean yield loss ( %), based on weed-free soybean maximum yield, from Amaranthus palmeri. Mean yield loss of soybean for 2016 and 2017 is plotted as a function of increasing A. palmeri density per square meter at the Horticultural Crops Research Station, Clinton, NC. Data were fit to a rectangular hyperbola model (equation: Y R = (ID)/[1 + (ID/A)]), where I = 20.28 (7.19), A = 49.42 (10.40), R 2 = 0.84; SE is given in parentheses.
Parameter A, the asymptote for the regression model that estimates the maximum estimated yield loss according to the rectangular hyperbola model, was 50 % and 49 % for D. sanguinalis and A. palmeri, respectively. Predicted values for maximum yield loss were considered reliable estimates for D. sanguinalis and A. palmeri, as the standard errors of the parameter estimates were less than half of the estimated values (Koutsoyiannis Reference Koutsoyiannis1973). Digitaria sanguinalis interference estimates for maximum yield loss have been reported as 62 % in snap bean (Aguyoh and Masiunas Reference Aguyoh and Masiunas2003) and 91 % to 100 % in bell pepper (Fu and Ashley Reference Fu and Ashley2006). Although maximum yield loss parameters indicate that A. palmeri is more competitive than D. sanguinalis at lower densities, similar yield reductions did occur in our study at 8 and 16 plants m−2 for A. palmeri and D. sanguinalis, respectively. Yield reductions, estimated as 50 % and 49 % for D. sanguinalis and A. palmeri, respectively, indicate that management of these weeds is needed to prevent significant yield loss. The greater interference of A. palmeri compared with D. sanguinalis in soybean may be due to morphological differences between these two weed species. Digitaria sanguinalis has small leaf blades and is lower in height than A. palmeri, which may allow D. sanguinalis to compete well with the crop, but does not allow for it to shade out the crop, as A. palmeri can. Biomass accumulation (plant−1 and m−2) of A. palmeri is almost double that of D. sanguinalis at densities less than 4 plants m−2. Greater biomass accumulation at densities less than 4 plants m−2 may be a contributing factor to higher yield reductions at lower densities for A. palmeri. Although the impact of A. palmeri on light and water was not measured in this study, competition for these resources may have contributed to yield reductions, as is seen in previous work in corn and soybean (Green-Tracewicz et al. Reference Green-Tracewicz, Page and Swanton2012; Massinga et al. Reference Massinga, Currie and Trooien2003). Biomass accumulation for D. sanguinalis was less than for A. palmeri, which may have limited the interference of D. sanguinalis at lower measured densities (≤4 plants m−2). However, at the higher densities measured in this study (≥4 plants m−2), D. sanguinalis had similar (4 plants m−2) or greater biomass (16 plants m−2) than A. palmeri. The similarity in biomass at higher densities (≥4 plants−2) in this study may have been a contributing factor to similar soybean yield reductions.
Results from this study suggest that control of A. palmeri is necessary at 1 plant m−2 and control must be implemented for D. sanguinalis at greater than 2 plants m−2 to prevent yield loss. This study also brings to light the competitive nature of D. sanguinalis, which can be controlled with POST herbicides but could be overlooked with the integration of new soybean seed technologies resistant to dicamba and 2,4-D, which are ineffective in controlling grasses such as D. sanguinalis. Therefore, an efficacious herbicide with grass activity should be included as part of a weed management program for soybean. This study focused on high planting densities of soybean, which could have limited the competitive nature of these weeds due to a high crop population density. In this study, soybean planting density was at a high seeding rate (321,000 seeds ha−1) and narrow row spacing (0.3 m between rows). Row spacing can have an effect on total weed biomass accumulation and soybean yield loss (Hock et al. Reference Hock, Knezevic, Martin and Lindquist2006). Wider row spacing than in the present study may alter the interference of A. palmeri and D. sanguinalis on soybean.
Decision support systems such as WebHADSS, Pocket HERB, and PAM (Palmer Amaranth Management) have been developed to determine thresholds for weeds and assist growers in making management decisions (Bennett et al. Reference Bennett, Price, Sturgill, Boul and Wilkerson2003). As a means of comparison, weed densities in the present study were entered into the WebHADSS system to estimate soybean yield loss. For A. palmeri, yield loss estimates from WebHADSS were 33 % and 70 % for 1 and 8 plants m−2, respectively. These estimates were higher than yield loss from 1 and 8 plants m−2 observed in the study (14 % and 38 %, respectively). Results from our study suggest that D. sanguinalis may be more competitive than indicated by the data used in WebHADSS. WebHADSS yield loss estimates were 1 % and 16 % for 1 and 16 D. sanguinalis plants m−2, which were lower than predicted losses from this study (9 % and 38 %, respectively). One of the issues with the WebHADSS system is that it does not allow for input of soybean stand density and row spacing. Furthermore, these systems need additional updates to reflect current crop varieties and more recent research. The planting density of soybean in this study could have contributed to greater soybean competition for light and other resources with A. palmeri than is assumed by the WebHADSS system. Additionally, the underestimation of yield loss due to D. sanguinalis interference may be due to lack of specificity of Digitaria species in the WebHADSS system, as this system does not allow for selection of specific Digitaria species.
Results from this study provide estimations for the effect of season-long interference of D. sanguinalis and A. palmeri on soybean and the impact of intraspecific and interspecific interference of these weeds in soybean. Digitaria sanguinalis and A. palmeri reduced soybean yield when present at 2 and 1 plants m−2, respectively. The presence of soybean resulted in reduced weed biomass across all weed densities, reducing weed growth. Furthermore, resistant A. palmeri or D. sanguinalis biotypes that are not controlled by herbicide applications show only moderate reductions (Chandi et al. Reference Chandi, Jordan, York, Milla-Lewis, Burton, Culpepper and Whitaker2012) or no reductions in fitness (Giacomini et al. Reference Giacomini, Westra and Ward2014; Wiederholt and Stoltenberg Reference Wiederholt and Stoltenberg1996), resulting in crop yield reductions from weeds. Therefore, it may be advantageous to use management practices such as increased seeding density and narrow row spacing to further reduce the competitiveness of weeds with soybean (Hock et al. Reference Hock, Knezevic, Martin and Lindquist2006; Howe and Oliver Reference Howe and Oliver1987).
Both interspecific and intraspecific interference of D. sanguinalis and A. palmeri were observed in our studies. Evidence of interspecific interference in our studies was the observed reduction in weed biomass with soybean and the decrease in soybean yield as weed density increased. Decreasing weed biomass per plant with increasing density without soybean indicated the impact of intraspecific competition of D. sanguinalis and A. palmeri. Future studies should consider using additional densities of D. sanguinalis and A. palmeri to allow for more precise estimation of intraspecific and interspecific interference between soybean and A. palmeri or D. sanguinalis under different environments. Additional studies should investigate the competitive nature of D. sanguinalis in other row-crop and horticultural cropping systems, as there is limited research concerning its interference. Finally, quantifying impacts of resistant biotypes, weed emergence timings, and varying management conditions (irrigation, tillage systems, row spacing, and fertilization regimes) on weed interference would provide insight into additional management strategies to limit weed interference.
Acknowledgments
The authors would like to thank and acknowledge the North Carolina Agricultural Research Service, North Carolina Cooperative Extension Service, the Department of Horticultural Science, and the North Carolina Department of Agriculture and Consumer Services for funding this research. Additionally, the authors would like to thank the staff of the Horticultural Crops Research Station, Clinton, NC, for management and harvesting assistance, with special thanks to Wesley Hairr, Glen Aman, Dusty Jolly, and Rodney Mozingo. Thanks to Cole Smith, Matthew Waldschmidt, Andrea Genna, Amanda Yonnoni, Anna Wyngaarden, Rachel Berube, and Lauren Deans for their work in plot maintenance, data collection, and harvesting. Thanks to David Wooten, William Garrett, and Mike Baker from Monsanto for supplying seed for the study. No conflict of interest has been declared.
