Screening preemergence herbicides for weed control in cassava

Abstract Weed competition severely constrains cassava root yield in sub-Saharan Africa; thus, good weed control measures, including the use of herbicides, are increasingly important. Herbicide trials were conducted at five locations across eastern, western, and north-central Nigeria over two cropping seasons (2014 and 2015). Nineteen premixed PRE herbicides applied at different rates were evaluated for efficacy on weeds and selectivity on cassava. Manual hoe-weeding at 4, 8, and 12 wk after planting (WAP) and two S-metolachlor + atrazine treatments commonly used by cassava growers were included for comparison. Six of the 19 PRE herbicide treatments (indaziflam + isoxaflutole, indaziflam + metribuzin, flumioxazin + pyroxasulfone, isoxaflutole, acetochlor + atrazine + terbuthylazine, and terbuthylazine + S-metolachlor) consistently provided 80% to 98% broadleaf and grass weed control up to 8 wk after treatment. Overall, PRE herbicide treatments and cassava yield were significantly positively correlated. Herbicide treatments terbuthylazine + S-metolachlor, flumioxazin + pyroxasulfone, diflufenican + flufenacet + flurtamone (respectively, 60 + 60 + 60, 120 + 120 + 120, 90 + 360 + 120, and 135 + 360 + 180 g ha–1), acetochlor + atrazine + terbuthylazine (875 + 875 + 875 g ha–1), S-metolachlor + atrazine (870 + 1,110 g ha–1), oxyfluorfen (240 g ha–1), indaziflam + isoxaflutole (75 + 225 g ha–1), indaziflam + metribuzin (75 + 960 g ha–1), and aclonifen + isoxaflutole (500 + 75 g ha–1) contributed to yields exceeding twice the Nigerian national average of 8.76 tonnes ha–1. These treatments had root yields of 1.4 to 2 times higher than plots that had been hoe-weeded three times. There were some adverse herbicide treatment effects such as delayed cassava sprouting and temporary leaf bleaching observed in indaziflam and diflufenican + flufenacet + flurtamone treatments, whereas sulfentrazone caused prolonged leaf crinkling. The PRE applications alone at rates safe for cassava did not provide adequate season-long weed control; supplemental POST weed control is needed about 10 WAP for satisfactory season-long control.


Introduction
Cassava is extensively cultivated in the humid and subhumid tropical regions of Africa (Lebot 2009), which produces more than 54% of the world's cassava output (FAOSTAT 2014). This crop is cultivated mainly by smallholders and medium-scale farmers in 37 countries (FAOSTAT 2011). Nigeria is a global leader in cassava production with an output of approximately 59.5 million tonnes from 6.79 million hectares under cassava cultivation (FAOSTAT 2017). This output accounts for approximately 62% of cassava production in West Africa (FAOSTAT 2011). Cassava is an important crop in sub-Saharan Africa (Nweke 2004), where it is a major staple food for more than 200 million people (Nweke and Emete 1999). It is the second most important staple food crop after maize in terms of calories consumed (Nweke 1994). Cassava plays a vital role in the food economy of many African countries, including Nigeria, where it remains a strategic crop for both food security and poverty alleviation (Donkor et al. 2017;FAO 2011). This crop is now also an essential source of industrial raw material for the production of starch, bioethanol, high-quality flour for pharmaceuticals, food, and beverages and has the potential to contribute to the economic growth of Nigeria and most cassava-producing countries in sub-Saharan Africa.
A major challenge to cassava production in Nigeria is the low root yield (8 to 12.6 tonnes ha −1 ) obtained by smallholders and medium-scale farmers (Donkor et al. 2017;Ekeleme et al. 2016;FAOSTAT 2017) compared with yields ranging from 20 to >35 tonnes ha −1 from Asian and Caribbean countries (Donkor et al. 2017;Hauser and Ekeleme 2017). Yields higher than 25 tonnes ha −1 have been achieved in Nigeria on research plots with appropriate crop management (Ekeleme et al. 2016;Hauser and Ekeleme 2017).
Poor weed control has been identified as a major cause of low yields in farmers' fields (Chikoye et al. 2001;Ekeleme et al. 2016;Howeler 2007). Although competition from weeds occurs at all periods of growth, the most damaging effects of weeds on cassava occur during two specific periods: the first 3 to 12 wk after planting (WAP) when the crop is in its early canopy-formation stage and the third month after planting when the storage roots commence bulking (Akobundu 1980;Chikoye et al. 2001;Melifonwu 1994;Onochie 1975). Several studies have stressed the importance of early weed control in the first 1 to 3 mo after planting to achieve high yields (Aye 2011;Howeler 2007;Tongglum et al. 1992).
Manual weeding is the most common method of weed control in cassava cultivation in Nigeria. Farmers carry out two to three hoe-weedings in the first growth cycle of cassava, but in areas where rhizomatous perennial weeds such as cogon grass (Imperata cylindrica L.) or sedges are dominant, additional hoeweeding may be required. Generally, manual hoe-weeding is labor-intensive and time-consuming, and in most cases, farmers do not follow the recommended weeding regimes of 3, 6, and 9; or 4, 8, and 12 WAP (Adigun and Lagoke 2003;Ekeleme et al. 2016;Joshua and Gworgwor 2000).
Smallholders and medium-scale farmers increasingly use herbicides to control weeds in cassava due to reduced manpower availability and high labor costs. Odoemenem and Otanwa (2011) reported that 68.9% of smallholder farmers used herbicides to control weeds in cassava in north central Nigeria. Of those farmers, 9.5% used a variety of PRE herbicide formulations, whereas 69.0% used POST herbicides such as glyphosate (52.6%) and paraquat (16.4%). Currently, a limited number of herbicides are registered for use in cassava production in Nigeria.
The most common PRE herbicides currently used by cassava farmers are formulations containing atrazine, diuron, and S-metolachlor. These herbicides are usually applied at high doses that are prohibitively expensive for smallholder farmers. It is therefore essential to provide farmers with efficient and sustainable weed management to enhance cassava yields. This could be achieved with PRE herbicides that are effective against weeds and environmentally safe. The objective of this research was to identify additional PRE herbicide options for weed control in cassava production ecosystems in Nigeria.

Site Description and Treatment Application
Trials were conducted in two cropping seasons (first season, April to December 2014; second season, August 2014 to April 2015) at five locations representing three agroecologies in Nigeria: Humid Forest, Humid Forest Transition (Derived Savanna), and Southern Guinea Savanna (Figure 1). The Humid Forest and Humid Forest Transition agroecologies have a growing season rainfall of >1,300 mm with a growing period of 211 to 270 d and a window for two planting periods (referred to here as "seasons").
The Southern Guinea Savanna has a growing season rainfall of 1,200 to 1,500 mm with a shorter growing period of 181 to 210 d and a single planting window (Ekeleme et al. 2003). The first season trials were established in April 2014 at three locations and the second season in August 2014 at two locations (Table 1).
Nineteen PRE herbicides (Table 2) and manual hoe-weeding at 4, 8, and 12 WAP were evaluated for weed control efficacy. A weedy check was included for determination of weed control (%) values. Except for isoxaflutole and metribuzin, which had only one recommended application rate, the other herbicides were evaluated at two or three rates. Application rates were selected to represent lower label-recommended rates and 1.5 times the recommended application rate except for herbicides supplied by Bayer Crop Science, which provided application rates for its products. In total, 49 PRE herbicide treatments were evaluated. At each site, trials were established in a randomized complete block design with three replications. All herbicide treatments were commercial formulations and S-metolachlor þ atrazine, which is commonly used by farmers, was included for comparison.
The treated plot size at each location was a 4.0 × 4.0 m square. An erect cassava variety, TME 419, was used in the trial. The planting material consisted of cassava stem cuttings measuring 25 cmknown as cassava stakes. The first season trial was on a scale of 10,000 plants ha −1 and 20,000 plants ha −1 were used in the second season trial. The increase in cassava plant numbers for the second planting season was aimed at achieving early canopy closure before the dry season commenced. Every site was treated with glyphosate to control perennial grasses and broadleaf weeds at 2 to 3 wk before field plowing. The experimental site was plowed and then harrowed 2 wk after plowing. The PRE herbicide treatments were applied 1 or 2 d after planting (DAP) with a hand-pumped CP 15 (COOPER PEGLER®) knapsack sprayer calibrated to deliver 250 L ha −1 of water at 240 kPa through a Cooper Pegler Hypro Polijet nozzle AN1.2 Green (EXEL GSA, ZI NORS ARNAS -BP 30 424, 69653 Villefranche Cedex). The overall herbicide efficacy in each plot was visually assessed at 8 wk after treatment (WAT) using a 0 to 100 scale, where 0 = no weed control, 10 to 49 = poor control, 50 to 69 = moderate control, 70 to 79 = fair/acceptable control, 80 to 89 = good control, and 90 to 100 = excellent control. Weed species in each plot were identified and counted in two 1-m 2 quadrants placed along a diagonal transect in each plot at each herbicide efficacy rating period. Weed species density data were used to estimate herbicide efficacy on major species as follows: Where WSP untreated and WSP treated is the weed species population in untreated plots and treated plots, respectively. To evaluate crop selectivity, phytotoxicity was assessed by visually rating crop damage on a scale of 0 (no phytotoxicity) to 10 (total plant death) at 2 and 4 WAT. All plots treated with herbicides were hoe-weeded at 10 wk after PRE herbicide treatment. Cassava stand establishment was assessed at 8 WAP and calculated as a percentage of planted stakes that sprouted. Cassava stakes that failed to sprout after the application of herbicides were not replaced. Fresh roots were harvested at 9 mo after planting at each site.

Statistical Analysis
Analysis of variance was used to examine the differences in treatment effects of two key yield variables: cassava stand count at 8 WAP and cassava fresh root yield at crop harvest. Significant treatment means were separated using the LSD at 5% probability. Where a location-by-treatment interaction effect was significant (P < 0.05), simple effect differences were evaluated among treatments to understand the nature of the interactions. Also, evaluated were Pearson linear correlation coefficients of cassava fresh root yield with herbicide control efficacy on various weed types (broadleaf weeds, grassy weeds, and all weeds) to discern the level of association between cassava root yield and weed control measures. Heat map presentation of the estimates of herbicide efficacy on all weeds, dominant weed species, cassava stand establishment, and root yield was used to strengthen the understanding and interpretation of the data matrix. Data on cassava stand establishment (%) was log 10 (x þ 1) transformed before analysis to stabilize the  variance. All data analyses were performed using SAS software (version 9.4; SAS Institute Inc., Cary, NC).

Results and Discussion
There was a strong seasonal influence on herbicide efficacy, cassava stand establishment, and root yield. In the first cropping season, location-by-treatment interaction was significant (P < 0.01) for herbicide efficacy, cassava stand establishment, and root yield; therefore, data are presented separately by season and location.
In the second cropping season, location by treatment was not significant (P > 0.20) for cassava stand establishment (%) or cassava stand population at crop harvest; therefore, data were pooled over location for these variables and combined data are presented. Data on herbicide efficacy against major weeds are presented by location.
= Excellent weed control (90% to 100%), = Good weed control (80% to 89%), = Fair/acceptable weed control (70% to 79%), = moderate weed control (50% to 69%), = poor weed control (10% to 49%). site (Southern Guinea Savanna), Olorunmaiye and Olorunmaiye (2008) observed that coat buttons, crabgrass, and itchgrass were not controlled by S-metolachlor þ atrazine or by metolachlor þ metobromuron when applied PRE alone or when followed by one or two manual hoe-weedings at 6 and 12 WAP in a cassava/maize intercrop. Several studies have identified resistance in some populations of itchgrass to some acetyl coenzyme-A carboxylase-inhibiting herbicides in Bolivia and Costa Rica (Avila et al. 2007;Castillo-Matamoros et al. 2016). At the International Center for Tropical Agriculture in Colombia, oxyfluorfen did not control goosegrass at the label rate of 0.5 kg ai ha −1 (Tonggulum and Leihner 2015). In the aforementioned study, goosegrass was controlled at a higher oxyfluorfen rate, but this caused unacceptable damage to cassava at 28 d after application. Flumioxazin þ pyroxasulfone provided excellent control of coat buttons, mint weed, ironweed, crabgrass, goosegrass, and itchgrass.

Herbicide Efficacy at the Oyo State Site
At the International Institute of Tropical Agriculture (IITA) site in Oyo State, indaziflam þ metribuzin, terbuthylazine þ S-metolachlor, sulfentrazone, S-metolachlor þ atrazine, and acetochlor þ atrazine þ terbuthylazine provided excellent (90% to 97%) control of grasses up to 8 WAT (Table 6, dark green). Indaziflam þ isoxaflutole and indaziflam þ metribuzin consistently had the highest efficacy against broadleaf and grass weeds relative to other tested herbicides at most locations. This trend may be attributed to indaziflam, which has been reported to provide seasonlong residual control of annual grasses and broadleaf weeds in many crops when applied PRE (Sebastian et al. 2017;Singh et al. 2011). In Florida, Singh et al. (2011) reported that indaziflam applied as a PRE herbicide provided 3 to 4 mo of residual weed control in citrus groves.
In general, flumioxazin þ pyroxasulfone consistently provided very good to excellent control of broadleaf and grass weeds at all sites (Table 6, dark green). Flumioxazin þ pyroxasulfone has been reported to provide excellent weed control at rates similar to those evaluated in this study. This herbicide has been shown to be effective in soybean fields against broadleaf weeds such as velvetleaf (Abutilon theophrasti Medik), redroot pigweed (Amaranthus retroflexus L.), smooth pigweed (Amaranthus hybridus L.), and lambsquarter (Chenopodium album L.; Mahoney et al. 2014). Curtis et al. (2011) reported its effectiveness against grasses such as annual bluegrass (Poa annua L.), perennial ryegrass (Lolium perenne L.), and tall fescue [Lolium arundinaceum (Schreb.)]. Long soil residual activity has been reported for flumioxazin þ pyroxasulfone (Bernards et al. 2010).

First Cropping Season
In the first cropping season, a significant treatment effect on cassava stand establishment was observed at the NRCRI (P = 0.0246), Table 6. Percentage of broadleaf and grass weeds controlled by different herbicide treatments at 8 wk after treatment in the second cropping season at two sites in Nigeria.  (Table 7). Cassava establishment among herbicide treatments at the three sites (NRCRI, UAM, and FUNAAB; coded yellow in Table 7) was below 80%. At the NRCRI and UAM sites, cassava establishment in the other herbicide treatments ranged from 80% to 98% except for dimethenamid-P þ pendimethalin treatment at the NRCRI site (Table 7, light brown). Cassava stakes in plots treated with all herbicides that contained indaziflam exhibited delayed sprouting at all sites. Delay in cassava sprouting was consistent in plots treated with indaziflam þ isoxaflutole and indaziflam þ isoxaflutole (Table 7, coded yellow). These two herbicide treatments contain a higher concentration of indaziflam.
In the majority of the plots treated with indaziflam-containing herbicides, we noted that buds on the exposed part of unsprouted cassava stakes were still fresh and green in color when they were scratched. Still, when the stakes were inspected at 3 WAP, we observed cassava shoots emerging from the buried portion of the stake. A possible explanation is that cassava shoots sprouting from the buried buds on the stake required more time to emerge, and in clay soil, emergence may be prolonged or hindered. Indaziflam has been shown to inhibit cellulose biosynthesis in plants (Brabham et al. 2014), and this may be responsible for Values followed by the same letter within a column are not significantly different at α = 0.05. c Hoe-weeded at 4, 8, and 12 wk after planting. the delayed sprouting. At the FUNAAB site, cassava establishment was generally poor due to poor soil drainage.

Second Cropping Season
In the second cropping season, cassava establishment was >80% at the NRCRI and IITA sites except in plots treated with indaziflam þ isoxaflutole and indaziflam þ metribuzin, which had 73% to 79% of the expected cassava population of 20,000 plants ha −1 (Table 7, yellow). In both cropping seasons, diflufenican þ flufenacet þ flurtamone caused temporary leaf bleaching for 2 to 3 wk. Sulfentrazone caused prolonged leaf crinkling at the tip of the cassava shoot. Legleiter and Johnson (2013) reported similar symptoms on soybean leaves in fields treated with sulfentrazone as a PRE herbicide. Crop injury from sulfentrazone has often been attributed to such conditions as wet soil or heavy rainfall (Legleiter and Johnson 2013;Swantek et al. 1998). Swantek et al. (1998 noted that several heavy rainfall events totaling 170 mm in 16 DAP caused significant injury to soybean in Keiser, Florida. Our study was conducted in areas that received a total annual rainfall of 1,100 to 1,500 mm.

Cassava Fresh Root Yield
Cassava fresh root yield in the first cropping season varied with location (P = 0.0047). The location-by-treatment effect was significant only in the first cropping season (P = 0.0084).
First Cropping Season A significant 7.72 to 12.3 tonnes ha −1 increase in fresh root yield was observed in herbicide-treated plots compared with hoeweeded plots at the NRCRI site. This result occurred in plots treated with terbuthylazine þ atrazine (Table 8, dark and light codes), and acetochlor þ atrazine þ terbuthylazine, diflufenican þ flufenacet þ flurtamone, oxyfluorfen, and indaziflam þ isoxaflutole (Table 8, light green). Indaziflam þ metribuzin, isoxaflutole, S-metolachlor þ atrazine, flumioxazin þ pyroxasulfone, aclonifen þ isoxaflutole, clomazone þ pendimethalin, dimethenamid-P þ pendimethalin, and diflufenican þ flufenacet þ flurtamone had a 4 to 7.3 tonnes ha −1 yield advantage over the hoe-weeded treatment (Table 8, light green). Overall, uncontrolled weed growth in the untreated plots led to a reduction in root yield of 28.5% to 66.4%. Cassava fresh root yield at the NRCRI site in 23 out of 45 herbicide treatments was 1.5 to more than 2 times the Nigerian national root yield average. FAOSTAT (2017) reports a national yield average of 8.76 tonnes ha −1 for Nigeria, which is lower than the 22 tonnes ha −1 average yields currently obtained in Asia.
Cassava root yield at the UAM site was generally lower than yields obtained from the other locations, mainly due to poor soil drainage at the site after rain (Table 8). Cassava root yield from plots treated with isoxaflutole þ cyprosulfamide, indaziflam þ isoxaflutole, prometryn þ S-metolachlor, and diflufenican þ flufenacet þ flurtamone (Table 8, yellow) more than doubled the yield from the hoe-weeded plot.

Second Cropping Season
Cassava root yield in the second cropping season was not affected by location (P = 0.2671). Location-by-treatment interaction did not affect root yield (P = 0.7145). A significant treatment effect was observed when treatments were averaged across location (P = 0.0011) mainly due to root yield from plots treated with aclonifen (Table 8, yellow) and the weedy treatment. Aclonifen was not effective in controlling weeds at this site, resulting in severe weed competition with cassava. Cassava root yields in 24 out of 45 herbicide treatments in the second cropping season were comparable to those in hoe-weeded treatments. Cassava root yield in plots treated with isoxaflutole, indaziflam þ isoxaflutole, sulfentrazone, and terbuthylazine þ S-metolachlor (Table 8, dark green) was significantly higher by 8.7 to 13.7 tonnes ha −1 than for plots treated with S-metolachlor þ atrazine at both rates. Cassava root yield in both seasons correlated positively with herbicide efficacy against broadleaf and grass weeds (Table 9). Generally, cassava root yield from the second cropping season was higher than the yield from the first cropping season. The cassava population was modeled after farmers' practice especially in the Humid Forest and Humid Forest Transition Savanna where small-hold farmers intercrop cassava with maize and vegetables in the first planting period (April to June), which usually receives more rainfall. In the planting period (August to October) with less rainfall, farmers tend to increase cassava population to compensate for the absence of a second crop in the season. Ekeleme et al. (2016) reported a similar season-dependent trend in the same agroecology in southwestern Nigeria where our study was conducted. Cassava yield from the second cropping season exceeded the national yield average by 1.8 to 3 times.
Cassava root yield in the hoe-weeded plots exceeded the average national yield of 8.76 tonnes ha −1 (FAOSTAT 2017) except at Benue and Ogun states. There, root yields from most herbicidetreated plots remained below the average national yield, suggesting that selection of an appropriate herbicide and dose rate for weed control in cassava may require the consideration of site-specific conditions. Manual weeding with appropriate timing, especially in the first cropping season, resulted in root yields that were equivalent to those from most herbicide treatments. In this study, hoeweeding was conducted at 4, 8, and 12 WAP. Several weeding regime recommendations (Alabi et al. 2004b;NACWC 1994) exist for smallholder cassava production systems, however, farmers often do not follow them because of scarcity and the high cost of labor, or lack of awareness of the scale of yield damage caused by weed competition with crops. Farmers often perform their first weeding via hoe after the period when cassava should be free of weeds (critical period); this can result in severe yield loss due to competition from weeds. A critical period of 2 to 6 and 8 to 12 WAP is when cassava should be free of weed competition in Nigeria (Akobundu 1980;Melifonwu 1994). These periods correspond to early canopy formation and initiation of the storage roots.
In conclusion, our work identified several PRE herbicides (indaziflam þ isoxaflutole, indaziflam þ metribuzin, flumioxazin þ pyroxasulfone, isoxaflutole, acetochlor þ atrazine þ terbuthylazine, terbuthylazine þ S-metolachlor, and aclonifen þ isoxaflutole) with excellent efficacy against broadleaf weeds and grasses for up to 8 WAT. These treatments plus one hoe-weeding at 10 WAP resulted in root yields that more than doubled the national root yield average in Nigeria. Because cassava is a long-duration crop, PRE herbicides at use rates that are safe on cassava do not provide season-long control through harvest at 12 mo after planting. Cassava requires POST weed control to supplement PRE herbicide treatments. In this experiment, plots treated with herbicides received one hoe-weeding at 10 wk after the PRE herbicide treatment. This suggests the need to screen potential POST herbicides for use together with the PRE herbicides identified here for weed control in cassava. Several PRE herbicide treatments plus one hoe-weeding at 10 WAP gave superior cassava root yield compared with hoe-weeding at 4, 8, and 12 WAP. We attributed this trend to early weed emergence in the hoe-weeded treatment. Although hoe-weeding at 4, 8, and 12 WAP is usually recommended to smallholder farmers as the appropriate weeding regime in cassava, field observation at all sites showed that the first hoe weeding at 4 WAP was too late for effective weed control. At 4 WAP, perennial and fast-growing weed species such as Mexican sunflower [Tithonia diversifolia (Hemsl.) A. Gray], cogon grass, giant potato, passionflower, guineagrass (Panicum maximum Jacq.), and woodrose [Merremia cissoides (Lam.) Hallier f.] had infested the field, resulting in intense competition with cassava. Cassava is susceptible to weed competition, especially at 2 to 3 WAP when the Table 8. Cassava fresh root yield as influenced by herbicide treatment and manual hoe-weeding in first and second cropping seasons 9 mo after planting. a Values followed by the same letter within a column are not significantly different at α = 0.05. growth rate is slow and at the root formation stage. A first hoeweeding at 3 WAP in the plowed and ridged field and at 2 WAP in the plowed but unridged field may result in better weed control and root yield than a first hoe-weeding at 4 WAP. The need for early and timely weed control in cassava makes the use of PRE herbicide a better option than manual hoe-weeding.